You are on page 1of 16

J. Math. Anal. Appl.

414 (2014) 434–449

Contents lists available at ScienceDirect

Journal of Mathematical Analysis and Applications


www.elsevier.com/locate/jmaa

Symplectic geometry and dissipative differential operators


Siqin Yao a , Jiong Sun a , Anton Zettl b
a
Math. Dept., Inner Mongolia University, Hohhot, 010021, China
b
Math. Dept., Northern Illinois University, DeKalb, IL 60115, USA

a r t i c l e i n f o a b s t r a c t

Article history: Everitt and Markus characterized the domains of self-adjoint differential operators in
Received 29 January 2013 terms of Lagrangian subspaces of complex symplectic spaces. In this paper we define
Available online 9 January 2014 Dissipative and strictly Dissipative subspaces for complex symplectic spaces and
Submitted by D. O’Regan
characterize the domains of dissipative and strictly dissipative differential operators
Keywords: in terms of these subspaces.
Dissipative differential operators © 2014 Elsevier Inc. All rights reserved.
Symplectic geometry
Dissipative subspaces

1. New concepts for complex symplectic spaces

It is well known that complex symplectic spaces provide important algebraic structures clarifying the
theory of boundary value problems of linear ordinary differential equations and the theory of the associated
self-adjoint linear operators on Hilbert spaces [2,3,14]. In this paper we show that these spaces can also be
used to characterize dissipative and strictly dissipative operators in Hilbert space and in investigating their
properties.
In Section 2 we define Dissipative, strictly Dissipative, and Lagrangian subspaces of finite dimensional
complex symplectic spaces S and develop their algebraic properties. These definitions and results for Dis-
sipative and strictly Dissipative subspaces seem to be new. In Section 3 we ‘connect’ the algebraic theory
of finite dimensional complex symplectic spaces and their subspaces with the theory of boundary value
problems of very general even order symmetric ordinary differential equations. Section 4 contains our main
results giving a natural one-to-one correspondence between:

(1) The set of all Dissipative extensions of the minimal operator Mmin and the set of all Dissipative subspaces
in the complex symplectic space S̃ = Dmax /Dmin .
(2) The set of all symmetric extensions of Mmin and the set of all Lagrangian subspaces of S̃.
(3) The set of all strictly Dissipative operators generated by M and the set of all strictly Dissipative
subspaces in the complex symplectic space S̃.

E-mail addresses: 89608930@qq.com (S. Yao), masun@imu.edu.cn (J. Sun), zettl@math.niu.edu (A. Zettl).

0022-247X/$ – see front matter © 2014 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jmaa.2014.01.019
S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 435

1.1. Definitions on complex symplectic space

In this subsection, we recall some definitions on complex symplectic spaces. These are taken from the
Everitt–Markus monograph [2] (see also [3]).

Definition 1. A complex linear space S, together with a complex-valued function on the product space S ×S,

u, v → [u : v], S×S →C

is a pre-symplectic space in case:

(i) sesquilinear

[c1 u + c2 v : w] = c1 [u : w] + c2 [v : w], [u : c3 v + c4 w] = c̄3 [u : v] + c̄4 [u : w],

for all u, v, w ∈ S, and cj ∈ C, j = 1, . . . , 4; and


(ii) skew-Hermitian

(ii) [u : v] = −[v : u], for all u, v ∈ S.

If in addition to properties (i) and (ii) we further require

(iii) non-degenerate

[u : S] = 0 implies u = 0,

then S, together with the non-degenerate, skew-Hermitian, sesquilinear form [:], is a symplectic space.

Definition 2. A linear subspace L in the complex symplectic space S is called Lagrangian in case [L : L] = 0,
that is,

[u : v] = 0 for all vectors u, v ∈ L.

Furthermore, a Lagrangian manifold L ⊂ S is said to be complete in case

u∈S and [u : L] = 0 imply u ∈ L.

Definition 3. Let S be a complex symplectic space with symplectic form [:]. Then linear subspaces S− and
S+ are symplectic ortho-complements in S, written as

S = S− ⊕ S + ,

in case

(i) S = span{S− , S+ }, (ii) [S− : S+ ] = 0.

In this case S− ∩ S+ = 0, so S is the direct sum of S− and S+ .


436 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

1.2. Dissipative and Accretive subspaces

In this subsection, we define Dissipative, strictly Dissipative, maximal Dissipative; Accretive, strictly
Accretive and maximal Accretive subspaces of complex symplectic spaces. These definitions seem to be new
but are minor variants of known definitions; we state them here for the convenience of the reader and for
use in stating and proving our main results in Section 4.

Definition 4. A linear subspace D in the complex symplectic space S is called Dissipative in case

[u : u]  0 for all vectors u ∈ D.

A linear subspace A in the complex symplectic space S is called Accretive in case

[u : u]  0 for all vectors u ∈ A.

