USA-Chile Workshop on Nonlinear Analysis,

Electron. J. Diff. Eqns., Conf. 06, 2001, pp. 313–342.

http://ejde.math.swt.edu or http://ejde.math.unt.edu
ftp ejde.math.swt.edu or ejde.math.unt.edu (login: ftp)

Generalized boundary value problems for

nonlinear elliptic equations ∗
Laurent Véron

Abstract
We give here an overview of some recent developments in the study of
the description of all the positive solutions of

−∆u + |u|q−1 u = 0 (1)

in a domain Ω where q > 1.

1 Introduction
Given a partial differential equation in a domain Ω of RN , a natural question is
to find a way to describe all its solutions by means of their possible boundary
value. For example, if Ω is smooth, any nonnegative harmonic function u in
Ω admits a boundary trace which is a Radon measure µ on ∂Ω and the Riesz-
Herglotz representation formula holds,
Z
u(x) = P (x, y)dµ(y) (2)
∂Ω

for any x ∈ Ω, where P (x, y) is the Poisson kernel in Ω. If Ω is not smooth, the
Poisson kernel is replaced by the Martin kernel and a representation formula
holds.
This article is concerned with is the description of the positive solutions of
(1) and this study is known as the nonlinear trace theory. A description as
in the linear case is still far out of reach, but thank to the works of Le Gall
[24, 25, 26], Marcus and Véron [28, 29, 30, 31, 32], Dynkin and Kuzneztsov
[11, 12, 13, 14, 15] this program is now well advanced. In this survey I want
to present the historical progression which led to the nonlinear trace problem,
and in particular present the preliminary works of Gmira and Véron (1989-
1991) [17] dealing with the measure boundary data and the isolated boundary
singularities, the question of existence and uniqueness of the large solutions in
particular in non-smooth domains and the role of the Borel measures and the
∗ Mathematics Subject Classifications: 35J60, 35J67, 35D05.

Key words: Dirichlet Problem, Radon and Borel measures.

2001
c Southwest Texas State University.
Published January 8, 2001.

313
314 Generalized boundary value problems

boundary capacities framework for representing all the positive solutions. We

shall also give applications to conformal deformations of hyperbolic space.
The associated boundary value problem (or generalised Dirichlet problem)
is the following,
−∆u + |u|q−1 u = 0, in Ω,
(3)
u = g on ∂Ω,
where g may be a function (regular or not, a Radon measure, a generalized
Borel measure).

2 The regular non-linear Dirichlet problem in

regular domains
By using convex analysis, Lp and Schauder’s regularity theory of elliptic equa-
tions, it is classical that for any g in C 2,α (∂Ω) there exists a unique u in C 2,α (Ω̄)
solution of (3). The extension to merely integrable boundary data is due to
Brezis [4]

Theorem 1 Assume g ∈ L1 (∂Ω), then there exists a unique function u ∈

L1 (Ω) ∩ Lq (Ω, ρΩ dx) , where ρΩ (x) = dist(x, ∂Ω such that
Z Z

∂ζ
−u∆ζ + |u|q−1 uζ dx = − gdHN −1 (4)
Ω ∂Ω ∂ν

∀ζ ∈ C01,1 (Ω̄). Moreover the mapping g 7→ u = PΩq (g) is increasing and

ku1 − u2 kL1 (Ω) + kρΩ (h(u1 ) − h(u2 )) kL1 (Ω) ≤ Ckg1 − g2 kL1 (∂Ω) (5)

where uj = PΩq (gj ), j = 1, 2 and h(r) = |r|q−1 r.

In this statement, dHN −1 denotes the (N − 1)-dimensional Hausdorff mea-

sure. Notice also that ζ ∈ C01,1 (Ω̄) ⇒ |ζ(x)| ≤ CρΩ (x), which gives its meaning
to the condition u ∈ Lq (Ω, ρΩ dx). The key point in Brezis proof is the following
result

Lemma 1 For any (f, g) ∈ L1 (Ω, ρΩ dx) × L1 (∂Ω) there exists a unique v ∈
L1 (Ω) such that
Z Z Z
∂ζ
− v∆ζdx = f ζdx − gdHN −1 (6)
Ω Ω ∂Ω ∂ν

∀ζ ∈ C01,1 (Ω̄). Moreover, if ζ ≥ 0,

Z Z Z
∂ζ
− |v|∆ζdx + |g|dHN −1 ≤ f ζsign(v)dx (7)
Ω ∂Ω ∂ν Ω

and the same estimate holds if one replaces |v| by v+ .

Laurent Véron 315

3 The a priori estimate of Keller and Osserman

One of the most striking aspects of the equation (1) is the fact that all the
solutions are locally uniformly bounded. More precisely, by using suitable local
radial supersolutions of (1), Keller [20] and Osserman [34] proved independently
that there always holds

|u(x)| ≤ C(N, q)ρΩ (x)2/(1−q) . (8)

One of the important consequence of this inequality consists in the construction

of positive solution of (3) with a boundary value g belonging to C(∂Ω; [0, ∞]).

Theorem 2 Let Ω be a Lipschitz bounded domain. Then for each g in

C(∂Ω; [0, ∞]), there exists at least one solution of (3).

Proof. For any positive integer k let uk be the solution of

q−1
−∆uk + |u|k uk = 0, in Ω,
uk = gk = min(k, g) on ∂Ω,

By the maximum principle, the sequence {uk } is positive and increasing. More-
over it is locally bounded thanks to (8). Therefore it converges to some solution
u of (1). If x0 ∈ ∂Ω is such that g(x0 ) < ∞, then the continuity of g and
the elliptic equations boundary regularity theory imply that the set of functions
{uk } remains equicontinuous in some neighborhood of x0 . Consequently the
limit function u achieves the value of g in a relative neighborhood of x0 on ∂Ω.
If g(x0 ) = ∞, then {uk (x0 )} is not bounded from above.
The Lipschitz regularity assumption can be relaxed and replaced by the
Wiener regularity criterion. The problem of uniqueness remains open up to 1993
when Kondratiev and Nikishkin [21] discovered that there may exist infinitely
many solutions in the framework of boundary data in the class of continuous
functions with possibly infinite value. This will be completely understood in
the framework of the boundary trace theory.

4 Measure boundary data

In this section Ω is a smooth bounded domain. The Poisson formula which
expresses the Poisson potential of a boundary Radon measure is the following
Z
Pµ (x) = P (x, y)dµ(y) (9)
Ω

The first extension of Theorem 1 to measure boundary data, is due to Gmira

and Véron [17] for the existence part (1989-1991) and Marcus and Véron [31]
for the stability part (1996).
316 Generalized boundary value problems

Theorem 3 Suppose that 1 < q < (N + 1)/(N − 1), then for any µ ∈ M(∂Ω)
there exists a unique u ∈ L1 (Ω) ∩ Lq (Ω, ρΩ dx) such that
Z Z

∂ζ
−u∆ζ + |u|q−1 uζ dx = − dµ (10)
Ω ∂Ω ∂ν

∀ζ ∈ C01,1 (Ω̄). Moreover the mapping g 7→ u = PΩq (µ) is increasing. If µn

converges weakly to µ when n goes to infinity, un = PΩq (µn ) converges to u =
PΩq (µ), locally uniformly in Ω.

Proof. Uniqueness follows from estimate (5). The proof of existence relies
on the following well-known estimates on the Poisson kernel: there exists C =
C(Ω) > 0 such that
−N −N
C −1 ρ(x)|x − y| ≤ P (x, y) ≤ Cρ(x)|x − y|

for all (x, y) ∈ Ω × ∂Ω. Consequently

kP (., y)kLp (Ω) ≤ Kp,Ω ∀1 < p < N/(N − 1) ∀y ∈ ∂Ω,

∗
kP (., y)kLp (Ω;ρdx) ≤ Kp,Ω ∀1 < p < (N + 1)/(N − 1) ∀y ∈ ∂Ω.

Let {gn } ⊂ L1 (∂Ω) such that gn → µ in the sense of measures, and set un =
PΩq (gn ). From the maximum principle,

−Pgn− ≤ un ≤ Pgn+ .

