1403 7732
1403 7732
1403 7732
1. introduction
Block operator matrices arise in various areas of mathematical physics such as
ordinary differential equations [22, 23, 26], theory of elasticity [15, 33, 34], hydrody-
namics [10, 11], magnetohydrodynamics [14], quantum mechanics [28], and optimal
control [13, 32]. The spectral properties of block operator matrices are of vital im-
portance as they govern for instance the solvability and stability of the underlying
physical systems. As a basis for spectral analysis, the multiplication or adjoint op-
eration and self-adjointness of block operator matrices with unbounded entries have
attracted considerable attention and have been investigated case by case, see, e.g,
[3, 4, 7, 18, 19] for the former and [5, 6, 12, 17, 25] for the latter, or the monograph
[29] for both of these topics. As has been pointed out in [18], what is essentially
trivial for bounded operators appears to become erratic for unbounded operators.
The purpose of this paper is to build a common framework for these problems.
To this end, let us first recall some notions on block operator matrices. Through-
out this paper, we will denote by X1 , X2 complex Banach spaces, X1∗ , X2∗ the adjoint
spaces (see [9, Section III.1.4]), and X := X1 × X2 the product space equipped with
the norm
1
k(x1 x2 )t k := (kx1 k2 + kx2 k2 ) 2 .
such that if the element f of (X1 × X2 )∗ is identified with the element (f1 f2 )t of
X1∗ × X2∗ , then
(f, x) = (f1 , x1 ) + (f2 , x2 )
(b) Suppose that A is closed with ρ(A) 6= ∅, and that D(A) ⊂ D(C). Then for
some (and hence for all) λ ∈ ρ(A),
I 0 A−λ 0
A−λ=
C(A − λ)−1 I 0 S2 (λ)
−1
I (A − λ) B
,
0 I
where S2 (λ) := D − λ − C(A − λ)−1 B is the second Schur complement of
A with domain D(S2 (λ)) = D(B) ∩ D(D).
Proof. To prove the first equality, we denote by RST the product of the three
linear operators on the right side. It is easy to see that
D(RST ) = D(ST ) = (D(A) ∩ D(C)) × D(D).
Hence, we have, by Theorem 2.1,
RST = R × (S × T ),
this proved the first equality. Similarly, the second equality holds.
Definition 2.2. Let
A B
A=
C D
be a block operator matrix on X with dense domain D1 ×D2 . Then the block operator
matrix
(A|D1 )∗ (C|D1 )∗
A× :=
(B|D2 )∗ (D|D2 )∗
is said to be the formal adjoint block operator matrix of A.
Theorem 2.2. Let A be a block operator matrix on X with dense domain. Then
A× = A∗ if and only if A∗ has a matrix representation.
Proof. We only need to prove the “if” part. Assume A∗ has a matrix representation.
Writing
A B
A=
C D
and D(A) = D1 ×D2 . Then it follows from (f, Ax) = (A× f, x) for all x ∈ D(A) and
all f ∈ D(A× ) that A× ⊂ A∗ (see also [18]). It remains to show D(A∗ ) ⊂ D(A× ).
Let
g1
g= ∈ D(A∗ ).
g2
By Lemma 2.1 we have Qk g ∈ D(A∗ ) for k = 1, 2, so that
(2.2) (Qk g, Ay) = (A∗ Qk g, y) for y ∈ D(A), k = 1, 2.
Writing
hk1
A∗ Qk g = , k = 1, 2.
hk2
By taking k = 1 in (2.2) we get
y1
(g1 , A|D1 y1 ) + (g1 , B|D1 y2 ) = (h11 , y1 ) + (h12 , y2 ), ∈ D(A).
y2
BLOCK OPERATOR MATRICES 5
It follows that
(g1 , A|D1 y1 ) = (h11 , y1 ) for y1 ∈ D(A|D1 ),
(g1 , B|D1 y2 ) = (h12 , y2 ) for y2 ∈ D(B|D1 ),
so that
(2.3) g1 ∈ D((A|D1 )∗ ) ∩ D((B|D1 )∗ ).
Similarly, by taking k = 2 in (2.2) we get
(2.4) g2 ∈ D((C|D1 )∗ ) ∩ D((D|D1 )∗ ).
From (2.3) and (2.4) we get g ∈ D(A× ), so that D(A∗ ) ⊂ D(A× ).
