IAMG Proceedings. The projective method approach in the electromagnetic integral equations solver. presenting author
|
|
- Franklin Johnston
- 5 years ago
- Views:
Transcription
1 The projective method approach in the electromagnetic integral equations solver M. KRUGLYAKOV 1 AND L. BLOSHANSKAYA 2 1 Lomonsov Moscow State University, Russia m.kruglyakov@gmail.com 2 SUNY New Paltz, USA presenting author Abstract We present a new parallel (MPI+OpenMP) solver for the 3D volumetric integral equations (IE) of electrodynamics. The solver requires only half of an amount of memory compared with the other IE solvers and has high degree of parallelization. The computational experiments, including the ones performed for the high conductivity contrast problems, show solver s efficiency. The developed software is distributed under the GPLv2 license. 1 Introduction We developed a new iterative numerical solver for the 3d integral equations (IE) of electrodynamics. While the other solvers are based on the collocation method we utilize the projective method for IE solving. Instead of obtaining the general matrices as in the collocation method, we compute the symmetric and the antisymmetric matrices. This allows us to store only the upper-triangular parts of these matrices. Hence, compared to the collocation method for the same number of cells, we reduce the RAM usage almost by half. Another feature of the proposed solver is a high degree of parallelism. The computational experiments performed by Bluegene/P and Lomonosov supercomputers from Moscow State University (MSU) show that it makes the best usage of nodes for calculation at a single frequency and a single source. The maximum number of the used nodes is 2max(N x,n y ), where N x and N y are the numbers of model cells in X and Y direction respectively. This paper is organized as follows. In Section 2 we briefly describe the contracting IE (CIE) approach and the construction of the system of linear equations approximating the CIE. In Section 3 we describe the way to reduce the memory requirments by half compared to the analogus IE solvers and our approach to matrix coefficients computing. We also demonstrate the experimental results on the parallelizm efficiency. In Section 4 we provide the example of computations based on the high (more than ) conductivity contrast COMMEMI3d3 model. 2 Projection method for CIE 2.1 Contracting integral equation Let σ(m) describe the 3D conductivity distribution in space. We further assume that the electromagnteic (EM) field is induced by the external electric currents J ext. Moreover, assume that the EM fields are time dependent as e iωt, where ω is angular frequency, i = 1. Then the electrical field E and the magnetic field H give the solution of the system of Maxwell s equations curlh = σe+j ext curle = iωµ 0 H (1) E(r),H(r) 0 as r σ = σ +iωεµ 0 ISBN (DVD) 673
2 The projective method approach in the electromagnetic integral equations... σ b (z) σ a (M) T Z Figure 1: Typical model Here ε is dielectric permeability and µ 0 is magnetic permeability. For simplicity we assume that iωεµ 0 1 and that the conductivity σ is a real value function. Let T R 3 be some bounded domain and σ(m) = σ b (z) for M T and σ(m) = σ a (M) as shown in the Figure 1. In this case E(M), H(M) are expressed in terms of the integrals E(M) = E N (M)+ Ĝ E (M,M 0 ) σ (M 0 )E(M 0 )dt M0 H(M) = H N (M)+ σ (M 0 ) = σ a (M 0 ) σ b (M 0 ) T T Ĝ H (M,M 0 ) σ (M 0 )E(M 0 )dt M0 Here ĜE, ĜH are electrical and magnetic Green s tensors respectively, Pankratov et al. (1995), Dmitriev et al. (2002). The terms E N, H N are called normal electric and magnetic fields correspondingly and form the solution of the system curlh N = σ b (z)e N +J ext curle N = iωµ 0 H N (3) E N (r),h N (r) 0 as r (2) We define the operator G m E as G m E V = σ b Ĝ E [2 σ b V]+V, (4) whereĝe isanintegraloperatorfromthefirstequationin(2). TheoperatorG m E isacontracting operator in L 2 [T], see Pankratov et al. (1995). Using (2) and (4) we obtain the so-called contracting integral equation (CIE) for E: where I is the identity operator. 674 ( I G m E b a ) Ẽ = σ b E N, Ẽ = ae, a = σ a +σ b 2 σ b b = σ a σ b 2 σ b, (5)
3 2.2 Projective method approach for CIE solving Suppose the domain T is divided in nonoverlapping cells T = T k, k = 1...N and σ b (M) = σb k, σ a (M) = σa k for M T k,k = 1...N. For each cell T k we define the function W k (M) as 1,M T k W k (M) = V k, V k = dt M, k = 1...N (6) 0,M T k T k Let W N bealinearspanofthevectorfunctionsw k, W k = (W kx,w ky,w kz ),k x,k y,k z = 1...N and P N be a projection operator from L 2 [T] to W N : [ F L 2 [T] P N [F] ] N F γ = γ (M)dT M α γ k W k, α γ k = T k, (7) V k where γ = x,y,z. Note that P N = 1. Applying P N to (5) we obtain the following operator equation in W N : k=1 W P N G m b E a W = W0 W 0 = P N (8) σ b E N Since b a < 1, Gm E is a contracting operator and P N = 1, it can be easily shown that (8) has a unique solution W in W N, see Singer (2008), Kruglyakov (2011). Moreover, W approximates E with the first order of d in L 2 [T], where d = max d k, d k is a diameter of the k=1...n cell T k, Kruglyakov (2011). Using the definition of W N we can express the components of W = (W x,w y,w z ) as W γ = N U γ k W k, γ = x,y,z (9) k=1 Using (7),(8),(9) and taking into account that σ a,σ b are piecewise functions, we obtain the following system of linear equations for the coefficients U n = (U x n,u y n,u z n), n = 1...N: where U n + N ˆγ k Kk n U k = U 0 n, (10) k=1 K n k = Î + 2 σ V b kσn ˆB b n k n B n k = Ĝ E (M,M 0 )dt M dt M0 T k T n ˆγ k = σk a σb k σa k +σb k σ U 0 n n = b V n T n E N (M)dT M Note that K k n, B k n, Î, ˆγk are 3 3 matrices, Î is an identity matrix, ˆγk is a diagonal matrix. It is clear that the system (10) has a unique solution. The main goal of this paper is to present a new effective high parallel solver for this system. ISBN (DVD) 675 (11)