Since Z
Pgn± = P (x, y)gn± (y)dHN −1 ,
∂Ω
0
we take f ∈ Lp (Ω) with p0 = p/(p − 1) and 1 ≤ p < N/(N − 1), and we have
Z Z Z
Pgn± (x)f (x)dx = P (x, y)gn± (y)dHN −1 (y)f (x)dx
Ω Ω ∂Ω
Z Z 
= P (x, y)f (x)dx gn± (y)dHN −1 (y)
Ω
∂Ω
Z
≤ Kp,Ω kf kLp (Ω) gn± (y)dHN −1 (y)
∂Ω

This implies
kPgn± kLp (Ω) ≤ Kp,Ω kgn± kL1 (∂Ω) ≤ K. (11)

Similarly
∗
kPgn± k ≤ Kp,Ω kgn± kL1 (∂Ω) ≤ K ∗ , (12)
Lp (Ω;ρdx)

for 1 ≤ p < (N + 1)/(N − 1). If we take q < p in (11), and 1 < p in (12),
q−1
we deduce that {un } and {|un | un } are uniformly integrable and therefore
Laurent Véron 317

weakly compact in L1 (Ω) and L1 (Ω; ρdx) respectively. From the Osserman-
Keller estimate {un } remains also locally uniformly bounded in Ω and by Ascoli’s
theorem there exist a sequence {unk } and a C 2 (Ω)-function u such that unk → u
locally uniformly and weakly in L1 (Ω). Moreover |unk | unk → |u| u weakly
q−1 q−1

in L1 (Ω; ρdx). Letting n → ∞ in the next exprssion

Z   Z
q−1 ∂ζ
−unk ∆ζ + |u|nk unk ζ dx = − gnk dHN −1 ,
Ω ∂Ω ∂ν

yields to (10).
The stability result is proved by the same device. If µn → µ weakly in
M(∂Ω) , it remains bounded in the total variation norm and therefore {P|µ|n }
remains bounded in Lp1 (Ω) ∩ Lp2 (Ω; ρdx) for any 1 ≤ p1 < N/(N − 1) and
1 ≤ p2 < (N + 1)/(N − 1). Since |PΩq (µn )| ≤ P|µn | , the needed compactness in
order to let {nk } in
Z   Z
q−1 ∂ζ
−unk ∆ζ + |unk | unk ζ dx = − dµnk .
Ω ∂Ω ∂ν

Therefore the full convergence result follows from uniqueness.

Remark. The problem (3) may not have a solution for any measure when
q ≥ (N + 1)/(N − 1). For example there exists no solution if µ = δa for some
a ∈ ∂Ω. The full treatment of the solvability of (3) is now well understood
thanks to the works of Dynkin and Kuznetsov ((N + 1)/(N − 1) ≤ q < 2), Le
Gall (q = 2) and Marcus and Véron (q > 2). This treatment involves the notion
of removable sets which are associated to Bessel capacities. It will be presented
in Section 9.

Remark. A more elaborated analytic tool (weighted Marcinkiewicz spaces)

allowed Gmira and Véron [17] to prove an existence and uniqueness result for
the general problem
−∆u + g(u) = 0, in Ω,
(13)
u = µ on ∂Ω,
where g is continuous and non-decreasing, µ ∈ M(∂Ω), and
Z
 
g(P|µ| )ρ(x)dx < ∞.
Ω

By a solution we mean an integrable function u defined in Ω such that g(u) ∈

L1 (Ω; ρdx) which satisfies, for every ζ ∈ C01,1 (Ω̄),
Z Z
∂ζ
(−u∆ζ + g(u)ζ) dx = − ζdµ. (14)
Ω ∂Ω ∂ν

The above integrability condition appears to be more general, however in the

particular case where g(r) = r|r|q−1 , it may not be satisfied even with µ =
318 Generalized boundary value problems

hdx ∈ L1 (∂Ω) when q ≥ (N + 1)/(N − 1). It is then natural to isolate the

singular part of the measure by writing the Lebesgue decomposition

µ = µS + µR

where µR is the regular part (with respect to the (N-1)-Hausdorff measure), and
µS , the singular one. If g satisfies a ∆2 -condition,

|g|(r + r0 ) ≤ θ(|g|(r) + |g|(r0 ))

for some θ > 0, whenever rr0 ≥ 0, Marcus and Véron [41](1996) proved the
existence of a solution (always unique) under the mere condition
Z
|g|(P|µ|S )ρ(x)dx < ∞.
Ω

The case of the exponential is different. For example if g(r) = e2r the existence
of a solution of (13) is insured if, for some p ∈ (1, ∞], there holds

exp(2PµS ) ∈ Lp/(p−1) (∂Ω),

(15)
exp(2µR ) ∈ Lp (∂Ω).

This was obtained by Grillot and Véron in 1997 [18].

5 Isolated singularities
As we have seen it above, if 1 < q < (N + 1)/(N − 1) and a ∈ ∂Ω, for any k > 0
the function uk,a = PΩq (kδa is a solution of (1) which vanishes on ∂Ω \ {a}.
Moreover, when k increases, it is the same with uk,a . From the Osserman-
Keller estimate, this sequence is locally uniformly bounded in Ω , therefore it
converges to some positive solution u∞,a of (1). By using some local estimate on
the boundary it can be checked that {uk,a } is equicontinuous on any compact
subset of Ω. Therefore u∞,a vanishes on ∂Ω \ {a}. This function u∞,a is a
solution of (1) has the maximal blow-up rate at a among the solutions vanishing
on ∂Ω \ {a}. As for the behaviour of uk,a near a it can be obtained from
perturbation theory, since the blow-up estimate on Poisson’s kernel yields to

0 ≤ PΩq (kδa ) ≤ Pkδa = kPδa = kP (x, a),

that is
−N
0 ≤ uk,a ≤ Ck|x − a| ρΩ
Finally it is possible to prove that the non-linear term is negligible near a in
some sense and
uk,a
lim = k. (16)
x→a P (x, a)

Always in the range 1 < q < (N + 1)/(N − 1), the function u∞,a has a much
stronger blow-up than uk,a . The precise expression of this blow-up is made
Laurent Véron 319

clearer by introducing spherical coordinates centered at a. In fact there exists

N −1
a unique function ω defined on the half unit sphere S+ , the equator of which
belongs to the plane Ta ∂Ω tangent to ∂Ω at the point a, such that
2/(q−1)
lim |x − a| u∞,a (x) = ω ((x − a)/|x − a|) . (17)
x→a

Moreover u∞,a is the unique solution of (1) vanishing on ∂Ω\ {a} which satisfies
(16) with k = ∞.
The following result due to Gmira and Véron [17]asserts that the behaviours
(16) and (17) characterise all the isolated singularities of the solutions of (1)
which coincide on ∂Ω \ {a} with a continuous function h ∈ C(∂Ω).

Theorem 4 Suppose ∂Ω is smooth, a ∈ ∂Ω, 1 < q < (N + 1)/(N − 1), and

u ∈ C(Ω̄ \ {a}) ∩ C 2 (Ω) is a positive solution of (1) in Ω which coincides with
h on ∂Ω \ {a}. Then
(i) either u = PΩq (h), and u is regular in Ω̄;
(ii) or there exists k > 0 such that u = PΩq (h + kδa ) and u ≈ uk,a in the
sense of (16);
(iii) or u = PΩq (h + ∞δa ) = limk→∞ u = PΩq (h + kδa ) and u ≈ u∞,a in the
sense of (17).

Remark. It is always possible to assume that u vanishes on the boundary

except the point a. Actually, if it is not the case, it follows by the maximum
principle that there exists a solution ũ of (1) vanishing on ∂Ω \ {a} and such
that
|u(x) − ũ(x)| ≤ sup |h(y)|, ∀x ∈ Ω.
y∈∂Ω

Remark. The above classification can be extended to changing sign solutions

when (N + 2)/N ≤ q < (N + 1)/(N − 1): in case (ii) there is no sign restriction
on k, and in case (iii), we have either u ≈ u∞,a or u ≈ −u∞,a .

Sketch of the proof of Theorem 4. The full proof is highly technical,

therefore we shall restrict ourselves to the case where ∂Ω is flat near a. By scaling
and translating it can be supposed that a = 0 and ∂Ω ⊃ T0 ∂Ω∩{x : |x| ≤ 1} We
shall not impose the positivity of u in order to see at what level this condition
versus (N + 2)/N ≤ q < (N + 1)/(N − 1) appears. Let (r, σ) ∈ R+ × S N −1 be
the spherical coordinates, then u solution of (1) satisfies

N −1 1
∂rr u + ∂r u + 2 ∆S N −1 u = |u|q−1 u (18)
r r
N −1 N −1
in (0, 1] × S+ , and vanishes on (0, 1] × ∂S+ . Setting t = ln(1/r) and
2/(q−1)
w=r u, then w is bounded and it satisfies
q−1
∂tt w + βq,N ∂t w + ∆S N −1 w + αq,N w = |w| w (19)
320 Generalized boundary value problems