∗
A C∗
Obviously, ⊂ A× ⊂ A∗ . Moreover, there is a block operator
B ∗ D∗ ∗
A C∗
matrix A such that A× = A∗ but 6= A∗ , see the following example
B ∗ D∗
(for the notions of differential operators see [21]).
Example 2.1. Given the following four differential operators on the Hilbert space
L2 (0, 1):
D(L) := H 2 (0, 1), Lf := −f ′′ ,
D(L0 ) := {f ∈ D(L) | f (k) (0) = f (k) (1) = 0, k = 0, 1}, L0 := L|D(L0 ) ,
D(LD ) := {f ∈ D(L) | f (0) = f (1) = 0}, LD := L|D(LD ) ,
D(LN ) := {f ∈ D(L) | f ′ (0) = f ′ (1) = 0}, LN := L|D(LN ) .
It is well known that L0 is closed, L = L∗0 , LD = L∗D , and LN = L∗N . For the
block operator matrix
LD LN
L :=
LN −LD
defined on the Hilbert space L2 (0, 1) × L2 (0, 1), we have
(a) L is
closed,
L∗D L∗N
(b) L = ( L× = L∗ .
L∗N −L∗D
In fact, we see from D(L0 ) = D(LD ) ∩ D(LN ) that
L0 L0 I I L0 0
L= = ,
L0 −L0 I −I 0 L0
where the latter equality follows from Theorem 2.1. By Lemma A.1,
L 0 I I
(2.5) L∗ = ,
0 L I −I
which implies D(L∗ ) = D(L) × D(L). Thus, by Theorem 2.1,
∗ L L
L = = L× .
L −L
But
L∗D L∗N
= L ( L∗ .
L∗N −L∗D
6 JIN AND ALATANCANG
I I √1
It remains to prove L is closed. Since is a self-adjoint unitary
I −I 2
operator, we have, by applying Lemma A.2 to (2.5),
∗
∗∗ I I L 0
L = = L.
I −I 0 L∗
If a densely defined operator A has a matrix representation, this need not be
true for A∗ even if A is closed; see the following example.
Example 2.2. Let X1 = X2 and let A be a closed densely defined operator on X1
with D(A) 6= X1 . Consider the block operator matrix
A 0
A :=
A 0
on X1 × X1 . It is easy to verify that A is a closed operator with domain D(A) =
D(A) × X1 . Furthermore, we see from Theorem 2.1 that
I 0 A 0
A= .
I 0 0 0
By Lemma A.1,
∗ A∗ 0 I I
A = ,
0 0 0 0
so that
∗ x1 ∗
D(A ) = ∈ X1 × X1 | x1 + x2 ∈ D(A ) .
x2
Since A is unbounded, we have D(A∗ ) 6= X1 . Taking x1 ∈ X1 \ D(A∗ ), then
x1 x1
∈ D(A∗ ) but 6∈ D(A∗ ).
−x1 0
Hence, by Lemma 2.1, A∗ has no block operator matrix representation.
3. self-adjointness
Let H1 , H2 be complex Hilbert spaces. Now we consider self-adjointness of block
operator matrices with unbounded entries acting on the Hilbert space H1 × H2 .
First we shall consider
necessary
conditions for a block operator matrix to be
A B
self-adjoint. Let A := be a block operator matrix acting on H1 × H2
C D
with dense domain D1 × D2 . Clearly, A is symmetric if and only if
(3.1) A|D1 ⊂ (A|D1 )∗ , B|D2 ⊂ (C|D1 )∗ , D|D2 ⊂ (D|D2 )∗ .
Furthermore, it follows from Theorem 2.2 and (3.1) the following assertion.
Proposition 3.1. If A is self-adjoint, then
A|D1 B|D2
A= = A× .
C|D1 D|D2
In addition, we point out that the entry operators of a self-adjoint block operator
matrix need not be closed, see the following example.
BLOCK OPERATOR MATRICES 7
Example 3.1. Let X := (L2 (0, 1) × L2 (0, 1)) × (L2 (0, 1) × L2 (0, 1)) Let M be the
differential operator on the Hilbert space L2 (0, 1) which is defined by
D(M ) := {f ∈ H 1 (0, 1) | f (0) = f (1) = 0}, M f := if ′ .
From the methods of differential operators we know that M is closed, Cc∞ (0, 1) is a
core of M , and M ∗ is determined by
D(M ∗ ) := H 1 (0, 1), M ∗ f := if ′ .
