On 1-Harmonic Functions
Characterizations of entire subsolutions for the 1-harmonic equation of a constant 1-tension field are given with applications in geometry via transformation group theory. In particular, we prove that every level hypersurface of such a subsolution is calibrated and hence is area-minimizing over R; a...
Saved in:
| Published in: | Symmetry, Integrability and Geometry: Methods and Applications |
|---|---|
| Date: | 2007 |
| Main Author: | |
| Format: | Article |
| Language: | English |
| Published: |
Інститут математики НАН України
2007
|
| Online Access: | https://nasplib.isofts.kiev.ua/handle/123456789/146897 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| Cite this: | On 1-Harmonic Functions / S.W. Wei // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 22 назв. — англ. |
Institution
Digital Library of Periodicals of National Academy of Sciences of Ukraine| id |
nasplib_isofts_kiev_ua-123456789-146897 |
|---|---|
| record_format |
dspace |
| spelling |
Wei, S.W. 2019-02-11T21:15:45Z 2019-02-11T21:15:45Z 2007 On 1-Harmonic Functions / S.W. Wei // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 22 назв. — англ. 1815-0659 2000 Mathematics Subject Classification: 53C40; 53C42 https://nasplib.isofts.kiev.ua/handle/123456789/146897 Characterizations of entire subsolutions for the 1-harmonic equation of a constant 1-tension field are given with applications in geometry via transformation group theory. In particular, we prove that every level hypersurface of such a subsolution is calibrated and hence is area-minimizing over R; and every 7-dimensional SO(2) × SO(6)-invariant absolutely area-minimizing integral current in R8 is real analytic. The assumption on the SO(2) × SO(6)-invariance cannot be removed, due to the first counter-example in R8, proved by Bombieri, De Girogi and Giusti. This paper is a contribution to the Proceedings of the 2007 Midwest Geometry Conference in honor of Thomas P. Branson. Research was partially supported by NSF Award No DMS-0508661, OU Presidential International Travel Fellowship, and OU Faculty Enrichment Grant. The author wishes to thank Professors H. Blaine Lawson Jr., and Wu-Yi Hsiang for their interest, and Professor Herbert Federer for his help in Theorem 6 which essentially derives from him. The author also wishes to express his gratitude to the referees and the editors for their comments and suggestions without which the present form of this article would not be possible. en Інститут математики НАН України Symmetry, Integrability and Geometry: Methods and Applications On 1-Harmonic Functions Article published earlier |
| institution |
Digital Library of Periodicals of National Academy of Sciences of Ukraine |
| collection |
DSpace DC |
| title |
On 1-Harmonic Functions |
| spellingShingle |
On 1-Harmonic Functions Wei, S.W. |
| title_short |
On 1-Harmonic Functions |
| title_full |
On 1-Harmonic Functions |
| title_fullStr |
On 1-Harmonic Functions |
| title_full_unstemmed |
On 1-Harmonic Functions |
| title_sort |
on 1-harmonic functions |
| author |
Wei, S.W. |
| author_facet |
Wei, S.W. |
| publishDate |
2007 |
| language |
English |
| container_title |
Symmetry, Integrability and Geometry: Methods and Applications |
| publisher |
Інститут математики НАН України |
| format |
Article |
| description |
Characterizations of entire subsolutions for the 1-harmonic equation of a constant 1-tension field are given with applications in geometry via transformation group theory. In particular, we prove that every level hypersurface of such a subsolution is calibrated and hence is area-minimizing over R; and every 7-dimensional SO(2) × SO(6)-invariant absolutely area-minimizing integral current in R8 is real analytic. The assumption on the SO(2) × SO(6)-invariance cannot be removed, due to the first counter-example in R8, proved by Bombieri, De Girogi and Giusti.
|
| issn |
1815-0659 |
| url |
https://nasplib.isofts.kiev.ua/handle/123456789/146897 |
| citation_txt |
On 1-Harmonic Functions / S.W. Wei // Symmetry, Integrability and Geometry: Methods and Applications. — 2007. — Т. 3. — Бібліогр.: 22 назв. — англ. |
| work_keys_str_mv |
AT weisw on1harmonicfunctions |
| first_indexed |
2025-11-25T10:18:16Z |
| last_indexed |
2025-11-25T10:18:16Z |
| _version_ |
1850512495294283776 |
| fulltext |
Symmetry, Integrability and Geometry: Methods and Applications SIGMA 3 (2007), 127, 10 pages
On 1-Harmonic Functions?