N −1
in [0, ∞) × S+ , where
 
q+1 2 2q
βq,N = 2 − N and αq,N = −N .
q−1 q−1 q−1
N −1 N −1
Since w is bounded in [0, ∞) × S+ and vanishes on [0, ∞) × ∂S+ , it fol-
β
lows from the regularity theory of linear elliptic equations that ∇α ∂
S t w remains
uniformly bounded for any |α| + |β| ≤ 3 (here ∇S is the covariant derivative
symbol). Multiplying (19) by ∂t w gives
Z
dE
βq,N (∂t w)2 dσ = (20)
N −1
S+ dt

with
Z  
1 2 1 αq,N 2 1
|w| − (∂t w)2 dσ
q+1
E= |∇S w| + |w| −
N −1
S+ 2 q+1 2 2

But βq,N 6= 0 since q 6= (N + 2)/(N − 2). Since E is bounded,

Z ∞Z
(∂t w)2 dσdt < ∞. (21)
N −1
0 S+

By differentiating (19) with respect to t and multiplying by ∂tt w, there also

holds Z ∞Z
(∂tt w)2 dσdt < ∞, (22)
N −1
0 S+

and by using the previous regularity estimates and (21), (22), it follows

lim ∂t w = lim ∂tt w = 0, (23)

t→∞ t→∞

N −1 N −1
in L2 (S+ ) and actually
S uniformly on S̄+ . Consequently, the ω-limit set
of the trajectory T = t≥0 {w(t, .)} is included into a non-empty, compact and
connected component of the set
n o
N −1 N −1 N −1
S = ϕ ∈ C 2 (S+
q−1
) : ∆S N −1 ϕ + αq,N ϕ = |ϕ| ϕ in S+ , ϕ = 0 on ∂S+

N −1
1- If αq,N ≤ N − 1 = λ1 (S+ ), which is equivalent to q ≥ (N + 1)/(N − 1),
S is reduced to the zero function. This is easily done by multiplying by ϕ, and
N −1
integrating over S+ .
N −1
2- If 2N = λ2 (S+ ) ≤ αq,N < λ1 , which is equivalent to (N + 2)/N ≤ q <
(N + 1)/(N − 1), the set S, besides the zero function, contains two nonzero
elements, ω and −ω, which keep constant sign ((ω > 0 for example), and are
rotationally invariant.
2 ¯ N −1
3- If 1 < q < (N + 1)/(N − 1) S ∩ C+ ((S+ ) = {0, ω}
Laurent Véron 321

It follows that either if u ≥ 0 or if (N + 2)/N ≤ q < (N + 1)/(N − 1), the

function w admits a limit ` at infinity, and ` ∈ {0, ω, −ω}.
If ` = 0 a fine analysis similar to the one performed by Chen, Matano and
Véron [7] yields to the existence of some ε > 0 such that

|w(t, σ)| ≤ Ce−εt

for t ≥ 0. This type of estimate is not easy to obtain. It is obtained by

contradiction and the fact that either αq,N is not an eigenvalue of −∆S in
W01,2 (S+
N −1
), or, if it is, the equivariance properties of w are not compatible with
the ones of the element of ker(−∆S − αq,N I). From this follows an improved
estimate of the following type

|w(t, σ)| ≤ C min(e−θεt , e(N −1−2/(q−1))t )

where θ = θ(N, q) > 1. This estimate is of linear type and easy to obtained by
a representation formula. Since θ, a final estimate of the type

|w(t, σ)| ≤ Ce(N −1−2/(q−1))t

is derived in a finite number of iterations. The projection of w onto ker(−∆S −

(N − 1)I) yields to the existence of some ϕ ∈ ker(−∆S − (N − 1)I) such that

lim e(2/(q−1)−N +1)t w(t, .) = ϕ(.)

t→∞

N −1
in the C 2 (S+ -topology. Finally, if ϕ = 0, comparison of u with ±εP (., 0)
implies that u is actually the zero-function.

Remark. If the assumptions on q and the sign of u are relaxed, the only
thing which can be proved is that the limit set of the trajectory T is included
into a connected component of the set S. However, in such cases exept for the
constant-sign solutions, and the zero function, the connected components of S
are continuum. Therefore it is not possible to assert that w converges precisely
to a particular element.

6 Large solutions
Let Ω be a domain with a compact boundary. By a large solution of (1) we
mean a positive solution u such that

lim u(x) = ∞. (24)

ρ(x)→0

Since it is classical to approximate Ω by an increasing sequence Ωn of smooth

subdomains, on each of them we can construct a positive solution un of (1) in
Ωn with infinite value on the boundary , we obtain a decreasing sequence of
solutions of (1), each of them dominating in Ωn any solution u of (1). Therefore
322 Generalized boundary value problems

the sequence {un } converges to the maximal solution ūΩ of (1) in Ω (this is
essentially the results of Keller and Osserman). The questions which are raised
are therefore
1- Existence of large solution or, equivalently, is the maximal solution a large
solution? This question is linked to the regularity of ∂Ω.
2- Uniqueness of the large solution.
The first question which is a problem linked to the regularity of ∂Ω, and it has
been solved in the case q = 2 by Dhersin and Le Gall [8] , by probabilistic
methods. If a ∈ ∂Ω, they proved that a Wiener-type criterion at the point a,
associated to some Bessel capacity is a necessary and sufficient condition for the
existence of a solution u of (1) such that

lim u(x) = ∞.
x→a

Their methods heavily relies on probability theory and the question remains
open to cover the full range of exponent q > 1. The following easy to prove
result is essentially due to Marcus and Véron [28] (although not explicitely
written in this reference, it clearly corresponds to their cone condition with
dimension 0).

Theorem 5 Suppose 1 < q < N/(N − 2), then there exists a large solution of
(1) in Ω.

Proof. At every point a ∈ ∂Ω the maximal solution ūΩ constructed above is

bounded from below by the explicit positive singular solution Ua of (1)
1/(q−1) 2/(1−q)
x 7→ Ua (x) = αq,N |x − a| . (25)

Notice that the existence of Ua is insured by the positivity of αq,N , which is

equivalent to 1 < q < N/(N − 2).
When q ≥ N/(N − 2) large solutions may exist provided ∂Ω satisfies some
weak geometric constraint. For example Marcus and Véron [28] proved

Theorem 6 Suppose 1 < q < (N − 1)/(N − 3), and Ω satisfies the exterior
segment property, then there exists a large solution of (1) in Ω.

The exterior segment property asserts that there exists ε > 0 such that, for
any a ∈ ∂Ω, there exists a segment Iε,a ⊂ Ωc with length ε and a as one of its
end-points. It is also proved in [28] that if q ≥ N/(N − 2), the blow-up rate of
the maximal solution may be much weaker than the one given by (25).
The first result of uniqueness of the large solution are due to Iscoe [19],
Bandle and Marcus [1, 2] and Véron [42]. The technique used by Bandle and
Marcus or Véron needs a C 2 -regularity of ∂Ω and relies on a precise blow-up
estimate of any large solution u at the boundary, namely

lim ρ2/(q−1) (x)u(x) = C(q) > 0 (26)

ρ(x)→0
Laurent Véron 323

From this estimate and the maximum principle it follows

u ≤ (1 + ε)ũ in Ω,
if u and ũ are two maximal solutions, and this holds for any ε > 0. Letting
ε → 0 and exchanging the role of u and ũ implies u = ũ. The result of Iscoe,
only stated in the case q = 2 gives uniqueness without the possible existence in
the case where Ω is star-shapped with respect to some point, say 0. It use the
equi-variant properties of equation (1)under the transformation

u 7→ uk , where uk (x) = k 2/(q−1) u(kx), ∀k > 0.

If u and ũ are two large solutions in Ω, then uk is a large solution in Ωk = k −1 Ω.

If k > 1, Ωk ⊂ Ω and therefore uk ≥ ũ. Letting k → 1 and exchanging again
the role of u and ũ implies uniqueness.
In 1995 Marcus and Véron [28] proved a uniqueness result extending Iscoe’s
one to a wide class of domains

Theorem 7 Suppose ∂Ω is locally the graph of a continuous function, then

there exists at most one large solution of (1) in Ω.

The proof of this result is not easy. The key point is that the assumption on
∂Ω implies that any two large solutions u and ũ satisfy

u(x)
lim = 1. (27)
ρ(x)→0 ũ(x)

Finally, in 1997 Marcus and Véron [29] introduced a new technique for prov-
ing uniqueness of a wide class of solutions of (1) with positively homogeneous
boundary data. In the case of uniqueness of large solutions, their results is the
following

Theorem 8 Suppose 1 < q < N/(N − 2) and ∂Ω = ∂ Ω̄c , then there exists
exactely one large solution of (1) in Ω.

Proof. The meaning of positively homogeneous boundary data is that they

are unchanged by any multiplication by positive real numbers: they take only
two possible values, 0 and ∞. Existence of the maximal solution ūΩ follows
from Theorem 5. Let u be another large solution and assume u 6= ūΩ . Then
u < ūΩ .
Step 1 There exists a constant K = K(N, q) > 1 such that

u(x)
K −1 ≤ ≤ K, ∀x ∈ Ω. (28)
ūΩ (x)

Since ∂Ω = ∂ Ω̄c ; for any a ∈ ∂Ω, there exists a sequence {an } ⊂ Ω̄c such that
an → a when n goes to infinity. Because u blows-up on ∂Ω and the explicit
324 Generalized boundary value problems

solution Uan (defined in (25)), is finite, Uan ≤ u, and letting n go to infinity,

Ua ≤ u in Ω. If x is any point in Ω, let a ∈ ∂Ω such that ρΩ (x) = |x − a|. Then

1/(q−1) 2/(1−q) 2/(1−q)

u(x) ≥ Ua (x) = αq,N |x − a| = αq,N ρΩ (x).