Definition 5. A Dissipative subspace D ⊂ S is said to be maximal Dissipative, if for any Dissipative subspace
D̄ such that D ⊆ D̄, we have D = D̄.
An Accretive subspace A ⊂ S is said to be maximal Accretive, if for any Accretive subspace Ā such that
A ⊆ Ā, we have A = Ā.

Definition 6. A Dissipative subspace D is called strictly Dissipative in case

[u : u] > 0, ∀u ∈ D, u
= 0.

An Accretive subspace A is called strictly Accretive in case

[v : v] < 0, ∀v ∈ A, v
= 0.

We denote a strictly Dissipative (Accretive) subspaces as Ds (As ).

Remark 1. (1) It is clear that a Lagrangian subspace is both Dissipative and Accretive.
(2) Linear subspaces of a complex symplectic space S need not be complex symplectic spaces, since the
induced symplectic form can be degenerate on them. For example Lagrangian subspaces and Dissipative but
not strictly Dissipative subspaces are pre-symplectic spaces but not symplectic spaces; strictly Dissipative
subspaces are symplectic spaces.
(3) For a symplectic orthogonal direct sum decomposition S = S− ⊕ S+ each of S− and S+ is itself a
complex symplectic space, since the symplectic form induced by [:] is non-degenerate on S− and S+ .

Example 1. As an interesting special example take the complex linear space S = C3 with the prescribed
symplectic products [e1 : e1 ] = i, [e2 : e2 ] = i, [a1 : a1 ] = −i, and all other symplectic products are zero for
the customary basis vectors

e1 = (1, 0, 0), e2 = (0, 1, 0), a1 = (0, 0, 1).

That is, we use the skew-Hermitian matrix H = diag{i, i, −i} to define the symplectic structure on C3 .
Define L = span{e2 + a1 } = {(0, c, c)}, for all c ∈ C, then L is a Lagrangian subspace, but not a complete
Lagrangian subspace, since [e2 : L] = 0 yet e2 ∈/ L.
S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 437

Define D = span{e1 , e2 + a1 } = span{(α, β, β)}, for all α, β ∈ C, then D is a Dissipative subspace since

 
 αe1 + β(e2 + a1 ) : αe1 + β(e2 + a1 ) = [αe1 : αe1 ]  0.

Furthermore D is a maximal Dissipative subspace.


Assume there is a Dissipative subspace

D∗ = D  span{ė}, ė ∈ S, ė ∈
/ D.

It is obvious that e1 , e2 + a1 , ė ∈ S and e1 , e2 + a1 and ė are linearly independent. Hence S = D∗ . Note


that S is not a Dissipative subspace, since [a1 : a1 ] < 0. This contradiction proves that D is maximal.

2. Finite dimensional complex symplectic spaces and their dissipative subspaces

In this section we develop the algebraic properties of the Dissipative and strictly Dissipative subspaces
of complex symplectic spaces which will be applied to differential operators in Hilbert space in Section 3.

Definition 7. In a complex symplectic space S, with symplectic form [:], and finite dimension D  1, define
the following symplectic invariants of S:

 
p = max dimension of complex linear subspace whereon [v : v]  0 ,
 
q = max dimension of complex linear subspace whereon [v : v]  0 ,

(p, q) is called the signature of S, consisting of a pair of integers: the positivity index p  0 and the negativity
index q  0.
In addition, we define the Lagrangian index and the excess Ex of S

= max{dimension of complex Lagrangian subspace of S},


Ex = p − q, excess of positivity over negativity index of S.

Lemma 1. Consider a complex symplectic space S, with symplectic form [:] and finite dimension D  1. Let
D ⊆ S be a Dissipative subspace. Then

(1) dim D  p.
(2) There are non-trivial dissipative subspaces if and only if p
= 0.

0, a Dissipative subspace D is maximal if and only if dim D = p.
(3) If p =

Proof. This follows immediately from Definition 4 and Definition 7. 2

Theorem 1. Consider a complex symplectic space S with symplectic form [:] and finite dimension D  1.
Choose any basis of S, with corresponding coordinates and skew-Hermitian nonsingular matrix H determined
by [:], and note that H is congruent to a diagonal matrix of the form diag{i, i, . . . , i, −i, −i, . . . , −i}. Then
conclude that the symplectic invariants of S are related to H by:

p = number of (+i) terms on the diagonal,


q = number of (−i) terms on the diagonal,
438 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

with 0  p, q  D, so the diagonal form for H is unique. Also we have

1  1
D = p + q, Ex = p − q, = min{p, q} = D − |Ex |  D.
2 2

Proof. See [2,3] for the proof. 2

Let {e1 , e2 , . . . , ep , a1 , a2 , . . . , aq } denote the basis of S which generates the symplectic form

iIp 0
H= .
0 −iIq

Theorem 2. Consider a complex symplectic space S with symplectic form [:] and finite dimension D  1,
and let p, q denote the positivity and the negativity indices of S. Let

S+ = span{e1 , . . . , ep },
S− = span{a1 , . . . , aq }.