Shihshu Walter WEI
Department of Mathematics, The University of Oklahoma, Norman, Ok 73019-0315, USA
E-mail: wwei@ou.edu
Received September 18, 2007, in final form December 17, 2007; Published online December 27, 2007
Original article is available at http://www.emis.de/journals/SIGMA/2007/127/
Abstract. Characterizations of entire subsolutions for the 1-harmonic equation of a con-
stant 1-tension field are given with applications in geometry via transformation group the-
ory. In particular, we prove that every level hypersurface of such a subsolution is calibrated
and hence is area-minimizing over R; and every 7-dimensional SO(2) × SO(6)-invariant
absolutely area-minimizing integral current in R8 is real analytic. The assumption on the
SO(2)×SO(6)-invariance cannot be removed, due to the first counter-example in R8, proved
by Bombieri, De Girogi and Giusti.
Key words: 1-harmonic function; 1-tension field; absolutely area-minimizing integral current
2000 Mathematics Subject Classification: 53C40; 53C42
1 Introduction
The study of 1-harmonic functions, or more generally that of p-harmonic maps is an area of an
active research that is related with many branches of mathematics. For instance, in a celebrated
paper of Bombieri, De Girogi and Giusti [3], a 1-harmonic function has been constructed to
provide a counter-example for interior regularity of the solution to the co-dimension one Plateau
problem in Rn for n > 7. Recall a C1 functions f : Rn → R is said to be 1-harmonic if it is
a weak solution of 1-harmonic equation
div
(
∇f
|∇f |
)
= 0 , (1.1)
where |∇f | is the length of the gradient ∇f of f , and for a C2 function f without a critical
point, div
(
∇f
|∇f |
)
is said to be the 1-tension field of f .
In this paper, characterizations of entire subsolutions for the 1-harmonic equation of a con-
stant 1-tension field are given in various aspects, and their relationships with calibration geomet-
ry are established (cf. Theorem 2, Corollary 3). As applications, we prove via transformation
group theory (cf. [9, 10, 13, 2, 21]) that the cone over S1×S5 is not minimizing in R8 but is sta-
ble; that any 7-dimensional SO(2)×SO(6)-invariant absolutely area-minimizing integral current
in R8 is real analytic; and that the only 7-dimensional SO(3)×SO(5)-invariant minimizing inte-
gral current with singularities in R8 is the cone over S2×S4, and is minimizing over R (cf. Theo-
rems 3–5). These results improved an early partial proof by numerical computation done by
Plinio Simoes [17] in his Berkeley thesis. The assumption on the SO(2) × SO(6)-invariance
cannot be removed, due to the first counter-example of Bombieri, De Girogi and Giusti that
the cone over S3( 1√
2
)×S3( 1√
2
) ⊂ S7(1) is area-minimizing in R8. It should be pointed out that
Fang-Hua Lin [14] proved that the cone over S1×S5 is one-sided area-minimizing and is stable by
a different method. By constructing 1-harmonic functions on hyperbolic space Hn, Hn ×Hn,
?This paper is a contribution to the Proceedings of the 2007 Midwest Geometry Conference in honor of
Thomas P. Branson. The full collection is available at http://www.emis.de/journals/SIGMA/MGC2007.html
mailto:wwei@ou.edu
http://www.emis.de/journals/SIGMA/2007/127/
http://www.emis.de/journals/SIGMA/MGC2007.html
2 S.W. Wei
Hn × SO(n, 1) and many other associated spaces, S.P. Wang and the author [19] show the
Bernstein Conjecture in these spaces to be false in all dimensions. In particular, these construc-
tions give the first set of examples of complete, smooth, embedded, minimal (hyper-)surfaces in
hyperbolic space Hn in all dimensions (cf. also Remark 3(ii)).
2 Fundamentals in geometric measure theory
For our subsequent development, we recall some fundamental facts, definitions, and notations,
for which the reference is Federer’s book [5] and paper [7].
Let N denote an n-dimensional Riemannian manifold and denote by Rloc
p (N) the set of p-
dimensional, locally rectifiable currents (of Federer and Fleming, cf. [8]) on N . For S ∈ Rloc
p (N),
denote the mass of S by M(S), and the boundary of S by ∂S, and is given by (∂S)(w) = S(dw),
where w is a smooth p-form and d is the exterior differentiation. From a calculus of variational
viewpoint, we make the following
Definition 1. A current T ∈ Rloc
k (N) is said to be stationary if d
dtM(φV
t∗(T ))|t=0 for all vector
fields V on N with compact support where φV
t is the flow associated with V , and stable if for
every vector fields V on N with compact support, there exists an ε > 0 such that M(T ) ≤
M(φV
t∗(T )) for |t| < ε.
We are primarily interested in minimizing currents.
Definition 2. A current T ∈ Rloc
k (N) is homologically (resp. absolutely) area-minimizing over Z
if for all compact sets K ⊂ M , we have M(φKT ) ≤ M((φKT ) + S) for all S ∈ Rloc
k (N) having
compact support and being the boundary of some current in Rloc
k+1(N) with compact support
(resp. the empty boundary)(here φK denotes the characteristic function on K).
