DDA for size parameter 100, Yurkin

279

Can the Discrete Dipole Approximation simulate scattering of particles with size parameter equal to 100? Maxim A. Yurkin,1,2 Valeri P. Maltsev,2,3 and Alfons G. Hoekstra1 1

Faculty of Science, Section Computational Science, of the University of Amsterdam, Kruislaan 403, 1098SJ, Amsterdam, The Netherlands, tel: +31-20-525-7530, fax: +31-20-525-7490 2 Institute of Chemical Kinetics and Combustion, Siberian Branch of the Russian Academy of Sciences, Institutskaya 3, Novosibirsk 630090 Russia, tel: +7-3832-333240, fax: +7-3832-342350 3 Novosibirsk State University, Pirogova Str. 2, 630090, Novosibirsk, Russia email: [email protected], [email protected] Abstract This study is devoted to exploring the applicability bounds of a modern Discrete Dipole Approximation (DDA) code running on a state-of-the-art massively parallel supercomputer. A number of DDA simulations of light scattering by spheres with different size parameters (x) and refractive indices (m) are performed and compared with the exact solution. Results of DDA performance and simulation errors are presented. We show that DDA is at least applicable in the range of refractive index from 1 to 2, and size parameter up to 130 (for m < 1.2). We are currently exploring larger values of m.

1

Introduction

The methods to solve the single light scattering problem can be divided into approximate and exact [1]. Approximate methods include, among others, Rayleigh and Geometrical Optics (GO) approximations, which are applicable, respectively, to particles small and large compared to the wavelength [2]. Exact methods are numerous, but only few of them are applicable to scatterers with complex internal structure. Those are volume integral methods, which solve the Maxwell’s equations either in the time domain (Finite Difference Time Domain, FDTD) or frequency domain (Discrete Dipole Approximation, DDA [3]). In this study we explore the applicability of DDA to simulate light scattering in intermediate size range (to order ten times the wavelength), where no approximation applies. Even when the size of the scatterer is large enough for GO, the internal structure with characteristic sizes comparable to the wavelength may greatly complicate the application of GO [4]. All methods, which are capable to simulate light scattering by complex inhomogeneous objects in this size range, face extreme computational complexity with increasing scatterer’s size, because they divide the scatterer in subvolumes that should be much smaller than the wavelength. We have recently demonstrated that DDA is applicable for biological particles in suspension with size parameters and relative refractive indices in the range 30 < x < 130 and 1.02 < m < 1.2, respectively [5]. In this study we extend that work to higher refractive indices (up to 2) to include many more relevant applications.

2

Simulation method

Our code, the Amsterdam DDA (ADDA), is capable of running on a cluster of computers (parallelizing a single DDA computation), which allows us to use practically unlimited numbers of dipoles, since we are not limited by the memory of a single computer [6, 7]. In this study we use the Lattice Dispersion Relation

Ninth Conference on Light Scattering by Nonspherical Particles

280

10

5

m

=

1.02

10

-3

1.05 4

1.1 1.2

10

10

relative error of

time, s

Qext

10

3

2

10

10

-4

-5

m 10

-6

=

1.02 1.05 1.1 1.2

10

10

1

40

60

80

100

120

size parameter, x

Figure 1: DDA simulation times for spheres with different x and m (on 64 P4-3.4 GHz processors). Two largest computations times for m = 1.2 are estimated.

-7

40

60

80 size parameter,

x

100

120

Figure 2: Relative errors of Qext for spheres with different x and m.

(LDR) [8] for dipole polarizability. We use two iterative methods: Bi-Conjugate Gradient STABilized (BiCGSTAB) [9] and Quasi Minimal Residual (QMR) [10]. The former performed better for very large indexmatching particles, described in Sect. 3. All the DDA simulations were carried out on the Dutch national computer cluster LISA [11]. Since the DDA method is equally applicable to particles of any shape and internal structure, we test it on simple objects, for which an exact analytical solution is available, and then extend the conclusions to arbitrary shaped scatterers. We use homogenous spheres as test objects and compare DDA simulations with Mie solutions. As a result, we obtain DDA performance statistics (hardware requirements, simulation time) and accuracy of both integral characteristics (cross sections and asymmetry parameter) and angular dependencies of all Mueller matrix elements for different x and m.

3

First results

Here, we present results obtained for the refractive index range of biological particles 1.02 < m < 1.2. Figure 1 shows DDA simulation times for different spheres (obtained on 64 P4-3.4 GHz processors). The size of a single dipole has been kept constant at λ/12. The total number of dipoles for the largest simulation was 6 · 107 . One can see that simulation time scales approximately cubical with the size parameter and rapidly increases with refractive index. Relative errors of Qext for these simulations are presented in Fig. 2. They are fairly small. Results for the asymmetry parameter are similar (data not shown). The angular dependence of S 11 is presented for two particular cases: x = 100, m = 1.02 and x = 130, m = 1.2 in Fig. 3 and Fig. 4, respectively. Those are the cases that give the largest errors in the range of biological particles. One can see that DDA results are in good agreement with the exact solution, however relative errors can be significant at specific angles (near the deep minima). Results for other Mueller matrix elements show the same trends (data not shown). At the conference we will show results for refractive index up to 2, and for size parameter up to 100. Simulation times are expected to be as large as one week on 128 P4-3.4 GHz processors. This number tells us that such massive DDA computations are only useful to serve as benchmark studies, i.e. providing data to test the quality of other, less expensive approximate methods.

