This file contains a running commentary on updates to all R-matrix related codes
- maintained by NRB -
13/04/22 stgicf (4.6), stgicfdamp (4.6)
***************************************
FDIP (top-up) routines were called for all J for printing CC vs CBe partial omegas,
despite default print level being off.
This could be problematic for fine energy mesh and/or low-J (FDIP2 failure).
Only call FDIP now if print level asks for it or if it is actually required for top-up.
Could still cause a problem on the coarse energy mesh, if unlucky.
Overflow in FDIP2 is now trapped.
*** stgf suite only ever called FDIP when top-up was required.
*** TBD: update stgf suite's FDIP2, for safety.
29/06/21
********
Removed tabs. Incremented all version numbers.
07/10/20 stgb (2.20), stgbb (2.7), stgbf (2.16)
***********************************************
Deeply-closed channels coding introduced in
16/06/97 stgb (2.10) did not pass outer-region channel normalization XVECT to the B-files.
Consequently,
09/07/97 stgbf (2.7) was incorrectly normalized for such deeply-closed channels
19/11/99 stgbb (2.6) was incorrectly normalized for such deeply-closed channels
from there-on until now. (Earlier versions just zeroed-out such channels.)
stgb now passes XVECT on an existing record so older/other codes ignore it.
Also, stgbb,stgbf check for presence of XVECT and if not found (old B-files)
then zero-out any deeply-channels in the outer region.
Note, stgb cannot zero-out such channels without crashing if MZPTS & AC allow
a sufficiently large RTWO. This the reason for moving-in. (And to include such
outer-region contributions.)
N.B. Dates are approx, versions are not.
23/10/19 stgicf (4.5), stgicfdamp (4.5)
***************************************
Attempt to correct JMNTWO (22/06/19) was incorrect.
If JMNTWO was correct and even, fine. But it tried to change it if was odd and correct.
16/10/19 stgicf (4.5), stgicfdamp (4.5)
***************************************
Fixed bug (mixed type in intrinsic MOD) which caused gfortran to fail to compile,
introduced by previous fix 22/06/19.
22/06/19 stgicf (4.5), stgicfdamp (4.5)
***************************************
Attempt to correct JMNTWO if mis-aligned with total spin viz one half-integer the other integer.
20/06/19 omgmrg.f
*****************
Now checks that symmetries match. omadd does this, but is needed in omgmrg as well
so as to catch a mis-match between AUTOSTRUCTURE OMGINF file and R-matrix OMEGA.
18/06/19 omadd.f
****************
Missed a de-allocate when points being dropped.
12/06/19 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
If user sets MZLMX=2, as they should, to reduce memory (since stgf by default only
includes dipole & quadrupole perturbing potentials) then LS top-up in large-L will fail
because a secondary dimension is using MZLMX to "determine" the max target orbital l.
Now "hardwired" for a max l=4 and quietly drop any higher. This top-up in large-L is
v. small and not done in the level resolved case.
08/06/19 stg2r 3.7
******************
Suppress false *** WARNING: MIXED EXCHANGE & NON-EXCHANGE RUN! in RELOP='TCC' run.
04/06/19 omadd.f
****************
Coding below only worked if infinite energy point was on file1 (array out of bounds.)
09/01/19 Utility codes
**********************
All web-based utility codes which process the OMEGA file:
om2omu.f, omu2om.f, omr2omc.f, omc2omr.f, omgmrg.f, omgmrgp.f, omadd.f,
omorder.f (= omreduce.f), omgcmp.f, omsumps.f, xtrct.f
are now fully allocatable, all dimensions determined by OMEGA.
Also, treatment of non-default elastic transitions has been made consistent throughout
the suite, at least as far is as possible for codes with no user input, see below.
If the number of transitions has been restricted by the user (beyond +/- elastic)
and elastic is not default, then user must flag so. Currently, this is not possible for
om2omu.f, omu2om.f, omr2omc.f, omc2omr.f - so don't restrict if you need to use these!
13/12/18 stg1r (1.4)
********************
Changed LAMAX default from 2*MAXLA to LAMAX=2 since p/stgf default is to ignore .gt.2.
Helpful for R-matrix script since it does not set LAMAX explicitly. For codes
compiled with MZLMX=2, p/stg2/3/r ignore .gt. 2, but p/stg1/r does not, of course.
07/12/18 stgicf (4.5), stgicfdamp (4.5)
***************************************
Broke-up an IF statement (in sr.rdlsne) into two parts, for compilers
that accessed the second part even when the first part terminated the IF
(the second part has array out of bounds if evaluated then).
17/08/18 adasexj 3.17
*********************
Did not correctly copy header of CA adf04_om to adf04_ups - AS CADW.
18/08/17 stgicf (4.5), stgicfdamp (4.5)
***************************************
Default TCC operation "changed" so that the term energies in TCCDW.DAT are
always matched against those in term.dat so as to determine the CC terms.
Previously, the user had to edit TCCDW.DAT to flag such, e.g. for use with
CC terms interspersed with CI/correlation terms. Only all CC below all CI
was automatic since energies in term.dat not read by default, only the
number of CC terms. Since energy matching is not an issue (or should not be)
since both files use energies determined by R-matrix, "unnecessary" matching
when all CC below CI is outweighed by being able to handle the mixed case
without user intervention, i.e. can be used more generally with R-matrix
Script - see related KCUT development below.
Note, if for some reason, the user wants to revert to ignoring term.dat energies,
remove term.dat file and enter its no. of CC terms (NTLS) as NTLSRD in TCCDW.DAT.
17/08/17 stgjk 2.28
*******************
KCUT selection did not work correctly on more complex cases.
16/08/17 stg2r 3.7
******************
MSYM(MZNC1->MZTAR) array out of bounds, only when writing term energies case TCC/BP
(as NAST gets expanded to all), but could overwrite memory...
24/07/17 stgjk 2.28
*******************
Added KCUT (and TOLB) option to &STGJB as per stg2.
24/07/17 stg2r 3.7
******************
Updated to work correctly with RELOP="TCC" - CI expansion gets expanded to become CC,
but minimal user list is the necessary symmetries, as before, i.e. KCUT has no effect.
20/07/17 stg2r 3.7
******************
If correlation (CI) terms lie below spectroscopic CC terms then the user has to
specify their skipping, either individually via ISKIP, or via energy ranges
such as ESKPL, ESKPH, ECORR. This can be tedious.
In addition, the R-matrix script uses the AS KCUT to flag a CC config subset
of the full CI expansion. This writes only the CC terms to TERMS which is then
used to construct dstg2. This "assumes" that ALL spec. (CC) terms lie below
ALL correlation terms, of same symmetry. If not, stgicf likely will give an
error, but BPRM...
Now, if user (or script) sets the AS KCUT in the dstg2 NAMELIST STG2B then
sr.stg2rd assumes the TERMS file has been used to construct the symmetry list
in dstg2 and also reads the target energies (and the preceding two
intger columns - free-formatted). The CC terms are then selected in sr.bound
by matching to the energies in TERMS of the same symmetry, to within TOLB Ryd,
which the user can adjust as necessary - default = 5.e-4 Ryd.
This all assumes that the configuration number is a good quantum number.
Occasionally, AS flags a TERM as correlation (and v.v.) which it clearly
is not - it sits amongst a sea of spectroscopic. The user then must edit
dstg2 manually to correct.
Any ISKIP data is ignored - just list the term energies you want!
28/06/17 adasexj 3.16
*********************
Made dimension test operation INTEGER*8.
24/02/17 stg1r
**************
Generate B-B and B-C dipole polarization integrals (SR.ONEELE) for non-core
(i.e. non-B&W) orbitals - previously, only C-C were evaluated.
20/02/17 stg1r/stg2r/stgjk/stg3r
********************************
Pass MAXNCB through to all (inner) stages, for test purposes - serial only.
08/02/17 stg1r/stg2r/stgjk
**************************
TEC version now on web - see below.
(Delayed as had not been ported into the parallel codes until recently.)
12/10/16 stg2r 3.6
******************
Flag direct access usage (memory need) in sr.setmx1. Previously, only
sr.chektp usage (for RK.DAT) was flagged.
07/04/15 adasexj 3.16
*********************
Factor 2 missing from characteristic temperature of Juttner relativistic
correction to Maxwellian distribution: Ryd vs a.u.
07/04/15 stgbf0damp (4.7)
*************************
Minor (TWG): clarified comments about continuum normalization and
corresponding PI coefficient. Use a more accurate value for the
fine-structure constant, overall coefficient changes by <0.1%.
13/02/15 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
28/10/14 development could go into an infinite loop for l=0: RZERO/2 got reset
to RZERO and could not satisfy .lt. RZERO/3 to go no further after RZERO/4.
This reset occurred when the phase was ~zero and the code attempted to use
a large step - not particularly numerically accurate either. Now ensure new
start is not reset so, and limit "large" step to improve accuracy in such cases.
04/11/14 stg3r (3.5)
********************
Buttle correction to dipole matrices: ABUT* and BBUT* (*=L and V).
Original coding of DMAT is consistent with their use by prebf.
When someone split this into DMAT, DMAT1 (original DMAT) and DMAT2, the
roles of ABUT and BBUT got reversed in DMAT2 wrt the the original DMAT
(now DMAT1) operation and their WRITE in the new DMAT.
To handle this, reverse ABUT and BBUT in the CALL DMAT2.
This changes the content of the Dnm files!
Note, a different approach is taken for the parallel implementation (pstgd_dip).
Unlike EIE, the Buttle correction to the dipole matrices is small - the
original work by Yu Yan predated DMAT2. In fact, the role reversal of
ABUT and BBUT (essentially applying the initial state Buttle correction
to the final state and vice versa) does not change the size of the effect.
Only prebf (for stgbf) has the capability to use these dipole corrections.
Because they are small, (OP) production work omited them.
The radiation damped codes cannot make any use of them - the reads simply
skip over them.
28/10/14 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
Case MQDT, large box, if s and c function series evaluation fails to converge
then moves inside box to 1.2*RINF (inner turning point), but for l>0 only.
For l=0, went to RZERO/2. If really large box then this is too large still
e.g. half an n=8 box is still larger than an n=5 box. Now allow for a
second halving if first fails to converge, i.e. tries RZERO/4, but no more.
Note on non-MQDT operation with large box: when closed channel outer region
(theta) solution is too large for outer region grid (i.e. outer turning point
.gt. "MZPTS") determines s and c at RZERO and forms theta from appropriate
lin. comb. prior to outward integration if needed (PERT='YES'). BUT, if
s and c fail to converge (likely!) does *NOT* move inside box (different coding)
and so drops channel. So, ensure MZPTS large enough to determine all
"necessary" n. E.g. 1200 is likely not enough points, 10000 is better (slower).
For an n=5 box there should be sufficient convergence of s and c for these
points not to be an issue.
23/10/14 stg2r 3.5
******************
Added ISHFTLS.ne.0 to ensure e-vectors are written, needed for TECs in stgjk.
07/10/14 adasexj 3.16
*********************
Minor: do not attempt to determine forbidden high energy behaviour when
collision strength is identically zero. (IRMPS=-1 was workaround.)
16/09/14 stgicf (4.4), stgicfdamp (4.4)
***************************************
Attempt to fix long standing problem of synchronizing non/ordering of
near degenerate terms betweem TCCDW.DAT and term.dat (i.e. TCCs and K-matrices).
SR.ORDER does not re-order if within 1.D-7, but this was a.u. in the inner region
and Ryd in the outer region. Now compensated-for.
Exact degeneracy (to 7 d.p.) was/is not a problem, and so remains a manual fix.
26/08/14 adasexj 3.13-15
************************
Relativistic Juttner correction factor applied to electron distribution
if IREL.ne.0. (Default=0, off.)
Can read, interpolate and convolute AS-DW/PWB type-5 files, if detected.
Default binned energy mesh adjusted to be (i) more efficient for small
resonance regions (ii) more accurate at low-T, esp. for small kappa.
No longer writes a file called "adf04", rather adf04_om and/or adf04_ups,
as appropriate.
Forbidden transitions default returned to 1/E^alpha extrapolation.
(alpha=2 was forced to avoid additional source differences when
comparing binned/non-binned etc.)
Numerous other (under the bonnet) improvements.
01/08/14 adasexj 3.10-12
************************
If XDRY.gt.1 then use Druyvesteyn distribution, instead of the default Maxwellian.
17/07/14 adasexj 3.8/9
**********************
If KAPPA.gt.1.5 then use a Kappa distribution, instead of the default Maxwellian.
12/07/14 adasexj 3.7
********************
NSTATES.LT.0 enables read (back in) of a (type 5) binned collision strength adf04 file.
(Ideally, written by adasexj...) Currently, the actual value is not used, just the sign.
(It needs NBIN=0, the default, since this flags what is to be calculated and written,
not what is being read.)
Only the NAMELIST is read. Any of the "usual" subsequent data is ignored, currently.
This development removes any dependence on the original OMEGA file, which can be siloed.
It reads from adf04_om (so mv adf04 adf04_om) and writes to adf04_ups (sound familiar?)
Note, all bin energies are now chosen pre-rounded to 3 s.f. so that no error is
introduced into the bin widths on reading the bin energies. (So, do not read a
binned adf04 file written by an earlier version, e.g. 3.5.)
Not averaging (the binned collision strengths) over the bin width in the first
place is an alternative strategy, but then if someone does a quadrature on
the tabulated "collision strengths"...
03/07/14 adasexj 3.x (x=1-6)
****************************
Generates an adf04 file of binned collision strengths, from an RM OMEGA
file, at a set of final (scattered electron) bin energies (so, type 5).
Default (NBIN=0) gives the usual convoluted upsilons without binning.
Set |NBIN|.NE.0 for the binned output.
NBIN.LT.0 usefully, in addition, loops back-up and convolutes
the binned collision strengths, and writes them to the end of the same adf04 file,
after the binned data. You can then compare with the NBIN=0 run.
If you use NBIN=+1 or -1 to bin then the code resets |NBIN| to a reasonable value.
A likely suitable energy mesh is set-up by default (STC):
C NBIN -- .GT. 0 NUMBER OF LOGARITHMIC SPACED BIN ENERGIES.
C .LT. 0 LOOP BACK-UP AND CONVOLUTE BINNED (NBIN=-NBIN).
C IF |NBIN| .LT. 10 THEN INTERNALLY RESET TO 101.
C DEFAULT: 0 (NO BINNED DATA)
C
C EMIN,MAX-- ENERGY RANGE OF NBIN.
C DEFAULTS: TKADAS(1)/5, ENAST(NSTATES) (UNSCALED RYD)
C
C NLIN -- NUMBER OF LINEARLY SPACED BIN ENERGIES TO FOLLOW LOG.
C DEFAULT: INITIALLY, ALL ALLOWED BY DIMENSION. THIS IS
C REDUCED SUBSEQUENTLY WHEN THE LAST ENERGY IS
C KNOWN FROM THE omega FILE.
C
C DELIN -- LINEAR STEP.
C DEFAULT: THE LAST STEP OF THE LOG BIN ENERGIES.
C
C MAXE -- FINAL BIN ENERGY. ONLY USED IF DELIN.EQ.0.
C DEFAULT: 0 (NOT USED THEN)
See comments at start of code for further details.
23/06/14 adasexj 2.x ( x=1-11)
******************************
Complete rewrite. Removed all legacy code (e.g. least squares fits).
Can reproduce default results of v1.11. Non-standard energy meshes may
give somewhat different results for high-T forbidden transitions.
Present results equally valid.
Added NLOWMN, NLOWMX, NUPMN, NUPMX -- can be used to restrict output
to a subset of transitions from lower states between
NLOWMN & NLOWMX to upper states between NUPMN & NUPMX.
Default: all inelastic.
Merged adasex branching. Flag LS via ITCC=-1.
20/05/14 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
An unfortunate user choice of MINLT in BP/DARC operation
could cause all 2J.gt.MINLT to be omitted.
Basically, 11-20 (non-MQDT) or 16-30 (MQDT), for default ISKP0.
12/03/14 stgjk (2.36)
*********************
Implement AS TEC operation (ISHFTLS etc).
18/12/13 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
Use theta for hybrid MQDT for NU.lt.FNUHYB, not just .AND.NU.lt.L.
FNUHYB.gt.0 now sets appropriate value for IOMSW.
FNUMIN only operates on S & C QDT channels, not theta.
24/09/13 adasexj (1.11)
***********************
Fixes for pure algebraic jK-recoupling (ITCC=0).
18/09/13 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
Fixes for MQDT operation: do not step too far inside the box else
integrating away from c-function divergence; do not switch to
K-matrix pert too soon - low channel-l can occur on high total-L
in complex atoms.
24/07/13 stgicf (4.4), stgicfdamp (4.4)
***************************************
Added TOLCC option to drop TCCs involving CC terms whose abs value.lt.TOLCC.
(All TCCs are automatically dropped involving terms only in the CI expansion.)
19/07/13 stgicf (4.4), stgicfdamp (4.4)
***************************************
Added new BLAS coding to MQDTK,TMTRIX & ZMQDTK,ZSMTRX. (Increases memory
somewhat, see notes in said routines if need to revert.)
26/06/13 stgjk (2.25)
*********************
Wrong dimension tested in COPYTP: AIJ is MZNC1, not MXNC1.
25/06/13 stgfdamp (4.7), stgbf0damp (4.7)
******************************************
BLAS version now uses DGEMM in SR.RINIT (as stgf below, 20/06/13).
Also SR.ZMQDTK and SR.PIABSK (stgbf0damp).
***Note: a BLAS version of SR.ZRMAT has also been implemented, but not fully
tested and so it is not part of the public distribution (guinea pigs welcome.)
25/06/13 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
SR.PETKMX and PETTMX now have BLAS coding for all IPERT/IQDT options.
Also, switch DSYMM to DGEMM since we are not using the other half matrix
for anything and the later is maybe 20% faster (it seems) likely due
to simpler accesing of memory in the (full) matrix multiplcation, which
is carried-out by both of course.
21/06/13 stgf (4.8)
*******************
Further BLAS code for matrix multiply in SR.MQDTK.
20/06/13 stgf (4.8)
*******************
BLAS version now uses DGEMM in SR.RINIT (like parallel) now that
DGEMM is called repeatedly to carry-out half-matrix multiply by
sub-block - do not need full WMAT transpose now.
DGEMM from highly optimized libraries is much faster (>>2) than DDOT or
DGEMV which can explicitly do the half-matrix multiply only.
16/06/13 stgf (4.8), stgfdamp (4.7), stgbf0damp (4.7)
*****************************************************
***IMPORTANT: the size of present day R-matrix calculations often
leads to much larger box sizes than was historic. The determination
of outer region solutions from Coulomb series at the box increasingly
fails. This is particularly true for MQDT operation. The previous
default was to drop channels for such failures as they were deemed
to be deeply closed (i.e. classically forbidden). Now, the 09/10/08
option is the *default*, i.e. for such cases, move inside the box, where
starting values can be determined and Numerov out to the box boundary.
This is controlled by the IOMSW parameter.
Old defaults: IOMSW=1 (IQDT.gt.0, i.e. MQDT) IOMSW=0 (IQDT.le.0, i.e. non-MQDT)
New defaults: IOMSW=11 (IQDT.gt.0, i.e. MQDT) IOMSW=10 (IQDT.le.0, i.e. non-MQDT)
14/06/13 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
Catch an extreme case of overflow in MQDT long-range integerals.
12/06/13 hmerge
***************
Re-vamped H.DAT concatenation program to handle an arbitrary mixture
of H-files with or without headers, allow user to extract a subset
of symmetries via IPWINIT, IPWFINAL and restrict the long range
multipoles via LAMAX.
