Blow-up phenomena for the constant scalar curvature and constant

boundary mean curvature equation

arXiv:1908.04815v1 [math.DG] 13 Aug 2019

Xuezhang Chen* and Nan Wu†

∗†
Department of Mathematics & IMS, Nanjing University, Nanjing 210093, P. R. China

Abstract
We first present a warped product manifold with boundary to show the non-uniqueness of the positive
constant scalar curvature and positive constant boundary mean curvature equation. Next, we construct a
smooth counterexample to show that the compactness of the set of “lower energy” solutions to the above
equation fails when the dimension of the manifold is not less than 62.
2010 MSC: Primary 53C21, 35J20; Secondary 35B33, 34B18.
Keywords: Manifold with boundary, conformal metrics, scalar curvature, boundary mean curvature.

1 Introduction
Let (M n , g) be a smooth compact Riemannian manifold with boundary ∂M and n ≥ 3. In 1999, Zheng-
Chao Han and Yan Yan Li [13] proposed a question of finding a conformal metric with positive constant
scalar curvature and any constant boundary mean curvature in the positive Yamabe constant conformal
class. In analytical terms, it corresponds to the existence of a positive solution to
 n+2
−∆g u + cn Rg u = c1 u n−2 , in M,
∂u n (1.1)
 + dn hg u = cu n−2 , on ∂M,
∂νg

for c1 ∈ R+ , c ∈ R, where cn = (n − 2)/[4(n − 1)] and dn = (n − 2)/2, Rg is the scalar curvature and
hg is the mean curvature of ∂M and νg is the outward unit normal on ∂M . This existence problem has
been studied by Z. C. Han and Y. Y. Li [12, 13], and recently by the first named author and his collaborators
[6, 5]. The closely related works are referred to J. Escobar [10, 9] etc. X. Chen, Y. Ruan an L. Sun [5]
introduced a “free” functional
Z   Z
4(n − 1)
I(M,g) [u] = |∇u|2g + Rg u2 dµg + 2(n − 1) hg u2 dσg
M n−2 ∂M
Z Z
4(n − 1) 2n
n−2
2(n−1)

− c1 u+ dµg − 4c u+n−2 dσg (1.2)

n M ∂M

for any u ∈ H 1 (M, g), where u+ = max{u, 0}. The authors applied the Mountain Pass Lemma to show
the existence of PDE (1.1) for all c ∈ R, in addition that the energy of solutions below a threshold Sc ,
except for the case that n ≥ 8, ∂M is umbilic, the Weyl tensor of M vanishes on ∂M and has an interior
non-zero point. Here the geometric meaning of Sc is the energy I(Rn+ ,|dx|2 ) of a single bubble u(0,1) (x);
see Section 3 or [5]. We will present an example in Section 2 to show the non-uniqueness of PDE (1.1)
* X. Chen is partially supported by NSFC (No.11771204), A Foundation for the Author of National Excellent Doctoral Dissertation
of China (No.201417) and start-up grant of 2016 Deng Feng program B at Nanjing University. Email: [email protected].
† N. Wu: [email protected].

1
when c1 , c ∈ R+ . Indeed, Han-Li [13] established the compactness of the full set of positive solutions to
PDE (1.1) for all c ≤ c̄ with any given positive constant c̄ (see [12, Conjecture 2] and [13, Theorem 0.1]),
provided that M is locally conformally flat with umbilic boundary, and is not conformally equivalent to the
standard hemisphere Sn+ .
Denote by Lg = −∆g +cn Rg the conformal Laplacian and Bg = ∂ν∂ g +dn hg the boundary conformally
covariant operator, respectively. Both Lg and Bg have the following conformally covariant properties: Let
g̃ = u4/(n−2) g, then for any ϕ ∈ C ∞ (M ), there hold
n+2 n
Lg (uϕ) = u n−2 Lg̃ (ϕ) and Bg (uϕ) = u n−2 Bg̃ (ϕ). (1.3)
The Yamabe constant is defined by
R R
M
Rg̃ dµg̃ + 2(n − 1) ∂M hg̃ dσg̃
Y (M, ∂M, [g]) = inf R n−2 .
g̃∈[g] ( M dµg̃ ) n
For the closed manifolds, the question of compactness of the full set of solutions to the Yamabe equation
was initiated by R. Schoen in 1988. A necessary condition is that the manifold is not conformally equivalent
to the standard sphere Sn . It has been extensively studied by R. Schoen [21, 22], Y.Y. Li and M. Zhu [17], O.
Druet [8], F. Marques [20], Y.Y. Li and L. Zhang [18, 19], etc. Eventually, the compactness for dimensions
3 ≤ n ≤ 24 with assuming positive mass theorem was established by M. Khuri, F. Marques and R.
Schoen [15]. For the non-compactness part, S. Brendle [3] discovered the first smooth counterexamples in
dimensions n ≥ 52. S. Brendle and F. Marques [4] extended the above counterexample to the remaining
dimensions 25 ≤ n ≤ 51.
For the manifolds with boundary, the blow-up phenomena in dimensions n ≥ 25 were discovered by
Almaraz [1] corresponding to c1 = 0, c > 0 in (1.1). Such blow-up phenomena in large dimensions also
appear in the Q-curvature equation (see J. Wei and C. Zhao [23] ) and in the fractional Yamabe problem
(see S. Kim, M. Musso and J. Wei [14]).
It is natural to expect that the blow-up phenomena of PDE (1.1) occur in large dimensions. Now we
confirm it in dimensions n ≥ 62 for the set of solutions whose energy of I(M,g) is below Sc .
Theorem 1.1. For n ≥ 62, there exists a smooth Riemannian metric g on Sn+ , such that Y (Sn+ , ∂Sn+ , [g]) >
0 and a sequence of positive smooth functions {vν ; ν ∈ N} with the following properties:
(i) g is not conformally flat;
(ii) ∂Sn+ is umbilic with respect to g;
(iii) for all ν, vν is a positive solution to PDE (1.1) with c1 , c > 0 and M = Sn+ ;
(iv) I(Sn+ ,g) [vν ] < Sc ;
(v) sup∂Sn+ vν → ∞ as ν → ∞.
Since the compactness results of PDE (1.1) are not abundant yet, the critical dimension of the non-
compactness is not a main issue in this paper and thus left to future study.
The paper is arranged as follows. We show the multiplicity of PDE (1.1) on a warped product manifold
with boundary, which is presented in Section 2. In Section 3, we describe how the problem can be reduced
to finding critical points of a certain function Fg (ξ, ǫ), where ξ is a vector in Rn−1 and ǫ is a positive real
number. In Section 4, we show that the function Fg (ξ, ǫ) can be approximated by an auxiliary function
F (ξ, ǫ). In Section 5, we prove that the function F (ξ, ǫ) has a strict local minimum point (0, 1). Finally,
in Section 6, we use a perturbation argument to find critical points of Fg (ξ, ǫ) and then show the non-
compactness.

Acknowledgement: Part of this work was carried out while both authors were visiting the Department of
Mathematics at Rutgers University, to which they are grateful for providing the very stimulating research
environment and supports.

2
2 Non-uniqueness: an example
The purpose of this section is to construct a warped product manifold with boundary, which demonstrates
the multiplicity of solutions of (1.1). This is somewhat inspired by the one for the Yamabe problem in [2,
p.178].
Proposition 2.1. If n ≥ 5 and c1 , c2 > 0, then PDE (1.1) admits at least two positive smooth solutions.
Proof. Suppose (M n1 , g1 ) is an n1 -dimensional (n1 ≥ 3) smooth compact manifold with boundary such
that Rg1 and hg1 are two positive constants, which is guaranteed by [5, Theorem 1.1] for almost all smooth
manifolds with boundary. Let (M n2 , g2 ) be an n2 -dimensional (n2 ≥ 2) smooth closed manifold with
positive constant scalar curvature Rg2 . Consider a warped product manifold (M n , g) = (M n1 ×M n2 , kg1 +
g2 ), where n = n1 + n2 ≥ 5 and k is a positive constant. Obviously, ∂M = ∂M n1 × M n2 and the second
fundamental form π on ∂M satisfies:
1 g1
πij = k 2 πij , πiI = 0, πIJ = 0 1 ≤ i, j ≤ n1 − 1, 1 ≤ I, J ≤ n2 ,
1
which means hg = k − 2 hg1 and Rg = k −1 Rg1 + Rg2 .
Next we claim that the following PDE
 n+2
−∆g u + cn Rg u = cn Rg u n−2 , in M,
∂u n (2.1)
 + dn hg u = dn hg u n−2 , on ∂M,
∂νg

has at least two positive smooth solutions.

To that end, first notice that 1 is a solution of (2.1). On the other hand, it follows from [5, Theorem
1.1] that there exists a positive smooth mountain critical point u0 of I(M,g) , which is the one in (1.2) with
c1 = cn Rg and c = dn hg , such that

I(M,g) [u0 ] < Sc (k),

where we use Sc (k) instead of Sc to emphasize k. If we replace (M, g) by (Rn+ , |dx|2 ) in (2.1) with its
positive solution denoted by
  n−2   n−2
n(n − 1) 4
2 2
hg
W (x) = with Tc = − ,
Rg 1 + |x − Tc en |2 2

which is also called a standard bubble. Furthermore, as k → ∞, by the dominated convergence theorem
we have

Sc (k) =I(Rn+ ,gRn ) [W ]

Z Z Z
4(n − 1) 2 n−2 2n 2(n−1)
= |∇W | dx − Rg W n−2 dx − 2(n − 2)hg W n−2 dσ
n − 2 Rn+ n R+n n
∂R+
Z Z
2 2n 2(n−1)
= Rg W n−2 dx + 2hg W n−2 dσ
n Rn+ ∂Rn+
 n Z  n
2 n(n − 1) 2 2
→ Rg2 dx
n Rg2 Rn+
1 + |x|2
 n  n
2 n(n − 1) 2 ωn n(n − 1) 2 ωn
= Rg2 = Rg2 := Sc (∞).
n Rg2 2 Rg2 n

3
For simplicity, we let

V1 = Vol(M n1 , g1 ), V̂1 = Vol(∂M n1 , g1 ), V2 = Vol(M n2 , g2 ).

Indeed, if k is large enough, then u0 is distinct from 1. This follows from

Z Z
n−2
I(M,g) [1] = Rg dµg + 2(n − 1) hg dσg − Rg Vol(M ) − 2(n − 2)hg Vol(∂M )
M ∂M n
2 n1 1 n1 −1
= (k −1 Rg1 + Rg2 )k 2 V1 V2 + 2k − 2 hg1 k 2 Vˆ1 V2
n
>Sc (∞) + 1 > Sc (k) > I(M,g) [u0 ],

if k is sufficiently large.