4 The projective method approach in the electromagnetic integral equations... 3 Implementation issues and features 3.1 Memory requirements The main challenge of using the integral equation approach for numerical solving of 3D EM problems is the necessity to solve the system of linear equations with dense matrices (10). Moreover, the challenge is not only to solve such systems, but to store their large matrices in the RAM. The standard approach, see Avdeev et al. (1997), is to use the following property of ĜE: Ĝ E (M,M 0 ) = ĜE(x x 0,y y 0,z,z 0 ) (12) To implement the method we now consider T R 3 to be a rectangular domain. As before T is divided in N = N x N y N z cells T k, k = 1,...,N, where N x,n y,n z is the number of cells in XYZ directions respectively. Suppose in addition that each T k has the same size h x h y in XY plane. Hence, B n k = Ĝ E (M,M 0 ) = B ( ) n k Ix n Ix,I k y n Iy,I k z,i n z k, (13) T n T k where Ix n,k = 1...N x, Iy n,k = 1...N y, Iz n,k = 1...N z, n,k = 1...N. Therefore B n k is a block Toeplitzmatrixinducedbytheblockvector(C y (N y 1),Cy (N y 2),...Cy N y 2,Cy N y 1 ). Eachblock C y i, i = (N y 1)...N y 1 is also a block Toeplitz matrix and is induced by the block vector (D (N i x 1),Di (N x 2),...Di N x 2,Di N ). The x 1 Di j is a 3 3 block matrix with the structure Q xx Q xy Q xz Dj i = Q(i,j) = Q yx Q yy Q yz (14) Q zx Q zy Q zz Here Q αβ are the matrices of the order N z, α,β = x,y,z, i = (N y 1)...N y 1, j = (N x 1)...N x 1. Let A be the matrix corresponding to the system of linear equations (10). We then can express A as A = S +R 1 BR 2, (15) } where S, R 1, R 2 are the diagonal matrices; B = { Bk n is the block Toeplitz matrix described above. In view of (15) it follows that we need only 36 N x N y Nz 2 16+O(N x N y N z ) bytes to store matrix A in double precision. Using the equivalence G xy = G yx this requirement can be reduced to 32 N x N y Nz 2 16+O(N x N y N z ) bytes as in Avdeev et al. (1997). Now we show how to reduce this memory requirement by half. Our result is based on the following Lemma 1. Lemma 1. If ĜE(M,M 0 ) is an electrical Green s tensor of any layered media, then in Cartesian coordinates it possesses the following symmetric and antisymmetric properties G E zx(x x 0,y y 0,z,z 0 ) = G E xz(x x 0,y y 0,z 0,z) G E zy(x x 0,y y 0,z,z 0 ) = G E yz(x x 0,y y 0,z 0,z) G E xx(x x 0,y y 0,z,z 0 ) = G E xx(x x 0,y y 0,z 0,z) G E yy(x x 0,y y 0,z,z 0 ) = G E yy(x x 0,y y 0,z 0,z) G E zz(x x 0,y y 0,z,z 0 ) = G E zz(x x 0,y y 0,z 0,z) G E xy(x x 0,y y 0,z,z 0 ) = G E yx(x x 0,y y 0,z,z 0 ) G E xy(x x 0,y y 0,z,z 0 ) = G E xy(x x 0,y y 0,z 0,z) (16) 676
5 This lemma is a trivial corollary from Lorentz reciprocity, see Ward and Hohmann (1988) and (12). Using (13) and (16) we obtain the following properties of the blocks of matrix Q: Q zx = Q T xz,q zy = Q T yz, Q xx = Q T xx,q yy = Q T yy,q zz = Q T zz, Q xy = Q T xy = Q yx = Q T yx, (17) where T indicates a matrix transpose. Therefore, we need to store only Q xz,q yz and upper diagonal parts of Q xx,q xy,q yy,q zz. Hence, we need 8 N x N y N z (2N z +1) 16 bytes to store B k n which is only one half of the memory requirements to Avdeev et al. (1997). 3.2 The computation of B k n The next computational challenge of projective approach is an evaluation of the coefficients B k n, n,k = 1...N, i.e. the double volumetric integrals of the ĜE in the RHS of (13). The components of ĜE are the improper integrals containing the Bessel functions. The integration in vertical direction is performed analytically using the fundamental function of layered media approach from Dmitriev et al. (2002). The main problem, however, is the integration over the horizontal domains. In this case we need to compute with the appropriate accuracy the fifth-order integrals over the fast-oscillating functions. To compute these integrals we change the order of integration and make the appropriate substitution. This allows us to convert the fifth-order integral to the convolution with the special kernel. Then we compute the spectrum of this kernel and build the quadrature formula based on Shannon s interpolation. It is important to emphasize that both the nodes and the weights in the obtained formula significantly depend on the integration domains. At the same time their computational cost is independent of the integration domains. Moreover, the integration over different horizontal domains is completely data independent, which we use in our parallel algorithm. 3.3 Parallelization The most essential part of any iterative method for solving a system of linear equations is the matrix-vector multiplication. Since matrix B is a block Toeplitz matrix, we can use the 2D Fast Fourier Transform (FFT) to speed up this operation, see Avdeev et al. (1997). Therefore, instead of matrices Q(i,j) we need to store their 2D discrete Fourier transformations Q(i,j). It is obvious that storing Q(i,j) requires the same amount of memory as storing Q(i,j) itself. The algorithm for the multiplitcation of block Toeplitz matrix B on some vector V W N can be described as follows: 1. Compute 3N z forward 2D FFT of vector V; 2. Compute 4N x N y algebraic matrix-vector multiplication to obtain vector Ṽ; 3. Compute 3N z backward 2D FFT of vector Ṽ. In the second step of the algorithm above all 4N x N y multiplications are data independent. This allows us to use the following scheme of distributed data storage and parallel algorithm of IE solver. Assume, for the sake of simplicity, that we have 2N y nodes. The distributed storage of matrix is organized as follows. The block-vector Q(k,j), j = 0...2N x 1 is stored at kth node, k = 1...2N y (see Table 1). ISBN (DVD) 677
