Baixe (ebook - math) - Differential Geometry In Physics (Lugo, 1998) e outras Notas de estudo em PDF para Física, somente na Docsity!
Differential Geometry in Phys
Gabriel Ingo
Department of Mathematical Sciences
University of North Carolina at Wilmington
si
Copyright 1995, 1998
his document was reproduced by the University of North Carolina al Wilminglon from a camera
ready copy supplied by le authors. Lhe lext was generated on an desktop computer using EIFX.
Copyright (1992, 1995
All rights reserved. No part of this publication may be reproduced, stored in a retricval system,
or transmitted, in any form or by any means, electronic. mechanical, photocopying. recording or
otherwise, without the written permission of the authors. Printed in the United States of Americ
dv
Preface
1 Vectors and Curves
1 Tangent Vec
12 Curvesin R
1.3 Tundamental Theorem of Curves
2 Differential Forms
2.10 IForms cce e e e
2.2 Tensors and Torms of Thigher Rank ..
23 Exterior Derivatives .
24 The Todge-s Operator «cc
3 Connections
8.1 Frames
32 Curvilinear Coordinates
33 €
3.4 Cartan Equations cc.
variant Derivative . cc.
4 Theory of Surfaces
41 Manifolds .lccccccccc e
42 The Tirst Tundamental Form
The Second Fundamental Form ..
44 Curvalute cc eee e e
Chapter 1
1.1 Tangent Vectors
1.1 Definition Euclidem n-space R?º is defincd as the set of ordered n-tuples p = (p1,...,9"),
where pº ER. forcachi=1,....m.
Given any two n-tuples p= (p!,..p), q=(g',...,4') and any real number c, we define two
operations:
p+tg o (prq +ç) (1.1)
cp = (ep! cp")
With the sum and the scalar multiplication of ordered n-tuples defined this way, Tuclidean space
acquires the structure of a vector space of n dimensions
1.2 Definition Let x) be the rcal valued functions in Rº such that zi(p) = pi for any poimt
p= (p!,...,p"). The functions 2º are then called the natural coordinates ol We the point p. Wen
el =yadal=s
the dimension ol the space n = 3, we olten write: x! =
1.3 Definition À real valued function in Rº is ol class C” if all the partial derivatives of the
function up to order 7 exist and are continuous, “The space of infinitely differentiable (smooth)
functions will be denoted by CS(R).
Tn advanced caleulus, vectors are usually regarded as arrows characterized by a direction and a
ion in space. Because of physical
amd mathematical reasons, it is advantageous to introduce a notion of vectors which does depend em
ample, if the vector is to represent a force acting on a rigid body, then the resulting
equations of motion will obviously depend on the point at which the force is applicd.
lu a later chapter we will also consider vectors on spaces which are curved, In Lhese cases the
position of the vectors is crucial, for instance, à unit vector pointing north at the eartl's equator, Is
not at all the same as a unit vector pointing north at the tropic ol Capricorn. This example should
help motivate the following definition.
length. Vectors as thus considered as independent of their loca:
location. Tor cx
1.4 Definition A tangent vector X, in Rº, is an ordered pair (X.p). We may regard X as an
ordinary advanced caleulus vector and p is the position vector of the foot the arrow.
tm these notes we will use the following index conventions.
Indices such as à, j,b,Lm, 1, vun from 1 to a
such as p,v, 7.0, rum from O tom
such as o, 0,7 8, run from | to 2.
CHAPTER 1. VECTORS AND CURVE
One may think of lhe parameter ! as representing time, and the curve « as representing the
trajectory ol a moving point particle.
1.10 Example Let
a(b) = (t+ by. ant + bo ast + ba).
This equation represents a straight line passing through the point p = (b1,ba. by), in the direction
of the vector v = (ay. a», a3).
1.11 Example Let
a(f) = (a cosut, asim ut, bt).
as the path des
wrapped around a circular cylinder of radius a.
This curve is called à ribed by
the hypothenuse of a triangle with slope b, which
The projection of the helix onto the ay-plane is a cirele and the curves risos at a constant rate in
the z-direction.
reular helix. Geometrically, wc may view the curve
Oecasionally we will revert to the position vector notation
(8) (18)
shieh is more prevalent in vector caleulus and clementary physics texthooks. Of course, what this
notation really means is
x(4) = (2"(8)
(8,
(8) = (2 o 0)(1), (1.9)
where x! are the coordinate slot functions in an open set in Rê.
1.12 Definition The derivative a/(t) of the curve is called the velocity vector and the
derivative a”(t) is
the speed of the curve. The components of the veloc
dx (del da? das
vg=-T-(E ma 1
= ( dt 7) (110)
de PO fd Do (dus
(o (D+) (111)
1 position vector given by
dal da? das
dx (E a a (1.12)
; called the acceleration. The length + = [Jo*(8)]| of the velocity vector is called
y vector are simply given by
and lhe speed is
The diferential of dx of the classic
a dO
is called an infinitesimal tangent vector, and lhe norm ||dx]| of the infinitesimal tangent vector
is called (he differential of arclenguh ds. Clearly we have
ds = |ldx|| = dt (1.13)
As we will see later in this text, the notion of infinitesimal objects needs to be treated in a more
rigorous malhematical setting. Al lhe same time, we must not discard the great intuitive value ol
this notion as envisioned by Lhe masters who invented of Calculus; even at Lhe risk of some possible
confusion! Thus, shercas in the more strict sensc of modern differential geometry, the velocit
vector is really a tangent vector and hence it should be viewed as a lincar derivation on the space
of functions, it is helpful to regard dx as a traditional vector srhich, at the infinitesimal level, gives
a lincar approximation to the curve.