DDA for size parameter 100, Yurkin

10

281 108

8

Mie

10

6

107

DDA

DDA

10

10

10

10

2

105 104

0

103

-2

102

0

30

60

90

120

150

180

scattering angle,

Figure 3: S 11 (θ) for a sphere with x = 100, m = 1.02. Both DDA (solid line) and Mie results (dotted line) are shown.

4

6

4

S11

S

11

10

Mie

0

30

60

90

120

150

180

scattering angle,

Figure 4: S 11 (θ) for a sphere with x = 130, m = 1.2. Both DDA (solid line) and Mie results (dotted line) are shown.

Conclusion

DDA is capable of simulating light scattering by large particles with m < 2, however it requires very large, computation times even on state-of-the-art massively parallel supercomputers. For biological particles (m < 1.2) DDA is proven to be applicable up to x = 130 with overall satisfactory accuracy. Accuracy of integral scattering quantities is very good, while angle-resolved quantities show larger relative errors (especially in minima). At the conference we will present the bounds of size parameter for practical applicability of DDA (with current code and hardware) for m < 2. We will show DDA performance and accuracy of both integral characteristics (cross sections and asymmetry parameter) and angular dependencies of all Mueller matrix elements.

References [1] F. M. Kahnert, “Numerical methods in electromagnetic scattering theory,” J. Quant. Spectrosc. Radiat. Transf. 79–80, 775–824 (2003). [2] C. F. Bohren and D. R. Huffman, Absorption and scattering of Light by Small Particles, (Wiley, New York, 1983). [3] B. T. Draine and P. J. Flatau, “Discrete-dipole approximation for scattering calculations,” J. Opt. Soc. Am. A11, 1491–1499 (1994). [4] A. Macke, “Monte-Carlo calculations of light scattering by large particles with multiple internal inclusions,” In: Light Scattering by Nonspherical Particles, Theory, Measurements, and Applications, M. I. Mishchenko, J. W. Hovenier, and L. D. Travis, (eds.), 309–322 (Academic Press, 2000). [5] M. A. Yurkin and A. G. Hoekstra, “Capabilities of the discrete dipole approximation for simulation light scattering by biological cells,” In: Proceedings of Optics of Biological Particles, 19–20 (Novosibirsk, Russia, 2005).

282

Ninth Conference on Light Scattering by Nonspherical Particles

[6] A. G. Hoekstra, M. D. Grimminck, and P. M. A. Sloot, “Large scale simulations of elastic light scattering by a fast discrete dipole approximation,” Int. J. Mod. Phys. C9, 87–102 (1998). [7] M. A. Yurkin, K. A. Semyanov, P. A. Tarasov, A. V. Chernyshev, A. G. Hoekstra, and V. P. Maltsev, “Experimental and theoretical study of light scattering by individual mature red blood cells with scanning flow cytometry and discrete dipole approximation,” Appl. Opt. 44, 5249–5256 (2005). [8] B. T. Draine and J. J. Goodman, “Beyond Clausius-Mossotti — wave-propagation on a polarizable point lattice and the discrete dipole approximation,” Astrophys. J. 405, 685–697 (1993). [9] R. Barrett, M. Berry, T. F. Chan, J. Demmel, J. Donato, J. Dongarra, V. Eijkhout, R. Pozo, C. Romine, and H. A. van der Vorst, Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods, (SIAM, 1994). [10] R. W. Freund, “Conjugate gradient-type methods for linear-systems with complex symmetrical coefficient matrices,” SIAM J. Sci. Stat. Comp. 13, 425-448 (1992). [11] “Description of the national computer cluster Lisa,” http://www.sara.nl/userinfo/lisa/ description/index.html (2005).

Can the Discrete Dipole Approximation simulate ...

of DDA performance and simulation errors are presented. We show ... Our code, the Amsterdam DDA (ADDA), is capable of running on a cluster of computers ...

237KB Sizes 1 Downloads 230 Views

Recommend Documents

Capabilities of the Discrete Dipole Approximation for ...
solution of the system of 3Nd complex linear equations, where. Nd is the number of .... [9] M. A. Yurkin, V. P. Maltsev, and A. G. Hoekstra, “The discrete dipole.

Features and capabilities of the discrete dipole approximation code ...
4Computational Science research group, Faculty of Science, University of Amsterdam, ... M.Y. is supported by the program of the Russian Government “Research ... [2] M. A. Yurkin and A. G. Hoekstra, “The discrete dipole approximation: an ...