6 The projective method approach in the electromagnetic integral equations... Node 1 Node 2... Node 2N y Q(0,0) Q(1,0)... Q(2Ny 1,0) Q(0,1) Q(1,1)... Q(2Ny 1,1).... Q(0,2N x 1) Q(1,2Nx 1)... Q(2Ny 1,2N x 1) Table 1: Matrix storage organization Using this storage organization, we developed the solver with the following features of parallelization: 1. The computation of the coefficients of matricies B, S, R 1, R 2 stored at different nodes is completely data independent. It is performed simultaneously for all these matrices. 2. The computation of Fourier transform of B is done by distributed code from well-known FFTW3 library; 3. The interative method is implemented using the distributed FGMRES implementation by Fraysse et al. (2003); 4. The distributed matrix-vector calculation is processed as described above using FFTW3 library for distributed FFT and OpenBlas for algebraic matix-vector calculations; 5. The hybrid MPI+OpenMP scheme is used at all stages described above. The computational experiments performed at Bluegene/P and HPC Lomonosov from MSU showed good increment in speed depending on the number of processes for different models (see Figure 2). The matrix calculation time includes time of FFT calculation of B. One can see that the speed increment is close to a linear Matrix calculation 10 3 Equation solving Time, s 10 2 Time, s Number of processes Number of processes Bluegene/P Bluegene/P Lomonosov N x =1024 N x =128 N x =128 N Linear speed up N y =1024 N y =1024 y =128 N z =10 N z =50 N z =50 Figure 2: Efficiency of parallelization 678
7 4 High conductivity contrast modeling The accurate modeling of the EM field in a high conductivity contrast media is one of the most complex problems of EM modeling. We used the COMMEMI3D3 model (see Figure 3) as one of the test models for our solver. This model consists of 7 blocks placed in layered media. Their resistivity is stated at the Figure 3 legend. The normal vertical section of the model consists of three layers with the resistivity of 10 3, 10 4 and 10 Ω m and with thickness of the first and the second layers of 1 and 6.5 km respectively. The maximum conductivity contrast in this model is more than D view 6 Top view 5 Z,km Y,km 300 Ω m Ω m X,km 100 Ω m Ω m Y,km X,km Ω m 0.1 Ω m 0.3 Ω m Profile Figure 3: Model COMMEMI3D3 The results of magnetotellurics (MT) modeling at profile y = 3.8 km for 1Hz are presented in Figures 4, 5. The solid lines are the results of modeling with different cell sizes by our IE solver. The green crosses are the result of modeling by the high-order finite elements solver from Grayver and Kolev (in review). One can see that for the large cells (black curves) the correspondence between the above methods is bad. However, decreasing the cell sizes results in a good fit between the methods. To decrease the cell sizes means to increase the number of cells. Thus the best fit requires more than cells and more than 2TB of RAM to store the matrix. This emphasizes the importance of good parallelization and decreasing of memory requirements for IE solvers ρ xy 10 3 ρ yx m m 50m 50m 50m 25m 25m 25m 6.25m 25m 12.5m 25m 6.25m 12.5m High-order FE Figure 4: Apparent resistivity ρ xy, ρ yx at profile y = 3830 for 1 Hz depends on cell size ISBN (DVD) 679
8 The projective method approach in the electromagnetic integral equations Φ xy 96 Φ yx m m 50m 50m 50m 25m 25m 25m 6.25m 25m 12.5m 25m 6.25m 12.5m High-order FE Figure 5: Impedance tensor phases Φ xy, Φ yx at profile y = 3830 for 1 Hz depends on cell size 5 Conclusion The presented solver for 3D integral equations of electrodynamics requires only half of an amount of RAM compared with the other existing 3D IE solvers. The high parallelization degree allows us to efficiently use the modern HPC systems and to perform computations for large scale models (up to 10 8 cells) with a complex conductivity distribution. The good fit with the high-order finite elements method demonstrates the high accuracy of our solver, including the solving of the high conductivity contrast problems. The developed solver named Gnu Integral Equation Modeling in ElectroMagnetic Geophysics (GIEM2G) is implemented as hybrid MPI+OpenMP software on Fortran language. The GIEM2G is an open source software distributed under the GPLv2 license and it is available at 6 Acknowledgements This work has been supported by the Russian Foundation for Basic Research under grant no Authors acknowledge the team of HPC CMC Lomonosov MSU for the access to Bluegene/P HPC and the team of the Lomonosov MSU Research Computing Center for the access to HPC Lomonosov Sadovnichy et al. (2013). References Avdeev, D. B., A. V. Kuvshinov, O. V. Pankratov, and G. A. Newman(1997). High-performance three-dimensional electromagnetic modelling using modified neumann series. wide-band numerical solution and examples. J. Geomagn. Geoelectr. 49, Dmitriev, V., A. Silkin, and R. Farzan (2002). Tensor green function for the system of maxwellś equations in a layered medium. Computational Mathematics and Modeling 13(2), Fraysse, V., L. Giraud, S. Gratton, and J. Langou (2003). A set of GMRES routines for real and complex arithmetics on high performance computers. 680
9 Grayver, A. and T. Kolev (in review). Large-scale 3d geo-electromagnetic modeling using parallel adaptive high-order finite element method. Geophysics. Kruglyakov, M. (2011). Modified integral current methods in electrodynamics of nonhomogeneous media. Computational Mathematics and Modeling 22(3), Pankratov, O., D. Avdeyev, and A. Kuvshinov (1995). Electromagnetic-field scattering in a heterogeneous Earth: A solution to the forward problem. Physics of the Solid Earth 31(3), Sadovnichy, V., A. Tikhonravov, V. Voevodin, and V. Opanasenko (2013). Lomonosov : Supercomputing at Moscow State University. In Contemporary High Performance Computing: From Petascale toward Exascale, Chapman& Hall/CRC Computational Science, pp Boca Raton, United States. Singer, B. (2008). Electromagnetic integral equation approach based on contraction operator and solution optimization in Krylov subspace. Geophys. J. Int. 175, Ward, S. H. and G. W. Hohmann (1988). 4. Electromagnetic Theory for Geophysical Applications, Chapter 4, pp ISBN (DVD) 681
Anna Avdeeva Dmitry Avdeev and Marion Jegen. 31 March Introduction to 3D MT inversion code x3di
Anna Avdeeva (aavdeeva@ifm-geomar.de), Dmitry Avdeev and Marion Jegen 31 March 211 Outline 1 Essential Parts of 3D MT Inversion Code 2 Salt Dome Overhang Study with 3 Outline 1 Essential Parts of 3D MT
More informationTu Efficient 3D MT Inversion Using Finite-difference Time-domain Modelling
Tu 11 16 Efficient 3D MT Inversion Using Finite-difference Time-domain Modelling S. de la Kethulle de Ryhove* (Electromagnetic Geoservices ASA), J.P. Morten (Electromagnetic Geoservices ASA) & K. Kumar
More information( ) R kj. = y k y j. y A ( ) z A. y a. z a. Derivatives of the second order electrostatic tensor with respect to the translation of ( ) δ yβ.