12. CURVES INR?
1£ fis any smooth function on Rº, we formally define a/(8) in local coordinates by We formula
(0)
exit
= Eua (114)
“The modern nolation is more precise, since il Lakes into account that the velocity las a vector part
as well as point of application. Given a point on the curve, the velocily of the curve acting on a
function, yields the direcional derivative of that function in the direction tangential to the curve al
the point in question.
The diagram below provides a more geometrical interpretation ol the the velocity vector lor-
mula (1.14). Lhe map o(!) from R to R$ induces a map au from lhe tangent space of R to the
tangent space of Rê. The image o(£) in TRÊ of the tangent vector É is what we call a'(£)
dl
ali)= (8.
Since a!(t) à
function fon Rº
a!(t) to the coordinate Tunclions 4”, we gel the components of the the Langent vector
a langent vector in R$, it acts ou [unctions in Rê. Lhe action ol «(!) on a
is the same as the action of & on Uhe composition fo a. In particular, il we apply
as ilustrated
a as
GETROS TRÊS o'(t)
4 4
RREO R
(8) (e) lares
The map a, on the tangent spaces induced by the curve a is called the push-forward. Many
authors use the notation da: to denote the push-forward. but we prefer to avoid this notation because
most students fresh out of advanced caleulus have not yet been introduced to the interpretation of
(1.15)
Uhe differential as à linear isomorphism on tangent spaces.
1.13 Definition
TE = 4(3) is a smooth, real valued function and a(f) is a curve in R$, wo sa
3(5) = o(t(5) is a reparametrization of a
À common reparametrization of curve is obtained by using the arclength as the parameter. Using
this reparametrization is quite natural, that the rate of change of
Ue arclengih is seliat we call speed
g that the curve
nec we know from ba;
=“ .
v= Slot (1.16)
The are length is obtained by integrating the above formula
“= flotopa= f (5) (5 +) & (L17)
Tn practice it is typically difficult to actually find an explicit arclength parametrization of a
curve since not only does one have caleulate the integral, but also one needs to be able to find the
inverse function t in terms of s. On the other hand. from a theoretical poimt of view, arclength
parametrizations are ideal since any curve so parametrized. has unit speed. The proof of this fact i
a simple application of the chain rule and the inversc function theorem.
Bs) = [lts
6 CHAPTER 1. VECTORS AND CURV
= afeto)
1
a(s) —L—
ro)
O st)
Iron
and any vector divided by its length is a unit vector. Leibnitz notation makes this even more self
exident,
dx dxdi
de dtds
dx
- Cê
GE
1.14 Example Let (t) = (acoswt, asincot, dt). Then
VÍt) = (Cawsinwl, du cost, b),
a
“) = /
v
[sia du
bo
= ct, where, c> Vau b
(au sin au)? + (au coswr)? + B2 du
Yhe helix of unit speed is then given by
su,
(5) = (a cos — esmn —
(9) = (oc0s E, ain E
Frenet Frames
Lei 3(5) be à curve paramelrized by arc length and lei (8) be lhe vector
T(s) = 8'(8). (1.18)
Yhe vector P'(s) is langential to the curve and it has unit lenglh. Herealter, we will call £ We unit
Tangent vector, Dillerentiating We relation
Fr=l (1.19)
we gel
9TT'=0, (1.20)
so we conelude that the vector T” is orthogonal to T. Let NX be a unit vector orthogonal to T. and
let x be the scalar such that
TU) = aN(s). (131)
We call N the unit normal to the curve, and & thc curvature. Taking the length of both sides of
last equation, and recalling that NX has unit length, wc deduee that
e= Pe) (1.22)
1,2, CURVES
“Then,
(é)
Ive =
= r3(sin" t+ cos?!)
(=rsint,rcost.0)
sint)2 + (rcost)2 + 02
=".
Therefore ds/dt = r and s = rt, which we recognize as the formula for the length of am are of circle
of radius t, subtended by a central angle whosc measure is £ radians. We conclude that,
EQ
os E
reos 5,0)
his is à very simple but important example, “The fack that for à circle of radius v the curvature
e has large curvalure and à large circle has small
isa = 1/r could not be more intuitive, À small cir
curvalure. As Uhe radius of the circle approaches infinity, the circle locally looks more and more like
a straight line, and the curvature approaches to 0, 1f one were walking along a greal circle on à very
large sphere (like the carth) one would be perecive the space to be locally flat.
1.18 Proposition Let o(t) be a curve of velocity V, acecleration A, speed 4 and curvature «,
then
V= a
de 2x
A = Cpyçan. 1.29
qr + ts (1.29)
Proof Let s(t) be the arclengih and let 2(5) be à unit speed reparametrization. Then a(t) =
(s(4)) and by the chain rule
V = off)
= 5"((0)()
= “T
A = (f)
= GTA RT (st (0
= GTA (aja
= GTA tax
10 CHAPTER 1. VECTORS AND CURVE
Equation 1.29 is iuportant in physics. The equation states that a particle moving along a curve
in space feels a component of acceleration along the direction of motion whenever there is a change
ol speed, and à centripetal acceleration in lhe direction ol the normal whenever it changes direction.
