the sum of two projections. Finally, in Section 5, we apply the theory of Section 4 to the case of nite element spaces. 2. Additive Algorithms for Par

Size: px
Start display at page:

Download "the sum of two projections. Finally, in Section 5, we apply the theory of Section 4 to the case of nite element spaces. 2. Additive Algorithms for Par"

Transcription

1 ON THE SPECTRA OF SUMS OF ORTHOGONAL PROJECTIONS WITH APPLICATIONS TO PARALLEL COMPUTING PETTER E. BJRSTAD y AND JAN MANDEL z Abstract. Many parallel iterative algorithms for solving symmetric, positive denite problems proceed by solving in each iteration, a number of independent systems on subspaces. The convergence of such methods is determined by the spectrum of the sums of orthogonal projections on those subspaces, while the convergence of a related sequential method is determined by the spectrum of the product of complementary projections. We study spectral properties of sums of orthogonal projections and in the case of two projections, characterize the spectrum of the sum completely in terms of the spectrum of the product. Key words. Orthogonal Projections, Parallel Computing, Domain Decomposition, Grid Renement, Schwarz Alternating Method. 1. Introduction. Recently there has been a strong revival of the interest in domain decomposition algorithms for elliptic problems; cf. e.g. Glowinski et al. [11], and Chan et al. [4]. A classical algorithm of this kind is the Schwarz alternating method [18]. It proceeds by computing the solution on subdomains in a sequential fashion, and is therefore not necessarily attractive in a parallel computing environment. Similarly, the FAC algorithm [16, 17], an iterative algorithm for grid renement problems, computes the solution to subproblems on a sequence of uniform grids. Alternative methods that may be more suitable for parallel computing, have recently been proposed. These so called additive methods proceed by computing the solution on all subdomains, or in the renement case on all grids, simultanously, thereby making the algorithms more suitable for parallel computers. The present work has been motivated by the observed success of the above mentioned algorithms despite a rather incomplete theoretical foundation. The convergence of these methods is determined by properties of the spectrum of certain sums of orthogonal projections. However, studies of sums of orthogonal projections per se seem to be missing from the literature, and more such tools are needed for the analysis of parallel iterative methods. In this paper, we collect some known propositions in a unied framework and complement this by new results. In Section 2, we recall several parallel iterative algorithms, the analysis of which leads to sums of orthogonal projections. In Section 3, we study sums of an arbitrary number of projections, and in Section 4, we completely characterize the spectrum of BIT 31 (1991) y Institutt for Informatikk, University of Bergen, Thormhlens gate 55, N-5008 Bergen, Norway. This work was supported in part by the Norwegian Research Council for Science and the Humanities under grant D and by The Royal Norwegian Council for Scientic and Industrial Research under grant IT z Computational Mathematics Group, University of Colorado at Denver, 1200 Larimer Street, Denver, CO This work was supported in part by the Air Force Oce of Scientic Research under grant AFOSR and by the National Science Foundation under grant DMS

2 the sum of two projections. Finally, in Section 5, we apply the theory of Section 4 to the case of nite element spaces. 2. Additive Algorithms for Parallel Solution of Linear Systems. Let H be a Hilbert space with inner product a(; ), and h; i another inner product on H extended to a duality pairing hf; ui, f 2 H 0, u 2 H in the usual way. Consider the variational problem (2.1) u 2 H : a(u; v) = hf; vi; 8v 2 H: We are mainly concerned with a discrete version of this problem, that is, the case when H is nite dimensional; however, most of the results hold in the general case. Let u be the solution of (2.1). Let V i, i = 1; : : : ; n be closed subspaces of H. Following to P. L. Lions [14, 15], the classical Schwarz alternating method for the iterative solution of (2.1) can be written in an abstract way as iterations u k 7 u k+1 dened by ) u i 2 V i : a(u i ; v i ) = hf; v i i? a(u k+(i?1)=n ; v i ); 8v i 2 V (2.2) i u k+i=n = u k+(i?1)=n ; i = 1; : : : ; n: + u i It is easy to see that it holds for the transformation of error that u? u k+1 = (I? P Vn ) (I? P V 1)(u? u k ); where P Vi is the orthogonal projection on V i. This method decomposes the problem (2.1) into a series of subproblems (2.2) on subspaces V i. A parallel version of this method [2, 8, 10, 14, 15] is dened by (2.3) u i 2 V i : a(u i ; v i ) = hf; v i i? a(u k ; v i ); 8v i 2 V i ; i = 1; : : : ; n; u k+1 = u k + P n u i; and it is easy to see that the error in (2.3) is transformed by u? u k+1 = I? P Vi (u? u k ): For obvious reasons, the method (2.2) is called the multiplicative method, and the method (2.3) is called the additive method. Dene A : H H 0 by hau; vi = a(u; v) and C : H 0 H by (2.4) CA = P Vi : Then the additive method (2.3) can be written in the standard form u k+1 = u k? C(Au k? f); 2

3 the mapping C being an approximate solver for Au = r, dened by Cr = u i ; u i 2 V i ; a(u i ; v i ) = hr; v i i; 8v i 2 V i : Note that by (2.4), CA is symmetric and positive denite on H. Since the additive method will fail to converge if the spectral radius (CA) > 2, which may well happen in the general case, the approximate solver C is often used as a preconditioner in the conjugate gradient method whose convergence properties are then determined by the spectrum (CA). In particular, the number of steps of the precondjugate gradient method required to solve the problem to a xed precision grows like ( max (CA)= min (CA)) 1=2. See [12] for more details. The analysis of such iterative methods thus leads to the problem of localizing the spectrum of P n P Vi. We proceed to give a few examples of highly interesting algorithms that can be put into this framework. The Schwarz' alternating method is obtained when H is a space of functions on a domain = S n i, and V i = fv 2 V : supp v i g: Additive FAC (AFAC). Following [17] we briey describe an additive algorithm for computing the composite solution on a grid having one level of renement. This algorithm generalizes in a straightforward fashion to many (nested) levels of renement. Let H = H 2h +H h, where H 2h and H h are nite element spaces such that H 2h H 1 0( 1 ), H h H 1 0( 2 ), 2 1, and H 2h \ H 1 0( 2 ) H h. Given u 2 H as the current approximation to the composite solution, the additive algorithm proceeds as follows [17]: u 2h 2 H 2h : a(u? u 2h ; v 2h ) = f(v 2h ); 8v 2h 2 H 2h Update u by u u? u 2h. The next step is u h 2 H h : a(u? u h ; v h ) = f(v h ); 8v h 2 H h w h 2 H 2h \ H h : a(u? w 2h ; v 2h ) = f(v 2h ); 8v 2h 2 H 2h \ H h : Update u by u u? u h + w 2h. Following [17] dene H 2h?harm 2h = fu 2h 2 H 2h : a(u 2h ; v 2h ) = 0; 8v 2h 2 H 2h \ H h g: With this denition, one can show [17] that the error propagates according to a formula of the same structure as in the additive Schwarz case: e k+1 = (I? (P H 2h?harm 2h + P Hh ))e k ; and the method is a particular case of (2.3) with V 1 = H h and V 2 = H 2h?harm 2h. For problem-specic analysis of AFAC, see [17] for the case of two levels and [9] for a general number of levels. 3

4 Douglas and Miranker [6] studied a method in which the subspaces V i are dened as spaces of functions which satisfy suitable symmetry and antisymmetry properties and gave algebraical conditions under which the subspaces V i are mutually orthogonal; then the method (2.3) reduces to a direct method. The \robust multigrid method" by Hackbush [13] is also of the form (2.3) with the subspaces V i being dened as ranges of suitable stencil operators (\prolongations") on a uniform grid. Douglas and Smith [7] studied several other methods within this framework and proved a convergence bound for Hackbusch's method. For convergence estimates in the case when the subproblems in (2.3) are solved only approximately, see Douglas and Mandel [5]. 3. The spectra of sums of projections. In this section, we give several results valid for arbitrary sums of projections. In the next section we give a more complete theory for the sum of only two projections. Because we will not need to refer to the problem (2.1) any more, denote the inner product in H by (u; v) rather than a(u; v). The corresponding norm is kuk = (u; u) 1=2. Our rst theorem is an extension of Lemma 2.3 in [17] to the innite dimensional case. It shows that the bounds on the spectrum of the sum of projections on linearly independent subspaces can be reduced to the bounds on the spectra of Gram matrices of unit vectors, one from each subspace. Theorem 3.1. Let H be a Hilbert space with inner product (; ), V i closed subspaces of H, and H = n V i. Let P Vi inf ( sup ( be the orthogonal projection onto V i. Then P Vi ) = inf kv i k=1 v i 2V i min G(v 1 ; : : : ; v n ); P Vi ) = sup kv i k=1 v i 2V i max G(v 1 ; : : : ; v n ); where G(v 1 ; :::; v n ) = (g ij ) is the Gram matrix, g ij = (v i ; v j ), i; j = 1; : : : ; n. Proof. Dene the inner product in V 1 V n by (u1 ; : : : ; u n ); (v 1 ; : : : ; v n ) = V1Vn Let A : V 1 V 2 :::: V n H be given by A : (v 1 ; v 2 ; : : : ; v n ) 7 Then the adjoint A : H V 1 V 2 ::: V n is v i : A : v (P V 1v; : : : ; P Vn v) because A(0; : : : ; 0; vj ; 0; : : :); w = (v j ; w) = (v j ; P Vj w) (u i ; v i ): = (0; : : : ; 0; v j ; 0; : : : ; 0 ; (0; : : : ; 0; P Vj ; 0; : : : ; 0) V1V n = (0; : : : ; 0; v j ; 0; : : : ; 0); A w V1V n : 4

