On the Domains of General Ordinary Differential Operators in the Direct Sum Spaces ()
1. Introduction
Jiangang, Zheng and Jiong Sun [1] considered the problem of Sturm-Liouville differential equation:
(1.1)
where
are complex functions,
and
a.e. on
,
are all locally integrable functions on
,
is the so-called spectral parameter. They studied the classification of Equation (1.1) according to the number of square-integrable solutions of Equation (1.1) in suitable weighted integrable spaces. This type of classification of differential equations plays an important role in the spectral theory of differential operators as it can tell us how to obtain the operator realizations associated with the differential equations.
Amos [2] considered the problem that all solutions of the second-order ordinary differential equation
are in
when τ is a second-order symmetric ordinary differential expression of the form
on
under sufficient conditions on the coefficients p and q. The case that not all solutions are in
was considered by Atkinson and Evans in ( [3], Theorem 1). Sobhy El-Sayed and others [4] extend their results for a second-order non-symmetric ordinary differential expression
with complex coefficients.
Everitt and Zettl [5] considered the problem of characterizing all self-adjoint differential operators which can be generated by a formally symmetric Sturm-Liouville differential expression
defined on two intervals
with boundary conditions at the endpoints. Their work was motivated by Sturm-Liouville problems which occur in the literature in which the coefficients have a singularity in the interior of the underlying interval. An interesting feature of their work is the possibility of generating self-adjoint operators in this way which are not expressible as the direct sum of self-adjoint operators defined in the separate intervals.
Jiong Sun [6] gives a characterization of the self-adjoint extensions of the minimal operator
generated in
by a formally-symmetric differential expression τ of arbitrary order
. If the minimal operator
has deficiency indices
, the domain of any self-adjoint extension of
is described in terms of
boundary conditions involving the square-integrable solutions of the differential equation
for
. Thus Sun Jiong has completely solved a problem of central importance which has evaded the efforts of mathematicians for the last two decades.
J. Knowles [7] and Zai-Jiu-Shang (1988) (see [8] ) gave a characterization of the boundary conditions which determine the domain of any J-self-adjoint extension of the minimal operator
with maximal deficiency index in the case when the field of regularity,
, of
was non-empty. This is achieved by using Sun Jiong’s results (1986) (see [6] ) with only one singular endpoint.
Evans and Sobhy El-Sayed [9] gave a characterization of all regularly solvable operators and their adjoints generated by a general differential expression in Hilbert space
in the case of one interval with one singular endpoint. Also, in [4] [10] - [20], Sobhy El-Sayed gives a characterization of all regularly solvable operators in the case of one interval with two singular endpoints a and b, and a characterization of Sturm-Liouville differential operators in direct sum spaces. The domains of these operators are described in terms of boundary conditions featuring
-solutions of
and
at both singular end points a and b. Their results include those of Sun Jiong [6] concerning self-adjoint realizations of symmetric expression τ when the minimal operator has equal deficiency indices, and Zai-Jiu Shang in [8] concerning the J-self-adjoint operators as a special case.
Our objective in this research is to generalize the results of Evans and Sobhy El-Sayed, Jiong Sun, Naimark, Zettl and Zai-Jiu-Shang’s results in [5] - [12] [21] [22] [23] for the general ordinary quasi-differential expressions
each of order n with complex coefficients generated by general Shin-Zettl matrices (see [9] [24] [25] ) in the direct sum spaces such that the operators defined on each of the separate intervals
. The left-hand endpoint of
is assumed to be regular but the right-hand end-point may be regular or singular.
2. Notation and Preliminaries
We begin with a brief survey of adjoint pairs of operators and their associated regularly solvable operators; a full treatment may be found in [4] [9] [10] - [20] [23] [24] [25] ( [26], Chapter III) and [27].
The domain and range of a linear operator T acting in a Hilbert space H will be denoted by
and
respectively and
will denote its null space. The nullity of T, written
, is the dimension of
and the deficiency of T, written
, is the co-dimension of
in H; thus if T is densely defined and
is closed, then
. The Fredholm domain of T is (in the notation of [9] [24] [26] ) the open subset
of
consisting of those values of
which are such that
is a Fredholm operator, where I is the identity operator in H. Thus
if and only if
has closed range and finite nullity and deficiency. The index of
is the number
, this being defined for
.
Two closed densely defined operators A and B acting in H are said to form an adjoint pair if
and consequently,
; equivalently,
for all
and
, where
denotes the inner-product on H.