The centripetal acceleration and any point is
e
7
where r às the radius of a circle which las imaximal tangential contact will the curve at te poiat
in question. “This langential circle is called the osculating circle. “Lhe osculating circle can be
envisioned by a limiting process similar to thai of the tangent 10 a curve in dillerential calculus.
Lei p be point ou the curve, and let q; aud q; two nearby points. [he three points determine a
cirele uniquely. This cirele is a “secant? approximation to the tangent circle. As the points qi and
q» approach the point p, the “secant” cirele approaches the osculating circle. The osculating circle
always lies in the the TN-plane. which by analogy, à
called the osculating plane.
1.19 Example (Telix)
Hs) = (acos E asin E here € EC
o
= (“sin
beta (o) =
no =
a
I
é atu!
Simplifying the last expression and substituting the vale of e, wc get
dh
,r ER
a = E
2 +52
ç
Notice tab if b = 0) the helis collapses to a circle in lhe
reduce to x = 1/a and 7 = 0. The ratio n/7 = au/b is particularly simple. Any curve where
«/7 = constant is called a helix, of wlich the circular Delix is a special case,
plane. lu this case the [ormulas above
1.20 Example (Plane curves) Let a(f) = (2(2),4(1),0). Them
ad = (8,0)
12. CURVES INR? 11
af = (AO)
O (PO)
lo! x 07]
E a
or
uyt—
1.21 Example (Comu Spiral) Let (5) = (2(4), 4(8),0), where
|
rj = | cosisd
) / Bo"
a $2
sj) = st tt. 1.3
v(s) / sin so é (1:30)
Then, using the fundamental thcorem of caleulus, wc have
the Comu spiral is given by
x
IH
The integrals (1.30) defining the coordinates of the Cormu spiral arc the classical Frenel Integrals.
These functions, as well as th al itself arise in the computation of the diffraction pattern of a
coherent beam of light by à straight edge.
In cuses where lhe given curve o(t) is not of unit speed, the following proposition provides
formulas to compute Lhe curvature and Lorsion in terms ol a
1,22 Proposition Il a(!) is a regular curve iu R$, then
vo axe!b
E > EA 13
' Ei (at)
om)
= 1.32
= Ps (1.32)
where (aaa!) is We triple vector produel [a x! 0 at”,
Proof:
q = 7
ofo= dP+eRa
= (NUS +.
= etuN'+
= werB+..
14
CHAPTER 1
VECTORS AND CURVES
Chapter 2
2.1 1-Forms
One of the most puzzling ideas in clementary caleulus is the idea of the differential. In the usual
definition, the differential of a dependent variable y = (2), is given im terms of the diffcrential of
the independent variable by dy = F'(z)de. The problem is with the quantity de. What does dz
mean? What is the difference between Ag and de? How much “smaller” than Az does de have
to be? “Lhere is no trivial resolution to his question. Most introductory calculus tests evade the
issue by treating de as am arbitravily small quantity (which lacks mathematical tigor) or by simply
referring to du as an infinitesimal (a Lerm introduced by Newton [or au idea that could not otherwise
be clearly defined at the time.)
Tn this section we introduce lincar algebraic tools that veill allow us to interpret the difierential
in terms of am lincar operator.
21 Definition Let p€ Rº, and let T,(Rº) be the tangent
map 6 from T(R?) into R. Wo recall that such a map mus
pace at p. A L-form at p is a lincar
'y the following properties
a) AX)ER, VXp ER” (2.1)
db) dlaMp +) = ad(Ap) + AM), Va,DER, Xp.%p E TUR!)
A 1-form is a smooth choice of a lincar map é as above for cach point in the space.
2.2 Definition Let f:Rº > R be a rcal-valucd € function. We define the diffcrential df of
the function as the 1-form such that
dx) =*(1) (2.2)
for every vector field in X in Rº.
Tn other words, at any point p. the differential df of a function is an operator which assigns to
à tangent vector X,, te directional derivative of the function in the direction of that vector
FO) = Nil) = Ho) -X(p) (2.3)
In particular, if we apply the dilferential of We coordinate Tunctions x Lo the basis vector fields,
we gel
(2.4)
The set of all lincar functionals on a vector space is called the dual of the vector space. Tt is
am standard thcorem in lincar algebra that the dual of a vector space is also à vector spacc of the
15
16 CHAPTYER 2. DIFFERENTIAL FORMS
same dimension. “Lhus, the space Z/RY of all L-lorms at p is a vector space which is the dual o
Lhe Langent space 7,Rº. “The space “1; =(Rº) is called the cotangent space of Rº' at the point p.
Equation (2.4) indicates Lhal the sel of diferential forms tda) pre (dar) "Jp + constitutes the basis
of the cotangemt space which is dual to the standard basis £ 1 ol lhe Langent space.
jr
s p rauges over all points ia Rº is called the cotangent bundle
'Yhe union of all the cotangent space
TR).
