A Note on Harmonic Functions on Surfaces

Author(s): Jean C. Cortissoz

Source: The American Mathematical Monthly, Vol. 123, No. 9 (November 2016), pp. 884-893
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/10.4169/amer.math.monthly.123.9.884
Accessed: 31-10-2016 18:17 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access
to The American Mathematical Monthly

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
 A Note on Harmonic Functions on Surfaces
Jean C. Cortissoz

Abstract. We review and give elementary proofs of Liouville-type properties of harmonic and
subharmonic functions in the plane endowed with a complete Riemannian metric, and prove a
gap theorem for the possible growth of harmonic functions when this metric has nonnegative
Gaussian curvature.

1. INTRODUCTION. Given a function u :  −→ R, where  is an open subset of

R2 , the Laplacian is defined, in rectangular coordinates, as

∂ 2u ∂ 2u
u = + ,
∂x2 ∂ y2

and we say that u is harmonic (resp. subharmonic) if u = 0 (u ≥ 0). The classi-
cal Liouville’s theorem in R2 states that a bounded harmonic function is constant (for
a beautiful proof of this fact, we recommend [9]). A stronger version says that if u
is a subharmonic function bounded above, then it must be constant; we refer to this
property of the plane as parabolicity, and, no doubt, it is an amazing fact that being a
solution to a partial differential identity or a partial differential inequality may deter-
mine the growth properties of a function.
Over the years, the analysis studied on Rn has been carried over to Riemannian
manifolds, a realm where a differential and an inner product structure coexist. In par-
ticular, a Laplacian operator acting over functions can be defined, and hence, it is a
framework in which the concept of harmonic, subharmonic, and superharmonic func-
tion have a natural extension. So it is also natural to ask which properties of harmonic
functions, such as Liouville’s theorem or parabolicity, are preserved in a Riemannian
manifold.
In this paper, we are interested in Liouville-type theorems and gap theorems on
surfaces with a pole (i.e., surfaces where polar coordinates can be defined). By a
Liouville-type property, we mean a theorem that states that if a harmonic function
is conveniently bounded, then it must be constant, and by a gap theorem we mean a
theorem that imposes restrictions on how fast a harmonic function must grow so that
it does not belong to a class of strictly lower growth.
Before starting to throw definitions, formulas and theorems at the reader, let us men-
tion some interesting results related to our work. Regarding Liouville-type theorems,
of great importance are the results of Ahlfors and Milnor, which in the case of sur-
faces endowed with a rotationally symmetric metric relates an intrinsic quantity, the
curvature, to the behavior of subharmonic functions and, to be more precise, to the
parabolicity of the surface. Green and Wu ([6]) extended the Ahlfors–Milnor theorem
to the case of surfaces with a pole. We will give a relatively simple proof of part of the
Ahlfors–Milnor–Greene–Wu parabolicity criterion in which our main tool will be the
strong maximum principle.
Regarding gap properties for harmonic functions, on the classical side, that is, in
the complex plane, it is well known that if the rate of growth of a harmonic function is
http://dx.doi.org/10.4169/amer.math.monthly.123.9.884
MSC: Primary 53C21

884 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 123

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
bounded by a power of the distance to a fixed point, then it must be a polynomial. This
is a consequence of the analyticity of harmonic functions in R2 and of the Cauchy esti-
mates. A most recent result has been proved by Ni and Tam in [4]: Here, the authors
show how fast, in a Kähler manifold with nonnegative bisectional curvature, a super-
linear harmonic function must grow. As a treat for the reader, we improve upon Ni and
Tam’s result in the case of a surface with a pole (see Theorem 4 below). This is the
only new result in this paper (at least to the best of our knowledge).
The main ideas of our proofs are contained in, and some could even say are trans-
planted, via a classical comparison theorem due to Sturm, from the beautiful book
Maximum Principles in Differential Equations [8], which we highly recommend. We
also hope that this note serves as an introduction, assuming as few prerequisites as
possible, to the study of harmonic functions in Riemannian geometry.