Using a dimension reduction technique, Federer proves that the support of an area-minimizing
integral current T [8] minus another compact set S whose Hausdorff dimension does not exceed
n − 8 is an (n − 1)-dimensional analytic manifold [6]. Hence, if n ≤ 7, then S = ∅. If n = 8,
S consists of at most isolated points [5, 5.4.16]. This result is optimal by the counter-example
due to Bombieri–De Giorgi–Giusti [3] that {x ∈ R2m : x2
1 + · · ·+ x2
m = x2
m+1 + · · ·+ x2
2m} is an
area-minimizing cone over the product of (m−1)-spheres
{
x ∈ R2m : x2
1 + · · ·+x2
m = x2
m+1 + · · ·
+ x2
2m = 1
2
}
in R2m for m ≥ 4.
The union of the groups Fm,K(U) = {R+∂T : R ∈ Rm,K(U), T ∈ Rm+1,K(U)} corresponding
to all compact K ⊂ U is the group Fm(U) of m-dimensional integral flat chains in an open
subset U of Rn. We denote the group ofm-dimensional integral flat chains, cycles and boundaries
by Fm(A) = Fm(Rn)∩{S : sptS ⊂ A}, Zm(A,B) = Fm(A)∩{S : ∂S ⊂ Fm(B) or m = 0}, and
Bm(A,B) = {R+ ∂T : R ∈ Fm(B), T ∈ Fm+1(A)} respectively. Similarly, we define and denote
Fm(A), Zm(A,B) and Bm(A,B) the vector space of m-dimensional real flat chains, cycles and
boundaries respectively, where B ⊂ A are compact Lipschitz neighborhood retract in U .
For every positive convex parametric integrand ψ, and every compact subset K of A, we
define Zm,K(A,B) = Zm(A,B) ∩ {R : sptR ⊂ K}, Bm,K(A,B) = Bm(A,B) ∩ {R : sptR ⊂ K},
Zm,K(A,B) = Zm(A,B) ∩ {R : sptR ⊂ K}, and Bm,K(A,B) = Bm(A,B) ∩ {R : sptR ⊂ K},
and make the following
Definition 3. An m-dimensional rectifiable current Q (resp. Q′) is said to be absolutely (resp.
homologically) ψ-minimizing in K with respect to (A,B) over Z if∫
Q
ψ = inf
{∫
S
ψ : S ∈ Fm,K(U), Q− S ∈ Zm,K(A,B)
}
(
resp.
∫
Q′
ψ = inf
{∫
S
ψ : S ∈ Bm,K(U), Q′ − S ∈ Bm,K(A,B)
} )
.
On 1-Harmonic Functions 3
Definition 4. An m-dimensional real flat chain Q (resp. Q′) is said to be absolutely (resp.
homologically) ψ-minimizing in K with respect to (A,B) over R if∫
Q
ψ = inf
{∫
S
ψ : S ∈ Fm,K(U), Q− S ∈ Zm,K(A,B)
}
(
resp.
∫
Q′
ψ = inf
{∫
S
ψ : S ∈ Bm,K(U), Q′ − S ∈ Bm,K(A,B)
} )
.
We will make comparisons between real and integral absolute (resp. homological) minimizing
currents in the subsequent Sections 3, 4, and 5.
3 Characterizations of subsolutions for 1-harmonic equation
of constant 1-tension field
We connect an entire subsolution of this sort, with a calibration. Recall a calibration is a closed
form with comass 1.
Lemma 1. Let M be a complete noncompact Riemannian manifold. For any x0 ∈ M and
any pair of positive numbers s, t with s < t, there exists a rotationally symmetric Lipschitz
continuous function ψ(x) = ψ(x; s, t) and a constant C1 > 0 (independent of x0, s, t) with the
properties:
(i) ψ ≡ 1 on B(x0; s), and ψ ≡ 0 off B(x0; t);
(ii) |∇ψ| ≤ C1
t− s
, a.e. on M. (3.1)
Proof. (cf. Andreotti and Vesentini [1], Yau [22], Karp [11]). �
Theorem 1. Let Ω be a domain in Rn containing a ball B(x0, r) of radius r, centered at x0,
and g : Ω → R be a continuous function with g ≥ 0, and c = inf
x∈B(x0, r
2
)
g(x). Let f : Ω → R be
a C1 weak solution of
div
(
∇f
|∇f |
)
= g(x) on Ω, (3.2)
then the inf imum c satisf ies
0 ≤ c ≤ C12n
r
,
where C1 is as in (3.1).
Proof. Let ψ ≥ 0 be as in Lemma 1, in which M = Rn, t = r, s = r
2 . Choose ψ to be a test
function in the distribution sense of (3.2). Then via the assumption on g, and Cauchy–Schwarz
inequality we have:∫
B(x0, r
2
)
cψ(x)dx ≤
∫
B(x0, r
2
)
g(x)ψ(x)dx
≤
∫
B(x0,r)
g(x)ψ(x)dx = −
∫
B(x0,r)
∇f
|∇f |
· ∇ψdx ≤
∫
B(x0,r)
|∇ψ|dx.
