readme

一套应用很广的光谱程序

                    -99999999.
   -1
 
except that the first two (optional) control cards will be absent (but
can be edited in), and except that the "-6" in columns 73-74 of the
principal control card will be zero if a zero was used in columns 71-75
of the RCN control card.  (As in RCN, the value -6 causes intermediate
calculational output to be sent to the monitor screen during the
course of the RCG calculation.)  Here the Eav values in Ry from RCN
(-1197.769, 1184.586, and 1185.615, resp.) have been changed by RCN2 to
values in kK (-130337.379, etc.), and then to values 0.0, 344.0732, and
231.1776 kK relative to the value for the first configuration. Values
of the other radial energy integrals have been scaled down by the
factors 0.90, 1.00, 0.90, etc., and are in kK, which will then be the
units of energy-matrix elements and eigenvalues, unless the "1000.0000"
on the second optional control card is changed--for example, to 8065.48
to obtain all energies in eV.  (The "0000000000" on the second optional
control card can be replaced by non-zero values to change the scale
factors from the ones specified on the RCN2 control card to these new
values.)
     NOTE:  For a first RCG calculation, the input file must consist of
(or begin with) a set of coefficient-of-fractional-parentage decks for
each set of subshells (e.g., s0, s1, s2, p0, p1, p2,...) that will be
involved in any of the configurations for which calculations are to be
made.  (Decks are not needed for singly occupied subshells if there
will be no more-highly occupied subshells of the same l.)  The file
ing11k contains such a set of cfp decks, and can be used for this first
calculation to compute files tape72, tape73, and tape74 for use in
subsequent RCG calculations.  [Some IBM systems do not allow file
numbers greater than 50.  For these systems, the "72" on the first line
of ing11k can be changed to "42" to produce files tape42, tape43, and
tape44; the statement id2def=72 near the beginning of the main program
of RCG must correspondingly be changed to id2def=42, and the OPEN
statements for units 72, 73, and 74 likewise changed appropriately.]
 
 
 
NOTE:  The file "rcng.UNIX" is a c-shell procedure file for automatically
running RCN, RCN2, and RCG in succession on UNIX-type systems.
If it is launched by the command
rcng.UNIX .KVI
the three output files will be named
out36.KVI, out2.KVI, and outg11.KVI  .
The file rcng.VMS is a similar command file for VMS systems; for usage,
see the file RCNwriteup.
 
 
 
RCE input
     If on the RCG control card a -1 is placed in columns 9-10 and the
card containing "-99999999." is preceded with an identical card, except
with 5's instead of 9's, then RCG will write a binary file tape2e and
an ASCII file outgine.  (These changes will be made automatically if a 1
is placed in column 10 of the RCN2 g5inp control card.)  If the name of
outgine is changed to ine20, RCE can be run with the RCG energy
parameters and theoretical energy levels as input.  RCE should then
converge straightforwardly in about three cycles with little or no
change in parameter values, which then provides a test run for the RCE
code.  Genuine RCE runs require that experimental energy levels be
edited by hand into ine20 in place of the theoretical eigenvalues.  An
RCE output file parvals contains final least-squares parameter values,
which can be edited into the RCG input file ing11 for a repeat run with
the new parameter values.
 
Program writeups
     Various possible additional RCG options can be input on the RCN2
control card and carried over automatically to RCG, or entered directly
onto the RCG control card, but they are too numerous and detailed to
describe here.  Likewise, notes on convergence problems, etc., in RCN
and RCE will not be given here.  For details, see the hard-copy program
writeups available from the author upon request:
  e-mail to        [e-mail]    or regular post to
     Robert D. Cowan
     2917-B Nickel
     Los Alamos, NM  87544
     USA
phone [phone].
 
Note: ASCII versions of the writeups are included in the ftp directory;
they are useful to a large extent, but are somewhat difficult to read in
places, because there are no sub- nor super-scripts, Greek letters, nor
script letters for coupled quantum numbers LSJ.  There are no sample
output listings in the RCN and RCG writeups, but these are contained
in files OutputK+5 and OutputSe+24 for a normal dipole-spectrum run
in K VI and a dielectronic-recombination calculation in Se XXV.
 