The field of regularity
of A is the set of all
for which there exists a positive constant
such that:
for all
(2.1)
or, equivalently, on using the Closed Graph Theorem, and
and
is closed.
The joint field of regularity
of A and B is the set of
which are such that
,
and both
and
may be finite. An adjoint pair of A and B is said to be compatible if
.
Now, we define a second order quasi-differential equations and quasi-derivatives.
Definition 2.1: Let the set
denotes the collection of all square matrices
of order 2 × 2 and satisfy the conditions:
1)
,
2)
, (2.2)
3)
(almost all
),
where
denote, in view of condition 3) in (2.2), the reciprocal function
(almost all
).
Given
, we define the quasi-derivatives
on I of a function
by:
(2.3)
where the prime ' denotes classical differentiation on I. Also, we define the linear manifold
by:
(2.4)
where the notations
and
, denote the linear space of functions with values in the complex field
, which are absolutely continuous and Lebesgue integrable, respectively over all compact sub-intervals of the interval
of the real line
.
The general linear ordinary quasi-differential equation of second-order given by:
on I. (2.5)
The quasi-differential Equation (2.5) is said to be Lagrange symmetric when the matrix
satisfies the additional conditions:
1)
,
2)
on I. (2.6)
As examples of the homogeneous quasi-differential Equation (2.5), i.e.,
on I, we have:
1) Let:
on I. (2.7)
Then
and (2.5) takes the form:
on I;
here
denote the classical derivatives
.
2) If
and are continuous on I, and if:
on I. (2.8)
Then
with
and the Equation (2.5) takes the form:
on I, (2.9)
which is the classical equation of the second-order with continuous coefficients.
3) The most general Lagrange symmetric equation of the second-order, see (2.3) and (2.6) is given by:
on I, (2.10)
where
almost everywhere on I,
and
. This yields the symmetric quasi-differential equation in the standard form:
on I, (2.11)
with:
If
on I then this equation reduces to the generalized Sturm-Liouville equation:
on I, (2.12)
for which
.
4) If
,
almost everywhere on I,
and
. Let,
on I, (2.13)
then the quasi-derivatives and the quasi-differential equation associated with A are defined as follows:
on I, (2.14)
(Evans’s differential expression, see [9] and [26] ). If
on I, Equation (2.14) reduces to (2.12).
We now turn to the quasi-differential expressions defined in terms of a Shin-Zettl matrix A on an interval I.
Definition 2.2: The set
of Shin-Zettl matrices on I consists of
-matrices
, whose entries are complex-valued functions on I which satisfy the following conditions:
(2.15)
For
, the quasi-derivatives associated with A are defined by:
(2.16)
The quasi-differential expression τ associated with the matrix A is given by:
(2.17)
this being defined on the set:
(2.18)
The formal adjoint
of
defined by the matrix
is given by:
, for all
; (2.19)
this being defined on the set:
(2.20)
where
, the quasi-derivatives associated with the matrix
,
for each
,
(2.21)
are therefore:
(2.22)
Note that:
and so
. We refer to [1] [4] [5] [7] [9] - [13] [20] - [27] for a full account of the above and subsequent results on quasi-differential expressions.
Definition 2.3: For
and
, we have the Green’s formula:
(2.23)
where,
(2.24)
see [4] [9] - [20] ( [23], Corollary 1) and [24] [26].
Let the interval I have end-points
(
) and let
be a non-negative weight function with
and
(for almost all
). Then
denotes the Hilbert function space of equivalence classes of Lebesgue measurable functions such that
; the inner-product is defined by:
(2.25)
The equation,
on I, (2.26)
is said to be regular at the left end-point
, if for all
,
(2.27)
Otherwise (2.26) is said to be singular at a. If (2.26) is regular at both end-points, then it is said to be regular; in this case we have,
(2.28)
We shall be concerned with the case when a is a regular end-point of (2.26), the end-point b being allowed to be either regular or singular. Note that, in view of (2.22) an end-point of I is regular (see [4] [9] - [20] [22] [23] ) for the Equation (2.26), if and only if it is regular for the equation,
on I. (2.29)
Note that, at a regular end-point a, say,
is defined for all
. Set,
(2.30)
The subspaces
and
of
are the domains of the so-called maximal operators
and
respectively, defined by:
and
.
