7 Metric and Affine Conformal Geometry

We consider the metric extrinsic geometry of f : M -+ R in. relation to the

quantities associated to

L:= M HPI.

For brevity we write < .,.

> instead of < >R-

7.1 Surfaces in Euclidean Space

Let N, R denote the left and right normal vector of f : M H, i.e.

*df =
Ndf =
-df R.

Proposition 7. The second fundamental form II(X, Y) (X df (Y))--L of

f is given by

II(X, Y) (*df (Y)dR(X) -

dN(X) * df (Y)). (7.1)

2

Proof. We know from Lemma 2 that v -+ N(x)vR(x) is an involution with

the tangent space as its fixed point set:

Ndf (Y)R =
df (Y) (7.2)
Its (-I)-eigenspace is the normal space, so we need to compute

1
II (X, Y) =
(X df (Y)
- -
NX -
df (Y) R).
2

But differentiation of (7.2) yields

dN(X)df (Y)R + NX -
df (Y)R + Ndf (Y)dR(X) = X -
df (Y),

or

X -
df (Y) -
NX -
df (Y)R =
dN(X)df (Y)R + Ndf (Y)dR(X)
=
-dN(X) * df (Y) + *df (Y)dR(X).

F. E. Burstall et al.: LNM 1772, pp. 39 - 46, 2002

© Springer-Verlag Berlin Heidelberg 2002
40 7 Metric and Affine Conformal Geometry

Proposition 8. The mean curvature vector 'H trace II is given by

2

'Rdf (*dR + RdR), dfR (*dN + NdN). (7-3)

2 2

Proof. By definition of the trace,

4'H jdfJ2 =
*dfdR -
dN * df -

df * dR + *dNdf (7.4)
=
-df (*dR + RdR) + (*dN + NdN)df, (7.5)

but

(*dN + NdN)df =
*dNdf -
dN * df = -dN A df =
-d(Ndf)
=
-df A dR =
-df (*dR + RdR).

If follows that

27ildfI2 =
-df (*dR + RdR),

and

2 ldfTf dR + dRR)Tf =
(*dR + RdR)Tf

Similarly for N.

*
Proposition 9. Let K denote the Gaussian curvature of (M, f < >R)
and let K' denote the normal curvature of f defined by

Kj- :=< Rj- (X, JX) , N >R)

where X E TpM, and E-Lp M are unit vectors. Then

1
KIdf 12 =
(< *dR, RdR > + < *dN, NdN >) (7.6)
2

K
1
ldf 12
-1 (< *dR, RdR *dN, NdN >)
=

2
> -

< (7.7)

Proof.

Kldfl4(X) =< II(X, X), I.T(jX, jX) > _III(X, jX)12.

Therefore
 7.1 Surfaces in Euclidean Space 41

4KIdf 14 =< *df dR -

dN * df, -df * dR + *dNdf >

-
< *df * dR -
*dN * df, -df dR + dNdf >

= < N (df dR + dNdf ), -df * dR + *dNdf >

< N(df * dR + *dNdf), -df dR + dNdf >

< df dR + dNdf, N(-df * dR + *dNdf) >

< df * dR + *dNdf, N(-dfdR + dNdf) >

=- < dfdR + dNdf, dfR * dR + N * dNdf >

+ < df dR + *dNdf, dfRdR + NdNdf >

< dfdR, dfR * dR > -

< dfdR, N dNdf >

< dNdf, df R * dR > -

< dNdf, N dNdf >

+ < df * dR,dfRdR > + < df * dR,NdNdf >

+ < *dNdf, dfRdR > + < *dNdf, NdNdf >

ldf 12 < dR, RdR > -

< dfdR, N * dNdf >

+ < dNdf, Ndf dR > -ldf 12 < dN, N * dN >

+ ldf 12 < *dR, RdR > + < df * dR, NdNdf >

< *dNdf , Ndf dR > +jdf 12 < *dN, NdN >

I dyl2(< dR,R*dR > + < dN,N* dN >

< *dR, RdR > -

< *dN, NdN >)
21dfI2 (< dR,R * dR > + < dN,N * dN >).
This proves the formula for K. Using (7.1) and the Ricci equation

Kj- =< N II(X, JX), II(X, X) -

II(JX, JX) >,

we find, after a similar computation,

1
4K jdfJ2 =< *dR -

RdR, RdR > -

< *dN -

NdN, NdN >

+ < df (*dR -

RdR), NdNdf > -

< (*dN -

NdN)df, dfRdR >

On this we use (7.5) to obtain (7.7).