But, by the Keller-Osserman estimate (8),

2/(1−q)
ūΩ (x) ≤ C(q, N )ρΩ (x).

1/(q−1)
Consequently (28) holds with K(N, q) = C(q, N )/αq,N ).
Step 2 End of the proof. Set

1
w =u− (ū − u).
2K
Because of Step 1,
 
1 1
+ u < w < u,
2 2K
which implies in particular that w blows-up on ∂Ω. Moreover
    q
1 1 q 1 1
∆w = 1+ uq − ū ≤ 1+ u− ū = wq
2K 2K 2K 2K

which means that w is a super solution for (1). Since ( 12 + 2K

1
)u is a subsolution,
there exists a large solution u1 of (1) which satisfies
   
1 1 1
+ u < u1 < u and ūΩ − u1 ≥ 1 + (ūΩ − u)
2 2K 2K

in Ω (see [35] for an a proof of this classical result under these conditions).
Iterating this process, we consruct a sequence {un } of large solutions of (1),
which always satisfy u0 = u,
ūΩ ≤ Kun ,

and
   
1 1 1
+ un−1 < un < un−1 and ūΩ − un ≥ 1 + (ūΩ − un−1 ).
2 2K 2K

Consequently
 n
1
ūΩ − un ≥ 1+ (ūΩ − u).
2K
Because ūΩ − u is locally bounded in Ω, we derive a contradiction by letting n
go to infinity.
Laurent Véron 325

Remark. It is clear that uniqueness cannot hold if Ω = Ω̃∪F , where F = {aj }

is finite. In that case, ∂Ω = ∂ Ω̃ ∪ F and a large solution u of (1) in Ω may have
a weak blow-up such as
u(x) = c|x − aj |2−N (1 + o(1)
for any c > 0 near aj (we assume N > 2, the cases N = 2, 1 needing only very
minor modifications). However, it has been noticed by Chasseigne and Vazquez
[6] that a large solution u is unique if one imposes the strongest blow-up on ∂Ω,
that is
lim ρ(x)N −2 u(x) = ∞.
ρ(x)→0

7 The boundary trace

For simplicity of exposition, in the next sections, we shall restrict ourself to
the case where Ω = B = {x ∈ RN : |x| < 1}, although all the results we
shall propose extend to a C 2 bounded domain. We recall that (r, σ) are the
spherical coordinates in RN . It is classical, and easy to check, that every positive
harmonic function u in B possesses a boundary trace given by a positive Radon
measure µ ∈ M(∂B) which is attained in the sense of weak convergence of
measures: Z Z
lim u(r, σ)ζ(σ)dσ = ζ(σ)dµ (29)
r→1 S N −1 S N −1
N −1
for every ζ ∈ C(S ) and u = Pµ . In this writing, we identify ∂B and S N −1 .
The trace result is still valid if harmonic is replaced by super-harmonic [9] and
a representation formula holds. Moreover, the positivity assumption on u can
be replaced by an integrability condition such as ∆u ∈ L1 (B; (1 − r)dx) and in
that case the boundary trace is a general Radon measure on ∂B.
The existence of a boundary trace for positive solutions of (1) is due to Le
Gall in 1993 [24, 25] when q = 2. Actually, in this pioneering work, Le Gall gives
a probabilistic representation of any positive solution of (1) in this case. The
notion of trace that we present is due to Marcus and Véron [29](1995); it is a
purely analytic presentation and is extendible to a wider class of nonlinearities.
Theorem 9 Suppose q > 1 and u is a positive solution of (1) in B. Then for
any ω ∈ ∂B we have the following alternative.
(i) Either for every relatively open neighbourhood U ⊂ ∂B of ω
Z
lim u(r, σ)dσ = ∞, (30)
r→1 U

(ii) or there exists a relatively open neighbourhood U ⊂ ∂B of ω such that for

every ζ ∈ C ∞ (U ) Z
lim u(r, σ)ζ(σ)dσ = `(ζ), (31)
r→1 U
where ` is a positive linear functional on C ∞ (U ).
326 Generalized boundary value problems

If V is an open domain of S N −1 , we denote by ϕV the first eigenfunction of −∆

in W01,2 (V ) with the normalization condition

max ϕV = 1.
V

The corresponding eigenvalue is quoted λV , and if the relative boundary of V

is C 2 , the Hopf lemma applies with νS the normal outward unit vector to V
(tangent to S N −1 ),
∂ϕV
< 0 on ∂V.
∂νS

Lemma 2 Let V be an open domain of S N −1 with a C 2 boundary, u a positive

solution of (1) in B and α > (N + 1)/(N − 1) . Then we have the following
dichotomy.
(i) Either
Z 1Z
V (1 − r)r drdσ = ∞,
uq ϕα N
0 V
and in that case Z
lim V )(r, σ)dσ = ∞,
(uϕα
r→1 V

(ii) or
Z 1 Z
V (1 − r)r drdσ < ∞,
uq ϕα N
0 V

and in that case, for any C 2 function on V which satisfies

|ζ| ≤ kϕα
V and |∆ζ| ≤ kϕα−2
V (32)

for some k > 0, the following limit exists

Z
lim u(r, σ)ζdσ = `(ζ),
r→1 V

and the mapping ζ 7→ `(ζ) is a positive linear functional defined on the subset
of C 2 functions which satisfy (32).

Proof. Step 1. There holds

Z
 N −1 α q/(q−1) −qα/(q−1)
Iα = ∆ ϕV  ϕV dσ < ∞.
S
V

From Hopf boundary Lemma

ϕV (σ) ≥ C1 distSN−1 (σ, ∂V )

for any σ ∈ V , where distSN−1 is the geodesic distance on S N −1 and C1 > 0.

Since
−1 α 2
∆N
S ϕV = −αλV ϕα V + α(α − 1)ϕV
α−2
|∇ϕV | ,
Laurent Véron 327

then  N −1 α q/(q−1)
q(α−2)/(q−1)

∆ ϕV  ≤ C2 distSN−1 (., ∂V ) ,
S

and finally
 N −1 α q/(q−1) −qα/(q−1)
q((α−2)−α)/(q−1)
∆ ϕV  ϕV ≤ C3 distSN−1 (., ∂V ) .
S

But α > (N + 1)/(N − 1), and the claim follows follows.

Step 2 Reduction of the equation. We shall assume N > 2 , the case N = 2
requiring only minor technical modifications. We transform the equation (18)
satisfied by u in polar coordinates by setting

rN −2
s= and u(r, σ) = r2−N v(s, σ),
N −2
and we get

∂2v 1
s2 + ∆ N −1 v − KsN/(N −2)−q v q = 0, (33)
∂s2 (N − 2)2 S

in (0, a) × S N −1 , where K = K(N ) > 0 and a = (N − 2)−1 , from which follows

Z Z Z
d2 1
s2 2 vϕα dσ + v∆ S N −1 ϕα
dσ − Ks N/(N −2)−q
v q ϕα
V dσ = 0.
ds V V
(N − 2)2 V V
V

We set Z Z 
X(s) = vϕα
V dσ and Y (s) = v q
ϕα
V dσ .
V V

Since, by Hölder’s inequality and Step 1,

Y (s) ≤ Iα1−1/q Y (s),

it infers
(i) s2 X 00 + JY − KsN/(N −2)−q Y q ≥ 0
(34)
(ii) s2 X 00 − JY − KsN/(N −2)−q Y q ≤ 0
where J = I 1−1/q (N − 2)−2 .
R1R
V (1 − r)r drdσ = ∞.
Step 3 End of the proof. Case 1: suppose 0 V uq ϕα N

Since (34-i) implies

X 00 ≥ AY q − B
on [a/2, a), where A > 0 and B > 0 do not depend on s, a double integration
gives
Z s
0
X(s) ≥ X(a/2) + (s − a/2)X (a/2) + A (s − τ )Y q (τ )dτ − B(s − a/2)2 /2,
a/2

and lim X(s) = ∞, which is assertion (i).

s→a
328 Generalized boundary value problems

R1R Ra
Case 2: suppose 0 V uq ϕα V (1 − r)r drdσ < ∞. Then a/2 (a − s)Y ds < ∞.
N q

Moreover X 00 ≤ AY q + B, and
Z s !
d2 B 2
X(s) + A (s − τ )Y (τ )dτ − (s − a/2)
q
≤ 0,
ds2 a/2 2

2
which infers that s 7→ X(s) admits
R a finite limit at infinity. If ζ is a C -function
satisfying (32), we set Xζ (s) = V vζdσ and derive
Z Z
d2 Xζ 1
s2 + v∆S N −1 ζdσ − Ks N/(N −2)−q
v q ζdσ = 0 (35)
ds2 (N − 2)2 V V

from (33). But

Z  Z 1/q Z 1−1/q
 
 v∆S N −1 ζdσ  ≤ k q α
v ϕV dσ
α−2q/(q−1)
ϕV dσ
 
V V V

and Z  Z
 
 v q ζdσ  ≤ k v q ϕα
  V dσ.
V V

Consequently
Z a Z  Z Z 
  a  
 v∆S N −1 ζdσ (a − s)ds < ∞ ,  v q ζdσ (a − s)ds < ∞.
   
a/2 V a/2 V

Integrating twice the equality (35) implies that s 7→ Xζ (s) admits a finite limit
at infinity and this limit depends linearly of ζ. moreover it is nonnegative if
such is the case of ζ. We call this limit `(ζ).

