Scattering by two spheres in contact: comparisons between discrete dipole approximation and modal analysis

TECHNICAL NOTE Scattering by two spheres in contact: comparisons between discrete-dipole approximation and modal analysis Piotr J. Flatau, Kirk A. Fuller, and Daniel W. Mackowski This paper applies two different techniques to the problem of scattering by two spheres in contact: modal analysis, which is an exact method, and the discrete-dipole approximation (DDA). Good agreement is obtained, which further demonstrates the utility of the DDA to scattering problems for irregular particles. The choice of the DDA polarizability scheme is discussed in detail. We show that the lattice dispersion relation provides excellent improvement over the Clausius-Mossoti polarizability parameterization. Key words: Scattering, irregular particles. 1. Introduction The discrete-dipole approximation (DDA) developed by Purcell and Pennypacker1 is a flexible and general technique for calculating the scattering and absorp­ tive properties of particles of arbitrary shapes. This method is the subject of active research.2-4 The DDA is especially suited to scattering by irregular particles characterized by medium values of the size parameter, x = 2πaeq/λ, where aeq is the radius of a sphere with a volume equivalent to that of the nonspherical scatterer and λ is the wavelength of the electromagnetic wave. There is a need for compari­ son of the DDA with exact solutions in order (1) to find the number of dipoles needed to model correctly the shape of a particle in the DDA approximation, (2) to decide on numerical convergence criteria, and (3) to find a proper dipole polarizability scheme through modification to the Clausius-Mossoti relationship. Also, since the DDA solution often involves an iterative solution of a linear problem with matrices of the order of 105, it reassuring to see that the method compares well with the exact solution. P. J. Flatau is with the Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California 92093- 0221; K. A. Fuller is with the Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado 80523; and D. W. Mackowski is with the Department of Mechanical Engineering, Auburn University, University, Alabama 36849. Received 28 September 1992. 0003-6935/93/183302-04$06.00/0. © 1993 Optical Society of America. 3302 APPLIED OPTICS / Vol. 32, No. 18 / 20 June 1993 An arbitrary cluster of spheres5-6 can be solved exactly by the multipole expansion method. Such a modal analysis provides an independent test of scatter­ ing by nonspherical particles. In a recent report, Wiscombe and Mugnai7 state: "Of great interest, perhaps, is the solution ... for an arbitrary cluster of spheres; indeed, it seems... that sphere clusters would be an excellent archetype for general nonspher­ ical particles. Apparently, however, the solution ... is nightmarish to put into practice." In agree­ ment with the first part of this quote, we use a cluster of spheres for testing the DDA. It seems that the practical part has been recently improved.5 Using a modified version of the work by Mackowski6 we were able to solve for a relatively large cluster of spheres. In this paper, the multipole solution for two spheres is used to test the DDA. 2. Two Spheres Recent investigations have studied the general solu­ tion to Maxwell's wave equation for clusters of spheres.5,6 The analysis involves a superposition technique, in that the total solution for a scattered electromagnetic field of a particle is constructed from a superposition of individual solutions. To satisfy the boundary conditions on each sphere, addition theorems are used to transform a spherical harmonic from one coordinate to another. Such a formulation leads to a set of linear equations for the expansion coefficients of the scattered fields of the constituent spheres. In matrix form the system of equations is

expressed as where vector a defines coefficients of the scattered field for sphere i, vector pi defines incident field coefficients for a sphere i, Tji represents the transla­ tion matrix from sphere j to i, and Ns is number of spheres. The sizes of a, p, and T depend on the order of multipole expansion. Far-field scattering and cross sections are determined by a. A detailed account has been given by Mackowski.6 An exact solution for multiple spheres, which uses multipole expansion, offers a unique opportunity to compare our DDA implementation with an independent solu­ tion for nonspherical targets. Figure 1 shows two pseudospheres in contact resolved on 32 × 32 × 32 dipole grid for each of the spheres. In Fig. 2 we have compared the exact solution for scattering by two spheres (solid curve) in contact with the solution derived from DDA (crosses) for m = 1.33 + O.OOli and a scattering parameter x = 5 (results not presented) and x = 10. Exact results for S11 and S22 are compared with S11 and S22 results, which were computed by the DDA for two perpendic­ ular scattering planes. For x = 5, DDA matches the exact solution well, and on a plot scaled as in Fig. 2 it matches exactly and DDA results would overlap. For x = 10 and orientation α = 0 (Fig. 2) the overall agreement is good although errors can be seen in the side and backscatter directions. In Section 3 we perform a more quantitative error analysis by show­ ing fractional error plots for several values of equiva­ lent size parameter. Fig. 1. Two pseudospheres composed of 2 × 32 × 32 × 32 point dipoles. Light is travelling along the x axis. Spheres are rotated (α = 30). Fig. 2. Scattering by a sphere with refractive index m = 1.33 + 0.001i for x = ka = 10 and α = 0. Exact results for S11 and S22 are compared with S11 and S22 results that were computed by the DDA for x = 10. 