As a corollary we have

Proposition 10. The pull-back of the 2-sphere area under R is given by

R*dA =< *dR, RdR >

Integrating this for compact M yields

1
T7r fA
M
Kjdf 12
2
(deg R + deg N).

In 3-space (R =
N) this is a version ofthe Gauss-Bonnet theorem.
42 7 Metric and Affine Conformal Geometry

Proposition 11. We obtain

(J-H12 -
K -

Kj-) ldf12 4
1 * dR -

RdRJ2

Inparticular, if f : M -+ Im H = R' then Kj- =

0, and the classical Willmore
integrand is given by

(I Ij 12 -

K)Idfl2 1 * dR -

RdRJ2. (7.8)
4

Proof. Equations (7.3), (7.6), (7.7) give

(I,HI2 -
K -

K-L)Idfl2 11
4
* dR + RdRj2_ < *dR, RdR >

11
4
* dR12 +
4
1IRdRI2 2
< *dR, RdR >

1 2
41 * dR -

RdRI

7.2 The Mean Curvature Sphere in Affine Coordinates

We now discuss the characteristic properties of S in affine coordinates. We

describe S relative to the frame

0 1
i.e. we write S = GMG-1, where

G
01 f)
First, SL C L is equivalent to S EV -- H2 having the following matrix
representation:

0
S=
(1f) (' -R)
01 -H 0 1f) (7.9)

where N, R, H : M -+ H. From S2 = _I

2
N 2=-l=R ,
RH HN. (7.10)

The choice of symbols is deliberate: N and R turn out to be the left and right
normal vectors, of f while H is closely related to its mean curvature vector
,

The bundle L, has the nowhere vanishing section (fl) E V (L). Using this

section, we compute
 7.2 The Mean Curvature Sphere in Affine Coordinates 43

(f) (*df)
=
,

-df
is ird(S (f
=
7rd( (f R)) 7r((-dfR) (f) (-dR)) =

0
+
1
-7r
0 R)
sj
(f)
1
=
(f) Ir((Ndf) (f) (-Hdf)) (Ndf
irSd
1 0
+
1
= Ir
0

Therefore *6 = S6 = JS is equivalent to

*df =
Ndf =
-df R,

and we have identified N and R.

For the computation of the Hopf fields, we need dS. This is a straight-
forward but lengthy computation, somewhat simplified by the fact that
GdG = dG G-dG. We skip the details and give the result:
=

-dfH + dN -dfR Ndf)

(
-

dS = G
-dH -dR + HdfP7 '

-NdfH + NdN 0
SdS = G
(HdfH + RdH -
HdN Hdf R + R dR) G-1.

Rom this we obtain

Q = SdS -
*dS

NdN *dN 0
( *dR)
-

= G G-1
*dH + HdfH + RdH -
HdN 2HdfR + RdR +

4A = SdS + *dS

NdN + *dN 2Ndf H 0

( *dR)
-

G G-1.
.
-
* dH + HdfH + RdH -
HdN RdR -

The condition QIL =

0, and the corresponding AH C L, which we have not
used so far-, have the following equivalents:

2Hdf dR -
R * dR, (7.11)
2dfH dN -
N * dN. (7.12)

Together with equations (7.3) we find

2Hdf = dR -
R dR =
-R(*dR + RdR) -2RTtdf, =

2dfH = dN -
N dN =
-N(*dN NdN)
+ 2NdfR = =
-2dfRR,
and therefore

H = -RN = -RR. (7.13)

44 7 Metric and Affine Conformal Geometry

Remark 9. Given an immersed holomorphic curve L the mean cur-

vature vector of f at x E M is determined by Sx. On the other hand, S"

is the mean curvature sphere of Sx, see Example 17. Therefore S,' and f
have, in fact, the same mean curvature vector at x, justifying the name mean
curvature sphere.