Then

(1) S = S+ ⊕ S− , [S+ : S− ] = 0.
(2) S+ is a maximal Dissipative subspace, also a strictly Dissipative subspace.
(3) S− is a maximal Accretive subspace, also a strictly Accretive subspace.

Proof. It is obvious that S = span{S+ , S− } and [S+ : S− ] = 0. So item (1) is true.



p
For every f = 1 αj ej ∈ S+ we have
 
[f : f ] = i |α1 |2 + · · · + |αp |2

So S+ is a Dissipative subspace, in fact a strictly Dissipative subspace.


Similarly, S− is an Accretive subspace, in fact a strictly Accretive subspace.
Now prove S+ is a maximal Dissipative subspace. Assume that there exists a Dissipative subspace

S ∗ = S+  span{u̇}, u̇ ∈ S, u̇ ∈
/ S+ ,

then u̇ = u̇+ + u̇− , u̇+ ∈ S+ , u̇− ∈ S− and u̇−


= 0.
Note that u̇− ∈ S ∗ but

[u̇− : u̇− ] < 0.

This contradicts the assumption that S ∗ is a Dissipative subspace. Therefore S+ is a maximal Dissipative
subspace.
Similarly S− is a maximal Accretive subspace. 2

Definition 8. Consider a complex symplectic space S, with symplectic form [:] and finite dimension D  1.
Let D ⊆ S be a Dissipative subspace.

(1) We call an element u ∈ D a Lagrangian element if [u : u] = 0.


(2) We call an element u ∈ D a Dissipative element if [u : u] > 0.
S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 439

Theorem 3. Consider a complex symplectic space S with symplectic form [:] and finite dimension D  1.
Let D ⊆ S be a Dissipative subspace with dim D = r  p. Then the set
 
DL = f : f ∈ D, [u : u] = 0 ⊆ D

of all the Lagrangian elements of D is a Lagrangian subspace of S.

Proof. For every f, g ∈ DL , α, β ∈ C, αf + βg ∈ D, so [αf + βg : αf + βg]  0. Since

[αf + βg : αf + βg] = |α|2 [f : f ] + |β|2 [g : g] + 2αβ̄[f : g] = 2αβ̄[f : g]

we have

αβ̄[f : g]  0.

Since α, β ∈ C are arbitrary, we obtain [f : g] = 0, then [αf + βg : αf + βg] = 0. Hence DL is a Lagrangian


subspace of S. 2

Theorem 4. Let the notations and hypotheses of Theorem 3 hold. Let dim DL = rL , and let {a1 , . . . , arL } be
a basis for DL . Then there is a basis for D

{e1 , . . . , ers , a1 , . . . , arL }, rs + rL = r,

such that

(1) Ds = span{e1 , . . . , ers } is a strictly Dissipative subspace, and

[Ds : DL ] = 0.

(2) The matrix of the symplectic form for D induced from S becomes

iIrs 0
HD = ,
0 0rL ×rL

here HD is a singular skew-Hermitian matrix.

Proof. Assume {e1 , . . . , ers , a1 , . . . , arL }, rs + rL = r is a basis for D, since ej ∈


/ DL , [ej : ej ] > 0,
j = 1, . . . , rs . Without loss of generality, choose the basis satisfying

[ej : ej ] = i, [ej : ek ] = 0, j
= k, j, k = 1, . . . , rs .

rs
(1) For every φ = 1 βj ej ∈ Ds we have
 
[φ : φ] = i |β1 |2 + · · · + |βrs |2 , [φ : φ]  0, [φ : φ] = 0 ⇔ φ = 0.

So Ds = span{e1 , . . . , ers } is a strictly Dissipative subspace.


For every f ∈ DL , φ ∈ Ds , and α, β ∈ C, since αf + βφ ∈ D, so

[αf + βφ : αf + βφ]  0.
440 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

From

[αf + βφ : αf + βφ] = |α|2 [f : f ] + |β|2 [φ : φ] + 2αβ̄[f : φ]


= |β|2 [φ : φ] + 2αβ̄[f : φ]

and since α and β are arbitrary we obtain that [f : φ] = 0, i.e. [DL : Ds ] = 0.


Item (2) follows immediately from item (1). 2

Example 2. Consider the complex symplectic space S = span{e1 , e2 , e3 , a1 , a2 , a3 } with customary basis
vectors

[ej : ej ] = i, [aj : aj ] = −i, j = 1, 2, 3,

and other symplectic products are zero. That is, we use the skew-Hermitian matrix H = diag{i, i, i,
−i, −i, −i} to define the symplectic structure on S = C6 .
Define D4 = span{ω1 , ω2 , ω3 }, ω1 = e1 + a1 , ω2 = e2 + a1 , ω3 = e3 .
It is obvious that [ω1 : ω1 ] = 0, [ω2 : ω2 ] = 0, and [ω3 : ω3 ] > 0. But D4 is not a Dissipative subspace,
since ω1 + ω2 ∈ D4 , and

[ω1 + ω2 : ω1 + ω2 ] = (1, 1, 0, 2, 0, 0)H(1, 1, 0, 2, 0, 0)∗ = −2i.

