Geom Dedicata (2013) 167:11–21

DOI 10.1007/s10711-012-9800-7

ORIGINAL PAPER

Two functionals connected to the Laplacian in a class

of doubly connected domains on rank one symmetric
spaces of non-compact type

Anisa M. H. Chorwadwala · M. K. Vemuri

Received: 5 March 2012 / Accepted: 8 November 2012 / Published online: 22 November 2012
© Springer Science+Business Media Dordrecht 2012

Abstract Let B1 be a ball in a non-compact rank-one symmetric space and let B0 be a

smaller ball inside it. It is shown that if y is the solution of the problem −u = 1 in B1 \ B0
vanishing on the boundary, then the Dirichlet-energy of y is minimal if and only if the balls
are concentric. It is also shown that the first Dirichlet eigenvalue of the Laplacian on B1 \ B0 is
maximal if and only if the two balls are concentric. The formalism of Damek-Ricci harmonic
spaces is used.

Keywords Shape optimization problem · Rank one symmetric spaces

of non-compact type · Dirichlet boundary value problem

Mathematics Subject Classification (1991) 35J20 · 49R05 · 58J32

1 Introduction

Let M be a Riemannian manifold with metric g and Laplace-Beltrami operator . We use the
analyst’s convention:  is a negative operator. Let B1 be an open (geodesic) ball in M. Let B0
be an open ball whose closure is contained in B1 . Consider the family F := { = B1 \B0 }
and the following problems for  ∈ F :
−u = 1 in ,
(1)
u = 0 on ∂,

A. M. H. Chorwadwala (B)
Indian Institute of Science Education and Research, Pashan, Pune 411021, India
e-mail: anisa23in@gmail.com

M. K. Vemuri
Department of Mathematics, West Virginia University, Morgantown, WV 26506-6310, USA

Present Address:
M. K. Vemuri
Chennai Mathematical Institute, H1, SIPCOT IT Park, Siruseri, Kelambakkam, Chennai 603103, India
e-mail: mkvemuri@gmail.com

123
12 Geom Dedicata (2013) 167:11–21

and

−u = λu in ,
(2)
u = 0 on ∂.

Let y ∈ H01 () be the unique weak solution of (1). By Theorem 4.8, page 105 of [2],
y ∈ C ∞ (). The energy functional E() for problem (1) is defined as
 
E() := ∇ y2 d x = y d x.
 

The eigenvalues of the operator − are strictly positive. The eigenfunctions correspond-
ing to the first eigenvalue λ1 are proportional to each other. They belong to C ∞ () and they
are either strictly positive or strictly negative on . Moreover, λ1 = inf { ∇ϕ2L 2 () | ϕ ∈
H01 (), ϕ2L 2 () = 1} (cf. Theorem 4.4, p.102, [2]). Let y1 := y1 () be the unique solution
of
 (2), corresponding to the first eigenvalue λ1 (= λ1 ()), characterized by y1 > 0 on  and
y 2 d V = 1.
 1
Throughout this article, S will denote a rank-one symmetric space of noncompact type.
We study the extreme of the energy functional E() for problem (1) and the first eigenvalue
λ1 () of problem (2) over the family F = { = B1 \B0 } in S. We state our main results:

Theorem 1 The energy functional E() for problem (1), attains its minimum over F if and
only if B0 and B1 are concentric.

Theorem 2 The first eigenvalue λ1 () of problem (2) attains its maximum over F if and
only if the balls are concentric.

In the case of a Euclidean space, these results are proved by Kesavan [10] (see also Ramm–
Shivakumar [11]). The proofs in [10,11] rely on shape differentiation [12] and the moving
plane method [3,8]. In the application of the moving plane method in [10,11], the commu-
tativity of the Laplacian and reflection in the hyperplane is used.
Anisa and Aithal [1] developed a shape calculus on Riemannian manifolds, and used it to
prove the analogues of the results of Kesavan and Ramm–Shivakumar on space-forms (com-
plete simply connected Riemannian manifolds of constant sectional curvature). The reflection
method works here just as in Euclidean space, because reflection in a hyperplane is an isome-
try in any space form, and so it commutes with the Laplacian. However, a moments reflection
shows that if every hyperplane reflection in a simply connected Riemannian manifold M is
an isometry, then M must be a space-form. We observe that because of the presence (in our
situation) of cylindrical symmetry about a geodesic, the arguments used in the reflection
method are still valid if we replace reflections by geodesic inversions.
If M = S, a noncompact rank-one symmetric space then it is a Damek-Ricci harmonic
space S = N A (semidirect product) where N is an H -type group and A = R+ acts on N by
“dilations”. We show that if the centers of both B0 and B1 lie on the A-axis then the analogous
of the results of [1,10,11] hold. Moreover, since the isometry group of S is two-transitive on
S the restriction on the centers of B0 and B1 may be dropped.