2.3 Proposition Let f be any smooth function in Rº and let (x1,...2”] be coordinate functions
in a neighborhood [ of a point p. Then. the differential df is given locally by the expression
df = Ee da (2.5)
i=l
BL
Ba
ble as a
Proof: The differential df
lincar combination of the basis clements f(dr!),...., (dz?) ). Therefore, to prove the proposition,
it sufficos to show that the expression 2.5 applied to am arbitrary tangent va with
defini To sec this, consider a tangomt vector X, = q (45) and apply the expression above
3 by definition a 1-form, so, at cach point, it must be expr
or, coincides
om 2 É)
(E (o e) (2.6)
(2)
(Ee (49)
The definition of differentials as lincar functionals on the space of vector fields is much more
sati
factory than the notion of infinitesimals, since the new definition is based on the rigorous
machinery of lincar algebra. Ta is am arbitrary 1-form, then locally
a =a(xdel+,... + au(x)de”, ç
where the cocfficients a; are CS functions. À 1-form is also called a covariant tensor of rank 1,
or just simply à covector. “Lhe coellicients (a1,..., 4) are called the covariant components ol
Lhe covector. We will adopt the convention to always write the covariant components of à covector
with the indices down. Physicisis often reler Lo Lhe covariant components of à I-form as à covariant
vector and this causes some contusion about le position of Lhe indices. We empliasize that not all
one forms are obtained by taking le dilerential of à [unclion. If there esisis à function f, such
that = df, then the one form «is called exact. In vector caleulus and elementary physics, exact
forms are important in understanding the path independence of line integrals of conservative vector
fields.
As we have already noted, the cotangent space TE(R”) of |-forms at a point p has a natural
vector space structure. Ne cam casily extend the operations of addition and scalar multiplication to
2.2. TENSORS AND FORMS OF HIGHER RANK 19
17 we now deline
bi = gb, (2.14)
we see that the equation above can be rewritten as
dby= gabi.
and we recover the expression for the inner product.
Equation (2.14) shows that the metric can be used as mechanism to lower indices, thus Lrans-
forming the contravariant components of a vector to covariant ones. Tf we let 9% be the inverse of
the matrix g;;. that is
ih si 5 E
sÉ 9 6; (2.15)
wc can also raise covariant indices by the equation
bi= qb; (2.16)
We have mentioned that the tangent and cotangent spaces of Tnclidean space at a particular poimt
are isomorphic. In vicw of the above discussion. we see that the metric accopts a dual interpretation:
one as bilincar pairing of two vectors
VR) TR) SR
and another as à linear isomorphism
STARS T(R")
Lab maps vectors Lo covectors and vice-versa.
mn clementary trcatments of caleulus authors often ignore the subtleties of differential I-forms
and tensor products and define the diffcrential of arclength as
ds? = gide'de',
although, what is really meant by such an expression is
aid o del. (2.17)
2.7 Example lu cylindrical coordinates, the dilTerential of arclengih is
ds? = dr + de 4 do? (2.18)
Tn this case the metric tensor has components
10 0
m=/0 7 01. (2.19)
001
2.8 Example Im spherical coordinatos
e = psinficosy
v = psinfsinê
= post, (2.20)
the diffcrential of arclengkh is given by
= do + de + pêsin” Odo”, (2.21)
la this case the metric tensor has components
100
sm=|0 20 . (2.22)
00 ping?
20 CHAPTYER 2. DIFFERENTIAL FORMS
Minkowski Space
An important object in mathematical physies is the so called Minkowski space sehich is can be
defined as the pair Let (M1,3,9) be the pair, where
Masg= tale, a)t ci eR)
and q is the bilinear map such ta
HI = LA ue PA (+ (ué (2.24)
Lhe matris representing Minkowski's mebric g is given by
9 =diag(=1,1,1,1),
ia which case, the dilferential of arclength is given by
ds = gude da”
= -dtodt+ del o dal 4 de cdr 4 dr? a de?
= de + (del 4 (da?)2 4 (day. (2.25)
Note: “Technically speaking, Minkowski's metrie is not really a metric since g(X,X) = 0 does
not imply that X = 0. Non-zero vectors with zero length are called Liglt-like vectors and they are
associated with with particles which travel at the speed of light (which sec have set equal to Lin our
tem of units.)
The Minkowski metric gaz and its matrix inverso q” are also uscd to raisc and loser indices in
the space in a manner completely analogous to Rº . Thus, for example, if À is a covariant vector
with components
Au= (poda,
then the contravariant components of À are
Wedge Products and n-Forms
2.9 Definition A map é: T(Rº) x Y(R7) — R is called alternatingil
AX) = (FX)
The alternaling property is reminiscent of determinants of square mabrices which change sign il
auy two column vectors are switched, In fact, Lhe determinant function is à perfect example ol au
alternating bilincar map on the space Max» of two by two matrices. Of course, for the definition
above to apply, onc has to view Max» as the space of column vectors.
2.10 Definition A 2-form Gisa map é: T(R?) x T(Rº) — R. which is alternating and
bilincar.
2.11 Definition Lei o and 3 be I-forms in Rº and let X and Y be any two vector fields, The
wedge product of the two Llorms is the map AS: TR”) x T(R") — R given by the equation
(en XI) = (MO) — (ICO)
2.2. TENSORS AND FORMS OF HIGHER RANK 21
2.12 Theorem If and 3 are I-lorms, then a A 2 is a 2-form,
Proof: : We break up the proof into the following two lemmas.
213 Lemma The wedge product of two 1-forms is alternating.
Proof: Let a and 3 be I-formsin Rº and let X and Y be any two vector ficlds. then
(CAST) = (080) = e0)0T)
—(0(F)800) — e(08())
(0 AB. X)
2.14 Lemma Lhe wedge product of ivo Llorms is bilinear,
Proof Consider I-forms, a, 2, vector fields X1,X2,Y and Tunciions [1,62 Then, since the
1-forms are linear funciionals, we gel
(en PM +
= fx + PAIS) aà + PP Xo)
E e(Ã) + P(A (O) — QL A) + P(A)
PANIC A P(A IO) + FAQ) + P(A)
PAIO + OBD] A Poa) 20) + 000 )H(X5)]
Fo A XI A Po A SIX», Y)
The proof of lincarity on the second slot is quite similar and it is left to the reader.