2. PRELIMINARIES. For the convenience of the reader, let us give a quick review
of a few concepts in Riemannian geometry. We will consider R2 equipped with polar
coordinates (r, θ) with respect to the origin, and endowed with a family of inner prod-
ucts of the form

g = dr 2 + ( f (r, θ))2 dθ 2 ,

where f : (0, ∞) × [0, 2π] −→ (0, ∞) is a smooth function such that

f (r, 0) = f (r, 2π) , lim f (r, θ) = 0 and lim f  (r, θ) = 1,

r →0+ r →0+

where we have used (and will use in what follows) f  to denote differentiation with
respect to r .
For those not familiar with Riemannian manifolds, g represents a way of measuring
vectors, and it is called a Riemannian metric. It defines an inner product for vectors
based at the point (r, θ) and is represented in the basis

∂ x y ∂
= i+  j, = yi − xj
∂r x +y
2 2 x + y2
2 ∂θ

in the following way. If we have vectors

∂ ∂
vj = a j + bj , j = 1, 2,
∂r ∂θ

based at the point (r, θ) (the point (r cos θ, r sin θ) in rectangular coordinates), then

g (v1 , v2 ) = a1 a2 + b1 b2 ( f (r, θ))2 .

Notice that with the choice f (r, θ) = r we obtain the usual inner product of vectors
in the plane. The pair (M, g) is called a Riemannian surface (as opposed to a Riemann
surface), and, as discovered by Gauss, Riemannian surfaces have an important intrinsic
quantity: the curvature. From the expression for a Riemannian metric given above, the
curvature can be computed as

f  (r, θ)
K g (r, θ) = − .
f (r, θ)

November 2016] HARMONIC FUNCTIONS ON SURFACES 885

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
In a Riemannian surface, the gradient of a function u : M −→ R can be defined since
given a nondegenerate scalar product, a metric dual of the derivative of a function can
be defined. In our case, the gradient of u can be computed as

∂u ∂ 1 ∂u ∂
∇g u = + 2 .
∂r ∂r f ∂θ ∂θ

We can also define a Laplacian, which is the operator of our utmost interest:

∂2 f ∂ 1 ∂2 1 ∂f ∂
g = + + − 3 .
∂r 2 f ∂r f ∂θ 2 f ∂θ ∂θ

Given a Laplacian, we can define a C 2 function u as harmonic if g u = 0, subhar-

monic if g u ≥ 0, and superharmonic if g u ≤ 0.
f
The term h g := in the expression for the Laplacian is rather important. It gives
f
the mean curvature of the circle of radius r with respect to the metric g. We will need to
estimate this term, and the tool we will employ is the following comparison theorem,
due to Sturm, which among geometers is known as the Laplacian comparison theorem.

f  h 
Theorem 1. Let h, f : [a, ∞) −→ (0, ∞) be C 2 functions. If (r ) ≤ (r ), r > a,
f h
f h f h
and (a) ≤ (a), then (r ) ≤ (r ) for r > a.
f h f h

The proof of this theorem is based upon the following observation:

   2
w w w
=− + .
w w w
    
f h f
Notice then that the hypothesis implies that (r ) ≤ (r ) whenever (r )
f h f
h
= (r ).
h
Another important tool in the arguments that follow is the maximum principle.

Theorem 2. Let  be a bounded open subset ofR2 , endow R2 with a Riemannian met-
ric g, and let u :  −→ R be a C 2 () ∩ C  function. Then we have the following.
(i) If g u ≤ 0 and u has a minimum in the interior of , then u is constant.
(ii) If g u ≥ 0 and u has a maximum in the interior of , then u is constant.

This theorem follows from the strong maximum principle for elliptic operators (see
Section 3.2 in [5], in particular Theorem 3.5).

3. BOUNDED HARMONIC FUNCTIONS. Before we state the main result of this

section, we shall make an observation. Let z : (a, ∞) −→ (0, ∞) be a C 2 function.
Direct calculation shows that, for α > a fixed, the function
 r
1
h (r ) = dσ
α z (σ )