Supporting information Derivatives of R with respect to the translation of fragment along the y and z axis: y = y k y j (S1) z ( = z z k j) (S2) Derivatives of S with respect to the translation of fragment
More informationFinite Element Method (FEM)
Finite Element Method (FEM) The finite element method (FEM) is the oldest numerical technique applied to engineering problems. FEM itself is not rigorous, but when combined with integral equation techniques
More informationDemonstration of the Coupled Evolution Rules 163 APPENDIX F: DEMONSTRATION OF THE COUPLED EVOLUTION RULES
Demonstration of the Coupled Evolution Rules 163 APPENDIX F: DEMONSTRATION OF THE COUPLED EVOLUTION RULES Before going into the demonstration we need to point out two limitations: a. It assumes I=1/2 for
More informationLOWELL WEEKLY JOURNAL
Y -» $ 5 Y 7 Y Y -Y- Q x Q» 75»»/ q } # ]»\ - - $ { Q» / X x»»- 3 q $ 9 ) Y q - 5 5 3 3 3 7 Q q - - Q _»»/Q Y - 9 - - - )- [ X 7» -» - )»? / /? Q Y»» # X Q» - -?» Q ) Q \ Q - - - 3? 7» -? #»»» 7 - / Q
More informationProgramming Project 2: Harmonic Vibrational Frequencies
Programming Project 2: Harmonic Vibrational Frequencies Center for Computational Chemistry University of Georgia Athens, Georgia 30602 Summer 2012 1 Introduction This is the second programming project
More informationMagnetotelluric (MT) Method
Magnetotelluric (MT) Method Dr. Hendra Grandis Graduate Program in Applied Geophysics Faculty of Mining and Petroleum Engineering ITB Geophysical Methods Techniques applying physical laws (or theory) to
More informationMagnetotelluric tensor decomposition: Part II, Examples of a basic procedure
GEOPHYSICS, VOL. 63, NO. 6 (NOVEMBER-DECEMBER 1998); P. 1898 1907, 5 FIGS. Magnetotelluric tensor decomposition: Part II, Examples of a basic procedure F. E. M. (Ted) Lilley ABSTRACT The decomposition
More informationEJL R Sβ. sum. General objective: Reduce the complexity of the analysis by exploiting symmetry. Garth J. Simpson
e sum = EJL R Sβ General objective: Reduce the complexity of the analysis by exploiting symmetry. Specific Objectives: 1. The molecular symmetry matrix S. How to populate it.. Relationships between the
More informationA DARK GREY P O N T, with a Switch Tail, and a small Star on the Forehead. Any
Y Y Y X X «/ YY Y Y ««Y x ) & \ & & } # Y \#$& / Y Y X» \\ / X X X x & Y Y X «q «z \x» = q Y # % \ & [ & Z \ & { + % ) / / «q zy» / & / / / & x x X / % % ) Y x X Y $ Z % Y Y x x } / % «] «] # z» & Y X»
More informationWaves in Linear Optical Media
1/53 Waves in Linear Optical Media Sergey A. Ponomarenko Dalhousie University c 2009 S. A. Ponomarenko Outline Plane waves in free space. Polarization. Plane waves in linear lossy media. Dispersion relations
More informationNeatest and Promptest Manner. E d i t u r ami rul)lihher. FOIt THE CIIILDIIES'. Trifles.
» ~ $ ) 7 x X ) / ( 8 2 X 39 ««x» ««! «! / x? \» «({? «» q «(? (?? x! «? 8? ( z x x q? ) «q q q ) x z x 69 7( X X ( 3»«! ( ~«x ««x ) (» «8 4 X «4 «4 «8 X «x «(» X) ()»» «X «97 X X X 4 ( 86) x) ( ) z z
More information3 Constitutive Relations: Macroscopic Properties of Matter
EECS 53 Lecture 3 c Kamal Sarabandi Fall 21 All rights reserved 3 Constitutive Relations: Macroscopic Properties of Matter As shown previously, out of the four Maxwell s equations only the Faraday s and
More informationMANY BILLS OF CONCERN TO PUBLIC
- 6 8 9-6 8 9 6 9 XXX 4 > -? - 8 9 x 4 z ) - -! x - x - - X - - - - - x 00 - - - - - x z - - - x x - x - - - - - ) x - - - - - - 0 > - 000-90 - - 4 0 x 00 - -? z 8 & x - - 8? > 9 - - - - 64 49 9 x - -
More informationSome Geometric and Algebraic Aspects of Domain Decomposition Methods
Some Geometric and Algebraic Aspects of Domain Decomposition Methods D.S.Butyugin 1, Y.L.Gurieva 1, V.P.Ilin 1,2, and D.V.Perevozkin 1 Abstract Some geometric and algebraic aspects of various domain decomposition
More informationA Brief Revision of Vector Calculus and Maxwell s Equations
A Brief Revision of Vector Calculus and Maxwell s Equations Debapratim Ghosh Electronic Systems Group Department of Electrical Engineering Indian Institute of Technology Bombay e-mail: dghosh@ee.iitb.ac.in
More informationi r-s THE MEMPHIS, TENN., SATURDAY. DEGfMBER
N k Q2 90 k ( < 5 q v k 3X3 0 2 3 Q :: Y? X k 3 : \ N 2 6 3 N > v N z( > > :}9 [ ( k v >63 < vq 9 > k k x k k v 6> v k XN Y k >> k < v Y X X X NN Y 2083 00 N > N Y Y N 0 \ 9>95 z {Q ]k3 Q k x k k z x X
More informationProblem Set 2 Due Tuesday, September 27, ; p : 0. (b) Construct a representation using five d orbitals that sit on the origin as a basis: 1
Problem Set 2 Due Tuesday, September 27, 211 Problems from Carter: Chapter 2: 2a-d,g,h,j 2.6, 2.9; Chapter 3: 1a-d,f,g 3.3, 3.6, 3.7 Additional problems: (1) Consider the D 4 point group and use a coordinate
More informationOWELL WEEKLY JOURNAL
Y \»< - } Y Y Y & #»»» q ] q»»»>) & - - - } ) x ( - { Y» & ( x - (» & )< - Y X - & Q Q» 3 - x Q Y 6 \Y > Y Y X 3 3-9 33 x - - / - -»- --
More informationLOWELL JOURNAL. MUST APOLOGIZE. such communication with the shore as Is m i Boimhle, noewwary and proper for the comfort
- 7 7 Z 8 q ) V x - X > q - < Y Y X V - z - - - - V - V - q \ - q q < -- V - - - x - - V q > x - x q - x q - x - - - 7 -» - - - - 6 q x - > - - x - - - x- - - q q - V - x - - ( Y q Y7 - >»> - x Y - ] [
More information13. LECTURE 13. Objectives