5 Consequently, AA = Because A is a bounded, one to one mapping of the Hilbert space V 1 V n onto H, its inverse is also bounded and thus (A A) = (A?1 (AA )A) = (AA ). Now write v 2 V 1 V n as v = (b 1 v 1 ; : : : ; b n v n ) with kv i k = 1, v i 2 V i, Then the Rayleigh quotient of v is P Vi : RQ(v) = (A Av; v) V 1:::V n = (v; v) V 1:::V n = P n i;j=1 b ib j (v i ; v j ) P n b 2 i (Av; Av) P n b2 i = bt G(v 1 ; : : : ; v n )b ; b t b where b = (b 1 ; :::; b n ) t. The rst part of the following theorem shows that for the spectrum of the sum of orthogonal projections to be bounded from below, it is sucient that the corresponding decomposition is bounded from above. It is due to P. L. Lions [14] and the proof is given here for completeness only. The second part of the theorem provides an analogous statement for the upper bound of the spectrum. Theorem 3.2. Let H be a Hilbert space with inner product (; ), V i closed subspaces of H, and H = P n V i. Let P Vi be the orthogonal projection onto V i. Then it holds: (i) If there exists a constant c 1 > 0 such that (3.1) then 8v 2 H 9v i 2 V i : v = v i ; kvk 2 c 1 n X inf n X P Vi c1 : kv i k 2 ; then (ii) If there is a constant c 2 such that 8v 2 H 8v i 2 V i ; v = v i : kvk 2 c 2 n X sup n X P Vi c2 : kv i k 2 ; Proof. (i) Let v = P n v i. Then kvk 2 = (v; v) = (v; v i ) = kp Vi vk kv i k ( (v; P Vi v i ) = kp Vi vk 2 ) 1=2 ( 5 (P Vi v; v i ) kv i k 2 ) 1=2 :

6 Using the assumption, we therefore get kvk 2 c 1?1 kp Vi vk 2 = c?1 1 (v; P Vi v): proving the rst part of the theorem. (ii) Let v 2 H and take w i = 0 if P Vi v = 0, and w i = P Vi v=kp Vi vk when P Vi v 6= 0. Then kp Vi vk 2 = (P Vi v; P Vi v) = (P Vi v; v) = (w i kp Vi vk; v) = (w i ; v)kp Vi vk; and, consequently, kp Vi vk = (w i ; v). Dene X : R n H by X : d 7 P n d i w i. Then X : H R n, X : u 7 f(w i ; u)g n. We now have X n v; P Vi v = kp Vi vk 2 = (w i ; v) 2 = kx vk 2 kx k 2 kvk 2 : But it holds for all d 2 R n that kxdk 2 = k w i d i k 2 c 2 n X kd i w i k 2 = c 2 n X so kxk c 1=2 2. Since kxk = kx k, it follows that X n v; P Vi v kx k 2 kvk 2 c 2 kvk 2 ; d 2 i = c 2 kdk 2 ; which concludes the proof. We should note that there is always the trivial upper bound sup P n P V i n, because all projections are orthogonal. Nontrivial upper bounds can often be obtained by dierent means; for example, for the additive Schwarz' method, it is easy to see that the upper bound can be taken to be the maximum number of subdomains having a common nonempty intersection [10]. In the case when the subspaces V i are linearly independent, the question arises, if the Lions' assumption (3.1) implies also a nontrivial upper bound, perhaps one independent of the number of subspaces n. In the case of two linearly independent subspaces, we show in the next section that the extreme points of the spectrum of the sum of the projections are symmetrical around one; however, in the general case the problem is open. 4. The case of two subspaces. In this section, let H be a Hilbert space, which is the sum of two closed subspaces, H = U + V, where possibly U \ V 6= f0g. Since all propositions hold when exchanging the roles of U and V, we may state and prove only one variant in such cases. The following lemma summarizes a number of properties we need in order to prepare for a decomposition of the space H. We note that (4.2) below is an abstract version of the specic result stated as Lemma 3.4 in [17]. Dene ~U = U \ (U \ V )? ; ~ V = V \ (U \ V )? : 6

7 Lemma 4.1. It holds that (4.1) U = U ~ (U \ V ); (4.2) P ~U?P V? = P U?P V?; (4.3) P U = P U\V + P U ~ ; (4.4) P U + P V = 2P U\V + P ~U + P ~V ; H = U ~ V ~ (U \ V ); U ~ V ~? (U \ V ); (4.5) P ~U P ~V = P UP ~V : (4.6) Proof. For (4.1), note that uniqueness of the decomposition follows from the fact that U ~ \(U \V ) = U \(U \V )? \(U \V ) = f0g. In order to prove existence, let u 2 U, u = P U\V u, and u = u + ~u. But ~u 2 U, since U \ V U, so ~u 2 (U \ V )? \ U = U. ~ Now we prove (4.2). Let v 2 V? and w = P U ~ v. For any z 2 U, write z = ~z + z, with ~z 2 U ~ and z 2 U \ V. Then from z 2 U \ V and w 2 U ~ (U \ V )?, it follows that (v; z) = (w; z) = 0: Because (w; ~z) = (v; ~z) by denition of a projection, it follows that (w; z) = (v; z) for all z 2 U, which implies that w = P U v. Consequently, P ~ U?v = P U?v. Equation (4.3) follows from (4.1) and ~ U? (U \ V ). Equation (4.4) follows immediately from (4.3). The second statement in (4.5) follows trivially from the denitions of ~ U and ~ V. To prove uniqueness in the rst statement in (4.5), note that ~U \ ~ V \ (U \ V )? = U \ V \ (U \ V )? = f0g: To prove existence, let w 2 H, w = u + v, u 2 U, v 2 V. Now by (4.1), u = ~u + u, ~u 2 ~ U, u 2 U \ V and in the same way, v = ~v + v, ~v 2 ~ V, v 2 U \ V. The proof of (4.5) is concluded by noting that w = ~u+(u+ v)+ ~v. The proof of (4.6) is completely analogous to that of (4.2). We now turn to measuring the angle between spaces and spectral radii of associated products of projections. Recall that the cosine of two vectors u; v 2 H is dened by cos(u; v) = and the cosine of two subspaces X; Y H by (u; v) kukkvk cos(x; Y ) = sup j cos(x; y)j: x2x y2y We have the following simple result, which was stated and proved, e.g., by Bank and Dupont [1] and Braess [3]. 7

8 Lemma 4.2. If H = X Y, where X and Y are closed subspaces of H, then (4.7) (P X?P Y?) cos 2 (X; Y ) We can further relate projections and the cosine of subspaces as follows. Lemma 4.3. If X; Y H are closed subspaces of H, then (4.8) (P X P Y ) = cos 2 (X; Y ): Proof. We have (P 2 X P Y u; P Y u) (P X P Y ) = kp Y P X P Y k = sup u2h kuk 2 = sup y2y (P X y; P X y) kyk 2 = cos 2 (X; Y ): (x; y) 2 = sup sup y2y x2x kxk 2 kyk 2 The following statement extends Lemma 2.2 in [17] to the general (innite dimensional) case. Lemma 4.4. If X; Y H are closed subspaces of H and H = X Y, then (4.9) cos(x? ; Y? ) = cos(x; Y ): Proof. From Lemmas 4.2 and 4.3, we get the inequality cos(x? ; Y? ) cos(x; Y ): Because orthogonal complements are closed, it will suce to show for the converse inequality that H = X? Y? : Since H = X Y, we have X? \ Y? = f0g and H = X? + Y? ; therefore, we need only to show that X? + Y? is closed. Let w 2 X? + Y?, that is, w n = u n + v n ; u n 2 X? ; v n 2 Y? ; kw n? wk 0; n 1: By [14, Theorem I.1], we have from H = X + Y that kp X?P Y?k < 1, so by Lemma 4.3, cos(x? ; Y? ) < 1: It follows that the sequences u n and v n are bounded and we can thus extract weakly convergent subsequences u nk * u 2 X? ; v nk * v 2 Y? ; k 1: 8