886 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 123

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
satisfies the identity

d 2h z  (r ) dh
+ = 0.
dr 2 z (r ) dr
In other words, it is harmonic with respect to the metric

g0 = dr 2 + z (r )2 dθ 2 ,

defined on the complement of the closed ball of radius a > 0. Let us now assume that
z satisfies
• limr →a + z (r ) = 0, and
• limr →a + z  (r ) > 0.
z 
Hence, if g = dr 2 + ( f (r, θ))2 dθ 2 is a metric on R2 such that K g (r, θ) ≥ − (r ),
z
z
then g h ≤ 0. Indeed, notice that the assumptions on z imply that lim (r ) = ∞,
r →a + z
f z
and hence, for a0 close to a, (a0 ) ≤ (a0 ). Thus, the hypothesis of Sturm’s com-
f z
f z
parison theorem holds, and in consequence, we have that (r ) ≤ (r ), for r > a.
f z
Therefore,

d 2h f  dh d 2h f 1 d 2h z  dh
g h = 2
+ = 2 + ≤ 2 + = 0.
dr f dr dr f z dr z dr
Given a continuous function v define

M (v; r ) = sup |v (r, θ)| .

θ∈[0,2π)

We are ready to show the following result.

Theorem 3. Let g = dr 2 + f (r, θ)2 dθ 2 be a metric on R2 . Let z be as above, and

z 
assume that K g (r, θ) ≥ − (r ) for r large enough. Then, any subharmonic function
z
u that satisfies
M (u; r )
lim inf =0
r →∞ h (r )
must be constant.

Proof. In what follows, we will denote by B R the ball of radius R centered at (0, 0).
Fix R1 > 0 so that h (r ) ≥ 0 for r ≥ R1 , with h defined as above, fix δ > 0 also small,
and define the function

wδ,η = u − δh (r ) − M (u; R1 ) .

By hypothesis, we can take R2 larger than R1 such that wδ,η ≤ 0 on both ∂ B R1 and
∂ B R2 . It is also clear from its definition that g wδ,η ≥ 0. From the maximum principle
it follows that

November 2016] HARMONIC FUNCTIONS ON SURFACES 887

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
 wδ,η ≤ 0

in the annulus of inner radius R1 and outer radius R2 . Since δ > 0 can be made arbi-
trarily small, we conclude that for all P in the annulus the estimate

|u (P)| ≤ M (u; R1 )

holds and, hence, that |u| ≤ M (u; R1 ) in the ball of radius R2 . Via the maximum
principle (as u attains its maximum at an interior point of B R2 ) we can conclude that u
is constant in the ball of radius R2 . Since R2 can be taken arbitrarily large, the theorem
follows.

The proof given above is presented in [8] in the case of a flat metric (see Theorem 19
in Section 2 of [8] and the example thereafter), but its generalization to other metrics,
as shown above, is straightforward. Besides, the previous theorem has the following
interesting consequence.

Corollary 1. Let u be a harmonic function on a surface such that for r ≥ r0 ≥ 1 its

1
curvature function satisfies K g ≥ − 2 . If
r log r

M (u; r )
lim inf = 0,
r →∞ log log r

then u is constant. In particular, if u is bounded, then it is constant.

This corollary is due to Greene and

 Wu (Theorem D in [6]), and its proof is now
r
quite simple. Take z (r ) = r log for r ≥ r0 . Notice that, since r0 ≥ 1, then
r0

z  1 1
− =−  ≤− 2 ≤ Kg.
z r 2 log r r log r
r0

z
On the other hand, limr →r + (r ) = +∞, and hence by Sturm’s comparison theorem,
0 z
if we define
 r  
1 r
h (r ) = dσ = log log ,
e·r0 z (σ ) r0
 
r
then g h ≤ 0. Notice then that log log r and log log have the same asymptotic
r0
behavior as r → ∞; by Theorem 3, the result follows.
We want to point out that, in higher dimensions, Liouville’s theorem has been
extended by Yau, via his gradient estimate (see below), to manifolds of nonnegative
Ricci curvature.