Hence,
cVol
(
B
(
x0,
r
2
))
≤ C1
r
Vol(B(x0, r))
yields the desired. �
4 S.W. Wei
Corollary 1. Let f : Rn → R be a C1 weak subsolution of 1-harmonic equation (1.1) with
constant 1-tension field c, i.e. 0 ≤ div
(
∇f
|∇f |
)
= c in the distribution sense. Then f is a 1-har-
monic function.
Corollary 2. There does not exist a C1 weak subsolution f : Rn → R of equation (3.2) with
lim
r→∞
inf
x∈B(x0,r)
g(x) > 0, for any x0 ∈ Rn.
Let A ⊂ Rn be an open set. We denote BVloc(A) = {f ∈ L1
loc(A): the distributional
derivatives Dif of f are (locally) measures}= {f ∈ L1
loc(A) : suppφn ⊂ K ⊂ A, φn → 0
uniformly, imply
(
∂
∂xi
f
)
φn → 0}. Let Df = (D1f, . . . , Dnf) denote the gradient of f in the
sense of distributions and |Df | the scalar measure defined by
∫
K |Df | = sup
∫
K
∑
i εi(x)Dif ,
where the supremum is taken over all sets {εi(x), i = 1, . . . , n} of C∞(K) functions which satisfy∑
ε2i (x) ≤ 1.
Definition 5. A function f ∈ BVloc(A) has least gradient in A if for every g ∈ BVloc(A), with
compact support K ⊂ A we have∫
K
|Df | ≤
∫
K
|D(f + g)|. (3.3)
Definition 6. Let E be a set in Rn and φE its characteristic function. E has an oriented
boundary of least area with respect to A, if (i) φE ∈ BVloc(A) and (ii) for each g ∈ BVloc(A)
with compact support K ⊂ A we have
∫
K |DφE | ≤
∫
K |D(φE + g)|.
Theorem 2. Let f ∈ H1,1
loc (Rn), and ∇f(x) 6= 0 for every x in Rn. Let Eλ = {x : f(x) ≥ λ},
and Sλ = {x : f(x) = λ}. We denote the set of integers by Z. Then the following thirteen
statements (1)–(13) are equivalent and each of them implies the fourteenth statement (14).
1. f : Rn → R is a C1 weak subsolution of (1.1) with constant 1-tension field.
2. f is a C1 weak solution of (1.1) on Rn.
3. f is a C1 1-harmonic function on Rn.
4. For each (a, t0) = (a1, . . . , an−1, t0) ∈ Sλ, there exists a neighborhood D of a in Rn−1, and
a unique real analytic function η : D → R such that η(a) = t0, f(x1, . . . , xn−1, η(x1, . . . ,
xn−1)) = λ and div
(
∇η√
1+|∇η|2
)
= 0 on D.
5. Each level hypersurface Sλ is minimal in Rn.
6. ∗df
|df | is a globally defined “weakly” closed form with comass 1.
7. f is a function of least gradient in Rn.
8. Each Eλ, λ ∈ R has an oriented boundary of least area with respect to Rn.
9. Each level hypersurface Sλ is absolutely area-minimizing in Rn over Z.
10. Each level hypersurface Sλ is absolutely area-minimizing in Rn over R.
11. Each level hypersurface Sλ is homologically area-minimizing in Rn over R.
12. Each level hypersurface Sλ is homologically area-minimizing in Rn over Z.
13. Each level hypersurface Sλ is stable in Rn.
14. If f ∈ C2(Rn), then ∗df
|df | is closed and the restriction ∗df
|df |
∣∣∣
Sλ
is its volume form, hence each
Sλ is real absolutely area-minimizing in Rn over R.
On 1-Harmonic Functions 5
Corollary 3. Every level hypersurface of a C2 subsolution of 1-harmonic equation on Rn+1 with
constant 1-tension field is calibrated and hence is area-minimizing over R.
Proof. (1) ⇔ (2) ⇔ (3) : This follows immediately from Corollary 1.
(2) ⇔ (4) : (⇒) Let f(x1, . . . , xn−1, t) = η(x1, . . . , xn−1)− t. The assertion follows from the
implicit function theorem and
0 =
∫ n−1∑
i=1
∂f
∂xi
∂ϕ
∂xi
|∇f |
+
∫ ∂f
∂t
|∇f |
∂ϕ
∂t
=
∫ n−1∑
i=1
∂η
∂xi√
1 + |∇η|2
∂ϕ
∂xi
(3.4)
for all ϕ ∈ C∞0 (D × R). The regularity of solutions of minimal surface equation implies that η
is real analytic and completes the proof. (⇐) This follows immediately from (3.3).
