342 Difference Schemes with Constant Coefficients
The value yj+i on the relevant layer is to be determined by the recursion
j = 0, 1, ....
For the sake of simplicity we are working in a parallelepiped G such
that { 0 < x°' < l°', a= 1, 2, ... , p} and on a certain grid wh = {xi=
(i 1 h 1 , ... ,iphp), i°' = 0, 1, ... ,Na, h°' = l°'N;;^1 } equidistant in each of
the directions x °'. We refer to the (2p + 1 )-point operator A of second-order
approximation with the values
p
( 4) A y = L AO! y,
a=l
where y is the value of the function y(x, t) at a fixed node x = (i 1 h 1 , ... ,
ip hp), y(±la) = y(x(±la), t), x(±la) = (i1 h1 J ... , i0!-1 hO!-]J (iO! ± 1) hO!,
ia+i ha+i, ... , ip hp) is the node adjacent to x (x( +la l stands on the right
from x and x(-la) stands on the left from x). With the aid of the well-
established relations
I h^12 = h2 1 + h2 2 + ... + h2 p ,
we calculate the residual
provided the asymptotic expansion <pj = f(x, tj) + 0 (I h 12 + T) holds.
Scheme (3) is conditionally stable in the space C with respect to the
initial data, the right-hand side and the boundary data. The maximum
principle for the difference problem (3) may be of help in establishing the
indicated properties with further reference to the canonical form
p
+L ~ (yj(x(+Ia))+yj(x(-Ia)))+r<pj.
a=l °'