The internal cause of [ω1 + ω2 : ω1 + ω2 ] < 0 is

[ω1 : ω2 ] = −i, [ω1 + ω2 : ω1 + ω2 ] = 2[ω1 : ω2 ].

I.e. the elements ω1 , ω2 don’t meet the internal relations between Lagrangian elements of a Dissipative
subspace.
Define D5 = span{ϑ1 , ϑ2 , ϑ3 }, ϑ1 = e1 + a1 , ϑ2 = e2 + a2 , ϑ3 = e2 . We have [ϑ1 : ϑ1 ] = 0, [ϑ2 : ϑ2 ] = 0,
and [ϑ3 : ϑ3 ] > 0. But D5 is not a Dissipative subspace. Since ϑ2 − ϑ3 ∈ D5 ,

[ϑ2 − ϑ3 : ϑ2 − ϑ3 ] = −i.

The internal cause is that ϑ2 , ϑ3 don’t meet the internal relations between Lagrangian elements and strictly
Dissipative elements of a Dissipative subspace.

Theorem 5. Consider a complex symplectic space S, with symplectic form [:], and finite dimension D  1.
Then D ⊆ S is a Dissipative subspace if and only if there exist a Lagrangian subspace DL ⊆ D and a strictly
Dissipative subspace Ds ⊆ D such that

D = DL  D s , (2.1)

where ‘’ means DL is symplectic orthogonal to Ds , i.e. [DL : Ds ] = 0.

Proof. The necessity is obtained directly from Theorem 4. To prove the sufficiency, take any u ∈ D.
If u ∈ DL , [u : u] = 0, i.e. [u : u] = 0.
If u ∈ Ds , [u : u] > 0.
If u ∈
/ DL , u ∈
/ Ds then u = f + g, f ∈ DL , g ∈ Ds , f
= 0, g
= 0, then

[u : u] = [f + g : f + g] = [g : g] > 0.

This completes the proof. 2


S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 441

Example 3. This example illustrates that there can be more then one maximal Dissipative subspace.
Consider the complex symplectic space S = C2 with customary basis vectors e1 = (1, 0), a1 = (0, 1)
satisfying [e1 : e1 ] = i, [a1 : a1 ] = −i, and other symplectic products are zero. That is, we use the
skew-Hermitian matrix H = diag{i, −i} to define the symplectic structure on S = C2 .
(1) Define D1 = span{e1 }. Then D1 is a maximal Dissipative subspace and a strictly Dissipative subspace.
(2) Define D2 = span{2e1 +a1 }. Then D2 is also a maximal Dissipative subspace and a strictly Dissipative
subspace.

Corollary 1. Consider a complex symplectic space S, with symplectic form [:] and finite dimension D  1.
Then

(1) if p
= 0, q = 0, there is only one maximal Dissipative subspace,
(2) if p
= 0, q =

0 there are more than one maximal Dissipative subspaces.

3. Applications of symplectic geometry to ordinary differential operators

In this section we state and prove our main results by showing that there is a natural one-to-one cor-
respondence between the set of all (strictly) Dissipative extensions of the minimal operator Mmin and the
set of all (strictly) Dissipative subspaces in the complex symplectic space S̃ = Dmax /Dmin . Our approach
is based on the GKN Theorem, see [2,14]. For another approach based on boundary triplets, see [5,6].

3.1. Symmetric expressions and the maximal domain

Let J = (a, b) be an interval with −∞  a < b  ∞ and let n = 2k be a positive even integer. We
consider differential and quasi-differential equations

M y = (−1)k y [n] = λwy, w ∈ Lloc (J), w > 0, a.e. on J, (3.1)

and study operators generated by (3.1) in the Hilbert space H = L2 (J, w). Here M is a general even order
symmetric quasi-differential expression of order n with real coefficients. See [7–9,15] for a detailed definition
of M . These symmetric expressions M include those studied by Naimark in his well-known book [11]. For
n = 2 and n = 4 they have the representations:
 
M y = − py  + qy,
   
M y = py  + ry  + qy.