123
Geom Dedicata (2013) 167:11–21 13

2 Damek-Ricci spaces

Let n be a two-step nilpotent Lie algebra with inner product ( | ). Let z be the center of n and
v = z⊥ . Then n is called an H-type algebra if for every unit vector X ∈ v

ad X : (ker ad X )⊥ → z
is a surjective isometry [7,9]. Given Z ∈ z, the skew-symmetric map J Z : v → v given by
(J Z X |Y ) = (Z |[X, Y ])
satisfies J Z2 = − Z 2 I . We use the Lie-theoretic exponential map to parametrize the ele-
ments of the H-type group N = exp n by pairs (X, Z ) ∈ v ⊕ z. The product on N is then
given by
 
1
(X, Z )(X , Z ) = X + X , Z + Z + [X, X ] ,
2
and the Lebesgue measure d X d Z is a Haar measure on N .
Let A = R+ act on N by the dilations (X, Z ) → (a 1/2 X, a Z ), and let S = N A be the
semi-direct product. The product on S is given by
 
1 1 1
(X, Z , a)(X , Z , a ) = X + a 2 X , Z + a Z + a 2 [X, X ], aa .
2
The Lie algebra of S is then s = v ⊕ z ⊕ R, on which we put the inner product
((X, Z , t)|(X , Z , t )) = (X |X ) + (Z |Z ) + tt .
This defines a unique left-invariant Riemannian metric g on S with volume element
dm L = a −Q−1 d X d Z da
where Q = (m/2) + k, m = dim v, and k = dim z (cf. [7]). Of course, dm L defines the
left Haar measure on the group S. It is shown in [6] that the sectional curvature of S is
non-positive, so by the Cartan-Hadamard theorem, for all p ∈ S the Riemannian exponential
map exp p : T p S → S is a diffeomorphism.
 
Via the map h(X, Z , a) = X, Z , a + 14 X 2 , we identify S with the generalized Siegel
domain D = {(X, Z , t) ∈ s | t > 41 X 2 }.
Let σ be the geodesic inversion of the symmetric space D around the base point o :=
(0, 0, 1), i.e., the map such that σ · Expo (T ) = Expo (−T ) for all tangent vectors T at o of D;
here Exp denotes the exponential map of the Riemannian manifold D. Let G denote the full
isometry group of D and g its Lie algebra. Let K be the stabilizer of the base point o. Then
G contains both S and σ . Let M be the centralizer of A in K and let m be its Lie algebra.
Then, by Lemma 5.5 of [4], Ad(M)|n is a group of all orthogonal automorphisms of n. By
Kostant’s double transitivity theorem, Ad(M) acts transitively on Sv × Sz where Sv and Sz
are unit spheres in v and z respectively (cf. Theorem 6.2 of [4]).
We also have a generalized Cayley transform from the unit ball

B = (X, Z , t) ∈ s | X 2 + Z 2 + t 2 < 1
onto D:
1  
C (X, Z , t) = 2(1 − t + J Z )X, 2Z , 1 − t 2 − Z 2 .
(1 − t)2 + Z 2

123
14 Geom Dedicata (2013) 167:11–21

The inverse transform C −1 : D → B is given by

1  
C −1 (Y, W, s) = (1 + s − JW )Y, 2W, −1 + s 2 + W 2 .
(1 + s)2 + W 2
Note that there is a typographical error in equation (4.5) (p. 229) of [7]. It is shown in [5]
that if we transfer the metric g to B using the map C −1 ◦ h, then the radii in B are geodesics
and the distance to the origin is a radial function on B. It follows that the distance in S
between (0, 0, 1) and (X, Z , t) depends only on X , Z  and t. Since the distance function
is left-invariant, it now follows that the distance in S between (0, 0, a) and (X, Z , t) also
depends only on X , Z  and t for any a ∈ R+ .