9 Consequently, u nk + v nk * u + v = w 2 X? + Y? ; k 1; which concludes the proof. The following theorem shows that there is a single relation of U and V. Theorem 4.5. It holds that number characterizing the cos(u? ; V? ) = cos( ~ U; V ) = cos(u; ~ V ) = cos( ~ U; ~ V ): Proof. We have cos 2 (U? ; V? ) = (P U?P V?) = (P ~ U?P V?) = cos2 ( ~ U? ; V? ) = cos 2 ( ~ U; V ) = (P V P ~ U ) = (P ~ V P ~ U ) = cos2 ( ~ U; ~ V ); using (4.8), (4.2), (4.8), (4.9), (4.7), (4.6) and (4.7) in this order. The following theorem is our rst localization of the spectrum of two projections. Theorem 4.6. Decompose H = (U \ V ) (U \ V )? : Then in the block notation corresponding to this decomposition, P U + P V = 2I 0 0 P ~ U + P ~ V with the rst block void if U \ V = f0g, and it holds inf (P ~ U + P ~ V ) = 1? cos( ~ U; ~ V ) sup (P ~ U + P ~ V ) = 1 + cos( ~ U; ~ V ) ; Proof. From (4.4), we know that P U + P V = 2P U\V + P U ~ + P V ~. It remains to show that we have the stated bounds of the spectrum of P U ~ + P V ~. But by Theorem 3.1, these bounds are the inmum and supremum of the eigenvalues of the 2 by 2 matrices 1 a a 1 with? cos( ~ U; ~ V ) a cos( ~ U; ~ V ), which are 1 a. In the case when the subspaces U and V are linearly independent, we recover from Theorems 4.5 and 4.6 a result of [17]. Corollary 4.7. If H is nite dimensional and H = U V, then (P U?P V?) = (I? P U? P V ) 2 : 9

10 In order to obtain more detailed information of the spectrum, we need to decompose our spaces U ~ and V ~ further. This will provide us with an abstract version of the decomposition given in [2] in the context of nite element spaces. Write U ~ = Up U b where U p = U ~ \ V ~? and U b = U ~ \ Up? : Similarly, V ~ = Vp V b ; with the subspaces V p and V b dened analogously. Reordering the subspaces, we can now write a decomposition of H, (4.10) H = (U \ V ) U p V p U b V b : Note that all subspaces in this decomposition are pairwise orthogonal except for the pair U b and V b. Let us use the notation P XjY to mean the ortogonal projection of the (sub)space Y to X, or, equivalently, the orthogonal projection operator onto X with the domain restricted to Y. With this notation we can state our complete decomposition. Theorem 4.8. In the block notation corresponding to the decomposition (4.10), it holds that (4.11) (4.12) P U + P V = 0 2I I I I P Ub jv b P Vb ju b I Proof. We use Theorem 4.6 and further decompose P ~ U + P ~ V on (U \ V )? = ~ U ~ V : Now P ~ V j ~ U : U p U b V p V b and (P U ~ + P V ~ )j U ~ V ~ = I U ~ P V ~ j U ~ P Uj ~ V ~ I V ~ : 1 C A P V ~ j U ~ = P V pjup P VpjUb = 0 0 ; P Vb ju p P Vb ju b 0 P Vb ju b because P VpjUp = 0, P VpjUb = 0, and P Vb ju p = 0 from U p? V ~ and V p? U. ~ Substituting into (4.12) along with the block form of identity the I U ~ and analogous expressions with U and V interchanged gives (4.11). From Theorem 4.8, we may deduce the complete structure of the spectrum. Corollary 4.9. The operator P U +P V subspaces: Eigenvalue 2 with the invariant subspace U \ V. Eigenvalue 1 with the invariant subspace U p V p. The rest of the spectrum is of the form 1 where 2 2 (P Ub jv b P Vb ju b ): has the following eigenvalues and invariant If H is nite dimensional, then the number of such eigenvalues dierent from 1 is at most 2 minfdim U b ; dim V b g. Consequently, the conjugate gradient method for the problem (2.1) preconditioned by the approximate solver (2.3) converges in at most minfdim U b ; dim V b g steps. 10

11 5. An application to nite element spaces. Here we briey explain the application of the theory to nite element spaces as used in domain decomposition algorithms [2], which motivated the general results above. Let H be a space of nite element functions with support on two overlapping subregions (1) and (2). We dene the subspaces U and V to be the corresponding spaces of nite element functions (with zero traces on the boundaries) dened on (1) and (2). Following [2] we use the notations 1 = n (2), where (2) is the closure of (2), 2 = n (1) and 3 = (1) \ (2). The region is thus also divided into three nonoverlapping subregions 1, 2, and 3 which are separated from each other by the curves (or surfaces)? 4 = 1 \ 3 and? 5 = 2 \ 3. With subvectors and subscripts corresponding to the degrees of freedom associated with the open sets 1 ; 2 and 3 and the curves (surfaces)? 4 and? 5, the entire discrete problem can be written as (5.1) Kx = 0 K K K K K 33 K 34 K 35 K T 14 0 K T 34 K 44 K 45 0 K T 25 K T 35 K T 45 K C A The stiness matrix K has been generated in the usual way by the bilinear form a(u; v). We make the unique correspondence between nite element functions u; v and the corresponding nodal values x and y. Similarly, y T Kx corresponds to a (u; v). We write any function u 2 H as the sum of 5 components u 1 to u 5 corresponding to the 5 components x 1 to x 5 of the vector x. We can then identify the spaces dened in Section 4: U: Functions u corresponding to x 1, x 3 and x 4, that is, all functions such that x 2 = 0 and x 5 = 0. V : Functions v corresponding to x 2, x 3 and x 5. U \ V : Functions u 3 with support in 3, that is, corresponding to x 3. (U \V )? : Functions which are discrete harmonic on 3, corresponding to all vectors x such that K 33 x 3 + K 34 x 4 + K 35 x 5 = 0. ~U: Vectors x with possibly only x 1, x 3, and x 4 nonzero such that K 33 x 3 +K 34 x 4 = 0. ~V : Only x 2, x 3, and x 5 may be nonzero and K 33 x 3 + K 35 x 5 = 0. ~V? : Vectors x such that K 22 x 2 + K 25 x 5 = 0 and for all y 3, y 5 such that K 33 y 3 + K 35 y 5 = 0, it holds x 1 x 2 x 3 x 4 x 5 1 C A = 0 b 1 b 2 b 3 b 4 b 5 1 C A (5.2) y T 3 (K 33 x 3 + K 34 x 4 + K 35 x 5 ) + y T 5 (K T 25x 2 + K T 35x 3 + K T 45x 4 + K 55 x 5 ) = 0: U p : Since u 2 U p implies that u 2 (U \ V )? the rst term in 5.2 is zero. In particular, (5.3) K 33 x 3 + K 34 x 4 = 0: Assuming that K 33 is invertible, it follows from (5.2) that y 5 is arbitrary and we get the necessary and sucient conditions for u 2 U p are x 2 = 0 and x 5 = 0 (since u 2 U), 11

12 and K T 35x 3 + K T 45x 4 = 0. Using (5.3), we may conclude that u 2 U p, x 2 = 0; x 5 = 0; K 33 K 34 K T 35 K T 45 x3 = 0: x 4 Thus if the matrix (5.4) K 33 K 34 K T 35 K T 45 has a trivial nullspace, then x 3 = 0 and x 4 = 0. This is the case if it is of full rank and the dimension of x 5 is at least as large as the dimension of x 4. In this case, functions from U p simply correspond to x such that only the subvector x 1 may be nonzero. U b : Here we restrict ourselves to the case when the matrix (5.4) has a trivial nullspace and K 33 and K 11 are nonsingular. Then functions in U b are given by an arbitrary component x 4, the component x 3 is determined from (5.3), x 1 from K 11 x 1 + K 14 x 4 = 0, and x 2 = 0 and x 5 = 0. In other words, a function from U b is then given by its values on? 4 and extended as a discrete harmonic function onto 1 and 3. In this particular case, the decomposition is in complete agreement with the conclusions in [2]. REFERENCES [1] R. E. Bank and T. F. Dupont, Analysis of a two-level scheme for solving nite element equations, Tech. Report CNA-159, Center for Numerical Analysis, University of Texas at Austin, [2] P. E. Bjrstad, Multiplicative and additive Schwarz methods: Convergence in the 2-domain case, in Domain Decomposition Methods for Partial Dierential Equations II, T. Chan, R. Glowinski, G. A. Meurant, J. Periaux, and O. Widlund, eds., Philadelphia, 1989, SIAM, pp. 147{159. [3] D. Braess, The contraction number of a multigrid method for solving the Poisson equation, Numerische Mathematik, 37 (1981), pp. 387{404. [4] T. Chan, R. Glowinski, G. A. Meurant, J. Periaux, and O. Widlund, eds., Domain Decomposition Methods for Partial Dierential Equations II, Philadelphia, 1989, SIAM. Proceedings of the second international symposium on domain decomposition methods for partial dierential equations, UCLA, January [5] C. C. Douglas and J. Mandel, The domain reduction method: High way reduction in three dimensions and convergence with inexact solvers, in Multigrid Methods, Proceedings of the Fourth Copper Mountain Conference on Multigrid Methods, J. Mandel, S. F. McCormick, J. E. Dendy, Jr., C. Farhat, G. Lonsdale, S. V. Parter, J. W. Ruge, and K. Stuben, eds., Philadelphia, 1989, SIAM. [6] C. C. Douglas and W. L. Miranker, Constructive interference in parallel algorithms, SIAM J. Numer. Anal., 25 (1988), pp. 376{398. [7] C. C. Douglas and B. F. Smith, Using symmetries and antisymmetries to analyze a parallel multigrid algorithm: the elliptic boundary value problem case, tech. report, IBM, Submitted to SIAM J. Num. Anal. [8] M. Dryja and O. B. Widlund, An additive variant of the Schwarz alternating method for the case of many subregions, Tech. Report 339, also Ultracomputer Note 131, Department of Computer Science, Courant Institute,