For the regular problem the minimal operators
and
, are the restrictions of
and
to subspaces:
(2.31)
respectively. The subspaces
and
are dense in
,
and
are closed operators (see [1] [4] [9] - [20] [22] ( [23], Section 3) and [24] [26] [27] [28] ).
In the singular problem we first introduce the operators
and
;
being the restriction of
to the subspace:
(2.32)
and with
defined similarly. These operators are densely-defined and closable in
; and we defined the minimal operators
and
to be their respective closures (see [1] [5] [9] [23] [24] [26] [28] [29] ). We denote the domains of
and
by
and
respectively. It can be shown that:
(2.33)
because we are assuming that a is a regular end-point. Moreover, in both regular and singular problems, we have
and
, (2.34)
see ( [14] Section 5) in the case when
and compare with treatment in [4] [9] - [20] [22] and ( [26], Section III.10.3) in general case. Note that
and
are closed and densely-defined operators on H.
3. The Operators in Direct Sum Spaces
The operators here are no longer symmetric but direct sums:
and
, (3.1)
on any finite number of intervals
, where
is the minimal operator generated by
in
and
denotes the formal adjoint of
, which form an adjoint pair of closed operators in
. Let H be the direct sum,
. (3.2)
The elements of H will be denoted by
with
,
,
. When
, the direct sum space
can be naturally identified with the space
where
on
. This is of particular significance when
may be taken as a single interval; see [13] [19] [20] [22] [28].
We now establish by [5] [10] [12] some further notation.
(3.3)
(3.4)
Also,
(3.5)
(3.6)
, (3.7)
where
and
the inner-product defined in (2.13). Note that
is a closed densely-defined operator in H.
We summarize a few additional properties of
in the form of a Lemma.
Lemma 3.1: We have:
1)
,
.
In particular,
,
,
2)
,
.
3) The deficiency indices of
are given by
for all
,
for all
Proof: Part 1) follows immediately from the definition of
and from the general definition of an adjoint operator. The other parts are either direct consequences of part 1) or follows immediately from the definitions.
Lemma 3.3: If
are regularly solvable with respect to
and
, then
is regularly solvable with respect to:
and
.
Proof: The proof follows from Lemmas 3.1 and 3.2.
Lemma 3.4: For
,
is constant and:
.
In the case with one singular end-point:
.
In the regular problem:
, for all
.
Proof: The proof is similar to that in ( [3], lemma 3.1), [10] [11] [12] and therefore omitted.
For
, we define
and m as follows:
(3.8)
Then
and by Lemma 3.4, m is constant on
, and:
(3.9)
For
, the operators which are regularly solvable with respect to
and
are characterized by the following theorem:
Theorem 3.5: For
, let r and m be defined by (3.8), and let
be arbitrary functions satisfying:
1)
is linearly independent modulo
and
is linearly independent modulo
;
2)
,
.
Then the set:
(3.10)
is the domain of an operator
which is regularly solvable with respect to
and
and:
(3.11)
is the domain of an operator
, moreover
.
Conversely, if S is regularly solvable with respect to
and
and
, then with r and m defined by (3.8) there exist functions
and
which satisfied 1) and 2) and are such that (3.10) and (3.11) are the domains of the operators S and
respectively.
S is self-adjoint (J-self-adjoint) if, and only if,
,
and
; S is J-self-adjoint if
, (J complex conjugate),
and
.
Proof: The proof is similar to that in [6] [8] [9] [10] [11] ( [26], Theorem III.3.6) and [30].
For
, define
and
be defined by (3.8). Let
,
be bases for
and
respectively; thus
and
(3.12)
Since
has closed range, so does its adjoint
and moreover:
Hence:
and
.
We can therefore define the following:
(3.13)
(3.14)
Next, we state the following results, the proofs are similar to those in [4] [10] [11] [12] [13] [19] - [23] and ( [26], Section 4).
Lemma 3.6: ( [23], Lemma 3.3). The sets
and
are bases of
and
respectively,
.
On applying ( [26], Theorem III.3.1), [10] [11] [12] [13] [19] [20] we obtain:
Corollary 3.7: Any
and
have the unique representations
(3.15)
(3.16)
A central role in the argument is played by the matrices.