3 Proof of the main results

Recall from Sect. 1 that S is a noncompact rank-one symmetric space and hence a Damek-Ric-
ci harmonic space S = N A (semidirect product) where N is an H -type group and A = R+
acts on N by “dilations”. In this section, we show that if the centers of both B0 and B1 lie on
the A-axis then the analogous of the results of [1,10,11] hold. Moreover, since the isometry
group of S is two-transitive on S the restriction on the centers of B0 and B1 may be dropped.
For x ∈ S and r ∈ R+ , B(x, r ) denotes the (open) geodesic ball in S with center x
and radius r . Fix r0 , r1 ∈ R+ such that 0 < r0 < r1 . Also, fix a ∈ (0, 1) such that
B((0, 0, a), r1 ) ⊂ S and d((0, 0, a), (0, 0, 1)) < r1 − r0 . Without loss of generality we
can take B1 = B((0, 0, a), r1 ). Let B0 be any ball of radius r0 with center on the A-axis
such that B0 ⊂ B1 . That is, B0 = B((0, 0, et ), r0 ) for some t ∈ R such that d(t) :=
d((0, 0, a), (0, 0, et )) < r1 − r0 . Now we want to study the behavior of E() associated to
problem (1), and of λ1 () associated to problem (2), as the center of B0 moves along the
A-axis in such a way that its closure is still inside B1 . For convenience, put p = (0, 0, a)
and q(t) = (0, 0, et ) ∈ S. Then B1 = B( p, r1 ) and d(t) = d( p, q(t)), t ∈ R. We need to
study the functionals E and λ1 only on domains (q(t)) := B1 \B(q(t), r0 ) for t such that
0 ≤ d(t) < r1 − r0 . Let us denote the set of such permissible real numbers t by E. That is,
E = {t ∈ R | 0 ≤ d(t) < r1 − r0 }. We define j, j1 : E → (0, ∞) as follows:
j (t) = E((q(t))) and j1 (t) = λ1 ((q(t))).
Fix t0 ∈ E. Put  := (q(t0 )) and B0 := B(q(t0 ), r0 ). It is easy to see that B0 (t) :=
B(q(t0 + t), r0 ) is nothing but (0, 0, et ) · B0 . We have, B0 (t) ⊂ B1 if and only if t0 + t ∈ E.
We consider domains in F of the form B1 \B0 (t) for t0 + t ∈ E.
Consider the vector field W (X, Z , a) = ( X2 , Z , a). Then W generates the left translations
of S by (0, 0, et ). Fix a neighbourhood (say N (B0 )) of B0 . Let ρ : S → R be a smooth
function satisfying ρ ≡ 1 in N (B0 ) and ρ = 0 on ∂ B1 . Put V = ρ W , and let { t } be the
one parameter group of diffeomorphisms of S generated by V . Then, for t sufficiently small,
t () = (q(t0 + t)). And hence, for t sufficiently close to 0, E( t ()) = j (t0 + t) and
λ1 ( t ()) = j1 (t0 + t). By Proposition 2.4 (resp. 3.1) in [1] j (resp. j1 ) is differentiable at
t0 . In particular, j (resp. j1 ) is differentiable at log a ∈ E. Note that since geodesic symmetry
about p is an isometry of S both j and j1 depend only on d(t). Hence j (log a) = 0 =
j1 (log a). Also, it is enough to study the behaviour of j, j1 on E := {t ∈ E | et > a}.
Henceforth we fix t0 ∈ E . By the choice of a we have 0 ∈ E . For the sake of
computational convenience we take t0 = 0; and consider  := B1 \B0 (0). That is,
 = B1 \B((0, 0, 1), r0 ). Consider the geodesic symmetry of B w.r.t to the origin (0,0,0)
namely I : B → B given by I (X, Z , a) = (−X, −Z , −a). Then σ , the geodesic inversion

123
Geom Dedicata (2013) 167:11–21 15

of the symmetric
  D around the base point (0, 0,
space 1) as described in Sect. 2, is nothing but
the map C ◦ I ◦ C −1 . That is, σ · (X, Z , a) = a 2 +Z
1
2
((−a + J Z )X, −Z , a). The geodesic
symmetry in S w.r.t the point (0, 0, 1) namely I := (h −1 ◦ σ ◦ h) : S → S is given by
I (X, Z , a) = 2 1 2 ((−a1 + J Z )X, −Z , a), where a1 = a + X4 .
2

a +Z 
1
Let H denote the hypersurface {(X, Z , a) ∈ S | a12 + Z 2 = 1} in S. Notice
 thatH is
nothing but the image of the hyperplane {(X, Z , 0) ∈ B} in B under the map h −1 ◦ C . We
 
will denote the domain  ∩ h −1 ◦ C ({(X, Z , a) ∈ B | a > 0}) by O. Then, we have the
following (See (A) in the appendix for proof):
I (∂ B0 ) = ∂ B0 , I (H) = H, I (O) ⊂ , (d I )x (V (x)) = −V (I (x))∀ x ∈ ∂ B0 .