3 Lyapunov-Schmidt reduction
From now on, let c1 = n(n − 2) and Tc = −c/(n − 2) be a negative real number for brevity. Given a pair
(ξ, ǫ) ∈ Rn−1 × (0, ∞) we define
  n−2
ǫ 2
u(ξ,ǫ) (x) = ,
ǫ2 + (xn − Tc ǫ)2 + |x′ − ξ|2
where x = (x′ , xn ) ∈ Rn+ . Then u(ξ,ǫ) satisfies
 n+2

−∆u(ξ,ǫ) = n(n − 2)u(ξ,ǫ) n−2
, in Rn+ ,
∂u (ξ,ǫ)
n (3.1)

 n−2
= (n − 2)Tc u(ξ,ǫ) , on ∂Rn+ .
∂xn
4/(n−2)
This implies that the metric u(ξ,ǫ) |dx|2 is Einstein, then there holds

1 
∂a u(ξ,ǫ) ∂b u(ξ,ǫ) − cn ∂a ∂b u2(ξ,ǫ) = |∇u(ξ,ǫ) |2 − cn ∆u2(ξ,ǫ) δab (3.2)
n
for 1 ≤ a, b ≤ n. Define
  n+2
ǫ 2
2ǫ(xi − ξi )
u(ξ,ǫ,i) (x) = ,
ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2 ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2
  n2
ǫ 2ǫ(xi − ξi )
û(ξ,ǫ,i) (x) =
ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2 ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2
for 1 ≤ i ≤ n − 1, and
  n+2
ǫ 2
(1 + Tc2 )ǫ2 − x2n − |x′ − ξ|2
u(ξ,ǫ,n) (x) = ,
ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2 ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2
  n2
ǫ (1 + Tc2 )ǫ2 − x2n − |x′ − ξ|2
û(ξ,ǫ,n) (x) = .
ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2 ǫ2 + (Tc ǫ − xn )2 + |x′ − ξ|2
Obviously, ku(ξ,ǫ,a)k 2n and ku(ξ,ǫ,a)k 2(n−1) are constant in ξ and ǫ for 1 ≤ a ≤ n.
L n+2 (Rn
+) L n (∂Rn
+)
Define ( )
Z
2n 2(n−1)
E= w∈L n−2 (Rn+ ) ∩L n−2 (∂Rn+ ) ∩ 1
Hloc (Rn+ ); 2
|∇w| < ∞
Rn
+

4
and
( Z Z )
E(ξ,ǫ) = w ∈ E; 2n wu(ξ,ǫ,i) − Tc wû(ξ,ǫ,i) = 0 for all 1 ≤ i ≤ n .
Rn
+ ∂Rn
+

R 1/2
We define a norm on E by kwkE = Rn
|∇w|2 . Clearly, u(ξ,ǫ) ∈ E(ξ,ǫ) .
+
It follows from [10, Theorem 3.3] that there exists an optimal constant K = K(n) > 0 such that

Z ! n−2
n Z ! n−2
n−1 Z
2n 2(n−1)
w n−2 + w n−2 ≤K |∇w|2 (3.3)
Rn
+ ∂Rn
+ Rn
+

for all w ∈ E.
Let π : η ∈ Sn 7→ x ∈ Rn+ be the stereographic projection (see [6, Fig.1 on p.9]), which is given by
 2xi

 ηi = , for 1 ≤ i ≤ n − 1,

 1 + (Tc − xn )2 + |x′ |2

 2(xn − Tc )
ηn = 2 ′ 2
,

 1 + (T c − xn ) + |x |

 ′ 2 2
ηn+1 = |x | + (xn − Tc ) − 1 .

1 + (Tc − xn )2 + |x′ |2

Let Σ = π −1 (Rn+ ) be a spherical cap equipped with the standard round metric gΣ = gSn . If we choose the
center of Σ as the north pole, the coordinate system ξ is changed to another coordinate system ζ by


 ζi = ηi , for 1 ≤ i ≤ n − 1,


ζ = − Tcp ηn + ηn+1
,
n
 1 + Tc2

 ηn − Tc ηn+1

ζn+1 = p .
1 + Tc2

Proposition 3.1. There exist a positive constant µ depending only on n and Tc , such that
"Z   Z #
1 2
4
n−2 2
2
n−2 2
|∇w| − n(n + 2)u(ξ,ǫ) w + nTc u(ξ,ǫ) w
2 Rn+ ∂Rn
+
" Z Z #2
(1 + (n − 2)µ)2n−1 n+2 n
+ 2n n−2
u(ξ,ǫ) w − Tc u(ξ,ǫ) w ≥ µkwk2E
n−2
(3.4)
n|Σ| − Tc |∂Σ| Rn
+ ∂R n
+

for all w ∈ E(ξ,ǫ) .

Proof. Given a function Φ defined on Σ, we define
n−2
φ(x) = 2 2 Φ ◦ π −1 (x)u(ξ,ǫ) (x), x ∈ Rn+ .

It follows from [12, Proposition 3.4] that there exists a positive constant µ depending on n and Tc , such that
Z Z 
1 2 2 2
Q2 (Φ0 ) := (|∇Φ0 |gΣ − nΦ0 ) + Tc Φ0
2 Σ ∂Σ
 Z Z 2
1
+ n Φ0 − T c Φ0
2(n|Σ| − Tc |∂Σ|) Σ ∂Σ

5
 Z Z 
n(n − 2) 2 n − 2
≥µ |∇Φ0 |2gΣ + Φ0 − Tc Φ20 (3.5)
Σ 4 2 ∂Σ

for all Φ0 ∈ span{1, ζ1 , ζ2 , · · · , ζn }⊥ , which denotes the orthogonal complement of span{1, ζ1 , ζ2 , · · · , ζn }

in H 1 (Σ). R
1
For any Φ ∈ span{ζ1 , ζ2 , · · · , ζn }⊥ , we set Φ = Φ0 + Φ̄, where Φ̄ = |Σ| Σ
Φ. Then by (3.5) we have
Z Z   Z Z 2
1 1 + (n − 2)µ
(|∇Φ|2gΣ − nΦ2 ) + Tc Φ2 + n Φ − Tc Φ
2 Σ ∂Σ 2(n|Σ| − Tc |∂Σ|) Σ ∂Σ
Z   Z 
2 n(n − 2) 2 n−2 2
−µ |∇Φ|gΣ + Φ − Tc Φ
Σ 4 2 ∂Σ
Z Z  Z
1 1
= (|∇Φ0 |2gΣ − nΦ20 ) + Tc Φ20 − (n|Σ| − Tc |∂Σ|)Φ̄2 + Tc Φ̄ Φ0
2 Σ ∂Σ 2 ∂Σ
 Z 2
1 + (n − 2)µ
+ (n|Σ| − Tc |∂Σ|)Φ̄ − Tc Φ0
2(n|Σ| − Tc |∂Σ|) ∂Σ
Z Z 
2 n(n − 2) 2 n − 2 2
−µ |∇Φ0 |gΣ + Φ0 − Tc Φ0
4 2
 Σ Z
∂Σ

n(n − 2) 2 n−2 2
− −(n − 2)Tc Φ̄ Φ0 + Φ̄ |Σ| − Tc Φ̄ |∂Σ|
∂Σ 4 2
Z Z   Z 2
1 1 + (n − 2)µ
= (|∇Φ0 |2gΣ − nΦ20 ) + Tc Φ20 + Tc Φ0
2 Σ ∂Σ 2(n|Σ| − Tc |∂Σ|) ∂Σ
Z Z 
2 n(n − 2) 2 n − 2 2 n(n − 2) 2
−µ |∇Φ0 |gΣ + Φ0 − Tc Φ0 + µΦ̄ |Σ| ≥ 0.
Σ 4 2 ∂Σ 4
By (1.3) we have
Z Z Z
∂φ
|∇φ|2 = − φ∆φ + φ
Rn
+ Rn
+ ∂Rn
+
∂ν
Z Z
= φL|dx|2 (φ) + φB|dx|2 (φ)
Rn
+ ∂Rn
+
Z Z
n(n − 2) 2 n − 2
= |∇Φ|2gΣ + Φ − Tc Φ2
Σ 4 2 ∂Σ

and
Z Z Z Z
4
2 1 2 1 2
2
n−2
u(ξ,ǫ) φ = Φ , u(ξ,ǫ) φ = n−2
Φ2 ,
n
R+ 4 Σ ∂Rn
+
2 ∂Σ
Z n+2
Z Z n
Z
n+2 n
n−2
u(ξ,ǫ) φ = 2− 2 Φ, n−2
u(ξ,ǫ) φ = 2− 2 Φ.
Rn
+ Σ ∂Rn
+ ∂Σ

Therefore, we combine these facts together to obtain the desired estimate.

Proposition 3.2. Consider a Riemannian metric in Rn+ of the form g(x) = exp(h(x)), where h(x) is a
trace-free symmetric two-tensor in Rn+ satisfying |h(x)| + |∂h(x)| + |∂ 2 h(x)| ≤ α ≤ 1, and hna (x) = 0
for all x ∈ Rn+ and h(x) = 0 for all x ∈ Rn+ \ B1+ (0). Then there exists a constant C, depending only on
n and Tc , such that
n+2
n−2
k∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ) k 2n + khg u(ξ,ǫ) k 2(n−1) ≤ Cα.
L n+2 (Rn
+) L n (∂Rn
+)

6
Proof. It is not hard to verify that hg = O(|h|), which together with [1, Proposition 2.3] yields the desired
estimate.
Proposition 3.3. Consider a Riemannian metric in Rn+ of the form g(x) = exp(h(x)), where h(x) is a
trace-free symmetric two-tensor in Rn+ satisfying |h(x)| + |∂h(x)| + |∂ 2 h(x)| ≤ α ≤ 1 and hna (x) = 0
for all x ∈ Rn+ and h(x) = 0 for all x ∈ Rn+ \ B1+ (0). Here α depends only on n and Tc . Then, given any
2n 2(n−1)
pair (ξ, ǫ) ∈ Rn−1 × (0, ∞) and any functions f ∈ L n+2 (Rn+ ), fˆ ∈ L n (∂Rn+ ), there exists a unique
function w := G(ξ, ǫ)(f, fˆ) ∈ E(ξ,ǫ) , such that
Z  4
 Z  2

n−2 n−2
h∇w, ∇ϕig + cn Rg wϕ − n(n + 2)u(ξ,ǫ) wϕ + dn hg ϕ + nTc u(ξ,ǫ) ϕ w
Rn
+ ∂Rn
+
Z Z
= fϕ + fˆϕ (3.6)
Rn
+ ∂Rn
+

for all ϕ ∈ E(ξ,ǫ) . Furthermore, there holds

kG(ξ, ǫ)(f, fˆ)kE ≤ Ckf k 2n + Ckfˆk 2(n−1) . (3.7)

L n+2 (Rn
+) L n (∂Rn
+)

Proof. By Propositions 3.1 and 3.2, and Hölder’s inequality, we can follow nearly the same lines in [3,
Corollary 3] that there exist two positive constants α and C, depending only on n and Tc , such that |h(x)| +
|∂h(x)| + |∂ 2 h(x)| ≤ α and
Z  
µ 4
kwk2E ≤ |∇w|2g + cn Rg w2 − n(n + 2)u(ξ,ǫ)
n−2
w2
2 Rn+
Z  
2
C
+ 2 n−2 2
dn hg w + nTc u(ξ,ǫ) w + A(w)2 (3.8)
n
∂R+ µ

for all w ∈ E(ξ,ǫ) , where

Z  n+2

n−2
A(w) = ∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n + 2)u(ξ,ǫ) w
Rn
+
Z  n 
n−2
− dn hg u(ξ,ǫ) + 2Tc u(ξ,ǫ) w.
∂Rn
+

Suppose that w ∈ E(ξ,ǫ) satisfies (3.6), then

Z  4
 Z  2

2 2 n−2 2 2 n−2 2
|∇w|g + cn Rg w − n(n + 2)u(ξ,ǫ) w + dn hg w + nTc u(ξ,ǫ) w
Rn
+ ∂Rn
+
Z Z
= fw + fˆw. (3.9)
Rn
+ ∂Rn
+

Since u(ξ,ǫ) ∈ E(ξ,ǫ) , we have

Z  n+2

n−2
∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n + 2)u(ξ,ǫ) w
Rn
+
Z  n 
Z Z
− n−2
dn hg u(ξ,ǫ) + 2Tc u(ξ,ǫ) w=− f u(ξ,ǫ) − fˆu(ξ,ǫ) .
∂Rn
+ Rn
+ ∂Rn
+

7
Then by (3.3), (3.8) and (3.9), we have
µ
kwk2E
2
Z Z
C
≤ fw + fˆw + A(w)2
Rn
+ ∂Rn
+
µ
≤kf k 2n kwk 2n + kfˆk 2(n−1) kwk 2(n−1)
L n+2 (Rn
+) L n−2 (Rn
+) L (∂Rn
+) L (∂Rn
+)
n n−2
!
C
+ kf k 2
2n ku(ξ,ǫ) k 2
2n + kfˆk2 2(n−1) ku(ξ,ǫ) k 2
2(n−1)
µ L n+2 (Rn
+) L n−2 (Rn
+) L n (∂Rn
+) L n−2 (∂Rn
+)
   
1 C
≤K kwkE kf k
2 2n + kfˆk 2(n−1) + 2 ˆ 2
kf k 2n n + kf k 2(n−1) .
L n+2 (Rn
+) L n (∂Rn
+) µ L n+2 (R+ ) L n (∂Rn
+)

Hence it follows from Young’s inequality that

µ
kwk2
4 E   
2K C
≤ kf k2 2n n + kfˆk2 2(n−1) + kf k2 2n n + kfˆk2 2(n−1) .
µ L n+2 (R+ ) L n (∂Rn +)
µ L n+2 (R+ ) L n (∂Rn
+)

This implies the uniqueness of the solutions to (3.6).

