Scattering and Absorption of Light by Small Particles

[Back to B.T. Draine's home page.] Scattering and Absorption of Light by Small Particles

The Discrete Dipole Approximation for Scattering and Absorption of Light by Irregular Particles

DDSCAT, a Fortran code for calculating scattering and absorption of light by irregular particles, has been jointly developed by Bruce T. Draine (Dept. of Astrophysical Sciences, Princeton University) and Piotr J. Flatau (Scripps Institution of Oceanography, UCSD). The current version is DDSCAT 7.0, which supersedes previous versions. DDSCAT 7.0 can calculate scattering and absorption by isolated particles (e.g., dust grains) but can also calculate scattering and absorption by one- and two-dimensional arrays of "target unit cells" that may have complex geometries; this has obvious applications to photonics and studies of arrays of nanostructures. DDSCAT 7.0 is publicly available (see below), and is now considered to be the standard version of DDSCAT. If you choose to use it, please send email to <draine@astro.princeton.edu> "registering" as a user; registered users of DDSCAT will be notified when updates to the code are made.

User Guide for DDSCAT.7.0

An extensive User Guide is available: "User Guide to the Discrete Dipole Approximation Code DDSCAT.7.0", by B.T. Draine and Piotr J. Flatau. A copy of this document (postscript or pdf) can be obtained from astro-ph/0809.0337 and can be cited as

Draine, B.T., and Flatau, P.J. 2008, "User Guide to the Discrete Dipole Approximation Code DDSCAT 7.0", http://arxiv.org/abs/astro-ph/0809.0337v1.

User Guide for DDSCAT.6.1

We continue to make available the UserGuide to the previous version of DDSCAT: "User Guide to the Discrete Dipole Approximation Code DDSCAT.6.1", by B.T. Draine and Piotr J. Flatau. A copy of this document (postscript or pdf) can be obtained from astro-ph/0409262 (gzipped postscript) or astro-ph/0409262 (pdf) and can be cited as

Draine, B.T., and Flatau, P.J. 2004, "User Guide to the Discrete Dipole Approximation Code DDSCAT 6.1", http://arxiv.org/abs/astro-ph/0409262v2.

Downloading the DDSCAT 7.0 Code

A single distribution is provided for DDSCAT 7.0 -- by appropriate editing of the Makefile, this distribution can be used to generate executables using either single- or double-precision arithmetic; either without or with MPI capability; either without or with OPENMPI support; and either without or with the Intel MKL library. To download, click here: ddscat7.0.7.tgz: gzipped tarfile containing complete source code and documentation for DDSCAT 7.0, including the programs DDfield and CALLTARGET..

Downloading the DDSCAT Code

Two distributions of DDSCAT 6.1 are available. Both include bug fixes (see below) up to and including 2008.06.23):

The "plain" distribution. This is easiest to install. It does not include support for MPI, FFTW, or netCDF, but many users do not require these. To download, click here: ddscat6.1_plain.tgz: gzipped tarfile containing complete source code and documentation for the plain distribution of DDSCAT 6.1

The "full" distribution. This includes support for MPI, FFTW, and netCDF. To download, click here: ddscat6.1_full.tgz: gzipped tarfile containing the complete source code and documentation for the full distribution of DDSCAT 6.1

History of recent releases:

DDSCAT 6.1 was released 2004 September 10.