Moreover, we have
(d I )x (n(x)) = n(I (x))∀ x ∈ ∂ B0 . (3)
Here, n denotes the outward unit normal of  on ∂. Since I is an isometry of S we have
gx (V, n) = gI (x) ((d I )x (V (x)), (d I )x (n(x))) = −gI (x) (V, n) .
Therefore, gI (x) (V, n) = −gx (V, n)∀ x ∈ ∂ B0 . For x ∈ ∂ B0 ∩ ∂ O, let x := I (x). Then,
it follows from Propositions 2.4 and 3.4 in [1] that
 2 2
∂y ∂y
j (0) = (x) − (x ) gx (V, n) d S
∂n ∂n
x∈∂ B0 ∩∂ O

and,
 2 2
∂ y1 ∂ y1
j1 (0) = − (x) − (x ) gx (V, n) d S.
∂n ∂n
x∈∂ B0 ∩∂ O

Recall here that, as described in Sect. 1, y and y1 are the unique solutions of (1) and (2)
respectively. Let Hc denote the complement of H in S. Then, we also have the following
(See (B) in the Appendix for proof):
gx (V, n) < 0 ∀ x ∈ ∂ B0 ∩ ∂ O ∩ Hc .
So, we only need to compare the absolute values of the normal derivatives at x and x of
both y as well as y1 .
Now, for u = y or y1 , the following is true: u ∈ C ∞ (). Since u > 0 in , by the Hopf
∂u
Lemma we have, (x) < 0 ∀ x ∈ ∂. Let ũ(x) := (u ◦ I )(x) ∀ x ∈ O. Then by (3),
∂n
∂ ũ ∂u
∂n (x) = ∂n (x ) < 0 ∀ x ∈ ∂ ∩ ∂ O . Define w on O as w = u − ũ. Using the facts that
Laplacian commutes with I , I (H) = H and that I preserves (X , Z , t) for (X, Z , t) ∈ H
it follows that w satisfies the following elliptic boundary value problem:
L(w) = 0 in O,
w = 0 on ∂ O ∩ (H ∪ ∂ B0 ) , (4)
w < 0 on ∂ O ∩ (∂ B1 ) ,

 if u = y, the solution of (1)

L= .
 + λ1 if u = y1 , the solution of (2)

123
16 Geom Dedicata (2013) 167:11–21

Thus if u = y, the solution of problem (1), then the maximum principle applied to (4)
gives w < 0 in O, and the Hopf Lemma applied to (4) implies ∂w
∂n (x) > 0 ∀ x ∈ ∂ O ∩ ∂ B0 .
This proves that j (0) > 0.
Now let us consider problem (2). Since w ≤ 0 on ∂ O we have w + ∈ H01 (O). And from
(4) we get
   
0= (−w) w+ d x −λ1 () ww+ d x = (−w)w+ d x −λ1 () ww+ d x
O O O ∩{w≥0} O ∩{w≥0}
   
+ 2 + 2 + 2
= ∇w  d x − λ1 () (w ) d x > ∇w  d x − λ1 (O) (w+ )2 d x.
O O O O

∇w + 2 d x
O
This implies that λ1 (O) > 
+ 2
, a contradiction. Thus, w + ≡ 0. That is,
O (w ) d x
w ≤ 0 in O. A s a result we get w = −λ1 w ≥ 0 on O. Now, by the maximum principle it
∂w
follows that w < 0 in O; and then by Hopf Lemma it follows that > 0 a.e. on ∂ O ∩ ∂ B0 .
∂n
This proves that j1 (0) < 0. 

Appendix

We continue with the notations as in the previous sections.