(4) ⇔ (5) : This is due to the fact that the graph of a solution to the minimal surface equation
on D is a minimal hypersurface in D × R.
(2) ⇔ (6) : This follows from the following: For every φ ∈ C∞0 (A),∫
A
∗df
|df |
∧ dφ =
∫
A
n∑
i,j=1
(−1)i−1
∂f
∂xi
|∇f |
dx1 ∧ · · · ∧ dxi−1 ∧ dxi+1 ∧ · · · ∧ dxn ∧ ∂φ
∂xj
dxj
=
∫
A
n∑
i=1
(−1)n−1
∂f
∂xi
∂φ
∂xi
|∇f |
dx1 ∧ · · · ∧ dxi ∧ · · · ∧ dxn.
(2) ⇒ (7): let us first assume that g ∈ C1
0 (A). Let h(t) =
∫
|D(f + tg)|. Then
h′(t) =
∫ (
n∑
i=1
∂(f+tg)
∂xi
∂g
∂xi
)
(
n∑
i=1
(
∂(f+tg)
∂xi
)2
) 1
2
.
Hence h′(0) = 0 by assumption. Furthermore,
h′′(t) =
∫ (
n∑
i=1
( ∂g
∂xi
)2
)(
n∑
i=1
(∂(f+tg)
∂xi
)2
)
−
(
n∑
i=1
∂(f+tg)
∂xi
∂g
∂xi
)2
[
n∑
i=1
(∂(f+tg)
∂xi
)2
] 3
2
≥ 0,
by the Cauchy–Schwarz inequality. Therefore
∫
|Df | = h(0) ≤ h(1) =
∫
|D(f + g)|. If g ∈
BVloc(A) with compact support K and let Dg = G1 + G2 where G1 is completely continuous
and G2 is the singular part of Dg with support Ng of measure zero. Then we have
∫
K |D(f +
g)| =
∫
K |Df + G1| +
∫
K |G2| because f ∈ H1,1
loc (A). Let gε = g ∗ ψε where ψε is a mollifier.
Then g ∈ C1
0 (A) and
∫
Kε
|Df | ≤
∫
Kε
|D(f + gε)| ≤
∫
Kε
|Df + G1 ∗ Ψε| +
∫
A |G2 ∗ Ψε|, where
Kε = {x ∈ A : dist(x,K) < ε}. Letting ε→ 0 completes the proof (cf. [3]).
(7) ⇒ (8) : This follows from Coarea formula for BV functions [15],
∫
K |Df | =
∞∫
−∞
(
∫
K |Dφλ|)dλ
together with two observations:
(i) If f1 and f2 satisfy (3.3), so does sup(f1, f2).
(ii) If fi ∈ BVloc(A), fi → f in L1
loc and each fi satisfies (3.3), so does also f ∈ BVloc(A) and
satisfies (3.3).
For detailed proof see [16].
6 S.W. Wei
(8) ⇒ (9) : Let φλ = φEλ
. Since for every x in Rn, ∇f(x) 6= 0, ∂Eλ = Sλ for Sλ 6= ∅.
It follows from a theorem of Miranda [15] that on any compact set K in Rn, the Hausdorff
(n− 1)-measure
Hn−1(K ∩ Sλ) =
∫
K
|Dφλ| ≤
∫
K
|D(φλ + g)| = Hn−1(K ∩ T )
for all sets T with ∂(K ∩ T ) = ∂(K ∩ Sλ).
(9) ⇒ (10) : It follows from Theorem 6.
(10) ⇒ (11) ⇒ (12) : Since absolute area-minimization over R ⇒ homological area-minimi-
zation over R ⇒ homological area-minimization over Z.
(12) ⇒ (13) ⇒ (5) : Since homological minimization over Z ⇒ stability ⇒ minimality. This
completes the proof of (1) ⇔ · · · ⇔ (13).
(2) ⇒ (14) : If f ∈ C2(A) then by (3.4) ∗df
|df | is closed. Now let e1, . . . , en−1 be an orthonormal
basis for the tangent space of Sλ at x0 and ν a unit normal vector at x0. We denote by tilde
“∼” the canonical isomorphism between a tangent space and its dual space. To show ∗df
|df | has
comass 1, note for any (n− 1)-vector field ξ,
∗df
|df |
(ξ) =
(
∗ ∇̃f
|∇f |
)
(ξ)
(
because
df
|df |
(X) =
Xf
|∇f |
=
〈
∇f
|∇f |
, X
〉)
= (∗ν̃)(ξ) = ( ˜e1 ∧ · · · ∧ en−1)(ξ) = 〈e1 ∧ · · · ∧ en−1, ξ〉.