For the existence part, thanks to the coercive estimate (3.8), it suffices to minimize the following func-
tional
Z  4

2 2 n−2 2
|∇w|g + cn Rg w − n(n + 2)u(ξ,ǫ) w − 2f w
Rn
+
Z  
2
C
+ dn hg w2 + nTc u(ξ,ǫ)
n−2
w2 − 2fˆw + A(w)2
∂Rn
+
µ

over all w ∈ E(ξ,ǫ) .

Proposition 3.4. Consider a Riemannian metric in Rn+ of the form g(x) = exp(h(x)), where h(x) is a
trace-free symmetric two-tensor in Rn+ satisfying |h(x)| + |∂h(x)| + |∂ 2 h(x)| ≤ α ≤ 1 and hna (x) = 0
for all x ∈ Rn+ and h(x) = 0 for all x ∈ Rn+ \ B1+ (0). Here α depends only on n and Tc . Then, given any
pair (ξ, ǫ) ∈ Rn−1 × (0, ∞), there exists a unique function v(ξ,ǫ) ∈ E such that v(ξ,ǫ) − u(ξ,ǫ) ∈ E(ξ,ǫ) , and
Z  
4
h∇v(ξ,ǫ) , ∇ϕig + cn Rg v(ξ,ǫ) ϕ − n(n − 2)|v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ
Rn
+
Z  
2
+ dn hg v(ξ,ǫ) + (n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ = 0 (3.10)
∂Rn
+

for all ϕ ∈ E(ξ,ǫ) . Moreover, there exists a positive constant C, depending only on Tc and n, such that

kv(ξ,ǫ) − u(ξ,ǫ) kE
n+2
n−2
≤Ck∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ) k 2n + Ckhg u(ξ,ǫ) k 2(n−1) . (3.11)
L n+2 (Rn
+) L n (∂Rn
+)

In particular, v(ξ,ǫ) 6≡ 0 if α is sufficiently small.

8
 2n 2(n−1)
Proof. Let G(ξ,ǫ) : L n+2 (Rn+ ) × L n (∂Rn+ ) → E(ξ,ǫ) be the solution operator constructed in Proposi-
tion 3.3, and we define a nonlinear operator Φ(ξ,ǫ) on E(ξ,ǫ) by

Φ(ξ,ǫ) (w)
 
n−2 n+2
n−2 n−2
=G(ξ,ǫ) ∆g u(ξ,ǫ) − Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ) ,− hg u(ξ,ǫ) +
4(n − 1) 2
 n+2 4
4
n−2 n−2
G(ξ,ǫ) n(n − 2)|u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) − n(n − 2)u(ξ,ǫ) − n(n + 2)u(ξ,ǫ) w,
n 2

2
n−2 n−2
−(n − 2)Tc |u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) + (n − 2)Tc u(ξ,ǫ) + nTc u(ξ,ǫ) w .

In particular, it follows from Propositions 3.2 and 3.3 that kΦ(ξ,ǫ) (0)kE ≤ Cα.
Using the pointwise estimates
 
 4 4 n + 2 n−2
4 
|u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) − |u(ξ,ǫ) + w|
e n−2 (u(ξ,ǫ) + w)
e − u e 
(w − w)
 n − 2 (ξ,ǫ)
4 4
≤C(|w| n−2 + |w|
e n−2 )|w − w|
e

and
 
 2 2 n 2 
|u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) − |u(ξ,ǫ) + w|
e n−2 (u(ξ,ǫ) + w)
e − n−2
e 
u(ξ,ǫ) (w − w)
 n−2
2 2
≤C(|w| n−2 + |w|
e n−2 )|w − w|
e

and Proposition 3.3, we obtain

kΦ(ξ,ǫ) (w) − Φ(ξ,ǫ) (w)ke E

4 4
≤C |u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) − |u(ξ,ǫ) + w| e n−2 (u(ξ,ǫ) + w)
e

n + 2 n−24
− u(ξ,ǫ) (w − w)e n+2
n−2 L
2n
(Rn
+)

2 2
+ C |u(ξ,ǫ) + w| n−2 (u(ξ,ǫ) + w) − |u(ξ,ǫ) + w| e n−2 (u(ξ,ǫ) + w)
e

n 2
− n−2
u(ξ,ǫ) e
(w − w) 2(n−1) n
n−2 L n (∂R+ )
 4 4

≤C kwk n−2 2n + kwk e n−2
2n kw − wke E
L n−2 (Rn
+) L n−2 (Rn
+)
!
2 2
n−2 n−2
+C kwk 2(n−1) + kwk
e 2(n−1) kw − wk
e E
L n−2 (∂Rn
+) L n−2 (∂Rn
+)

for w, w
e ∈ E(ξ,ǫ) . Hence, if α is sufficiently small, then the contraction mapping principle implies that
Φ(ξ,ǫ) has a unique fixed point w(ξ,ǫ) within E(ξ,ǫ) . Hence v(ξ,ǫ) = u(ξ,ǫ) + w(ξ,ǫ) is the desired solution,
and not identically zero, which follows from (3.11) and Proposition 3.2.
Given a pair (ξ, ǫ) ∈ Rn−1 × (0, ∞), we define the following energy functional

Fg (ξ, ǫ)
Z   Z
2n
:= |∇v(ξ,ǫ) |2g + cn Rg v(ξ,ǫ)
2
− (n − 2)2 |v(ξ,ǫ) | n−2 + dn 2
hg v(ξ,ǫ)
Rn
+ ∂Rn
+

9
 Z Z Z
(n − 2)2 2(n−1) n−2
2n
n−2
2(n−1)
n−2
+ Tc |v(ξ,ǫ) | n−2 − 2(n − 2) u(ξ,ǫ) + Tc u(ξ,ǫ) . (3.12)
n−1 ∂Rn
+
n
R+ n−1 ∂Rn
+

Proposition 3.5. The function Fg is continuously differentiable. Moreover, if (ξ, ǫ) is a critical point of Fg ,
then the function v(ξ,ǫ) is a positive smooth solution of
 n+2

−∆g v(ξ,ǫ) + cn Rg v(ξ,ǫ) = n(n − 2)v n−2 , in Rn+ ,
(ξ,ǫ)
∂v(ξ,ǫ) n (3.13)

 n−2
− dn hg v(ξ,ǫ) = (n − 2)Tc v(ξ,ǫ) , on ∂Rn+ .
∂xn
Proof. By definition of v(ξ,ǫ) , we can find real numbers ba = ba (ξ, ǫ), 1 ≤ a ≤ n, such that
Z  
4
h∇v(ξ,ǫ) , ∇ϕig + cn Rg v(ξ,ǫ) ϕ − n(n − 2)|v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ
Rn
+
Z  
2
+ dn hg v(ξ,ǫ) ϕ + (n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ
∂Rn
+
n Z Z !
X
= ba (ξ, ǫ) 2n ϕu(ξ,ǫ,a) − Tc ϕû(ξ,ǫ,a)
a=1 Rn
+ ∂Rn
+

for all test function ϕ ∈ E. This implies

n Z Z !
X
∂ǫ Fg = 2 ba (ξ, ǫ) 2n ∂ǫ v(ξ,ǫ) u(ξ,ǫ,a) − Tc ∂ǫ v(ξ,ǫ) û(ξ,ǫ,a)
a=1 Rn
+ ∂Rn
+

and
n Z Z !
X
∂ξj Fg = 2 ba (ξ, ǫ) 2n ∂ξj v( ξ, ǫ)u(ξ,ǫ,a) − Tc ∂ξj v( ξ, ǫ)û(ξ,ǫ,a)
a=1 Rn
+ ∂Rn
+

for 1 ≤ j ≤ n − 1. On the other hand, we have

Z Z
0 = 2n (v(ξ,ǫ) − u(ξ,ǫ) )u(ξ,ǫ,a) − Tc (v(ξ,ǫ) − u(ξ,ǫ) )û(ξ,ǫ,a) ,
Rn
+ ∂Rn
+

since v(ξ,ǫ) − u(ξ,ǫ) ∈ E(ξ,ǫ) . Differentiating the above equation with respect to ǫ and ξk , we obtain
Z Z
0 =2n ∂ǫ (v(ξ,ǫ) − u(ξ,ǫ) )u(ξ,ǫ,a) − Tc ∂ǫ (v(ξ,ǫ) − u(ξ,ǫ) )û(ξ,ǫ,a)
Rn
+ ∂Rn
+
Z Z
+ 2n (v(ξ,ǫ) − u(ξ,ǫ) )∂ǫ u(ξ,ǫ,a) − Tc (v(ξ,ǫ) − u(ξ,ǫ) )∂ǫ û(ξ,ǫ,a)
Rn
+ ∂Rn
+
Z Z
=2n (v(ξ,ǫ) − u(ξ,ǫ) )∂ǫ u(ξ,ǫ,a) − Tc (v(ξ,ǫ) − u(ξ,ǫ) )∂ǫ û(ξ,ǫ,a)
Rn
+ ∂Rn
+
Z Z
+ 2n u(ξ,ǫ,a)∂ǫ v(ξ,ǫ) − Tc û(ξ,ǫ,a) ∂ǫ v(ξ,ǫ) + c̄n (2ǫ)−1 δna ,
Rn
+ ∂Rn
+

where 1 ≤ a ≤ n, c̄n is a nonzero constant independent of ξ and ǫ, and

Z Z
0 =2n (v(ξ,ǫ) − u(ξ,ǫ) )∂ξk u(ξ,ǫ,a) − Tc (v(ξ,ǫ) − u(ξ,ǫ) )∂ξk û(ξ,ǫ,a)
Rn
+ ∂Rn
+

10
 Z Z
+ 2n ∂ξk v(ξ,ǫ) u(ξ,ǫ,a) − Tc ∂ξk v(ξ,ǫ) û(ξ,ǫ,a) − c̄a (2ǫ)−1 δak ,
Rn
+ ∂Rn
+

where each c̄i , 1 ≤ i ≤ n − 1 is a nonzero constant independent of ξ and ǫ.