 
 
Program information
     The programs rcn.f, rcn2.f, rcg.f, and rce.f are all in FORTRAN77,
complete with all necessary subroutines except for standard intrinsic
functions such as min, max, exp, log, etc. (Generic function names have
been used so as to require no changes for either "single-precision"
64-bit- or double-precision 32-bit-machine calculations.)  The programs
were originally developed in the 1960's on the IBM 7030 ("Stretch"),
and over the course of the years many new features were added and the
programs adapted to other IBM mainframes, CDC 6600, 7600, and 205
computers, CRAYs, various VAXs and MicroVAXs, and finally SUN and
IBM RISC System/6000 workstations, and Macintosh and PC desk
computers.  The programs have also been used on a Silicon Graphics
Origin 200.  Other people have found the programs readily adaptable to
Apollos, Hewlett Packards, and other computers.  Each program contains
a timing routine "seconds," with sections appropriate to CRAYs, VAXs or
Macintoshes, SUNs, IBM RISCs, and PCs; the appropriate section must be
selected, or a new section written for the computer to be used (or t
simply set to 0.0, as times are used for information only).
 
     Compilation has generally been done with minimum options and op-
timization--for example, cf77 on CRAYs, with or without the g_floating
option on VAXs; f77 on SUNs; on Power Macs, the Language Systems MPW
script RUNPPC with -t72 option added; and on PCs the GNU g77 compiler.
     Relative computing times on different machines are roughly
 
                          RCN         RCN2       RCG        RCE
      CRAY(Y-MP):         1.0         1.0        1.0        1.0
      SUN(Sparc2):        15           4          2
      SUN(Sparc20):        4
      SUN Ent. 3500        1           5         0.3        0.2
      IBM930:              4          1.3        0.2
      IBM550:             2.5         0.7        0.2
      IBM320H(32H):        4          1.2        0.3
      MAC 6500/300        2.4         20          4          6
      MAC G3/350          2.8         75          7
      Pow. Cen. Pro 240   3.2         12          6          5
      IBM Aptiva E275      1          16          1
      SGI Origin 200      0.3         1.7        0.7
 
[The SUN Enterpise 3500 has four 336 MHz processors, the Power
Center Pro MAC clone has a 240 MHz processor, the Aptiva has two
475 MHz processors, and the Silicon Graphic has four 180 MHz
processors.]  Absolute computing times for three-configuration
K VI (K 5+) runs are on the CRAY about 1.8, 0.03, 0.4, and 0.2 sec
for RCN, RCN2, RCG, and RCE, respectively, the RCE time being for
test-run convergence in three cycles on each parity.
     For the dimensions used, file sizes on the 240, in Mbytes, are
approximately
 
                                  RCN        RCN2       RCG        RCE
       Source file                0.16       0.07       0.4        0.08
       Executable file            0.40       0.26       0.78       0.24
       Minimum RAM for execution  2.0        1.2        8.0        3.0
 
Note:  The current versions of rcn.f and rcn2.f contain no equivalence
statements, but rcg.f does have such.
     The file rcg-noeq.f is a versions of rcg.f from which all equi-
valence statements have been removed, as FORTRAN 90 is expected to frown
on such statements.  The minimum RAM for execution of this version is
about 40% greater than for the version with equivalences present.
      Considerable recoding was required in rcg, in addition to simple
removal of the equivalence statements, but these changes were made in an
older version of rcg, and so rcg-noeq.f does not contain a number of
refinements that were introduced later into rcg.f.
     The file rcglg.f is a version of rcg.f with larger dimensions--
large enough to handle f(n)d configurations for all n.  Note that
for such configurations, the files tape72, tape73, and tape74 must
be computed using the rcg input file "cfp" (with comment cards
removed) instead of using ing11k.