Release Notes for DDSCAT version 4b.1

                                 B.T.Draine
                      Princeton University Observatory
                              Princeton NJ 08544
                        (draine@astro.princeton.edu)

                                     and

                               Piotr J. Flatau
                       Scripps Institution of Oceanography
		           California Space Institute
                  La Jolla Shores Dr., La Jolla CA 92093-0221
                           (pflatau@macao.ucsd.edu)

                          last update: 93.12.15

                          1. Significant Changes


   Release 4b.1 (93.07.09) has a minor change in GETFML with regard to the
method of computation of Q_sca.  Once a converged solution is found:
   1. the scattering efficiency Q_s1 is computed by averaging the
      scattering over ICTHM values of THETA and IPHM values of PHI.
   2. Q_s1 is used to estimate the albedo A=Q_s1/Q_ext.
   3. a second estimate of the scattering cross section Qs2 is computed
      by taking Q_s2=Q_ext-Q_abs
   4. since both Q_ext and Q_abs are in principle calculated accurately,
      their difference is a good estimate for Q_sca unless Q_ext and Q_abs
      are nearly the same (i.e., A is small).  Therefore, the following
      formula is used for the best estimate of Q_sca:

          Q_sca = [Q_s1 + Q_s2*(A/.03)**2 ]/[1.+(A/.03)**2]

   This is a rather arbitrary formula which smoothly goes from Q_sca=Q_s1 when
A << .03, to Q_sca = .0009*Q_s1+.9991*Q_s2 when A = 1.  The transition
value A=.03 was chosen via the following reasoning.  We suppose that normally
Q_ext and Q_abs will both be computed to an accuracy of approximately 0.01%
for a solution converged with an error criterion ERR=1.E-5; if so, then when
A>.03 their difference Q_s2 would be evaluated with an error of less than
approximately sqrt(2)*.01%/.03 = 0.5% .  Thus for A >> .03 we heavily weight
Q_s2.  Under normal circumstances when A is small the scattering pattern will
be approximately dipolar, in which case the evaluation of Q_s1 should be quite
accurate.  Thus for A << .03 we heavily weight Q_s1.

                             2. Minor Changes

For the benefit of users wishing to generate their own "shape.dat" files,
subroutine REASHP (file reashp.f) has been modified so that, as provided,
it will read "shape.dat" files of the form generated by the various shape
generating routines called by CALLTARGET.  This has no effect except when
the "FRMFIL" option is specified in "ddscat.par".

                                 3. Bugs

The first release of DDSCAT.4b (93.03.12) contained a typo in the
routine EXTEND.  This was corrected 93.03.13.

Ralf Stognienko has reported a bug when using INIT=1 to start the CCG
iteration with |X0> obtained by fourth-order expansion in dipole polarizability.
This problem was created when changing the code to use FFT methods.
The problem was corrected 93.05.17 by a minor change in routine ddaccg.f .

Calculated values of the orientationally-averaged backscattering cross
section Q_bk (in the files waarbbori.avg) were incorrect due to failure
to reset sums to zero for second and subsequent targets.  This was found
and DDSCAT.f corrected 93.06.03 by BTD.

Here in Princeton we ran into a problem with appears to be due to a "bug"
in either SunOS 4.1.1 or in the Sun Fortran compiler (version 1.4).  The
symptom is that when compiled on some of our Sparcstations, the value of
IDVOUT was not properly initialized in spite of a DATA statement in
DDSCAT.f  This problem was only observed on machines which were using the
Sun "tmpfs" software, which allows cohabitation of /tmp and swap on the
same disk partition, so it is possible that the problem is with tmpfs.
In any case, one line of code (re-)initializing IDVOUT was added 93.09.28
by BTD; the problem appears to have been suppressed.

Problems were encountered on Silicon Graphics systems when calculating for
more than one value of orientation angle PHI.  This has been eliminated by
a SAVE statement in subroutine GETFML (BTD, 93.11.23).  Additional problems
were encountered on Silicon Graphics systems when calculating for more than
one wavelength.  This has been eliminated by including additional variables
in the SAVE statement in subroutine DIELEC (BTD, 93.12.06).

Problem concerning option THRELL was corrected in subroutine TARGET (BTD,
93.12.14).

Problem with computation of PPOL (degree of linear polarization of scattered
light for incident unpolarized light) was communicated by M. Wolff (U of
Wisconsin) 93.12.13.  Incorrect expression was being used by DDSCAT to
compute this quantity, leading to error except for problems with sufficient
symmetry that |f_12|^2=|f_21|^2.  DDSCAT was corrected to use correct
expression (BTD, 93.12.15).

At this time there are no known bugs in DDSCAT.4b.1 -- if you encounter any,
please communicate them to draine@astro.princeton.edu