Also fixed a bug (from PJS) when surface amplitudes were split -
output H.DAT was flagged incorrectly because the split was "removed".
Now, any split is passed thru unchanged, and flagged correctly.
06/06/13 stgicfdamp.f (4.4)
***************************
Gailitis average was (always) incorrect for transitions from excited states.
Attempt to catch pathological case of complex diagonalization of identity,
used in Gailitis average.
If DR was switched-on (non default) then electron-impact excitation cross
sections were wrong - would not normally do this since the two processes
optimally use very different energy meshes.
06/06/13 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
Attempt to catch pathological case of complex diagonalization of identity,
used in Gailitis average.
23/05/13 stgf (4.7)
*******************
Fix for SR.THETA generating negative Wronskian, case nu.lt.l+1.
27/03/13 omadd.f (1.12)
***********************
If IDROP in use, then infinite energy point on FILE1 was lost. (Obvious failure.)
25/03/13 stg2r.f (3.5)
**********************
RELOP='TCC' gave incorrect message about a mixed exchange and non-exchange run,
when IDWOUT=2.
15/03/13 omadd.f (1.12)
***********************
Now test for any mis-match between file2 and file1 energy meshes.
(File2 must span file1.) Introduce IDROP for user (>0) or code (<0)
to drop |IDROP| final energy points, excluding any infinite energy.
07/03/13 stg1r (2.39)
*********************
Trivial: BASORB/FINDEE error message gave incorrect n-value to be
treated as Lagrange orthogonalization.
08/11/11 adasexj (1.10), adasex (1.9)
*************************************
25/08/10 update did not work for new style (short) config strings. (KB)
10/05/11 omgmrgp.f, omgmrg.f, omu2om.f, omadd.f, omc2omr.f
***********************************************************
If E0=ENAT(2) in stgf suite then may or may not get a threshold collision
strength, depending on round-off. Post-processors assume present. This
causes unformatted columnwise read to fail. Added "err=" to escape
gracefully. But best to start E0 half an energy step below ENAT(2).
07/01/11 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
The 26/02/09 update for IMESH=3 now means that IMESH=3 prunes near-threshold
energies by default for neutrals (as case IMESH=1). This is likely desirable,
but there was no facility for the experienced/foolish user to override because
ABVTHR and BELTHR (=0 effectively switches-off pruning) was (only) in the /MESH1/
NAMELIST. Now added to /MESH3/. IMESH=2 does not need to prune, by definition.
04/11/10 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
Array out-of-bounds if multipole dimension (MZLMX) less than no. input (LAMAX).
This was caused by attempt to limit reads rather than stop, as in the past.
25/08/10 adasexj (1.10), adasex (1.9)
*************************************
The configurations in the adasex/j.in.form support files were not Eissner or
Standard Configuration format. This means that automatic ADAS processing
was not possible. For adf04 to be integrated seamlessly with ADAS, changed
AS configuration format write to Standard. Modified adasex/j.f so that they
could read new style (as well as old style still) including extended range,
which they couldn't do before. They just write what they read i.e. they
do not covert old style to new. (Also, ported in the irates option from CPB.)
27/07/10 stgbf0damp (4.6)
*************************
Port from parallel code for consistancy.
20/06/10 stgbf0damp (4.6)
*************************
Tweak tolerances for MQDT to try and cope with breakdown of
deeply-closded channel representation when Auger damping present.
20/06/10 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
Do not interpolate when IEQ=-1. (Unnecessary problems with IOMIT.)
21/05/10 stgf (4.7)
*******************
Minor.
21/05/10 stgfdamp (4.6), stgbf0damp (4.6)
*****************************************
Floating point exception (divide by zero) for NTYP2OR radiation
for deeply closed channels - effective quantum number <0.5 was
rounded down to integer 0 - now skip.
10/04/10 stgicf (4.3)
*********************
Elastic collision strengths "obviously" wrong.
16/02/10 stgicfdamp_blas (4.3)
******************************
A user set WORK*8 array for a BLAS routine should have been ZWORK*16.
Once the number of open-channels was greater than half-dimensioned
(NMCJ/2) then memory went out of bounds (but not array bounds).
08/10/09 stg2r (3.5), stglib (2.13)
***********************************
Bug in LNOEX=-1 which affected monopole transitions.
04/09/09 stg1r (2.30) stg2r (3.5)
*********************************
Allow LNOEX=-1 flag to switch-off all cont-cont exchange Slater interactions.
If used at low-L, bound-cont 1- and 2-body interations are NOT switched-off.
Best set in STG1B. Should only use MINLT, in STG2B, such that there are
no bound-bound and bound-continuum contributions anyway (as is the case
for the explict NX codes.)
02/09/09 stg2r (3.5)
********************
For total L .gt. LNOEX in LS-coupling, neglect exchange and form only
the diagonal sub-blocks of the (N+1)-electron Hamiltonian for each
target spin, since the off-diagonal blocks which couple different target
spins for a given total spin are identically zero. (This mimics
the behaviour of the explicit NX codes, without the heavy disc I/O.)
Save for the highest total/target spin, the sub-blocks are roughly
a factor of two smaller since only one, not two, target spins need
to be considered. (The highest total spin was always "small" because
only one target spin can contribute.) The total spin is flagged .lt. 0
as in the NX case. stg3r is transparent to this. stgf picks it up as
with the NX codes. So, stgicf requires INOEXCH=1 to be set, as per NX,
since it must re-construct the complete K-matrix for spin-orbit mixing.
("Obviously", exchange and non-exchange symmetries cannot be mixed
in the same stg2r run for stgicf.)
01/09/09 stg1r (2.39), stg2r (3.5), stglib (2.13)
*************************************************
Added LNOEX (the maximum exchange lambda multipole) to stg1r (&STG1B)
and slightly, but critically, adjusted the usage of the corresponding
variable in stg2r/stglib. Previously, if exchange integrals were
"dropped" in stg1r due to inaccuracy, they were in fact merely zeroed-out.
Memory was set aside to index/store them and their zero-value written to file.
If LNOEX was set in stg2r, independent of whether they had been dropped
in stg1r, then exchange multipoles were explicitly dropped in stg2r -
no angular algebra was determined for them. Now, if LNOEX is set in
stg1r then exchange multipole (integrals) .gt. LNOEX are explicitly
never indexed, stored or written to file, and this value of LNOEX
is passed to stg2r. stg2r now checks for consistency between LNOEX
in stg2r and stg1r, but only if LNOEX was set in stg1r.
Thus, IF you set LNOEX (in the new stg1r!) you MUST update stg2r *AND*
stglib (because stglib is where LNOEX is used, and LNOEX was interpreted
slightly differently there because of its original introduction
in association with STGDW and IDWOUT.) Otherwise, the codes are backwards
compatible. Note, for a given total orbital angular momentum L, then
the exchange multipoles are L, L+/-1,... So, dropping exchange multipoles
.gt. LNOEX is approximately the same as dropping exchange at L .gt. LNOEX.
(Since exchange should be negligible when it is dropped, the presence or
absence of a multipole or two either side of LNOEX is ignorable. In the
case of STGDW it was implemented more rigorously. CC has hijacked LNOEX,
but its usage differs slightly from DW hence the slight hand-waving.
The root issue is the fact that Slater integrals are computed independent
of total L, as indeed is the initial angular algebra before its recoupling,
i.e. the total L is not known when the decision to in/exclude is made.)
27/08/09 stglib (2.13)
**********************
Fix GENSUM INTEGER*4 bug (carry-over from DARC).
12/08/09 stg1nx.f (1.11)
************************
Trivial format expansion for large dimension problems.
26/02/09 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
For IMESH=3, implement same set-up as IMESH=1 for MQDT: case when
a fine mesh is generated externally, rather than the odd (coarse)
energy, as per MCW's "tree" method.
22/12/08 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
Fix for case where initial step H is zero (divide by zero).
15/12/08 stgb (2.19)
********************
Read blocked WMAT from pstg3r.
28/11/08 stg2nx (1.11)
**********************
FINDER failed to find basis orbital with correct nodes for large box size.
Re-sync'ed with stg1r FINDER which did work.
09/10/08 stgf (4.7), stgfdamp (4.6), stgbf0damp (4.6)
*****************************************************
For deeply closed-channels, series solution of unphysical s- and c-functions
can fail to converge at RZERO (especially K-shell holes) and these channels
are then omitted from MQDT operation. Now, code can move inwards and try again
and if converged Numerov back out to RZERO. This is the new default operation
for stgbf0damp with Auger damping - which is the only case where it has
been noticed to make a difference so far - else operates as before in stgf etc.
Can switch-on the new operation elsewise by setting IOMSW=11,10,-11 instead
of 1,0,-1, respectively. "Recall", IOMSW=1 is default (except stgbf0damp now
where it is 11, when IAUGER.gt.0) which drops MQDT channels with nu.lt.l.
IOMSW=0 (or 10) attempts to include nu.lt.l, but this can (eventually) run into
linear dependence problems and so there is the hybrid method (IOMSW=-1,-11).
See Gorczyca and Badnell JPB v33, pp2511-23 (2000) for a detailed study.
***It is not recommended that the user change MOD(IOMSW,10) from the default.
09/10/08 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Archive accumulated changes.
10/08/08 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Output DARC K-mx for differential code.
01/08/08 stg1r v2.38
********************
Keith (KAB) intoduced FINDEE in v2.37 for cases where FINDER failed.
But, FINDEE did not check for correct number of nodes.
07/07/08 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
A do-loop indexing bug was introduced into ZRMAT on first moving from v4.2 to
v4.3 datestamped 22/03/04 and was still present in all v4.4 and the early v4.5.
A partial correction was made to v4.5 dated 14/01/06. It has now been
fully-corrected and reproduces v4.2 results. Prior to the first correction,
the bug led to array out-of-bounds. Post first, could lead to broad false
(NTYP2I) DR/PI resonance(s) - the effect on corresponding excitation was just
the wiping out of an isolated resonance.
26/06/08 stg1r (2.38)
*********************
Re-instated Keith's development for accurate treatment of high-L exchange
intgerals by RS [JPB39, 3873 (2006)] after fixing the B&W crash cause - see
18/03/07.
10/06/08 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Compare QETEST with EINCR (coarse) for extreme cases (need QETEST<EINCR).
Minor mods for use with DARC.
30/05/08 stg3r (3.4)
********************
Fix longstanding problem of inconsistency between ORDER acting on
target energies and channel energies when they are near degenerate,
within the tolerance of ORDER for not swapping. Only ORDER on target
energies, then sum over channel ang. mom. to determine channel order.
Occurs for hydrogenic pseudostate problem, work around was to set
targets at 1.e-6 separation. (Note, first attempt of 28/05 was wrong.)
21/05/08 stg3r (3.4)
********************
Make DSYEVD the default LAPACK diagonalizer as DSYEVR fails IEEEOK test
with optimized compiler libraries more often these days, or so it seems
(e.g. SUN, AMD).
06/05/08 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
Attempt to switch-off NTYP1 damping in classically forbidden region,
where MQDT then gives false "resonance". This is a fix, not a solution.
02/05/08 stg3r (3.4)
********************
Allow pseudo-resonance transformation reduction matrices to overflow
to disc (dipole operation needs them all) for small memory processors,
e.g. on parallel machines.
01/05/08 stg3r (3.3)
********************
Buttle correction to orbitals (for dipole matrix) was transformed incorrectly
by pseudo-resonance elimination code. DMAT/DMAT2P now re-worked so that
a separate DMAT2 is no longer necessary memorywise. But will be retained
while DMAT2P is still being tested (altered) so that non-pseudo-state
dipole problem is not (accidently) affected.
30/04/08 stg3r (3.3)
********************
Rename the new DMAT2 as DMAT2P and only enter for dipole pseudo-resonance
elimination and revert to original DMAT2 elsewise. Then, dimensions
associated with DMAT2P (nsizea, nsizeh) can be set as small as desired,
even =1 if not in use. Note, Hamiltionian pseudo-resonance elimination
does NOT use (nsizea, nsizeh), rather it uses the maximal dimensions, and
so can still be used (even for large cases) independent of the dipole code.
30/04/08 stg3r (3.2)
********************
Array out of bounds when processing dipoles with no pseudo-resonance
elimination, following addition of Tom's dipole pseudo-resonance elimination
coding. Has a large memory requirement, even when not being used - under
investigation to reduce.
25/04/08 stgbf0damp (4.5)
*************************
PIABSS failure (no inverse found) eliminated.
25/04/08 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
The BLAS version of stgfdamp,stgbf0damp did not work for PERT='YES'
Should have given complete nonsense for results.
19/04/08 adasexj (1.9), adasex (1.8)
************************************
ILOW=1 (default) determination of high energy behaviour of forbidden
transitions (those for which no limit point exists) used an absolute
number of points to separate the two energies examined. Obviously,
it really is a function of the coarse mesh step. If a very coarse
step was used then code stopped unnecessarily. Of course, a very coarse
step may not be advisable (numerical failure, quadrature accuracy)
but this is not the place to stop. Also, not going to high enough
energy could result in too few points. But, again, the catch should
be, and is, on the highest temperature requested for said energies.
Now, within reason, just steps back one-third of highest finite energy,
as it did before, but does not stop unless absolutely too few points, <10.
23/11/07 stg3r (3.2)
********************
Added a STOP if stg2 symmetries were incorrectly ordered for Tom's dipole
pseudo-resonance elimination. Also, LAPACK version missed a symmetry
update in TAPERD (dipole pseudo. elimin. only.)
21/11/07 stgfdamp (4.5), stgbf0damp (4.5)
******************************************
For IQDT.LE.0, NTYP2OF damping is not explicitly evaluated at every energy step,
for speed, rather a scaled quantity was calculated on a coarse energy mesh and
extrapolated on the fine energy mesh. This could give rise to small sawtooth
features in the recombination cross section. It has been replaced by
linear interpolation, and the sawtooths disappear in the test case.
(Initiated by Tom and Dragan, at WMU.)
06/11/07 stg3r (3.2)
********************
Added Tom's dipole pseudo-resonance elimination coding - see DMAT2.
28/08/07 stgjk (2.25)
*********************
TCCDW.DAT was unreadable when number of CI terms exceeded 999.
10/07/07 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Did not correctly skip read of long-range multipole coefficients when
memory exceeded (LAMAX.gt.MZLMX).
23/05/07 omadd.f (1.11)
***********************
Set IEMIN and/or IEMAX to extract omegas for the range of energies specified
by these energy indexes (IE). If IBIG=1 then the infinite energy point
will still be retained. A normal complete i.e. OMEGA/U file is the output,
but with MXE adjusted to the new number of reduced energies represented.
18/03/07 stg1r (2.38) -> (2.37)
*******************************
Revert back to previous version of stg1r for now as KAB's change of 11/01/07
crashes B&W.
21/02/07 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
Number of real expansion coefficients for theta was overwritten by the
number of complex expansion coefficients for ztheta (same number 99.9%
of the time). Only used by non-MQDT (IQDT=0), normal MQDT operation
(IQDT>0) unaffected.
22/01/07 stgb (2.19)
********************
Scale determinant so as to avoid overflow.
17/01/07 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
COUL: case weakly closed-channel set large RTWO and open channel with
RINF.gt.RZERO then radial function tabulation exceeded dimension MZPTS
(also started open-channel inward integration at unnecessarily large RTWO).
Only likely encountered with (large) RMPS calculation.
15/01/07 stgjk (2.25)
*********************
Did not catch all changes to SPINBB - KAB now checked.
13/01/07 stgjk (2.25)
*********************
KAB/PJS: Phase saga continues. Reverse prime and unprime in BSPNO.
First "reversed" in early 1995. SPINOR then put back to original in late 1995.
So, BSPNO (and SPINOR) are now back to where they were late 1994/early 1995.
***BUT, CALL SETUPE(MZ,MX, interchanges MX,MZ in SPINBB.
Note, NRB/PJS fixed RECUD back in late 2002.
11/01/07 stg1r (2.38)
*********************
KAB's treatment of high L exchange added (by Keith) [JPB 39, 3873 (2006)].
06/11/06 stgb (2.19)
********************
Add and correct BLAS routine usage.
02/11/06 stgb (2.19)
********************
Re-arrange expressions involving determinant so as to avoid overflow
(e.g. deta*detb/detc - deta*(detb/detc)).
30/05/06 grasp0, stg2d
**********************
TNSRJJ bug-fix from Charlotte Froese Fischer (also in pstg2d).
14/02/06 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Read blocked WMAT from pstg3r.
10/02/06 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
Bug in neutral damping - divide by zero (TWG).
27/01/06 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Allow IDIP=1 to write strength.dat independently of IPRKM.
14/01/06 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
Loop index error in ZRMAT. O.K. if not array out of bounds.
***07/07/08 UPDATE.
This bug was introduced on moving from v4.2 to v4.3 22/03/04.
However, this 14/01/06 fix is NOT complete, it corrected the upper
bound of the do-loop but not the lower one.
This was corrected in v4.5 on 07/07/08.
05/10/05 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
"Obvious" conflict (not to SUN compiler) between use of CC, CCP as
array and non-array.
28/09/05 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Again, F-file C-function not defined correctly when low-L follows
high-L (because INOUT not set and so picks-up old value).
05/09/05 stgf (4.6), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
IRAD>0 (write of F-files for stgbf, stgff) S-function not determined
(undefined) at second point (RZERO+H) when point of inflection (RINF)
moves outside of RZERO, - spotted by MJS.
03/09/05 stg1r (2.37)
*********************
KAB's new SR.FINDEE mods for "difficult" cases where BASORB fails
to find the correct number of nodes.
01/09/05 stg1r (2.37)
*********************
Bug in new dielectric polarization potential, ONE not defined.
01/09/05 stg1r (2.37), stg2r (3.4), stg3r (3.1)
***********************************************
Minor, for NAG compiler.
12/08/05 stgff (1.8)
********************
Signs wrong in converting from K- to S-matrix in DTOT.
09/08/05 stg1r (2.37)
*********************
Re-implemented dielectric polarization potential, controlled by
ALFD, RCUT and IPOLFN (=1, Norcross, =2, Baylis).
09/08/05 stgff (1.8)
********************
Renamed SR.MATMUL to SR.MXMUL to avoid linkage warnings or, for some
compilers, failure.
08/08/05 stgff (1.8)
********************
Phase check (re-)instated in SR.AB for jK-coupling. Was suppressed
because could not get it to agree with inner region R-matrix. Problem
was RECUPD uses initial state odd parity and final state even parity,
but LS (STG2D) uses initial state even parity and final state odd
parity (as does STGFF), so needed an extra factor (-1)**(J-J').
05/08/05 stgd (1.9)
*******************
Top-up extended to jK-coupling.
05/08/05 stgff (1.8)
********************
MJS changed output to BF.L-LP files to Z-scaled line strength (was a
quasi-cross section).
04/08/05 stgd (1.8)
*******************
Extended to jK-coupling for Breit-Pauli runs.
(N.B. jj-coupling for a DARC run is not coded, ditto for stgff.)
03/08/05 stgd (1.7), stgff (1.8)
********************************
First "public" release for testing of the line broadening/damping codes.
stgff has been extensively re-written taking into account efficiencies
implemented in other outer region codes. It has also been extended to
neutral atoms and jK-coupling.
stgd has been completely restructured and now runs as a follow-on to
a stgf run (with IPRKM=5), as opposed to standalone from H.DAT.
This enables one to make use of the current stgf, including for neutrals.
stgd is currently LS-coupling only (jK-coupling being worked-on).
BOTH CODES SHOULD ONLY BE USED FOR PARTIALS, AS THERE ARE ISSUES TO
BE RESOLVED WITH TOP-UP.