In particular ∗df
|df |(e1 ∧ · · · ∧ en−1) = 1, ∗df
|df |(ξ) ≤ 1 and ∗df
|df |
∣∣∣
Sλ
= volume element of Sλ. By
the formalism of Stokes theorem, for any integral current T with ∂T = ∂(Sλ ∩Br)
M(Sλ ∩Br) = (Sλ ∩Br)
(
∗df
|df |
)
= T
(
∗df
|df |
)
=
∫
∗df
|df |
(−→
Tx
)
d||T ||(x) ≤
∫
d||T || = M(T ),
where
−→
T is the field of oriented unit tangent planes to T . �
Remark 1. In Theorem 2, if one replace Rn with an open subset A in Rn, then assertions
(2)⇔ · · · ⇔ (13) ⇒ (14) remain to be true.
Remark 2. Concerning the assertion (2) ⇒ (7), a stronger theorem can be found in [3]: Let
A ⊂ Rn be an open set and let f ∈ H1,1
loc (A). Suppose that (i) Hn({x ∈ A : |∇f | = 0}) = 0,
(ii) Hn−1(N) = 0 where N is a closet set in A, (iii)
∫
A−N |∇f |
−1
n∑
i=1
∂f
∂xi
∂φ
∂xi
dx = 0 for every
φ ∈ C1
0 (A−N). Then f has least gradient with respect to A.
Remark 3. (i) The assertion (7) ⇒ (9) is due to Miranda.
(ii) Connecting the assertions (5), (6), and (12) on Riemannian manifolds, S.P. Wang and the
author [19] prove that if each level hypersurface of a smooth function f : M → R on an oriented
Riemannian manifold M with nowhere vanishing ∇f , is minimal, then there exists a closed form
with comass 1 on M and hence each level hypersurface is homologically area-minimizing over R.
Corollary 4. Let A be an open subset in Rn, N be a closed subset in A with Hn−1(N) = 0.
Then the graph of any weak solution of the minimal surface equation
n∑
i=1
∂
∂xi
(
∂f
∂xi√
1+|∇f |2
)
= 0
on A−N is in fact absolutely area-minimizing in A× R ⊂ Rn+1 over R.
On 1-Harmonic Functions 7
Proof. Applying (3.4) in which “f(x1, . . . , xn−1, t) = η(x1, . . . , xn−1) − t” is replaced with
“F (x1, . . . , xn, t) = f(x1, . . . , xn) − t”, and Remark 2, we have that F is a C1 1-harmonic
function in A. By Theorem 2, the zero level set S0 = {(x1, . . . , xn, t) : t = f(x1, . . . , xn)} is
absolutely area-minimizing in A× R ⊂ Rn+1 over R. �
4 Further applications
A natural question arises: Are Bombieri–De Giorgi–Giusti and Lawson cones the only SO(m)
×SO(n)-invariant singular absolutely area-minimizing integral currents in Euclidean
space Rm+n+2? The answer is affirmative. Combining the theory of 1-harmonic functions
developed, and the techniques of transformation groups in [10, 13, 2], and [21], evolved from the
ideas in [9], one obtains the following:
Theorem 3. The cone C(Sm × Sn) over Sm × Sn is the unique singular absolutely area-
minimizing hypersurface in the class of SO(m + 1) × SO(n + 1)-invariant integral currents
in Rm+n+2 over R for m+ n > 7 or m+ n = 6, |m− n| ≤ 2. (It is known that the cone is not
even stable otherwise.)
Proof. Assume m = n. Let Lie group G = SO(n + 1) × SO(n + 1) acting on manifold
Rn+1 × Rn+1 in the standard way, i.e. assigning
(
(A,B), (x, y)
)
∈ G× R2n+2 to (A · x,B · y) ∈
R2n+2, where “·” is the matrix multiplication. Then the collection X of principle orbits is
given by X = {(x, y) ∈ R2n+2 : |x||y| 6= 0}, where “| · |” is the length of “·” in Rn+1. The
orbit space which is stratified, can be represented as R2n+2/G = {(u, v) ∈ R2 : u, v ≥ 0} =
X ∪ {(u, v) ∈ R2 : u = 0, v > 0} ∪ {(u, v) ∈ R2 : u > 0, v = 0} ∪ {(0, 0)}. The canonical
metric on R2n+2/G (compatible with the fibration over each stratum) is the usual flat one
ds20 = du2 + dv2. The canonical projection π : R2n+2 → R2n+2/G is given by π(x, y) = (|x|, |y|),
and let X/G = π(X). Then the length of a curve σ in (X/G, ds20) is the length of any orthogonal
trajectory through the corresponding orbits in X, and 2n-dimensional volume of π−1((u, v))
(which is diffeomorphic to Sn × Sn) is proportional to unvn, for (u, v) ∈ X/G. Thus if we
choose the metric ds2 = u2nv2n(du2 + dv2) on R2n+2/G , then by Fubini’s theorem, the length
of a curve σ in (R2n+2/G, ds2) is equal to (2n + 1)-dimensional volume of hypersurface π−1σ
(with possible singularities) in R2n+2, up to a constant factor. It follows that σ is a length
minimizing geodesic “downstairs” (in (R2n+2/G, ds2)), if and only if π−1σ is area-minimizing in
the class of G-invariant (2n+1)-dimensional currents “upstairs” (in (R2n+2, dx2
1+· · ·+dx2
2n+2)),
or equivalently, π−1σ is area-minimizing in (R2n+2, dx2
1 + · · · + dx2
2n+2) in general (cf. [13], [2,
p. 174, 6.4] and [21]). Furthermore, if a length minimizing geodesic σ meets the boundary
{(u, v) ∈ R2 : u = 0, v > 0} ∪ {(u, v) ∈ R2 : u > 0, v = 0}, it meets the boundary orthogonally
by the first variational formula for the arc-length functional, and the corresponding π−1σ is
a regular, embedded and analytic hypersurface in R2n+2. If σ meets the vertex {(0, 0)}, then
π−1σ is singular. Therefore, it suffices to show that any curve in R2n+2/G, other than the
diagonal ray emanating from the origin is not absolutely length minimizing with respect to the
metric ds2 = u2nv2n(du2 + dv2).