Let LD be the same as in Example 2.1 and let M0 := M |D(LD ) . For the block
operator matrix
LD 0 M0 0
0 M0 0 LD A B
A := =:
M0 0 LD 0 B A
0 LD 0 M0
on the Hilbert space X, we claim that
(a) A is self-adjoint,
(b) A, B are not closed and A ( A∗ , B ( B ∗ .
In fact, it is easy to see that
I 0 0 0 LD M0 0 0 I 0 0 0
0 0 I 0 M0 LD 0 0 0 0 I 0
A= 0 I 0 0 0
=: EBE.
0 M0 LD 0 I 0 0
0 0 0 I 0 0 LD M0 0 0 0 I
By interpolation
theorem ofSobolev spaces (see[1, Theorem 5.2]) and Lemma A.3,
M0 LD LD M0
the operators and are self-adjoint. Hence, B is self-
LD M0 M0 LD
adjoint (see [29, Proposition 2.6.3]). Since E = E −1 = E, it follows from Lemma
∗
A.2 and Lemma A.3 that A is self-adjoint. This proved the first claim. The second
claim follows from the following four equalities:
∗ LD 0 LD 0
A = , A= ,
0 M∗ 0 M
M∗ 0 M 0
B∗ = , B= .
0 LD 0 LD
Next we consider sufficient conditions for a block operator matrix to be (essen-
tially) self-adjoint. In view of (3.1) and Example 3.1, through out the rest of this
section we make the following basic assumptions:
(i) A, B, C, D are densely defined and closable,
(ii) A ⊂ A∗ , B ⊂ C ∗, D ⊂ D∗ ,
A B
(iii) A := is densely defined on H1 × H2 .
C D
Further assumptions will be formulated where they are needed.
Proposition 3.2. A is self-adjoint if one of the following statements holds:
(a) A, D are self-adjoint, C is A-bounded with relative bound < 1, and B is
D-bounded with relative bound < 1.
(b) B is closed, C = B ∗ , A is C-bounded with relative bound < 1, and D is
B-bounded with relative bound < 1.
8 JIN AND ALATANCANG
Proof. We prove the claim in case (a); the proof in case (b) is analogous. Writing
A 0 0 B
A= + =: S + T .
0 D C 0
Then S is self-adjoint and T is symmetric (see [29, Proposition 2.6.3]). Furthermore,
by the assumptions T is S-bounded with relative bound < 1. By applying Lemma
A.3 to S, T , we complete the proof.
Proposition 3.3. A is essentially self-adjoint if one of the following statements
holds:
(a) A, D are self-adjoint, and for some a ≥ 0,
kCxk2 ≤ akxk2 + kAxk2 , x ∈ D(A),
kByk2 ≤ akyk2 + kDyk2 , y ∈ D(D).
(b) B is closed, C = B ∗ , and for some a ≥ 0,
kAxk2 ≤ akxk2 + kCxk2 , x ∈ D(C),
kDyk2 ≤ akyk2 + kByk2 , y ∈ D(B).
Proof. We prove e.g. case (a). By the assumptions, we have, for all (x, y)t ∈
D(A) × D(D),
√
0 B x
≤ a
x
+
A 0 x
.
C 0 y
y
0 D y
Consequently, the assertion follows from the Wüst theorem (see [30, Theorem 4]).
Theorem 3.1. The following statements hold.
(a) If D = D∗ , D(D) ⊂ D(B), then A is self-adjoint if and only if
(A − B(D − λ)−1 C)∗ = A − B(D − λ)−1 C
for some (and hence for all) λ ∈ ρ(D).
(b) If A = A∗ , D(A) ⊂ D(C), then A is self-adjoint if and only if
(D − C(A − λ)−1 B)∗ = D − C(A − λ)−1 B
for some (and hence for all) λ ∈ ρ(A).