02/08/05 stgf (4.6)
*******************
IPRKM=5 writes elastic (physical) S-matrices to (direct access) files SNM
for STGD damping constant code. (N.B. stgf only, i.e. undamped serial - but
inc. BLAS.)
12/07/05 stgicf 4.3 stgicfdamp 4.3
**********************************
Set IPRINT.LT.0 to suppress writes to screen (default 0).
02/07/05 stgicfdamp 4.3
***********************
Introduce IREADK to specify reading of real K-mx (IREADK.lt.0). This enables
real exchange K-matrices, which are only Auger damped, to be read. (NX real
K-mx read was always possible via INOEXCH.LT.0; this is still present and
any conflict with IREADK is tested-for.) Note, there was a bug in the
commented-out code to read and store multiple energy real exchange K-mx files
NOT split by symmetry (not serial default).
20/05/05 stg1r (2.36)
*********************
STOP if MZNR2 or MZLR2 dimensions exceeded. NRANG2 (MAXC) or LRANG2 (MAXLT)
were previously reduced instead. This was confusing to novice users.
09/05/05 stg1nx 1.11
********************
There was a bug in TENSOR - it did not take account of seniority in testing
whether the spectator shells are orthogonal. Both L and S of each shell were
checked as identical in the two configurations but not seniority.
The error can only occur in the special case where the two configurations are
identical in each spectator shell except that the seniority in a shell differs.
If you are not using configs of the form 3d^n, for 3.le.n.le.7, then unaffected.
The exchange code (RMATRX1) used a later version of TENSOR which has this
problem fixed but NX used the original TENSOR as published.
(Thanks to PJS for highlighting the discrepancy between NX and X results
and to VMB for tracking it to TENSOR.)
29/04/05 stg3nx 1.11
********************
Stop any NBUG from dstg3 from overwriting NBUG set (or not set) in dstgnx;
i.e. exchange NBUG settings in dstg3 are ignored. NX NBUG must be set in
dstgnx.
28/04/05 stg2nx 1.11
********************
Add NBUG varaibles to dstgnx NAMELIST.
20/04/05 stgicf (4.3), stgicfdamp (4.3)
***************************************
STOP if zeroed-out TCC too large.
15/03/05 stgjk (2.24)
*********************
If levels tagged for skipping lie above ESKPH but there are no higher
levels (below ECORR) of the same symmetry then the wrong levels are
included in the CC expansion (because there is no subsequent higher
CC level to tag with skip info). Note, mdstjk flags this as a
checksum error, but the dstgjk input file looks "o.k.".
stgjk can now handle this and mdstgjk does not flag a checksum error.
04/11/04 stg1r (2.36)
*********************
Tests for use of target box functions.
02/11/04 stg1r (2.36)
*********************
Use R-matrix boundary defined by externally tabulated radial mesh when
it defines a box function i.e. do not redetermine internally.
02/09/04 stgf (4.5), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Speed-up perturbation integrals RZERO to RTWO (mainly non-MQDT).
25/08/04 stgfdamp (4.5), stgbf0damp (4.5)
*****************************************
Partitioned energy (EZERO) missing.
31/07/04 stg2r (3.4), stg3r (3.1)
*********************************
KAB's latest approach to eliminating weakly coupled N+1 terms (which he calls
configurations...)
15/07/04 stgf (4.5), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Dipole top-up in jj-coupling (DARC and DARC interface) was not coded-for.
19/06/04 stgb (2.18)
********************
E-files not being written - KB.
25/05/04 stg3r (3.1)
********************
Corrections to below.
22/05/04 stg3r (3.1)
********************
Added alternative input of energy corrections - see STG3RD for details.
13/05/04 stgf (4.5), stgfdamp (4.5), stgbf0damp (4.5)
*****************************************************
Read partitioned R-matrix H.DAT produced by stg3r v3.1 or later.
13/05/04 stg3r (3.1)
********************
Uses a partitioned R-matrix when IPRCENT.lt.100 (.gt. 50 recommended) is
specified in NAMELIST STG3A. Requires use of stgf v4.5 or later.
27/04/04 stgb (2.18)
********************
Allow read and processing of DARC "H.DAT". Also, processing of
H.DAT from DARC interface, i.e. containing K-1 (not K of jK)
in the L2P array, as opposed to small l. Basically, READ1,2 SCALE1,2
from stgf plus energy derivative of Buttle correction (INTBUTD) dereived
analagously from INTBUT. IDENTFY labelling left as
27/04/04 stgf (4.4), stgfdamp (4.4), stgbf0damp (4.4)
*****************************************************
Code decoupling of core radiative rates in RAD for jj-coupling, case DARC
and DARC interface. Use DARC DSTGH.DAT name for DARC "H.DAT" - temporarily
used an alternative for testing purposes.
24/04/04 stgf (4.4), stgfdamp (4.4), stgbf0damp (4.4)
*****************************************************
Switch to reading of K-1 in L2P array, as opposed to an additional kappa
array, from DARC interface. (DARC DSTGH writes kappa in "L2P" array still.)
K=2*kappa (kappa.gt.0) and K=-2*kappa-1 (kappa.lt.0). K-1 is then the
exact analogue of channel l used in LS/BP case to reference basis
orbitals, e.g. surface amplitudes (ENDS), energies (EIGENS) and
Buttle correction (COEFF). Thus no need to make any changes to their uses
in DARC/interface case. Note, in pure DARC run L2P array is converted
from kappa to K-1 in READ2.
24/04/04 stg3r (2.23)
*********************
DARC interface now writes K-1 in L2P array so no need for any change to
stg3r use of ENDS but need to flag this in H.DAT, by setting LRANG2=-LRANG2.
This is detected in stg3r itself by use of ICODE=25 from DARC interface.
This finally solves the differences between pure DARC run and DARC
interface. (Small l was in L2P array and this was accessed by ENDS, should
have switched to kappa resolution as per stgf Buttle.) Use of K-1 in L2P
array makes this transparent now.
22/04/04 stgf (4.4), stgfdamp (4.4), stgbf0damp (4.4)
*****************************************************
Allow processing of H.DAT from DARC interface to use original DARC Buttle
correction (DBUT.DAT) as opposed to Seaton fit via (BUTFIT).
21/04/04 stgicf_blas (4.3)
**************************
Fixed bug in BLAS version only. Introduced in stgicf v4.3 affects
parallel (2.3) as well v2.2/4.2 o.k.
19/04/04 stgf (4.4), stgfdamp (4.4), stgbf0damp (4.4)
*****************************************************
Allow read and processing of DARC "H.DAT". Also, processing of
H.DAT from DARC interface, i.e. containing kappa (not K of jK).
Needed to reference Buttle correction by kappa, not small l.
09/04/04 stgicf (4.3), stgicfdamp (4.3)
***************************************
Variable (EINCR) not defined case IMODE=0. If taken to be zero by compiler
then results are correct still. If not (e.g. g77 or Intel with -O3) then
NaNs likely to arise in collision strengths. NOTE: repeat of problem of
04/02/04 where EINCRH was not defined, but EINCRH is defined in terms of
EINCR!
26/03/04 stgf (4.3), stgfdamp (4.3), stgbf0damp (4.3)
*****************************************************
Re-worked COUL for open-channels when PERT='YES' to integrate inwards
to RZERO from R2(I) and outwards from R2(I) to RTWO. Previous long
inwards integration gave small errors to S,SP,C,CP at RZERO which
caused large errors in weak elastic K-/S-matrices. In turn, (ICFT)
transforming these LS matrices, gave rise to large spikes in weak
forbidden transitions.
22/03/04 stg2nx 1.11
********************
Reversed orbital and radial indices on orbital storage for RS (CPB).
Large speed-up in big cases.
20/03/04 stgf (4.3), stgfdamp (4.3), stgbf0damp (4.3)
*****************************************************
Reversed indices on WMAT (and AVECT1, DDEC, ZCOEF for damped codes).
Large speed-up in big cases on certain architectures (e.g. Itanium 2).
17/03/04 stg2r (3.4)
********************
STOP rather than just WARNING when LRANG2 too small.
16/03/04 stg1r (2.36)
*********************
Ensure fine enough radial mesh for STO- when fast NX1.DAT run being
made, i.e. with small MAXLT and MAXC.
12/03/04 stgjk (2.24)
*********************
Write N2HDAT=1 to RECUPH for consistency with stg3r.
05/03/04 stg2r (3.4)
********************
Added KAB's option (ICUT=-1) to drop weak N+1 configs (ICUT=0 default still).
04/02/04 stgicf (4.2) stgicfdamp (4.2)
**************************************
Variable (EINCRH) not defined case IMODE=0. If taken to be zero by compiler
then results are correct still. If not (e.g. g77 or Intel with -O3) then
NaNs likely to arise in collision strengths.
04/02/04 stgicf (4.2) stgicfdamp (4.2)
**************************************
Restrict UNIT numbers to .LT. 100, for g77 compiler.
02/02/04 stgjk (2.24)
*********************
Allowance for sign reversal on dipole velocity (Mar 03) introduced
incompatibile lengths in COMMON/LRPOT/.
29/01/04 stgf (4.2), stgfdamp (4.2), stgbf0damp (4.2), stgicfdamp (4.2)
***********************************************************************
REAL -> DBLE
07/11/03 stglib (2.13)
**********************
GENSUM bug-fix from KAB:
COLD IF (I.EQ.1) GOTO 460
IF (I.LT.J) GOTO 460 !KAB 5/11/03
16/10/03 stg1r (2.36), stg2r (3.3), stg3r (2.23)
************************************************
Added the default FORM='UNFORMATTED' explicitly to
OPEN (STATUS='SCRATCH',ACCESS='DIRECT' etc for sloppy f77 compilers.
18/08/03 stg1r (2.36)
*********************
Allow NPSMN(l), NPSMX(l) to independently specify MIN and MAX
pseudo n for given l. Extension of old NPS(l) parameter. (Don)
18/08/03 stgf (4.2), stgfdamp (4.2), stgbf0damp (4.2)
***************************************************
Added output for BP differential code. (Don)
08/08/03 stgf (4.2), stgfdamp (4.2), stgbf0damp (4.2),
stgicf (4.2), stgicfdamp (4.2)
***************************************************
Catch overflow for graceful failure in FDIP. Failure expected and
coded for, but ungraceful hardware failure to be avoided.
08/08/03 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
ITOP not being (re-)set (because not being passed-on from MAIN).
Thus, near-degenerate non-dipole top-up not using our preferred
methodology, but probably o.k. all the same.
05/08/03 stgf (4.2)
****************
Inaccessible statement moved - case neutral MQDT, which stops anyway
if there are any closed channels!
04/08/03 stgb (2.17)
********************
Fix array out of bounds (MZEST).
23/05/03 stg2r (3.3)
********************
KAB's writing of target LS polarizabilities.
13/03/03 stgbf (2.8)
********************
Updated for BP velocity as per stgbb (below).
Must use updated prebf. (Attempt is made to see if vel. array
present and BACKSPACE if apparently not, not guaranteed failsafe.)
LS unaffected.
13/03/03 prebf (2.8)
********************
Pass thru BP velocity array. BP MUST use updated stgbf as the new
DVEC file will be mis-read by old stgbf. DNM o.k. because new
info at end of each file. But single DVEC has new info distrib.
thru out.
13/03/03 stgbb (2.6)
********************
Initial BP velocity to zero for LS case.
11/03/03 stgjk (2.24), stg3r (2.23), stgbb (2.6)
************************************************
Allow for LS reversal in stgjk on velocity.
Hence reversed BP velocity correct in stgbb.
10/03/03 stgbb (2.6)
********************
Reverse sign on DP contribution.
08/03/03 stgjk (2.24)
*********************
Sign reversal on velocity dipoles (relative to length) when matrix
elements reversed.
07/02/03 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
Scaling factors in ZMQDTK were hard wired, now replaced by ZFSCL and FSCL.
06/02/03 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
Neutral MQDT simplified. Scaling introduced via ZFSCL and FSCL,
complex coefficient and energy factor, on K-matrix (IQDT=2) only so far.
Dipole matrices not yet adjusted (stgbf0damp).
31/01/03 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
PERT now added for neutral MQDT.
Corrected MQDTS routine for use on neutrals (i.e. don't!).
Still to factor out threshold energy dependence, so do not interpolate
(or use in stgicfdamp) yet.
25/01/03 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
MQDT extended to neutrals (no PERT yet).
25/01/03 stgf (4.2)
*******************
Minor
21/01/03 stgbf (2.15)
*********************
Inserted PARAMETER (MZMSH0=MZMSH). Delete and add specific value for
MZMSH0 (.lt. MZMSH) to PARAM file if this proves too large (only used
by stgbf).
15/01/03 stg3r (2.23)
*********************
Write headers to H.DAT when IPWINIT.GT.1 if no prior H.DAT exists,
i.e. H.DAT not being appended.
13/01/03 stg2r (3.3)
********************
Fix for LNOEX.
Write sizeH.dat for pstg3r. Previously, info was read from rout2r.
When this was changed to sizeH.dat in pstg2r/pstg3r it got left out of
stg2r.
07/01/03 stgb (2.17), stgf (4.2), stgfdamp (4.2), stgbf0damp (4.2)
*****************************************************************
Fixed bug in evaluation of function derivative following Numerov
integration - neutral atom only.
20/12/02 stgjk (2.24)
*********************
First attempt to fix bug in RECUD phase factor.
20/12/02 stgb (2.17)
********************
Fixed case where symmetry in H.DAT not found by stgb.
(This was because default stgb operation uses symmetry order
in dstgb, not H.DAT as before, for damping codes.)
19/12/02 stg1r (2.36), stg2r(3.3), stgjk (2.24), stg3r (2.23)
*************************************************************
Transparent mods to allow reading of passing files from Werner's codes.
17/12/02 stgbb (2.6)
********************
Fix from KB to allow odd-even parity ordering. Also, added missing
AC from THETBB.
12/12/02 stgicf (4.2) stgicfdamp (4.2)
**************************************
Dimension exceeded can cause crash before reaching STOP test.
01/11/02 stgfdamp (4.2), stgbf0damp (4.2)
*****************************************
Neutral case added.
01/11/02 stgf (4.2)
*******************
Archived 4.1.
17/10/02 stgbb (2.6)
********************
Read of B00 not terminating correctly when -1 -1 -1 used to terminate,
as opposed to EOF.
22/07/02 stg1r (2.36)
*********************
Minor fix for gf77.
06/07/02 stg2r (2.43)
*********************
Minor formatting.
04/07/02 stg2r (2.43) stg3r (2.23)
**********************************
Minor re-organization of some dimensioning.
03/07/02 stg2r (2.43) stg3r (2.23)
**********************************
Adjust for consistency with parallel codes viz STG2HX.DAT -> STG2HXXX.DAT,
split AMAT.DAT to AMATUXXX.DAT (and unformatted).
28/06/02 stg1r (2.36)
*********************
Minor.
31/05/02 stg2r (2.43)
*********************
Improved some checks.
02/05/02 stg3r (2.22), stg3nx (1.11)
************************************
Modify EPSI for DSYEVR.
01/05/02 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Supress over cautious warning messages.
29/04/02 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Continuing high-L RMPS (esp neutrals) problem, this new coding occasionally
causes the outer region mesh to be incorrectly defined for certain channels
at certain enregies and ang. mom. (specifically, if the redefined mesh
still cannot fit on the grid allowed by the dimensions). This has been
corrected.
07/02/02 stgfdamp (4.1), stgbf0damp (4.1) and stgicfdamp (4.2)
***************************************************************
Approx tan(pi*znu) by i when imaginary part of effective quantum number gets
large (Allan Whiteford).
07/01/02 stg1r (2.36)
*********************
Made SR.TABORB radial tabulation to file 'radout' consistent with input for
AUTOSTRUCTURE. Access via NBUG5=2 or RADOUT='YES' (in NAMELIST/STG1A/).
STO output does not have number of orbitals, needs to be added manually
on line 1 of 'radout' file.
06/12/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Continuing high-L RMPS (esp neutrals) problem, cannot allow non-oscillatory
(negative energy) solutions to use excessively large step otherwise Numerov
can overflow. Re-instated testing of these channels, but using a weaker condition
than for oscillatory solutions. Furthermore, drop these deeply-closed channels
if their presence means that outside the box contribution from open-channels
cannot be determined. At high-L, virtually all of the open-channel interaction
is from outside of the box so it is crucial to retain it.
27/11/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Fix for high-L, RMPS in neutrals. Outer-region step should not be
determined by negative energy solutions (since non-oscillatory).
Otherwise, get H excessively small and outer region tabulation is
severely restricted. These deeply-closed channels have negligible
effect on open-channels, as we have seen previously in MQDT.
27/10/01 stg3nx (1.11)
**********************
Minor.
20/10/01 stg1r (2.36), stg2r(2.43), stgjk (2.24)
************************************************
Tidy-up precision etc.
19/10/01 stglib.f (2.13)
************************
Tidy-up precision etc.
18/10/01 stg3r (2.22), stg3nx (1.11)
************************************
Added optional (i.e. commented-out) code for using LAPACK/BLAS library
to carry-out diagonalization of H(N+1). DSYEVR uses 4x memory, but is not
suitable for all architectures, while DSYEVD uses 6x memory and should
work on all systems. A factor 5 speed-up can be obtained with highly
optimized LAPACK/BLAS libraries, e.g. SUN Performance library or ATLAS BLAS.
Tidy-up precision etc.
17/10/01 stg1nx.f (1.10), stg2nx.f (1.10)
*****************************************
Tidy-up precision etc.
23/07/01 stgicf (4.2), stgicfdamp (4.2)
***************************************
In optional read of NTLSRD from TCCDW.DAT, test and reset if NTLSRD.LE.NTLS.
29/06/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1) and
*****************************************************
29/06/01 stgicf (4.2), stgicfdamp (4.2)
***************************************
Modification to OMEGA scratch file operation (Circa March 2001) for parallel
code was incorrect, crashed serial code as well.
09/06/01 stglib (2.12)
**********************
Minor - avoid floating point exception on T3E/IBM-SP.
07/06/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Variable not being initialized in Numerov routines, case of
a single step only (Dario).
30/05/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1) and
*****************************************************
30/05/01 stgicf (4.2), stgicfdamp (4.2)
***************************************
Re-instated option of LS K-/S-matrix files *not* split apart by symmetry.
Of no interest to serial code, but useful for parallel code when using
a large number of processors and a small limit on the global number of
open files. Default is to split by symmetry. These stgicf/damp (4.2) are
thus back compatible with stgf suite 3.x, provided the default is overriden.
29/05/01 stgicfdamp (4.1)
*************************
Fixed Gailitis average & DR - last working version was v3.7 of 09/03/01.
25/05/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1)
*****************************************************
Radiative width was not always being re-calculated when a DR final state
moved from bound to autoionizing as ETOT increased, hence overestimating
DR because autoionizing states were sometimes treated as recombined (stable).
24/05/01 stgf (4.1), stgfdamp (4.1), stgbf0damp (4.1) and
*****************************************************
24/05/01 stgicf (4.1), stgicfdamp (4.1)
***************************************
stgf/damp/bf0damp LS K-/S-matrices for stgicf/damp are split apart by symmetry.
(So, 3.x stgf suite is compatible with 3.x stgicf suite, and 4.x with 4.x.)
21/05/01 stgicf (3.7), stgicfdamp (3.9)
***************************************
Speeded-up the jk->IC recoupling again.
19/05/01 stgicf (3.7), stgicfdamp (3.9)
***************************************
Further reduced memory requirement by moving fully to packed vector
storage. stgicf k(phys)->T substantially faster for large NCHOP.
Re-coupling (jK->IC) a little slower.
12/05/01 stgicf (3.6), stgicfdamp (3.8)
***************************************
Substantially reduced memory requirement for large channel (>1000)
cases.
