Difference schemes as operator equations. General forinulations 127
Stability of the scheme is to be understood as the property that the inverse
Ah^1 acting from B 1 ~^2 ) into B~^1 ) is uniformly bounded in h:
(24) II Ah^1 II < M where M > 0 does not depend on h.
Combination of (23) and (24) gives estimate (22):
II Yh ll(h) <II Ahl II. II 'Ph ll(h) < M II 'Ph ll(h)'
That is to say, the meaning of stability of scheme (21) is that a solution (21)
depends continuously on the right-hand side and this dependence is uniform
in the parameter h. This implies that a small change of the right-hand side
results in a small change of the solution. If the scheme is solvable and
stable, it is correct. Note that the uniqueness of the scheme (21) solution
is a consequence of its solvability and stability and, hence, we might get
rid of the uniqueness requirement in condition (1). Indeed, assume to the
contrary that there were two solutions to equation (21), say ·fh and yh f- ·fh.
By the linearity property of the operator Ah, their difference zh = yh - ·fh
should satisfy the homogeneous equation
Ah zh =Ah Cfh - Yh) =Ah 'fh - Ah Yh ='Ph - 'Ph= 0 ·
Because scheme (21) is stable, inequality (22) holds true and, therefore,
implying that Yh = y 1 ,.
To prove the stability of (21), we need an a priori estimate of the form
(22). A derivation of some a priori estimates for the operator equation (21)
will be carried out in Section 4. A difference scheme Ah Yh = 'Ph is said
to be ill-posed if at least one of the conditions (1)-(2) we have imposed
above is violated..
Suppose that a solution Yh of problem (21) exists for any 'Ph E B,;^2 l, so
that yh^1 = Ah^1 'Ph·· Since B~^1 ) and B~^2 ) are finite-dimensional spaces, the
inverse operator Ah^1 from B,;^2 ) into B~^1 ) is bounded and its norm equals
II Ah^1 II= j\lfh, where Mh is a positive constant depending on the parameter
h. If the scheme is stable, there exists a constant M > 0 independent of h
such that M h < M for all I h I < h 0. The meaning of instability of scheme
(21) and, hence, of its ill-posedness is that Mh ---+ oo as I h I---+ 0, it being
understood that the above constant M does not exist. For an ill-posed
problem only an estimate of the form
(25)