13 [9], On the optimality of an additive iterative renement method, in Multigrid Methods, Proceedings of the Fourth Copper Mountain Conference on Multigrid Methods, J. Mandel, S. McCormick, J. E. Dendy, Jr., C. Farhat, G. Lonsdale, S. V. Parter, J. W. Ruge, and K. Stuben, eds., Philadelphia, 1989, SIAM. [10], Some domain decomposition algorithms for elliptic problems. To appear in Proceedings of the Conference on Iterative Methods for Large Linear Systems, Austin, Texas, October 1988, to celebrate the Sixty-fth Birthday of David M. Young, Jr., [11] R. Glowinski, G. H. Golub, G. A. Meurant, and J. Periaux, eds., Domain Decomposition Methods for Partial Dierential Equations, Philadelphia, 1988, SIAM. Proceedings of the rst international symposium on domain decomposition methods for partial dierential equations, Paris, January [12] G. H. Golub and C. F. Van Loan, Matrix Computations, John Hopkins University Press, second ed., [13] W. Hackbusch, A new approach to robust multi-grid solvers, in ICIAM '87, Proceedings of the First International Conference on Industrial and Applied Mathematics (Paris, 1987), Philadelphia, 1988, SIAM, pp. 114{126. [14] P. L. Lions, On the Schwarz alternating method. I., in Domain Decomposition Methods for Partial Dierential Equations, R. Glowinski, G. H. Golub, G. A. Meurant, and J. Periaux, eds., Philadelphia, 1988, SIAM, pp. 1{42. [15], On the Schwarz alternating method. II., in Domain Decomposition Methods for Partial Dierential Equations II, T. Chan, R. Glowinski, G. A. Meurant, J. Periaux, and O. Widlund, eds., Philadelphia, 1989, SIAM. (Proceedings of the Second International Symposium on Domain Decomposition Methods for Partial Dierential Equations, UCLA, January 1988.). [16] J. Mandel and S. McCormick, Iterative solution of elliptic equations with renement: The model multi-level case, in Domain Decomposition Methods for Partial Dierential Equations II, T. Chan, R. Glowinski, G. A. Meurant, J. Periaux, and O. Widlund, eds., 1989, pp. 93{102. [17], Iterative solution of elliptic equations with renement: The two-level case, in Domain Decomposition Methods for Partial Dierential Equations II, T. Chan, R. Glowinski, G. A. Meurant, J. Periaux, and O. Widlund, eds., 1989, pp. 81{92. [18] H. A. Schwarz, Gesammelete Mathematische Abhandlungen, vol. 2, Springer, Berlin, 1890, pp. 133{143. First published in Vierteljahrsschrift der Naturforschenden Gesellschaft in Zurich, volume 15, 1870, pp.272{

XIAO-CHUAN CAI AND MAKSYMILIAN DRYJA. strongly elliptic equations discretized by the nite element methods.

XIAO-CHUAN CAI AND MAKSYMILIAN DRYJA. strongly elliptic equations discretized by the nite element methods. Contemporary Mathematics Volume 00, 0000 Domain Decomposition Methods for Monotone Nonlinear Elliptic Problems XIAO-CHUAN CAI AND MAKSYMILIAN DRYJA Abstract. In this paper, we study several overlapping

More information

Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence)

Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence) Math 413/513 Chapter 6 (from Friedberg, Insel, & Spence) David Glickenstein December 7, 2015 1 Inner product spaces In this chapter, we will only consider the elds R and C. De nition 1 Let V be a vector

More information

MULTIPLICATIVE SCHWARZ ALGORITHMS FOR SOME NONSYMMETRIC AND INDEFINITE PROBLEMS. XIAO-CHUAN CAI AND OLOF B. WIDLUND y

MULTIPLICATIVE SCHWARZ ALGORITHMS FOR SOME NONSYMMETRIC AND INDEFINITE PROBLEMS. XIAO-CHUAN CAI AND OLOF B. WIDLUND y MULTIPLICATIVE SCHWARZ ALGORITHMS FOR SOME NONSYMMETRIC AND INDEFINITE PROBLEMS IAO-CHUAN CAI AND OLOF B. WIDLUND y Abstract. The classical Schwarz alternating method has recently been generalized in several

More information

ADDITIVE SCHWARZ FOR SCHUR COMPLEMENT 305 the parallel implementation of both preconditioners on distributed memory platforms, and compare their perfo

ADDITIVE SCHWARZ FOR SCHUR COMPLEMENT 305 the parallel implementation of both preconditioners on distributed memory platforms, and compare their perfo 35 Additive Schwarz for the Schur Complement Method Luiz M. Carvalho and Luc Giraud 1 Introduction Domain decomposition methods for solving elliptic boundary problems have been receiving increasing attention

More information

A MULTIGRID ALGORITHM FOR. Richard E. Ewing and Jian Shen. Institute for Scientic Computation. Texas A&M University. College Station, Texas SUMMARY

A MULTIGRID ALGORITHM FOR. Richard E. Ewing and Jian Shen. Institute for Scientic Computation. Texas A&M University. College Station, Texas SUMMARY A MULTIGRID ALGORITHM FOR THE CELL-CENTERED FINITE DIFFERENCE SCHEME Richard E. Ewing and Jian Shen Institute for Scientic Computation Texas A&M University College Station, Texas SUMMARY In this article,

More information

/00 $ $.25 per page

/00 $ $.25 per page Contemporary Mathematics Volume 00, 0000 Domain Decomposition For Linear And Nonlinear Elliptic Problems Via Function Or Space Decomposition UE-CHENG TAI Abstract. In this article, we use a function decomposition

More information

SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS

SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS SPECTRAL PROPERTIES OF THE LAPLACIAN ON BOUNDED DOMAINS TSOGTGEREL GANTUMUR Abstract. After establishing discrete spectra for a large class of elliptic operators, we present some fundamental spectral properties

More information

QUASI-UNIFORMLY POSITIVE OPERATORS IN KREIN SPACE. Denitizable operators in Krein spaces have spectral properties similar to those

QUASI-UNIFORMLY POSITIVE OPERATORS IN KREIN SPACE. Denitizable operators in Krein spaces have spectral properties similar to those QUASI-UNIFORMLY POSITIVE OPERATORS IN KREIN SPACE BRANKO CURGUS and BRANKO NAJMAN Denitizable operators in Krein spaces have spectral properties similar to those of selfadjoint operators in Hilbert spaces.

More information

ANDREA TOSELLI. Abstract. Two-level overlapping Schwarz methods are considered for nite element problems

ANDREA TOSELLI. Abstract. Two-level overlapping Schwarz methods are considered for nite element problems OVERLAPPING SCHWARZ METHODS FOR MAXWELL'S EQUATIONS IN THREE DIMENSIONS ANDREA TOSELLI Abstract. Two-level overlapping Schwarz methods are considered for nite element problems of 3D Maxwell's equations.

More information

MULTIGRID PRECONDITIONING IN H(div) ON NON-CONVEX POLYGONS* Dedicated to Professor Jim Douglas, Jr. on the occasion of his seventieth birthday.

MULTIGRID PRECONDITIONING IN H(div) ON NON-CONVEX POLYGONS* Dedicated to Professor Jim Douglas, Jr. on the occasion of his seventieth birthday. MULTIGRID PRECONDITIONING IN H(div) ON NON-CONVEX POLYGONS* DOUGLAS N ARNOLD, RICHARD S FALK, and RAGNAR WINTHER Dedicated to Professor Jim Douglas, Jr on the occasion of his seventieth birthday Abstract

More information

An additive average Schwarz method for the plate bending problem

An additive average Schwarz method for the plate bending problem J. Numer. Math., Vol. 10, No. 2, pp. 109 125 (2002) c VSP 2002 Prepared using jnm.sty [Version: 02.02.2002 v1.2] An additive average Schwarz method for the plate bending problem X. Feng and T. Rahman Abstract

More information

Convergence of The Multigrid Method With A Wavelet. Abstract. This new coarse grid operator is constructed using the wavelet

Convergence of The Multigrid Method With A Wavelet. Abstract. This new coarse grid operator is constructed using the wavelet Convergence of The Multigrid Method With A Wavelet Coarse Grid Operator Bjorn Engquist Erding Luo y Abstract The convergence of the two-level multigrid method with a new coarse grid operator is studied.