11/04/01 stgf (3.5), stgfdamp (3.5), stgbf0damp (3.5)
*****************************************************
Minor.
31/03/01 stg3r (2.21)
*********************
Minor mod for parallel code.
09/03/01 stgicf (3.5), stgicfdamp (3.7)
***************************************
By default, the (dipole) infinite energy info is now only written to the
end of OMEGANOTOP. The program omadd (V1.10) uses this to differentiate
between dipole and non-dipole transitions when looking at top-up factors.
See the stgicfadas.txt notes, which have also been updated to take account
of the infinite energy limits coming from AUTOSTRUCTURE (including Born
limits for the non-dipole transitions). This is why the OMEGA produced by
stgicf/damp does not contain infinite energy info now, since it is dipole
only.
08/03/01 ***IMPORTANT*** stgicf (3.5), stgicfdamp (3.7)
*******************************************************
The fix for (near) degenerate dipole top-up introduced on 26/11/00
introduced an error which effectively switched-off the dipole top-up.
Versions between 26 NOV 00 and 21 FEB 01 inclusive should not be used.
A reduced energy plot of any dipole transition should clearly be
inconsistent with the limit point, and we all check this don't we.
On a related issue, it turns out that the OMEGANOTOP file infact has
contained top-up for the dipole transitions since v2.6, introduced
in Feb 2000. Comparisons for non-dipole transitions were unaffected.
02/03/01 stgf (3.5), stgfdamp (3.5), stgbf0damp (3.5)
*****************************************************
Incorporated Seaton's changes to the specification of constants in the Numerov
routines etc. (Implemented slightly differently to ensure correct precision
obtained when using IMPLICIT REAL*8 etc. rather than setting it via the
compiler.) The original constants were given to 8 s.f, consequently some,
hopefully, small differences may arise compared to results from v3.4.
28/02/01 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Fix for COULS at high-L (>60).
21/02/01 stgicf (3.5), stgicfdamp (3.7)
***************************************
Added parity label to 2J+1 target symmetries to enable infinite energy
Born to be distinguished from dipole. (stgf suite already did so.)
21/02/01 stg3nx (1.10)
**********************
Minor.
17/02/01 stg2r (2.42)
*********************
Revised dimension check (Don).
04/01/00 stg1r (2.35), stg2r (2.42), stgjk (2.23), stg3r (2.21)
***************************************************************
Added MAXLD/J2MAXD to limit max L/2J for which dipole data is calculated.
This is independent of MAXLT/J2MAX which limits the scattering symmetries.
Useful for damped excitation/recombination where NTYP2I damping only
occurs for low-L/J but we require scattering to much higher-L/J. Especially
useful in ICFT runs where it is tedious to split dipole and non-dipole
inner region runs. The DNM files obtained are identical to those obtained
from a full run except for: (i) in LS coupling the last few will be ordered
slightly differently due to the omission of a DNM with L>MAXLD which occurs
before a DNM with both L<=MAXLD. (ii) in BP mode the last few DNM files are
apparently different. Infact this is due to some fossil LS-coupling data
that is passed-on by stgjk without re-coupling, and to point (i), and so could
never be used in BP mode. Infact we don't use it in LS-coupling either.
It was used by STG4 to generate (LS) dipole polarizabilities.
01/12/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Speed-up recoupling in KMTRIC by loop interchange (Keith Butler).
14/12/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Added code for use by parallel version only, commented-out in serial code.
26/11/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Fix to dipole top-up for exact degenerate levels.
25/11/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Minor mods for g77.
24/10/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Important mods for MQDT operation on Crays.
12/10/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Use explicit REAL*16 in coulfg for large cases, especially on a Cray.
26/09/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Minor.
26/09/00 stg2r (2.41), stg3r (2.20)
***********************************
Don & Dario's coding of the splitting of STG2H.DAT into multiple files
for running in limited operating systems. Set N2HDAT.gt.1 (.le.10) to
break-up in to N2HDAT files.
16/09/00 stg1r (2.34)
*********************
Switch-off diagonal one-body relativistic corrections except when the
new (ISMIT.NE.0) Schmidt orthogonalization is being used for that L.
13/09/00 stg2nx (1.9)
*********************
Switch-off diagonal mass-velocity terms - gives errors on inelastic BP
cross sections. See also 26/05/00. Default is all off (RELOP='NO').
11/09/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Tweaked near-degenerate non-dipole top-up.
11/09/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Tweaked near-degenerate non-dipole top-up.
07/09/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Case of writing formatted (non-zero) omegas columnwise:
modified number of omegas written to be consistent with rounded energies.
07/09/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Case of writing formatted (non-zero) omegas columnwise:
modified number of omegas written to be consistent with rounded energies.
19/08/00 stgicf (3.5), stgicfdamp (3.7)
***************************************
Replaced matrix inversions by AX=B solution in MQDT routines.
19/08/00 stgf (3.4), stgfdamp (3.4), stgbf0damp (3.4)
*****************************************************
Replaced matrix inversions by AX=B solution in MQDT routines and in
ABINV, which also incorporates KAB's partitioning of the non-MQDT problem.
12/08/00 stgicf (3.4), stgicfdamp (3.6)
***************************************
Added facility to read-in LS K-matrices for batches of energies so as
to reduce I/O. The number read is NMLSE+1 and the additional memory used is
NOMCL*NMLS*NMLSE words. (Use is made of the original memory to store the
first energy read, NMLSE are then held.)
02/08/00 stgf (3.3), stgfdamp (3.3), stgbf0damp (3.3)
*****************************************************
Added explicit symmetric coding for K->S, includng matrix inversions.
31/07/00 stg2r (2.40), stgjk (2.22)
***********************************
Modified ESKPL test to handle certain cases.
27/07/00 stg3r (2.19)
*********************
Further reduced memory requirement of HAMNEW.
No longer need to worry about NSIZEH.
26/07/00 stg1r (2.34)
*********************
Put in warning if all B&W occuppation numbers are zero in the radial file
as code reverts back to using defaults. Set IMAXN=-1 to force them all to
be zero.
25/07/00 stg2r (2.40), stg3r (2.19)
***********************************
Format of AMAT.DAT file written by stg2r needed changing for large N+1 cases
as stg3r failed to read it correctly (but stopped safely). Re-wrote HAMNEW
to use less memory. However, if NSIZEH was previously set much smaller than
MXMAT then memory will increase. If not using HAMNEW then set NSIZEH=1.
22/07/00 stgicf (3.4), stgicfdamp (3.6)
***************************************
Re-orthonormalize (as opposed to re-normalize) TCCs.
Allow number of IC target levels to be less than number of jk levels.
Replaced both BLAS and non-BLAS matrix inversion routines with those
for a symmetric matrix as opposed to a general matrix.
19/07/00 stgbf (2.15)
*********************
M11 added to SAVE in READXY (PJS).
18/07/00 stgicf (3.3), stgicfdamp (3.5)
***************************************
Use of observed term energies failed for default NTLSRD=0 in TCCDW.DAT.
stgicf (3.2), stgicfdamp (3.4) unaffected.
18/07/00 prebf (2.8), stgbf (2.15)
**********************************
ABORT to ABORTT for gnu f77 compiler (PJS).
Fixed bug in stgbf, FLAFGL in CALL READXY (PJS).
15/07/00 stgicf (3.3), stgicfdamp (3.5)
***************************************
Reading of NX K-/S-matrix data could halt because of apparent inconsistancy
of input data with what was expected. Only occurred for certain systems.
Now fixed.
15/07/00 stg2r (2.42)
*********************
Added TITLE to SAVE in SR.STG2RD (PJS).
10/07/00 stgicf (3.3)
*********************
In the commented-out BLAS routines, IPIV should be IPIVOT.
07/07/00 stgfdamp (3.2), stgbf0damp (3.2)
*****************************************
Minor mod to the evaluation of NTYP2OR radiation.
27/06/00 stgbf (2.15), prebf (2.8)
**********************************
Updated for consistancy with 08/04/00 stgb (2.17) writing extra info to B00.
27/06/00 stgf (3.2), stgfdamp (3.2), stgbf0damp (3.2)
*****************************************************
Bug-fix for IMODE=-1 PERT='YES' operation.
24/06/00 stgf (3.2), stgfdamp (3.2), stgbf0damp (3.2)
*****************************************************
Control READ/WRITE of unphysical K/S-matrix via IMODE (as per stgicf) rather
than IQDT viz. IMODE=0 WRITE, IMODE=1 READ, IMODE=-1 no WRITE/READ (single
pass). Use IQDT only to specify the type of MQDT operation (K-, S- etc.)
Default, IMODE=0.
15/06/00 stgf (3.2), stgfdamp (3.2), stgbf0damp (3.2), stgicf (3.3), stgicfdamp (3.5)
*************************************************************************************
Optionally write unformatted collision strength file OMEGAU, if PRINT='UNFORM'.
Added utility routines om2omu.f and omu2om.f to convert. Other utilities, viz.
omadd.f, omgmrg.f, xtrct.f, adasex.f adasexj.f can all handle unformatted files.
12/06/00 stgf (3.2), stgfdamp (3.2), stgbf0damp (3.2)
*****************************************************
Could fail to determine all radiative rates in LS coupling if two
terms never appear in the same N+1 symmetry.
08/06/00 stgicf (3.3), stgicfdamp (3.5)
***************************************
Given the generalization of the use of CI terms interspersed with CC terms,
elimination of CI components from the TCCs has been made more automatic by
comparing the the first NTLSRD term energies in TCCDW.DAT (after energy
ordering) with the NTLS in term.dat. Those terms in TCCDW.DAT that find a
match in term.dat are taken to be CC and those that don't find a match
are taken to be CI. The default is unchanged, i.e., the first NTLS are
taken to be CC. The new option is accessed via the NTLSRD being set non-zero
on line 1 of the TCCDW.DAT file after "LS TERMS". The default is blank/zero.
If this automatic matching is a problem then it can still be done manually
by setting NTLSRD= -|NTLSRD| and tagging the term numbers (in TCCDW.DAT)
negative for the CI. Anything above |NTLSRD| is still taken to be CI.
07/06/00 stg2r (2.40), stgjk (2.22)
***********************************
Added ESKPL, ESKPH, ECORR to specify energy range of terms/levels to be omitted
from the CC expansion. Those that lie between ESKPL and ESKPH Ryd relative
to the ground are omitted. In this case, one may need to specify ECORR, the
energy above which ALL terms/levels are treated as CI since it cannot now be
determined internally. These options are consistent with the ISKIP option
which can be used in conjunction with them. The defaults remain unchanged.
07/06/00 stg3r (2.19)
*********************
Added early dimension check on MXDM so as not to bomb-out in DMAT1
after many hours of CPU.
26/05/00 stg1r (2.34)
*********************
Reversed change 1/ of 29/01/00 i.e. switch-off of diagonal BP integrals
which gave rise to errors in elastic channels only because this is
inconsistent with new Schmidt orthohgonalization (ISMITN=1) and affects
INELASTIC transitions. Ditto stg2nx (1.9).
23/05/00 stg3r (2.19), stg3nx (1.10)
************************************
In solving the obscure bug of 12/04/00 I inadvertently created one:
the channels for a fixed term were ordered by descending l instead
of ascending, the latter is required by the LS top-up in stgf.
22/05/00 stgf (3.2), stgfdamp (3.2), stgbf0damp (3.2) Minor
*****************************************************
Additional tests.
18/05/00 stg2r (2.40) Minor
*********************
Added KAB's code to printout gf-values (IBUG5=-1), minus the test
IF(CFADD.EQ.ZERO) GO TO 180 for safety.
13/04/00 stgg (1.20)
********************
Update projection subroutine to handle case of pseudo and physical
orbitals of same n but different l. Prior to this it was assumed
that all l of a given n were either physical or pseudo. Note,
AUTOSTRUCTURE has also been updated to provide the overlaps for
such a scenario.
12/04/00 stg3r (2.19), stg3nx (1.10)
************************************
Use of a 1.e-7 tolerance in SR.ORDER gives rise to non-energy ordered
states in hydrogenic systems which can cause subsequent problems esp. in
stg3r as ORDER in TAPERD uses a.u. while WRITOP uses Ryd. This causes an
inconsistency between term and channel indexing if the separation is
between 5.e-8 and 1e-7 a.u. Use NAST to read in slightly separated
term energies to ensure the order that you want. There is no problem
with stg3nx but it needs to order target terms consistently with stg3r.
08/04/00 stgfdamp (3.1), stgbf0damp (3.1)
*****************************************
Changed damping defaults and added more checking of user input for damping.
Previously, all damping was OFF by default (but DR was ON, which meant that
elastic transitions were present in the OMEGA file). Now all damping for
excitation is ON (NTYP1=1, NTYP2I=1, NTYP2OR=1) but DR is OFF (NDRMET=0
NTYP2OF=0). The code looks for appropriate B and D files and checks user
input. The user is then instructed to switch off damping if required files
are missing or to generate them if damping is required. Hopefully, this
will mean that damping is not "missed" by inexperienced users. Code also
attempts to determine NMIN, as required by NTYP2OR/F, from a modified B00
file produced by stgb 2.17 or later. Obviously, if NTYP2I=0 this is not
possible but the user is asked for a valid NMIN or told switch off the
NTYP2O damping - again, their choice.
08/04/00 stgf (3.1) Minor
*******************
Minor
08/04/00 stgb (2.17)
********************
Write extra info to end of B00 for stgfdamp.
07/04/00 stgicfdamp (3.4)
*************************
Corrected bug in NTYP1 rad. Added Gailitis average (via QNMAX of MESH1 as
per stgf/damp). Allow re-coupling of unphysical S-matrix produced by
stgf/damp. Gives good agreement with BP for dn=0 DR.
07/04/00 stgf (3.1), stgfdamp (3.1), stgbf0damp (3.1)
*****************************************************
Write complex K- and or S- matrix for stgicfdamp.
07/04/00 stgicf (3.2)
*********************
Minor, for consistancy with stgicfdamp.
07/04/00 stgjk (2.22) minor
***************************
Output TCCDW.DAT TCCs with more d.ps for closer unitarity test in stgicfdamp.
04/04/00 stgicf (3.2), stgicfdamp (3.4)
***************************************
Allow IMODE=1 to use a subset of the coarse enegry mesh.
01/04/00 stgicfdamp (3.4)
*************************
Added determination of DR collision strengths.
01/04/00 stgicf (3.2), stgicfdamp (3.3)
***************************************
Bug-fixes for ITCC=0 operation and for IMODE=1 operation.
Note, ITCC=1 and IMODE=0,-1 unaffected - these denote our main operation.
31/03/00 stgicf (3.2), stgicfdamp (3.3)
***************************************
Added elastic transitions - required for DR via unitarity.
29/03/00 stgicfdamp (3.2)
*************************
Added inner electron (NTYP1) damping.
29/03/00 stgf (3.1), stgfdamp (3.1), stgbf0damp (3.1) Minor
***********************************************************
Tweak evaluation of radiative widths.
24/03/00 stg2r (2.40)
*********************
Switch dimension in /CUPPLE/ MZNC1 to MZNC2.
17/03/00
Wrote JRELOP(1) to NX2.DAT for NX code to flag use of mass-velocity term.
18/03/00 stg2nx (1.9)
*********************
Added mass-velocity term, for "high-l" continuation run only viz. only
continuum-continuum contribution (and no bound contribution to it).
13/03/00 stgicf (3.1), stgicfdamp (3.1)
***************************************
Minor re-ordering (both). File name mods to stgicfdamp.
11/03/00 stgicfdamp (3.1)
*************************
First release of damped version of stgicf (based on 3.1). Reads and
transforms damped (i.e. complex) unphysical K-matrix produced by
stgfdamp, found on zkmtls.dat. The damped K-matrix includes all NTYP2
damping, but not NTYP1. NTYP1 damping will be via a complex effective
quantum number in stgicfdamp - NOT YET CODED.
In general, the non-exchange K-matrices will be real still and could
be produced by stgf on kmtls.dat. To allow for this, INOEXCH.GT.0
looks on zkmtls.dat while INOEXCH.LT.0 looks on kmtls.dat.
11/03/00 stgicf (3.1)
*********************
Synchronization with stgicfdamp.
11/03/00 stgfdamp (3.1), stgbf0damp (3.1)
*****************************************
Write complex K-matrix to zkmtls.dat (when IPRKM.EQ.4) for stgicfdamp.
****************************************************************************
CODE RENAMING: stgic (2.7) becomes stgicf (3.1)
****************************************************************************
11/03/00 stgic (2.7)
********************
Re-organization and tidy-up ready for initial development of damped version.
08/03/00 stgf (3.1), stgfdamp (3.1), stgbf0damp (3.1) Minor
***********************************************************
Made the use of Coulomb-Bethe omegas the default for dipole top-up.
Added alternative (smooth) switching from non-degenerate to degenerate
case in non-dipole top-up, as per stgic.
04/03/00 stgbf0damp (3.1) Minor
*******************************
Changed total photoabsorption file name to XPATOT to emphasise that it is
not the same as the total photoionization when damping is present.
XPIPAR contains partial photoionization to each electron target continuum
still and XPISUM still contains the sum over the partials which may, or
may not, be equal to the total photoionization. XPISUM can contain more
partials than are in XPIPAR.
02/03/00 stgic (2.6)
********************
Additional test for determination of allowed multipoles for non-dipole top-up.
Fixed TOP3 not being initialized if no inelastic open channels present at the
first energy.
26/02/00
Alternative switch to (non-dipole) degenerate limit - when energy ratio
equals [Jmax/(Jmax+1)]**(2*lambda-1). Detailed logging of top-up fractions.
22/02/00 stgbf0damp (3.1)
*************************
Allow NTYP2I=0 when IPHOTO.NE.0, i.e. case of photoionization/recombination
with NTYP1 damping switched-on (NODAMP=1 switches it all off).
22/02/00 stg2r (2.39)
*********************
Mods to NOICC operation.
19/02/00
KAB's elimination of weakly mixing CI terms.
18/02/00 stg2r (2.38)
*********************
Connor's NOICC.GT.0 option to calculate the no. of IC channels from an LS run.
17/02/00 stgbf (2.15)
*********************
Initialize some switches for IQDT.GT.0 and IPRINT.GE.0 combination.
15/02/00 stg1r (2.34), stg2r(2.37)
**********************************
Do no write or overwrite NX*.DAT files when spin-orbit operator is on viz.
RELOP='YES', 'TCC' or IRELOP(3)=1.
15/02/00 stgbf (2.15) minor
*********************************
Minor mods for Cray.
15/02/00 stgf (3.1), stgfdamp (3.1), stgbf0damp (3.1)
*****************************************************
Minor mods for Cray.
12/02/00
Tweak CORINT for efficiency at high-L (MQDT perturbations).
14/02/00 stgb (2.16), stgbf (2.15) minor
**********************************
Initialize a couple of variables that were not being initialized.
12/02/00 stg3nx (1.10)
**********************
BUFFER 1 in RDRINT needed a test re-instated (no error would have resulted
because the code would have halted execution because of it).
11/02/00 stgic (2.6)
********************
Determine multipole of transition. Then use this to enable non-dipole
top-up to switch over to the denerate energy case.
09/02/00 stgic (2.5)
********************
Added improved CBe routines. Also, Don's JMNTWO specification for an
exchange (R-matrix) run.
****************************************************************************
CODE RENAMING: stgfdamp (2.14) becomes stgbf0damp (3.1)
CODE RENAMING: stgfrad (2.19) becomes stgfdamp (3.1)
(stgf (2.33) just becomes (3.1))
****************************************************************************
05/02/00 stgf (2.33), stgfrad (2.19), stgfdamp (2.14)
*****************************************************
Improved testing for breadown of dipole Coulomb integrals used by CBe
top-up. In addition, the values of all failures should infact be
vanishingly small and so they are now set to zero rather than switching
back to using CC partial omegas.