Note the ‘extra’ parentheses i.e. the bracket [·,·] in the n = 4 case. This is due to the weak Lloc integrability
conditions on the coefficients. The symmetric quasi-differential expressions M defined in the above cited
references to which our results apply are much more general than those studied by Naimark in [11] and
assume no smoothness conditions on the coefficients. In particular the coefficients need not be continuous.
The maximal and minimal operators associated with a symmetric expression M and a positive weight
function w in the Hilbert space H are defined as follows:
 
Dmax = y ∈ L2 (J, w): y [r] ∈ ACloc (J), w−1 M y ∈ L2 (J, w), r = 1, . . . , n − 1 ,
Mmax y = w−1 M y, y ∈ Dmax ,

Mmin = Mmax ,
Dmin = D(Mmin ).
442 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

Lemma 2. Let Mmin and Mmax be defined as above. Then Dmin and Dmax are dense in H, Mmin and Mmax
∗ ∗
are closed operators in H, Mmin = Mmax , Mmin = Mmax and Mmin is a symmetric operator in H.

Proof. This is well known, see [4] or [10]. 2

Lemma 3. For any y, z in Dmax we have

b
{z̄M y − yM z} = [y, z](b) − [y, z](a).
a

Proof. See [4] or [10]. 2

Now we denote the deficiency indices of the minimal operator Mmin by (d+ , d− ). Namely,

d− = dim N (Mmax + i) = dim R(Mmin − i)⊥ ,


d+ = dim N (Mmax − i) = dim R(Mmin + i)⊥ .

Then we have the well-known von Neumann formula [11–13] for the decomposition of the maximal domain
Dmax :

Dmax = Dmin  N (Mmax + i)  N (Mmax − i). (3.2)

Here N (S) denotes the null space of the operator S.

Theorem 6. Let (d− , d+ ) denote the deficiency indices of M . Consider the equation

M y = λwy, (3.3)

then:

(1) There exist d+ linearly independent solutions on (a, b) which lie in N (Mmax − i) satisfying

1
ui , uj  = δij , i, j = 1, . . . , d+ .
2

(2) There exist d− linearly independent solutions v1 , v2 , . . . , vd− on (a, b) which lie in N (Mmax +i) satisfying

1
vi , vj  = δij , i, j = 1, . . . , d− .
2

(3)

ui , vj  = 0, i = 1, . . . , d+ , j = 1, . . . , d− .

Proof. Since the space N (Mmax + i)  N (Mmax − i) is a (d− + d+ )-dimensional subspace of the Hilbert
space H, there exist orthogonal bases

u̇1 , u̇2 , . . . , u̇d+ , v̇1 , v̇2 , . . . , v̇d−

satisfying u̇1 , u̇2 , . . . , u̇d+ lie in N (Mmax − i), and v̇1 , v̇2 , . . . , v̇d− lie in N (Mmax + i). Let
S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 443

1 1
ui = √ u̇i , vj = √ v̇j , i = 1, 2, . . . , d+ , j = 1, 2, . . . , d− ,
2 2

then it follows that the bases u1 , u2 , . . . , ud+ , v1 , v2 , . . . , vd− satisfy the properties (1), (2), (3). 2

Theorem 7. Let the notations and hypotheses of Theorem 6 hold. Then

Dmax = Dmin  {u1 , u2 , . . . , ud+ }  {v1 , v2 , . . . , vd− }. (3.4)

Proof. This follows immediately from the von Neumann formula [11–13] and Theorem 6. 2

3.2. Dissipative and accretive extensions of symmetric operators

Definition 9. A linear operator T on a Hilbert space H with dense domain D(T ) is called symmetric if
T ⊂ T ∗ , i.e. if for every f, g ∈ D(T ),

(T f, g) = (f, T g).

Definition 10. (See [1,5].) A linear operator TD on H with dense domain D(TD ) is called dissipative if

(TD f, f )  0 for all f ∈ D(TD ).

A linear operator TA on H with dense domain D(TA ) is called accretive if

(TA f, f )  0 for all f ∈ D(TA ).

Definition 11. (See [5].) A dissipative (accretive) operator TD (TA ) is called maximal dissipative (maximal
accretive) if it has no non-trivial, that is, different from TD (TA ) itself, dissipative (accretive) extensions.

Next we define strictly dissipative (strictly accretive) extensions of symmetric operators.

Definition 12. A dissipative extension TD of a symmetric operator T is called strictly dissipative if

(TD f, f ) > 0 for every f ∈ D(TD ) \ D(T ).

An accretive extension TA of a symmetric operator T is called strictly accretive if

(TA f, f ) < 0 for every f ∈ D(TA ) \ D(T ).

We denote the strictly dissipative (accretive) extension as TsD (TsA ).

Remark 2. We note that:

(1) Since a linear operator T is accretive if and only if −T is dissipative, all results concerning dissipative
operators can be immediately transferred to accretive operators.
(2) A symmetric operator is both dissipative and accretive.

Lemma 4. Let the notations and hypotheses of Theorem 6 hold. Then there are dissipative extensions of
Mmin if and only if d+
= 0.
444 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

Proof. Let r  d+ , and define

T = M |D , D1 = Dmin ⊕ span{u1 , . . . , ur }. (3.5)

Then T is a dissipative extension of Mmin . In fact, for every f + u ∈ D, f ∈ Dmin , u ∈ span{u1 , . . . , ur },


 
T (f + u), f + u = (M f, f ) + (iu, u) + (M f, u) + (M u, f )
= (M f, f ) + iu2 + (f, M u) + (M u, f )
= (M f, f ) + iu2 − i(f, u) + i(u, f ).