Lemma 3.8: Let,
, (3.17)
and:
, (3.18)
Then,
(3.19)
In view of Lemma 3.6 and since
,
, we may suppose, without loss of generality, that the matrices,
, (3.20)
satisfy:
(3.21)
If we partition
as:
(3.22)
and let:
(3.23)
(3.24)
Then (3.21) yields the result:
(3.25)
Lemma 3.9: Let
be the linear span
, where
satisfy the following conditions for
and some
;
(3.26)
and let
be the linear span of
with (3.21) satisfied. Then,
(3.27)
If
and
be the linear spans of
and
respectively, then:
(3.28)
4. The Boundary Conditions Featuring
-Solutions
We shall now characterize all the operators which are regularly solvable with respect to
and
in terms of boundary conditions featuring
-solutions of the equations
and
on any finite number of the intervals with one regular end-point and the other may be regular or singular. The results in this section are extension of those in [1] [4] [5] [7] [9] - [22] [27] [28] [29].
Theorem 4.1: Let
, let
and m be defined by (3.8), and let
be defined in (3.13) and (3.14) respectively, and arranged to satisfy (3.21). Let
and
,
be numerical matrices which satisfy the following conditions:
1)
and
.
2)
,
,
being the Kronecker delta.
The set of all
such that,
, (4.1)
is the domain of an operator
which is regularly solvable with respect to
and
and
is the set of all
which are such that:
. (4.2)
Proof: Let,
(4.3)
and set,
(4.4)
Then
, by [5] [10] and [12] we may choose
such that for
and some
(4.5)
This gives:
by (2.11). Also, since
on
,
. Then,
The boundary condition (4.1) therefore coincides with that in (3.10). Similarly (4.2) coincides with (3.11) on making the following choices:
(4.6)
(4.7)
and by [5] [10] and [12] we may choose
such that for
and some
,
(4.8)
It remains to show that the above functions
and
are linearly independent modulo
and
respectively and satisfy conditions 1) and 2) in Theorem 3.5. First, suppose that
is not linearly modulo
that is, there exist constants
not all zero, such that
. Then, from (2.24), (4.6) and (4.8),
On noting that:
.
But
has rank n and so we infer that:
. (4.9)
Since
, we have that
for all
.
Hence,
on using the notation in (3.23). In view of (3.25), we conclude that:
(4.10)
We obtain from (4.9) and (4.10) that:
which contradicts the assumption that
has rank r.
It follows similarly that,
is linearly independent modulo
.
Finally, we prove 2) in Theorem 3.5,
By (4.5) and (4.8),
. (4.11)
Next, we see that:
Hence,
(4.12)
From 2), (4.11) and (4.12) it follows that condition 2) in Theorem 3.5 is satisfied. The proof is therefore complete.
The converse of Theorem 4.1 is
Theorem 4.2: Let
be regularly solvable with respect to
and
, let
, let
and m be defined by (3.8), and suppose that (3.21) is satisfied. Then there exist numerical matrices
,
,
and
such that conditions 1) and 2) in Theorem 4.1 are satisfied and
is the set of
satisfying (4.1) while
is the set of
satisfying (4.2).
Proof: Let
and
satisfy the second part of Theorem 3.5. From (3.27) and (3.28), we have:
(4.13)
for some
and complex constants
and
. Let:
, (4.14)
(4.15)
Then,
Moreover, for all
,
,
, and hence, from (4.13),
Therefore, we have shown that the boundary conditions (4.1) coincide with those in (3.10). Similarly (4.2) and the conditions in (3.11) can be shown to coincide if we choose,
, (4.16)
and:
, (4.17)
where the
are the constants uniquely determined by the decomposition:
(4.18)
derived from Lemma 3.9. Next, we prove that 1) and 2) in Theorem 4.7 are consequences of the fact that
and
are linearly independent modulo
and
respectively. Suppose that:
Then there exist constants
not all zero, such that:
(4.19)
This implies that,
and as
non-singular, it follows that
, satisfies
. (4.20)
We also, infer from (4.19) that:
Consequently, on substituting (4.18), we obtain:
(4.21)
For arbitrary
it follows that
. This fact and (4.20) together imply that
and hence that
is linearly independent modulo
contrary to assumption. We have therefore proved that
has Rank r. The proof of
is similar. From (4.14) and (4.16),
On using (4.13), (4.18) and the fact that
on
,
and
if either
and
or
and
, we obtain:
The proof is therefore complete.
Remark 4.3: Assume that
is formally J-symmetric, that is
, where J is the complex conjugation. Then the operator
is the J-symmetric and
and
form an adjoint pair with:
(4.22)
Since
if and only if
(
), it follows from Lemma 3.1 that for all
,
is constant
, say, so in (3.8) and (3.9),
with
.