13. LECTURE 13 Objectives I can use Clairaut s Theorem to make my calculations easier. I can take higher derivatives. I can check if a function is a solution to a partial differential equation. Higher
More informationLowell Dam Gone Out. Streets Turned I n t o Rivers. No Cause For Alarm Now However As This Happened 35 Years A&o
V ()\\ ))? K K Y 6 96 Y - Y Y V 5 Z ( z x z \ - \ - - z - q q x x - x 5 9 Q \ V - - Y x 59 7 x x - Y - x - - x z - z x - ( 7 x V 9 z q &? - 9 - V ( x - - - V- [ Z x z - -x > -) - - > X Z z ( V V V
More information( ). One set of terms has a ω in
Laptag Class Notes W. Gekelan Cold Plasa Dispersion relation Suer Let us go back to a single particle and see how it behaves in a high frequency electric field. We will use the force equation and Maxwell
More informationFinite Element Method in Geotechnical Engineering
Finite Element Method in Geotechnical Engineering Short Course on + Dynamics Boulder, Colorado January 5-8, 2004 Stein Sture Professor of Civil Engineering University of Colorado at Boulder Contents Steps
More information2012 SEG SEG Las Vegas 2012 Annual Meeting Page 1
Frequency-domain EM modeling of 3D anisotropic magnetic permeability and analytical analysis Jiuping Chen, Michael Tompkins, Ping Zhang, Michael Wilt, Schlumberger-EMI, and Randall Mackie, Formerly Schlumberger-EMI,
More informationRelativistic Electrodynamics
Relativistic Electrodynamics Notes (I will try to update if typos are found) June 1, 2009 1 Dot products The Pythagorean theorem says that distances are given by With time as a fourth direction, we find
More informationNIELINIOWA OPTYKA MOLEKULARNA
NIELINIOWA OPTYKA MOLEKULARNA chapter 1 by Stanisław Kielich translated by:tadeusz Bancewicz http://zon8.physd.amu.edu.pl/~tbancewi Poznan,luty 2008 ELEMENTS OF THE VECTOR AND TENSOR ANALYSIS Reference
More informationFast computation of Hankel Transform using orthonormal exponential approximation of complex kernel function
Fast computation of Hankel Transform using orthonormal exponential approximation of complex kernel function Pravin K Gupta 1, Sri Niwas 1 and Neeta Chaudhary 2 1 Department of Earth Sciences, Indian Institute
More informationWednesday, April 12. Today:
Wednesday, April 2 Last Time: - The solid state - atomic arrangement in solids - why do solids form: energetics - Lattices, translations, rotation, & other symmetry operations Today: Continue with lattices,
More informationArtificial Intelligence & Neuro Cognitive Systems Fakultät für Informatik. Robot Dynamics. Dr.-Ing. John Nassour J.
Artificial Intelligence & Neuro Cognitive Systems Fakultät für Informatik Robot Dynamics Dr.-Ing. John Nassour 25.1.218 J.Nassour 1 Introduction Dynamics concerns the motion of bodies Includes Kinematics
More informationDetermination of Locally Varying Directions through Mass Moment of Inertia Tensor
Determination of Locally Varying Directions through Mass Moment of Inertia Tensor R. M. Hassanpour and C.V. Deutsch Centre for Computational Geostatistics Department of Civil and Environmental Engineering
More informationSimple Examples on Rectangular Domains
84 Chapter 5 Simple Examples on Rectangular Domains In this chapter we consider simple elliptic boundary value problems in rectangular domains in R 2 or R 3 ; our prototype example is the Poisson equation
More informationTopological insulator part I: Phenomena
Phys60.nb 5 Topological insulator part I: Phenomena (Part II and Part III discusses how to understand a topological insluator based band-structure theory and gauge theory) (Part IV discusses more complicated
More informationComplex Variables. Chapter 1. Complex Numbers Section 1.2. Basic Algebraic Properties Proofs of Theorems. December 16, 2016
Complex Variables Chapter 1. Complex Numbers Section 1.2. Basic Algebraic Properties Proofs of Theorems December 16, 2016 () Complex Variables December 16, 2016 1 / 12 Table of contents 1 Theorem 1.2.1
More informationGeophysical Journal International
Geophysical Journal International Geophys. J. Int. (06) 06, 78 79 Advance Access publication 06 June 4 GJI Marine geosciences and applied geophysics doi: 0.093/gji/ggw37 Contraction pre-conditioner in
More informationL bor y nnd Union One nnd Inseparable. LOW I'LL, MICHIGAN. WLDNHSDA Y. JULY ), I8T. liuwkll NATIdiNAI, liank
G k y $5 y / >/ k «««# ) /% < # «/» Y»««««?# «< >«>» y k»» «k F 5 8 Y Y F G k F >«y y
More informationTEST CODE: MMA (Objective type) 2015 SYLLABUS
TEST CODE: MMA (Objective type) 2015 SYLLABUS Analytical Reasoning Algebra Arithmetic, geometric and harmonic progression. Continued fractions. Elementary combinatorics: Permutations and combinations,
More informationLecture10: Plasma Physics 1. APPH E6101x Columbia University
Lecture10: Plasma Physics 1 APPH E6101x Columbia University Last Lecture - Conservation principles in magnetized plasma frozen-in and conservation of particles/flux tubes) - Alfvén waves without plasma
More informationProblem Set 2 Due Thursday, October 1, & & & & # % (b) Construct a representation using five d orbitals that sit on the origin as a basis:
Problem Set 2 Due Thursday, October 1, 29 Problems from Cotton: Chapter 4: 4.6, 4.7; Chapter 6: 6.2, 6.4, 6.5 Additional problems: (1) Consider the D 3h point group and use a coordinate system wherein
More informationMIS S BALLS, L.L.A.]