(A) (d I )x (V (x)) = −V (I (x)) ∀ x = (X, Z , a) ∈ ∂ B0 .
Proof. We have
1
σ · (Y, W, b) = ((−b + JW )Y, −W, b)
b2 + W 2
and,
  1
h −1 σ h (X, Z , a) = ((−a1 + J Z )X, −Z , a) ,
a12 + Z 2

X 2  
where a1 = a + . Let I := h −1 σ h . Let α := a12 + Z 2 . Recall that in the neigh-
4
bourhood N (B0 ) of B0 , the vector field V is given by V (X, Z , a) = ( X2 , Z , a). Then, for
x = (X, Z , a) ∈ ∂ B0 we have
 
1 1
V (I (x)) = 2 [−a1 + J Z ]X, −Z , a .
a1 + Z 2 2

1
The Jacobian matrix J of I is given by J = [c1 c2 c3 ] where,
α2
⎛  2  ⎞
−αa1 Id +a1 − Z 2 (X ⊗ X ) + α J Z − a1 (J Z X ⊗ X )
1
2
c1 = ⎝ a1 (Z ⊗ X ) ⎠
−aa1 X, •
⎛ ⎞ ⎛ 2 ⎞
2a1 (X ⊗ Z ) + α J• X − 2(J Z X ⊗ Z ) (a1 − Z 2 )X − 2a1 J Z X
c2 = ⎝ −αId + 2(Z ⊗ Z ) ⎠ and c3 = ⎝ 2a1 Z ⎠
−2aZ , • α − 2aa1

123
Geom Dedicata (2013) 167:11–21 17

Let W = (W1 , W2 , W3 ) denote (d I )x (V (x)). Then,

   
1 X
W2 = 2 a1 (Z ⊕ X ) − α Z + 2 (Z ⊕ Z ) Z + 2 a1 a Z
α 2
 
1 X
= 2 a1  , X Z − α Z + 2Z , Z Z + 2 a a1 Z
α 2
1  a1 
= 2 X 2 − α + 2Z 2 + 2 a a1 Z
α 2
   
1 X 2
= 2 2 a1 a + − α + 2Z 2 Z
α 4
1   2   1 Z
= 2 2 a1 + Z 2 − α Z = 2 [2 α − α] Z = ,
α α α
 
1 X 2
W3 = 2 −a a1 − 2 a Z 2 + a α − 2 a 2 a1
α 2
 
a X 2
= 2 −a1 − 2 Z 2 + α − 2 a a1
α 2
 
a X 2
= 2 −a1 − Z 2 + a12 − 2 a a1
α 2
  
a X 2
= 2 − Z 2 + a12 − 2 a1 a +
α 4
a   −a
= 2 − Z  − a1 =2 2
.
α α
Finally,

1 −α a1   X 2 α X 2
W1 = 2 X − a12 − Z 2 X + J Z X − 4 a1 J Z X + 2 a1 Z 2 X
α 2 4 2 2

 
+ α J Z X − 2 Z 2 J Z X + a a12 − Z 2 X − 2 a a1 J Z X
 
1 −α a1  2 
= 2 + a1 − Z 2 a1 + 2 a1 Z 2 X
α 2

1 
+ α − a1 X  + 2 α − 4 Z  − 4 a a1 J Z X
2 2 2
2
 
1   α  1 
= 2 a1 − + a12 + Z 2 X + 3 α − 4 Z 2 − 4 a12 J Z X
α 2 2
 
1 α 1
= 2 a1 X + [3 α − 4 α] J Z X
α 2 2
1 1
= (a1 α − α J Z ) X = − (−a1 + J Z ) X.
2 α2 2α
Hence the proof. 

(B) gx (V, n) < 0 ∀ x ∈ ∂ B0 ∩ ∂ O ∩ Hc .
Proof. Suppose that (X, Z , s) ∈ n ⊕ a and X 2 + Z 2 + a 2 = 1. The geodetic γ : R →
N A such that γ (0) = (0, 0, 1) and γ̇ (0) = (X, Z , a) is given by the formula
 
2θ (1 − aθ ) 2θ 2 2θ 1 − θ2
γ (τ ) = X− J Z X, Z, ,
χ χ χ χ

123
18 Geom Dedicata (2013) 167:11–21

where θ = tanh(τ/2) and χ = (1 − aθ )2 + Z 2 θ 2 (see proposition 2.2 on page 14 of [5]).