On the Ahlfors–Milnor–Greene–Wu parabolicity criterion: an application of

Hadamard’s three circles theorem. A surface M is called parabolic if any subhar-
monic function (i.e., u ≥ 0) bounded above is constant. Milnor in [7] showed that

888 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 123

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
given a rotationally symmetric metric on R2 , it is parabolic if for large enough r the
1
curvature of the metric is larger than or equal to − 2 .
r log r
We use the maximum principle to give a proof of a generalization of Milnor’s cri-
terion (which is a theorem due to Greene and Wu). We assume that we have endowed
R2 with a metric of the form g = dr 2 + ( f (r, θ))2 dθ 2 whose curvature is larger than
1
or equal to − 2 .
r log r
First, we have the following version of Hadamard’s three circles theorem. Let R > 0
1
be such that K g ≥ − 2 for r ≥ R, let u be a subharmonic function, and define
r log r
Mr = max u (r, θ) .
θ∈[0,2π)

Notice that if r2 > r1 , then Mr2 ≥ Mr1 by the maximum principle. So let R < r1
< r < r2 , and define



  
1 log r2 log Rr
ϕ (r ) =  Mr1 log   + Mr2 log .
log log r2 log Rr log r1
log r1

r
It is easy to check that g ϕ ≤ 0 (indeed, verify that g log log ≤ 0 and use the
R
fact that Mr1 ≤ Mr2 ); hence, u − ϕ is subharmonic, and also u − ϕ ≤ 0 on both ∂ B1
and ∂ B2 . We can conclude via the maximum principle that

Mr ≤ ϕ (r ) .

This last inequality is our version of Hadamard’s three circles theorem. Now assume
that u is bounded above. By taking r2 → ∞, we obtain the estimate

Mr ≤ Mr1 .

But then, since r > r1 and r is arbitrary, by the maximum principle u must be constant.
We can conclude that the surface is parabolic. Again, we must point out that our proof
follows closely the arguments given in Chapter 2, Section 12 in [8].

4. A GAP THEOREM FOR SURFACES WITH NONNEGATIVE GAUSSIAN

CURVATURE. It is an exercise in complex analysis to prove the following result.
Given a holomorphic function f such that | f (z)| ≤ C |z|k + B, then f is a polynomial
of degree at most k. From this, we can conclude that there are gaps between the
possible growth that a holomorphic and, in consequence, harmonic functions can have.
For instance, a harmonic function of subquadratic growth (i.e., k < 2) must be of at
most linear growth (i.e., k must be less or equal to 1).
In this section, we prove a gap theorem on the possible growth of harmonic func-
tions on a complete noncompact surface M of nonnegative Gaussian curvature. To be
more precise, we will show the following gap theorem.

Theorem 4. Let g = dr 2 + f (r, θ)2 dθ 2 be a metric on R2 with nonnegative Gaussian

curvature. Let u : M −→ R be a harmonic function. If for all δ > 0
M (u; r )
lim inf = 0,
r →∞ r 1+δ

November 2016] HARMONIC FUNCTIONS ON SURFACES 889

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
then there is a constant C > 0 such that |u (r, θ)| ≤ Cr (i.e., it must be of linear
growth).

Let us comment on the significance of Theorem 4 in view of what is known. Ni and

Tam in [4] showed that on a Kähler manifold with nonnegative bisectional curvature a
harmonic function must be of linear growth if it satisfies

u (r, θ)
lim sup =0
r →∞ r 1+δ

for all δ > 0. This estimate is sharp in the following sense. It is known that there are
complete noncompact surfaces of nonnegative Gaussian curvature that support har-
monic functions that grow like r 1+δ for 0 < δ < 1. Theorem 4 can be compared to
Ni and Tam’s theorem since in dimension 2 every orientable Riemannian manifold is
Kähler.

Proof of Theorem 4. First, we must recall, without a proof (one of which is via the
maximum principle), an important estimate due to Yau (see [10]) for the gradient of
a harmonic function defined on a surface of nonnegative Gaussian curvature. This is
a far-reaching generalization of Cauchy’s estimate for the gradient of a holomorphic
function defined on an open subset of the complex plane. To simplify the notation, in
what follows, we will suppress the subindex g indicating the dependence on the metric
of the gradient and the Laplace operator.

Theorem 5. Let u :  ⊂ R2 −→ R be a positive harmonic function with R2 being

endowed with a Riemannian metric of nonnegative Gaussian curvature. Then the esti-
mate
C
|∇ log u| ≤
r
holds on any ball of radius r contained in .

An immediate consequence of Yau’s estimate and the maximum principle is the

following.