& N k k QY GN ( x - N N & N & k QY GN x 00 - XX N X ± - - - - ---------------- N N G G N N N Y NG 5 880N GN N X GN x ( G ) 8N ---- N 8 Y 8 - N N ( G () G ( ) (N) N? k [ x-k NNG G G k k N NY Y /( Q G (-)
More informationMicroscopic-Macroscopic connection. Silvana Botti
relating experiment and theory European Theoretical Spectroscopy Facility (ETSF) CNRS - Laboratoire des Solides Irradiés Ecole Polytechnique, Palaiseau - France Temporary Address: Centre for Computational
More informationRigid body dynamics. Basilio Bona. DAUIN - Politecnico di Torino. October 2013
Rigid body dynamics Basilio Bona DAUIN - Politecnico di Torino October 2013 Basilio Bona (DAUIN - Politecnico di Torino) Rigid body dynamics October 2013 1 / 16 Multiple point-mass bodies Each mass is
More information6. 3D Kinematics DE2-EA 2.1: M4DE. Dr Connor Myant
DE2-EA 2.1: M4DE Dr Connor Myant 6. 3D Kinematics Comments and corrections to connor.myant@imperial.ac.uk Lecture resources may be found on Blackboard and at http://connormyant.com Contents Three-Dimensional
More informationFast Inversion of Logging-While-Drilling (LWD) Resistivity Measurements
Fast Inversion of Logging-While-Drilling (LWD) Resistivity Measurements David Pardo 1 Carlos Torres-Verdín 2 1 University of the Basque Country (UPV/EHU) and Ikerbasque, Bilbao, Spain. 2 The University
More informationElectromagnetic wave propagation. ELEC 041-Modeling and design of electromagnetic systems
Electromagnetic wave propagation ELEC 041-Modeling and design of electromagnetic systems EM wave propagation In general, open problems with a computation domain extending (in theory) to infinity not bounded
More informationDiscretization of PDEs and Tools for the Parallel Solution of the Resulting Systems
Discretization of PDEs and Tools for the Parallel Solution of the Resulting Systems Stan Tomov Innovative Computing Laboratory Computer Science Department The University of Tennessee Wednesday April 4,
More informationMultivariable Calculus and Matrix Algebra-Summer 2017
Multivariable Calculus and Matrix Algebra-Summer 017 Homework 4 Solutions Note that the solutions below are for the latest version of the problems posted. For those of you who worked on an earlier version
More informationTheory and Applications of Dielectric Materials Introduction
SERG Summer Seminar Series #11 Theory and Applications of Dielectric Materials Introduction Tzuyang Yu Associate Professor, Ph.D. Structural Engineering Research Group (SERG) Department of Civil and Environmental
More informationChapter 2. Vectors and Vector Spaces
2.1. Operations on Vectors 1 Chapter 2. Vectors and Vector Spaces Section 2.1. Operations on Vectors Note. In this section, we define several arithmetic operations on vectors (especially, vector addition
More informationSimple medium: D = ɛe Dispersive medium: D = ɛ(ω)e Anisotropic medium: Permittivity as a tensor
Plane Waves 1 Review dielectrics 2 Plane waves in the time domain 3 Plane waves in the frequency domain 4 Plane waves in lossy and dispersive media 5 Phase and group velocity 6 Wave polarization Levis,
More informationIn this section, mathematical description of the motion of fluid elements moving in a flow field is
Jun. 05, 015 Chapter 6. Differential Analysis of Fluid Flow 6.1 Fluid Element Kinematics In this section, mathematical description of the motion of fluid elements moving in a flow field is given. A small
More informationA new computation method for a staggered grid of 3D EM field conservative modeling
Earth Planets Space, 54, 499 509, 2002 A new computation method for a staggered grid of 3D EM field conservative modeling Elena Yu. Fomenko 1 and Toru Mogi 2 1 Geoelectromagnetic Research Institute RAS,
More informationAn analytical method for the inverse Cauchy problem of Lame equation in a rectangle
Journal of Physics: Conference Series PAPER OPEN ACCESS An analytical method for the inverse Cauchy problem of Lame equation in a rectangle To cite this article: Yu Grigor ev 218 J. Phys.: Conf. Ser. 991
More informationACCEPTS HUGE FLORAL KEY TO LOWELL. Mrs, Walter Laid to Rest Yesterday
$ j < < < > XXX Y 928 23 Y Y 4% Y 6 -- Q 5 9 2 5 Z 48 25 )»-- [ Y Y Y & 4 j q - Y & Y 7 - -- - j \ -2 -- j j -2 - - - - [ - - / - ) ) - - / j Y 72 - ) 85 88 - / X - j ) \ 7 9 Y Y 2 3» - ««> Y 2 5 35 Y
More informationECE 6340 Intermediate EM Waves. Fall Prof. David R. Jackson Dept. of ECE. Notes 6
ECE 6340 Intermediate EM Waves Fall 016 Prof. David R. Jackson Dept. of ECE Notes 6 1 Power Dissipated by Current Work given to a collection of electric charges movg an electric field: ( qe ) ( ρ S E )
More informationPolyhedral Mass Properties (Revisited)
Polyhedral Mass Properties (Revisited) David Eberly, Geometric Tools, Redmond WA 98052 https://www.geometrictools.com/ This work is licensed under the Creative Commons Attribution 4.0 International License.