Then
⎛  ⎞
(1 − aθ )2 − θ 2 Z 2 − 2θ (1 − aθ )J Z X
1 − θ2 ⎝ ⎠.
γ̇ (τ ) = [2(1 − aθ ) − χ] Z
χ2
(a − θ )(1 − aθ ) − θ Z  2

In the neighbourhood N (B0 ) of B0 , the vector field V is given by V (X, Z , a) = ( X2 , Z , a).

Then, for γ (τ ) ∈ N (B0 ) we have,
1 
V (γ (τ )) = θ [1 − aθ − θ J Z ]X, 2θ Z , 1 − θ 2 ,
χ
Since n = −γ̇ (τ0 ) for some τ0 > 0 we need to prove that gγ (τ ) (V, γ̇ (τ )) > 0 ∀ γ (τ ) ∈
∂ B0 ∩ ∂ O ∩ Hc . Notice that for γ (τ ) ∈ ∂ O ∩ Hc , a > 0.
The product on S is given by
 
1 1 1
L (X,Z ,a) (X , Z , a ) = (X, Z , a) · (X , Z , a ) = X +a X , Z + a Z + a [X, X ], aa .
2 2
2
So, for (X , Z , a ) ∈ T(X ,Z ,a ) S we get
⎛⎞ ⎛ √ ⎞
√aX
  X
⎜ a ⎟
d L (X,Z ,a) (X , Z , a ) ⎝ Z ⎠ = ⎝a Z + [X, X ]⎠ .
a 2
aa

As the Riemannian metric g on S is left invariant we have

 
g p (W1 , W2 ) = g L q p d(L q ) p W1 , d(L q ) p W2 ∀ p, q ∈ S, W1 , W2 ∈ T p S,

that is, L q is an isometry. Therefore,

 
g p (W1 , W2 ) = ge d(L p−1 ) p W1 , d(L p−1 ) p W2 ∀ p ∈ S, W1 , W2 ∈ T p S,

where e := (0, 0, 1), the identity element of S. Thus, for γ (τ ) ∈ N (B0 ) we have,
     
gγ (τ ) (V, γ̇ (τ )) = ge d L γ (τ )−1 γ (τ ) (V (γ (τ ))) , d L γ (τ )−1 γ (τ ) (γ̇ (τ )) .

1  √ 
For p = (X , Z , a ) ∈ S, p −1 = − a X , −Z , 1 . So,
a
⎛ ⎞ ⎛ √ ⎞
X aX
1 ⎜ 1 ⎟
d(L p−1 ) p ⎝ Z ⎠ = ⎝ Z − [X , X ]⎠ .
a a 2
a

Therefore, for p ∈ N (B0 ),

⎛ √ ⎞ ⎛ ⎞
a X
⎜
1 ⎜
X ⎟ ⎜ 2 √a ⎟
2 ⎟ ⎜ ⎟
d(L p−1 ) p (V ( p)) = ⎜ 1 ⎟=⎜ 1 ⎟.
a ⎝ Z − [X , X ]⎠ ⎝ Z ⎠
4 a
a 1

123
Geom Dedicata (2013) 167:11–21 19

Thus, for γ (τ ) ∈ N (B0 ) we have,

⎛ θ [1 − a θ − θ J ] X ⎞
Z
  
⎜ χ 1−θ ⎟
  ⎜ 2
⎟
d L γ (τ )−1 γ (τ ) (V (γ (τ ))) = ⎜
⎜ 2θ Z
⎟;
⎟
⎝ ⎠
1 − θ2
1

and
⎛√ ⎞
1 − θ2  
  1 ⎜ √ (1 − aθ ) − θ Z  − 2θ (1 − aθ )J Z X ⎟
2 2 2

d L γ (τ )−1 γ (τ ) (γ̇ (τ )) = ⎜ χ ⎟.
χ ⎝ [2(1 − aθ ) − χ] Z + θ 2 [X, J Z X ] ⎠
(a − θ )(1 − aθ ) − θ Z  2

Therefore,
     
gγ (τ ) (V, γ̇ (τ )) = ge d L γ (τ )−1 γ (τ ) (V (γ (τ ))) , d L γ (τ )−1 γ (τ ) (γ̇ (τ ))
     