More information

Luca F. Pavarino. Dipartimento di Matematica Pavia, Italy. Abstract

Luca F. Pavarino. Dipartimento di Matematica Pavia, Italy. Abstract Domain Decomposition Algorithms for First-Order System Least Squares Methods Luca F. Pavarino Dipartimento di Matematica Universita dipavia Via Abbiategrasso 209 27100 Pavia, Italy pavarino@dragon.ian.pv.cnr.it.

More information

boundaries are aligned with T h (cf. Figure 1). The union [ j of the subdomain boundaries will be denoted by. Figure 1 The boundaries of the subdo

boundaries are aligned with T h (cf. Figure 1). The union [ j of the subdomain boundaries will be denoted by. Figure 1 The boundaries of the subdo The Condition Number of the Schur Complement in Domain Decomposition * Susanne C. Brenner Department of Mathematics University of South Carolina Columbia, SC 29208 Dedicated to Olof B. Widlund on the occasion

More information

Vector Space Basics. 1 Abstract Vector Spaces. 1. (commutativity of vector addition) u + v = v + u. 2. (associativity of vector addition)

Vector Space Basics. 1 Abstract Vector Spaces. 1. (commutativity of vector addition) u + v = v + u. 2. (associativity of vector addition) Vector Space Basics (Remark: these notes are highly formal and may be a useful reference to some students however I am also posting Ray Heitmann's notes to Canvas for students interested in a direct computational

More information

Multilevel and Adaptive Iterative Substructuring Methods. Jan Mandel University of Colorado Denver

Multilevel and Adaptive Iterative Substructuring Methods. Jan Mandel University of Colorado Denver Multilevel and Adaptive Iterative Substructuring Methods Jan Mandel University of Colorado Denver The multilevel BDDC method is joint work with Bedřich Sousedík, Czech Technical University, and Clark Dohrmann,

More information

4.3 - Linear Combinations and Independence of Vectors

4.3 - Linear Combinations and Independence of Vectors - Linear Combinations and Independence of Vectors De nitions, Theorems, and Examples De nition 1 A vector v in a vector space V is called a linear combination of the vectors u 1, u,,u k in V if v can be

More information

MULTIGRID PRECONDITIONING FOR THE BIHARMONIC DIRICHLET PROBLEM M. R. HANISCH

MULTIGRID PRECONDITIONING FOR THE BIHARMONIC DIRICHLET PROBLEM M. R. HANISCH MULTIGRID PRECONDITIONING FOR THE BIHARMONIC DIRICHLET PROBLEM M. R. HANISCH Abstract. A multigrid preconditioning scheme for solving the Ciarlet-Raviart mixed method equations for the biharmonic Dirichlet

More information

ON THE CONVERGENCE OF A DUAL-PRIMAL SUBSTRUCTURING METHOD. January 2000

ON THE CONVERGENCE OF A DUAL-PRIMAL SUBSTRUCTURING METHOD. January 2000 ON THE CONVERGENCE OF A DUAL-PRIMAL SUBSTRUCTURING METHOD JAN MANDEL AND RADEK TEZAUR January 2000 Abstract In the Dual-Primal FETI method, introduced by Farhat et al [5], the domain is decomposed into

More information

Multispace and Multilevel BDDC. Jan Mandel University of Colorado at Denver and Health Sciences Center

Multispace and Multilevel BDDC. Jan Mandel University of Colorado at Denver and Health Sciences Center Multispace and Multilevel BDDC Jan Mandel University of Colorado at Denver and Health Sciences Center Based on joint work with Bedřich Sousedík, UCDHSC and Czech Technical University, and Clark R. Dohrmann,

More information

Multigrid and Domain Decomposition Methods for Electrostatics Problems

Multigrid and Domain Decomposition Methods for Electrostatics Problems Multigrid and Domain Decomposition Methods for Electrostatics Problems Michael Holst and Faisal Saied Abstract. We consider multigrid and domain decomposition methods for the numerical solution of electrostatics

More information

ETNA Kent State University

ETNA Kent State University Electronic Transactions on Numerical Analysis. Volume 11, pp. 1-24, 2000. Copyright 2000,. ISSN 1068-9613. ETNA NEUMANN NEUMANN METHODS FOR VECTOR FIELD PROBLEMS ANDREA TOSELLI Abstract. In this paper,

More information

Lecture notes: Applied linear algebra Part 1. Version 2

Lecture notes: Applied linear algebra Part 1. Version 2 Lecture notes: Applied linear algebra Part 1. Version 2 Michael Karow Berlin University of Technology karow@math.tu-berlin.de October 2, 2008 1 Notation, basic notions and facts 1.1 Subspaces, range and

More information

Linear Algebra 2 Spectral Notes

Linear Algebra 2 Spectral Notes Linear Algebra 2 Spectral Notes In what follows, V is an inner product vector space over F, where F = R or C. We will use results seen so far; in particular that every linear operator T L(V ) has a complex

More information

SOLVING MESH EIGENPROBLEMS WITH MULTIGRID EFFICIENCY

SOLVING MESH EIGENPROBLEMS WITH MULTIGRID EFFICIENCY SOLVING MESH EIGENPROBLEMS WITH MULTIGRID EFFICIENCY KLAUS NEYMEYR ABSTRACT. Multigrid techniques can successfully be applied to mesh eigenvalue problems for elliptic differential operators. They allow

More information

A SHORT NOTE COMPARING MULTIGRID AND DOMAIN DECOMPOSITION FOR PROTEIN MODELING EQUATIONS

A SHORT NOTE COMPARING MULTIGRID AND DOMAIN DECOMPOSITION FOR PROTEIN MODELING EQUATIONS A SHORT NOTE COMPARING MULTIGRID AND DOMAIN DECOMPOSITION FOR PROTEIN MODELING EQUATIONS MICHAEL HOLST AND FAISAL SAIED Abstract. We consider multigrid and domain decomposition methods for the numerical

More information

Multispace and Multilevel BDDC

Multispace and Multilevel BDDC Multispace and Multilevel BDDC Jan Mandel Bedřich Sousedík Clark R. Dohrmann February 11, 2018 arxiv:0712.3977v2 [math.na] 21 Jan 2008 Abstract BDDC method is the most advanced method from the Balancing

More information

MAT 445/ INTRODUCTION TO REPRESENTATION THEORY

MAT 445/ INTRODUCTION TO REPRESENTATION THEORY MAT 445/1196 - INTRODUCTION TO REPRESENTATION THEORY CHAPTER 1 Representation Theory of Groups - Algebraic Foundations 1.1 Basic definitions, Schur s Lemma 1.2 Tensor products 1.3 Unitary representations

More information

Linear Algebra Massoud Malek

Linear Algebra Massoud Malek CSUEB Linear Algebra Massoud Malek Inner Product and Normed Space In all that follows, the n n identity matrix is denoted by I n, the n n zero matrix by Z n, and the zero vector by θ n An inner product

More information

A Balancing Algorithm for Mortar Methods

A Balancing Algorithm for Mortar Methods A Balancing Algorithm for Mortar Methods Dan Stefanica Baruch College, City University of New York, NY 11, USA Dan Stefanica@baruch.cuny.edu Summary. The balancing methods are hybrid nonoverlapping Schwarz

More information

Linear Algebra. Session 12

Linear Algebra. Session 12 Linear Algebra. Session 12 Dr. Marco A Roque Sol 08/01/2017 Example 12.1 Find the constant function that is the least squares fit to the following data x 0 1 2 3 f(x) 1 0 1 2 Solution c = 1 c = 0 f (x)

More information

Domain Decomposition Algorithms for an Indefinite Hypersingular Integral Equation in Three Dimensions

Domain Decomposition Algorithms for an Indefinite Hypersingular Integral Equation in Three Dimensions Domain Decomposition Algorithms for an Indefinite Hypersingular Integral Equation in Three Dimensions Ernst P. Stephan 1, Matthias Maischak 2, and Thanh Tran 3 1 Institut für Angewandte Mathematik, Leibniz

More information

Figure 1. Figure. Anticipating te use of two-level domain decomposition preconditioners, we construct a triangulation of in te following way. Let be d

Figure 1. Figure. Anticipating te use of two-level domain decomposition preconditioners, we construct a triangulation of in te following way. Let be d Lower Bounds for Two-Level Additive Scwarz Preconditioners wit Small Overlap * Susanne C. Brenner Department of Matematics University of Sout Carolina Columbia, SC 908 Summary. Lower bounds for te condition

More information

A theorem on summable families in normed groups. Dedicated to the professors of mathematics. L. Berg, W. Engel, G. Pazderski, and H.- W. Stolle.

A theorem on summable families in normed groups. Dedicated to the professors of mathematics. L. Berg, W. Engel, G. Pazderski, and H.- W. Stolle. Rostock. Math. Kolloq. 49, 51{56 (1995) Subject Classication (AMS) 46B15, 54A20, 54E15 Harry Poppe A theorem on summable families in normed groups Dedicated to the professors of mathematics L. Berg, W.