More informationIMPLICATION ALGEBRAS
Discussiones Mathematicae General Algebra and Applications 26 (2006 ) 141 153 IMPLICATION ALGEBRAS Ivan Chajda Department of Algebra and Geometry Palacký University of Olomouc Tomkova 40, 779 00 Olomouc,
More informationHigher order derivative
2 Î 3 á Higher order derivative Î 1 Å Iterated partial derivative Iterated partial derivative Suppose f has f/ x, f/ y and ( f ) = 2 f x x x 2 ( f ) = 2 f x y x y ( f ) = 2 f y x y x ( f ) = 2 f y y y
More informationA Brief Introduction to Magnetotellurics and Controlled Source Electromagnetic Methods
A Brief Introduction to Magnetotellurics and Controlled Source Electromagnetic Methods Frank Morrison U.C. Berkeley With the help of: David Alumbaugh Erika Gasperikova Mike Hoversten Andrea Zirilli A few
More informationChapter Two: Numerical Methods for Elliptic PDEs. 1 Finite Difference Methods for Elliptic PDEs
Chapter Two: Numerical Methods for Elliptic PDEs Finite Difference Methods for Elliptic PDEs.. Finite difference scheme. We consider a simple example u := subject to Dirichlet boundary conditions ( ) u
More informationLesson Rigid Body Dynamics
Lesson 8 Rigid Body Dynamics Lesson 8 Outline Problem definition and motivations Dynamics of rigid bodies The equation of unconstrained motion (ODE) User and time control Demos / tools / libs Rigid Body
More informationNumerical simulation of the Gross-Pitaevskii equation by pseudo-spectral and finite element methods comparison of GPS code and FreeFem++
. p.1/13 10 ec. 2014, LJLL, Paris FreeFem++ workshop : BECASIM session Numerical simulation of the Gross-Pitaevskii equation by pseudo-spectral and finite element methods comparison of GPS code and FreeFem++
More informationEfficient domain decomposition methods for the time-harmonic Maxwell equations
Efficient domain decomposition methods for the time-harmonic Maxwell equations Marcella Bonazzoli 1, Victorita Dolean 2, Ivan G. Graham 3, Euan A. Spence 3, Pierre-Henri Tournier 4 1 Inria Saclay (Defi
More informationMath 671: Tensor Train decomposition methods II
Math 671: Tensor Train decomposition methods II Eduardo Corona 1 1 University of Michigan at Ann Arbor December 13, 2016 Table of Contents 1 What we ve talked about so far: 2 The Tensor Train decomposition
More informationLevi s Commutator Theorems for Cancellative Semigroups
Levi s Commutator Theorems for Cancellative Semigroups R. Padmanabhan W. McCune R. Veroff Abstract A conjecture of Padmanabhan, on provability in cancellative semigroups, is addressed. Several of Levi
More informationManipulator Dynamics 2. Instructor: Jacob Rosen Advanced Robotic - MAE 263D - Department of Mechanical & Aerospace Engineering - UCLA
Manipulator Dynamics 2 Forward Dynamics Problem Given: Joint torques and links geometry, mass, inertia, friction Compute: Angular acceleration of the links (solve differential equations) Solution Dynamic
More informationarxiv: v3 [math-ph] 24 Aug 2010
Inverse Vector Operators Shaon Sahoo Department of Physics, Indian Institute of Science, Bangalore 5600, India. arxiv:0804.9v [math-ph] 4 Aug 00 Abstract In different branches of physics, we frequently
More information2.1 NUMERICAL SOLUTION OF SIMULTANEOUS FIRST ORDER ORDINARY DIFFERENTIAL EQUATIONS. differential equations with the initial values y(x 0. ; l.
Numerical Methods II UNIT.1 NUMERICAL SOLUTION OF SIMULTANEOUS FIRST ORDER ORDINARY DIFFERENTIAL EQUATIONS.1.1 Runge-Kutta Method of Fourth Order 1. Let = f x,y,z, = gx,y,z be the simultaneous first order
More informationFast Multipole Methods: Fundamentals & Applications. Ramani Duraiswami Nail A. Gumerov
Fast Multipole Methods: Fundamentals & Applications Ramani Duraiswami Nail A. Gumerov Week 1. Introduction. What are multipole methods and what is this course about. Problems from physics, mathematics,
More informationAN FFT-BASED APPROACH IN ACCELERATION OF DISCRETE GREEN S FUNCTION METHOD FOR AN- TENNA ANALYSIS
Progress In Electromagnetics Research M, Vol. 29, 17 28, 2013 AN FFT-BASED APPROACH IN ACCELERATION OF DISCRETE GREEN S FUNCTION METHOD FOR AN- TENNA ANALYSIS Salma Mirhadi *, Mohammad Soleimani, and Ali
More informationMultipole moments. Dipole moment. The second moment µ is more commonly called the dipole moment, of the charge. distribution and is a vector
Dipole moment Multipole moments The second moment µ is more commonly called the dipole moment, of the charge distribution and is a vector µ = µ x ˆx + µ y ŷ + µ z ẑ where the α component is given by µ
More informationAn improved parallel Poisson solver for space charge calculation in beam dynamics simulation
An improved parallel Poisson solver for space charge calculation in beam dynamics simulation DAWEI ZHENG, URSULA VAN RIENEN University of Rostock, Rostock, 18059, Germany JI QIANG LBNL, Berkeley, CA 94720,
More informationDomain Decomposition Preconditioners for Spectral Nédélec Elements in Two and Three Dimensions
Domain Decomposition Preconditioners for Spectral Nédélec Elements in Two and Three Dimensions Bernhard Hientzsch Courant Institute of Mathematical Sciences, New York University, 51 Mercer Street, New
More informationIntel Math Kernel Library (Intel MKL) LAPACK
Intel Math Kernel Library (Intel MKL) LAPACK Linear equations Victor Kostin Intel MKL Dense Solvers team manager LAPACK http://www.netlib.org/lapack Systems of Linear Equations Linear Least Squares Eigenvalue
More informationCourse no. 4. The Theory of Electromagnetic Field
Cose no. 4 The Theory of Electromagnetic Field Technical University of Cluj-Napoca http://www.et.utcluj.ro/cs_electromagnetics2006_ac.htm http://www.et.utcluj.ro/~lcret March 19-2009 Chapter 3 Magnetostatics
More informationFFT-based Galerkin method for homogenization of periodic media
FFT-based Galerkin method for homogenization of periodic media Jaroslav Vondřejc 1,2 Jan Zeman 2,3 Ivo Marek 2 Nachiketa Mishra 4 1 University of West Bohemia in Pilsen, Faculty of Applied Sciences Czech
More informationGypsilab : a MATLAB toolbox for FEM-BEM coupling
Gypsilab : a MATLAB toolbox for FEM-BEM coupling François Alouges, (joint work with Matthieu Aussal) Workshop on Numerical methods for wave propagation and applications Sept. 1st 2017 Facts New numerical
More informationPolarimetry of homogeneous half-spaces 1