Proof of Theorem 9. If ω ∈ S N −1 and there exists an open neighbourhood

V such that Lemma 1-i holds, there existence of a nonegative Radon measure
µV such that
Z Z
lim u(r, σ)ζ(σ)dσ = ζdµ ∀ζ ∈ Cc (V ) (36)
r→1 V V

If such a neighbourhood does not exist, we are in case (i).

Let R be the set of the ω ∈ S N −1 such that (ii) holds; R is relatively open
and there exists a unique (non-negative) Radon measure µ such that µ V = µV .
The set S = S N −1 \ R is closed.

Definition. The couple (S, µ) is called the boundary trace of u. The set S is
the singular part of this trace and the measure µ on R, the regular part of the
trace. We denote
tr∂B (u) = tr(u) = (S, µ).
Laurent Véron 329

For convenience it is often useful to introduce the Borel measure framework.

Actually there is a one to one correspondence between the family CM of couples
(S, µ) where S is a compact subset of §N −1 and µ a positive Radon measures on
R = S N −1 \ S and the set Breg +
(S N −1 ) of outer regular, positive Borel measures
on S N −1 . If β ∈ Breg
+
, the singular set Rβ and the singular part Sβ are defined
as follows:

Rβ = {ω ∈ S N −1 : ∃U rel. open neighborhood of ω s.t. β(U ) < ∞},

Sβ = {ω ∈ S N −1 : ∀U rel. open neighborhood of ω s.t. β(U ) = ∞}.

+
The correspondence CM ↔ Breg (S N −1 ) is given by

µ(A) if A ⊆ R,
M((S, µ)) = µ̄ where µ̄(A) = (37)
∞ if A ∩ S =
6 ∅,

for any Borel subset A of S N −1 , and

M−1 (β) = (Sβ , β Rβ ). (38)

With this correspondence, we denote

+
Tr(u) = M(tr(u)) ∈ Breg (S N −1 ),

and a more general formulation for (3) is therefore what we call the generalized
Dirichlet problem, namely

−∆u + uq = 0, in B,
+ (39)
Tr(u) = µ̄ ∈ Breg (S N −1 ).

The problem is completely different according 1 < q < (N + 1)/(N − 1) (the

subcritical case) or q ≥ (N + 1)/(N − 1) (the supercritical case).

8 Generalized Dirichlet problem: the subcriti-

cal case
The following estimate links the local integral blow-up estimate corresponding
to singular boundary points and pointwise blow-up estimate.

Theorem 10 Suppose 1 < q < (N + 1)/(N − 1), u is a non negative solution

of (1) in B with tr(u) = (S, µ) and ω ∈ S. Then

lim u(r, ω) = ∞,
r→1

and more precisely

u(x) ≥ Uω (x), ∀x ∈ B, (40)
where Uω is defined in (25).
330 Generalized boundary value problems

Proof. A scaling-concentration argument is used. Since ω ∈ S, for any η > 0

Z
lim u(r, σ)dσ = ∞,
r→1 Dη (ω)

where Dη (ω) = {σ ∈ S N −1 : distSN−1 (σ, ω) < η} We recall that uk (x) =

k 2/(q−1) u(kx), and uk is a soluton of (1) in B1/k = k −1 B. For 0 < ε < 1 we
denote Z
Mε,η = u(1 − ε, σ)dσ.
Dη (ω)

For m > 0 large enough and η small enough there exists ε = ε(η, m) > 0 such
that m = Mε,η . Let χDη (ω) be the characteristic function of the set Dη (ω) and
wη be the solution of (1) in B with boundary value u1−ε (1, .)χDη (ω) . Clearly
wη ≤ u1−ε . Since limη→0 ε = 0, then limη→0 wη = PBq (mδω ) = um,ω . Moreover
limε→0 u1−ε = u. Therefore um,ω ≤ u, for any m > 0, which implies (40).

Remark. The proof below cannot work in the supercritical case, and actually
the result does not hold as we shall see it below.
It is proved by Le Gall [24, 25] in the case N = 2 = q, and by Marcus and
Véron [29] in the general case, that the boundary data problem (39) is well
posed. The method of Le gall is mainly a probabilistic one while Marcus and
Véron use only analytic tools.

Theorem 11 Suppose 1 < q < (N + 1)/(N − 1). Then the correspondence

u 7→ Tr(u) which assigns to each nonnegative solution u of (1) in B, its boundary
+
trace in Breg (S N −1 ) is one to one.

Since the proof is long and very technical, we shall only give a sketch of it,
pointing out the main steps.
Lemma 3 There exits a minimal solution uS,µ .

Sketch of the proof. Let {yk }k∈N be a dense sequence in S and

Sε = {σ ∈ S N −1 : distSN−1 (σ, S) ≤ ε}.

We set
X
n
µn = n δyk + χS1/n
c µ.
k=0

Since the closure of S1/n

c
is a compact subset of of S c , µn is a bounded Radon
measure. By comparison principle and Theorem 3, the sequence {un } = {PBq (µn )}
is increasing. Moreover, it follows from Lemma 1 and Theorem 10 that if u is
any solution of (39), then un ≤ u. When n → ∞, un increases and converges to
the minimal solution uS,µ .

Lemma 4 There exists a maximal solution uS,µ .

Laurent Véron 331

Sketch of the proof. We denote by µn the Borel measure with singular

set S1/n and regular part µn = χS c1/n µ Since µn is the increasing limit of the
sequence of Radon measures {gn,k } defined by
gn,k = kχS 1/n + χS c µ,
1/n

we first construct the solution un,k = PBq (gn,k )

of (1) and after (as k → ∞)
a solution un of (39) with Tr(un ) = µn . Moreover the sequences {un } is non-
increasing, and, if u is any solution of (39), then un ≥ u. When n → ∞, un
decreases and converges to the maximal solution uS,µ .
Lemma 5 There always holds
uS,µ − uS,µ ≤ uS,0 − uS,0 .

Sketch of the proof. For r, s > 0 set

 rq −sq
r−s if r 6= s,
Φ(r, s) =
qsq−1 if r = s.
By convexity,
 
r0 ≥ s0 , r1 ≥ s1
=⇒ Φ(r1 , s1 ) ≥ Φ(r0 , s0 ).
r1 ≥ r0 , s1 ≥ s0 ,
We set (with the notations of Lemmas 3-4)
un = un,µ and un = un,µ ,
and
Zn,µ = un,µ − un,µ and λn,µ = Φ(un,µ , un,µ ).
Then
∆Zn,µ = λn,µ Zn,µ
and
∆(Zn,µ − Zn,0 ) − λn,µ (Zn,µ − Zn,0 ) = (λn,µ − λn,0 )Zn,0 .
Since
un,µ ≥ un,µ , un,µ ≥ un,0 and un,0 ≥ un,0 , un,µ ≥ un,0 ,
λn,µ ≥ λn,0 . Therefore
∆(Zn,µ − Zn,0 ) − λn,µ (Zn,µ − Zn,0 ) ≥ 0.
Moreover Zn,µ = Zn,0 on ∂B. Therefore
Zn,µ ≤ Zn,0 in B.
Actually an approximation argument is needed to take into account the fact
that the solutions have infinite values in S 1/n . Conclusion follows by letting
n → ∞.
Lemma 6 There exits L = L(N, q) > 1 such that
uS,0 ≤ LuS,0 .
332 Generalized boundary value problems

Sketch of the proof. Case 1 Suppose ω ∈ S. Since

Uω (r, ω) ≤ uS,0 (r, ω) ≤ uS,0 (r, ω) ≤ ūB (r, ω)

where ūB is the maximal solution of (1) in B, and

Uω (r, ω) ≥ K(N, q)ūB (r, ω),

the estimate holds.