215 Corollary If q and 3 arc I-forms, then
ang=-9A0 (2.27)
last result tells us that secdge products have characteri similar to cross products of
vectors im the sense that both of these products are anti-commutative. This means that we need to
be careful to introduce à minus sign every time sec interchange the order ofthe operation. Thus, for
example, we have
de Adel = —de A de
ifi j, vhercas
de nda = dei nd =0
since any quantity which is equal Lo the negative of itself must vanislt “Lhe similarity between wedge
producis is even more suiking iu the next proposition but we emphasize again that wedge products
are by far much more powerful than cross products, because wedge products cam be computed in
any dimension.
2.16 Proposition Leia = Ade ud d = Bida! be any two Llorms in Rº. Then
ang=(ABde A de (2.28)
Proof: Let X and Y be arbitrary vector fields, [hem
(a A BUAS) (asd OB) — (Ad OB da (A)
(As By) [da (dar 7) — dr (dai (0)]
(As Bj) (dr A de (XY)
24 CHAPTYER 2. DIFFERENTIAL FORMS
2.22 Proposition
a) di NS A
b) d=dod=0
O) daeng)=dang+(-IJondo VacA den (2.34)
Proof:
a) Obvious from equation (2.32).
b) Tirst we prove the proposition for a = fe A. We have
d(da) =
de? A de”
of
7 aca de A de
that
Now, suppose that a is represented locally as in equation (2.32). Tt follows from
d(da) = d(dAs, iu) Ada A deli. der =0
c) Leta Ee A" de A”. Lhen we can write
e = As idade? A. dei
8 By jileldei A. dela,
(2.35)
By definition,
0n8= Ana Ba cp Ad A Adro) A (dei AoA dedo)
Now we take take the exterior derivative of the last equation taking into account that d(fy) =
Edy + gdf for any functions f and g. We get
ond) = (MA) Bida (Aa MB cs dae ncia dei) A (dad An dai)
= [As (da Ade AB, A (dra A dei] +
= Asa A (det A A de A (DAR A (dit AA deio)]
= dang+(-IJa nda. (2.36)
Lhe (= 1)" factor comes in because Lo pass lhe term dB, 5, Uurough p L-forms of the type de”, one
has to perform p transpositions.
2.23 Example Leto = Pe, gde + Qle,g)d8. Then,
JP o o
dos CE) ndo + (Ca 4 22) pay
dy de Oy
da =
oP o
= Pyndo + Cor a dy
dy E
o) àP
0 din dy.
gy
This example is related to Green's theorem in RZ.
2.4, THE HODGE-+* OPERATOR 25
2.24 Example Leio = M(e, y)de+ Ne, y)dy, and suppose that do = 0. Then, by the previous
example,
ON OM
—- E yde A dy.
de pg it A dy
that Nº
da
Thus, da = 0 if? My vehich impli F and Mo for some Cl function f(a
a=tdr+fydf= df
The reader should also be familiar with this example in the context of exact differential cquations
of first order, and conservative force fields.
2.25 Definition À dillerential form à is called exact il da = 0.
2.26 Definition À dillerential form à is called closed if there exist a [orm 2 such that a = dj.
Since dod=0, it is clear Uial a closed [orm is also exact, “The converse is not at all obvious in
general and we state it here srithout proof.
2.27 Poincare's Lemma Mm a simply connected space
it is closed.
The assumption hypothesis that the space must be simply connected is someehat subtle. The
condition is reminiscent ol Cauchy's integral theorem [or [unctions of a comples variable, which
uch as Rº ). if a differential is exact then
slutes that if f() is holomorphie function and C is à simple closed curve, then,
f fl)d:=0
er
This theorem does not hold if the region bounded by the curve C is not simply connected. The
standard example is the integral of the complex IHorm w = (L/:)dz around the unit circle €'
bounding a punctured disk. Jo this case,
2.4 The Hodge-* Operator
One of the important lessons that students learn in lincar algebra is that all vector space of finite
dimension n arc isomorphie to cach other. Thus, for instance. the space Py of all real polynomials
in 2 of degree 3, and the space Max» of real 2 by 2 matrices, arc basically no different than the
Euclidean vector space Rá in terms of their vector space properties. We have already encountered
a number of vector spaces of finite dimension in these notes. A good example of this is the tangent
space T;R$. The “vector” part af + a2Z + aºZ can be mapped to a regular advanced calentus
vector ali + a2j + aºk, by roplacing É by É byjand £ by kk. Of course, wo must not confuso
a tangent vo h a Tuclideam vector which is just an ordered triple
but as far their vector space properties. there is basically no difference.
We have also observed that the tangent space TR”
or which is a lincar operator w
isomorphie to the cotangent space TR”.
á
Im this case, the vector space isomorphism maps the standard bs Vito their duals
fdz'). This isomorphism then transforma a contravariant vector to a covariamt vector.
Another interesting example is provided by the spaces Ai (R$) and AÇ(R?), both of which have
dimension 3. Tt follows that th smorphic. Tn this casc the isomorphism i
given by the map
vecto
» two spaces must be
dr — dyAde
26 CHAPTYER 2. DIFFERENTIAL FORMS
dy > —dende
dz > dendy
(2.38)
More generally, we have seen that We dimension of the space o! m-forms in Rº is given by the
(1) n nt
m)=(n—-m =
(um)!