01/02/00 stgfrad (2.19), stgfdamp (2.14) Minor
****************************************
Allow execution to continue with warning if D file(s) missing.
29/01/00 stg1r (2.34)
*********************
1/Switch-off diagonal continuum-continuum 1-body relativistic corrections as
they have an unphysically large effect on the elastic K-matrix elements
(possibly because a non-relativistic Buttle correction is still being used).
2/Check for breakdown of exchange integrals earlier.
3/Stop if code tries to switch Buttle correction method between l's (stgf
can't handle this). Added MJS variable (=0) to force quadratic if required
(e.g. when BSTO.eq.ZERO).
25/01/00 stg2nx (1.8), stg3nx (1.10)
************************************
Fixed bugs to BUFFER 1 in GENINT (stg2nx) and LOC1CC, RDRINT (stg3nx).
21/12/99 stgf (2.33), stgfrad (2.19), stgfdamp (2.14)
*****************************************************
NOMWRT .LT. 0 writes -NOMWRT collision strengths to OMEGA columnwise
(upper half matrix) excluding zeroes. NOMWRT .GT. 0 (default) remains
unchanged: NOMWRT collision strengths row-wise, including zeroes.
The latter is for consistancy with our current post-processing programs
while the former is IP "standard".
20/12/99 stgjk (2.22)
*********************
Dimension of DUMMY in COMMON/SCRACH/ increased to 3*MXJC.
20/12/99 stgfdamp (2.13)
************************
Total photoabsorption now switched-on by IPHOTO.GE.1000 and the cross section
output to XPITOT. Number of electron continua is now IPHOTO modulo 1000.
The partials are in XPIPAR still but the sum of the partials to IPHOTO (modulo
1000) is now in XPISUM. If IPHOTO (modulo 1000) .GT. MZPHT then MZPHT partials
are resolved but the sum is over IPHOTO (modulo 1000) still. Before, the
sum was obtained by summing over the resolved partials at the end of the run.
Now it is incremented as each partial is obtained and is not limited by MZPHT.
18/12/99
MQDT treatment of damped photoionization now on same general footing as
damped excitation viz .the unphysical dipole matrices can use K- or S-mx
normalization before being closed-off (in the former case the change to
S-mx normalization is then via the physical S-mx) and can be interpolated
in a single or double pass.
18/12/99 stgfrad (2.18)
***********************
Auger damping added as in stgfdamp.
16/12/99 stgfdamp (2.12)
************************
Photoabsorption generalized as below. crosspi3 now XPIABS. Still switched-on
by IAUGER.NE>0 for now. This may change.
14/12/99
Damped photoionization from first NPISYM (STGB) symmetries (see below) and
first NPIEB (STGB) energies of said symmetries. Output is now to XPITOT (was
crosspi) for photoionization resolved by initial state but summed over all
final (electron continuum) states and XPIPAR (was crosspi2) for the
partial final state resolved. Note, IPHOTO.LT.0 now switches on partial
recombination and IPHOTO.GT.0 partial photoionization to at most IPHOTO
electron cintinua. IPHOTO=0 still switches off access to SR.PHOTO.
14/12/99 stgb (2.16)
********************
Case IORDER=1 (default) process symmetries according to order listed by IOPT2
rather than order in H.DAT (IORDER=0). The latter is (slightly) faster but
the former is more convenient for the damped codes, especially photoionization.
11/12/99 stgic (2.4)
********************
Write an OMEGANOTOP file (approx OMEGA without top-up) when top-up is on,
so as to assist assessment of accuracy of effective collision strengths.
06/12/99
Re-organized non-dipole top-up - use either energy or omega ratio (default).
Tag omegas negative if top-up exceeds TOPR factor (1.3 = 30% etc), default
is large i.e. off. Note, if top-up is applied to a NX run then this test
of the top-up contribution is relative to the NX sum only. If a substantial
contribution comes from lower J of the exchange run then this warning is
unnnecessarily cautious.
10/12/99 stgbf (2.15)
*********************
Allow user to over-ride DELTA, point at which velocity dropped and length-to-
acceleration conversion no longer carried-out due to possible cancellation errors.
08/12/99
Evaluation of beta failed in DSTOR if there was more than one final symmtery.
This bug was introduced when it was extended to handle BP betas.
03/12/99 stgic (2.3)
********************
Remove all but trivial memory dependence on the number of Jp symmetries.
01/12/99 stgic (2.2)
********************
Only read LS K-matrices (per energy) for those LSp required by the current
Jp symmetry.
29/11/99 stgic (2.1)
********************
Inter-changed energy and Jp symmetry loop. Identical results to V1.10
except that non-dipole top-up uses the ratio of final to incident
energy rather than successive partial omegas in the geometric sum.
30/11/99 stgic (1.10)
********************
Minor.
27/11/99 stgfdamp (2.11)
************************
Partial photorecombination from NDRMET metastables.
27/11/99 stgf (2.32), stgfrad (2.18), stgfdamp (2.11)
*****************************************************
Bug in SR.OMEG - overflow to disk - fixed. (You would have got zeroes.)
19/11/99 stgbb (2.6)
********************
Extended to handle neutral atoms (+electron).
Also added stgb fix of 6/97 for deeply-closed channels.
19/11/99 stgbf (2.15)
*********************
Extended to handle neutral atoms (+electron).
18/11/99 stgfdamp (2.10)
************************
Photoionization to resolved final (target) states now being accumulated
over multiple final symmetries.
15/11/99 prebf (2.8)
********************
Modified to handle neutral atoms (trivial).
13/11/99 stgb (2.15)
********************
Extended to handle neutral atoms (+electron).
12/11/99 stgf (2.32), stgfrad (2.18), stgfdamp (2.10) Minor
*****************************************************
Re-ordered some do-loops for efficiency.
11/11/99 stgf (2.31) Minor
********************
Made elastic output default for neutrals.
25/10/99 stg1nx (1.9)
*********************
Fix bug in double buffering for large CI/small buffer.
25/10/99 stg3r (2.19), stg3nx (1.9)
***********************************
Added ISORT to STG3A NAMELIST. For ISORT=1, observed energies are expected
in calculated energy order and not the order they were stored in stg2/stgjk.
I.e. terms/levels can be read-in in any order in stg2/stgjk or resorted in stg2.
As always, the first term listed in stg2 should be the ground one.
11/09/99 stgb (2.14), stgbf (2.14), stgbb (2.6) Minor
***********************************************
Minor mods to avoid compiler warnings.
11/09/99 stgf (2.31), stgfrad (2.17), stgfdamp (2.9) Minor
****************************************************
Minor mods to avoid compiler warnings.
11/09/99 stg1nx (1.9), stg2nx (1.8), stg3nx (1.9) Minor
*************************************************
Minor mods to avoid compiler warnings.
11/09/99 stgic (1.10) Minor
*********************
Minor mods to avoid compiler warnings.
13/08/99
File name modified: adasexj.in -> adasexj.in.form to avoid overwriting.
07/09/99 stgf (2.31)
********************
Extended to handle neutral atoms.
11/08/88 stg2r (2.37)
*********************
stg2r failed for bound orbital angular momentum .ge. 8 (case of extreme
pseudo-state expansion.
07/08/99 stgb (2.14)
********************
KAB's latest mods to IDFORM/IDENTFY.
30/07/99 stgf (2.30), stgfrad (2.17), stgfdamp (2.9), stgic (1.10)
******************************************************************
Try to catch failures in FDIP and do not replace CC partial omegas by CBe
in those cases.
27/07/99 stgf (2.30), stgfrad (2.17), stgfdamp (2.9)
****************************************************
The previous change (14/07/99) to avoid overwriting the strength.dat
file also "avoided" writing the terminator to jbinls causing the NX stgic
run to fail - sorry.
15/07/99 stgic (1.10)
*********************
Don's "tidied-up" version, plus a further correction.
14/07/99 stgf (2.30), stgfrad (2.17), stgfdamp (2.9) Minor
***********************************************************
Don't overwrite strength.dat (when IPRKM.EQ.4) if no line strengths
have been calculated (i.e. if IBIGE.NE.1).
09/07/99 stgic (1.9)
********************
Don's correction of his first attempt at separating LRGLAM & IDIP operation.
08/07/99 stg3r (2.19), stglib (2.12) Minor
******************************************
Mods for g77 compiler.
26/06/99 stg1r (2.33)
*********************
Corrected spin-orbit bug affecting STO- operation only.
Removed ISMITN from non-namelisted input since free-formatted input
means that standard CPC files won't run. Can easily be added by user,
or use namelisted input instead.
24/06/99 stgf (2.30), stgfrad (2.17), stgfdamp (2.9)
****************************************************
Incorporated hybrid-MQDT approach for deeply-closed channels (nu<l) which
is accessed via IOMSW=-1. Uses theta for channels with FKNU<FNUHYB, subject
to FKNU<LLCH. Default is still IOMSW=1 - omit deeply closed channels.
22/06/99 stg2r (2.37)
*********************
Use STO- instead of S.S. to read L,2S+1,PARITY, instead of 2S+1,L,PARITY,
as per CPC version.
18/06/99 stgf (2.29), stgfrad (2.16), stgfdamp (2.8)
****************************************************
Added Don's option out outputting partial cross sections - see ISGPT,ITRMN,ITRMX.
17/06/99 stg1r (2.33)
*********************
Allow usual S.S. NAMELIST input along with STOs instead of AS/SS radial file.
Use STO- instead of S.S. and add NZ/NZED,NELC,MAXORB to NAMELIST STG1B.
If MAXORB.GT.0 then read nl pairs, order unimportant so as per STG2 easiest
even though, of course the STO data that follows is ordered 1s,2s,3s...2p,3p..
etc. The STO data is as per RECORD 9 of 1995 CPC write-up.
09/06/99 stgic (1.9)
********************
Don's bug fix for systems where target levels are half-integer J. Systems with
target levels of integer J are unaffected. Also, separated LRGLAM operation
from that of IDIP/IBIGE.
28/05/99 stgf (2.29)
********************
Added Tom's eigenphase sum as a subroutine, accessed when IPRINT.GT.0.
28/05/99 stg1r (2.33) (Minor)
*********************
Add ISMITN to (non-namelist) input option list for running analytic orbitals.
06/05/99 stg2r (2.37)
*********************
(Hopefully) solved long-standing bug for photoionization when: RK ON FILE.
06/05/99 stg3r (2.18)
*********************
Minor.
22/04/99 stgf (2.29), stgfrad (2.16), stgfdamp (2.8) (Minor)
****************************************************
Removed a rather unnecessary test which had the side effect of unnecessarily
restricting the fine-energy mesh range subset of the coarse energy mesh range
allowable. Production runs use same energy range so no effect there. Also,
increased BP ISKP0 to J=15 (was L/2*J=15, now L/J=15). Was only a problem if
interpolating K/S for high J, which is not normally done since no resonances
are present there.
01/04/99 stgf (2.29), stgfrad (2.16), stgfdamp (2.8)
****************************************************
Incorporated alternative method to evaluate non-divergent perturbation
integrals in MQDT viz. use only non-divergent part of closed channel
s and c functions i.e. theta. This is an alternative to CORINT (the
default, INTPQ=0). Set INTPQ=1 to access the new method.
11/03/99 stgfrad (2.15), stgfdamp (2.7)
***************************************
Reduce memory requirements as in prebf.
10/03/99 prebf (2.7), stgbb (2.5)
*********************************
Reduce memory requirements as per KB/WE but with crucial fixes to errors
found in the original source.
06/03/99 stgb (2.14)
********************
Minor.
23/02/99 stg1r (2.33)
*********************
Evaluation of the Buttle correction to the orbital was incorrect for
l-continuum .gt. l-atomic in RMPS mode (ISMITN=1). This would be of little
consequence, since we never make use of it, except for the fact that it
causes a crash on the T3E as it tries to read-access a non-existent storage
location.
06/02/99 stgbf (2.14)
*********************
Mods for f90 - may not compile else.
11/12/98 stgic (1.9)
********************
A note on using observed energies:
If itcc.eq.0: if observed term energies are used in stg3 then nothing further
is needed. If not, observed term energies can replace the calculated in file
term.dat, but **not** if it changes the energy order of a term.
If itcc.ne.0: if observed term energies are used in stg3 then they should be
entered into the tccdw.dat file as well, since it will still contain the
calculated term energies, and **must** be entered if the energy order of a
term is changed by stg3. Observed level energies can also be entered into
tccdw.dat to replace the calculated ones, the energy order of a level can
change.
05/12/98 stgjk (2.21), stg3r (2.18), stgf(2.28)
***********************************************
Minor, for Cray.
02/12/98 stg1nx (1.9), stg2nx (1.8), stg3nx (1.9)
*************************************************
Minor, but note the modified PARAM files.
02/12/98 stg1r (2.33), stg2r (2.37)
***********************************
Minor.
02/12/98 stgic (1.8)
********************
Minor, increased dimensions.
30/11/98 stgf (2.28)
********************
Bug-fix to stgf that affected 2.27 and earlier versions of 2.28 (stgfrad
and stgfdamp are unaffected). Affected non-MQDT operation only, but both
PERT='YES' and PERT='NO', over a limited energy range where a closed channel
outer turning point moved-out beyond RTWO. The variable ONE was not defined!
26/11/98 stgf (2.28), stgfrad (2.14), stgfdamp (2.6)
****************************************************
No longer interpolates when then is an imbalance between channels omitted
at the two interpolation energies. Instead use one or the other (the one
with the deepest channel omitted, that is still kept by the other).
Also, tightened up convergence test of Coulomb functions from series.
24/11/98 stg3r (2.18)
*********************
Added ESHIFT, energy in Rydbergs by which all target states are shifted. The
net effect is to shift (only) correlation resonances by -ESHIFT. In &STG3B.
08/10/98 stgic (1.8)
********************
Bug-fix to phase error in determining the recoupled line strengths. Collision
strengths, apart from infinite energy, are unaffected.
27/08/98 stgf (2.27), stgfrad (2.13), stgfdamp (2.5)
****************************************************
Introduced a much more flexible choice for RTWOC, the value of r at which
CORINT is entered to evaluate the finite part of the MQDT open-closed and
closed-closed multipole integrals. v2.26 uses RTWO (IRAD=0) or RZERO (IRAD=1).
Both have their problems. If RINF, for a given channel, is "significantly"
larger than RZERO then the direct solution of the S.E. in CORINT breaks down
leading to large errors. If R2, for a given channel, is "significantly"
smaller than the global RTWO then the Bode's rule integration from RZERO to
RTWO gives large errors because it contains the divergent part of the integral.
v2.27 chooses RTWOC at each energy for each total symmetry based on an
examination of the inner and outer turning points of each closed channel
solution and, where necessary, each open channel solution as well.
Parallel changes have been made to stgfrad/stgfdamp.
Also, introduced codeing so that the c-function is always integrated towards
the divergence. Only when a closed-channel RINF exceeded RZERO was there a
problem, and then only if PERT='YES'. Note, both MQDT and non-MQDT operations
were affected.
Parallel changes have been made to stgfrad/stgfdamp.
20/08/98 stgf (2.26), stgfrad (2.12), stgfdamp (2.4)
****************************************************
The line strengths from stgf following a NX run were incorrect. This affected
the infinite energy point (inc. stgic). The problem arises because there are no
coupled channels connecting target states of different spin. SR.RAD failed
to take account of this and so thought that there should be rates between
target states of different spin. This would not be a problem as such except,
unfortunately, that the calculation of the line strengths is only completed
following a successful determination of ALL (as RAD thinks) radiative
transitions (otherwise there is always a "premature" return from RAD).
I have modified RAD to take account of the NX circumstances. I have also
changed it to fail safe so that any line strengths that it has determined will
always be correct.
Also, modified writes for IPRKM=4 so that stgic (1.8) does not need
nz, nelc and npw specified, and zeroes for low-E/high-L K-matrix that would
normally be skipped in stgf, and split kmtls.dat into two files.
Finally, fix for evaluation of c-function when RINF exceeds RZERO and the
integration is outwards - only affects high-L closed-channels i.e. the odd
energy point - see next release for full solution.
11/08/98 stg1nx (1.8)-minor, stg2nx (1.7)
*****************************************
Updates for Cray. stg1nx minor. stg2nx needed a SAVE statement in ROOT, could
have caused problems depending on compiler switch settings.
06/08/98 stgic (1.8)
********************
Introduces an imode=-1 option for production runs. Results are calculated in
a single pass interpolating from the stgf coarse mesh onto the user-specified
fine mesh with no passing files written or scratch-file I/O. (The two step
route, imode=0 followed by imode=1 is heavy on I/O and generates large passing
files in large production runs.)
27/07/98 stgic (1.7)
********************
Based on Don's (1.6) of 23/5/98, introduces Burgess dipole top-up at
LRGLAM (best set=twice max total J - 2). The top-up is made using CC partial
omegas. If, in addition, LCBE is set positive then Coulomb-Bethe partial omegas
are calculated at 2*J=LCBE and higher, the ratios to CC printed (if IPRTOP=1)
and the top-up is done with the CBe partial omegas.
18/07/98 stgf (2.26), stgfrad (2.12), stgfdamp (2.4)
****************************************************
As you may be aware, the JBIN file can get very large in stgf production runs
(e.g. Co2+). It has always been possible to run stgf in a single pass mode
via the IEQ variable. This controlled how often the unphysical K/S-matrix was
updated on the specified fine mesh but it wasn't interpolated and this
led to small steps in the cross section at the energies where the K/S-matrix
was updated (trivially, JBIN was still written). I have now modified stgf/rad/damp
so that the K/S-matrix is interpolated on a single pass run, and no JBIN file
is written. This gives results identical to the two run strategy. One
should still use the (default) two run strategy to determine a suitable
coarse and fine mesh, using a representative partial wave and energy range, but
switch to a single pass for production runs to avoid generating JBIN. (To do
this, specify the fine mesh as usual and then set IEQ equal to the number of
coarse energy points that you want. These are distributed linearly across the
energy range spanned by the fine mesh.) Of course, the drawback to not writing
JBIN is obvious: if you decide you want a narrower mesh then the whole
calculation has to be repeated!
15/07/98 stgf (2.26), stgfrad (2.12), stgfdamp (2.4)
****************************************************
Re-organised BP Burgess dipole top-up. Before (see 11/07/98) the top-up used
the maximum possible channel small l (LITLAM) for a given total 2*J given
by LRGLAM. This meant that several higher J partials needed to be cycled
thru to get the K/J top-up i.e. LRGLAM<<MAXLT. Now top-up uses the smallest
possible small l at LRGLAM. At most, one higher J can contribute to the
K/J top-up and its neglect is a negligible effect in practice. (You can
set LRGLAM=MAXLT-2 to see this.) This doesn't change the maximum J that
you need since you should still top-up at approx the same small l, but is
less likely to cause error. Also, before and now, the quadrupole top-up
takes place at LRGLAM so one gets the maximum benefit now.
11/07/98 stgf (2.26), stgfrad (2.12), stgfdamp (2.4)
****************************************************
Extended the Burgess sum-rule dipole top-up (in small l) to Breit-Pauli.
No top-up in K/J (analagous to the LS top-up in large L) since this is small,
but tedious to code for. If you want to top-up in K/J set LRGLAM.LT.LRGL2.
The top-up is applied at LRGLAM and the higher partial waves are cycled
through and any relevant K/J contributions are extracted. The maximum BP LRGL2
than can contribute is LRGLAM+2*MAX(LAT)-2 (i.e. LRGLAM+2*MIN(LAT)+2 for a given
dipole transition) where LRGL2,LRGLAM,LAT contain twice the relevant J-values.