Proof. Proof of (a). Let λ ∈ ρ(D). By applying Corollary 2.1 to (A − λ) we have
I B(D − λ)−1 S1 (λ) 0
(3.2) A−λ =
0 I 0 D−λ
I 0
,
(D − λ)−1 C I
where S1 (λ) := A − λ − B(D − λ)−1 C. We see from D(D) ⊂ D(B), B ⊂ C ∗
that D(D) ⊂ D(C ∗ ), so that (D − λ)−1 C is bounded on its domain D(C) (see [3,
Proposition 3.1]). Moreover, in (3.2), the domain of the middle factor is equal to
(D(A) ∩ D(C)) × D(D), and so we can replace (D − λ)−1 C = ((D − λ)−1 C)|D(C)
by
(D − λ)−1 C = ((D − λ)−1 C)∗∗
= (C ∗ (D − λ)−1 )∗ (by Lemma A.1)
−1 ∗
= (B(D − λ) ) .
BLOCK OPERATOR MATRICES 9
It follows that
I B(D − λ)−1 S1 (λ) 0
(3.3) A−λ =
0 I 0 D−λ
I 0
.
(B(D − λ)−1 )∗ I
In the factorization (3.3), the first and last factor are bounded and boundedly
invertible, and therefore by Lemma A.1 and Lemma A.2,
S1 (λ)∗
I B(D − λ)−1 0
(3.4) A∗ − λ =
0 I 0 D−λ
I 0
.
(B(D − λ)−1 )∗ I
Furthermore, in (3.3) we can replace λ by λ and obtain
I B(D − λ)−1 S1 (λ) 0
(3.5) A−λ =
0 I 0 D−λ
I 0
.
(B(D − λ)−1 )∗ I
We conclude from (3.4),(3.5) that A∗ = A if and only if S1 (λ)∗ = S1 (λ).
Proof of (b). Let λ ∈ ρ(A). Similar to the proof of (a) we have A∗ = A if and
only if S2 (λ)∗ = S2 (λ), where S2 (λ) := D − λ − C(A − λ)−1 B.
Corollary 3.1. Let A = A∗ , D = D∗ . Then A is self-adjoint if one of the following
holds:
(a) C is A-bounded with relative bound < 1 and B is D-bounded with relative
bound ≤ 1,
(b) C is A-bounded with relative bound ≤ 1 and B is D-bounded with relative
bound < 1.
Proof. We prove the claim in case (a); the proof in case (b) is analogous. It is
enough to prove
(3.6) (A − B(D − iλ)−1 C)∗ = A − B(D + iλ)−1 C for some λ > 0.
Step 1. We start from the claim that for λ > 0 large enough, B(D − iλ)−1 C is
A-bounded with relative bound < 1. Since C is A-bounded with relative bound
< 1, it is enough to prove for each ε > 0 there exists λ > 0 such that
kB(D − iλ)−1 k < 1 + ε.
We observe that for x ∈ D(D) and λ > 0,
(3.7) k(D − iλ)xk2 = kDxk2 + λ2 kxk2
since D is self-adjoint. By the assumption that B is D-bounded with relative bound
≤ 1, there exists a(ε) ≥ 0 such that
ε
(3.8) kBxk ≤ (1 + )kDxk + a(ε)kxk, x ∈ D(D),
2
so that for x ∈ D(D), we have, using (3.7) twice and then (3.8),
ε a(ε)
kBxk ≤ (1 + + )k(D − iλ)xk.
2 λ
10 JIN AND ALATANCANG
Step 2. In this step, we show that for λ > 0 large enough, (B(D − iλ)−1 C)∗
is A-bounded with relative bound < 1. Since D(D) ⊂ D(B) and B is closable,
B(D − iλ)−1 is everywhere defined and closed, so that it is bounded by the closed
graph theorem. Thus, by Lemma A.1,
(B(D − iλ)−1 C)∗ = C ∗ (B(D − iλ)−1 )∗ = C ∗ (D + iλ)−1 B ∗ .
Now (D + iλ)−1 B ∗ is bounded on D(B ∗ ) since D(B) ⊃ D(D) (see [3, Proposition
3.1]), so that (C ∗ (D + iλ)−1 B ∗ )|D(B ∗ ) = C ∗ (D + iλ)−1 B ∗ = B(D + iλ)−1 B ∗ . Then
it follows from D(A) ⊂ D(C), C ⊂ B ∗ that
(3.9) (B(D − iλ)−1 C)∗ |D(A) = (B(D + iλ)−1 C)|D(A) .
Thus, by Step 1, (B(D − iλ)−1 C)∗ is A-bounded with relative bound < 1.
Step 3. Now (3.6) follows from Step 1 and Step 2 by applying Lemma A.3 and
(3.9).