APR) = AR) (2:39)
Lo describe the isomorphism between Lhese two spaces, we will first need to introduce the totally
ymmetric Levi-Civita permutation symbol which is defined as follows
Ra RO
binomial coefliciem (7). Since
it must be the case that
+ if(i, oosim) à5 am even permutation of(1,...,m)
= ifi 1) is an odd permutation of(1,...,m) (2.40)
0 otherwise
In dimension 3, there are ouly 3 (3!=6) nonvanishing components of esp in
Cs =
fis =
=1 (241)
The permutation symbols are us
matrix, lhen, using equation (2.41), the reader can easily verify Uia
ful in the theory of determinants. Tn fac =(a;)isadx3
det 4 = |A| = eqisistfaias (2.42)
Vhis formula [or determinants extends in au obvious manner to 2 x à matrices. À more Lhorough
discussion of the Levi-Civita symbols will appear later in these notes.
lu Rº, the Levi-Civita symbol with some or all the indices up is numerically equal to the
permutation symbol will all indices down
since the Euclidean met
On the other hand, im Minkow:
==. Thus, in calMA 3)
because go = 4º
iniriais
Ciaivisia = —€
since any permutation of (0, 1, 2,8] must contain a O.
2.28 Definition “Lhe Hodge-+ operator is a linear map +: ALR) — AVR") defined in
standard local coordinates by the equation
1
a A A der et im deimt A A datos (2.43)
(em)! nto
“
Since the forms dxiz A. ..Adxi» constitute a basis of the vector space AV (R?) and the +-operator
umed to be a lincar map, equation (2.43) completely é
ceifies the map for all m-forms.
2.4, THE HODGE-+* OPERATOR 29
4, Lea = Bidx, en
ade =>V.B
Tt is also possible to define and manipulate formulas of classical vector caleulus using the per-
mutation symbols. For example, let a = (41.49.44) and B = (By, Bo. Bs) be any tro Euclidem
vectors. Then it is casy to sec that
(A x BJa = AB,
and =
GA;
VxBh=d :
( ) + Bud
To derive many classical vector identilies in this formalisim, il is necessary to first establish the
following identity (see Ex. ())
EM eutm = 6,6] — 6;6] (2.54)
232 Example
AX(BxCO = TA mBxCOm
Am (BC)
PE Am Bh)
emitir Ar B; Cs)
(816% — POA Bj
= BA"Ca— Gba
Or. reveriting in vector form
Ax(BxC)=B(A-C)-C(A.B) (2.55)
Maxwell Equations
“Lhe classical equations of Maxwell describing electromagnetic phenomena are
V.E=47p VxB=40]3 +
vV.B=0 vxE=-2É
õ
(2.56)
We would like to formulate these cquations
(4,21, 22,28) be local coordinates in Minkow:
in the language of differential forms. Let 2! =
pace Ms. Define the Maxwell 2-form P by the
equation
1
T= uodetAde”, (uv =0,1,23). (2.57)
wliere
—y —Ey
B. By 25
dm (2.58)
-B 0
Written in complete detail, Maxwell's 2-[orm is given by
FP = —Rodtadel- Editada? EdtA de +
Bedxl a dx? — Bydel a de 4 Bode? Ade”. (2.59)
30 CHAPTYER 2. DIFFERENTIAL FORMS
We also define Lhe source current I-form
1 = Jude” = pdt+ dude! + dada? + dade”. (2.60)
j are equivalent Lo the equations
2.33 Proposition Maxwell's Equations
dr = 0,
dep = dg4rxd, (2.61)
Proof: “Lhe proof is by direct computation using the definitions of the exterior derivaive and
the Hodge-* operalor.
a det dia de! +
de Ade? A dia de!
Ade! Adin de? — E A det adia dae? +
Ade! Adi A da? — Ade? n dia da? +
) 3
Coe ade dal A dal —
a
0B,
Cos nata del A das —
dt
ã
Ada A dela da? —
de! A del A de? +
9B
Ya A dia du? +
dz? JH
dE, 0By 1 3
Bdzê à Jdt nda! pa” +
v 0Ey OB 2
E geo gd! A dinda?
Therefore, dF = O iff
08, By
del À dez
which is the same as
v.B=0,
and
24, THE HODGE+ OPERATOR st
velicl means Uiat ,
-vxE-D=0. (2.62)
Vo verily Ue second set o! Maxwell equalions, we first compute the dual of the current density
I-form (2.60) using the results [rom example 2.4. We get
+] =-pdel ada da? + Ide A dta de? + Jodt A del A de? 4 Jadel A dt A de? (2.63)
Wo could now proceed to compute d+ P, but perhaps it is more clegant to notice that P E AM),
and so, according to example (2.4), E splits into E = E, + $.. In fact, we see from (2.58) that the
components of FP, are those ol —E and lhe components ol F- constitute Lhe maguetic field vector
B. Using the results of example (2.4), we can immediately write the components o[ +
aP = Badtndal + BydtA de 4 BodtA de +
Edu! nda? — Eyde! A de? 4 Ed? A do”, (2.64)
or equivalently,
0 Bo By By
Ms E (2.65)
“By E 0 E
-B ER 0
Since the ellect ol the dual operator amounts to exchanging
E -— -B
Bs +E,
we can infer from cquations (2.62) and (2.63) that
v.E=4mp
and,
sa CHAPTER 3. CONN
Tence
Given a framo vectors e;, we can also introduce the corrosponding dual coftame forms 6; by
requiving
eo 2
0(e;) = 0; (3.3)
since the dual coltame is a set ol Lforms, they can also be expressed in of local coordinates as
linear combinations
6 = Bida
Lt follows from equation( 3.3), that
Ole) = Bidr(0A!,)
= BiAtdr(à)
= BiAtõs
8, =
Lherefore we conclude that BA = £,so B = AT! = 47, In other words, wlen the frames are
orthonormal we have
e = WA
p= Adaf (3.4)
3.3 Example Consider te translormation [rom Cartesiau to cylindrical coordinates
2 = reos0.