3. Choice of Polarizability Definition: Comparisons with Two-Sphere Cluster Since DDA results depend directly upon the dipole polarizabilities, it is important to determine the optimal method for choosing such polarizabilities. Recently Draine and Goodman4 established the lat­ tice dispersion relation (LDR) method for prescribing the dipole polarizabilities. They solved a closely related problem: choosing the dipole polarizabilities such that an infinite lattice of point dipoles will propagate electromagnetic plane waves with the same dispersion relation as a medium of specified dielectric function ε. In order to compare their new prescription for polarizability with those proposed previously, they report the results of extensive DDA calculations for absorption and scattering by single spheres, and compare exact Lorenz-Mie theory results with DDA calculations. They find that for moderate | ε — 1| their prescription for the dipole polarizabilities gener­ ally provides more accurate results for scattering and absorption by spheres; this superiority is presumed to extend to nonspherical targets. The unique aspect of this prescription, which they were not able to test thoroughly, is that it depends not only on refractive index m but also on the direction of propagation and the polarization state of the incident radiation. In this section we extend their analysis to include comparisons with the exact solution for non- 20 June 1993 / Vol. 32, No. 18 / APPLIED OPTICS 3303

spherical particles composed of two spheres in contact. We conclude that, indeed, the DDA calculations of absorption and scattering should be based on the LDR prescription for the dipole polarizabilities. The Draine and Goodman4 polarizability can be written as with where α(0) is the Clausius-Mossotti polarizability: Here, mi is the refractive index at lattice site i, and the dependence on direction and polarization enters solely through the quantity S. There are two other choices of polarizability used in the current literature on the subject. Goedecke and O'Brien8 and Hage and Greenberg9 showed that the fundamental DDA equation can be derived from a simple discretization of an integral formulation of the scattering problem, the digitized Green's function/volume integral equa­ tion formulation (DGF/VIEF) method. Using the optical theorem, Draine10 showed that when k0d was finite the polarizabilities αi should include a radiative reaction correction [the Clausius-Mossoti plus radia­ tive reaction (CMRR) method]. Figure 3 presents the comparison of a two-sphere, multipole solution with the DDA for α = 30, which corresponds to the configuration in Fig. 1. Results are shown for three different prescriptions for the dipole polarizabilities: CMRR, DGF/VIEF, and LDR as functions of size parameter x = k0aeff. Refractive index m = 1.33 + 0.01i is used and the dipoles are placed on a 2 × 32 × 32 × 32 cubic lattice. For the calculations light propagates along the x axis and is unpolarized. The LDR method is superior to DGF/VIEF and CMRR for a broad range of size parameter values. For α = 30, the LDR method gives excellent agreement, which indicates that the directional corrections in Eq. (2) are indeed an im­ provement. Figure 4 presents a comparison of a two-sphere multipole solution with the DDA for the same physi­ cal parameters as in Fig. 3 but for fixed x = 5 with α as a function of orientation. Again the LDR is consis­ tently better in comparison with the other two schemes. The fractional error for scattering effi­ ciency (Fig. 4) is close to 1% in comparison with the 3304 APPLIED OPTICS / Vol. 32, No. 18 / 20 June 1993 Fig. 3. (a) Qsca and Qabs for two spheres of refractive index m = 1.33 + 0.0li in contact, (b) fractional error in computed value of Qsca. (c) Qabs for the N = 2 × 34,512 two pseudospheres of Fig. 1. Results are shown for three different prescriptions for the dipole polarizabilities: CMRR, VIEF, and LDR as functions of x = k0aeff, α = 30. multipole solution, and for absorption efficiency (re­ sults not presented) the error is less than 3%. 4. Summary This paper presents further evidence that DDA is a sound method for the study of electromagnetic scatter- Fig. 4. (a) Dependence of Qsca on orientation for the two spheres of Fig. 3 with x = 5, (b) fractional error in computed value of Qsca as a function of a for the two pseudospheres of Fig. 1.