Since −i(f, u) + i(u, f ) is real, (T (f + u), f + u)  0. So T is a dissipative extension of Mmin . 2

3.3. Complex symplectic spaces S̃ related to M

Next we define and investigate the structures of a complex symplectic space S̃, with its Lagrangian
subspaces L̃, Dissipative subspaces D̃ and Accretive subspaces Ã, which arise in connection with boundary
value problems associated with an even order symmetric differential expression M with real coefficients and
Eq. (3.1).
Define the endpoint space S̃ by

S̃ = Dmax /Dmin , (3.6)

and note that it is a complex vector space of dimension (d+ + d− ). Further denote the natural projection Ψ
of Dmax onto S̃ by

Ψ : Dmax → S̃, f → Ψ f = {f + Dmin }, (3.7)

and let

f˜ = Ψ f, so f˜ ∈ S̃. (3.8)

The pre-image of a subset Ũ ∈ S̃ is denoted by Ψ −1 Ũ ⊆ Dmax .


Let D = d+ + d− , then the endpoint space S̃ becomes a complex symplectic D-space, with the symplectic
product [:] inherited from Dmax as follows

[f˜ : g̃] = [f + Dmin : g + Dmin ] := [f : g],

where the skew-Hermitian form [f : g] is defined for f, g ∈ Dmax by



[f : g] = M (f ), g − f, M (g) = [f, g]ba .

Hence
 
Dmin = f ∈ Dmax : [f : Dmax ] = 0 .

Lemma 5. Let the notation and hypotheses of Theorem 6 hold, then


   
[ui : uj ] d+ ×d+ = iId+ , [vi : vj ] d− ×d− = −iId− . (3.9)

Here Id+ , Id− are unit matrices.


S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 445

Proof. From Lemma 3,

b b
 
[ui : uj ] = ūj M (ui ) − ui M uj = {ūj iwui − ui iwuj } = 2iui , uj ,
a a

b b
   
[vi : vj ] = v̄j M (vi ) − vi M vj = v̄j (−iwvi ) − vi (−iwvj ) = −2ivi , vj ,
a a

then from Theorem 6 we have


 
[ui : uj ] d+ ×d+ = iId+ ,
 
[vi : vj ] d− ×d− = −iId− . 2

Lemma 6. The symplectic invariants of a complex vector space S̃ with the skew-Hermitian form [:] defined
above satisfy

p = d+ , q = d− , dim S̃ = d+ + d− , and Ex = d+ − d− .

Proof. From Proposition 1 in [2], we know that the symplectic invariants of a complex vector space S̃ are
related to the deficiency indices d± of a symmetric differential expression M by p = d+ , q = d− . 2

Theorem 8. Let the notation and hypotheses of Theorem 6 hold, and let S̃ = Dmax /Dmin be the complex
vector space defined above. Then

(1) S̃ = span{ũ1 , . . . , ũd+ , ṽ1 , . . . , ṽd− }.


+ −
(2) S̃ is symplectic isomorphic to a complex symplectic space Cd +d .
(3) For the basis {ũ1 , . . . , ũd+ , ṽ1 , . . . , ṽd− }, the associated skew-Hermitian matrix is as follows

iId+ 0
H= .
0 −iId−

Proof. By von Neumann’s formula (3.2) and Theorem 7, we obtain (1); from Example 1 in [3], item (2)
follows. Note that ∀uj ∈ N (Mmax − i), vk ∈ N (Mmax + i),

b b
[uj : vk ] = v̄k M uj − uj M vk = v̄k (iuj ) − uj (−ivk ) = 0.
a a

Connecting with Lemma 5, the associated skew-Hermitian matrix is



iId+ 0
H= . 2
0 −iId−

3.4. A direct sum representation of S̃

In this subsection we give a symplectic orthogonal direct sum decomposition of the complex symplectic
space S̃ in terms of Dissipative and Accretive subspaces.
446 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

Theorem 9. Let S̃ = Dmax /Dmin be the complex vector space associated with M . Let

S̃+ = span{ũ1 , ũ2 , . . . , ũd+ },


S̃− = span{ṽ1 , ṽ2 , . . . , ṽd− }.

Then

(1) S̃ = S̃+ ⊕ S̃− , [S̃+ : S̃− ] = 0.


(2) S̃+ is a maximal Dissipative subspace, also a strictly Dissipative subspace.
(3) S̃− is a maximal Accretive subspace, also a strictly Accretive subspace.