Case 2 Suppose ω ∈ R. The way for obtaining this estimate is a bit more tech-
nical, however, if we assume that B is replaced RN + and ∂B by the hyperplane
H = RN −1 , the argument goes as follows. By a change of coordinates, it can be
assumed that 0 = ω is the origin. Let Dρ ⊂ H be the largest open (N-1)-ball
with center 0 such that Dρ ⊆ R (ρ is finite since S 6= ∅), and let a ∈ ∂Dρ ∩ S.
If we denote x = (x1 , x0 ) the current point in R+ × RN −1 , then

u∞,ω (x1 , 0) ≤ uS,0 (x1 , 0) ≤ uS,0 (x1 , 0) ≤ uDρc ,0 (x1 , 0).

In order to estimate the bounds of the quotient

u∞,ω (x1 , 0)
Qρ (x1 ) = ,
uDρc ,0 (x1 , 0)

when x1 runs from 0 to ∞, a scaling argument shows that it can be assumed

that ρ = 1. For small values of x1 the lower bounds is positive and given by
Hopf’s lemma, while for large values of x1 , the lower bound follows from the
fact that
2/(q−1)
lim x1 u∞,ω (x1 , 0) = ω(x1 /|x1 |),
x1 →∞

(in this formula ω(.) is the spherical function defined in (17)) while
 1/(q−1)
2/(q−1) 2(q + 1)
lim x1 uDρc ,0 (x1 , 0) = .
x1 →∞ (q − 1)2

Consequently there exists L1 > 0 such that

uS,0 (x1 , 0) ≤ LuS,0 (x1 , 0).

In the real situation where we are dealing with B and not with RN + , we proceed
by contradiction and scaling, reducing the situation to the half space case.

Proof of Theorem 10. It follows the proof of Theorem 8 (in a simplified way).
By Lemma 5 it is sufficient to prove uniqueness in the case where tr(u) = (S, 0).
Assuming that
uS,0 6= uS,0 =⇒ uS,0 < uS,0 ,
then
1
w = uS,0 − (uS,0 − uS,0 )
2L
Laurent Véron 333

is a supersolution of (1) dominated by uS,0 , but dominating the subsolution

(1/2 + 1/(2L))uS,0 . Therefore there exists a solution ũ of (1) such that

1 1
( + )u ≤ ũ ≤ w < uS,0 .
2 2L S,0
It follows that tr(ũ) = (S, 0), which contradicts the minimality of uS,0 .

9 Generalized Dirichlet problem: the supercrit-

ical case
As we have seen it in Sections 3-4, neither any Radon measure, nor any closed
subset of ∂B are eligible for being the regular or the singular part of the bound-
ary trace (S, µ) of a positive solution of (1) inB. Roughly speaking, a Radon
measure is admissible if it does not charge too thin sets, while a closed subset
is admissible for being a singular set if it is not too ramified. Moreover there is
a compatibility requirement between S and µ. Those different notions will be
made rigourous with the help of boundary Bessel capacities. The results that we
present are due to Le Gall [25] in the case q =2, Dynkin and Kuznetsov in the
case 1 < q < 2 [11, 12, 13]and Marcus and Véron [32]in the case q > 2. Because
of the technical difficulties of the various aspects of the theory of supercritical
boundary trace we shall essentially restrict ourself to present the main results
in the case q > 2, with some ideas of how to obtain them. Up to now a unified
proof covering all the cases q ≥ (N + 1)/(N − 1) is still missing.

Removable sets
Defintion. (i) A subset E ⊂ ∂B is said q-removable if any non-negativefunction
u ∈ C 2 (B) ∩ C(B̄ \ E) which satisfies (1) in B and vanishes on ∂B \ E is iden-
tically zero.
(ii) A subset E ⊂ ∂B is said conditionally q-removable if any non-negative
function u ∈ C 2 (B) ∩ C(B̄ \ E) which satisfies (1) in B and vanishes on ∂B \ E
is such that u ∈ Lq (B; (1 − r)dx).
The main result concerning the removability of set is the following

Theorem 12 Assume q ≥ (N + 1)/(N − 1) and E ⊂ ∂B is a Borel set. Then

the following assertions are equivalent.
(i) E is q-removable
(ii) E is conditionally q-removable
(iii) C2/q,q0 (E) = 0.
Moreover an arbitrary set A ⊂ ∂B is q-removable if and only if every closed
subset of A is q-removable.

In the statement of the Theorem, C2/q,q0 (E) denote the Bessel capacities of
order 2/q and exponent q. In the space Rd the Bessel capacities Cα,p (E) (α > 0,
334 Generalized boundary value problems

1 < p < ∞) is defined by

Z 
Cα,p (E) = inf f p dx : f ≥ 0, Gα ∗ f ≥ 1 on E
Rd

where Gα denotes the Bessel potential of order α. Another way to define ca-
pacities is to use Besov spaces. For h ∈ Lp (Rd ) 0 < α < 1 and 1 ≤ p, q ≤ ∞,
put
−d/q−α
Tαp,q (h)(t) = |t| kh(t + .) − h(.)kLp
for t ∈ Rd and denote

Bαp,q (Rd ) = h ∈ Lp (Rd ) : Tαp,q (h) ∈ Lq (Rd ) ,

with norm
p,q
khkBαp,q = khkLp + kT kα (h)Lq .
When α = 1, Tαp,q (h) is replaced by

T1p,q (h) = |t|−d/q−1 kh(t + .) + h(−t + .) − 2h(.)kLp

When α > 1 the spaces are defined by induction. From this spaces (which
coincide with the classical Sobolev spaces when α is not an integer and p = q),
we define the capacity of a compact subset E ∈ Rd by CBαp,q (E) by

CBαp,q (E) = inf kf kBαp,q : f ∈ S(Rd ), f ≥ 1 on E .

The CBαp,q -capacity of an open set is defined by the supremum of the capacities
of its compact subsets, and the capacity of a general set by the infimum of the
capacities of the open sets in which it is contained.
When α > 0, 1 < p < ∞, and p ≤ d/α the following equivalence holds: there
exists K = K(d, p, α) > 0 such that for any subset of Rd ,

K −1 Cα,p (E) ≤ CBαp,p (E) ≤ KCα,p (E).

On a submanifold of RN , the capacity is defined by using local charts.

In the forthcoming lemmas, we give the strategy for proving Theorem 12.

Lemma 7 Suppose that q ≥ (N + 1)/(N − 1) and let E ⊂ ∂B be compact and

such that C2/q,q0 (K) = 0. Then K is conditionally q-removable.

Sketch of the proof. In the case q > 2 Marcus and Véron [32] proceed as
follows: If η ∈ C 2 (∂B) is such that 0 ≤ η ≤ 1 is identically 1 is a neighborhood
0
of K set ζ = (1 − Pη )2q ϕ1 , where ϕ1 is the positive first eigenfunction of −∆
in W01,2 (B). Then it is proved by approximation that
Z
(−u∆ζ + uq ζ)dx = 0. (41)
B
Laurent Véron 335

From (41) and Hölder’s inequality,

Z Z 1/q Z 1/q0
0 q0
uq ζdx ≤ uq ζdx ζ −q /q |∆ζ| dx .
B B B

Using sharp regularizing estimates of the Poisson potential,

Z
0 q0
ζ −q /q |∆ζ| dx ≤ C1 kηkW 2/q,q0 + C2 .
B

From the assumption on K, there exists a sequence of functions ηk ∈ C 2 (∂B)

such that 0 ≤ ηk ≤ 1 and kηk kW 2/q,q0 → 0 as k → ∞. Since this implies in
particular that 1 − Pηk → 1, it follows
Z
uq ϕ1 dx ≤ C2 ,
B

which implies the claim.

Lemma 8 Any conditionally q-removable closed subset E ⊂ ∂B is q-removable.

Proof. If E is not q-removable, there exists a nonnegative nonzero solution

u of (1) vanishing on ∂B \ E. Therefore T r(u) is a Borel measure which is
zero outside E. Since E is conditionally q-removable, uq ∈ L1 (B; (1 − r)dx).
Therefore the singular part of the boundary trace of u is empty and
T r(u) = µ ∈ M+ (∂B).
For n ≥ 1, ũn = nu is a supersolution of (1) with boundary trace nµ. Therefore
there exists a nonnegative solution un of (1) such that T r(un ) = nµ, and in
particular Z Z
∂ϕ1
(λ1 un + uqn )ϕ1 dx = −n dµ
B ∂B ∂ν
Letting n → ∞ implies that the increasing sequence {un } converges to some
solution u∞ in B with T r(u∞ ) = 0 outside E. Moreover,
Z Z
(λ1 u∞ + u∞ )ϕ1 dx = ∞ =⇒
q
uq∞ ϕ1 dx = ∞.
B B
1
This contradicts the fact that uq∞ ∈ L (B; (1 − r)dx) because E is conditionally
q-removable.