v = sino,
2 =
Using the chain rule for partial derivatives, we have
à à à
— = cost sing
> cos dm + sin 2y
8 sô à
% = -ruiafço + rosto
9.
dr O
Trom these cquations we casily verify that the quantiti
RR
“Sm
o Vê
O = 0
ô
=
3.2, CURVILINEAR COORDINATES 35
are à lriplet of mutually orthogonal unit veclors and tus constitute an orthonormal frame.
3.4 Example For spherical coordinates( 2.20)
r = psinficosó
y = psinêsing
5 = pcoso,
the chain rule leads to
ô . Fe] . . Fe] Fe]
— = sinfcosê— +sinÊsind— + cos6-—
p de 3
Fe] 8 à + bai Fe] +
= posfcwsd—+pcosbsind D+
E PE ÉCOS Ea | PCS CA a
Fe] . . à . É
Lo= cpsnfsindl + psinbcosó
dó de
ln this case, the vectors
78
ro
psind da
tutc an orthonormal frame.
also const
The fact that the chain rule in the two situations above leads to orthonormal frame
of the level surfaces x = constant. Since
one expects the gradients of the surfaces
Lo also be orthogonal. Lransformatioas of Ui type are called triply orthogonal systems.
coincidental. The results are related to the orthogonalit;
z
the level surfaces arc orthogonal whenever they intersect
3.2 Curvilinear Coordinates
Orthogonal transformations such as Spherical and cylindrical coordinates appear ubiquitously in
mathematical physics because Lhe geometry of a large umber of problems in Uhis area exhibit sym-
metry weith respect to am axis or to the origin. Tn such situations, transformation to the appropriate
tem often result in considerable simplification of the ficld couations involved in the
problem. Tt has been shown that the Laplace operator sehich enters into all three of main classi
ficlds, the potential, the heat and the wave equations, is separable in 12 coordinate systems. A sim-
ple and cfficient method to calenlate the Laplacian in orthogonal coordinates can be implemented
by the use of differential forms.
coordinate
3.5 Example lu spherical coordinates the differential of arc lenguh is given by (see equation 2.21)
ds? = dp? + de 4 pê sin” Gdg”,
Let
6º = do
o pd
08 = psintdo. (3.6)
30 CHAPTER 3. CONN
CIO,
Note thai these three Llorms constitute the dual coframe to the orthonormal [tame which just
derived in equation( 3.5). Consider à scalar lield / = /(9,0,6). We now calculate the Laplacian
ol f in spherical coordinates using lhe methods ol section 2.4.
dilferential df and express the result in terms of the colrame,
af 91 of
To do this, we [irst compute the
d= pdpt Dqdo+ ção
= Cl dll, O Cl
dp "pool psmide
The components df in the coftame represent the gradient in spherical coordinates. Continuing with
the scheme of section 2.4, we first apply the Todge-s operator. Then we rewrite the resulting 2-form
in terms of wedges of coordinate differentials so that sc can apply the definition of the exterior
derivative.
)
df = - puorr
Pe
2
- Ponllanasop EB oeo A do
1... õ
= cairão Bout no
of 2 of
dadf = Dot smo Lap ndo ndo — E qoinaço Lda dp A da + E Jde A dp A do
. of 61) 4
= l|sin6>—(p? Sl donde nd
E melo + gpfsno E Ts PGE o
Finally, rewriling lhe differentials back in terms of lhe the coframe, we gel
Dm
Temo [lg dp E
| ,
sin é
The derivalion of lhe expression [or the spherical Laplacian through the use of differential forms
is elegant and leads naturally to the operator in Sturm Liouville form.
The process above can also be carried out [or general orthogonal translormations. À change of
coordinates 4º = 4 (uº) leads to an orthogonal trans[ormalion il in the new coordinale system vu”,
the line metrie
dad =
3 5
sin E) + E 4] BABA.
So, lhe Laplacian of f is given by
NO [,0f
ap 10 [oDf
Vi=20 |” 5)
of, Ff
Oo) + 3a
de = qu(du + gold) 4 gusldus)* (3.8)
ouly has diagonal entries. lu this case, we choose the colrame
1 = Vgudul = dal
8º = Vgade = hudo?