Therefore, putting these facts together, we conclude that
− c̄n bn (ξ, ǫ) = ǫ∂ǫ Fg +
n
" Z Z #
X
2ǫ ba (ξ, ǫ) 2n ∂ǫ u(ξ,ǫ,a) (v(ξ,ǫ) − u(ξ,ǫ) ) − Tc ∂ǫ û(ξ,ǫ,a) (v(ξ,ǫ) − u(ξ,ǫ) )
a=1 Rn
+ ∂Rn
+

and for 1 ≤ i ≤ n − 1,
c̄i ai (ξ, ǫ) = ǫ∂ξi Fg +
n
" Z Z #
X
2ǫ ba (ξ, ǫ) 2n ∂ξi u(ξ,ǫ,a) (v(ξ,ǫ) − u(ξ,ǫ) ) − Tc ∂ξi û(ξ,ǫ,a) (v(ξ,ǫ) − u(ξ,ǫ) ) .
a=1 Rn
+ ∂Rn
+

Hence, if (ξ, ǫ) is a critical point of Fg , then

n
X
|c̄a ||ba (ξ, ǫ)|
a=1
 X
n
≤C kv(ξ,ǫ) − u(ξ,ǫ) k 2n + kv(ξ,ǫ) − u(ξ,ǫ) k 2(n−1) |ba (ξ, ǫ)|.
L n−2 (Rn
+) L n−2 (∂Rn
+) a=1

By (3.3) and (3.11) we have

kv(ξ,ǫ) − u(ξ,ǫ) k 2n + kv(ξ,ǫ) − u(ξ,ǫ) k 2(n−1)
L n−2 (Rn
+) L (∂Rn
+)
n−2

≤Ckv(ξ,ǫ) − u(ξ,ǫ) kE ≤ Cα.

Thus, if α is sufficiently small, we obtain
n
X
|ba (ξ, ǫ)| = 0.
a=1

Consequently, we have
Z  
4
h∇v(ξ,ǫ) , ∇ϕig + cn Rg v(ξ,ǫ) ϕ − n(n − 2)|v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ
Rn
+
Z  
2
+ dn hg v(ξ,ǫ) ϕ + (n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ = 0
∂Rn
+

for all ϕ ∈ E.
Finally, we follow the same lines in [3, Proposition 6] that v(ξ,ǫ) ≥ 0 in Rn+ . Together with v(ξ,ǫ) ≡
6 0
by Proposition 3.4, the strong maximum principle and the Hopf boundary point lemma give v(ξ,ǫ) > 0 in
Rn+ . By the regularity theory of P. Cherrier [7], we show that v(ξ,ǫ) is smooth.

4 An estimate for the energy of a bubble

We first introduce a multi-linear form W : Rn−1 ×Rn−1 ×Rn−1 ×Rn−1 → R satisfying the same algebraic
properties of the Weyl tensor on ∂Rn+ . Moreover, we assume
n−1
X
(W ikjl + W iljk )2 > 0.
i,j,k,l=1

11
If x = (x′ , xn ) ∈ Rn+ , then we identify x′ with (x′ , 0) ∈ ∂Rn+ and define
Hij (x) = Hij (x′ ) = W ikjl xk xl and Hna (x) = 0, (4.1)
as well as H ab (x) = f (|x′ |2 )Hab (x), where f (s) is a polynomial of degree d for 0 ≤ d < (n − 6)/4 and
is to be determined later. Then H is symmetric, trace-free, independent of the variable xn , and satisfies
xa Hab (x) = xi Hib (x) = ∂a Hab (x) = ∂i Hib (x) = 0.
We define a Riemannian metric g = exp(h) in Rn+ , where h is a trace-free symmetric two-tensor in Rn+ and
hna = ∂n hab (x) = 0 for all x ∈ Rn+ , and satisfies
(
hab (x) = µλ2d f (λ−2 |x′ |2 )Hab (x), if |x| ≤ ρ,
hab (x) = 0, if |x| ≥ 1.

Here 0 < µ ≤ 1, 0 < λ ≤ ρ ≤ 1. This gives hab (x) = O(µ(λ + |x|)2d+2 ). In addition, we require that
|h| + |∂h| + |∂ 2 h| ≤ α, where α is the constant given in Proposition 3.4. The boundary ∂Rn+ is totally
geodesic with respect to g, since the second fundamental form vanishes on ∂Rn+ , explicitly
 
n 1 ∂gin ∂gjn ∂gij
πij = Γij = + − = 0.
2 ∂xj ∂xi ∂xn
Applying Proposition 3.4 to each pair (ξ, ǫ) ∈ Rn−1 × (0, ∞), we choose v(ξ,ǫ) to be the unique element
of E such that v(ξ,ǫ) − u(ξ,ǫ) ∈ E(ξ,ǫ) and
Z  
4
h∇v(ξ,ǫ) , ∇ϕig + cn Rg v(ξ,ǫ) ϕ − n(n − 2)|v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ
Rn
+
Z
2
+ (n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) ϕ = 0
∂Rn
+

for all ϕ ∈ E(ξ,ǫ)

 .
Let Ω = (ξ, ǫ) ∈ Rn−1 × (0, ∞); |ξ| < 1, 12 < ǫ < 1 . Similar to [3, Proposition 7 and Corollary 8]
and [4, Proposition 5 and Corollary 6], for any pair (ξ, ǫ) ∈ λΩ we obtain
n+2
n−2 2−n
n−2
k∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ) k 2n ≤ Cµλ2d+2 + Cλ 2 ρ 2 , (4.2)
L n+2 (Rn
+)

n+2
n−2
k∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ) + µλ2d f (λ−2 |x′ |2 )Hij (x)∂i ∂j u(ξ,ǫ) k 2n
L n+2 (Rn
+)
(2d+2)(n+2) n−2 2−n
≤Cµ2 λ n−2 + Cλ 2 ρ 2 (4.3)
and together with Proposition 3.4,
kv(ξ,ǫ) − u(ξ,ǫ) k 2n + kv(ξ,ǫ) − u(ξ,ǫ) k 2(n−1)
L n−2 (Rn
+) L (∂Rn
+)
n−2

n−2 2−n
≤Cµλ2d+2 + Cλ 2 ρ 2 . (4.4)
By Proposition 3.3 with h = 0, we define the function w(ξ,ǫ) as the unique element of E(ξ,ǫ) satisfying
Z  4
 Z 2
n−2 n−2
h∇w(ξ,ǫ) , ∇ϕi − n(n + 2)u(ξ,ǫ) w(ξ,ǫ) ϕ + nTc u(ξ,ǫ) w(ξ,ǫ) ϕ
Rn
+ ∂Rn
+
Z
=− µλ2d f (λ−2 |x′ |2 )Hij (x)∂i ∂j u(ξ,ǫ) ϕ (4.5)
Rn
+

for all ϕ ∈ E(ξ,ǫ) . In particular, w(0,ǫ) = 0, since xi Hij (x) = 0 for any x ∈ Rn+ .

12
Proposition 4.1. The function w(ξ,ǫ) is smooth and satisfies that given any (ξ, ǫ) ∈ λΩ, |∂ k w(ξ,ǫ) (x)| ≤
n−2
Cλ 2 µ(λ + |x|)2d+4−k−n for all x ∈ Rn+ , k = 0, 1, 2.
Proof. By definition of E(ξ,ǫ) , there exist real numbers b̄a (ξ, ǫ), 1 ≤ a ≤ n such that
Z  4
 Z 2
n−2 n−2
h∇w(ξ,ǫ) , ∇φi − n(n + 2)u(ξ,ǫ) w(ξ,ǫ) φ + nTc u(ξ,ǫ) w(ξ,ǫ) φ
Rn
+ ∂Rn
+
Z
=− µλ2d f (λ−2 |x′ |2 )Hij (x)∂i ∂j u(ξ,ǫ) (x)φ
Rn
+
n Z Z !
X
+ b̄a (ξ, ǫ) 2n u(ξ,ǫ,a) φ − Tc û(ξ,ǫ,a) φ (4.6)
a=1 Rn
+ ∂Rn
+

for all φ ∈ E. Hence it follows from standard elliptic estimates that w(ξ,ǫ) is smooth. Since

kµλ2d f (λ−2 |x′ |2 )Hij (x)∂i ∂j u(ξ,ǫ) (x)k 2n ≤ Cµλ2d+2 ,

L n+2 (Rn
+)

then by (3.3) and (3.7) we have

kw(ξ,ǫ) k 2n + kw(ξ,ǫ) k 2(n−1) ≤ Ckw(ξ,ǫ) kE ≤ Cµλ2d+2 .

L n−2 (Rn
+) L (∂Rn
+)
n−2

Choosing φ = u(ξ,ǫ,a) in (4.6), we obtain

n
X
|b̄a (ξ, ǫ)| ≤ Cµλ2d+2 .
a=1

Hence, we have
 
 4 
∆w(ξ,ǫ) + n(n + 2)u n−2 w(ξ,ǫ) 
 (ξ,ǫ) 
 2d 
= µλ f (λ−2 |x′ |2 )Hij (x)∂i ∂j u(ξ,ǫ) (x) − 2nb̄a (ξ, ǫ)u(ξ,ǫ,a) 
n−2
≤Cµλ 2 (λ + |x|)2d+2−n

for all x ∈ Rn+ , and

   n


 ∂ 2   X  n
 w + nT u n−2
w 
c (ξ,ǫ) (ξ,ǫ)  =  −T c b̄ a (ξ, ǫ)û (ξ,ǫ,a)  ≤ Cµλ 2 (λ + |x|)
2d+2−n
 ∂xn (ξ,ǫ)  
a=1

for all x ∈ ∂Rn+ . We let r = (λ + |x0 |)/2 for any fixed x0 ∈ Rn+ . Then λ + |x| ≥ r for any x ∈ Br+ (x0 ).
Based on the above facts, we obtain
2
n−2
u(ξ,ǫ) (x) ≤Cr−1 , ∀ x ∈ Br+ (x0 ),
 
 ∂ 2  n
 n−2  2 2d+2−n , ∀ x ∈ Br+ (x0 ) ∩ ∂Rn+ ,
 ∂xn w(ξ,ǫ) − nTc u(ξ,ǫ) w(ξ,ǫ)  ≤Cµλ r
 
 4 
∆w(ξ,ǫ) + n(n + 2)u n−2 w(ξ,ǫ)  ≤Cµλ n−2 2 r 2d+2−n , ∀ x ∈ Br+ (x0 ).
 (ξ,ǫ) 

By [?, Theorems 8.25 and 8.26]e have

n−2
r 2 |w(ξ,ǫ) (x0 )|

13

n+2 4
≤Ckw(ξ,ǫ) k 2n + Cr 2 ∆w(ξ,ǫ) + n(n + 2)u n−2 w(ξ,ǫ)
L n−2 (Br+ (x0 )) (ξ,ǫ)
L∞ (Br+ (x0 ))

n ∂
2

+ Cr 2 w(ξ,ǫ) + nTc u(ξ,ǫ) w(ξ,ǫ)
n−2

∂xn
L∞ (Br+ (x0 )∩∂Rn
+)
n−2
2d+2+ 2−n n n
≤Cµλ2d+2 + Cµλ 2 r 2 + Cµλ 2 r2d+2− 2
≤Cµλ2d+2 ,

since d < (n − 6)/4. Then we obtain

n−2
sup (λ + |x|) 2 |w(ξ,ǫ) (x)| ≤ Cµλ2d+2 .
x∈Rn
+

By Green’s representation formula, we have

w(ξ,ǫ) (x)
Z
1
=− (|x − y|2−n + |x∗ − y|2−n )∆w(ξ,ǫ) (y)dy
(n − 2)|Sn−1 | Rn
+
Z
1 ∂
− (|x − y|2−n + |x∗ − y|2−n ) w(ξ,ǫ) (y)dy,
(n − 2)|Sn−1 | ∂Rn
+
∂yn

for any x ∈ Rn+ , where x∗ = (x1 , x2 , · · · , xn−1 , −xn ). From this we obtain

sup (λ + |x|)β |w(ξ,ǫ) (x)|

x∈Rn
+
 
 ∂ 
≤ C sup (λ + |x|) β+2
|∆w(ξ,ǫ) (x)| + C sup (λ + |x|) β+1  
x∈Rn x∈∂Rn
 ∂xn w(ξ,ǫ) (x)
+ +

for all 0 < β < n − 2. Since

4 n−2
n−2
|∆w(ξ,ǫ) (x)| ≤ n(n + 2)u(ξ,ǫ) (x)|w(ξ,ǫ) (x)| + Cµλ 2 (λ + |x|)2d+2−n , ∀ x ∈ Rn+ ,

and
 
 ∂  2
n
  n−2 2d+2−n
, ∀ x ∈ ∂Rn+ ,
 ∂xn (ξ,ǫ)  ≤ n|Tc |u(ξ,ǫ) (x)|w(ξ,ǫ) (x)| + Cµλ (λ + |x|)
w (x) 2

we conclude that

sup (λ + |x|)β |w(ξ,ǫ) (x)|

x∈Rn
+
n
≤Cλ2 sup (λ + |x|)β−2 |w(ξ,ǫ) (x)| + Cλ sup (λ + |x|)β−1 |w(ξ,ǫ) (x)| + Cµλβ+2d+3− 2
x∈Rn
+ x∈∂Rn
+

for all 0 < β ≤ n − 4 − 2d. Iterating this inequality, we obtain

n−2
sup (λ + |x|)n−2d−4 |w(ξ,ǫ) (x)| ≤ Cµλ 2 .
x∈Rn
+

Differentiating the equation (4.5) twice and repeating the argument above, we obtain the estimates of the
first and second derivatives of w(ξ,ǫ) .