More information

Chapter 4 Euclid Space

Chapter 4 Euclid Space Chapter 4 Euclid Space Inner Product Spaces Definition.. Let V be a real vector space over IR. A real inner product on V is a real valued function on V V, denoted by (, ), which satisfies () (x, y) = (y,

More information

IN p-version AND SPECTRAL ELEMENT METHODS MARIO A. CASARIN

IN p-version AND SPECTRAL ELEMENT METHODS MARIO A. CASARIN DIAGONAL EDGE PRECONDITIONERS IN p-version AND SPECTRAL ELEMENT METHODS MARIO A. CASARIN Abstract. Domain decomposition preconditioners for high-order Galerkin methods in two dimensions are often built

More information

Splitting of Expanded Tridiagonal Matrices. Seongjai Kim. Abstract. The article addresses a regular splitting of tridiagonal matrices.

Splitting of Expanded Tridiagonal Matrices. Seongjai Kim. Abstract. The article addresses a regular splitting of tridiagonal matrices. Splitting of Expanded Tridiagonal Matrices ga = B? R for Which (B?1 R) = 0 Seongai Kim Abstract The article addresses a regular splitting of tridiagonal matrices. The given tridiagonal matrix A is rst

More information

Preconditioning in H(div) and Applications

Preconditioning in H(div) and Applications 1 Preconditioning in H(div) and Applications Douglas N. Arnold 1, Ricard S. Falk 2 and Ragnar Winter 3 4 Abstract. Summarizing te work of [AFW97], we sow ow to construct preconditioners using domain decomposition

More information

The Algebraic Multigrid Projection for Eigenvalue Problems; Backrotations and Multigrid Fixed Points. Sorin Costiner and Shlomo Ta'asan

The Algebraic Multigrid Projection for Eigenvalue Problems; Backrotations and Multigrid Fixed Points. Sorin Costiner and Shlomo Ta'asan The Algebraic Multigrid Projection for Eigenvalue Problems; Backrotations and Multigrid Fixed Points Sorin Costiner and Shlomo Ta'asan Department of Applied Mathematics and Computer Science The Weizmann

More information

Lecture 23: 6.1 Inner Products

Lecture 23: 6.1 Inner Products Lecture 23: 6.1 Inner Products Wei-Ta Chu 2008/12/17 Definition An inner product on a real vector space V is a function that associates a real number u, vwith each pair of vectors u and v in V in such

More information

with Applications to Elasticity and Compressible Flow Daoqi Yang March 20, 1997 Abstract

with Applications to Elasticity and Compressible Flow Daoqi Yang March 20, 1997 Abstract Stabilized Schemes for Mixed Finite Element Methods with Applications to Elasticity and Compressible Flow Problems Daoqi Yang March 20, 1997 Abstract Stabilized iterative schemes for mixed nite element

More information

Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) 1.1 The Formal Denition of a Vector Space

Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) 1.1 The Formal Denition of a Vector Space Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) Contents 1 Vector Spaces 1 1.1 The Formal Denition of a Vector Space.................................. 1 1.2 Subspaces...................................................

More information

Elementary linear algebra

Elementary linear algebra Chapter 1 Elementary linear algebra 1.1 Vector spaces Vector spaces owe their importance to the fact that so many models arising in the solutions of specific problems turn out to be vector spaces. The

More information

Adaptive Coarse Space Selection in BDDC and FETI-DP Iterative Substructuring Methods: Towards Fast and Robust Solvers

Adaptive Coarse Space Selection in BDDC and FETI-DP Iterative Substructuring Methods: Towards Fast and Robust Solvers Adaptive Coarse Space Selection in BDDC and FETI-DP Iterative Substructuring Methods: Towards Fast and Robust Solvers Jan Mandel University of Colorado at Denver Bedřich Sousedík Czech Technical University

More information

84 G. Schmidt (1.3) T : '! u := uj ; where u is the solution of (1.1{2). It is well known (cf. [7], [1]) that the linear operator T : H ( )! e H ( ) i

84 G. Schmidt (1.3) T : '! u := uj ; where u is the solution of (1.1{2). It is well known (cf. [7], [1]) that the linear operator T : H ( )! e H ( ) i Numer. Math. 69: 83{11 (1994) Numerische Mathematik c Springer-Verlag 1994 Electronic Edition Boundary element discretization of Poincare{Steklov operators Gunther Schmidt Institut fur Angewandte Analysis

More information

Optimal Interface Conditions for an Arbitrary Decomposition into Subdomains

Optimal Interface Conditions for an Arbitrary Decomposition into Subdomains Optimal Interface Conditions for an Arbitrary Decomposition into Subdomains Martin J. Gander and Felix Kwok Section de mathématiques, Université de Genève, Geneva CH-1211, Switzerland, Martin.Gander@unige.ch;

More information

Stability of implicit extrapolation methods. Abstract. Multilevel methods are generally based on a splitting of the solution

Stability of implicit extrapolation methods. Abstract. Multilevel methods are generally based on a splitting of the solution Contemporary Mathematics Volume 00, 0000 Stability of implicit extrapolation methods Abstract. Multilevel methods are generally based on a splitting of the solution space associated with a nested sequence

More information

290 J.M. Carnicer, J.M. Pe~na basis (u 1 ; : : : ; u n ) consisting of minimally supported elements, yet also has a basis (v 1 ; : : : ; v n ) which f

290 J.M. Carnicer, J.M. Pe~na basis (u 1 ; : : : ; u n ) consisting of minimally supported elements, yet also has a basis (v 1 ; : : : ; v n ) which f Numer. Math. 67: 289{301 (1994) Numerische Mathematik c Springer-Verlag 1994 Electronic Edition Least supported bases and local linear independence J.M. Carnicer, J.M. Pe~na? Departamento de Matematica

More information

A Nonoverlapping Subdomain Algorithm with Lagrange Multipliers and its Object Oriented Implementation for Interface Problems

A Nonoverlapping Subdomain Algorithm with Lagrange Multipliers and its Object Oriented Implementation for Interface Problems Contemporary Mathematics Volume 8, 998 B 0-88-0988--03030- A Nonoverlapping Subdomain Algorithm with Lagrange Multipliers and its Object Oriented Implementation for Interface Problems Daoqi Yang. Introduction

More information

The mortar element method for quasilinear elliptic boundary value problems

The mortar element method for quasilinear elliptic boundary value problems The mortar element method for quasilinear elliptic boundary value problems Leszek Marcinkowski 1 Abstract We consider a discretization of quasilinear elliptic boundary value problems by the mortar version

More information

Geometric Multigrid Methods

Geometric Multigrid Methods Geometric Multigrid Methods Susanne C. Brenner Department of Mathematics and Center for Computation & Technology Louisiana State University IMA Tutorial: Fast Solution Techniques November 28, 2010 Ideas

More information

Chapter 6: Orthogonality

Chapter 6: Orthogonality Chapter 6: Orthogonality (Last Updated: November 7, 7) These notes are derived primarily from Linear Algebra and its applications by David Lay (4ed). A few theorems have been moved around.. Inner products

More information

New Multigrid Solver Advances in TOPS

New Multigrid Solver Advances in TOPS New Multigrid Solver Advances in TOPS R D Falgout 1, J Brannick 2, M Brezina 2, T Manteuffel 2 and S McCormick 2 1 Center for Applied Scientific Computing, Lawrence Livermore National Laboratory, P.O.

More information

08a. Operators on Hilbert spaces. 1. Boundedness, continuity, operator norms

08a. Operators on Hilbert spaces. 1. Boundedness, continuity, operator norms (February 24, 2017) 08a. Operators on Hilbert spaces Paul Garrett garrett@math.umn.edu http://www.math.umn.edu/ garrett/ [This document is http://www.math.umn.edu/ garrett/m/real/notes 2016-17/08a-ops

More information

31. Successive Subspace Correction method for Singular System of Equations

31. Successive Subspace Correction method for Singular System of Equations Fourteenth International Conference on Domain Decomposition Methods Editors: Ismael Herrera, David E. Keyes, Olof B. Widlund, Robert Yates c 2003 DDM.org 31. Successive Subspace Correction method for Singular

More information

LECTURE VI: SELF-ADJOINT AND UNITARY OPERATORS MAT FALL 2006 PRINCETON UNIVERSITY

LECTURE VI: SELF-ADJOINT AND UNITARY OPERATORS MAT FALL 2006 PRINCETON UNIVERSITY LECTURE VI: SELF-ADJOINT AND UNITARY OPERATORS MAT 204 - FALL 2006 PRINCETON UNIVERSITY ALFONSO SORRENTINO 1 Adjoint of a linear operator Note: In these notes, V will denote a n-dimensional euclidean vector

More information

514 YUNHAI WU ET AL. rate is shown to be asymptotically independent of the time and space mesh parameters and the number of subdomains, provided that

