You are on page 1of 5

J. Math. Anal. Appl.

393 (2012) 311–315

Contents lists available at SciVerse ScienceDirect

Journal of Mathematical Analysis and


Applications
journal homepage: www.elsevier.com/locate/jmaa

Estimates of the principal eigenvalue of the p-Laplacian


Jiří Benedikt a , Pavel Drábek b,∗
a
Department of Mathematics, Faculty of Applied Sciences, University of West Bohemia, Univerzitní 22, 306 14 Plzeň, Czech Republic
b
Department of Mathematics and N.T.I.S., Faculty of Applied Sciences, University of West Bohemia, Univerzitní 22, 306 14 Plzeň, Czech Republic

article info abstract


Article history: We provide estimates from below and from above for the principal eigenvalue of the
Received 23 February 2012 p-Laplacian on a bounded domain. We apply these estimates to study the asymptotic
Available online 17 April 2012 behavior of the principal eigenvalue for p → +∞.
Submitted by Mr. V. Radulescu
© 2012 Elsevier Inc. All rights reserved.
Keywords:
Eigenvalue problem for p-Laplacian
Estimates of principal eigenvalue

1. Introduction

In this paper, we consider the eigenvalue problem for the p-Laplacian

−∆p u = λ|u|p−2 u in Ω ,

(1)
u=0 on ∂ Ω ,
where Ω is a bounded open subset of RN , N ≥ 1, p > 1. This eigenvalue problem has been intensively studied during the
last three decades. In the case of one dimension it is well understood; the first references are [1,2]. In higher-dimensional
cases, the structure of the set of all eigenvalues of (1) is not known. This is a long-standing open problem. However, the
properties of the first (principal) eigenvalue of (1) and of the corresponding eigenfunction are well known, thanks to several
pioneering works [3–5]. The work reported in [4] also studies the radial symmetric case in more detail. Since 1990, many
papers have appeared about (1) and its various modifications.
We denote by λ1 (Ω , p) the principal eigenvalue of (1). It is well known that

|∇ u|p dx
λ1 (Ω , p) = min Ω , (2)

|u|p dx
1 ,p
where the minimum is taken over all u ∈ W0 (Ω ), u ̸= 0.
In order to illustrate our main results, we consider first the problem in one dimension,
  ′
− |u′ |p−2 u′ = λ1 |u|p−2 u in (−R, R),
u(−R) = u(R) = 0.
It is known (see, e.g., [6, p. 244]) that
 p
1 π
λ1 ((−R, R), p) = (p − 1) . (3)
Rp p sin πp

∗ Corresponding author.
E-mail addresses: benedikt@kma.zcu.cz (J. Benedikt), pdrabek@kma.zcu.cz (P. Drábek).

0022-247X/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.jmaa.2012.03.054
312 J. Benedikt, P. Drábek / J. Math. Anal. Appl. 393 (2012) 311–315

Hence

lim λ1 ((−R, R), p) = +∞ for 0 < R ≤ 1


p→+∞

and

lim λ1 ((−R, R), p) = 0 for R > 1.


p→+∞

In higher dimensions, an analogue of the precise formula (3) is not known. In our work, we present upper bounds and lower
bounds for λ1 (Ω , p). In particular, these bounds allow study of the asymptotic behavior of λ1 (Ω , p) as p → +∞.
Let Ω = BN (0, R) be an open ball in RN of radius R > 0 and centered at the origin. Then it follows from our Theorems 2
and 4 that, for all p > 1,

Np (p + 1)(p + 2) · · · (p + N )
≤ λ1 (BN (0, R), p) ≤ . (4)
Rp N !Rp
Hence

lim λ1 (BN (0, R), p) = +∞ for 0 < R ≤ 1


p→+∞

and

lim λ1 (BN (0, R), p) = 0 for R > 1.


p→+∞

Note that limp→+∞ λ1 (BN (0, R), p) = +∞ for 0 < R < 1 and limp→+∞ λ1 (BN (0, R), p) = 0 for R > 1 also follow from
[7, Lemma 1.5]. Our contribution consists in the fact that we also prove that limp→+∞ λ1 (BN (0, R), p) = +∞ for the critical
radius RC = 1 and that the estimate for λ1 (BN (0, R), p), R > 0, p > 1, is given by inequalities (4).
Besides the case of Ω being a ball in RN , our estimates apply to more general bounded open subsets Ω of RN . The first
inequality in (4), i.e. the estimate from below for λ1 (Ω , p), holds for any Ω such that |Ω | = |BN (0, R)|. This fact easily
follows from the Schwarz symmetrization argument; see [8]. We also generalize the first inequality from (4) to