14
Proposition 4.2. There holds

kv(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) k 2n + kv(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) k 2(n−1)

L n−2 (Rn
+) L (∂Rn
+)
n−2

n (2d+2)n n−2 2−n

≤Cµ n−2 λ n−2 + Cλ 2 ρ 2

for all (ξ, ǫ) ∈ λΩ.

Proof. Consider the functions
n
X
B1 = ∂i [(g ik − δik )∂k w(ξ,ǫ) ] − cn Rg w(ξ,ǫ)
i,k=1

and
n
X
B2 = µλ2d f (λ−2 |x|2 )Hik (x)∂i ∂k u(ξ,ǫ) .
i,k=1

By definition of w(ξ,ǫ) , we have

Z  4
 Z 2
n−2 n−2
h∇w(ξ,ǫ) , ∇ϕig + cn Rg w(ξ,ǫ) φ − n(n + 2)u(ξ,ǫ) w(ξ,ǫ) ϕ + nTc u(ξ,ǫ) w(ξ,ǫ) ϕ
Rn
+ ∂Rn
+
Z
=− (B1 + B2 )ϕ
Rn
+

for all ϕ ∈ E(ξ,ǫ) . Since w(ξ,ǫ) ∈ E(ξ,ǫ) , we obtain

w(ξ,ǫ) = −G(ξ,ǫ) (B1 + B2 , 0).

By definitions of v(ξ,ǫ) and u(ξ,ǫ) we have

v(ξ,ǫ) − u(ξ,ǫ) = G(ξ,ǫ) (B3 + B4 , A1 ),

where
n+2
n−2
B3 =∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ)
n+2 4
4
n−2 n−2
B4 =n(n − 2)|v(ξ,ǫ) | n−2 v(ξ,ǫ) − n(n − 2)u(ξ,ǫ) − n(n + 2)u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) ),

and n 2
2
n−2 n−2
A1 = −(n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) + (n − 2)Tc u(ξ,ǫ) + nTc u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) ).
Thus, we obtain

v(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) = G(ξ,ǫ) (B1 + B2 + B3 + B4 , A1 ).

It follows from (3.11) that

kv(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) kE

≤CkB1 + B2 + B3 + B4 k 2n + CkA1 k 2(n−1) .
L n+2 (Rn
+) L n (∂Rn
+)

Following the same lines in [4, Corollary 8] and [3, Proposition 7], together with Proposition 4.1 and (4.3)
we obtain
(2d+2)(n+2) n−2 2−n
kB1 k 2n ≤Cλ n−2 µ2 + Cρ2d+2 µλ 2 ρ 2 ,
L n+2 (Rn
+)

15
 (2d+2)(n+2) n−2 2−n
kB2 + B3 k 2n ≤Cλ n−2 µ2 + Cλ 2 ρ 2
L n+2 (Rn
+)

and by (4.2), (4.4),

n+2
kB4 k 2n ≤kv(ξ,ǫ) − u(ξ,ǫ) k n−2
2n
L n+2 (Rn
+) L n−2 (Rn
+)
(2d+2)(n+2) n+2 n+2 −2−n
≤Cλ n−2 µ n−2 + Cλ 2 ρ 2 ,
n
n−2
kA1 k 2(n−1) ≤Ckv(ξ,ǫ) − u(ξ,ǫ) k 2(n−1)
L n (∂Rn
+) L n−2 (∂Rn
+)
(2d+2)n n n −n
≤Cλ n−2 µ n−2 + Cλ 2 ρ 2 .
Therefore, putting these facts together, we obtain the desired estimate.
Proposition 4.3. There holds
Z
  

 |∇v(ξ,ǫ) |2g − |∇u(ξ,ǫ) |2g + cn Rg (v(ξ,ǫ)
2
− u2(ξ,ǫ) )
 Rn+
Z  4 2n

4 2n
n−2 n−2
+ n(n − 2) (|v(ξ,ǫ) | n−2 − u(ξ,ǫ) )u(ξ,ǫ) v(ξ,ǫ) − (|v(ξ,ǫ) | n−2 − u(ξ,ǫ) )
Rn
+
Z n
X
− µλ2d f (λ−2 |x|2 )Hab (x)∂a ∂b u(ξ,ǫ) w(ξ,ǫ)
Rn
+ a,b=1
Z 2
2
n−2
− (n − 2)Tc (|v(ξ,ǫ) | n−2 − u(ξ,ǫ) )u(ξ,ǫ) v(ξ,ǫ)
∂Rn
+
Z 
2(n−1) 
2(n−1)
n−2 
+(n − 2)Tc (|v(ξ,ǫ) | n−2 − u(ξ,ǫ) )
∂Rn
+

(4d+4)(n−1) 2(n−1) n−2 2−n
≤Cλ n−2 µ n−2 + Cµλ2d+2+ 2 ρ 2 + Cλn−2 ρ2−n
for (ξ, ǫ) ∈ λΩ.
Proof. By definition of v(ξ,ǫ) , we have
Z
 
0= h∇v(ξ,ǫ) , ∇(v(ξ,ǫ) − u(ξ,ǫ) )ig + cn Rg v(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
Rn
+
Z
4
− n(n − 2) |v(ξ,ǫ) | n−2 v(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
Rn
+
Z
2
+ (n − 2)Tc |v(ξ,ǫ) | n−2 v(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) ).
∂Rn
+

By (3.1), (4.3) and (4.4), an integration by parts gives

Z
  

 h∇u(ξ,ǫ) , ∇v(ξ,ǫ) ig − |∇u(ξ,ǫ) |2g + cn Rg u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
 Rn+
Z n+2
n−2
− n(n − 2)u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
Rn
+
Z n
n−2
+ (n − 2)Tc u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
∂Rn
+

16
 
Z n
X 

− µλ f (λ |x| )Hab (x)∂a ∂b u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) )
2d −2 2
Rn+ a,b=1 
Z
 n+2
 n−2
= (∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u(ξ,ǫ)
 Rn+

n 
X 
+ µλ2d f (λ−2 |x|2 )Hab (x)∂a ∂b u(ξ,ǫ) )(u(ξ,ǫ) − v(ξ,ǫ) )
a,b=1 

n+2
≤kv(ξ,ǫ) − u(ξ,ǫ) k n−2
2n · ∆g u(ξ,ǫ) − cn Rg u(ξ,ǫ) + n(n − 2)u n−2
L (Rn
+)
(ξ,ǫ)

Xn

+ µλ2d f (λ−2 |x|2 )Hab (x)∂a ∂b u(ξ,ǫ)

a,b=1 n+2
2n
n
L (R+ )
2n(2d+2) n−2 2−n
≤Cλ n−2 µ3 + Cλ2d+2 µλ 2 ρ 2 + Cλn−2 ρ2−n .

On the other hand, by Proposition 4.2 we have

 
Z Xn 
 
 µλ f (λ |x| )Hab (x)∂a ∂b u(ξ,ǫ) (v(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) )
2d −2 2
 n
 R+ a,b=1 
≤Cλ2d+2 µkv(ξ,ǫ) − u(ξ,ǫ) − w(ξ,ǫ) k 2n
L n−2 (Rn
+)
(4d+4)(n−1) 2(n−1) n−2 2−n
≤Cλ n−2 µ n−2 + Cλ2d+2+ 2 µρ 2 + Cλn−2 ρ2−n .

Putting these facts together, we obtain the desired estimate.

Proposition 4.4. There hold
Z   Z  
 4
2 2n
 4
n−2
2n
n−2 
 |v(ξ,ǫ) | n−2 − u(ξ,ǫ) u(ξ,ǫ) v(ξ,ǫ) − |v(ξ,ǫ) | n−2 − u(ξ,ǫ) 
 Rn+ n Rn+ 
2n (4d+4)n
≤Cµ n−2 λ n−2 + Cλn ρ−n

and
Z   Z  
 2
1 2(n−1)
 2
n−2
2(n−1)
n−2 
 |v(ξ,ǫ) | n−2 − u(ξ,ǫ) u(ξ,ǫ) v(ξ,ǫ) − |v(ξ,ǫ) | n−2 − u(ξ,ǫ) 
 ∂Rn+ n − 1 ∂Rn+ 
2(n−1) (4d+4)(n−1)
≤Cµ n−2 λ n−2 + Cλn−1 ρ1−n

for (ξ, ǫ) ∈ λΩ.

Proof. We only need to prove the second assertion, since the first one is similar to [3, Proposition 12]
together with (4.4). Observe that
   
 2 2
1 2(n−1) 2(n−1) 
 |v(ξ,ǫ) | n−2 − u n−2
u v − |v | n−2 − u n−2 
 (ξ,ǫ) (ξ,ǫ) (ξ,ǫ)
n−1
(ξ,ǫ) (ξ,ǫ) 
2(n−1)
≤C|v(ξ,ǫ) − u(ξ,ǫ) | n−2 .

17
This together with (4.4) implies
Z   Z  
 2
1 2(n−1)
 2
n−2
2(n−1)
n−2 
 |v(ξ,ǫ) | n−2 − u(ξ,ǫ) u(ξ,ǫ) v(ξ,ǫ) − |v(ξ,ǫ) | n−2 − u(ξ,ǫ) 
 ∂Rn+ n − 1 ∂Rn+ 
2(n−1) 2(n−1) (4d+4)(n−1)
≤Ckv(ξ,ǫ) − u(ξ,ǫ) k n−2
2(n−1) ≤ Cµ n−2 λ n−2 + Cλn−1 ρ1−n .
L n−2 (∂Rn
+)

This proves the assertion.

Proposition 4.5. There holds
Z   Z
 2n 2(n−1)
 2 2 n−2 n−2
 |∇u(ξ,ǫ) |g + cn Rg u(ξ,ǫ) − n(n − 2)u(ξ,ǫ) + (n − 2)Tc u(ξ,ǫ)
 Rn+ ∂Rn
+

Z n Z n 
1 X cn X 
− hac hbc ∂a u(ξ,ǫ) ∂b u(ξ,ǫ) + (∂c hab ) u(ξ,ǫ) 
2 2
2 Bρ+ (0) a,b,c=1 4 +
Bρ (0) a,b,c=1 
≤Cρ2d+2 λ4d+4 µ3 + Cλn−2 ρ2−n (4.7)

for (ξ, ǫ) ∈ λΩ.

