Page 185 - Basic Structured Grid Generation
P. 185

174  Basic Structured Grid Generation

                                                                3
                          When M is a two-dimensional surface in E , we can use surface co-ordinates u α
                                                                       ij
                                   i
                        in place of x , and the surface metric a αβ  in place of γ . With N again a unit square
                        with G ij = δ ij ,weget
                                                               k
                                                                  k
                                               1      αβ  γ  ∂ξ ∂ξ √     1  2
                                           E =       a (u )          a du du ,             (6.80)
                                                               α
                                               2   M        ∂u ∂u β
                        where all indices are summed from 1 to 2. Thus we have a variational problem of the
                        form (6.11). The two Euler-Lagrange eqns (6.12) may be expressed here as
                                           √         	      √
                                        ∂(e a)     ∂     ∂(e a)
                                               −                   = 0,   i = 1, 2,        (6.81)
                                          ∂ξ i   ∂u γ  ∂ ∂ξ /∂u γ
                                                           i
                                                                                 i
                        with summation over γ . Since no terms in e depend explicitly on ξ ,and
                                             k
                                   ∂        ∂ξ ∂ξ k      γ  ∂ξ  k  ∂ξ k  γ  ∂ξ i  γ  ∂ξ i  γ
                                                    = δ ik δ   +     δ ik δ =   δ +     δ ,
                                   i   γ     α   β       α   β     α    β     β α      α β
                              ∂ ∂ξ /∂u     ∂u ∂u           ∂u    ∂u         ∂u      ∂u
                        we obtain the equations
                          1 ∂    √   αβ     ∂ξ  i  γ  ∂ξ i  γ      1 ∂    √     γβ  ∂ξ i  αγ  ∂ξ i
                                  aa        δ +      δ    =          a a       + a         = 0.
                          2 ∂u γ         ∂u β α  ∂u α β     2 ∂u γ         ∂u β      ∂u α
                                                αβ
                          Using the symmetry of a , we have, finally,
                                              ∂    √   γβ  ∂ξ i
                                                    aa        = 0,   i = 1, 2.             (6.82)
                                             ∂u γ        ∂u β
                                             √
                          Pre-multiplying by 1/ a shows by eqn (3.160) that the harmonic mapping in this
                        case must satisfy the pair of Beltramian equations
                                                        1
                                                                2
                                                      B ξ =   B ξ = 0                      (6.83)
                        which are the same as eqns (5.104) in the absence of the control functions P , Q.
                          Returning to the more general form (6.76) with N a unit square or cube and G kl =
                        δ kl , a similar procedure shows that the Euler-Lagrange equations are

                                           1  ∂    √   jk  ∂ξ i
                                          √    j    γγ     k  = 0,  i = 1,...,n,           (6.84)
                                           γ ∂x          ∂x
                        which can be expressed as
                                                2 i
                                               ∂ ξ      1 ∂ξ i  ∂  √   jk
                                            jk
                                           γ         + √          ( γγ   ) = 0.            (6.85)
                                                             k
                                                j
                                              ∂x ∂x k    γ ∂x ∂x j
                          This equation can be inverted to produce a different partial differential equation
                                    i
                                      k
                        satisfied by x (ξ ), by making use of the identity
                                                               2 l
                                                2 i
                                                                     i
                                               ∂ ξ   ∂x l     ∂ x  ∂ξ ∂ξ  m
                                                        =−                ,
                                                                     j
                                                   k
                                                j
                                                                  i
                                                              m
                                              ∂x ∂x ∂ξ i    ∂ξ ∂ξ ∂x ∂x  k
   180   181   182   183   184   185   186   187   188   189   190