kp
λ1 (Ω , p) ≥ (5)
Rp
for Ω ⊂ Bk (0, R) × RN −k , where Bk (0, R) is the open ball in Rk of radius R > 0 and centered at the origin of Rk , k ∈
{1, 2, . . . , N }. For k = 1, the estimate (5) holds for any Ω situated between two parallel hyperplanes of the distance 2R.
In particular, for R ≤ 1, limp→+∞ λ1 (Ω , p) = +∞ regardless of the measure of Ω .
On the other hand, the second inequality in (4) also holds for any open bounded set in RN which contains an inscribed
ball of radius R > 0.
However, as regards the asymptotic behavior of λ1 (Ω , p) for p → +∞, our results do not cover the case of Ω such that
the largest inscribed ball in Ω has radius R = 1, but Ω cannot be squeezed between two parallel hyperplanes of distance 2.
An equilateral triangle having a maximal inscribed disc of radius 1 is a concrete example of such a domain Ω in R2 . The
solution of this open problem, together with our results and the results of [7], would provide a complete characterization of
limp→+∞ λ1 (Ω , p) by means of the radius of the largest ball, inscribed in Ω .

2. Main results

The following lemma is a special version of [9, Theorem 2.1]. Its proof is based on Picone’s identity for the p-Laplacian.

Lemma 1. Suppose that v ∈ C 1 (Ω ), v > 0 in Ω , ∆p v ∈ C (Ω ), and that there exists λ > 0 such that v satisfies

− ∆p v ≥ λv p−1 in Ω . (6)

Then

λ ≤ λ1 (Ω , p).

Theorem 2. Let R > 0, k ∈ {1, 2, . . . , N }, and let Ω be a bounded open subset of Bk (0, R) × RN −k . Then, for any p > 1,
kp
λ1 (Ω , p) ≥ .
Rp
J. Benedikt, P. Drábek / J. Math. Anal. Appl. 393 (2012) 311–315 313

Proof. For x = (x1 , x2 , . . . , xN ), we write


def  1
|x|k = x21 + x22 + · · · + x2k 2 .
We choose
 R
def  p−1 1 def kp
v(x) = Rp − (R − t )p dt , x ∈ Ω, λ= ,

|x|k Rp
and prove that the assumptions of Lemma 1 are satisfied. Obviously, v > 0 in Ω and λ > 0. For all x ∈ Ω , we have
 R
p−1  R
p−1
kp  p−1 1 kp p
λv p−1 (x) = Rp − (R − t )p

dt ≤ R p−1 dt
Rp | x| k Rp | x| k
kp
= Rp (R − |x|k )p−1 .
Rp
It remains to prove that

− ∆p v(x) ≥ kp(R − |x|k )p−1 , (7)


x ∈ Ω.
Observe that v depends only on |x|k , and hence
R
    
d  p−1 1 x1 x2 xk
∇v(x) = Rp − (R − t )p , ,..., , 0, . . . , 0
 
dt  ·
ds s s=|x|k |x|k |x|k |x|k
 
1 x1 x2 xk
= −(Rp − (R − |x|k )p ) p−1 , ,..., , 0, . . . , 0 ,
|x|k |x|k |x|k
provided that |x|k ̸= 0 and ∇v(x) = 0 if |x|k = 0. From now on, we distinguish between |x|k ̸= 0 and |x|k = 0.
Case |x|k ̸= 0. We have
 
x1 x2 xk
|∇v| p−2
∇v = −(R − (R − |x|k ) ) p p
, ,..., , 0, . . . , 0 .
|x|k |x|k |x|k
A calculation yields
∂ ∂v x2i x2i
   
 1
p−2
= −p (R − |x|k ) p−1
− R − (R − |x|k )
p p

|∇v| 1−
∂ xi ∂ xi 2
|x|k |x|k |x|2k
for i ∈ {1, 2, . . . , k}, and
∂ ∂v
 
|∇v|p−2 =0
∂ xi ∂ xi
for i ∈ {k + 1, . . . , N }. Hence
 k−1
−∆p v = −div |∇v|p−2 ∇v = p (R − |x|k )p−1 + Rp − (R − |x|k )p .
  
|x|k
We observe that
 R
Rp − (R − |x|k )p = pt p−1 dt ≥ p(R − |x|k )p−1 |x|k .
R−|x|k

Consequently,
−∆p v ≥ p (R − |x|k )p−1 + (k − 1)p(R − |x|k )p−1 = kp(R − |x|k )p−1 ,
i.e., (7) holds for |x|k ̸= 0.
def
Case |x|k = 0. Since |∇v|p−2 ∇v = 0 if |x|k = 0, we use the definition of the derivative in order to evaluate
|∇v|p−2 ∂∂vxi − 0 − Rp − (R − |x|k )p |xx|ik
 
∂ ∂v
 
p−2
|∇v| = lim = lim
∂ xi ∂ xi xi → 0 xi
xi → 0 xi
− R − (R − |xi |)p
p
 
= lim
xi → 0 |xi |
l′ H ôpital
= lim −p (R − |xi |)p−1 = −pRp−1 .
xi → 0
314 J. Benedikt, P. Drábek / J. Math. Anal. Appl. 393 (2012) 311–315

Hence

−∆p v = kpRp−1 = kp (R − |x|k )p−1 ,


i.e., the equality holds in (7). The proof is finished. 