Proof. By equation (3.1) of u(ξ,ǫ) , we have
Z Z 2n
Z 2(n−1)
2 n−2 n−2
|∇u(ξ,ǫ) | = n(n − 2) u(ξ,ǫ) − (n − 2)Tc u(ξ,ǫ) .
Rn
+ Rn
+ ∂Rn
+

Then the LHS of (4.7) becomes


Z   1Z n
 X
 |∇u | 2
+ c n Rg u 2
− |∇u | 2
− hac hbc ∂a u(ξ,ǫ) ∂b u(ξ,ǫ)
 n (ξ,ǫ) g (ξ,ǫ) (ξ,ǫ)
2 Bρ+ (0)
 R+ a,b,c=1

Z n 
cn X 
+ (∂c hab ) u(ξ,ǫ)  .
2 2
4 Bρ (0)
+

a,b,c=1

Notice that
 
 Xn 
 ab 
g (x) − δab (x) + hab (x) − 1 h ac (x)h bc (x)
 2 
 a,b,c=1 
≤C|h(x)|3 ≤ Cµ3 (λ + |x|)6d+6

for x ∈ Bρ+ (0). This implies


Z Z Xn


 |∇u | 2
− |∇u | 2
+ hab ∂a u(ξ,ǫ) ∂b u(ξ,ǫ)
 n (ξ,ǫ) g (ξ,ǫ)
 R+ Rn
+ a,b=1

Z X n 
1 
− hac hbc ∂a u(ξ,ǫ) ∂b u(ξ,ǫ) 
2 Bρ+ (0) 
a,b,c=1
Z Z
≤ Cλn−2 µ3 (λ + |x|)6d+6+2−2n + Cλn−2 (λ + |x|)2−2n
Bρ+ (0) Rn +
+ \Bρ (0)

18
 ≤ Cµ3 ρ2d+2 λ4d+4 + Cλn−2 ρ2−n .

Since ∂a hab (x) = 0 in Bρ+ (0), it follows from [4, Proposition 4] that
 
 Xn 
 1 
Rg (x) + 2
(∂c hab (x)) 
 4
 a,b,c=1 
≤C|h(x)|2 |∂ 2 h(x)| + C|h(x)||∂h(x)|2 ≤ Cµ3 (λ + |x|)6d+4

for x ∈ Bρ+ (0). This implies

 
Z Z n 
 1 X 
 2
(∂l hik (x)) u(ξ,ǫ) 
2 2
 n Rg (x)u(ξ,ǫ) + 4 +
 R+ Bρ (0) i,k,l=1 
Z Z
≤Cλn−2 µ3 (λ + |x|)6d+6+2−2n + Cλn−2 (λ + |x|)4−2n
Bρ+ (0) Rn +
+ \Bρ (0)

≤Cµ3 ρ2d+2 λ4d+4 + Cλn−2 ρ4−n .

Since hab (x) is trace-free, by (3.2) we obtain

n
X n
X
hab ∂a u(ξ,ǫ) ∂b u(ξ,ǫ) = cn hab ∂a ∂b u2(ξ,ǫ) ,
a,b=1 a,b=1

then
Z n
X Z n
X
hab ∂a u(ξ,ǫ) ∂b u(ξ,ǫ) = cn hab ∂a ∂b u2(ξ,ǫ) .
Rn
+ a,b=1 Rn
+ a,b=1

Again by ∂a hab (x) = 0 in Bρ+ (0), we obtain

 
Z n  Z
 X 
 hab ∂a u(ξ,ǫ) ∂b u(ξ,ǫ)  ≤ u2(ξ,ǫ) ≤ Cλn−2 ρ4−n .
 n +
 R+ a,b=1  Rn
+ \Bρ (0)

Then the desired estimate follows from all the above facts.
Consequently, collecting Propositions 4.3- 4.5 together, we arrive at the following key estimate.
Corollary 4.6. Let Fg (ξ, ǫ) be the function defined in (3.12), then for any (ξ, ǫ) ∈ λΩ, there holds

 Z n Z n
 X X
Fg (ξ, ǫ) − 1 h h ∂ u ∂ u
ac bc a (ξ,ǫ) b (ξ,ǫ) +
cn
(∂c hab )2 u2(ξ,ǫ)
 2 + 4 +
 Bρ (0) a,b,c=1 Bρ (0) a,b,c=1

Z n 
X 
−µλ2d f (λ−2 |x′ |2 )w(ξ,ǫ) Hab (x)∂a ∂b u(ξ,ǫ) 
Rn
+ a,b=1 
2(n−1) (4d+4)(n−1) n−2 2−n
≤Cµ n−2 λ n−2 + Cµλ2d+2+ 2 ρ 2 + Cλn−2 ρ2−n ,

where w(ξ,ǫ) satisfies (4.5).

19
5 Finding a critical point of an auxiliary function
We define
Z n−1
X Z n−1
X
1 cn
F (ξ, ǫ) = H il H jl ∂i u(ξ,ǫ) ∂j u(ξ,ǫ) − (∂l H ij )2 u2(ξ,ǫ)
2 Rn
+ i,j,l=1
4 Rn
+ i,j,l=1

Z n−1
X
+ H ij ∂i ∂j u(ξ,ǫ) z(ξ,ǫ)
Rn
+ i,j=1

for (ξ, ǫ) ∈ Rn+ × (0, ∞), where z(ξ,ǫ) (x) = µ−1 w(ξ,ǫ) (x), which satisfies
Z 4
Z 2
n−2 n−2
h∇z(ξ,ǫ) , ∇ϕi − n(n + 2)u(ξ,ǫ) z(ξ,ǫ) ϕ + nTc u(ξ,ǫ) z(ξ,ǫ) ϕ
Rn
+ ∂Rn
+
Z
=− f (|x′ |2 )H ij (x)∂i ∂j u(ξ,ǫ) ϕ (5.1)
Rn
+

for all test function ϕ ∈ E(ξ,ǫ) .

Next we show that the function F (ξ, ǫ) has a strict local minimum. Throughout this section we use
indices 1 ≤ i, j, k, l, m, p, q, r ≤ n − 1 .
Since H ab (−x) = H ab (x) for any x ∈ Rn+ , the function F (ξ, ǫ) satisfies F (ξ, ǫ) = F (−ξ, ǫ) for all
(ξ, ǫ) ∈ Rn−1 × (0, ∞). This implies

∂ ∂2
F (0, ǫ) = F (0, ǫ) = 0 (5.2)
∂ξp ∂ǫ∂ξp

for all ǫ > 0 .

Proposition 5.1. There hold
Z n−1
X
(∂l Hik (x))2 xp xq
Sn−2
r (0) i,k,l=1

n−1
X
2rn+2
= |Sn−2 | (W ipkl + W ilkp )(W iqkl + W ilkq )
(n − 1)(n + 1)
i,k,l=1
n−1
X
rn+2
+ |Sn−2 | (W ijkl + W ilkj )2 δpq
(n − 1)(n + 1)
i,j,k,l=1

and
Z n−1
X
Hik (x)2 xp xq
Sn−2
r (0) i,k=1

n−1
X
2rn+4
= |Sn−2 | (W ipkl + W ilkp )(W iqkl + W ilkq )
(n − 1)(n + 1)(n + 3)
i,k,l=1
n−1
X
rn+4
+ |Sn−2 | (W ijkl + W ilkj )2 δpq .
2(n − 1)(n + 1)(n + 3)
i,j,k,l=1

Proof. The proof is similar to [3, Proposition 16].

20
Proposition 5.2. There holds
Z n−1
X
(∂l H ik (x))2 xp xq
Sn−2
r (0) i,k,l=1
  
= 2(W ipkl + W ilkp )(W iqkl + W ilkq ) (n + 3)f (r2 )2 + 8r2 f (r2 )f ′ (r2 ) + 4r4 f ′ (r2 )2
 
+ (W ijkl + W ilkj )2 δpq (n + 3)f (r2 )2 + 4r2 f (r2 )f ′ (r2 ) + 2r4 f ′ (r2 )2
|Sn−2 |rn+2
· .
(n − 1)(n + 1)(n + 3)
Proof. Since

∂l H ik (x) = f (|x′ |2 )∂l Hik (x) + 2xl f ′ (|x′ |2 )Hik (x),

and by Euler’s formula we obtain

n−1
X
(∂l H ik (x))2
i,k,l=1
n−1
X n−1
X
′ 2 2 2 ′ 2 ′ ′ 2
=f (|x | ) (∂l Hik (x)) + 4f (|x | )f (|x | ) Hik (x)xl ∂l Hik (x)
i,k,l=1 i,k,l=1
n−1
X
+ 4|x′ |2 f ′ (|x′ |2 )2 Hik (x)2
i,k=1
n−1
X   n−1
X
=f (|x′ |2 )2 (∂l Hik (x))2 + 4 2f (|x′ |2 )f ′ (|x′ |2 ) + |x′ |2 f ′ (|x′ |2 )2 Hik (x)2 .
i,k,l=1 i,k=1

Hence, the assertion follows from Proposition 5.1.

Corollary 5.3. There holds
Z n−1
X
(∂l H ik )2 (x)
Sn−2
r (0) i,k,l=1

n−1
X
|Sn−2 |rn  
= (W ijkl + W ilkj )2 (n + 1)f (r2 )2 + 4r2 f (r2 )f ′ (r2 ) + 2r4 f ′ (r2 )2 .
(n − 1)(n + 1)
i,j,k,l=1

Proposition 5.4. There holds

F (0, ǫ)
n−1
X Z ∞Z ∞
cn |Sn−2 |ǫn−2
=− (W ijkl + W ilkj )2 (ǫ2 + (t − Tc ǫ)2 + r2 )2−n
4(n − 1)(n + 1) 0 0
i,j,k,l=1
 
· r (n + 1)f (r ) + 4r2 f (r2 )f ′ (r2 ) + 2r4 f ′ (r2 )2 drdt.
n 2 2

Pn−1
Proof. Since i,j=1 H̄ij (x)∂i ∂j u(0,ǫ) (x) = 0, then z(0,ǫ) = 0, and by symmetry we have
Z n−1
X
H il (x)H kl (x)∂i u(0,ǫ) (x)∂k u(0,ǫ) (x) = 0.
Sn−2
r (0) i,k,l=1

21
Then we have
Z
cn
F (0, ǫ) = − (∂l H ik )2 (x)u2(0,ǫ) (x)
4 Rn
+
Z ∞ Z ∞ Z
cn
=− (∂l H ik )2 (x)u2(0,ǫ) (x)dσr (x)drdxn .
4 0 0 Sn−2
r (0)

Hence, the result follows from Corollary 5.3.

