692 Methods for Solving Grid Equations
determination of ktl. In so doing the passage from the kth iteration
to the (k + l)th iteration necessitates the storage not only for the value
t (i 1 h 1 , i 2 h 2 ), but also for the value~ (i 1 h 1 , i 2 h 2 ).
- A higher-accuracy scheme in a rectangle. In Chapter 4, Section 5, the
Dirichlet problem
( 45) ~^11 = - f ( x) , J: E^0 , u =^11 ( .'t) , x E r ,
was completely posed, for which the difference schen1e of fourth-order ac-
curacy was composed in the following way:
( 46) A'y=-<p(x), xEwh, y=μ(x), XE/1i,
where
C\'=l,2,
h2 h2
<p --f+- 12 1 A^1. f+- 12 2 A^2.. f
0
Let now H = Q be a space of all grid functions given on the grid w h
and vanishing on the boundary ih. Having involved the operators A y =
-A'y and Ry= -Ay specified for any y EH, we deduce instead of (46)
that A y = <p, where <p f:- <p only at the near-boundary nodes. Applying the
estimates obtained in Chapter 4 for the operator A yields
and reveals the coefficients to be c 1 = ~ and c 2
arranges itself as a sum
- The operator R
Being concerned with the same quantities b, ~ and w 0 as in Section 5, we
- 0 0 0 0
might agree to consider / 1 = ~ / 1 and / 2 = I 2 , where / 1, I 2 were specified
in Section 4:
0 (j 0 (j
l 1 = 2 (1 + J1]) ' i2=4v0'