Corollary 3.2. Let A = A∗ , D = D∗ . Then A is self-adjoint if one of the following
holds:
(a) C is A-bounded with relative bound 0, and D(D) ⊂ D(B).
(b) D(A) ⊂ D(C), and B is D-bounded with relative bound 0.
Proof. We prove the claim in case (a); the proof in case (b) is analogous. Let
λ ∈ ρ(A). We need to prove
(3.10) (D − C(A − λ)−1 B)∗ = D − C(A − λ)−1 B.
Step 1. First we claim that C(A−λ)−1 B is D-bounded with relative bound 0. Since
C is A-bounded with relative bound 0, for each ε > 0, there exists b1 (ε, λ) ≥ 0,
such that
kC(A − λ)−1 xk ≤ εkxk + b1 (ε, λ)k(A − λ)−1 xk, x ∈ H1 ,
so that for x ∈ D(B),
kC(A − λ)−1 Bxk ≤ εkBxk + b1 (ε, λ)k(A − λ)−1 Bxk
≤ εkBxk + b2 (ε, λ)kxk,
where the last inequality follows from the fact that (A− λ)−1 B is bounded on D(B)
(since D(B ∗ ) ⊃ D(C) ⊃ D(A)). Furthermore, since D is closed and D(D) ⊂ D(B),
there are a, b ≥ 0 such that
kBxk ≤ akDxk + bkxk, x ∈ D(D),
so that
kC(A − λ)−1 Bxk ≤ εakDxk + b3 (ε, λ)kxk, x ∈ D(D).
Step 2. We have, with arguments similar to the ones used in the proof of (3.9),
(C(A − λ)−1 B)∗ |D(D) = C(A − λ)−1 B|D(D) ,
so that by Step 1, (C(A − λ)−1 B)∗ is D-bounded with relative bound 0.
Step 3. Finally, (3.10) follows from Step 1 and Step 2 by applying Lemma A.3.
The following analogue of Theorem 3.1 can be proved in the same way.
Theorem 3.2. The following statements hold.
BLOCK OPERATOR MATRICES 11
Remark 3.1. It follows from Corollary 3.3 and Theorem 3.1 that the linearized
Navier-Stokes operator considered in [6] is essentially self-adjoint and it is not
closed.
By the techniques used in the proofs of [2, Theorem 3.1] and the correspond-
ing corollaries therein, with some slight modifications, we can prove the following
theorem and related corollaries.
12 JIN AND ALATANCANG
Lemma A.3. ([8, Corollary 1]) Let T be closed densely defined operators from X
to Y . Suppose S is a T -bounded operator such that S ∗ is T ∗ -bounded, with both
relative bounds < 1. Then S + T is closed and (S + T )∗ = S ∗ + T ∗ .
Acknowledgment. Project supported by Natural Science Foundation of China
(Grant Nos. 11371185, 11361034) and Major Subject of Natural Science Foundation
of Inner Mongolia of China (Grant No. 2013ZD01). The authors would like to
express their sincere thanks to Tin-Yau Tam for his useful comments.
References
[1] R. A. Adams, J. J. F. Fournier, Sobolev spaces, Second edition, Academic Press, Amsterdam,
2003.
[2] Alatancang, G. Jin, D. Wu, On symplectic self-adjointness of Hamiltonian operator matrices,
SCIENCE CHINA Mathematics, to appear.
[3] F. V. Atkinson, H. Langer, R. Mennicken, A. A. Shkalikov, The essential spectrum of some
matrix operators, Math. Nachr. 167 (194) 5–20.
[4] K-J. Engel, Matrix representation of linear operators on product spaces, Rend. Circ. Mat.
Palermo (2) Suppl. 56 (1998) 219–224.
[5] D. Eschwé, M. Langer, Variational principles for eigenvalues of self-adjoint operator functions,
Integr. Equ. Oper. Theory 49 (3) (2004) 287–321.
[6] M. Faierman, R. J. Fries, R. Mennicken, M. Möller, On the essential spectrum of the linearized
Navier-Stokes operator, Integr. Equ. Oper. Theory 38 (1) (2000) 9–27.
[7] S. Hassi, A. Sandovici, H. de Snoo, H. Winkler, Extremal extensions for the sum of nonneg-
ative selfadjoint relations, Proc. Amer. Math. Soc. 135 (10) (2007) 3193–3204.
[8] P. Hess, T. Kato, Perturbation of closed operators and their adjoints, Comment. Math. Helv.