By Proposition 5.4, we rewrite

(n − 2)|Sn−2 |
F (0, ǫ) = − (W ijkl + W ilkj )2
16(n − 1)2 (n + 1)
2d
X Z ∞Z ∞
· αq r2q+n ǫn−2 (ǫ2 + (t − Tc ǫ)2 + r2 )2−n drdt,
q=0 0 0

where αq are constants defined by

2d
X
αq sq := (n + 1)f (s)2 + 4sf (s)f ′ (s) + 2s2 f ′ (s)2 . (5.3)
q=0

Then we obtain
(n − 2)|Sn−2 |
F (0, ǫ) = − (W ijkl + W ilkj )2
16(n − 1)2 (n + 1)
X2d Z ∞Z ∞
r2q+n
· αq ǫ2q+4 drdt,
q=0 0 0 (1 + (t − Tc )2 + r2 )n−2
(n − 2)|Sn−2 |
=− (W ijkl + W ilkj )2
16(n − 1)2 (n + 1)
X2d Z ∞ Z ∞
2q+4 1 r2q+n
· αq ǫ n−5−2q dt dr
q=0 0 (1 + (t − Tc )2 ) 2 0 (1 + r2 )n−2

X2d
(n − 2)|Sn−2 | 2 2q + n + 1 n − 5 − 2q
=− (W ijkl + W ilkj ) αq cq ǫ2q+4 B( , ).
16(n − 1)2 (n + 1) q=0
2 2

where Z ∞
5+2q−n
cq = (1 + (t − Tc )2 ) 2 dt, q ∈ N and 0 ≤ q ≤ 2d. (5.4)
0
For clarity, we rewrite

(n − 2)|Sn−2 | n−1 n−5

F (0, ǫ) = − 2
B( , )(W ijkl + W ilkj )2 I(ǫ2 ), (5.5)
32(n − 1) (n + 1) 2 2

where
 
2d
X Yq
cq αq sq+2 n − 1 + 2j .
I(s) = (5.6)
q=0 j=0
n − 5 − 2j

22
Proposition 5.5. There holds
∂2
F (0, ǫ)
∂ξp ∂ξq
2(n − 2)2 |Sn−2 |
=− (W ipkl + W ilkp )(W iqkl + W ilkq )
(n − 1)(n + 1)(n + 3)
Z ∞Z ∞
ǫn−2
· rn+4 (2f (r2 )f ′ (r2 ) + r2 f ′ (r2 )2 )drdt
0 0 (ǫ2 + (t − Tc ǫ)2 + r2 )n
(n − 2)2 |Sn−2 |
− (W ijkl + W ilkj )2 δpq
2(n − 1)(n + 1)(n + 3)
Z ∞Z ∞
ǫn−2
· rn+4 (2f (r2 )f ′ (r2 ) + r2 f ′ (r2 )2 )drdt
0 0 (ǫ2 + (t − Tc ǫ)2 + r2 )n
(n − 2)2 |Sn−2 |
+ (W ijkl + W ilkj )2 δpq
4(n − 1)2 (n + 1)
Z ∞Z ∞
ǫn−2
· rn+4 f ′ (r2 )2 drdt.
0 0 (ǫ + (t − Tc ǫ)2 + r2 )n−1
2

Proof. As in [3, Proposition 21], similarly we obtain

∂2
F (0, ǫ)
∂ξp ∂ξq
Z
ǫn−2
=(n − 2)2 H pl (x)H ql (x)
Rn
+
(ǫ2 + (xn − Tc ǫ)2 + |x′ |2 )n
Z
(n − 2)2 ǫn−2
− 2 + (x − T ǫ)2 + |x′ |2 )n
(∂l H ik (x))2 xp xq
4 Rn+
(ǫ n c
Z
(n − 2)2 ǫn−2
+ (∂l H ik (x))2 δpq .
8(n − 1) Rn+ (ǫ2 + (xn − Tc ǫ)2 + |x′ |2 )n−1

This together with Proposition 5.2 and Corollary 5.3 gives the desired assertion.
For brevity, we let
2d−1
X
2f (s)f ′ (s) + sf ′ (s)2 := βq s q . (5.7)
q=0

By definition (5.4) of cq , a direct computation yields

Z ∞Z ∞
ǫn−2
rn+4 (2f (r2 )f ′ (r2 ) + r2 f ′ (r2 )2 )drdt
0 0 (ǫ2 + (t − Tc ǫ)2 + r2 )n
1 n+3 n−3
:= B( , )J(ǫ2 ),
2 2 2
where
2d−1
X B( n+2q+5 , n−2q−5 )
J(s) = βq cq sq+2 2
n+3 n−3
2

q=0
B( 2 , 2 )
 
2d−1
X Yq
cq βq sq+2 n + 3 + 2j 
= . (5.8)
q=0 j=0
n − 5 − 2j

23
 In order to show that F (ξ, ǫ) has a strict local minimum at (0, 1), By (5.2),(5.5) and Proposition 5.5, our
Pd
strategy is to find some polynomials f (s) := i=0 ai si , ai ∈ R for 1 ≤ i ≤ d, such that I(1) > 0, I ′ (1) =
0, I ′′ (1) < 0 and J(1) < 0.
Before proceeding to find such polynomials f , we first need the following elementary result.
Lemma 5.6. Let Tc < 0, n ≥ 25 and cq be defined in (5.4), there holds
n − 2q − 7 cq+1 n − 2q − 6
(1 + Tc2 ) ≤ ≤ (1 + Tc2 )
n − 2q − 8 cq n − 2q − 8
for 0 ≤ q ≤ 2d. In particular,
(n − 6)(n − 10) c2 (n − 10)(n − 7)
≥ 1 ≥ .
(n − 8)(n − 9) c0 c2 (n − 8)2
Proof. Let a = −Tc > 0 and define
Z ∞
1
Iα (a) := (1 + r2 )−α dr for α > .
a 2
An integration by parts gives
2α 1
Iα (a) = Iα+1 (a) + (1 + a2 )−α a. (5.9)
2α − 1 2α − 1
Notice that
k−1
Y k−1
Y 2α + 2i
2α + 2i π
lim Iα+k (a) ≤ lim (1 + a2 )−α−k+1
k→∞
i=0
2α + 2i − 1 2 k→∞ i=0 2α + 2i − 1
Y
k−1
1

1+
π 2α + 2i − 1
= lim i=0 = 0.
2 k→∞ (1 + a2 )α−(k−1)
From this, we iterate (5.9) to obtain
∞
X k−1
Y 2α + 2i
1
Iα (a) = a (1 + a2 )−(α+k) .
2α + 2k − 1 i=0 2α + 2i − 1
k=0

Then we have
Iα (a)
(1 + a2 )Iα+1 (a)
∞
X k−1
Y 2α + 2i
1
(1 + a2 )−(α+k)
2α + 2k − 1 i=0 2α + 2i − 1
= k=0
∞ k−1
.
X 1 Y 2α + 2i + 2
(1 + a2 )−(α+k)
2α + 2k + 1 i=0 2α + 2i + 1
k=0

Since
k−1
Y 2α + 2i
1
2α + 2k − 1 i=0 2α + 2i − 1 2α(2α + 2k + 1)
= ,
1
k−1
Y
2α + 2i + 2 (2α − 1)(2α + 2k)
2α + 2k + 1 i=0 2α + 2i + 1

24
then
2α Iα (a) 2α + 1
≤ ≤ .
2α − 1 (1 + a2 )Iα+1 (a) 2α − 1

In particular, it follows from (5.4) and the above inequality that for 0 ≤ q ≤ 2d,

cq = I n−2q−5 (a) and

2

n − 2q − 7 cq+1 n − 2q − 6
(1 + Tc2 ) ≤ ≤ (1 + Tc2 ).
n − 2q − 8 cq n − 2q − 8
And the remained estimates follow from the above estimate.
Now we choose d = 1 and let f (s) = a0 + a1 s. Then by (5.3) we obtain

α0 = (n + 1)a20 , α1 = 2(n + 3)a0 a1 , α2 = (n + 7)a21 .

Differentiating (5.6) with respect to s, we obtain

 
X2 Yq
n − 1 + 2j 
I ′ (s) = (q + 2)cq αq sq+1
q=0 j=0
n − 5 − 2j

2c0 α0 (n − 1) 3c1 α1 (n − 1)(n + 1) 2 4c2 α2 (n − 1)(n + 1)(n + 3) 3

= s+ s + s
n−5 (n − 5)(n − 7) (n − 5)(n − 7)(n − 9)
 
2(n + 1)(n − 1) 2 3(n + 3) 2 2(n + 3)(n + 7) 2 3
= c0 a0 s + c1 a0 a1 s + c2 a1 s .
n−5 n−7 (n − 7)(n − 9)

We set a1 = −1 and define

3(n + 3) 2(n + 3)(n + 7)
pn (a0 ) := c0 a20 − c1 a0 + c2 ,
n−7 (n − 7)(n − 9)

then
2(n + 1)(n − 1)
I ′ (1) = pn (a0 ).
n−5
Notice that the discriminant d(pn ) of pn is given by
 
(n + 3)2 2 (n + 7)(n − 7)
d(pn ) = 9c 1 − 8 c 0 2 .
c
(n − 7)2 (n + 3)(n − 9)

By Lemma 5.6, we have

 
(n + 3)2 2 (n + 7)(n − 8)2
d(pn ) ≥ c 9 − 8 .
(n − 7)2 1 (n + 3)(n − 9)(n − 10)

Define

q(n) = 9(n + 3)(n − 9)(n − 10) − 8(n − 8)2 (n + 7),

then

q ′ (n) = 3n2 − 144n + 681.

25
Notice that q ′ (x) > 0 for x ≥ 62 and q(62) > 0, q(61) < 0, then d(pn ) > 0 for n ≥ 62.
Hence, we can choose
" s #
(n + 3)c1 (n + 7)(n − 7) c0 c2
a0 = 3+ 9−8 (5.10)
2(n − 7)c0 (n + 3)(n − 9) c21

such that I ′ (1) = 0. From this and Lemma 5.6, we obtain

 
1 2 (n + 3)(n + 7)
I(1) = c0 a0 − c2
3 (n − 7)(n − 9)
 
c2 n + 3 9(n + 3) c21 n+7
≥ −
3 n − 7 4(n − 7) c0 c2 n−9
 
c2 n + 3 9(n + 3)(n − 10) n + 7
≥ − >0
3 n−7 4(n − 8)2 n−9
for n ≥ 62.
Lemma 5.7. There holds I ′′ (1) < 0 for n ≥ 62.
Proof. By definition (5.6) of I, we have
 
2(n + 1)(n − 1) 6(n + 3) 6(n + 3)(n + 7)
I ′′ (1) = c0 a20 − c1 a0 + c2 .
n−5 n−7 (n − 7)(n − 9)
Notice that
 
2(n − 1)(n + 1) (n + 3) (n + 3)(n + 7)
I ′′ (1) − I ′ (1) = −3c1 a0 + 4c2 .
n−5 (n − 7) (n − 7)(n − 9)
By definition (5.10) of a0 and I ′ (1) = 0, we have
 
′′ 2(n − 1)(n + 1) 9 c21 (n + 3)2 (n + 3)(n + 7)
I (1) ≤ − + 4c2
n−5 2 c0 (n − 7)2 (n − 7)(n − 9)
(n − 1)(n + 1)
=− d(pn ) < 0
(n − 5)c0
for n ≥ 62. This implies the desired result.
Lemma 5.8. There holds J(1) < 0 for n ≥ 62.
Proof. By definition (5.8) of J, letting a1 = −1 and a0 be chosen as in (5.10) we obtain β0 = −2a0 ,
β1 = 3 by virtue of (5.7), and
2(n + 3) 3(n + 3)(n + 5) 3
J(s) = − c0 a0 s2 + c1 s .
n−5 (n − 5)(n − 7)
Thus we have
" s #
(n + 3)c1 (n + 7)(n − 7) c0 c2
J(1) = 6 − (n + 3) 9 − 8 .
(n − 5)(n − 7) (n + 3)(n − 9) c21

By Lemma 5.6 we have

" s #
(n + 3)c1 8(n + 7)(n − 8)2
J(1) ≤ 6 − (n + 3) 9 − .
(n − 5)(n − 7) (n + 3)(n − 9)(n − 10)

26
We consider
 
1 2 36
P(n) := α(n + 3)(n − 9)(n − 10) − (n + 7)(n − 8) with α = 9− 2 .
8 65

Then a direct computation shows that P ′′ (n) = 6(α − 1)n + 18 − 32α > 0 for n ≥ 62 and P ′ (62) >
0, P(62) > 0. This implies that P(n) > 0 for n ≥ 62. Observe that
s
(n + 7)(n − 8)2 √
(n + 3) 9 − 8 > (n + 3) 9 − 8α ≥ 6,
(n + 3)(n − 9)(n − 10)

where the last inequality follows from n ≥ 62 and the choice of α. This yields J(1) < 0 for n ≥ 62.
Combing the above facts with Lemmas 5.7-5.8 we arrive at
Proposition 5.9. Let n ≥ 62, then there exists a polynomial f (s) = −s + a0 with
" s #
(n + 3)c1 (n + 7)(n − 7) c0 c2
a0 = 3+ 9−8
2(n − 7)c0 (n + 3)(n − 9) c21

such that I(1) > 0, I ′ (1) = 0, I ′′ (1) < 0 and J(1) < 0. This implies that F (ξ, ǫ) has a strict local
minimum at the point (0, 1).