Corollary 2. Let u : R2 → R be a harmonic function, where R2 is endowed with a

Riemannian metric of nonnegative Gaussian curvature. Suppose further that for some
δ > 0,
M(u; r )
lim infr →∞ = 0.
r 1+δ
Then, there exists a sequence rδ,k → ∞ so that we have the estimate
δ
|∇u| ≤ rδ,k on B(rδ,k ).

We leave the proof of this corollary to the interested reader. As a hint, notice that
Yau’s result cannot be directly applied to any harmonic function u (since it may take
on negative values). Once a sequence rδ,k → ∞ such that

  rδ,k
1+δ
M u, rδ,k < ,
3C

890 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 123

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
where C > 0 as in Yau’s estimate, has been fixed, apply Theorem 5 to
 
vk = u + M u; rδ,k + ,
 
with  > 0 arbitrary on B rδ,k , and use the maximum principle to obtain the corollary.
Before engaging in the proof of Theorem 4, we shall show a weaker result. To this
end, let us introduce Bochner’s identity

1
 |∇u|2 = |Hess u|2 + g (∇u, ∇u) + K g g (∇u, ∇u) .
2

We recommend the reader not familiar with this formula to prove it in the case of
Rn , where the term involving the curvature does not appear, and g is the usual inner
product. Recall that, in this case, Hess u, the Hessian of u, is the matrix of second
derivatives of u. The general case in a Riemannian manifold follows from the non-
conmutativity of the covariant derivatives, which is measured by the curvature (for
a proof of this formula, see Lemma 1.36 and Exercise 1.37 in [2], and beware that
Hess u = ∇∇u).
Now observe that, if u is harmonic and K g ≥ 0, then from Bochner’s identity
follows that |∇u|2 is subharmonic. Let us assume that

M (u; r )
lim inf  = 0.
r →∞ r log r

Then, by Yau’s estimate,

 
M |∇u|2 ; r
lim inf = 0,
r →∞ log r

so |∇u|2 is constant by Theorem 3 (take z (r ) = r , and the hypothesis hold since

K g ≥ 0), and hence, u must be of linear growth.
However, we can do much better. It is time to give a proof of Theorem 4. Following
the work of Ni and Tam, instead of using Bochner’s identity, we will make use of the
following identity, which is valid for any harmonic function u defined on a surface:
 
  2 |Hess u|2 + 2K g |∇u|2 1 + |∇u|2
 log 1 + |∇u| =
2
 2 . (1)
1 + |∇u|2

Not being as well known as Bochner’s, we shall give a proof of this identity in the last
paragraphs of this paper, so let us then continue with the proof of Theorem 4. 
Again, identity (1) implies in the case of nonnegative curvature that log 1 + |∇u|2
is subharmonic. Now use the hypothesis that if for δ > 0

M (u; r )
lim inf = 0,
r →∞ r 1+δ

then by Corollary 2, there is a sequence rδ,k → ∞ such that

  δ
log 1 + |∇u|2 ≤ log rδ,k .

November 2016] HARMONIC FUNCTIONS ON SURFACES 891

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
Since δ > 0 can be as small as we wish, as a consequence we can deduce that
   
M log 1 + |∇u|2 , r
lim inf = 0,
r →∞ log r
 
and from Theorem 3 we can conclude that log 1 + |∇u|2 is constant. Hence, |∇u| is
constant, i.e., u is of linear growth.

Proof of formula (1). As is customary, we pick a local orthonormal frame e1 , e2

around the point where we will be performing our computations. A subindex i will
denote covariant differentiation with respect to (or in the direction of, as you prefer) ei .
We shall enforce Einstein’s summation convention, i.e., we always add over repeated
subindices. For instance, we have

u = u j j = u 11 + u 22 , and |Hess u|2 = u i j u i j = u 211 + u 212 + u 221 + u 222 .

We are ready to begin:

 
   2
log 1 + |∇u|2 j j = ui j ui
1 + |∇u|2 j

4
= − 2 u i j u k j u i u k
1 + |∇u|2
2 2
+ u u +
2 ij ij
ui j j ui .
1 + |∇u| 1 + |∇u|2

The term u i j u k j u i u k when written in expanded form is

u 211 u 21 + 2u 12 u 11 u 1 u 2 + u 212 u 22 + u 222 u 22 + 2u 21 u 22 u 1 u 2 + u 221 u 21 ,

which, since u 11 = −u 22 and u 21 = u 12 , reduces to

u 211 u 21 + u 212 u 22 + u 222 u 22 + u 221 u 21 .

