Next: 1.3.4 Global minima and
Up: 1.3 Preview of infinite-dimensional
Previous: 1.3.2 First variation and
Contents
Index
1.3.3 Second variation and second-order
conditions
A real-valued functional
on
is called bilinear if it is linear
in each argument (when the other one is fixed). Setting
we then
obtain a quadratic functional, or quadratic form,
on
. This is a direct generalization of the corresponding familiar
concepts in finite-dimensional vector spaces.
A
quadratic form
is called the second
variation of
at
if for all
and all
we have
|
(1.39) |
This exactly corresponds to our previous second-order
expansion (1.12) for
given by (1.35).
Repeating the same argument we used earlier to prove (1.14),
we easily establish the following
second-order necessary condition for optimality:
If
is a local minimum of
over
, then for all admissible
perturbations
we have
|
(1.40) |
In other words, the second variation
must be positive semidefinite on the space of admissible perturbations.
For local maxima, the inequality in (1.40) is reversed.
Of course, the usefulness of the condition will depend on our ability
to compute the second variation of the functionals that we will want to study.
What about a second-order sufficient
condition for optimality? By analogy with the second-order sufficient
condition (1.16) which we derived for the
finite-dimensional case, we may guess that we need to combine the first-order
necessary condition (1.37) with the strict-inequality counterpart
of the second-order
necessary condition (1.40), i.e.,
|
(1.41) |
(this should again hold for all admissible perturbations
with respect to a subset
of
over which we want
to be a minimum).
We would then hope to show that for
the second-order term in (1.39)
dominates the higher-order term
, which would imply that
is a strict local minimum (since the first-order term is 0).
Our earlier proof of sufficiency of (1.16) followed the same
idea. However, examining that proof more closely, the reader
will discover that in the present
case the argument does not go through.
We know that there exists an
such that
for all nonzero
with
we have
. Using this inequality and (1.37), we
obtain from (1.39) that
. Note
that this does not
yet prove that
is a (strict) local minimum of
. According
to the definition of a local minimum, we must show that
is the lowest value
of
in some ball around
with respect to the selected norm
on
. The problem is that
the term
and hence the above
depend on the choice of
the perturbation
. In the finite-dimensional case we took the minimum of
over all perturbations of unit length, but we cannot do that here because
the unit sphere in the infinite-dimensional space
is not compact and the
Weierstrass Theorem does not apply to it (see Section 1.3.4 below).
One way to resolve the above difficulty would be as follows. The first step is to strengthen the condition (1.41) to
|
(1.42) |
for some number
.
The
property (1.42) does not automatically follow
from (1.41), again because we are in an infinite-dimensional space. (Quadratic forms
satisfying (1.42)
are sometimes called uniformly positive definite.) The second step is to modify the definitions of the
first and second
variations by explicitly requiring that the higher-order terms decay uniformly
with respect to
. We already mentioned such an alternative definition of the
first variation via the expansion (1.38). Similarly, we could define
via the following
expansion in place of (1.39):
|
(1.43) |
Adopting these alternative definitions and assuming
that (1.37) and (1.42) hold, we could easily complete
the sufficiency proof by noting that
when
is small enough.
With our current definitions of the first
and second variations in terms of (1.33)
and (1.39), we do not
have
a general second-order sufficient condition for optimality. However, in
variational problems that we are going to study,
the functional
to be minimized will take a specific form. This additional structure
will allow us to derive conditions under which second-order terms dominate
higher-order terms, resulting in optimality. The above discussion
was given mainly for illustrative purposes, and will not
be directly used in the sequel.
Next: 1.3.4 Global minima and
Up: 1.3 Preview of infinite-dimensional
Previous: 1.3.2 First variation and
Contents
Index
Daniel
2010-12-20