DDSCAT 6.1 introduces two versions of DDSCAT:
- the "plain" version for new users or users who do not require MPI, FFTW, or netCDF
- the "full" version, with support for MPI, FFTW, and netCDF for users who require one or more of these features.
DDSCAT 6.1 implements an automatic procedure for calculating angular averages for the scattered radiation. User now specifies a parameter ETASCA. The number of scattered directions is proportional to [(3+x)/ETASCA]^2 , where x=2*pi*aeff/lambda is the scattering parameter. The number of scattering angles used therefore depends on the ratio of target size to wavelength.
DDSCAT 6.1 includes new target options: ANIREC, for calculating scattering by anisotropic rectangular prisms; CYLCAP, for calculaing scattering by cylinders with hemispherical endcaps.
DDSCAT 6.1 now calculates the second moment in addition to the first moment of cos(theta) for the scattered radiation, where theta is the scattering angle.
Bug fixes:
- 2004.12.29: An error in routine writesca.f was found and corrected. The error affected output to the files waarbbkccc.sca (written when IWRKSC=1 is specified in ddscat.par), causing incorrect values of the Muller matrix elements S_ij to be written for the individual orientations when doing an orientational average over more than one target orientation. The bug did not affect any of the other calculated results. In particular, the orientationally-averaged Muller matrix elements (written to the output files waarbbori.avg) were correct, and the total and scattering cross sections for the individual target orientations were also correct. The bug was in the output routine WRITESCA (file writesca.f ), and consisted simply in printing the wrong variable. Users can download new versions of the complete "plain" distribution ddscat6.1_plain.tgz or the "full" distribution ddscat6.1_full.tgz of DDSCAT 6.1 . The new versions of version.f and writesca.f can be extracted from the gzipped tarfile with the command "tar xvfz ddscat6.1_plain.tgz src/version.f src/writesca.f" or "tar xvfz ddscat6.1_full.tgz src/version.f src/writesca.f" (quotation marks not included).
- 2005.01.17: The files mpi_bcast_char.f, mpi_bcast_int.f, mpi_bcast_real.f, mpi_bcast_cplx.f, and mpi_bcast_int2.f, which had been inadvertently omitted from ddscat6.1_full.tgz, are now included in ddscat6.1_full.tgz
- 2006.09.13: Error in tarcyl.f fixed. NAT was inadvertently not initialized to zero at the beginning of the calculation (most compilers set it to zero anyway).
- 2006.11.28: An error in routine writesca.f was found and corrected. The error resulted in incorrect values for linear polarization P to be written to files waaarbbkccc.sca waaarbbori.avg . All cross sections and Mueller matrix elements S_{ij} were evaluated and printed correctly. If the average being taken was over random orientations, you can recalculate the linear polarization from S_{11} and S_{12} in existing output files by simply taking P= - S_{12}/S_{11}.
- 2008.06.23: An error in routine tarcylcap.f was corrected. The array ICOMP had inadvertently not been set for lattice sites in the cylinder "caps". All calculations carried out for target option CYLCAP were unfortunately incorrect. Thanks to Stuart Prescott and Paul Mulvaney for reporting this bug on 2008.06.23.

DDSCAT.6.0 was released 2003 September 2.

DDSCAT.6.0 provides support for MPI (Message Passing Interface), enabling parallel calculations for different target orientations;
DDSCAT.6.0 provides support for FFTW (Fastest Fourier Transform in the West), as an alternative to the GPFA (Generalized Prime Factor Algorithm) code in DDSCAT.5a;
DDSCAT.6.0 allows the user to select which elements of the Mueller scattering matrix are to be printed out;
DDSCAT.6.0 provides a new target option -- LYRSLB -- to generate a layered slab.
Bug fixes:
- 2003.10.30: Corrected typo in routine writenet.f (used for writing netCDF output file dd.cdf)
- 2004.02.25: Corrected typo in routine tarprsm.f (Note: version of tarprsm.f in DDSCAT.5a10 release was correct).

DDSCAT.5a10 was released 2000 August 9. It provides a new target option -- NSPHER -- to create targets consisting of the unions of N spheres (possibly overlapping) of arbitrary sizes and locations. It also uses a more recent version of the LAPACK code used by subroutine PRINAXIS.

A new target option -- PRISM3 -- to generate a triangular prism was added to DDSCAT.5a10 on 2002.02.12
Bug fixes:
- 2001.04.21: Bug in subroutine tarnsp.f (used by NSPHER option) has been fixed. This bug would not have affected most prior calculations with option NSPHER except in special case where first sphere extends further in +y direction in target coordinates than any of the other N-1 spheres.
- 2003.06.06: Bug in routine rotate.f has been fixed. Prior to this fix, the routine miscalculated the initial target orientation when target axis A1 in the "Target Frame" had A1(3) < 0 and/or (A1(2)*A2(3)-A1(3)*A2(2)) < 0 . Under these circumstances, the routine rotated the target to orientation angles (PHI + constant) and (BETA + constant) instead of PHI and BETA. This should not have had any effect on orientational averaging if the user specified appropriate ranges for the angles PHI and BETA. Note also that many of the target options (e.g., ELLIPS) provide axes A1 and A2 for which this bug did not apply.