= d L γ (τ )−1 γ (τ ) (V (γ (τ ))) , d L γ (τ )−1 γ (τ ) (γ̇ (τ ))
⎛⎛ ⎞
θ [1 − a θ − θ J Z ] X ⎞T ⎛ √ ⎞T
⎜⎜    1 − θ 2   ⎟
⎟ (1 − aθ)2 − θ 2 Z 2 − 2θ(1 − aθ)J Z X ⎟ ⎟
1 ⎜
⎜⎜
⎜
χ 1 − θ2 ⎟ ⎜ √χ
⎟ ⎜ ⎟ ⎟
= ⎜⎜ 2θ Z ⎟ ,⎝ ⎠ ⎟
χ ⎜⎝ ⎠ [2(1 − aθ) − χ ] Z + θ 2 [X, J Z X ] ⎟
⎝ 1 − θ2 ⎠
(a − θ)(1 − aθ) − θZ 2
1
⎡⎛ √ ⎞
1 ⎣⎝ θ [1 − a θ − θ J Z ] X 1 − θ2  
=    , √ (1 − aθ) − θ Z  − 2θ(1 − aθ)J Z X ⎠
2 2 2
χ χ 1 − θ2 χ
⎤
 
2θ Z 2⎦
+ , [2(1 − aθ) − χ ] Z + θ [X, J Z X ] + (a − θ)(1 − aθ) − θZ 
2
1 − θ2
⎡ ⎡ ⎤
1 ⎣θ ⎣  
= (1 − a θ) (1 − aθ)2 − θ 2 Z 2 X 2 + 2 θ 2 (1 − a θ)Z 2 X 2 ⎦
χ χ
⎤
2θ 2 θ3 2⎦
+ [2(1 − aθ) − χ ] Z  +
2
(Z , [X, J Z X ]) + (a − θ)(1 − aθ) − θZ 
1 − θ2 1 − θ2

1 θ    
= (1 − a θ) (1 − aθ)2 − θ 2 Z 2 X 2 + 2 θ 2 (1 − a θ)Z 2 X 2
χ χ

2θ 2 θ3
+ [2(1 − aθ) − χ ] Z 2 + (J Z X, J Z X ) + (a − θ)(1 − aθ) − θZ 2
1 − θ2 1 − θ2

1 θ  
= (1 − aθ)3 X 2 − θ 2 (1 − aθ) Z 2 X 2 + 2 θ 2 (1 − aθ) Z 2 X 2
χ χ

2 θ3 θ Z 2  
+ Z 2
X 2
+ 4(1 − aθ) − 2 χ − (1 − θ 2
) + (a − θ)(1 − aθ)
1 − θ2 1 − θ2

1 θ   2 θ3
= (1 − aθ)3 X 2 + θ 2 (1 − aθ) Z 2 X 2 + Z 2 X 2
χ χ 1 − θ2

123
20 Geom Dedicata (2013) 167:11–21


θ Z 2  
+ 4(1 − aθ) − 2 χ − (1 − θ 2 ) + (a − θ)(1 − aθ)
1−θ 2

1 θ (1 − aθ) X 2   2 θ3
= (1 − aθ)2 + θ 2 Z 2 + Z 2 X 2
χ χ 1 − θ2

θ Z 2  
+ 4(1 − aθ) − 2 χ − (1 − θ 2
) + (a − θ)(1 − aθ)
1 − θ2

1 θ Z 2  
= θ (1 − aθ) X 2 + 4(1 − aθ) − 2 χ − 1 + θ 2 + 2 θ 2 X 2
χ 1 − θ2

+ a (1 − aθ) − θ (1 − aθ)
⎡ ⎤
1 ⎣   θ Z 2  
= θ (1 − aθ) X  − 1 +
2
4(1−aθ)−2 χ − 1 + θ + 2 θ X  + a (1−aθ)⎦
2 2 2
χ 1 − θ2

1
= − θ (1 − aθ) Z 2 − a 2 θ (1 − aθ) + a (1 − aθ)
χ

θ Z 2  
+ 4(1 − aθ) − 2 χ − 1 + θ 2
+ 2 θ 2
X  2
1 − θ2
 
1 θ Z 2  
= −θ (1 − aθ) Z 2 + a (1 − aθ)(1 − aθ) + 4(1 − aθ) − 2 χ − 1 + θ 2
+ 2 θ 2
X  2
χ 1 − θ2
 
1 θ Z  2 
= a (1 − a θ) (1−a θ) + 4(1 − aθ) − 2 χ − 1 + θ 2 − (1 − aθ) (1−θ 2 ) + 2 θ 2 X 2
χ 1−θ 2
 2  
1 θ Z  
= a (1 − a θ)2 + 4(1 − aθ) − 2 χ − 1 + θ 2
− (1 − aθ) (1 − θ 2
) + 2 θ 2
X  2
χ 1 − θ2
 