Remark. It is clear that if a subset E ∈ ∂B is q-removable, it is conditionnaly

q-removable.
By sharp linear estimates on the Poisson potential, the following result holds
Lemma 9 Suppose q ≥ 2, then µ 7→ PB (µ) maps M+ (∂B) ∩ W −2/q,q (∂B) into
Lq (B; (1 − r)dx), and there holds
kPµ kLq (B;(1−r)dx) ≤ C(q)kµkW −2/q,q (∂B) .
336 Generalized boundary value problems

As an important consequence we have

Corollary 1 Suppose q ≥ 2 and µ ∈ M+ (∂B) ∩ W −2/q,q (∂B). Then there
exists u = PBq (µ).

Proof. By Lemma 9, Pµ ∈ Lq (B; (1 − r)dx). For k > 0 we denote {uk } the

solution of
−∆uk + (min uk , k)q = 0, in B,
uk = µ on ∂B,
Clearly uk ≤ Pµ which implies that the sequence {(min uk , k)q } is uniformly
integrable in L1 (B; (1 − r)dx). When k increases, uk decreases and converges
to some u which belongs to L1 (B) ∩ Lq (B; (1 − r)dx). Letting k → ∞ in the
weak formulation (which is valid for every ζ ∈ C01,1 (B̄))
Z Z
∂ζ
(−uk ∆ζ + ζ(min uk , k)q )dx = − dµ,
B B ∂ν
infers that u = PBq (µ).
From a dual definition of the C2/q,q0 -capacity of a closed subset E ⊆ ∂B,
follows the implication

C2/q,q0 (E) > 0 =⇒ ∃µ ∈ M+ ∩ W −2/q,q (∂B) s.t. µ(E) = µ(∂B) > 0. (42)

Consequently there holds

Lemma 10 Suppose q ≥ max(2, (q + 1)/(q − 1). If E ⊆ ∂B is closed and such

that C2/q,q0 (E) > 0, then E is not conditionally q-removable.

Proof. By Lemma 9 there exists µ ∈ M+ ∩ W −2/q,q (∂B) such that µ(E) =

µ(∂B) > 0. By Corollary 1, u = PBq (µ) exists and tr(u) = (∅, µ), which implies
in particular that u vanishes on ∂B \ E.
The proof of Theorem 12 is completed by the next result whose proof is too
technical to be presented here.

Lemma 11 A set A ⊂ ∂B is q-removable if and only if every closed subset of

A is q-removable.

q-traces
Definition. A Radon measure µ on ∂B is called a q-trace if there exists a
solution u of (1) such that u = PBq (µ). The set of q-traces is denoted by
Mq (∂B).
The main result concerning q-traces is

Theorem 13 Assume q ≥ (N + 1)/(N − 1). A measure µ on ∂B is a q-trace

if and only if for any Borel subset A ⊆ ∂B

µ(A) = 0 whenever C2/q,q0 (A) = 0. (43)

Laurent Véron 337

Besides what has been proved in the preceding subsection, the two next
lemmas are needed, which both are mere adaptations of results of Baras and
Pierre [3], and Meyers [33].
Lemma 12 Let µ ∈ M ∩ W −2/q,q (∂B). Then µ does not charge the sets with
C2/q,q0 -capacity zero.

Lemma 13 Suppose µ ∈ M+ (∂B) does not charge the sets with C2/q,q0 -capacity
0
zero. Then there exists a sequence {µn } ⊂ W 2/q,q (∂B) ∩ M+ (∂B), such that
{µn } converges in increasing to µ.

Sketch of the proof of Theorem 13. We shall restrict ourselves to the

cases q ≥ 2 and the measures are nonnegative, the general case needing some
approximation argument. Let µ ∈ Mq+ (∂B) and E ⊂ ∂B be a Borel subset such
that C1/q,q0 (E) = 0. Set µE = χE µ. Since 0 ≤ µE ≤ µ, the construction given
in Corollary 1 yields µE ∈ Mq+ (∂B) and PBq (µE ) ≤ PBq (µ). But C1/q,q0 (E) = 0
and Theorem 12 implies that E is q-removable. therefore PBq (µE ) = 0 and
consequently µE = 0.
Conversely, let µ ∈ M+ (∂B) which does not charge the sets with C2/q,q0 -
capacity zero. Then there exists an increasing sequence of positive measures
{ρn } belonging to W −2/q,q (∂B) and converging to µ. Then ρn ∈ Mq+ (∂B), and
the sequence {un } = {PBq (ρn )} is increasing and converges to some u. Moreover
Z Z
∂ϕ1
(λ1 un + un )ϕ1 dx = −
q
dρn . (44)
B ∂B ∂ν

Because the right-hand side of (44) is convergent, the same holds for the left-
hand side (by the Beppo-Levi theorem). Therefore u ∈ Lq (B, (1−r)dx)∩L1 (B),
which is sufficient to derive that u = PBq (µ) from the weak formulation of the
fact that un = PBq (ρn ).

The generalized Dirichlet problem

+
Given a Borel measure µ̄ ∈ Breg (∂B) with M−1 (µ̄) = (S, µ) ∈ CM , we recall
that
Sε = {σ ∈ S N −1 : distSN−1 (S N −1 , S) ≤ ε}.
and
µε = χSεc µ.
Let uSε be the maximal solution of (1) in B with tr(uSε ) = (Sε , 0). Because of
the construction it is easy to see that

0 < ε < δ =⇒ uSε ≤ uSδ .

When ε → 0, uSε converges locally uniformly in B to a nonnegative solution u∗

of (1). Let Sq∗ be the singular part of the boundary trace of u∗ . Clearly Sq∗ ⊆ S.
338 Generalized boundary value problems

If R = ∂B \ R and µ ∈ M+ (R) is such that µK = χK µ is a q-trace for

any compact subset K ⊂ R, we denote uK S = PBq (µK ). If Kn is an increasing
sequence of compact subset of R such that Kn = R, the sequence {uKn } is
increasing and converges to a solution ǔ of (1). Let ∂ν S be the singular part of
the boundary trace of ǔ. Again it is easy to verify that ∂ν S ⊆ S.
The following result is the main result concerning the solvability of (39) in
the supercritical case.
+
Theorem 14 Assume q ≥ (N + 1)/(N − 1) and let µ̄ ∈ Breg (∂B), with regular
set R = Rµ̄ , regular part µ = µ̄ Rµ̄ and singular part S = Sµ̄ = ∂B \ Rµ̄ . Then
problem (39) possesses a maximal solution ūµ̄ if and only if the following two
conditions are satisfied:
(i) For every Borel subset A ⊂ R, C2/q,q0 (A) = 0 =⇒ µ(A) = 0.
(ii) S = Sq∗ ∪ ∂ν S.
We shall not give the proof, although it is interesting to note that a key
inequality which give the condition (ii) for solving (39) is the following
max(ǔ, u∗ ) ≤ ūµ̄ ≤ ǔ + u∗ , (45)
∗
in which formula u and ǔ have been defined above.
A striking difference between the subcritical case and the supercritical case
is the loo of uniqueness. It was first proved by Le Gall [26] in the case q = 2 and
then by Marcus and Véron [32] in the general case q ≥ (N + 1)/(N − 1) that
there may exists infinitely many solutions of (39) with a given singular trace.
In particular, for every γ > 0, there exists a solution uγ with tr(uγ ) = (∂B, 0)
and such that uγ (0) < γ. Moreover, by sharpening the construction of the uγ
it is proved in [32] that for any ε > 0 there exists a Borel subset A ⊂ ∂B with
meas(∂B \ A) < ε and a solution u of (39) such that T r(u) = (∂B, 0) and
lim u(r, σ) = 0 for every σ ∈ A. This clearly indicates that the formulation
r→1
of boundary traces in terms of the usual topology on ∂B is not sufficient to
describe the variety of phenomena which may occur at the boundary and that a
sharper notion is needed. A first and important step towards this direction has
been made by Dynkin and Kuznetsov [15][22, 23] in a series of recent papers. In
these works they introduce a topology thiner than the usual one in Rd and they
introduce a new class of problems in which some uniqueness is proved. However
the problem of finding The notion of boundary trace which gives rise to a one
to one and onto correspondence between the set of all positive solutions of (39)
and their boundary trace is still open.