MAXLT interacts with LRGLAM as you would expect (both in LS and BP mode).
The sum rule is a difference between partial omegas and hence can be sensitive
to cancellation. In BP especially, it might be more accurate to use the Coulomb
Bethe (dipole) matrix elements to evaluate the top-up, via the long-present but
little-used LCBE parameter.
07/07/98 stgfrad (2.12), stgfdamp (2.4)
***************************************
Dario spotted an inconsistancy between a dimension in a calling program
(ZRMAT) and in a subprogram (VERT) - VERT was called with the wrong
dimension info. Actually, the variable (B1) should have been originally
dimensioned as indicated by the call to VERT i.e. (MZDEC,MZDEC) and not
(MZCHF,MZCHF), which at best is unnecessarily large and at worst causes
a misaligment in memory, but that is very machine/compiler dependent.
26/06/98 stgjk (2.20)
*********************
Incorporated the ISKIP feature of stg2r into stgjk. Works exactly the
same, but tag parity as negative to read the subsequent iskip value
(use 2 for even parity instead of 0 then).
25/06/98 stgf (2.26), stgfrad (2.12), stgfdamp (2.4) (minor)
************************************************************
Updates to output to kmatls.dat for stgic operation. Minor bug-fix.
23/05/98 stgic (1.6)
********************
Extended to handle stgf K-matrices resultant from inner region run
with the NX R-matrix codes.
15/05/98 stgb (2.13) (minor)
****************************
AC->1.E-5 for consistancy with stgf,stgbf etc.
05/05/98 stgf (2.26)
********************
Revert to old MQDT RZERO to infinity integrals when IRAD.GT.0 (writing data
for stgbf) since RTWO is set artificially large and RZERO to RTWO diverges.
(This is o.k. since stgbf is a low-L problem.) See 2.27 for full solution.
02/05/98 stgjk (2.20)
*********************
Incorporated Tom's shifting of target term energies so as to get
correct spin-orbit mixing for TCCs. Similar to stg3r use, e.g. use NSHIFT
to switch-on.
30/04/98 jajom (1.2)
********************
Minor mods by Don to improve running. This code will be retired shortly,
it being superceded by stgic.
30/04/98 stglib (2.11)
**********************
A bug in subroutine ORDTRI of the NJGRAF branch has been fixed by
Stephen Bailey of QUB.
29/04/98 stgf (2.25), stgfrad (2.11), stgfdamp (2.3) (minor)
************************************************************
Fix minor discontinuity in interpolation when channel dropped from
one energy but not the other. Interpolated K-mx smooth now but S-mx not
so good. In general, interpolating K-matrix is more reliable in "extreme"
cases for MQDT. (Although S-mx is faster....)
22/04/98 stg1r (2.32) minor
***************************
Minor (transparent) bug-fix.
21/04/98 prebf (2.6) stgbf (2.13)
*********************************
Allow unequal IPERT from stgb and stgf.
18/04/98 stgf (2.25), stgfrad (2.11), stgfdamp (2.3)
****************************************************
Reworked perturbing potentials for both MQDT and non-MQDT operation.
The MQDT perturbing integrals (as originally coded by Tom for the low-L
radiation damped problem) failed spectacularly at high-L (>~15).
The non-MQDT perturbing integrals were occasionally in error over a
narrow energy range just below thershold due to an instability in the
outward integration. Later note: I was being optimistic here, see later
versions, in particular 2.27.
31/03/98 stgf (2.24), stgfrad (2.10), stgfdamp (2.2)
****************************************************
If you use MQDT via the K-matrix (IQDT=2) with PERT='YES' for high-L/J
then re-running on the fine energy mesh can give garbage because the
codes attempt to perturb the T-matrix but have none of the required info.
This mainly affects stgf since the damped codes, by default, perturb the
K-matrix so as to ensure a unitary (undamped) S-matrix. The stgf default
was left as its historical perturbation of the T-matrix. All codes now
perturb the K-matrix when PERT='YES' is used. This was a one line change.
12/03/98 stgf (2.24), stgfrad (2.10), stgfdamp (2.2)
****************************************************
There is a bug in all versions of stgf that we have ever
had/used (!). The routine WSQ (and hence WSQ2) incorrectly evaluates
the Racah coefficient for the case of A=B and C=D. This could affect
top-up and radiation damping. It occurs for la=la' and l=l'. It is
possible that we have never encountered this transition for damping, or
that the radiative rate may already have been decoupled from an earlier
symmetry. LS data is unaffected.
04/03/98 stgbf (2.13)
*********************
Phase factor in LS BETA evaluation - Lan Vo Ky.
28/02/98 stgbf (2.12)
*********************
Tom's bug-fix in BETA for zero cross sections.
24/02/98 stgic (1.5)
********************
First "public" release of Don's program to read unphysical k-matrices in LS
coupling generated by stgf, determine the unphysical k-matrices in pure jK
coupling or intermediate coupling and finally use these to determine and print
the collision strengths between individual levels. See the comments at the
beginning of the code for the simple input and the document stgicadas.txt
for help on usage.
07/02/98 stgb (2.12), stgbb (2.4)
*********************************
Minor adjustments for Cray operation.
07/02/98 stg1nx (1.7), stg2nx (1.6), stg3nx (1.8)
*************************************************
Minor adjustments for Cray operation.
05/02/98 stgf (2.24), stgfrad (2.10), stgfdamp (2.2)
****************************************************
Bug-fixes to top-up and MQDT operation that occur in certain circumstances.
29/01/98 stg3r (2.17)
*********************
Replaced JACORD by DIAG in HAMNEW and HAMNEW2.
Made a few minor changes to avoid (Cray) compiler warnings.
28/01/98 stg1r (2.31), stg2r (2.36), stgjk (2.19)
*************************************************
Made TCC generation more efficient for large LS cases, also use PARAM.TCC.
Added some extra dimension tests. Made a few minor changes to avoid
(Cray) compiler warnings.
28/01/98 stglib (2.10)
**********************
Minor.
20/01/98 stgdwic (1.5)
**********************
Dario's bug-fix to limit J's to those for which ALL LS partials exist.
22/12/97 stgfdamp (2.1)
***********************
Added SR.PHOTO to stgfrad to make new stgfdamp code for further development
of damped partial photorecombination/ionization etc. Unlike stgbf, the
(unphysical) dipole matrices are NOT interpolated so interpolation in MQDT
mode is not allowed for now (it wasn't in the old stgfrad either).
16/12/97 stgfrad (2.9)
**********************
This is a major revision of stgfrad. Infact, a new version derived from the new
stgf (2.23) so much cleaner coding and easier to follow. This also means that
the MQDT is an improvement over the old stgfrad (2.8) since it interpolates
either the unphysical S-matrix (IQDT=1) or the unphysical K-matrix (IQDT=2),
unlike the old one which just updated the (K-matrix) "periodically".
IQDT=100 or 200 parallels IQDT=1 or 2 but the JBIN is never read, only written.
Optionally (IOMSW=0), "deeply" closed channels (n<l) are not dropped.......
It has been tested-out against 2.8. Differences seen are due
to round-off or a more accurate numerical treatment by 2.9 over 2.8.
The stgfrad v2.9 and onwards will be pure radiation damping codes.
The Auger damping and resolved (damped) photoionization/recombination will
appear in a "new" code - stgfdamp. stgfdamp is the code under which all new
development will be carried-out.
16/12/97 stgf (2.23)
********************
Parallel changes to stgfrad 2.9 i.e. K-matrix interpolation etc.
03/12/97 prebf (2.5), stgbf (2.10) (~Minor)
**********************************
For each symmetry, if try to start from an excited STGB state (M11.GT.1) then
wrong number of records were skipped, resulting in garbage presumably. Default
usage is M11.EQ.1 so not a problem.
07/11/97 stgbf (2.9)
********************
Added preconvolution capability to OP production (IPRINT=-1) routines OPVMAT &
OPDOUT. No changes to OPSTRA, it already had MQDT capability.
14/10/97 stgbf (2.8)
********************
This is a major new release of stgbf which makes use of the recent developments
to stgf (2.21, but 2.22 is needed). If stgf has been run in MQDT mode (IQDT=1)
then stgbf runs in MQDT mode as well. Since only the original coarse energy
mesh is fed from stgf, information must be supplied as to the new fine energy
mesh to be used by stgbf. Currently, this is via the NAMELIST STGBF variables
MXE, E0 and EINCR i.e. the same parameters that define the linear mesh of stgf.
A subroutine mesh uses this info to tabulate the new mesh. Alternative meshes
could be added in the future as linearity is not assumed. The unphysical
S and D matrices are interpolated at the new physical energies. This is handled
internally, no second run is necessary - unlike stgf. In addition, if EWIDTH
is set .GE. 0.0 then the TOTAL photoionization cross section is preconvoluted
with a Lorentzian of said width (all energies are z-scaled). EWIDTH should
be set .GE. EINCR. A good set of values to start with is EINCR=1.E-4 in stgf
together with EINCR=1.E-5 and EWIDTH=1.E-5 in stgbf. Note, photorecombination
data is meaningless above the first excited state.
Currently, no beta asymmetry parameter data can be calculated in MQDT mode.
26/09/97 stgf (2.22)
********************
Changed MQDT default updating of S-matrix to be at every energy, to be safe.
Introduced (much) faster matrix inversion (real & complex) and diagonalization
(complex) routines for non-BLAS operation.
Also, modified output required by stgbf (2.8) and later.
19/09/97 prebf (2.4) Minor
********************
Eliminate false warnings occuring in BP mode.
11/09/97 stgf (2.21)
********************
Please find attached a new version of stgf (2.21) for your delectation.
The main new feature (over 2.20) is an MQDT option to treat ALL closed
channels as open. As a by-product, type-I radiation damping can be included
to gave damped excitation and DR. The controlling option is IQDT. IQDT=0 runs
stgf in non-MQDT mode, as before. IQDT=1 activates MQDT mode. The variable
IEQ controls how often the unphysical S-matrix is calculated, namely at
IEQ equally spaced energies across the energy range defined by your mesh -
the default is 100. The unphysical S-matrix is automatically written to the
file JBIN. Subsequent runs attempt to read JBIN and interpolate the S-matrix
at the scattering energies requested. Delete JBIN to allow a new one to be
written. Note, the "first" run that generates the S-matrix on a coarse mesh
(via IEQ) will still calculate collision strengths on whatever (fine) energy
mesh you give it in the normal fashion, using the latest S-matrix to hand. This
can give a small step-like nature to the calculated collision strengths.
Second and subsequent runs which interpolate the S-matrix will be "smoother".
The point of MQDT of course is that all of the time goes on evaluating the
unphysical S-matrix at a few energies (since it is slowly varying) while
the physical S-matrix (and hence collision strengths) can be evaluated
at many more energies relatively cheaply. The Gailitis average works by
calling MQDT to contract the S-matrix first and then a call to SQDT to
average over channels attached to the lowest target state.
Long-range dipole perturbing potentials are included when PERT='YES'.
Please test it out (both IQDT=0 qnd IQDT=1) and let me know of any problems
or suggestions for improvement. IQDT=0 and PERT='NO' should only give
very small differences from existing versons of stgf, due to re-ordering
of code. If PERT='YES' (and IQDT=0) larger differences may be noted from
earlier versions. This due to the fact that all versions prior to (2.21) did
NOT symmetrize the K-matrix despite claiming to do so. Note, you really
should use PERT='YES' or 'NO' to switch the perturbing potentials on and off.
IPERT takes on a number of possible values and the code chooses the most
appropriate one given the rest of your input options. If you set IPERT
explicitly without appreciating the comments concerning its use then you
might not get the desired result.
If you're interesting in tracking the code then, as always, everything
happens in sr.react - or rather everything is called by sr.react now
since I have modularized it, if that's the right word. See sr.mqdt and
sr.sqdt in particular. The comments in program main also give more details
on the new options and some I haven't mentioned (e.g. NDRMET) which are
straightforward. Updated makefiles and PARAM files are attached as well.
03/09/97 stg1nx(1.6),stg2nx(1.5),stg3nx(1.7)
********************************************
Coded, and added NPTYIN option to NAMELIST STGNX, case of running single
(N+1)-electron parity. Useful for when initial target of interest is an
S-state (factor 2 faster). Only read in CONTinuation mode (as we always go).
NPTYIN=-1 (default) gives both parities as before.
NPTYIN=0 gives even parity for even L and odd parity for odd L (e.g. H-like).
NPTYIN=1 gives odd parity for even L and even parity for odd L (e.g. N-like).
22/08/97 stgfrad (2.8)
**********************
Important bug-fix due to Tom concerning initialization of DDEC in SR.READB.
20/08/97 stgf (2.20)
********************
IQDT=-1 evaluates detailed SQDT, damped (Type-I) & undamped, excitation and DR.
If NU.GT.QNMAX (any mesh) then Gailitis average applied to both excitation and
DR, damped & undamped (IQDT=0 or -1). SQDT means only channels associated
with lowest closed term treated by QDT. This will be relaxed in the next
release.
26/07/97 stgf (2.19)
********************
ALPHA and REACT incorporate I.P. efficiencies, along with some of my own.
18/07/97 stg2r (2.35) Minor
*********************
Fixed inconsistency between initialization and namelist for LRNGDW.
Increased an internal DW dimension.
17/07/97 jajom (1.1)
********************
The Iron Project version of JAJOM has been installed, under ../rmatrix/misc.
It has been customized for our OMEGA output, PARAM etc.
NO user input is required. It reads JBIN and JJDAT from a previous (LS)
run of stgf (see 2.18 below). If no TCC.DAT file is present it just
recouples. If TCC.DAT is present from a short stg1r/stg2r/stgjk run
(e.g. RELOP='TCC') then jajom term couples the LS T-matrix.
Hint: delete RECUPH.DAT before making an LS run as stg3r will pick-up
on it - stgf checks for LS on a jajom run.
***WARNING*** I haven't looked at JAJOM with MQDT. So do not use/generate
a JBIN file via IQDT=1 in stgf. Use the original IOPT1=10 or 11 method.
(Also, currently, stgf then writes a slightly different format for JBIN for
JAJOM as opposed to that required for (MQDT) stgf/stgbf use.)
17/07/97 stgf (2.18)
********************
Write T-matrix (actually S=1-T) to file (JBIN) in JAJOM format and basic
data for JAJOM (on file JJDAT) when IOPT1.GE.10. Based on KAB's SR.OUTJJ.
The LS OMEGA still written unless IOPT1=11. IRAD is reset to zero, i.e. no
radiative data is allowed - I guess this could be relaxed......
10/07/97 stgf (2.17)
********************
Store OMEGA internally, with overflow to disk, to reduce I/O. Specify
internal memory in megawords with PARAMETER MZMEG. Based on KAB's SR.OMEG.
About 20% speed-up is the best I've seen. Set MZMEG=0 to write all OMEGA
to disk as before.
09/07/97 stgbf (2.7)
********************
Obvious bug on UNIT numbers being written for BETA asymmetry parameter fixed.
Pointed out by Tom.
07/07/97 stgfrad (2.7)
**********************
Re-calculate radiative loss at each energy testing to see if final state
is stable against autoionization i.e. contributes to DR, and adjust as
necessary. Also, NODAMP option can switch-off damping inside the box, really
a check against stgbf.
05/07/97 stgb (2.11) IP Update
********************
Incorporated Iron Project update, based-on OP 5.2, rather than OP 5.1
which was the last IP stgb (if you follow). Actually nothing of direct
interest. OP 5.2 has a more detailed effort to identify (i.e. label) the
bound states found. The actual IP update was to convert an OP ID file
into an E file, whatever they are! Just keeps things up to date in case
changes of interest to us are made at a later date.
30/06/97 stgg (1.19)
********************
AGAIN fixed the projection bug, must have saved the wrong file (1.18)
because the change wasn't present!
Added IUP option to generate de-excitation cross sections (for Alfred's
planned experiment). IUP=1 (default) upwards, IUP=0 downwards - energy
scale is for the initial state in each case i.e. switched between up and down.
Fixed minor bug when processing excitation to pseudo-state data i.e.
direct ionization, original code was set-up for excitation-autoionization
and did some setting up to enable a Lotz direct estimate to be added. This
conflicted with projection when convoluted with a Gausssian.
16/06/97 stgb (2.10)
********************
Zeroed-out radial functions (written to file for stgbf) for deeply closed channels
where they are vanishingly small and move in from RTWO to generate non-zero functions
for stgb use. stgbf needs to do likewise.
04/06/97 stgg (1.18)
********************
Fixed bug in projection of pseudo-states for ionization.
03/06/97 stgjk (2.18) IMPORTANT
*********************
Due to the problems with the SUN f77 4.0 compiler, I hadn't been
updating stgjk. One of the PARAMETER statements defining the length
of a common block had gotten out of sync with the corresponding value
in stglib. You MUST use stgjk(2.17) with stglib(2.8) or, preferably,
stgjk(2.18) with stglib(2.9) - see the first line of each code to check
on the version number, if it's not there then Mitch must have deleted it
again!
03/06/97 stg2r (2.34) Minor
*********************
Added IBUG3 and IBUG4 to NAMELIST STG2A.
31/05/97 stgbf (2.6)
********************
Added option to select length (IGAUGE=0) or velocity gauge (IGAUGE=1,
default) results for photorecombination. OMEGDR changed to OMEGPR!
Recoded sr.numfc and sr.numb for deeply closed channels (zero-out),
this should eliminate problems that occurred there.
23/05/97 stg1r (2.30) minor
*********************
Activated a warning test that was incorrect and hence inactive.
21/05/97 stgbf (2.5)
********************
Have coded for photorecombination via detailed balance of partial
photoionization (undamped!). It is up to the user to ensure all
possible initial photoionization states have been included so as
to get photorecombination summed over all final states for the
energy range of interest. Only runs when IPRINT=-2 selected.
Output is to a standard OMEGDR file which can be copied to OMEGA
and processed as you would an excitation collision strength.
Note, you need a new stgf as well.
Also made (BP) bug-fix (omitted by Werner) and minor Belfast fix.
21/05/97 stgf (2.15)
********************
Added output of atomic weights (ISAT(I),LAT(I),I=1,NAST) to F00 for
output to OMEGDR (via stgbf) - see above.
Reset all FS,FC=0.0 when vanishingly small (used by stgbf else).
13/05/97 stg2r (2.33)
*********************
Added my old Opacity Project code skip option. If target spin is -(2S+1)
then READ(*,*)LSKP where LSKP is the number of terms of the SAME
symmetry that must be skipped to get to the energetic one that is
required for the CC expansion. All terms (unless NCUT.NE.0) are included
in the CI expansion. If NCUT (via IKIP) is used to eliminate terms
from the CI expansion as well then any skipping must be done with
reference to tbe the term list that results AFTER NCUT/IKIP has
been applied. In summary, NCUT/IKIP eliminates terms (CSF's) from
the CI list (and hence the mixing) and thus the (possible) CC list
while LSKP retains the term(s) in the CI but omits them from the CC
list - thanks to Don for this summary.
10/05/97 stgg (1.17)
********************
Added subroutine PROJECT which projects cross sections to pseudo-states
back onto physical states. Requires a MIXDW.DAT file and a suitable
OVRLAP file from AUTOSTRUCTURE.
