192 Homogeneous Difference Schemes
By the same token,
so that
( 44)
1
h T' z
ri-1/2
1/J, * = 12r h2 ( qu - f)i I + 0( h2) ,
z
giving 11 r1/i* lie= O(h^2 ).
0
It is clear that the function 77; =a; ur,i - (ku');- 1 ; 2 = O(h^2 ), n1eaning
0
77i = ri-1/2 77; ,
0 2
(45) 77i = 0( h ).
Comparison of formulae (35) and ( 43) yields iJ = h, v. This provides sup-
port for the view that z; = O(h), because iJ = O(h^2 ) as stated above.
In order to estimate the enor z = y - u, we shall need an auxiliary
lemma.
Lemma Let z be a solution of problem (41) and let v be a solution of the
same problem for d; = 0, i = 1, 2,, .. , N - 1, q 0 = 0. Then the inequality
is valid:
(46) 11 z 11 c = 111ax I Z; I :::; 2 11 v 11 c ·
0'5oz<N
To prove this assertion, it suffices to use only the lemma of Chapter 1,
Section 1, Subsection 8 and to write down the equations for z and z - v in
the form ( 41).
The function vi can be recovered in explicit fonn from the conditions
Wi+1 - W;
----=-1fi,, i=l,2, ... ,N-1,
hr· z
that is,
where 1/ii is specified by formula ( 42) for di - 0. Summing the equations
wk+ 1 =wk - h rk 1/ik over k = 1, 2, ... , j yields
( 47)
j
W1+1 = w1 - L hrk1fk,
k=l
h2
W1 = -S ZJ.