514 YUNHAI WU ET AL. rate is shown to be asymptotically independent of the time and space mesh parameters and the number of subdomains, provided that Contemporary Mathematics Volume 218, 1998 Additive Schwarz Methods for Hyperbolic Equations Yunhai Wu, Xiao-Chuan Cai, and David E. Keyes 1. Introduction In recent years, there has been gratifying progress

More information

Numerische Mathematik

Numerische Mathematik Numer. Math. (1996) 75: 59 77 Numerische Mathematik c Springer-Verlag 1996 Electronic Edition A preconditioner for the h-p version of the finite element method in two dimensions Benqi Guo 1, and Weiming

More information

STABILITY OF INVARIANT SUBSPACES OF COMMUTING MATRICES We obtain some further results for pairs of commuting matrices. We show that a pair of commutin

STABILITY OF INVARIANT SUBSPACES OF COMMUTING MATRICES We obtain some further results for pairs of commuting matrices. We show that a pair of commutin On the stability of invariant subspaces of commuting matrices Tomaz Kosir and Bor Plestenjak September 18, 001 Abstract We study the stability of (joint) invariant subspaces of a nite set of commuting

More information

33 RASHO: A Restricted Additive Schwarz Preconditioner with Harmonic Overlap

33 RASHO: A Restricted Additive Schwarz Preconditioner with Harmonic Overlap Thirteenth International Conference on Domain Decomposition ethods Editors: N. Debit,.Garbey, R. Hoppe, J. Périaux, D. Keyes, Y. Kuznetsov c 001 DD.org 33 RASHO: A Restricted Additive Schwarz Preconditioner

More information

Ole Christensen 3. October 20, Abstract. We point out some connections between the existing theories for

Ole Christensen 3. October 20, Abstract. We point out some connections between the existing theories for Frames and pseudo-inverses. Ole Christensen 3 October 20, 1994 Abstract We point out some connections between the existing theories for frames and pseudo-inverses. In particular, using the pseudo-inverse

More information

A Finite Element Method for an Ill-Posed Problem. Martin-Luther-Universitat, Fachbereich Mathematik/Informatik,Postfach 8, D Halle, Abstract

A Finite Element Method for an Ill-Posed Problem. Martin-Luther-Universitat, Fachbereich Mathematik/Informatik,Postfach 8, D Halle, Abstract A Finite Element Method for an Ill-Posed Problem W. Lucht Martin-Luther-Universitat, Fachbereich Mathematik/Informatik,Postfach 8, D-699 Halle, Germany Abstract For an ill-posed problem which has its origin

More information

Linear Algebra. Paul Yiu. Department of Mathematics Florida Atlantic University. Fall A: Inner products

Linear Algebra. Paul Yiu. Department of Mathematics Florida Atlantic University. Fall A: Inner products Linear Algebra Paul Yiu Department of Mathematics Florida Atlantic University Fall 2011 6A: Inner products In this chapter, the field F = R or C. We regard F equipped with a conjugation χ : F F. If F =

More information

Two new enriched multiscale coarse spaces for the Additive Average Schwarz method

Two new enriched multiscale coarse spaces for the Additive Average Schwarz method 346 Two new enriched multiscale coarse spaces for the Additive Average Schwarz method Leszek Marcinkowski 1 and Talal Rahman 2 1 Introduction We propose additive Schwarz methods with spectrally enriched

More information

ROB STEVENSON (\non-overlapping") errors in each subspace, that is, the errors that are not contained in the subspaces corresponding to coarser levels

ROB STEVENSON (\non-overlapping) errors in each subspace, that is, the errors that are not contained in the subspaces corresponding to coarser levels Report No. 9533, University of Nijmegen. Submitted to Numer. Math. A ROBUST HIERARCHICAL BASIS PRECONDITIONER ON GENERAL MESHES ROB STEVENSON Abstract. In this paper, we introduce a multi-level direct

More information

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces.

Math 350 Fall 2011 Notes about inner product spaces. In this notes we state and prove some important properties of inner product spaces. Math 350 Fall 2011 Notes about inner product spaces In this notes we state and prove some important properties of inner product spaces. First, recall the dot product on R n : if x, y R n, say x = (x 1,...,

More information

INRIA. B.P. 105 Le Chesnay Cedex France. Moulay D. Tidriri. NASA Langley Research Center. Hampton VA Abstract

INRIA. B.P. 105 Le Chesnay Cedex France. Moulay D. Tidriri. NASA Langley Research Center. Hampton VA Abstract Convergence Analysis of Domain Decomposition Algorithms with Full Overlapping for the Advection-Diusion Problems. Patrick LeTallec INRIA Domaine de Voluceau Rocquencourt B.P. 05 Le Chesnay Cedex France

More information

if <v;w>=0. The length of a vector v is kvk, its distance from 0. If kvk =1,then v is said to be a unit vector. When V is a real vector space, then on

if <v;w>=0. The length of a vector v is kvk, its distance from 0. If kvk =1,then v is said to be a unit vector. When V is a real vector space, then on Function Spaces x1. Inner products and norms. From linear algebra, we recall that an inner product for a complex vector space V is a function < ; >: VV!C that satises the following properties. I1. Positivity:

More information

20. A Dual-Primal FETI Method for solving Stokes/Navier-Stokes Equations

20. A Dual-Primal FETI Method for solving Stokes/Navier-Stokes Equations Fourteenth International Conference on Domain Decomposition Methods Editors: Ismael Herrera, David E. Keyes, Olof B. Widlund, Robert Yates c 23 DDM.org 2. A Dual-Primal FEI Method for solving Stokes/Navier-Stokes

More information

The antitriangular factorisation of saddle point matrices

The antitriangular factorisation of saddle point matrices The antitriangular factorisation of saddle point matrices J. Pestana and A. J. Wathen August 29, 2013 Abstract Mastronardi and Van Dooren [this journal, 34 (2013) pp. 173 196] recently introduced the block

More information

The Best Circulant Preconditioners for Hermitian Toeplitz Systems II: The Multiple-Zero Case Raymond H. Chan Michael K. Ng y Andy M. Yip z Abstract In

The Best Circulant Preconditioners for Hermitian Toeplitz Systems II: The Multiple-Zero Case Raymond H. Chan Michael K. Ng y Andy M. Yip z Abstract In The Best Circulant Preconditioners for Hermitian Toeplitz Systems II: The Multiple-ero Case Raymond H. Chan Michael K. Ng y Andy M. Yip z Abstract In [0, 4], circulant-type preconditioners have been proposed

More information

Abstract. scalar steady and unsteady convection-diusion equations discretized by nite element, or nite

Abstract. scalar steady and unsteady convection-diusion equations discretized by nite element, or nite Local Multiplicative Schwarz Algorithms for Steady and Unsteady Convection-Diusion Equations Xiao-Chuan Cai Marcus Sarkis y Abstract In this paper, we develop a new class of overlapping Schwarz type algorithms

More information

Chapter 1 Foundations of Elliptic Boundary Value Problems 1.1 Euler equations of variational problems

Chapter 1 Foundations of Elliptic Boundary Value Problems 1.1 Euler equations of variational problems Chapter 1 Foundations of Elliptic Boundary Value Problems 1.1 Euler equations of variational problems Elliptic boundary value problems often occur as the Euler equations of variational problems the latter

More information

2 G. D. DASKALOPOULOS AND R. A. WENTWORTH general, is not true. Thus, unlike the case of divisors, there are situations where k?1 0 and W k?1 = ;. r;d

2 G. D. DASKALOPOULOS AND R. A. WENTWORTH general, is not true. Thus, unlike the case of divisors, there are situations where k?1 0 and W k?1 = ;. r;d ON THE BRILL-NOETHER PROBLEM FOR VECTOR BUNDLES GEORGIOS D. DASKALOPOULOS AND RICHARD A. WENTWORTH Abstract. On an arbitrary compact Riemann surface, necessary and sucient conditions are found for the

More information

MATRICES ARE SIMILAR TO TRIANGULAR MATRICES

MATRICES ARE SIMILAR TO TRIANGULAR MATRICES MATRICES ARE SIMILAR TO TRIANGULAR MATRICES 1 Complex matrices Recall that the complex numbers are given by a + ib where a and b are real and i is the imaginary unity, ie, i 2 = 1 In what we describe below,

More information

Where is matrix multiplication locally open?

Where is matrix multiplication locally open? Linear Algebra and its Applications 517 (2017) 167 176 Contents lists available at ScienceDirect Linear Algebra and its Applications www.elsevier.com/locate/laa Where is matrix multiplication locally open?