Corollary 3. Let R ∈ (0, 1] in Theorem 2. Then

lim λ1 (Ω , p) = +∞.
p→+∞

Theorem 4. Let R > 0. Then, for any p > 1,


(p + 1)(p + 2) · · · (p + N )
λ1 (BN (0, R), p) ≤ .
N !Rp

def
Proof. Choose v(x) = R − |x|. It follows from the variational characterization of λ1 (BN (0, R), p) that

BN (0,R)
|∇v|p dx |BN (0, R)|
λ1 (BN (0, R), p) ≤ = R
v p dx

BN (0,R) 0
(R − t )p |SN (0, t )|dt
RN |BN (0, 1)|
= R .
|SN (0, 1)| 0
(R − t )p t N −1 dt
The substitution t = Rτ yields
 R  1
(R − t ) t
p N −1
dt = R p+N
(1 − τ )p τ N −1 dτ = Rp+N B(p + 1, N ),
0 0

where
Γ (y)Γ (z )
B(y, z ) =
Γ (y + z )
is the Beta function. Furthermore,
N N N
π2 2π 2 2π 2
|BN (0, 1)| =  =  N  and |SN (0, 1)| =  N  .
Γ 1+ 2
N
NΓ 2 Γ 2

Consequently,

RN |BN (0, 1)| Γ (p + 1 + N )


= ,
N Γ (N )Γ (p + 1)Rp
R
|SN (0, 1)| 0
(R − t ) p t N −1 dt

which proves our statement. 

Corollary 5. Let R > 1 in Theorem 4, and let BN (0, R) ⊆ Ω . Then

lim λ1 (Ω , p) = 0.
p→+∞

3. Visualization of the results

In this section, we visualize the inequalities (4) by the shaded areas in Fig. 1. In the dimension N = 1, we illustrate the
estimate (4) on Ω = (− 12 , 21 ), (−2, 2), (−1, 1). In this case, λ1 is a function of p, given by (3). The corresponding graphs lie
in the shaded regions. In higher dimensions N = 2, 3, 4, the estimate (4) is visualized for Ω being the unit ball. The dots in
the shaded areas correspond to approximate values of λ1 for p = 1.1, 1.2, . . . , 4.0, which were evaluated in [10].

Acknowledgments

The first author was supported by the project KONTAKT, grant no. ME 10093. The second author was supported by the
European Regional Development Fund (ERDF), project ‘‘NTIS—New Technologies for Information Society’’, European Center
of Excellence, CZ 1.05/1.1.00/02.0090.
J. Benedikt, P. Drábek / J. Math. Anal. Appl. 393 (2012) 311–315 315

Fig. 1. Dependence of λ1 on p.

References

[1] P. Drábek, Ranges of a-homogeneous operators and their perturbations, Časopis Pěst. Mat. 105 (1980) 167–183.
[2] Á. Elbert, A half-linear second order differential equation, in: Qualitative Theory of Differential Equations, Vol. I, II (Szeged, 1979), in: Colloq. Math.
Soc. János Bolyai, vol. 30, North-Holland, Amsterdam, New York, 1981, pp. 153–180.
[3] A. Anane, Simplicité et isolation de la première valeur propre du p-Laplacien avec poids, C. R. Acad. Sci. Paris Sér. I Math. 305 (1987) 725–728.
[4] T. Bhattacharya, Some results concerning the eigenvalue problem for the p-Laplacian, Ann. Acad. Sci. Fenn. Math. 14 (1989) 325–343.
[5] P. Lindqvist, On the equation div(|∇ u|p−2 ∇ u) + λ|u|p−2 u = 0, Proc. Amer. Math. Soc. 109 (1990) 157–164.
[6] R.J. Biezuner, G. Ercole, E.M. Martins, Computing the first eigenvalue of the p-Laplacian via the inverse power method, J. Funct. Anal. 257 (2009)
243–270.
[7] P. Juutinen, P. Lindqvist, J.J. Manfredi, The ∞-eigenvalue problem, Arch. Ration. Mech. Anal. 148 (1999) 89–105.
[8] B. Kawohl, Rearrangements and Convexity of Level Sets in PDE, in: Lecture Notes in Mathematics, vol. 1150, Springer-Verlag, Berlin, 1985.
[9] W. Allegretto, Y.X. Huang, A Picone’s identity for the p-Laplacian and applications, Nonlinear Anal. 32 (1998) 819–830.
[10] R.J. Biezuner, J. Brown, G. Ercole, E.M. Martins, Computing the first eigenpair of the p-Laplacian via inverse iteration of sublinear supersolutions.
Preprint arXiv:1011.3172v2.

You might also like