23/04/97 stg1r (2.29)
*********************
Alternative (STG1B) specification of pseudo-orbitals. In addition to specifying
the range through NMIN,NMAX,LMIN,LMAX; if NMAX is .LE. 0 (default, as all are)
then the upper N-range of the pseudo orbitals NL is given by NPS(L), L=0,1,2..
for s-,p-,d-.. orbitals i.e. it is L-dependent. NMIN,LMIN,LMAX are required, as
before. Remark, one can of course combine the old ISMIT(K)=10*N+L to specify
the odd extra orbitals beyond the common NMAX BUT one must remember to use
K .GT. K0, where K0 is the last K indexed by NMIN,NMAX,LMIN,LMAX
i.e. K0=(NMAX-NMIN+1)*(LMAX-LMIN).
02/04/97 stgg (1.16) Minor
********************
Mods to use of DW omega file via OMEGDW or as OMEGA.
01/04/97 stgf (2.14)
********************
Every transition topped-up by lowest (positive) multipole. Very important
for pseudo-state ionization. Previously, only dipole & quadrupole
transitions were topped-up. Also, can now use LMX to specify largest
perturbing multipole included when IPERT=1/PERT='YES', default=2 - just the
original dipole and quadrupole. Only one multipole per channel is calculated,
thus an octupole (lambda=3) potential is not included if there already exists
a dipole potential.
15/03/97 stgdw (2.20)
*********************
Don's mods to output partial and total cross sections.
01/03/97 stg1r (2.28) Minor
*********************
Case IDWOUT=2 (DW only) allow MAXE to be set so as to enforce use
of a finer radial mesh. Note, for DW with pseudo-states it is likely to
be necessary to set the ISMITN and pseudo-orbital range still else the
(dummy) single basis orbital will start at high-n and there may be
problems in FINDER trying to find it. The initial energy estimate is geared
to low-n. Higher-n then depend on the previous e-energy.
24/02/97 stgfrad (2.6)
**********************
Includes Tom's PHOTO subroutine for partial recombination/ionization.
Dimensions now INCLUDEd via PARAM file as rest of codes.
22/02/97 stg1r (2.27) stg2r(2.32)
*********************************
Increase dimension parameter for MXRKBB. This is difficult to set.
It uses MXORB. However, the original MXORB wildly overestimates when
MZLR1<<MZNR1 (see 19/12/96 below) and inflates memory of this and
other arrays. Resetting MXORB to it the "correct" value when MZLR1<<MZNR1
leads to MXRKBB being about a factor 3 too small when right at dimension
limits. If MXRKBB is exceeded still, let me know. Note, you can do a
very fast dimension check (<1min) in stg1r using NBUG=7.
21/02/97 stgg (1.15)
********************
Can now read AUTOSTRUCTURE ayld file, for excitation-autoionization or
second stage of REDA.
20/02/97 stgdw (2.19)
*********************
Select partial wave using variable LSPI, e.g. = 30101 for triplet P odd.
31/01/97 stgf (2.13) Minor
********************
Correct error message output when dimension exceeded.
20/12/96 stgdw (2.18)
*********************
Added NCUT to exclude orbitals with N.GT.NCUT from continuum
distorted wave potential.
20/12/96 stgg (1.14)
********************
Treat R-matrix DR OMEGA file, including convolution of metastables,
correctly. Handle pseudo-state ionization (R-matrix and DW).
19/12/96 stgdwic (1.4)
**********************
Don's version of 7/17/96 (sic) plus bug-fix by Alan (DUM=0.5) on S6J1.
19/12/96 stg1r (2.26)
*********************
Reduce memory requirement case MZLR1<<MZNR1.
Also, MAXC no longer reset to at least MAXNHF since solved stg2 prob.
19/12/96 stg2r (2.31)
*********************
Reduce memory requirement case MZLR1<<MZNR1.
Solved problem of NRANG2.LT.MAXNHF (re-dimensioned NSTO).
So can now set MAXC/NRANG2=1 for DW or NX operation.
19/12/96 stglib (2.9)
********************
Reduce memory requirement case MZLR1<<MZNR1.
19/12/96 stg3r (2.16) Minor
********************
Minor mods, remove AMP's.
19/12/96 stgdw (2.17)
*********************
Mods for Don's latest stgdwic.
12/12/96 stgbb (2.3)
********************
No preprocessing required.
12/12/96 stgf (2.12)
********************
f90 mods: DOUBLE COMPLEX -> COMPLEX*16 and added AIMAG function
as IMAG intrinsic function not available.
Also, write S as well as Arad.
12/12/96 stgbf (2.4)
********************
No-preprocessing required. Dimensions INCLUDEd from PARAM file.
Terminate any SLp's that are input with -1-1-1 etc. or EOF.
Not 0 0 0, for consistancy with rest of outer region codes.
Note: IFLAGV=1 (default). Set IFLAGV=0 to switch off VMAT integral
that causes problems with deeply closed channels.
06/12/96 prebf (2.3)
********************
No-preprocessing required. Dimensions INCLUDEd from PARAM file.
Added Namelist PREBF (for IPRINT and IBUT). Terminate any SLp's
that are input with -1-1-1 etc. or EOF. Not 0 0 0, for consistancy
with rest of outer region codes.
05/12/96 stgg (1.13)
********************
Fixed DR from metastables.
30/11/96 stg1nx(1.5),stg2nx(1.4),stg3nx(1.6) Minor
********************************************
Mods for SUN f90 and DEC f77.
29/11/96 stg1r (2.29)
*********************
Returned default mesh to IPTS=23
29/11/96 stg3r(2.15) Minor
********************
Added CLOSE statement for DEC.
29/11/96 stgg (1.12)
********************
Added output file for GNUPLOT.
28/11/96 stgdw (2.16)
*********************
Increased dimension of NMSH2D and changed file omegdw to OMEGDW.
25/11/96 stg1r (2.24)
*********************
The (SUN) f90 compiler and some other (CRAY?) f77 compilers do not allow
a NAMELIST and subroutine to have the same name. I have been aware of
this for a number of years and have avoided such a clash. But the O.P.
created a NAMELIST STG1 and the I.P. created a subroutine STG1.
As per stg2r, stg3r etc., I have now split NAMELIST STG1 into two, namely
STG1A and STG1B, where "controlling" options are in STG1A and variable
values in STG1B. Thus,
5456 NAMELIST /STG1A/IOUT,IOUTDA,INDATA,IPSEUD,IDWOUT,
5457 A NBUG1,NBUG3,NBUG5,NBUG7,NBUG8,NBUG9,KRELOP,LAM,
5458 B RELOP,RAD,IRELOP,ISMITN
5459 C
5460 NAMELIST /STG1B/MAXLS,MAXPW,MAXE,LAMAX,MAXC,IBC,LCB,ISMIT,
5461 B MAXLA,MAXLT,NMIN,NMAX,LMIN,LMAX,IMAX,TOLE,EMIN,NELCOR,IPTS
I have also made the default radial mesh finer. I found an example
(which took a long time to figure out) where the default mesh gave
complete garbage. I have added the variable IPTS which controls the
mesh fineness (IPTS=no. of points per half cycle * sqrt(2)). The
default WAS 23 and I have increased the default to 33.
25/11/96 stg1nx (1.4), stg2nx (1.3), stg3nx (1.5)
*************************************************
No preprocessing required. Dimensions INCLUDEd via a PARAM file.
Also, default Buttle correction now uses Seaton's fit. Set MJS=0
to recover least squares.
22/11/96 stgb (2.9)
*******************
No preprocessing required. Dimensions INCLUDEd via a PARAM file.
21/11/96 stgb (2.8)
*******************
Allow execution to proceeed when closed channel wavefunction is
zero everywhere outside of box.
19/11/96 stgg (1.11)
********************
Reduce storage requirements, only store omegas to be plotted.
18/11/96 stgg (1.10)
********************
Added NPLOTI, NPLOTF to pick out transition from initial state NPLOTI to
final state NPLOTF.
30/10/96 stg1r (2.23)
*********************
New SHMITT routine added to generate orthogonal set by diagonalizing matrix
of overlaps. To access set ISMITN=1. IMPORTANT, ISMITN=-N is now used to
access the SHMITN routine, starting continuum basis at n=N when pseudo-states
present.
31/10/96 stgf (2.11)
********************
Allow calculation to finish when EOF encountered on H.DAT (due
to final symmetry in stg2/stg3 having no coupled channels,
nothing is written but the previous MORE=1 has already been set).
18/10/96 stgg (1.9)
*******************
omega -> OMEGA, omegadw -> OMEGDW
18/10/96 stgb (2.7) stg3nx (1.4)
********************************
tapeh -> H.DAT
18/10/96 stgf (2.10)
********************
Skip read of long-range coefficients (CF) if LAMAX=0. Case failed before.
16/09/96 stg1r (2.21)
*********************
Implemented numerical schmidt orthogonalization for large pseudo-state basis -
set ISMITN=1, use NMIN,NMAX,LMIN,LMAX to specify ISMIT(I)=NL compactly.
12/09/96 stgfrad (2.5)
**********************
Remove QDT option since have IQDT and Gailitis not required by damped code.
Mesh now uses constant dE after QNMAX where dE=2z**2/dn at QNMAX. INCLUDE
dimensions via PARAM file. Some IQDT tests made more robust.
Mods due to Alan Price.
14/06/96 stgf(2.9)
******************
No preprocessing of dimensions now. They are read from file PARAM -
not the same one as the inner-region! Inner and outer region codes
are normally kept in separate directories.
Changed file names tapeh->H.DAT and omega->OMEGA for consistancy with Belfast.
Also, added parity to spin in the OMEGA file (LS case, was already there on J).
12/06/96 stg1r,stg2r,stgjk,stg3r,stglib
***************************************
NO preprocessing of dimensions now. They are read from file PARAM. Two
sample files PARAM.CC and PARAM.DW are available and can be copied to
PARAM as required and the dimensions tweaked. tapeh->H.DAT in stg3r.
12/06/96 stg1r(2.19)
********************
Allow Blume & Watson screening by other electrons WITHIN a sub-shell i.e.
TOCC->TOCC-1.
11/06/96 stgf(2.8)
******************
Added infinite energy omega (dipole only).
06/06/96 stgjk(2.16)
********************
Tom's final fix of the indexing problem in RECUD for photoionization.
Note: only requires stg2 (N+1) symmetries ordered such that ALL even parities
precede ALL odd.
05/06/96 stgbb(2.2)
*******************
Again fix numerical failure for weak/strong closed channels.
05/06/96 stgbb(2.1)
*******************
Initial workstation implementation of Belfast code.
29/05/96 stg1r(2.18)
********************
Fix CIV3 option (so that it actually works!) for John Tully.
24/05/96 stgb(2.4), stgf(2.7)
*****************************
Solved numerical failure for case of weakly and strongly closed
channels (mainly stgb) required for inner-shell photoionization.
21/05/96 stgjk,stg3r,stgb,stgf,stgbf (minor)
********************************************
Minor changes to eliminate compiler warnings an errors with f90.
19/04/96 FARM
*************
Fixed a bug from Val Burke in subroutine sleqn of directory phv. Can cause
crashes or possibly huge k-matrix elements/omegas. The results of "successful"
runs are claimed to be unaffected.
19/04/96 stgf(2.5)
******************
Inline TANDTD in TLAG for efficiency.
08/02/96 stgfrad(2.4)
*********************
Tom added NCUTOFF for DR.
29/01/96 stg1r(2.17) minor
**************************
Fixed bug (introduced by me in v2.15) which gives "wrong" (!) Blume &
Watson screening when occupation numbers are not specified in the radial
file. If they were specified then code was fine.
18/11/95 stg3r(2.12)
********************
Tom's energy shifts of N+1 CFG eigenvalues added. Useful for
photoionization. See comments in code (NSHIFT) for details of use.
11/11/95 stg2r(2.29)/stgdw(2.15)
********************************
stg2r, fix nl writes to MIXDW.DAT for model potential operation.
stgdw v2.15 can now follow a model potential run of stg1r/stg2r.
Continuum waves are distorted by a direct core model potential only,
and direct plus local exchange for the valence orbitals.
Can also read-in observed term energies, rydbergs relative to ground,
in same order as specified in stg2r (after any sorting) i.e. as listed
in MIXDW.DAT.
07/11/95 stg2r(2.28)
********************
Fixed obvious bug of 15/10/95, nl's were being written to unit=2 instead
of unit=22, causing read error from stgdw v2.14.
31/10/95 stgjk(2.14)
********************
Tom fixed an important bug affecting photoionization (and hence our
damped code).
15/10/95 stg1r(2.16),stg2r(2.27),stgjk(2.13),stg3r(2.11),stglib(2.7),stgdw(2.14)
********************************************************************************
Cosmetic changes and esoteric bigfixes from Belfast, except
stg2r/stgdw.f which now write/read the orbital definition
list so that there can be no mismatch with the radial file
(orbital numbers are not read by stg2r but stgdw did read them).
Adding model potential to stgdw has still to be done but code
now stops with an error message if this case is attempted.
Added Bayliss alternative to Norcross for polarization potential.
14/10/95 preprmx.f
******************
A preprocessor that works for all the current inner and outer
region codes is now available under /misc together with dimension
files ppn (inner), ppa (outer) and ppnx (NX-LS) for people/places
without the tedi editor.
04/10/95 FARM
*************
Several bugfixes, this solves some problems we had with large
cases in N and O with odd values of omega being overwritten
with numbers of order d-311.
23/09/95 stg1r
**************
Reads orbital occupation numbers from the radial file for use by Blume&Watson.
Zeroes/blanks default to original R-matrix coding which assumes for orbital
of pricipal quantum number n that there are complete closed shells upto n-1.
Thus there is no 2s**2 for 2p but 3p can have 2p**6 even if there is only 2p**4!
Hence new coding.
Example of new orbital header card:
-7 2 2 0 2.0 -97.975267 310
The 2.0 (F5.1) is the occupation number.
22/09/95 stg1r (minor)
**********************
Now accepts semi-relativistic bound orbitals from AUTOSTRUCTURE and
switches-off appropriate mass-velocity and Darwin integrals. Reads
a label from the radial file to determine orbital type, default non-rel.
08/09/95 stgjk
**************
Belfast reversed bugfix in SPINOR BUT NOT IN BSPNO. Doesn't affect Fe25+
for example but does affect Xe53+. The BSPNO bugfix was what corrected
the long-range coefficients and hence our core radiative rates.
06/09/95 stg2r IMPORTANT
************************
On 17/08/95 I attempted to fix a bug in stg2r - the pseudo-resonance overlap
matrix was being evaluated in BP mode, but of course nothing was set for it.
In jumping around this evaluation I accidently skipped one line too many, this
unfortunately had a rather non obvious effect on BP results.
30/08/95 stg2r
**************
Skip writes in DMEL when there are zero-coupled channels for a given
symmetry elsewise stgjk aborts. Reorder symmetries into odd and even
when INAST=0, required for stgjk dipole matrices.
29/08/95 stg2r stgjk
********************
Don corrected write format error to tccdw in stgjk. Also changed sign of
largest component of e-vector to be positive, for TCC comparisons with MCHF.
26/08/95 stgdw(2.13)
********************
Takes symmetry order input from stg2r and outputs energy order K-,T-
matrices for difls and stgic. This completes the outstanding developments.
23/08/95 stgdw(2.12)
********************
A new release of stgdw which hopefully incorporates everyones favo(u)rite bits.
Browse the comments to see what I mean.
18/08/95 stg2r stgjk
********************
Added RELOP='MVD' to stg2r to be used in conjunction with TCC run and stg(dw)ic.
Modified TCC output via stgjk.
17/08/95 stg1r stg2r stgjk stg3r
********************************
stgjk IMPORTANT bug-fix for TCC's
stg1r stg2r stg3r minor mods, including latest Belfast fixes.
16/08/95 stgg.basic
*******************
A simplified version of my stgg is now available. It plots collision
strengths from an omega file, nothing more. Requires the NCAR library.
14/08/95 stg2r stgdw(2.11)
**************************
Joint release of stg2r/stgdw (v2.11) which treats L>LNOEX efficiently
by skipping every other total spin. Specify MINST,MAXST as normal (i.e.
as required for L<=LNOEX) everything should be handled transparently.
DO NOT MIX THIS RELEASE WITH ANY EARLY VERSION OF STG2R/STGDW.
11/08/95 stgdw IMPORTANT
************************
Exchange overlap assumed one electron orbital energies POSITIVE.
IMPACT format should be NEGATIVE, required by AUTOSTRUCTURE and
recommended for R-matrix. Changed sign of overlap so NEGATIVE
input assumed, Tom has changed Fischer's code. AS was always negative.
11/08/95 stgdw
**************
Input (namelist) variables tweaked a little.
Elastic omega's and T-mx evaluated now.
Test on whether input orbitals too big for NMSHBD now works correctly.
08/08/95 NX codes
*****************
stg1nx, stg2nx, stg3nx updated to be consistent with new B.P. codes.
If you want to use them with the old OP codes you will need to
cp/mv the OP files tapenx1 -> NX1.DAT and tapenx2 -> NX2.DAT
07/08/95 stg1r stgjk
********************
stg1r: Allow RELOP='TCC' input (equivalent to 'YES')
stgjk: IDWOUT=2 correctly terminates after BOUNDJ (RELOP='TCC'
not actually needed when IDWOUT=2)
05/08/95 stg1r stg2r stg3r (all minor)
**************************************
stg1r: mods to mesh and finder for numerical problems encountered with
neutrals.
stg2r: mods to DW coding for CC operation.
stg3r: pseudo-resonance removal OFF by default so as to think about TOLER
08/07/95 stg2r/stgdw RELEASE
********************************
The final major revision of the stg2r/stgdw suite has now taken place.
All obvious major CPU time and memory bottlenecks have been removed.
Only bug fixes and tweaks should remain.
07/07/95 stg2r
**************
collision algebra now stored and summed internally, leads to much
smaller file for stgdw and faster execution for stgdw- no duplication
or searching for duplication.
06/07/95 stgdw
**************
Added automatic use of direct access file for when number of channels gets
too large to store all continuum orbitals intenally.
Use of symmetry ordered data now tested by block for duplicate coefficients,
no longer a bottleneck.
Fixed a couple of bugs.
Current version: 2.9
05/07/95 stgdw stg2r
********************
stgdw (V2.8) can now take SYMMETRY ORDERED data from STG2. This
update is user transparent. stg2r has some minor mods for IDWOUT=2
and symmetry order. Thus you should update stg2r on updating stgdw.
In large cases it is extremely important to use symmetry ordered
terms in stg2r. Note below the ISORT parameter to help you with this.
01/07/95 stg2r stgdw
********************
Reduced dimension requirement of stg2r for IDWOUT=2 by adding &CHD:
set=1 for IDWOUT=2 runs (large) and =&CHF for CC. Code checks for
correct setting. Also, modified tagging of metastables so that can
run symmetry order for stg2r with IDWOUT=2, note stgdw still needs
energy order for now.
Reduced dimension requirement of stgdw by adding NCHUD: set=NCHMD
for (large) non-unitarized runs and =NCHFD for unitarized runs.
This means that both stg2r and stgdw can deal with cases involving
many thousands of channels and still only require 40-80Mb of memory.
29/06/95 stg1r minor
********************
Added NBUG9 to NAMELIST. Also helpful warning message about
explicitly setting LAMAX in NAMELIST when MAXLA large causes
memory/waste problems.
26/06/95 FARM bug
*****************
A bug in FARM (gal/coul.f) has been corrected following notice from
Valerie Burke. It ONLY affects IONS.
22/06/95 stg2r stg3r
*******************
Dimension bug in Tom's pseudoresonance code fixed. In principle
this could have affected TOLER=-1 runs but in practice only seems
to have been a problem for large runs with TOLER=0.01.
Have rewritten stg3r so that can use INAST.gt.0 with TOLER=0.01,
useful for debugging! Tidied things up a bit as well.
20/06/95 stg1r stg2r stgdw
**************************
Updated to allow DW to work when model potential (IPSEUD=1) used to
reduce size of "core" problem.