1 θ Z 2  
= a (1 − a θ)2 + (1 − a θ) (3 + θ 2
) − 2 χ − 1 + θ 2
+ 2 θ 2
X  2
χ 1 − θ2
 
1 θ Z 2  
= a (1 − a θ)2 + 3 + θ 2
−3aθ −aθ 3
−2(1−aθ) 2
− 2 θ 2
Z  2
−1 + θ 2
+ 2 θ 2
X 2
χ 1 − θ2

1 θZ 2 
= a(1−aθ)2 + 3 + θ 2 − 3aθ − aθ 3 − 2 + 4 a θ − 2 a 2 θ 2
χ 1 − θ2


− 2 θ 2 Z 2 − 1 + θ 2 + 2 θ 2 X 2
 
1 θ 2 Z 2    
= a (1 − a θ)2 + 2 θ 1 − a 2
− Z 2
− aθ 2
+ a + 2 θ X 2
χ 1 − θ2
 
1 θ 2 Z 2  
= a (1 − a θ)2 + 4 θ X 2
+ a(1 − θ 2
)
χ 1 − θ2
 
1   θ 2 Z 2  
= a (1 − a θ)2 + θ 2 Z 2 + 4 θ X 2
χ 1−θ 2

4 θ 3 X 2 Z 2
=a+   ≥ a > 0.
χ 1 − θ2

123
Geom Dedicata (2013) 167:11–21 21

References

1. Anisa, M.H.C., Aithal, A.R.: On two functionals connected to the Laplacian in a class of doubly con-
nected domains in space-forms. Proc. Indian Acad. Sci. Math. Sci. 115(1), 93–102 (2005). MR 2120602
(2005k:58034)
2. Aubin, T.: Nonlinear analysis on manifolds. Monge-Ampère equations. Grundlehren der Mathematis-
chen Wissenschaften (Fundamental Principles of Mathematical Sciences), vol. 252. Springer, New York
(1982). MR 681859 (85j:58002)
3. Berestycki, H., Nirenberg, L.: On the method of moving planes and the sliding method. Bol. Soc. Brasil.
Mat. (N.S.) 22 (1), 1–37 (1991). MR 1159383 (93a:35048)
4. Cowling, M., Dooley, A., Korányi, A., Ricci, F.: An approach to symmetric spaces of rank one via groups
of Heisenberg type. J. Geom. Anal. 8(2), 199–237 (1998). MR 1705176 (2000m:53071)
5. Cowling, M., Dooley, A.H., Korányi, A., Ricci, F.: H -type groups and Iwasawa decompositions, Adv.
Math. 87(1), 1–41 (1991). MR1102963 (92e:22017)
6. Damek, E.: Curvature of a semidirect extension of a Heisenberg type nilpotent group. Colloq. Math.
53(2), 249–253 (1987). MR 924069 (89d:22007)
7. Damek, E., Ricci, F.: Harmonic analysis on solvable extensions of H -type groups. J. Geom. Anal. 2(3),
213–248 (1992). MR 1164603 (93d:43006)
8. Gidas B., Ni, W.M., Nirenberg, L.: Symmetry and related properties via the maximum principle. Comm.
Math. Phys. 68(3), 209–243 (1979). MR 544879 (80h:35043)
9. Kaplan, A.: Fundamental solutions for a class of hypoelliptic PDE generated by composition of quadratic
forms Trans. Amer. Math. Soc. 258(1), 147–153 (1980). MR 554324 (81c:58059)
10. Kesavan, S.: On two functionals connected to the Laplacian in a class of doubly connected domains. Proc.
R. Soc. Edinb. Sect. A 133(3), 617–624 (2003). MR 1983689 (2004g:35058)
11. Ramm, A.G., Shivakumar, P.N.: Inequalities for the minimal eigenvalue of the Laplacian in an annulus.
Math. Inequal. Appl. 1(4), 559–563 (1998). MR 1646670 (99f:35149)
12. Sokołowski, J., Zolésio, J.-P.: Introduction to Shape Optimization. Springer Series in Computational
Mathematics, vol. 16. Springer, Berlin. Shape sensitivity analysis (1992). MR 1215733 (94d:49002)

123