8 S.W. Wei
Now let Γ = {(u0(t), v0(t))} be the geodesic through (1, 0) in (R2n+2/G, ds2), and Γλ =
{(λu0(t), λv0(t))}, λ > 0. In [3], a 1-harmonic function was constructed in such a way that the lift
of family {Γλ} of these homothetic geodesics are level hypersurfaces in (R2n+2, dx2
1+· · ·+dx2
2n+2).
Hence Γλ is absolutely length minimizing in (R2n+2/G, ds2) (cf. also Theorem 2, Remark 2). Now
suppose Theorem 3 were not true. Then there would exist a curveQP ⊂ Γλ transverse to a length
minimizing curve OP . It follows that the length l(OP ) of OP would satisfy l(OP ) = l(QP ).
Consider the curve OPR where R is on the curve Γλ, and l(OPR) = l(QPR). Then the curve
OPR would be a geodesic, and hence smooth at P . This is a contradiction. Similarly, one can
show the remaining case m 6= n. �
Theorem 4. The cone C(S1 × S5) over S1 × S5 is not absolutely area-minimizing, although it
is stable.
Proof. Suppose, on the contrary, that the cone were absolutely area-minimizing. Then consider
Lie group G = SO(2)×SO(6) acting on manifold R2×R6 in the standard way. By the previous
argument, this would imply the line segment OP were length-minimizing in (R8/G, ds2), where
ds2 = u2v6(du2 + dv2). On the other hand, based on the study of Simoes’ thesis [17], [13]
and [21], the level curve (uλ, vλ) in the u, v-plane is absolutely length-minimizing. Argue as
before, the curve OPR would be smooth at P . This is a contradiction. The stability of the cone
follows from Simons’ work [18]. �
Theorem 5. Any 7-dimensional SO(2)× SO(6)-invariant absolutely area- minimizing integral
current in R8 is real analytic.
Proof. By the argument given in the proof of Theorem 3, it suffices to show that any curve
in R2n+2/G, from the origin is not absolutely length minimizing with respect to the metric
ds2 = u2v6(du2 +dv2). By Theorem 4, the diagonal ray emanating from the origin is not length
minimizing. Similarly, if there were an absolutely length minimizing curve starting from the
origin lying above v =
√
5u, then this would lead to an irregularity of a geodesic, a contradic-
tion. �
5 Comparison theorem
It is known that each level hypersurface of a function of least gradient defined on an open
subset A ⊂ Rn is absolutely area-minimizing in A over Z. It is tempting to ask it if is absolutely
area-minimizing in A over R. This motivates our discussion on comparison between real and
integral absolute (or homological) minima. In general they are distinct. Examples are given by
Almgren [7, 5.11], Federer [7] and Lawson [12]. Furthermore, in the case of 1-dimensional (or
co-dimension 1) integral flat chains, Federer [7] has shown that real and integral homological
(or absolute) minimizing are the same.
Let M be a locally Lipschitz neighborhood retract in Rn (i.e. there exists a locally Lipschitz
map which retracts a neighborhood of M onto M), M be an open subset of M , and A be an
open subset of Rn. Using the assumption on vanishing topology, an exhaustion of M by an
increasing sequence of compact set Ki ⊂M , we obtain the following:
On 1-Harmonic Functions 9
Theorem 6. (1) Let Tn−1 denote a codimension 1 integral absolutely area-minimizing rectifiable
current in M with homology group Hn−1(M) = 0. Then Tn−1 is absolutely area-minimizing in M
if and only if Tn−1 is absolutely area-minimizing in A; and if and only if Tn−1 is real absolutely
area-minimizing in A. (2) Let H1(M) = 0. T 1 is a homologically area-minimizing rectifiable
current of degree 1 of M if and only if T 1 is real homologically area-minimizing in M .