19/06/95 stgdw
**************
Last tweak? v2.6
17/06/95 stgdw
**************
Numerics and memory optimized, results in good speed-up and space for
big problem. Also, IPERT added for even faster runs.
14/06/95 stg2r
**************
Added NMETA for IDWOUT=2. For NMETA<<NAST the speed-up is enormous as prior
to this stg2r was still evaluating all NCHAN(NAST)*NCHAN(NAST) interactions,
which were then ignored by stgdw for NMETA<<NAST. This reduces the stg2r run
to neglible time (~ 1 min). As stg1r is also trivial (< 1 min) when IDWOUT=2
99% of the time required for a DW run is now in stgdw - speed this up next?
As noted before, stgdw requires the terms in stg2r to be energy ordered. This
is inefficient but trivially so when IDWOUT=2. If a CC run is required as well
it is better to run IDWOUT=0 with the terms symmetry ordered, the time reduction
is approximately a factor NAST/NGROUPS where NGROUPS is the number of distinct
SLp's. For "large" NAST this factor can be substantial. Thus, IDWOUT=1 is not
currently recommended. I have added a parameter ISORT to namelist STG2A, when
non-zero it will re-order the input terms to symmetry order but at the
expense of expanding the CC expansion to the CI expansion - which is
what is required in B.P. mode of course.
13/06/95 stg2r
**************
Dimensions exceeded in ODH0 due to failure to initialize NONZER &
NDIML to zero for each partial wave - now corrected. Error present
since 13/04/95. Could affect all runs with stg2r.
08/06/95 stg1r stg2r stgdw
**************************
Minor changes for DW operation. Go back to unscaled energies in stgdw.
05/06/95 stgdw
**************
Minor changes: assume coeff's from algdw unique and reactance mx symmetric
03/06/95 stg2r
**************
DW writes for algdw are now unformatted (to algdwu).
02/06/95 stgdw (mega changes)
***************
A new parameter NMETA specifies the number of initial states (ground
plus metastables since these are the important transitions for
plasma modelling and experiment, excitation between excited states
is unimportant until relatively high densities as the bulk of the
population is in the ground and metastables - radiative transitions
being the dominant depopulating mechanism for the excited states) that
omega's are calculated for. In general, we would set NMETA << NAST/NTERM
for large NTERM and this should be highly efficient. I plan to do
some timing comparisons for the Cl6+ E-A where NTERM=40 or so.
The parameter NOMWRT output on the omega file gives the number of
transitions in the omega vector, it depends straightforwardly on
NMETA, NTERM and whether elastic transitions have been output (using
the ELAS parameter). There should be no need for the user to set it,
although it can be read-in.
A new parameter NTHRSH specifies whether threshold omega's are
calculated (IF .NE. 0) or not (IF .EQ. 0). These are calculated
for all possible transitions specified by NMETA & NTERM so could
be time consuming for large NMETA. The unitarization is basically
just for the 2 target states (since no other omega's are calculated
at the threshold energy) so if N-state unitarization is important it
maybe more accurate to set NTHRSH=0 and assume constant omega's.
The threshold omega's are written to the omega file in position 1 with
energy set equal to 0.0000 and KM/MXE is incremented by 1 (IF NTHRSH.NE.0).
I have removed the splining (sorry Tom) as I believe it is better done
in a subsequent code which graphs the omega's and/or adds resonances
and/or multiplies them by auger yields (such as the AUTOSTRUCTURE
post-processor which I have been developing in tandem as per my last note)
stgdw can now read either a formatted algdw file or an unformatted
algdwu file, no user action is required other than to ensure that one of
these two files is non-empty. I will shortly change stg2r to output
an unformatted algdwu file (only - unless anyone objects, stg2 will be
left as formatted algdw). For large cases I believe algdw(u) will
become large enough for unformatted I/O to a significant time saving.
I plan to leave the mixdw file as a formatted file as I don't think
it will ever get excessively large.
The IPRTM parameter can now be set negative for a very terse output
to routdw for production runs.
I think there was a small error in the elastic transitions. There used
to be a test IF(NCHOP.LE.1) then skip to the next energy as there is no
contribution to the omegas in this case. This is true for NCHOP=0 but
for NCHOP=1 the tan(phase) should have been accounted-for.
For ease/consistency with the R-matrix codes, the input scattering
energies EE are now expected to be in z-scaled Rydbergs.
Note, stgdw requires that the target symmetries in stg2/stg2r be
listed in energy order. stgdw now checks for this, we could re-index
to allow non-energy order but life is too short. In the past it
was necessary to group like symmetries together in stg2r for a B.P.
run. This is NO LONGER required, stg2r will re-order them in B.P.
mode and leave them as input in LS mode - which is where we generate
mixdw/aldw. This means you should always input the target terms in
energy order in stg2r to make the dstg2 useable by all options.
31/05/95 stg1r stg2r stglib stgjk
*********************************
IDWOUT=2 coded for fast DW operation, LRNGDW need not be set by user.
Note, KCOR now passed from stg1r to stg2r correctly for algdw and stgdw.
stglib required changes to evaluation of GAM(MA) to cope with IDWOUT=2.
The BP codes now do everything the OP codes do except for the polarization.
05/05/95 stgbf stgf
*******************
stgbf now calculates beta in jk-coupling, stgf has extra write of
K quantum number for this.
18/04/95 stgb, stgf, stgbf
**************************
New IP-BP stgb (no ID unit11 writes tho').
Werner's prebf and stgbf (IP-BP) includes beta param.
Dimensions &SA1 and &SA2 now required....
stgf writes for stgbf beta uncommented (use IBETA to switch).
This has Tom's NOMWRT param & my reduced dimension set.
makeb and makebf makefiles updated.
17/04/95 stg2r, stglib * stgjk
******************************
Tom's DW output (stg2r stglib) included but still need LRNGDW and stg1r
to be done (then only polarization remains TBD).
stgjk, Tom's alternate TCC output file.
13/04/95 stg2r & stg3r
**********************
Tom's supression of pseudo-resonances included IF TOLER.GT.1.E-19 in
stg3r NAMELIST/STG3B/, use NBUG9.gt.0 to get debug printout on fort.78.
stg2r automatically writes to file amout (RELOP.EQ.'NO' only, for now), set
IBUG5.gt.0 to get debug printout on fort.67.
12/04/95 stg2r & stg3r
**********************
stg2r: Use RELOP='TCC' if you require TCC in stgjk (RELOP='TCC' again).
stg3r: This uses NAST not NEST now for the observed target energies.
Also, NAST .LT. 0 merges states; use NAST .GT. 0 to act like our old
NEST .LT. 0.
11/04/95 stg1r and f77
**********************
If you are using f77 version 2.x you need to increase the default
allocations to -Nl40 (and maybe -Nx400). f77 v3.x defaults are
sufficiently large.
10/04/95 IRON PROJECT B.P. CODES
********************************
Minor bug fixes to stg1r, stg2r & stglib. IMPORTANT spin-orbit bug-fix
to stgjk which corrects the long-range coefficients (finally).
10/04/95 Opacity Project codes
******************************
These now include Tom's method for supressing pseudo-resonances.
21/11/94 IRON PROJECT B.P. CODES
********************************
The Breit-Pauli codes from the Iron Project are now available.
The I/O between stg's has changed but the UNIT 5 & 6 I/O has
been modified to conform to our requirements, so you should
notice no difference in actual operation.
The PREPROCESSING has changed however. This is now in the style
of the asymptotic codes and could probably use our rmxppa.x code.
However, I prefer to do away with "preprocessing" and just insert
the dimensions with your favourite editor. Here is an example using TEDI:
tedi stg1r
u;ppnt
The file ppnt contains the editing instructions and dimensions, and notes.
The dimensioned code will be in stgn.f; the source code (stg1r) remains
unchanged. Note, the file ppat can be used for all of the asymptotic codes.
WARNING: It is easy to inflate the size of stgjk and stg3r, and to a lesser
extent stg1r. See comments in ppnt.
I get exactly the same answers for the omega file as with the old BP codes.
stgjk seems much FASTER than the old stgjj.
The new BP codes have the same functionality as the old codes, i.e. model
potential operation now works and the NX writes have been inserted. Still,
there is no polarization potential or writes for stgdw. The Opacity Project
codes should be retained for this purpose for now. It is probably worth
keeping the old BP codes in directory /bpold say just incase future studies
throw-up a case that needs to be checked against the old codes.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
14/11/94 stgfrad (1.5)
PARAMETERS as per stgf of 16/10/94
06/10/94 stgf (1.8)
H=0 IPERT=0 bug fixed. Also, PARAMETERS inserted so can edit dimensions
in rather than use preprocessor.
17/09/94 Dirac-Fock R-matrix codes
**********************************
As from Belfast, completely untested. Past experience has shown that
their Unix shell scripts are a pain. If the codes can be implemented as I
have done for LS & B.P. codes etc, it would be appreciated by all.
Note: a version a stgf for DF R-matrix is now present.
03/09/94 stg1r
Evaluation of bound orbital mass-velocity term when a core model potential
is being used is believed to be incorrect. I have coded what I think is
correct!
13/08/94 stgjj, stg3r, stgf, makef
Allow for rad decays in B.P. mode, either by decoupling K or reading
radiative rates (from arad) explicitly - merits see me.
NOT in stgfpar yet!
25/07/94 stgf, stgfpar bugfix
Near threshold closed-channel bug fixed and numerics improved.
Elastic transitions available using ELAS parameter, default = 'NO'
6/23/94 Tom modified stg3r so that the ground state can be shifted by a
negative energy. I deleted the option that when EST(I).LT.0
(for NEST.NE.0), then EST is in cm-1 relative to ground state.
To shift the ground state, just have NEST.LT.0 and EST(1).LT.0
in Rydbergs, EST(I)-EST(1)=relative energies of higher states
in Rydbergs.
6/23/94 Tom modified stgf so that IOPT1=2 can be used for Breit-Pauli mode.
In that case, IS=0, so that is not a good flag for when to stop
reading slpi list. Instead, just specify IS=999, IL,IP arbitrary
in dstgf to specify end of slpi list (see comments in stgf code).
20/05/94 FARM makefiles
The FARM makefiles have been updated to work on NERSC Cray's.
Don't forget to switch timings in asy/asy.f!
16/05/94 *******FARM********FARM************FARM**************FARM***********
*****************************************************************************
FARM is now on russell ( under, ../rmatrix/nigel ) for testing. Initially,
feedback to me changes wanted so I can coordinate them; also, any apparent
errors so I can get back to Val Burke.
If you don't have the gmake command available copy it, & set path if not in /farm.
TO INSTALL (on russell):
cp /export/home/atomic/rmatrix/nigel/farm.tar.gz ~/rmatrix
cd ~/rmatrix
gunzip farm.tar
tar xf farm.tar
cd farm
cp /export/home/atomic/rmatrix/nigel/gmake gmake
./gmake MX=sunV
END
If all has gone well you should see an executable called stgfarm.x and a
sample input dataset dstg.farm.
The I/O structure of FARM is fundamentally different from stgf and stgfvpm.
To overcome this I have adopted the following procedure. A parameter "iop"
is used to switch between the original FARM I/O and our usual I/O. Set iop.eq.0
and the I/O is as in the FARM write-up. iop.ne.0 (default) uses our "iprint"
and "iprrm" variables to control I/O. I assume we will always use iop.ne.0.
Then, our "omega" file will always be output, as will rmatel when iprrm=1
(iprrm=2 has also been coded). "iprint" takes on the values -2,-1,0,+1
and follows the spirit of out usual IPRINT variable. DO NOT use the "prtflg"
variables discussed in the farm write-up when iop.ne.0 since "iprint" and "iprrm"
will set-them. Unpredicatable results will occur if the two I/O methods are
mixed. The rest ofthe input, however, is as described in the farm write-up.
Energies/ranges/increments and L's are relatively straighforward.
Also, lrglx (highest L/2J for which exchange has been included i.e. nspn.gt.0)
must be set, unlike stgf/vpm which takes it from nspn.
Similarly, jj=.true. must be set for B.P. operation, unlike stgf/vpm
which deduces it from nspn=0.
FARM sets-up a number of parameters using jj and lrglx before it has read the
first nspn case and so it may not be easy to change this operation.
Also, how could we determine a future non-exchange B.P. run (-0 !)?
The lrglx parameter is rather a pain, partial waves are omitted if it it set
incorrectly.
I have tested-out a number of cases and have noted the following.
For above threshold neutrals FARM is a factor of 3 to 7 faster than the VPM method.
Case of 6CC Cr.
For positive ions, however, it is a factor of 4 to 5 slower than STGF (with
PERT='YES') with all channels open. In the resonance region it is a factor
of 10 to 15 slower! The case was 8CC-LS Kr6+.
Points to note. The code is all in lower case, hence namelist input.
Should you have the urge to make changes REMEMBER the default is
implicit integer (a-z) so all reals must be declared explicitly.
Further, most information is stored in the two vectors ix(1500000) and
x(1500000) which are infact equivalenced! You have been warned.
***********************END***************FARM**************************************
13/05/94 stgfvpm
stgfvpm gave me garbage for reactance matrices on SUN. The kmtrx (& kmtt) are
integer arrays in sr.wrtkm ! Presumably o.k. on Cray since both are *8,
otherwise Don would have noticed. Have changed to tkmtrx (& tkmtt), i.e. real.
Note: sr.kmat does not appear to be used, k-matrices are obtained from vpm.
Have taken the opportunity to move the omega re-indexing in sr.writom
outside of the energy loop since it offended my sensibilities.
07/05/94 stg1 stg(f)vpm
stg1 replaced by Don's and minor bug for MAXLA=0 fixed.
stg(f)vpm replaced by Don's BUT with Buttle coefficient DO LOOP index
increased from 50 to 100 consistent with Don's larger dimensions.
Also, separate dimension tests inserted for XSPART/XTOT for
write to unit=3 crosstot. IPTOT .GT. 0 -> .NE. 0 ?
05/05/94 Don stg1,vpm
I have modified stg1 to used either the Norcross polarization potential
or the Baylis potential: V= -(1/2)alpha*r^2/(r^2+rcut^2)^3. The
STG1 input variable IPLFN is used to determine which function is to
be used: IPLFN=1 for Norcross; IPLFN=2 for Baylis. Note the default
is now set to Baylis. I have also made some very changes in
STGVPM.f, after determining that the changes that Nigel made to
write files for the differential and partial cross section codes are
correct; of course, the output for the partial cross section code
does not work on the Sun. The primary change is in the subroutine
WRITOM. It will now properly print the omega file, including the
the elastic collision strengths when ELAS='YES'. I have also increased
the dimension on the number of channels from 40 to 60.
03/05/94 stgf (minor)
If neutral STOP!
Moved dimension tests about a bit.
30/04/94 Photoionization stgb,stgf,stgbf
Minor fixes to photoionization operation (only IRAD=1 IPERT=0 worked).
Warning, watch dimension &MNP in stgbf!
28/04/94 B.P. Codes BUG
stg2r contained a bug, the unit number for the rkints direct access
file was not defined (defaulted to zero) this caused the rkints file
to be corrupted (on SUN, - CRAY ?) when stg2r execution completed.
The results from this run were o.k. but any subsequent re-run of
stg2r using the corrupted file gave weird errors. If stg2r always followed
stg1r, or the rkints file was read-only, or rkints from stg1r was stored
as a separate file and then re-instated then there was no problem.
---------------------------------------------------------------------------
Also: stg1r and stg2r now write files for the NX code, needed if using
model potential for the core. Minor bug in stg3r fixed, wouldn't recognize
LRGL input for INAST.GT.0; should always use INAST=0 in anycase.
23/04/94 stgdw & stg2
ERROR in stgdw? In SR.CROSS label 171 should be 25 lines higher
on IF(NUNIT.EQ.0)THEN
Added write of MAXORB by stg2 to mixdw file for DW operation.
Further adapted stgdw to operate when NRAD.GT.MAXORB; note
previous problem was due to NRAD.LT.MAXORB. stgdw had more
severe requirements than R-matrix tested-for. stgdw should
now be consistent with what is allowed by R-matrix.
20/04/94 stgdw
errors corrected for case of vacant orbital numbers.
top-up extended.
15/04/94 stgf
Fatal error in dipole top-up corrected, caused by my quadrupole top-up - sorry.
21/03/94 stg2
GAMMA -> LOG(GAMMA)
19/03/94 stg1 stg2
simplifying DW operation
18/03/94 stg2
High-l fixes for DW
17/03/94 rmatrix.dvi
A tex file of a document describing the QUB codes is available
as rmatrix.dvi it is about 125 pages long, happy reading.
17/03/94 stg*nx
The NX codes have been updated to utilize model potentials
as a continuation of a stg*r run.
17/03/94 stgfvpm.f
Dons IPRRM writes implemented
16/03/94 stg1 stg2
minor changes
09/03/94
stgfvpm.f: skip evaulation of vanishingly small omegas (high-L).
04/03/94
High-L bug fix to stgf
Top-up on vpm (stgfvpm.f) fixed.
Output rewritten, use IPRINT=-2 and -1
21/02/94 Iron Project Breit-Pauli R-matrix codes.
*************************************************
Use same (but revised) preprocessor as Opacity Project codes and
same (but revised) ppn's. OP codes all updated for consistency with
IP codes. The library stglib uses the dimensions pp2.
IP codes to be run as
stg1r
stg2r | stglib
stgjj | stglib
stg3r | stglib
stgf
Sample datasets are provided for the Kr6+ problem
If RELOP='NO' go 1,2,3; if RELOP='YES' go via jj. stgf is revised OP code.
I have constructed a preprocessable stglib which uses pp2, seems o.k.!
All input has been streamlined a la OP codes i.e. namelisted and free
formatted. Test-run on C2+ looks o.k., fine-structure transitions required
more basis orbitals (25) for convergence than LS operation (15).
(Dipole) top-up is crude and untested; LS TOP cannot be applied.
Model potential operation is eccentric.
10/02/94 stg2
Introduce MINLT,MAXLT,MINST,MAXST for automatic generation of
total SLp's, useful for IDWOUT=2 high-L problem.
03/12/93 non-exchange R-matrix codes installed
preprmxnx.f
ppnx
stg1nx
stg2nx
stg3nx
To be run in sequence 1,2,3 after stg1 and stg2 have been run
for at least one partial wave to produce tapenx1 tapenx2 for the
non-exchange code.
All three use dstgnx (UNIT5) and dstg3 needs to be available,
but a trivial generic one can be used if there are no energy shifts.
30/10/93 stgf
More high-L fixes, this time to Coulomb series.
29/10/93 stgf
More high-L fixes, this time to open-channel long-range integrals.
27/10/93 stgf
Introduction of variables MINLT & MAXLT into NAMELIST STGF to
select only those SLp with L.GE.MINLT & L.LE.MAXLT, useful for
non-exhange code where the use of IOPT1 would be tedious to select
100 cases say.
25/10/93 stgf
Top-up of quadrupole transitions now done as well as dipole when
LRGLAM specified .GT. 0.
High-l tweaked including some REAL*16 for SUN which should become
DOUBLE PRECISION on CRAY. Note on SUN that now use DOUBLE COMPLEX
to be consistent with d.p. in rest of program, replace with COMPLEX
on CRAY.
22/10/93 stg3
Bug-fixed that occurred when energy levels flipped by energy
corrections (NEST .LT. 0) when INAST .GT. 0.
INAST=0 enabled, i.e. a single pass is made through tape13 reading
all partial wave data.
19/10/93 stg1
One- and two-body polarization potentials included a la' Mitroy,
specify unique ALFD, RCUT and KCOR (last closed-shell orbital no.)
in NAMELIST &STG1 of dstg1.