More information

1 Constrained Optimization

1 Constrained Optimization 1 Constrained Optimization Let B be an M N matrix, a linear operator from the space IR N to IR M with adjoint B T. For (column vectors) x IR N and y IR M we have x B T y = Bx y. This vanishes for all y

More information

S MALASSOV The theory developed in this paper provides an approach which is applicable to second order elliptic boundary value problems with large ani

S MALASSOV The theory developed in this paper provides an approach which is applicable to second order elliptic boundary value problems with large ani SUBSTRUCTURNG DOMAN DECOMPOSTON METHOD FOR NONCONFORMNG FNTE ELEMENT APPROXMATONS OF ELLPTC PROBLEMS WTH ANSOTROPY SY MALASSOV y Abstract An optimal iterative method for solving systems of linear algebraic

More information

Near convexity, metric convexity, and convexity

Near convexity, metric convexity, and convexity Near convexity, metric convexity, and convexity Fred Richman Florida Atlantic University Boca Raton, FL 33431 28 February 2005 Abstract It is shown that a subset of a uniformly convex normed space is nearly

More information

Numerische Mathematik

Numerische Mathematik umer. Math. 73: 149 167 (1996) umerische Mathematik c Springer-Verlag 1996 Overlapping Schwarz methods on unstructured meshes using non-matching coarse grids Tony F. Chan 1, Barry F. Smith 2, Jun Zou 3

More information

PRODUCT OF OPERATORS AND NUMERICAL RANGE

PRODUCT OF OPERATORS AND NUMERICAL RANGE PRODUCT OF OPERATORS AND NUMERICAL RANGE MAO-TING CHIEN 1, HWA-LONG GAU 2, CHI-KWONG LI 3, MING-CHENG TSAI 4, KUO-ZHONG WANG 5 Abstract. We show that a bounded linear operator A B(H) is a multiple of a

More information

Peter Deuhard. for Symmetric Indenite Linear Systems

Peter Deuhard. for Symmetric Indenite Linear Systems Peter Deuhard A Study of Lanczos{Type Iterations for Symmetric Indenite Linear Systems Preprint SC 93{6 (March 993) Contents 0. Introduction. Basic Recursive Structure 2. Algorithm Design Principles 7

More information

A Balancing Algorithm for Mortar Methods

A Balancing Algorithm for Mortar Methods A Balancing Algorithm for Mortar Methods Dan Stefanica Baruch College, City University of New York, NY, USA. Dan_Stefanica@baruch.cuny.edu Summary. The balancing methods are hybrid nonoverlapping Schwarz

More information

ADJOINTS, ABSOLUTE VALUES AND POLAR DECOMPOSITIONS

ADJOINTS, ABSOLUTE VALUES AND POLAR DECOMPOSITIONS J. OPERATOR THEORY 44(2000), 243 254 c Copyright by Theta, 2000 ADJOINTS, ABSOLUTE VALUES AND POLAR DECOMPOSITIONS DOUGLAS BRIDGES, FRED RICHMAN and PETER SCHUSTER Communicated by William B. Arveson Abstract.

More information

MAT Linear Algebra Collection of sample exams

MAT Linear Algebra Collection of sample exams MAT 342 - Linear Algebra Collection of sample exams A-x. (0 pts Give the precise definition of the row echelon form. 2. ( 0 pts After performing row reductions on the augmented matrix for a certain system

More information

M.A. Botchev. September 5, 2014

M.A. Botchev. September 5, 2014 Rome-Moscow school of Matrix Methods and Applied Linear Algebra 2014 A short introduction to Krylov subspaces for linear systems, matrix functions and inexact Newton methods. Plan and exercises. M.A. Botchev

More information

The Dirichlet-to-Neumann operator

The Dirichlet-to-Neumann operator Lecture 8 The Dirichlet-to-Neumann operator The Dirichlet-to-Neumann operator plays an important role in the theory of inverse problems. In fact, from measurements of electrical currents at the surface

More information

Domain Decomposition Preconditioners for Spectral Nédélec Elements in Two and Three Dimensions

Domain 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 information

1. Subspaces A subset M of Hilbert space H is a subspace of it is closed under the operation of forming linear combinations;i.e.,

1. Subspaces A subset M of Hilbert space H is a subspace of it is closed under the operation of forming linear combinations;i.e., Abstract Hilbert Space Results We have learned a little about the Hilbert spaces L U and and we have at least defined H 1 U and the scale of Hilbert spaces H p U. Now we are going to develop additional

More information

OUTLINE 1. Introduction 1.1 Notation 1.2 Special matrices 2. Gaussian Elimination 2.1 Vector and matrix norms 2.2 Finite precision arithmetic 2.3 Fact

OUTLINE 1. Introduction 1.1 Notation 1.2 Special matrices 2. Gaussian Elimination 2.1 Vector and matrix norms 2.2 Finite precision arithmetic 2.3 Fact Computational Linear Algebra Course: (MATH: 6800, CSCI: 6800) Semester: Fall 1998 Instructors: { Joseph E. Flaherty, aherje@cs.rpi.edu { Franklin T. Luk, luk@cs.rpi.edu { Wesley Turner, turnerw@cs.rpi.edu

More information

2 Garrett: `A Good Spectral Theorem' 1. von Neumann algebras, density theorem The commutant of a subring S of a ring R is S 0 = fr 2 R : rs = sr; 8s 2

2 Garrett: `A Good Spectral Theorem' 1. von Neumann algebras, density theorem The commutant of a subring S of a ring R is S 0 = fr 2 R : rs = sr; 8s 2 1 A Good Spectral Theorem c1996, Paul Garrett, garrett@math.umn.edu version February 12, 1996 1 Measurable Hilbert bundles Measurable Banach bundles Direct integrals of Hilbert spaces Trivializing Hilbert

More information

October 25, 2013 INNER PRODUCT SPACES

October 25, 2013 INNER PRODUCT SPACES October 25, 2013 INNER PRODUCT SPACES RODICA D. COSTIN Contents 1. Inner product 2 1.1. Inner product 2 1.2. Inner product spaces 4 2. Orthogonal bases 5 2.1. Existence of an orthogonal basis 7 2.2. Orthogonal

More information

MATH 235: Inner Product Spaces, Assignment 7

MATH 235: Inner Product Spaces, Assignment 7 MATH 235: Inner Product Spaces, Assignment 7 Hand in questions 3,4,5,6,9, by 9:3 am on Wednesday March 26, 28. Contents Orthogonal Basis for Inner Product Space 2 2 Inner-Product Function Space 2 3 Weighted

More information

AMG for a Peta-scale Navier Stokes Code

AMG for a Peta-scale Navier Stokes Code AMG for a Peta-scale Navier Stokes Code James Lottes Argonne National Laboratory October 18, 2007 The Challenge Develop an AMG iterative method to solve Poisson 2 u = f discretized on highly irregular

More information

Solving Large Nonlinear Sparse Systems

Solving Large Nonlinear Sparse Systems Solving Large Nonlinear Sparse Systems Fred W. Wubs and Jonas Thies Computational Mechanics & Numerical Mathematics University of Groningen, the Netherlands f.w.wubs@rug.nl Centre for Interdisciplinary

More information

1 Vectors. Notes for Bindel, Spring 2017 Numerical Analysis (CS 4220)

1 Vectors. Notes for Bindel, Spring 2017 Numerical Analysis (CS 4220) Notes for 2017-01-30 Most of mathematics is best learned by doing. Linear algebra is no exception. You have had a previous class in which you learned the basics of linear algebra, and you will have plenty

More information

Algebra II. Paulius Drungilas and Jonas Jankauskas

Algebra II. Paulius Drungilas and Jonas Jankauskas Algebra II Paulius Drungilas and Jonas Jankauskas Contents 1. Quadratic forms 3 What is quadratic form? 3 Change of variables. 3 Equivalence of quadratic forms. 4 Canonical form. 4 Normal form. 7 Positive

More information

10.6 ITERATIVE METHODS FOR DISCRETIZED LINEAR EQUATIONS

10.6 ITERATIVE METHODS FOR DISCRETIZED LINEAR EQUATIONS 10.6 ITERATIVE METHODS FOR DISCRETIZED LINEAR EQUATIONS 769 EXERCISES 10.5.1 Use Taylor expansion (Theorem 10.1.2) to give a proof of Theorem 10.5.3. 10.5.2 Give an alternative to Theorem 10.5.3 when F

More information

INTERGRID OPERATORS FOR THE CELL CENTERED FINITE DIFFERENCE MULTIGRID ALGORITHM ON RECTANGULAR GRIDS. 1. Introduction

INTERGRID OPERATORS FOR THE CELL CENTERED FINITE DIFFERENCE MULTIGRID ALGORITHM ON RECTANGULAR GRIDS. 1. Introduction Trends in Mathematics Information Center for Mathematical Sciences Volume 9 Number 2 December 2006 Pages 0 INTERGRID OPERATORS FOR THE CELL CENTERED FINITE DIFFERENCE MULTIGRID ALGORITHM ON RECTANGULAR

More information

Springer-Verlag Berlin Heidelberg

Springer-Verlag Berlin Heidelberg SOME CHARACTERIZATIONS AND PROPERTIES OF THE \DISTANCE TO ILL-POSEDNESS" AND THE CONDITION MEASURE OF A CONIC LINEAR SYSTEM 1 Robert M. Freund 2 M.I.T. Jorge R. Vera 3 Catholic University of Chile October,

More information