5. Discussion
In [5] Everitt and Zettl discussed the possibility of generating self-adjoint operators which are not expressible as the direct sums of self-adjoint operators defined in the separate intervals. In this section we extend this case to the case of general ordinary differential operators, i.e., we discuss the possibility of the regularly solvable operators which are not expressible as the direct sums of regularly solvable operators defined in the separate intervals
. We will refer to these operators as “New regularly solvable operators” if
is a regular end point and
is singular, then by ( [26], Theorem III.10.13) the sum:
for all
,
If and only if the term in (3.11) at the end point
is zero,
. By Lemma 3.4, for:
, we get in all cases:
. (5.1)
When each interval has at most one singular end-point,
. (5.2)
In the case when all end-points are regular,
, for all
.(5.3)
Let,
And:
.
Then by part 3) in Lemma 3.1, we have that
.
We now consider some of the possibilities:
Example 1.
. This is the minimal case in (5.1) and can only occur when all four end-points are singular. In this case
is itself regularly solvable and has no proper regularly solvable extensions, see Edmunds and Evans ( [26], Chapter III) [10] [13] [19] [20].
Example 2.
with one of
and
is equal to n and all the others are equal to zero. We assume that
and
. The other possibilities are entirely similar. In this case we must have seven singular end-points and one regular. There are no new regularly solvable extensions and we have that,
, where
is regularly solvable extension of
, i.e., all regularly solvable extensions of
can be obtained by forming sums of regularly solvable extensions of
,
. These are obtained as in the “one interval” case.
Example 3. Six singular end-points and
. We consider two cases:
1) One interval has two regular end-points, say,
, and each one of the others has two singular end-points. Then,
, where
is regularly solvable extension of
, generates all regularly solvable extensions of
.
2) There are two intervals, say,
rand
each one has one regular and one singular end-point and each one of the others has two singular end-points. In this case
, and
generates all regularly solvable extensions of
. The other possibilities in the cases 1) and 2) are entirely similar.
Example 4: Five singular end-points and
. We consider two cases:
1) There are two intervals, say,
and
, such that
has two regular end-points and
has one regular and one singular end-points, and each one of the others has two singular end-points. In this case
and
, then,
, which is similar to 2) in Example 3.
2) There are three intervals, say,
and
each one has one regular and one singular end-point, and the fourth has two singular end-points. In this case
and
, then
, and hence
generates all regularly solvable extensions of
. The possibilities are entirely similar.
Example 5: Four singular end-points and
. We consider three cases:
1) There are two intervals, say,
and
, such that each one has two regular end-points and each one of the others has two singular end-points. In this case
and
, then,
.
2) There are two intervals, say,
and
, such that each has one regular and one singular end-point, and the others
and
has two regular and two singular end-points respectively. In this case
,
and
, then
as in Example 4 2).
3) Each interval has one regular and one singular end-points. In this case
. Then “mixing” can occur and we get new regularly solvable extensions of
. For the sake of definiteness assume that the end-points
and
are singular end-points and
and
are regular end-points. The other possibilities are entirely similar.
For
and
with
,
, condition (3.11)) reads:
(4.22)
Also, for
and
with
,
, condition (3.12)) reads:
(4.23)
and condition 2) in Theorem 3.5 reads:
(4.24)
By ( [3], Theorem III.10.13), the terms involving the singular end-points
and
are zero so that (4.22), (4.23) and (4.24) reduces to:
and:
. Thus, the boundary conditions are not separated for the four intervals and hence, the regularly solvable operators cannot be expressed as a direct sum of regularly solvable operators defined in the separate intervals
. We refer to Everitt and Zettl’s papers [5] [10] [12] [13] [19] [20] for more examples and more details.
Conclusion: We have characterized that all regularly solvable operators and their adjoints are generated by a general ordinary quasi-differential expression
in the direct sum Hilbert spaces
. The domains of these operators are described in terms of boundary conditions involving
-solutions of the equations
and its adjoint
on the intervals
. This characterization is an extension of those obtained in the case of one interval with one and two singular endpoints of the interval (a, b), and is a generalization of those proved in the case of self-adjoint and J-self-adjoint differential operators as a special case, where J denotes complex conjugation.
Acknowledgements
I am grateful to the Public Authority of Applied Education and Training (PAAET) in Kuwait for supporting the scientific researches and encouragement to the researchers.