Proof. It is obvious that S̃ = span{S̃+ , S̃− } and [S̃+ : S̃− ] = 0. So item (1) is true. From Lemma 5

[ũi : ũj ] = [ui : uj ] = 2iui , uj  = iδij , i, j = 1, . . . , p,

i.e.,

[ũi : ũj ]  0, [ũi : ũi ] > 0.

So from Definition 4 and Definition 6, S̃+ is a Dissipative subspace, in fact a strictly Dissipative subspace.
Similarly, S̃− is an Accretive subspace, in fact a strictly Accretive subspace.
Now prove S̃+ is a maximal Dissipative subspace. Assume that there exists a Dissipative subspace

S̃ ∗ = S̃+  span{u̇}, u̇ ∈ S̃, u̇ ∈


/ S̃+ ,

then u̇ = u̇+ + u̇− , u̇+ ∈ S̃+ , u̇− ∈ S̃− and u̇−


= 0.
Note that u̇− ∈ S̃ ∗ but

[u̇− : u̇− ] < 0.

This contradicts the assumption that S̃ ∗ is a Dissipative subspace. Therefor, S̃+ is a maximal Dissipative
subspace.
Similarly, S̃+ is a maximal Accretive subspace. 2

4. (Strictly) dissipative and symmetric extensions of the minimal operator

Consider S̃ = Dmax /Dmin . In this subsection we establish:

(1) A one-to-one correspondence between the set {TD } of all Dissipative extensions of Mmin and the set
{D̃} of all Dissipative subspaces in the complex symplectic space S̃.
(2) A one-to-one correspondence between the set {TL } of all symmetric extensions of Mmin and the set {L}
of all Lagrangian subspaces of S̃.
(3) A one-to-one correspondence between the set {TsD } of all strictly Dissipative operators generated by
M and the set {D̃s } of all strictly Dissipative subspaces in the complex symplectic space S̃.

Lemma 7. Let M be a symmetric differential expression studied in Eq. (3.1) and let the notation and
hypotheses of Theorems 6 and 8 hold. Then
S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 447

(1) Mmin has symmetric extensions if and only if S̃ has Lagrangian subspaces.
(2) There is a one-to-one correspondence between the set {TL } of all symmetric extensions of Mmin and
the set {L} of all Lagrangian subspaces.

Theorem 10. Let M be a symmetric differential expression studied in Eq. (3.1) and let the notation and
hypotheses of Theorems 6 and 8 hold. Then there exists a natural one-to-one correspondence between the
set {TsD } of all strictly dissipative operators generated by M and the set {D̃s } of all strictly Dissipative
subspaces in the complex symplectic space S̃ = Dmax /Dmin .

Proof. Take the correspondence TsD ↔ D̃s as given by the injective surjection

Ψ : {D̃s } → {TsD }, (4.1)

which is defined in terms of the natural projection

Ψ̃ : Dmax → S̃, (4.2)

according to

Ψ̃ D(TsD ) = D̃s , and D(TsD ) = Ψ̃ −1 D̃s . (4.3)

Take any strictly Dissipative r-space Ds ∈ {D̃s }, and select a basis

ũj1 , ũj2 , . . . , ũjr

for Ds , where ũjs = Ψ̃ ujs for some functions ujs ∈ Dmax , s = 1, 2, . . . , r. Define a linear manifold Ds ⊆
L2 (J, w)

Ds :≡ Dmin  span{uj1 , uj2 , . . . , ujr }, (4.4)

then Ds is a domain of the dissipative operator TsD = Mmax |Ds .


In fact, for every

ϕ + u ∈ Dmin  span{uj1 , uj2 , . . . , ujr },

where ϕ ∈ Dmin , u ∈ span{uj1 , uj2 , . . . , ujr } we have


 
TsD (ϕ + u), ϕ + u = (TsD ϕ + TsD u, ϕ + u)
= (TsD ϕ, ϕ) + (TsD u, u) + (TsD ϕ, u) + (TsD u, ϕ)
= (Mmin ϕ, ϕ) + (Mmax u, u) + (Mmin ϕ, u) + (Mmax u, ϕ).

Note that

(Mmin ϕ, ϕ) = 0, (Mmax u, u) > 0,

and from

(Mmin ϕ, u) = (ϕ, Mmax u) = (Mmax u, ϕ),


448 S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449

we get
 
 (Mmin ϕ, u) + (Mmax u, ϕ) = 0.

So
 
 TsD (ϕ + u), ϕ + u > 0.

Hence TsD = Mmax |Ds is a dissipative extension, further, a strictly dissipative extension of the minimal
operator Mmin .
Now, let us demonstrate that the map Ψ of (4.1) is injective. For this purpose consider two different
strictly Dissipative r-spaces D̃s1 and D̃s2 in S̃, say with a vector ũ0 ∈ D̃s1 but ũ0 ∈
/ D̃s2 . Then there is a
representation function u0 ∈ Ds1 such that Ψ u0 = ũ0 and hence u0 ∈ Ψ −1 D̃s1 = Ds1 . Yet u0 ∈ / Ψ −1 D̃s2 so
u0 ∈/ D̃s2 . Therefore Ds1
= Ds2 , so the map Ψ of (4.1) is injective.
On the other hand, from
 
[f˜ : f˜] =  (M f, f ) − (f, M f ) = 2(M f, f ), (4.5)

the map Ψ of (4.1) is surjective.