Polarimetry of homogeneous half-spaces 1 M. Gilman, E. Smith, S. Tsynkov special thanks to: H. Hong Department of Mathematics North Carolina State University, Raleigh, NC Workshop on Symbolic-Numeric Methods
More informationTheoretische Physik 2: Elektrodynamik (Prof. A-S. Smith) Home assignment 9
WiSe 202 20.2.202 Prof. Dr. A-S. Smith Dipl.-Phys. Ellen Fischermeier Dipl.-Phys. Matthias Saba am Lehrstuhl für Theoretische Physik I Department für Physik Friedrich-Alexander-Universität Erlangen-Nürnberg
More informationElectromagnetic Wave Propagation Lecture 3: Plane waves in isotropic and bianisotropic media
Electromagnetic Wave Propagation Lecture 3: Plane waves in isotropic and bianisotropic media Daniel Sjöberg Department of Electrical and Information Technology September 2016 Outline 1 Plane waves in lossless
More informationElectromagnetically Induced Flows in Water
Electromagnetically Induced Flows in Water Michiel de Reus 8 maart 213 () Electromagnetically Induced Flows 1 / 56 Outline 1 Introduction 2 Maxwell equations Complex Maxwell equations 3 Gaussian sources
More informationTHEORETICAL ANALYSIS AND NUMERICAL CALCULATION OF LOOP-SOURCE TRANSIENT ELECTROMAGNETIC IMAGING
CHINESE JOURNAL OF GEOPHYSICS Vol.47, No.2, 24, pp: 379 386 THEORETICAL ANALYSIS AND NUMERICAL CALCULATION OF LOOP-SOURCE TRANSIENT ELECTROMAGNETIC IMAGING XUE Guo-Qiang LI Xiu SONG Jian-Ping GUO Wen-Bo
More informationRealizations of Loops and Groups defined by short identities
Comment.Math.Univ.Carolin. 50,3(2009) 373 383 373 Realizations of Loops and Groups defined by short identities A.D. Keedwell Abstract. In a recent paper, those quasigroup identities involving at most three
More informationFinite-difference simulation of borehole EM measurements in 3D anisotropic media using coupled scalar-vector potentials
GEOPHYSICS VOL. 71 NO. 5 SEPTEMBER-OCTOBER 006 ; P. G5 G33 1 FIGS. 10.1190/1.567 Finite-difference simulation of borehole EM measurements in 3D anisotropic media using coupled scalar-vector potentials
More informationCP1 REVISION LECTURE 3 INTRODUCTION TO CLASSICAL MECHANICS. Prof. N. Harnew University of Oxford TT 2017
CP1 REVISION LECTURE 3 INTRODUCTION TO CLASSICAL MECHANICS Prof. N. Harnew University of Oxford TT 2017 1 OUTLINE : CP1 REVISION LECTURE 3 : INTRODUCTION TO CLASSICAL MECHANICS 1. Angular velocity and
More informationMath 671: Tensor Train decomposition methods
Math 671: Eduardo Corona 1 1 University of Michigan at Ann Arbor December 8, 2016 Table of Contents 1 Preliminaries and goal 2 Unfolding matrices for tensorized arrays The Tensor Train decomposition 3
More informationNONLINEAR STRUCTURAL DYNAMICS USING FE METHODS
NONLINEAR STRUCTURAL DYNAMICS USING FE METHODS Nonlinear Structural Dynamics Using FE Methods emphasizes fundamental mechanics principles and outlines a modern approach to understanding structural dynamics.
More informationMECH 5312 Solid Mechanics II. Dr. Calvin M. Stewart Department of Mechanical Engineering The University of Texas at El Paso
MECH 5312 Solid Mechanics II Dr. Calvin M. Stewart Department of Mechanical Engineering The University of Texas at El Paso Table of Contents Preliminary Math Concept of Stress Stress Components Equilibrium
More informationGG303 Lecture 6 8/27/09 1 SCALARS, VECTORS, AND TENSORS
GG303 Lecture 6 8/27/09 1 SCALARS, VECTORS, AND TENSORS I Main Topics A Why deal with tensors? B Order of scalars, vectors, and tensors C Linear transformation of scalars and vectors (and tensors) II Why
More informationBoolean Algebra. Examples: (B=set of all propositions, or, and, not, T, F) (B=2 A, U,, c, Φ,A)
Boolean Algebra Definition: A Boolean Algebra is a math construct (B,+,.,, 0,1) where B is a non-empty set, + and. are binary operations in B, is a unary operation in B, 0 and 1 are special elements of
More informationk incident k reflected k transmitted ε 1 µ 1 σ 1 ρ 1 ε 2 µ 2 σ 2 ρ 2
TECHNICAL REPORT UMR EMC LABORATORY 1 Perfectly Matched Layers Used as Absorbing Boundaries in a Three-dimensional FDTD Code David M. Hockanson Abstract The Finite-Dierence Time-Domain (FDTD) method is
More informationNDT&E Methods: UT. VJ Technologies CAVITY INSPECTION. Nondestructive Testing & Evaluation TPU Lecture Course 2015/16.
CAVITY INSPECTION NDT&E Methods: UT VJ Technologies NDT&E Methods: UT 6. NDT&E: Introduction to Methods 6.1. Ultrasonic Testing: Basics of Elasto-Dynamics 6.2. Principles of Measurement 6.3. The Pulse-Echo
More informationE E D E=0 2 E 2 E (3.1)
Chapter 3 Constitutive Relations Maxwell s equations define the fields that are generated by currents and charges. However, they do not describe how these currents and charges are generated. Thus, to find
More informationModule I: Electromagnetic waves
Module I: Electromagnetic waves Lectures 10-11: Multipole radiation Amol Dighe TIFR, Mumbai Outline 1 Multipole expansion 2 Electric dipole radiation 3 Magnetic dipole and electric quadrupole radiation
More informationGroups and Symmetries
Groups and Symmetries Definition: Symmetry A symmetry of a shape is a rigid motion that takes vertices to vertices, edges to edges. Note: A rigid motion preserves angles and distances. Definition: Group
More informationAxioms of Kleene Algebra
Introduction to Kleene Algebra Lecture 2 CS786 Spring 2004 January 28, 2004 Axioms of Kleene Algebra In this lecture we give the formal definition of a Kleene algebra and derive some basic consequences.
More informationHEAT TRANSFER MATERIALS. in COMPOSITE SEIICHI NOMURA A. HAJI-SHEIKH. The University of Texas at Arlington
HEAT TRANSFER in COMPOSITE MATERIALS SEIICHI NOMURA A. HAJI-SHEIKH The University of Texas at Arlington Heat Transfer in Composite Materials D EStech Publications, Inc. 439 North Duke Street Lancaster,
More informationHigh-Performance Three-Dimensional Electromagnetic Modelling Using Modified Neumann Series. Anisotropic Earth
J. Geomag. Geoelectr., 49, 1541-1547, 1997 High-Performance Three-Dimensional Electromagnetic Modelling Using Modified Neumann Series. Anisotropic Earth Oleg V. PANKRATOV, Alexei V. Kovsnnvov, and Dmitry
More informationHydra: Generation and Tuning of parallel solutions for linear algebra equations. Alexandre X. Duchâteau University of Illinois at Urbana Champaign
Hydra: Generation and Tuning of parallel solutions for linear algebra equations Alexandre X. Duchâteau University of Illinois at Urbana Champaign Collaborators Thesis Advisors Denis Barthou (Labri/INRIA
More information