45 (1970) 524–529.
[9] T. Kato, Perturbation theory for linear operators, Corrected printing of the second edition,
Springer, Berlin, 1980.
[10] N. D. Kopachevsky, S. G. Krein, Operator approach to linear problems of hydrodynamics,
Vol. I: self-adjoint problems for an ideal fluid, Birkhäuser, Basel, 2001.
[11] N. D. Kopachevsky, S. G. Krein, Operator approach to linear problems of hydrodynamics,
Vol. II: nonself-adjoint problems for viscous fluids, Birkhäuser, Basel, 2003.
[12] M. Kraus, M. Langer, C. Tretter, Variational principles and eigenvalue estimates for un-
bounded block operator matrices and applications, J. Comput. Appl. Math. 171 (1-2) (2004)
311–334.
[13] C. R. Kuiper, H. J. Zwart, Connections between the algebraic Riccati equation and the
Hamiltonian for Riesz-spectral systems, J. Math. Systems Estim. Control 6 (4) (1996) 1–48.
[14] A. E. Lifschitz, Magnetohydrodynamics and spectral theory (Developments in Electromag-
netic Theory and Applications, 4), Kluwer Academic, Dordrecht, 1989.
[15] C. W. Lim, X. S. Xu, Symplectic elasticity: theory and applications, Applied Mechanics
Reviews 63 (5) (2011), article ID 050802, 10 pages.
[16] R. E. Megginson, An introduction to Banach space theory (Graduate Texts in Mathematics,
183), Springer, New York, 1998.
[17] R. Mennicken, A. A. Shkalikov, Spectral decomposition of symmetric operator matrices,
Math. Nachr. 179 (1996) 259–273.
[18] M. Möller, F. H. Szafraniec, Adjoints and formal adjoints of matrices of unbounded operators,
Proc. Amer. Math. Soc. 136 (6) (2008) 2165–2176.
[19] R. Nagel, Towards a “matrix theory” for unbounded operator matrices, Math. Z., 201 (1989)
57–68.
[20] B. Sz.-Nagy, C. Foias, H. Bercovici, L. Kérchy, Harmonic analysis of operators on Hilbert
space, Revised and enlarged edition, Springer, New York, 2010.
[21] M. A. Naimark, Linear differential operators, Part II: Linear differential operators in Hilbert
space, Frederick Ungar, New York, 1968.
[22] J. Qi, S. Chen, Essential spectra of singular matrix differential operators of mixed order in
the limit circle case, Math. Nachr. 284 (2-3) (2011) 342–354.
[23] J. Qi, S. Chen, Essential spectra of singular matrix differential operators of mixed order, J.
Differ. Equ. 250 (12) (2011) 4219–4235.
14 JIN AND ALATANCANG
[24] M. Schechter, The conjugate of a product of operators, J. Funct. Anal. 6 (1970) 26–28.
[25] M. Strauss, Spectral estimates and basis properties for self-adjoint block operator matrices,
Integr. Equ. Oper. Theory 67 (2) (2010) 257–277.
[26] H. Sun, Y. Shi, Self-adjoint extensions for linear Hamiltonian systems with two singular
endpoints, J. Funct. Anal. 259 (8) (2010) 2003–2027.
[27] A. E. Taylor, D. C. Lay, Introduction to functional analysis, Second edition, John Wiley and
Sons, New York, 1980.
[28] B. Thaller, The Dirac equations (Theoretical and Mathematical Physics), Springer, Berlin,
1992.
[29] C. Tretter, Spectral theory of block operator matrices and applications, Imperial College
Press, London, 2008.
[30] R. Wüst, Holomorphic operator families and stability of selfadjointness, Math. Z. 125 (1972),
349–358.
[31] C. Wyss, Perturbation theory for Hamiltonian operator matrices and Riccati equations, PhD
thesis, University of Bern, 2008.
[32] C. Wyss, Hamiltonians with Riesz bases of generalised eigenvectors and Riccati equations,
Indiana Univ. Math. J., 60 (5) (2011) 1723–1766.
[33] W. Yao, W. Zhong, C. W. Lim, Symplectic elasticity, World Scientific, New Jersey, 2009.
[34] W. Zhong, X. Zhong, Method of separation of variables and Hamiltonian system, Numer.
Methods Partial Differential Equations 9 (1) (1993) 63–75.