Equations (7.11), (7.12) simplify the coordinate expressions for the Hopf
fields, which we now write as follows

Proposition 12.

dN + N dN 0
( 0) G-1,
*
4*Q=G -2dH + w
(7.14)

0 0
4*A= G
(w dR + R * dR) G-', (7.15)

f
where G
(01), and w = dH + H * dfH + R * dH -
H * dN.

Using (7.12) we can rewrite

w = dH + R * dH + 1H(NdN
2
-

*dN).

Proof. We only have to consider the reformulation of w. But

H * dfH -
H * dN IH * (dN -
N * dN) -
H * dN
2

-
1H* (dN+N*dN) H(NdN -

*dN).
2 2

7.3 The Willmore Condition in Affine Coordinates

We use the notations of the previous Proposition 12, and in addition abbre-
viate

v, = dR+R*dR.

Note that

V = -dR +. *dRR -- -dR -

R * dR = -v.

Proposition 13. The Willmore integrand is given by

1 1
< A A *A > =

16
JRdR -

*dR12 =

4
(IHI2 -
K -

K-L)JdfJ2.
For f : M -4 R, this is the classical integrand
1
< A A *A >= (Ih 12 -

K)Idfl2.
4
 7.3 The Willmore Condition in Affine Coordinates 45

Proof.

< A A *A > traceR(-A' -

(*A)) =

4
traceR(A2)
8
1
4
4 Re( 1V)
4
2

16
IV12 =

16
JdR + R * dR12
16
jRdR -

*dR12.
Now see Proposition 11 and, for the second equality, (7.8).
We now express the Euler-Lagrange equation d * A = 0 for Willmore
surfaces in affine coordinates. If we write 4 * A =
GMG-1, then

4d * A =
G(G-1 dG A M + dM + M A G-1 dG)G-1,
and again using G-'dG = dG we easily find
df A df A
4d * A = G
( dw
w

dv + w
v

A df) G-1.

Most entries of this matrix vanish:

Proposition 14. We have

df Aw=O (7.16)
df Av=O (7.17)
dv+wAdf =-(2dH-W)Adf =0. (7.18)
Proof. We have

df A w =
df A dH + df A R * dH + Idf A H(NdN
2
-

*dN)

=df AdH+dfRA*dH+ IdfH A (NdN

2
-

*dN)
1
df A dH -

*df A *dH+ dfH A (NdN -

*dN),
2
io
but

*dfH =
df (-R)H -df HN
2
*(NdN -

*dN) =
(N * dN -
N dN) =
-N(NdN -

*dN).
Hence, by type, the second term vanishes as well, and we get (7.16).
A similar, but simpler, computation shows (7.17)
Next, using (7.11), we consider

dv+wAdf =d(dR+R*dR)+wAdf
=
d(-2Hdf) + w A df
=
(-2dH + w) A df

=
(-dH + R * dH + 1H(NdN -

*dN)) A df.
2 "o
46 7 Metric and Affine Conformal Geometry

Again we show *a =
aN, ON. Then .(7.18) will follow by type.
Clearly

*(NdN -

*dN) N * dN + NdNN =
(NdN -

*dN)N,

showing *0 flN. Further

*a -
aN * dH -
RdH + dHN -

R(*dH)N
* dH -

d(RH) + (dR) H + d(HN) -

HdN -

R(d(HN) -

HdN)
=RH =RH

+R 2* dH + (dR)H -
HdN -
R * ((dR)H + RdH -

HdN)
(dR)H -
HdN -
R * (dR)H + RH * dN)
(dR -
R * dR)H -

H(dN -
N * dN)
2HdfH -

H(2dfH)
= 0.

As a corollary we get:

Proposition 15.

1 0 0 fdw -fdwf)
d*A=
4
G
(dw 0) G-1
4 (dw -dw

with w = dH + R * dH + !H(NdN
2
-

*dN).
Therefore f is Willmore if and only if dw 0.

Example 20 (Willmore Cylinder). Let -y : RIm H be a unit-speed curve,

and f : R2 -+ H the cylinder defined by

f (S' t) =
-Y(S) + t

with the conformal structure J-L = -2-. Then using Proposition 15, we obtain,
as at
after some computation, that f is (non-compact) Willmore, if and only if

Ir.3+ K11 -r.7, 2

=
0, (r,2-ol = 0.
2

This is exactly the condition that -y be a free elastic curve.