Now we need to be a little bit careful. Notice that for i = 1, 2,

u 11 u i = −u 22 u i ,

so we get that

u 211 |∇u|2 = u 211 u 21 + u 222 u 22 and u 222 |∇u|2 = u 211 u 21 + u 222 u 22 .

Hence,
 2   
u 11 + u 222 |∇u|2 = 2 u 211 u 21 + u 222 u 22 ,

which leads to
   
1 + |∇u|2 u i j u i j = 2 u 211 u 21 + u 222 u 22 + 2u 212 u 21 + 2u 212 u 22 + u i j u i j ,

892 
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 123

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms
from which we obtain
4 2 2u i j u i j
−  ui j uk j ui uk + u u =
2 ij ij 2 .
1 + |∇u| 2 2 1 + |∇u| 1 + |∇u|2

It is from u i j j u i that we obtain the curvature term. Indeed, the Ricci identity (Lemma
1.36 in [2]) tells us that

u j = (u) j + Ri j u i ,

where Ri j is the Ricci tensor of the metric. In the case of a surface, Ri j = K g gi j , where
K g is the Gaussian curvature, and hence, we have that

2 2 2K g |∇u|2
u u =
2 ijj i
R u u =
2 ki k i
,
1 + |∇u| 1 + |∇u| 1 + |∇u|2

and formula (1) follows.

ACKNOWLEDGMENT. The author wishes to thank the referees, the editors, and Professor Jaime Lesmes
for all their suggestions, which helped improve this manuscript, and the University of the Andes for providing
a great research environment.

REFERENCES

1. J. Cheeger, T. H. Colding, W. P. Minicozzi, Linear growth harmonic functions on complete manifolds

with nonnegative Ricci curvature, Geom. Funct. Anal. 5 (1995) 948–954.
2. B. Chow, P. Lu, L. Ni, Hamilton’s Ricci Flow. Graduate Studies in Mathematics, Vol 77, American
Mathematical Society, Science Press, New York, 2006.
3. P. Li, L.-F.Tam, Linear growth harmonic functions on a complete manifold, J. Differential Geom. 29
(1989) 421–425.
4. L. Ni, L.-F. Tam, Plurisubharmonic functions and the structure of complete Kähler manifolds with non-
negative curvature, J. Differential Geom. 64 (2003) 457–524.
5. D: Gilbarg, N. S. Trudinger, Elliptic Partial Differential Equations of Second Order. Reprint of the 1998
edition. Classics in Mathematics, Springer-Verlag, Berlin, 2001.
6. R. E. Greene, H. Wu, Function Theory on Manifolds which Possess a Pole. Lecture Notes in Mathematics,
Vol 699, Springer, Berlin, 1979.
7. J. Milnor, On deciding whether a surface is parabolic or hyperbolic, Amer. Math. Monthly 84 (1977)
43–46.
8. M. H. Protter, H. F. Weinberger, Maximum Principles in Differential Equations. Prentice-Hall, Engle-
wood Cliffs, NJ, 1967.
9. E. Nelson, A proof of Liouville’s theorem, Proc. Amer. Math. Soc. 12 (1961) 995.
10. S.-T.Yau, Harmonic functions on complete Riemannian manifolds, Comm. Pure Appl. Math. 28 (1975)
201–228.

JEAN C. CORTISSOZ received his Ph.D. in mathematics from Cornell University. He held a visiting posi-
tion at the University of Toledo in Ohio (where great memories of friendship remain: Deb, McKayla, Ray,
Henry, and Rao) before joining the faculty at Universidad de los Andes in 2006.
Departamento de Matemáticas, Universidad de los Andes, Bogotá DC, COLOMBIA.
[email protected]

November 2016] HARMONIC FUNCTIONS ON SURFACES 893

This content downloaded from 132.239.1.231 on Mon, 31 Oct 2016 18:17:25 UTC
All use subject to http://about.jstor.org/terms