ing by irregular particles.11,12 The choice of the polarizability scheme is justified on the basis of the multipole solution for a cluster of two spheres in contact. Results show that the DDA, together with the conjugate-gradient fast-Fourier-transform algo­ rithm, makes possible accurate calculations of scatter­ ing and absorption by dielectric targets for a medium range of size parameters. It is shown that the LDR polarizability works well in the case of nonspherical particles. This extends recent results of Draine and Goodman,4 wherein only spherical targets were tested. This research was supported by U.S. Air Force grant AFOSR-91-0269 and National Science Founda­ tion grant ATM-8812353. Computations were per­ formed on the National Center for Atmospheric Research CRAY-YMP computer, and the workstation was supported by the U.S. Air Force Office of Scien­ tific Research. The National Center for Atmo­ spheric Research is partially supported by the Na­ tional Science Foundation. References 1. E. M. Purcell and C. R. Pennypacker, "Scattering and absorp­ tion of light by nonspherical dielectric grains," Astrophys. J. 186,705-714(1973). 2. P. J. Flatau, G. L. Stephens, and B. T. Draine, "Light scattering by rectangular solids in the discrete-dipole approxi­ mation: a new algorithm exploiting the block-Toeplitz structure," J. Opt. Soc. Am. A 7, 593-600 (1990). 3. J. J. Goodman, B. T. Draine, and P. J. Flatau, "Application of fast-Fourier-transform techniques to the discrete-dipole approximation," Opt. Lett. 16, 1198-1200 (1991). 4· B. T. Draine and J. J. Goodman, "Wave propagation on a polarizable point lattice and the discrete dipole approximation," Astrophys. J. (to be published). 5. K. A. Fuller, "Optical resonances and two-sphere systems," Appl. Opt. 30, 4716-4731 (1991). 6. D. W. Mackowski, "Analysis of radiative scattering for multi­ ple sphere configurations," Proc. R. Soc. London Ser. A 433, 599-614 (1991). 7. W. J. Wiscombe and A. Mugnai, Single Scattering from Nonspherical Chebyshev Particles: A Compendium of Calcu­ lations, NASA Ref. Pub. 1157 (Goddard Space Flight Center, NASA, Greenbelt, Maryland, 1986). 8. G. H. Goedecke and A. G. O'Brien, "Scattering by irregular inhomogeneous particles via the digitized Green's function algorithm," Appl. Opt. 27, 2431-2438 (1988). 9. J. I. Hage and J. M. Greenberg, "A model for the optical properties of porous grains," Astrophys. J. 361, 251 (1990). 10. B. T. Draine, "The discrete-dipole approximation and its application to interstellar graphite grains," Astrophys. J. 333, 848-872 (1988). 11. The FORTRAN program DDSCAT.4b is available from P. J. Flatau on request. Direct queries to Internet address pflatau@macao.ucsd.edu or drain@astro.princeton. edu. 12. P.J. Flatau, "Scattering by irregular particles in anomalous diffraction and discrete dipole approximations," Tech. Rep. 517 (Colorado State University, Fort Collins, Colo. 1992). 20 June 1993 / Vol. 32, No. 18 / APPLIED OPTICS 3305