Partial Differential Equations - Modelling and ... - ResearchGate
Partial Differential Equations - Modelling and ... - ResearchGate Partial Differential Equations - Modelling and ... - ResearchGate
152 G. Bencteux et al. m 1 n C 1 0 C = m p N b= (p+1) n/2 0 C p N = m + ... + m 1 p Fig. 1. Block structure of the matrices C. n n 0 0 H 1 0 H = 0 H p D = N = (p+1) n/2 b N = (p+1) n/2 b Fig. 2. Block structure of the matrices H and D. A detailed justification of the choice of this structure is given in [BCHL07]. Let us only mention here that the decomposition is suggested from the localization of electrons and the use of a localized basis set. Note that each block overlaps only with its first neighbors. Again for simplicity, we expose the method in the case where overlapping is exactly n/2, but it could be any integer smaller than n/2. The resulting minimization problem can be recast as { p∑ inf Tr ( H i C i C t ) i , Ci ∈M n,mi (R), m i ∈ N, Ci t C i = I mi ∀ 1 ≤ i ≤ p, i=1 } p∑ Ci t TC i+1 =0 ∀ 1 ≤ i ≤ p − 1, m i = N . (6) i=1 In the above formula, T ∈M n,n (R) is the matrix defined by { 1 if k − l = n T kl = 2 , 0 otherwise, (7) H i ∈M n,n (R) is a symmetric submatrix of H (see Figure 2), and
Domain Decomposition Approach for Computational Chemistry 153 Tr H 1 H p C 1 0 0 C p C 1 0 0 C p t = p Σ i=1 Tr H i Ci C i t C 1 0 C 1 0 C p t 0 0 C p = 0 CT t i Ci+1 t Ci Ci . In this way, we replace the N(N+1) 2 global scalar constraints C t C = I N involving vectors of size N b ,bythe ∑ p m i(m i+1) i=1 2 local scalar constraints Ci tC i = I mi and the ∑ p−1 i=1 m im i+1 local scalar constraints Ci tTC i+1 = 0, involving vectors of size n. We would like to emphasize that we can only obtain in this way a basis of the vector space generated by the lowest N eigenvectors of H. This is the very nature of the method, which consequently cannot be applied for the search for the eigenvectors themselves. Before we describe in details the procedure employed to solve the Euler– Lagrange equations of (6) in a greater generality, let us consider, for pedagogic purpose, the following oversimplified problem: inf { 〈H 1 Z 1 ,Z 1 〉 + 〈H 2 Z 2 ,Z 2 〉, Z i ∈ R N b , 〈Z i ,Z i 〉 =1, 〈Z 1 ,Z 2 〉 =0 } . (8) We have denoted by 〈·, ·〉 the standard Euclidean scalar product on R N b . The problem (8) is not strictly speaking a particular occurrence of (6), but it shows the same characteristics and technical difficulties: a separable functional is minimized, there are constraints on variables of each term and there is a cross constraint between the two terms. The bottom line for our decomposition algorithm is to attack (8) as follows. Choose (Z1,Z 0 2) 0 satisfying the constraints and construct the sequence (Z1 k ,Z2 k ) k∈N by the following iteration procedure. Assume (Z1 k ,Z2 k )isknown, then Local step: Solve { ˜Zk 1 = arg inf { 〈H 1 Z 1 ,Z 1 〉, Z 1 ∈ R N b , 〈Z 1 ,Z 1 〉 =1, 〈Z 1 ,Z2 k 〉 =0 } , ˜Z 2 k = arg inf { 〈H 2 Z 2 ,Z 2 〉, Z 2 ∈ R N b , 〈Z 2 ,Z 2 〉 =1, 〈 ˜Z 1 k ,Z 2 〉 =0 } ; (9) Global step: Solve α ∗ = arg inf { 〈H 1 Z 1 (α),Z 1 (α)〉 + 〈H 2 Z 2 (α),Z 2 (α)〉, α ∈ R } (10) 0 where Z 1 (α) = ˜Z 1 k + α ˜Z 2 k √ , Z 2(α) = −α ˜Z 1 k + ˜Z 2 k √ , (11) 1+α 2 1+α 2
- Page 105 and 106: 98 I. Sazonov et al. In the first s
- Page 107 and 108: 100 I. Sazonov et al. Fig. 1. An ex
- Page 109 and 110: 102 I. Sazonov et al. H z 1 exact F
- Page 111 and 112: 104 I. Sazonov et al. (a) (b) Fig.
- Page 113 and 114: 106 I. Sazonov et al. Scattering Wi
- Page 115 and 116: 108 I. Sazonov et al. 6.4 Scatterin
- Page 117 and 118: 110 I. Sazonov et al. (a) (b) Fig.