Thus Ψ of (4.1) defines a one-to-one correspondence between {TsD } and {D̃s }. 2

Theorem 11. Let M be a symmetric differential expression studied in Eq. (3.1) and let the notation and
hypotheses of Theorems 6 and 8 hold. Then there exists a natural one-to-one correspondence between the
set {TD } of all dissipative operators TD generated by M and the set {D̃} of all Dissipative subspaces D̃ in
the complex symplectic space S̃ = Dmax /Dmin .

Proof. Take the correspondence TD ↔ D̃ as defined in (4.1)

Ψ : {D̃} → {TD }.

Take any Dissipative but not strictly Dissipative subspaces D̃ ∈ {D̃}. Then from Theorem 5, there exists
D̃L
= ∅ such that

D̃ = D̃L  D̃s . (4.6)

Assume

rL = dim D̃L , rs = dim D̃s ,

and

D̃L = span{w̃1 , w̃2 , . . . w̃rL }, D̃s = span{χ̃1 , χ̃2 , . . . , χ̃rs }.

Let

DL = Dmin  span{w1 , w2 , . . . , wrL }

then TL = Mmax |DL is a symmetric extension of Mmin . Let

D = DL  span{χ1 , χ2 , . . . , χrs }

then TD = Mmax |DL is a strictly dissipative extension of the symmetric operator TL .


S. Yao et al. / J. Math. Anal. Appl. 414 (2014) 434–449 449

From Lemma 7 and Theorem 10 it follows that there is a one-to-one correspondence between {TD } and
{D̃}. 2

Acknowledgments

The authors thank the anonymous referee whose suggestions improved the presentation of this paper.
The work of the first and second authors is supported by the National Nature Science Foundation of
China (Grant No. 11161030), and the first author is supported by the Program of Higher Level Talents
of Inner Mongolia University (SPH-IMU). The third author was supported by the Ky and Yu-fen Fan
US–China Exchange Fund through the American Mathematical Society.

References

[1] B.P. Allahverdiev, Dissipative Sturm–Liouville operators in limit-point case, Acta Appl. Math. 86 (2005) 237–248.
[2] W.N. Everitt, L. Markus, Boundary Value Problems and Symplectic Algebra for Ordinary Differential and Quasi-
differential Operators, Math. Surveys Monogr., vol. 61, American Mathematical Society, 1999.
[3] W.N. Everitt, L. Markus, Complex symplectic geometry with applications to ordinary differential operators, Trans. Amer.
Math. Soc. 351 (12) (1999) 4905–4945.
[4] W.N. Everitt, A. Zettl, Generalized symmetric ordinary differential expressions I: The general theory, Nieuw Arch. Wiskd.
(3) 27 (1979) 363–397.
[5] V.I. Gorbachuk, M.L. Gorbachuk, Boundary Value Problems for Operator Differential Equations, Kluwer Academic Pub-
lishers, Boston, London, 1990.
[6] A. Gorianov, V. Mikhailets, K. Pankrashkin, Formally self-adjoint quasi-differential operators and boundary value prob-
lems, arXiv:1205.1810v3.
[7] X. Hao, J. Sun, A. Wang, A. Zettl, Characterization of domains of self-adjoint ordinary differential operators II, Results
Math. 61 (2012) 255–281.
[8] X. Hao, J. Sun, A. Zettl, Real-parameter square-integrable solutions and the spectrum of differential operators, J. Math.
Anal. Appl. 376 (2011) 696–712.
[9] X. Hao, J. Sun, A. Zettl, The spectrum of differential operators and square-integrable solutions, J. Funct. Anal. 162 (2012)
1630–1644.
[10] M. Möller, A. Zettl, Semi-boundedness of ordinary differential operators, J. Differential Equations 115 (1995) 24–49.
[11] M.A. Naimark, Linear Differential Operators, Ungar, New York, 1968 (English transl.).
[12] J. Weidmann, Linear Operators in Hilbert Spaces, Grad. Texts in Math., Springer-Verlag, Berlin, 1980.
[13] J. Weidmann, Spectral Theory of Ordinary Differential Operators, Lecture Notes in Math., vol. 1258, Springer-Verlag,
Berlin, 1987.
[14] S. Yao, J. Sun, A. Zettl, Self-adjoint domains, symplectic geometry, and limit-circle solutions, J. Math. Anal. Appl. 397
(2013) 644–657.
[15] A. Zettl, Formally self-adjoint quasi-differential operators, Rocky Mountain J. Math. 5 (1975) 453–474.

You might also like