6 Proof of Theorem 1.1

Proposition 6.1. For n ≥ 62, let g = exp(h) is a smooth Riemannian metric on Rn+ , where h is a symmetric
trace-free two tensor on Rn+ satisfying

2 −2 ′ 2

hij (x) = µλ f (λ |x | )Hij (x), in Bρ+ ,
hab (x) = 0, in Rn+ \ Bρ+ ,


hna (x) = 0, in Rn+ ,

where f (s) = a0 − s with the constant a0 given in Proposition 5.9, 0 < µ ≤ 1, 0 < λ ≤ ρ ≤ 1,
1 ≤ i ≤ n − 1, 1 ≤ a, b ≤ n and Hab is defined in (4.1). Assume that |h(x)| + |∂h(x)| + |∂ 2 h(x)| ≤ α for
all x ∈ Rn+ . If α and µ−2 λn−10 ρ2−n are sufficiently small, then there exists a positive smooth solution of
 n+2
∆g v − cn Rg v + n(n − 2)v n−2 = 0, in Rn+ ,
∂v n (6.1)
 = (n − 2)Tc v n−2 , on ∂Rn+ .
∂xn
Moreover, there exists C = C(n, Tc ) > 0, such that
2−n
sup v ≥ Cλ 2
+
Bλ (0)

and
Z Z
2n n−2 2(n−1)
2(n − 2) v n−2 − Tc v n−2

Rn
+
n−1 ∂Rn
+
Z Z
n−2
2n
n−2 2(n−1)
n−2
<2(n − 2) u(0,1) − Tc u(0,1) .
Rn
+
n−1 ∂Rn
+

27
Proof. It follows from Proposition 5.9 that (0, 1) is a strict local minimum point of F (ξ, ǫ). Hence, we can
find an open set Ω′ ⊂ Ω such that (0, 1) ∈ Ω′ and F (0, 1) < inf ′ F (ξ, ǫ) < 0. By Corollary 4.6 with
(ξ,ǫ)∈∂Ω
d = 1, we have
 
Fg (λξ, λǫ) − λ4d+4 µ2 F (ξ, ǫ)
2(n−1) (4d+4)(n−1) n−2 2−n
≤ Cµ n−2 λ n−2 + Cµλ2d+2+ 2 ρ 2 + Cλn−2 ρ2−n

for all (ξ, ǫ) ∈ Ω, equivalently,

 −4d−4 −2 
λ µ Fg (λξ, λǫ) − F (ξ, ǫ)
2 4d+4 n−4d−6 2−n
≤Cµ n−2 λ n−2 + Cµ−1 λ 2 ρ 2 + Cµ−2 λn−4d−6 ρ2−n

for all (ξ, ǫ) ∈ Ω. If µ−2 λn−4d−6 ρ2−n is sufficiently small, then we have

Fg (0, λ) < inf Fg (λξ, λǫ) < 0.

(ξ,ǫ)∈∂Ω′

¯ ǭ) ∈ Ω′ such that

Consequently, there exists (ξ,
¯ λǭ) =
Fg (λξ, inf Fg (λξ, λǫ) < 0.
(ξ,ǫ)∈Ω′

It follows from Proposition 3.5 that the function v = v(λξ̄,λǭ) obtained in Proposition 3.4 is a positive
smooth solution to (6.1). By definition of Fg we have
Z Z
2n n−2 2(n−1)
2(n − 2) v n−2 − Tc v n−2
Rn
+
n−1 ∂Rn+
Z Z
¯ λǭ) + 2(n − 2)
2n
n−2
n−2 2(n−1)

=Fg (λξ, u(λξ̄,λǭ) − Tc u(λn−2

ξ̄,λǭ)
,
Rn
+
n−1 ∂Rn
+

whence
Z Z
2n n−2 2(n−1)
2(n − 2) v n−2 − Tc v n−2

Rn
+
n−1 ∂Rn
+
Z Z
2n
n−2
n−2 2(n−1)
n−2
<2(n − 2) u(0,1) − Tc u(0,1) .
Rn
+
n−1 ∂Rn
+

By (3.11) and Proposition 3.2 we estimate

kv − u(λξ̄,λǭ) k 2n
+
≤ kv − u(λξ̄,λǭ) k 2n ≤ Cα.
L n−2 (Bλ (0)) L n−2 (∂Rn
+)

Then,
n−2
|Bλ+ (0)| 2n sup v ≥ kvk 2n
+
≥ −Cα + ku(λξ̄,λǭ) k 2(n−1) .
+ L (n−2) (Bλ (0)) L (n−2) +
(Bλ (0))
Bλ (0)

Hence, if α is sufficiently small, then we obtain

2−n
sup v ≥ Cλ 2 .
+
Bλ (0)

This completes the proof.

28
Theorem 6.2. Let n ≥ 62, then there exists a smooth Riemannian metric g on Rn+ with the following
properties:
+
(a) gab (x) = δab for Rn+ \ B1/2 (0);

(b) g is not conformally flat;

(c) ∂Rn+ is totally geodesic with respect to the induced metric of g;
(d) there exists a sequence of positive smooth functions {vν ; ν ∈ N} satisfying
 n+2
∆g vν − cn Rg vν + n(n − 2)vνn−2 = 0, in Rn+ ,
 ∂vν = (n − 2)Tc vνn−2 ,
n
on ∂Rn+ ,
∂xn
for all ν. Moreover, there hold
Z Z
2n
n−2 2(n−1)

2(n − 2) vνn−2 − Tc vν n−2

Rn
+
n−1 ∂Rn
+
Z Z
2n
n−2
n−2 2(n−1)
n−2
<2(n − 2) u(0,1) − Tc u(0,1)
Rn
+
n−1 ∂Rn
+

for all ν, i.e. I(Rn+ ,g) [vν ] < Sc , and sup vν → ∞ as ν → ∞.

B1+ (0)

Proof. Let χ be a smooth cut-off function in R such that 0 ≤ χ(t) ≤ 1 for t ∈ R, χ(t) = 1 for t ≤ 1 and
χ(t) = 0 for t ≥ 2. We define a trace-free symmetric two-tensor in Rn+ by
∞
X
hab (x) = χ(4N 2 |x − xN |)2−N f (2N |x′ − xN |2 )Hab (x − xN ),
N =N0

where xN = ( N1 , 0, · · · , 0) ∈ ∂Rn+ . Observe that h is smooth and satisfies han (x) = 0 and ∂n hab (x) = 0
on ∂Rn+ . We choose α > 0 to be the constant in Proposition 6.1 and N0 sufficiently large, then hab (x) = 0
for |x| ≥ 12 and |h(x)| + |∂h(x)| + |∂ 2 h(x)| ≤ α for all x ∈ Rn+ . Thus, the desired assertion follows from
Proposition 6.1 with λ = 2−N/2 , ρ = (2N )−2 , µ = 2−N .

References
[1] S. Almaraz, Blow-up phenomena for scalar-flat metrics on manifolds with boundary, J. Differential
Equations 251 (2011) ,1813-1840.
[2] T. Aubin, Some nonlinear problems in Riemannian geometry, Springer Monographs in Mathematics,
Springer-Verlag, Berlin, 1998.
[3] S. Brendle, Blow-up phenomena for the Yamabe equation, J. Amer. Math. Soc. 21 (2008), no. 4,
951-979.
[4] S. Brendle and F. Marques, Blow-up phenomena for the Yamabe equation II, J. Differential Geom. 81
(2009), no. 2, 225-250.
[5] X. Chen, Y. Ruan and L. Sun, The Han-Li conjecture in constant scalar curvature and constant bound-
ary mean curvature problem on compact manifolds, preprint (2018), 49 pages, arXiv:1805.09597.

29
 [6] X. Chen and L. Sun, Existence of conformal metrics with constant scalar curvature and constant
boundary mean curvature on compact manifolds, Commun. Contemp. Math. 21 (2019), no. 3,
1850021, 51 pp.
[7] P. Cherrier, Problèmes de Neumann non linéaires sur les variétés Riemannienes, J. Funct. Anal. 57
(1984) 154-206.
[8] O. Druet, Compactness for Yamabe metrics in low dimensions, Internat. Math. Res. Notices. 23 (2004)
1143-1191.
[9] J. Escobar, Conformal deformation of a Riemannian metric to a constant scalar curvature metric with
constant mean curvature on the boundary, Indiana Univ. Math. J. 45(4), (1996), 917-943.
[10] J. Escobar, Uniqueness theorems on conformal deformation of metrics, Sobolev inequalities, and an
eigenvalue estimate, Comm. Pure Appl. Math., 43 (1990), 857-883.
[11] D. Gilbarg and N. Trudinger, Elliptic partial differential equations of second order, Reprint of the
1998 edition. Classics in Mathematics. Springer-Verlag, Berlin, 2001.
[12] Z. C. Han and Y. Y. Li, The existence of conformal metrics with constant scalar curvature and constant
boundary mean curvature, Comm. Anal. Geom. 8 (2000), no. 4, 809-869.
[13] Z. C. Han and Y. Y. Li, The Yamabe problem on manifolds with boundary: existence and compactness
results, Duke Math. J. 99 (1999), no. 3, 489-542.
[14] S. Kim, M. Musso and J. Wei, A non-compactness result on the fractional Yamabe problem in large
dimensions, J. Funct. Anal. 273 (2017), no. 12, 3759-3830.
[15] M. Khuri, F. Marques, and R. Schoen, A compactness theorem for the yamabe problem, J. Differential
Geom. 81 (2009), no. 1, 143-196.
[16] Y. Y. Li and M. Zhu, Uniqueness theorems through the method of moving spheres, Duke Math. J. 80
2 (1995), 383-417.
[17] Y. Y. Li and M. Zhu, Yamabe type equations on three-dimensional Riemannian manifolds, Commun.
Contemp. Math. 1 (1999) 1-50.
[18] Y.Y. Li and L. Zhang, Compactness of solutions to the Yamabe problem II, Calc. Var. PDE. 24 (2005)
185-237.
[19] Y.Y. Li and L. Zhang, Compactness of solutions to the Yamabe problem III, J. Funct. Anal. 245 (2007)
438-474.
[20] F. Marques, A priori estimates for the Yamabe problem in the non-locally conformally flat case, J.
Differential Geom. 71 (2005) 315-346.
[21] R.M. Schoen, Variational theory for the total scalar curvature functional for Riemannian metrics
and related topics, Topics in the calculus of variations (ed. by Mariano Giaquinta), Lecture Notes in
Mathematics, 1365, Springer Verlag, 1989, 120-154.
[22] R. Schoen, On the number of constant scalar curvature metrics in a conformal class, Differential
geometry (ed. by H. Blaine Lawson, Jr., and Keti Tenenblat), Pitman Monographs and Surveys in
Pure and Applied Mathematics, 52, Longman Scientific & Technical, 1991, 311-320.
[23] J. Wei and C. Zhao, Non-compactness of the prescribed Q-curvature problem in large dimensions,
Calc. Var. Partial Differential Equations 46 (2013), no. 1-2, 123-164.

30