- Page 119 and 120: 112 I. Sazonov et al. [MHP96] K. Mo
- Page 121 and 122: 114 R. Sanders and A.M. Tesdall I R
- Page 123 and 124: 116 R. Sanders and A.M. Tesdall imp
- Page 125 and 126: 118 R. Sanders and A.M. Tesdall alo
- Page 127 and 128: 120 R. Sanders and A.M. Tesdall (a)
- Page 129 and 130: 122 R. Sanders and A.M. Tesdall D C
- Page 131 and 132: 124 R. Sanders and A.M. Tesdall 8.6
- Page 133 and 134: 126 R. Sanders and A.M. Tesdall 0.3
- Page 135 and 136: 128 R. Sanders and A.M. Tesdall [TR
- Page 137 and 138: 132 S. Lapin et al. Ω R γ Ω 2 Γ
- Page 139 and 140: 134 S. Lapin et al. ∫ ∂ 2 ∫
- Page 141 and 142: 136 S. Lapin et al. Ω R γ Fig. 3.
- Page 143 and 144: 138 S. Lapin et al. 4 Energy Inequa
- Page 145 and 146: 140 S. Lapin et al. 5 Numerical Exp
- Page 147 and 148: 142 S. Lapin et al. Fig. 6. Contour
- Page 149 and 150: 144 S. Lapin et al. Fig. 9. Obstacl
- Page 151 and 152: Domain Decomposition and Electronic
- Page 153 and 154: Domain Decomposition Approach for C
- Page 155: Domain Decomposition Approach for C
- Page 159 and 160: Domain Decomposition Approach for C
- Page 161 and 162: Domain Decomposition Approach for C
- Page 163 and 164: Domain Decomposition Approach for C
- Page 165 and 166: Domain Decomposition Approach for C
- Page 167 and 168: Domain Decomposition Approach for C
- Page 169 and 170: Numerical Analysis of a Finite Elem
- Page 171 and 172: Numerical Analysis of a Finite Elem
- Page 173 and 174: Numerical Analysis of a Finite Elem
- Page 175 and 176: Numerical Analysis of a Finite Elem
- Page 177 and 178: Numerical Analysis of a Finite Elem
- Page 179 and 180: Numerical Analysis of a Finite Elem
- Page 181 and 182: so that |u| 2 1,ω h ≤ 2 Numerica
- Page 183 and 184: Numerical Analysis of a Finite Elem
- Page 185 and 186: Numerical Analysis of a Finite Elem
- Page 187 and 188: Numerical Analysis of a Finite Elem
- Page 189 and 190: 188 A. Bonito et al. of the model a
- Page 191 and 192: 190 A. Bonito et al. Fig. 1. The sp
- Page 193 and 194: 192 A. Bonito et al. 3 1 16 4 1 1 1
- Page 195 and 196: 194 A. Bonito et al. ∫ v n+1 h
- Page 197 and 198: 196 A. Bonito et al. −pn +2µD(v)
- Page 199 and 200: 198 A. Bonito et al. each of its pa
- Page 201 and 202: 200 A. Bonito et al. The normal vec
- Page 203 and 204: 202 A. Bonito et al. with initial c
- Page 205 and 206: 204 A. Bonito et al. Fig. 8. Jet bu
152 G. Bencteux et al.<br />
m<br />
1<br />
n<br />
C 1<br />
0<br />
C =<br />
m<br />
p<br />
N b= (p+1) n/2<br />
0<br />
C p<br />
N = m + ... + m<br />
1 p<br />
Fig. 1. Block structure of the matrices C.<br />
n<br />
n<br />
0<br />
0<br />
H 1<br />
0<br />
H =<br />
0<br />
H p<br />
D =<br />
N = (p+1) n/2<br />
b<br />
N = (p+1) n/2<br />
b<br />
Fig. 2. Block structure of the matrices H <strong>and</strong> D.<br />
A detailed justification of the choice of this structure is given in [BCHL07].<br />
Let us only mention here that the decomposition is suggested from the localization<br />
of electrons <strong>and</strong> the use of a localized basis set. Note that each block<br />
overlaps only with its first neighbors. Again for simplicity, we expose the<br />
method in the case where overlapping is exactly n/2, but it could be any<br />
integer smaller than n/2.<br />
The resulting minimization problem can be recast as<br />
{ p∑<br />
inf Tr ( H i C i C t )<br />
i , Ci ∈M n,mi (R), m i ∈ N, Ci t C i = I mi ∀ 1 ≤ i ≤ p,<br />
i=1<br />
}<br />
p∑<br />
Ci t TC i+1 =0 ∀ 1 ≤ i ≤ p − 1, m i = N . (6)<br />
i=1<br />
In the above formula, T ∈M n,n (R) is the matrix defined by<br />
{<br />
1 if k − l = n<br />
T kl =<br />
2 ,<br />
0 otherwise,<br />
(7)<br />
H i ∈M n,n (R) is a symmetric submatrix of H (see Figure 2), <strong>and</strong>