Conformal deformations of hyperbolic space

The problem of conformal deformation of Riemannian metrics is one of the most
interesting field of applications of semilinear elliptic equations. We perform the
identification of the hyperbolic N-space HN with (B, gH ), where
!2
2
gH (x) = 2 η(x)
1 − |x|
Laurent Véron 339

with ηij = δij . Given K ∈ C ∞ (B) with N > 2, three classical problems arising
from conformal geometry are the following
I- Does there exists a positive function v on such that the metric gv = v 4/(N −2)
has scalar curvature K ?
II- Is gv a complete Riemannian metric ?
III- If so is v unique in the class of complete Riemannian metrics conformal to
gH with scalar curvature K ?
It is classical that the function v satisfies

CN ∆v + K(x)v (N +2)/(N −2) = 0 in B (46)

with CN = 4(N − 1)/(N − 2). This problem has been thoroughly studied by
Loewner and Nirenberg [27](1974), Ni (1982), Aviles and McOwen (1985-1988)
and more recently (1993-94) by Ratto, Rigoli and Véron [35]. An interesting case
which is associated to supercritical trace problems occurs when K is nonpositive
(at least near ∂B ). The completeness assumption means that the geodesic
distance from inside up to the boundary is always infinite, which in this case is
equivalent to
Z 1
v (N −2)/2 (γ(t))dt = ∞ (47)
0
0,1
for any γ ∈ C ([0, 1]; B̄), with γ([0, 1)) ⊂ B, and γ(1) ∈ ∂B. This means that
v has some kind of blow-up near ∂Ω. Many existence results concerning this
equation have been proven. Some uniqueness results also hold under a strong
blow-up assumption, namely

lim v(x) = ∞,
|x|→1

and a positivity assumption on K (see [35] for details), but the problem of
uniqueness of v under the mere completeness assumption (47) remains com-
pletely open, even when K = 1.

References
[1] Bandle C. & Marcus M., Asymptotic behavior of solutions and their deriva-
tive for semilinear elliptic problems with blow-up on the boundary, Ann.
Inst. H. Poincaré 12, 155-171 (1995).
[2] Bandle C. & Marcus M., Large solutions of semilinear elliptic equations:
existence, uniqueness and asymptotic behavior, J. Anal. Math. 58, 9-24
(1992).
[3] Baras P. & Pierre M.; Singularités éliminables pour des équations semi-
lineaires, Ann. Inst. Fourier Grenoble 34, 185-206 (1984).
[4] Brezis H., Une équation semi-linéaire avec conditions aux limites dans L1 ,
unpublished paper. See also [41]-Chap. 4.
340 Generalized boundary value problems

[5] Brezis H. & Strauss W., Semilinear second order ellitic equations in L1 ,
J. Math. Soc. Japan 25, 565-590 (1973).
[6] Chasseigne E. & Vazquez J.L., Well-posedness of fast- diffusion equations
for optimal classes of data, Preprint Univ. Autônoma de Madrid (2000).
[7] Chen X.Y., Matano H. & Veron L.; Anisotropic singularities of nonlinear
elliptic equations, J. Funct. Anal. 83, 50-97 (1989).
[8] Dhersin J.S. & Le Gall E., Wiener test for super Brownian motion and
the Brownian snake, Prob. Theory relat. Fields 108, 103-129 (1997).
[9] Doob J., Classical Potential Theory and its Probabilistic Counterpart,
Springer-Verlag, Berlin/New-York (1984).
[10] Dynkin E.B., A probabilistic approach to one class of nonlinear differential
equations, Prob. Theory Relat. Fields 89, 89-115 (1991).
[11] Dynkin E.B. & Kuznetsov S.E., Superdiffusions and removable singularities
for quasilinear P.D.E., Comm. Pure, Appl. Math. 49, 125-176 (1996).
[12] Dynkin E.B. & Kuznetsov S.E., Linear additive functionals of superdiffu-
sion and related nonlinear P.D.E., Trans. A.M.S. 348, 1959-1988 (1996).
[13] Dynkin E.B. & Kuznetsov S.E., Trace on the boundary for solutions of
nonlinear differential equations, Trans. A.M.S. 350, 4499-4519 (1998).
[14] Dynkin E.B. & Kuznetsov S.E., Solutions of nonlinear differential equtions
on a Riemannian manifold and their trace on the Martin boundary , Trans.
A.M.S. 350, 4521-4552 (1998).
[15] Dynkin E.B. & Kuznetsov S.E., Fine topology and fine trace on the bound-
ary associated with a class of quasilinear differential equations, Comm. Pure
Appl. Math. 51, 897-936 (1998).
[16] Gilbarg D. & Trudinger N.S., Partial Differential Equations of Second Or-
der, 2nd Ed. Springer-Verlag, Berlin/New-York (1983)
[17] Gmira A. & Véron L., Boundary singularities of solutions of some nonlinear
elliptic equations, Duke Math. J 64, 271-324 (1991).
[18] Grillot M. & Véron L., Boundary trace of solutions of the Prescribed Gaus-
sian curvature equation, Proc. Roy. Soc. Edinburgh 130 A, 1-34 (2000).
[19] Iscoe I.,On the support of measure-valued critical branching Brownian mo-
tion, Ann. Prob. 16, 200-221 (1988).
[20] Keller J.B., On solutions of ∆u = f (u), Comm. Pure Appl. Math. 10,
503-510 (1957).
[21] Kondratiev V. & Nikishkin A.,On positive solutions of singular value prob-
lemsfor the equation ∆u = uk , Russian J. Math. Phys. 1, 123-138 (1993).
Laurent Véron 341

[22] Kuznetsov S.E., σ-moderate solutions of Lu = uα and fine trace on the

boundary C.R. Acad. Sci. Paris 326 Ser. I, 1189-1194 (1998).

[23] Kuznetsov S.E., On uniqueness of a solution of Lu = uα with given trace,

Electr. Comm. in Probability (to appear).

[24] Le Gall J.F., Les solutions positives de ∆u = u2 dans le disque unité, C.R.
Acad. Sci. Paris 317 Ser. I, 873-878 (1993).

[25] Le Gall J.F., The brownian snake and solutions of ∆u = u2 in a domain,

Prob. Theory Rel. Fields 102, 393-432 (1995).

[26] Le Gall J.F., A probabilistic approach to the trace at the boundary for solu-
tions of a semilinear parabolic partial differential equation, J. Appl. Math.
Stoch. Anal. 9, 399-414 (1996).

[27] Loewner C. & Nirenberg L., Partial differential equations invariant under
conformal or projective transformations, in Contributions to Analysis, 245-
272, Academic Press, Orlando Flo (1974).

[28] Marcus M. & Véron L., Uniqueness and asymptotic behaviour of solutions
with boundary blow-up for a class of nonlinear elliptic equations, Ann. Inst.
H. Poincaré 14, 237-274 (1997).

[29] Marcus M. & Véron L., Traces au bord des solutions positives d’équations
elliptiques non-linéaires, C.R. Acad. Sci. Paris 321 Ser. I, 179-184 (1995).

[30] Marcus M. & Véron L., Traces au bord des solutions positives d’équations
elliptiques et paraboliques non-linéaires: résultats d’existence et d’unicité,
C.R. Acad. Sci. Paris 323 Ser. I, 603-608 (1996).

[31] Marcus M. & Véron L., The boundary trace of positive solutions of semi-
linear elliptic equations: the subcritical case, Arch. Rat. Mech. Anal. 144,
201-231 (1998).

[32] Marcus M. & Véron L., The boundary trace of positive solutions of semi-
linear elliptic equations: the supercritical case, J. Math. Pures Appl. 77,
481-524 (1998).

[33] Meyers N.G., A theory of capacities for potential of functions in Lebesgue

classes, Math. Scan. 26, 255-292 (1970).

[34] Osserman R., On the inequality ∆u ≥ f (u), Pacific J. Math. 7, 1641-1647

(1957).

[35] Ratto A., Rigoli M. & Véron L., Scalar curvature and conformal deforma-
tion of hyperbolic space J. Funct. Anal. 121, 15-77 (1994).

[36] Sheu Y.C., Removable singularities for solutions of some nonlinear differ-
erential equations, Duke Math. J. 74, 701-711 (1994).
342 Generalized boundary value problems

[37] Stein E., Singular Integrals and Differentiability Properties of Functions,

Princeton Univ. Press 30 (1970).
[38] Treibel H., Interpolation Theory, Function Spaces, Differential Operators,
North-Holland Publ. Co (1978).
[39] Véron L., Singular solutions of some nonlinear elliptic equations, Nonlinear
Anal. T.,M. & A 5, 225-242 (1981).
[40] Véron L., Comportement asymptotique des solutions d’équations elliptiques
semi-linéaires dans RN , Ann. Mat. Pura Appl. 127, 25-50 (1981).

[41] Véron L., Singularities of Solutions of Second Order Quasilinear Equa-

tions, Pitman Research Notes in Math. 353, Addison Wesley Longman Inc.
(1996).
[42] Véron L., Semilinear elliptic equations with uniform blow-up on the bound-
ary, J. Anal. Math. 59, 231-250 (1992).

Laurent Véron
Laboratoire de Mathématiques et Physique Théorique
CNRS UMR 6083
Faculté des Sciences et Techniques
Parc de Grandmont
F 37200 Tours France
e-mail: veronl@univ-tours.fr