We have the following immediate
Corollary 5. The level hypersurface of a function of least gradient in an open subset A of Rn
is absolutely area-minimizing over R.
Corollary 6. Let N be a closed set in A ⊂ Rn with Hn−1(N) = 0. The graph of any weak
solution of the minimal surface equation
n∑
i=1
∂
∂xi
(
∂f
∂xi√
1+|∇f |2
)
= 0 on A−N is in fact absolutely
area-minimizing in A× R ⊂ RN+1 over R.
Corollary 7. All the examples we find in [21] are absolutely area-minimizing over R.
Acknowledgements
Research was partially supported by NSF Award No DMS-0508661, OU Presidential Inter-
national Travel Fellowship, and OU Faculty Enrichment Grant. The author wishes to thank
Professors H. Blaine Lawson Jr., and Wu-Yi Hsiang for their interest, and Professor Herbert
Federer for his help in Theorem 6 which essentially derives from him. The author also wishes to
express his gratitude to the referees and the editors for their comments and suggestions without
which the present form of this article would not be possible.
References
[1] Andreotti A., Vesentini E., Carleman estimate for the Laplace–Beltrami equation on complex manifolds,
Inst. Hautes Études Sci. Publ. Math. 25 (1965), 81–130.
[2] Brothers J.E., Invariance of solutions to invariant parametric variational problems, Trans. Amer. Math. Soc.
262 (1980), 159–179.
[3] Bombieri E., de Giorgi E., Giusti E., Minimal cones and the Bernstein problem, Invent. Math. 7 (1969),
243–268.
[4] Federer H., Some theorems on integral currents, Trans. Amer. Math. Soc. 117 (1965) 43–67.
[5] Federer H., Geometric measure theory, Springer, Berlin – Heidelberg – New York, 1969.
[6] Federer H., The singular sets of area minimizing rectifiable currents with codimension one and of area
minimizing flat chains modulo two with arbitrary codimension, Bull. Amer. Math. Soc. 76 (1970), 767–771.
[7] Federer H., Real flat chains, cochains and variational problems, Indiana Univ. Math. J. 24 (1974), 351–406.
[8] Federer H., Fleming W.H., Normal and integral currents, Ann. of Math. (2) 72 (1960), 458–520.
[9] Hsiang W.Y., On the compact homogeneous minimal submanifolds, Proc. Natl. Acad. Sci. USA 56 (1966),
5–6.
[10] Hsiang W.Y., Lawson H.B. Jr., Minimal submanifolds of low cohomogeneity, J. Differential Geom. 5 (1971),
1–38.
[11] Karp L., Subharmonic functions on real and complex manifolds, Math. Z. 179 (1982), 535–554.
[12] Lawson H.B., The stable homology of a flat torus, Math. Scand. 36 (1975), 49–73.
[13] Lawson H.B., The equivariant Plateau problem and interior regularity, Trans. Amer. Math. Soc. 173 (1972),
231–249.
[14] Lin F.-H., Minimality and stability of minimal hypersurfaces in RN , Bull. Austral. Math. Soc. 36 (1987),
209–214.
10 S.W. Wei
[15] Miranda M., Sul minimo dell’integrale del gradiente di una funzione, Ann. Scuola Norm. Sup. Pisa (3) 19
(1965), 626–665.
[16] Miranda M., Comportamento delle successioni convergenti di frontiere minimali, Rend. Sem. Mat. Univ.
Padova 39 (1967), 238–257.
[17] Simoes P., A class minimal cones in Rn, n ≥ 8, that minimizes area, Thesis, Berkeley, 1973.
[18] Simons J., Minimal varieties in riemannian manifolds, Ann. of Math. (2) 88 (1968), 62–105.
[19] Wang S.P., Wei S.W., Bernstein conjecture in hyperbolic geometry, in Seminar on Minimal Submanifolds,
Editor E. Bombieri, Ann. of Math. Stud., no. 103, 1983, 339–358.
[20] Wei S.W., Minimality, stability, and Plateau’s problem, Thesis, Berkeley, 1980.
[21] Wei S.W., Plateau’s problem in symmetric spaces, Nonlinear Anal. 12 (1988), 749–760.
[22] Yau S.T., Some function theoretic properties of complete Riemannian manifolds and their applications to
geometry, Indiana Univ. Math. J. 25 (1976), 659–670.
1 Introduction
2 Fundamentals in geometric measure theory
3 Characterizations of subsolutions for 1-harmonic equation of constant 1-tension field
4 Further applications
5 Comparison theorem
References
|