The discrete-dipole-approximation code ADDA ...
Feb 1, 2011 - line options is given in the relevant parts of this paper. The full list ... However, simulation data for large scatterers ..... a coarse way by bigger dipoles (cubes), but then use ..... storage space for the Fourier-transformed matrix

Accuracy of the discrete dipole approximation for gold ...
2Novosibirsk State University, 2 Pirogova St., 630090, Novosibirsk, Russia ... Science Research Group, Faculty of Science, University of Amsterdam,. Science ...

The discrete dipole approximation for simulation of light ...
The code is written for distributed memory systems using the message passing interface (MPI).2 ... is incident electric field, ¯Gij is the free-space Green's tensor. (complex symmetric), and .... to the ''rule of thumb'' (see main text for explanati

Features and capabilities of the discrete dipole approximation code ...
ADDA [1] is an open-source parallel implementation of the discrete dipole approximation. (DDA [2]), capable of simulating light scattering by particles of arbitrary ...

Application of the discrete dipole approximation to ...
Application of the discrete dipole approximation to extreme refractive indices: filtered coupled dipoles ... routinely simulated using modern desktop computers.

Convergence of the discrete dipole approximation. I ...
of a dipole d when the latter is in the range of DDA applicability. Moreover ... In a follow-up paper18 .... where d Ed corresponds to Eq. (25) or (26), and the er- ror.

Convergence of the discrete dipole approximation. II ...
to O N log N by advanced numerical techniques.2,5 Still, the usual application strategy for DDA is single computa- tion, where a discretization is chosen on the ...

Accuracy of the discrete dipole approximation for ...
a Institute of Chemical Kinetics and Combustion SB RAS, Institutskaya 3, ...... [11] H. Yoo, J. E. Millstone, S. Li, J. Jang, W. Wei, J. Wu, G. C. Schatz, and C. A..

Rigorous and Fast Discrete Dipole Approximation ... - ACS Publications
Nov 30, 2015 - problems is possible10,17,25−30 but introduces two additional issues. The first one is the ..... large computer cluster. OpenCL mode allows ... laptop processor (Intel Core i7-2630QM), while the extrapolated results with ...

Comparison between discrete dipole implementations ...
The geometry of the scatterer is read from a file and all the parameters of .... unlimited number of dipoles, since ADDA is not limited by the memory of a single ... symmetry of the interaction matrix is used to decrease storage requirement of its ..

Comparison between discrete dipole implementations ...
in astronomy and in some technological applications has greatly increased in the last years. ..... [16] or the more advanced package 'fastest Fourier transform in the west' (FFTW) [26]. ...... science and is owned by the Ministry of Education.

Systematic comparison of the discrete dipole ...
domain (FDTD) method for simulating light scattering of spheres in a range of size ..... Foundation through the grant "Best PhD-students of Russian Academy of ...

Current capabilities of the discrete dipole ...
3 computational box, resulting in linear system of 2⋅108 equations. .... Pentilla et al (JQSRT special issue dedicated to ELS-9). Eremin: Which numerical method ...

Does the Human Motor System Simulate Pinocchio's ... - SAGE Journals
1Institute of Neuroscience, National Yang Ming University, Taipei, Taiwan; 2Laboratories for Cognitive Neuroscience,. National Yang Ming University, Taipei, ...

How to Simulate the Whispering-Gallery-Modes of ...
Upon introducing a system of cylindrical coordinates (see top right Fig. ..... 27, page 205, for example) Js = n×H. The averaged-over-a-cycle power dissipated in ...

Simulate Reality for the Orchestration of Deployed ...
C.2.1 [Computer Communication Networks]: Network. Architecture and Design. General Terms. Design .... simulations through the calibration of communication mod- els with real measurements; 2) identifying and ...... ances such as microwave ovens or the

How to simulate the quasistationary state - Physical Review Link ...
Jan 21, 2005 - Departamento de Física, ICEx, Universidade Federal de Minas Gerais, 30123-970 Belo Horizonte, Minas Gerais, Brazil. Received 30 July 2004 ...

Does the Human Motor System Simulate Pinocchio's ...
sent in the individual go/no-go task because only one response code was formed and no conflict between spatial codes occurred. Address correspondence to Chia-Chin Tsai, Institute of Neurosci- ence, National Yang Ming University, 155, Sec 2, Li Long S

How to Simulate the Whispering-Gallery-Modes of ...
makes it advantageous to solve for H (or, equivalently, the magnetic flux density B = μH –with a constant global magnetic permeability μ), as opposed to the electric field ..... If the source were an infinitesimal(finite) multi-pole, then a surfa

How To Simulate 1000 Cores
2008 Hewlett-Packard Development Company, L.P.. The information contained herein is subject to change without notice. How To Simulate 1000 Cores.

Observation of the magnetic-dipole fine-structure transition in ... - naomib
based on electron spectrometry or tunable laser photodetachment threshold ... Electron spectrometry experiments do not achieve the same ultimate accuracy as ...

Observation of the magnetic-dipole fine-structure transition in ... - naomib
s-wave features, and in cases where the levels are not effectively populated in the ion source. [5]. Multiphoton approaches utilizing pulsed laser technology can largely .... extrapolation from neutral I [13, 14] by accounting for the difference in w