DDSCAT.5a9 was released 1998 December 23. It fixes a bug in DDSCAT.5a8, which resulted in incorrect evaluation of elements of the Mueller scattering matrix (other than S_11, which was correct) for scattering planes other than phi=0.

A new version of orient.f was released 1999 March 16. Orientational averages are now evaluated as described in the UserGuide; prior to this date averaging over cos(Theta) was evaluated by dividing the range of cos(Theta) into NTHETA equal intervals, evaluating the scattering at the midpoint of each interval, and taking the mean. With the new version of orient.f, Simpson's rule is now used for the quadrature when an odd value of NTHETA is specified by the user.

Benchmark Calculations

Draine (2000) has published benchmark calculations for scattering by tetrahedral targets. Two target refractive indices have been considered: m=1.33+0.01i and m=1.70+0.10i . Scattering was calculated for one target orientation and two incident polarizations. In each case solutions were iterated until the DDA equations were satisfied with a fractional error < 1.e-5 . CPU times are reported below for a 600 MHz Pentium III system running Red Hat Linux 6.2, using the Portland Group pgf77 compiler.

CPU time (s) on 600 MHz Pentium III System
m=1.33+0.01i tetrahedron

N x=5 x=10 x=15

4030 31. -- --

13426 141. 267. --

31598 356. 639. 1106.

61432 792. 1480. 2476.

105832 1298. 2292. 4226.

m=1.7+0.1i tetrahedron
N x=5 x=10

13426 377. --

31598 939. 2104.

61432 2115. 4429.

105832 3585. 7335.

CPU time (s) on 600 MHz Pentium III System
m=1.33+0.01i tetrahedron
N	x=5	x=10	x=15
4030	31.	--	--
13426	141.	267.	--
31598	356.	639.	1106.
61432	792.	1480.	2476.
105832	1298.	2292.	4226.

m=1.7+0.1i tetrahedron
N	x=5	x=10
13426	377.	--
31598	939.	2104.
61432	2115.	4429.
105832	3585.	7335.

For users wishing to repeat one or more of these calculations using DDSCAT.5a10 (which may be useful to reassure you that DDSCAT.5a10 is working properly on your computer), sample ddscat.par files are available for downloading:

ddscat.par file for m = 1.33+0.01i, N = 4030 dipole tetrahedron, x = 5.
ddscat.par file for m = 1.33+0.01i, N = 13426 dipole tetrahedron, x = 5, 10
ddscat.par file for m = 1.33+0.01i, N = 31598 dipole tetrahedron, x = 5, 10, 15
ddscat.par file for m = 1.33+0.01i, N = 61432 dipole tetrahedron, x = 5, 10, 15
ddscat.par file for m = 1.33+0.01i, N = 105832 dipole tetrahedron, x = 5, 10, 15
m1.33_0.01 refractive index table for m = 1.33+0.01i
ddscat.par file for m = 1.70+0.10i, N = 13426 dipole tetrahedron, x = 5
ddscat.par file for m = 1.70+0.01i, N = 31598 dipole tetrahedron, x = 5, 10
ddscat.par file for m = 1.70+0.10i, N = 61432 dipole tetrahedron, x = 5, 10
ddscat.par file for m = 1.70+0.10i, N = 105832 dipole tetrahedron, x = 5, 10
m1.70_0.10 refractive index table for m = 1.70+0.10i

Please consult the DDSCAT UserGuide (see above) for an explanation of the parameters in the ddscat.par file.

For additional information on other techniques for numerical calculation of electromagnetic scattering by finite targets, see also the SCATTERLIB library of light scattering codes maintained by Piotr Flatau.

Back to B.T. Draine's home page.