5 = Vgsadu = hadu?
ecally called the wcights. Please
+ for one and this will cause small
The quantities (Ay, fa. hs) are cl note that in the interest of
comnecting to classical terminology wc have exchanged two índio
discrepancies with the index sunimation convention. We will revert to using a summation symbol
COVARIANT DERIVADIVE 39
ln Euclidean space, the components of the standard [rame vectors are constant, aud thus their
rates of change in auy direction vanish, Let e; be arbitrary [rame field with dual forms 8º. “The
covarianl derivaives o[ te Irame vectors in lhe directions of a vector X, will in general yield new
vectors. The new vectors must be linear combinations of the the basis vectors
Two = u(Ne +wi(N)es Hui (Nes
Vrxto = wdA)e +uN)es +wi(X)es
Vres — w(A)e +uwdlX)es +wb(A es (3.14)
“The coeflicients can be more succincily expressed using the compact index notation
Vixe = ejui(X) (3.15)
Tt follows immediately that
(NX) = (VT xe). (3.16)
Equivalenty, one can take the inner product of both sides of equation (3.15) with ep Lo gel
<TrxeaD = <ewi(A)e>
= wU(X) <ere >
= (X)gj
Tenee.
< Fxenen >= wpi(N) (3.17)
The left hand side of the last equation is the inner product of two vectors, so the expre
represents an array of functions. Consequently, the right hand side also represents an array of
functions. In addition, both expressions are linear on X, since by definition Vx is linear on NX. We
conclude that the right land side can be interpreted as a matrix in which, each entry is a I-forims
acting on the vector X to yield a function. Lhe matrix valued quantity w”, is called the conncetion
form .
3.10 Let Tx be a Koszul connection and let fes) be a frame. The Christoffel
ymbols associated with the connection in the given frame are the functions T£, given by
Tec; = Then (3.18)
The
the fi
Lo the Christoltel symbols as the connection once again giving rise to possíble confusion. The precise
relation between the Cliristollel symbols and lhe connection I-forms is captured by the equalions
“hristofrel
ame vector:
mbols are the cocfficients which give the representation of the rate of change of
in the diveo
jon of the frame vectors themselves. Many physicists therefore refer
(3.19)
or equivalently
=14,6 (3.20)
Tn a general frame in R? there arc n2 entries in the connection 1-form and n2 Chris
The number of independent componen
toffel symbols.
is reduced if one assumes that the frame is orthonormal.
3.11 Proposition Let and e; be an orthonormal frame and Vx be a Koszul connection compatible
with the metric . Then
jiu; (3.21)
40 CHAPTER
Proof: Since il is given Lat < ee; >= 6, we lave
0 = Va<ee>
<Tyee>+<eTre>
<utee>+< cute >
= wÊ<ept D+ <em>
= wiga + wligia
= wo
thus proving that w is indeed anti
3.12 Corollary The Christoffel symbols of a Koszul connection in an orthonormal frame are
aulisynumetric on the lower indices; that is
(3.22)
We conclude that in an orthonormal frame in R the number of independent coefficients of the
connection Iform is (L/2)n(n — 1) since by anlisymmetry, the diagonal entries are zero, and one
only needs lo count (he number of entries in We upper triangular part of he nx n matrix vo;
Similarly, Ule number of independent Christollel symbols gets reduced to (1/2)n?(n— 1). In We case
of au orthonormal frame in Rº, where g, is diagonal, w”, is also antisymmetric, so the connection
5
equations become
fa 0 si (O) vl(N) ei
Tx lc [=| =ub(X) wi (N) e» |. (3.23)
es = UN) 0 es
Comparing the Trenct frame equation( 1.27), ec notice the obvious similarity to the general frame
cquations above. Clearly, the Trenet frame is a special case im vhich the basis
vectors have been
adapted to a curve resulúng iu a simpler connection in which some of the coeflicients vanish. A
further simplification occurs in the Frenet frame since here Lhe equalions represent the rate of change
of the frame only along (he direction of the curve rather than an arbitrary direction vector X.
3.4 Cartan Equations
Perhaps, the most important contribution to the development of Differential Gicometry is the work
of Cartan culminating into funous equatioas of suructure which we discuss in this chapter.
First Structure Equation
3.13 Theorem Let (c;] be a frame with connection w É and dual coframe 6º. Then
dE und =0 (3.24)
Proof; Lei
be a frame, and let 6º be the corresponding coframe. Since 0 (c;). we have
— -1 pi
(= (AY dad,
3.4, CARTAN EQUATIONS
Let X be au arbitrary vector field, Then
al
Tra = Tr(ó4,)
(O) = (4)
= dat WD(X)
= MATA (A)
(A) (ATA
Tonee,
af = (ATYA(A!).
or in matrix notation
wu = ATÍgA (3.25)
On the other hand, taking the exterior derivative of 9º. sec find
dé = dAyjAda
= dATY;A Ao
de = dATANO.
But, since ATIA = 7, we have (ATA = —ATÍdA = —w, hence
d=—w AB. (3.26)
Tn other words
dê wi AR =
Second Structure Equation
Let 8º be a colrame in Rº with connection w*,. Taking the exterior derivative ol the first equation
ol structure and recalling the properties (2.34) we gel
ala) + dll; A 0) =0
di; A Bi win doi =0,
Substituting recursively from the first equation of structure, we get
dan win (ow Ad) =0
dan pule Nu AGi=0
(dot; + wi nula =0
i É e
di, udp Aus =0,
We now introduce the following
3.14 Definition The curvature Q of a connection w is the matrix valued 2-form
di gi Á & 2 9=
DÊ, = dai, ot Noob, (2.27)
Thcorem Let 8 be a colrame with connection « in Rº . Lhe lhe curvature form vanishes
(3.28)
3.15
O=dutwunw=o
44 CHAPTYER 4, PHEORY OF SUREAC
4.3 Definition A dilferentiably smooth surface in R$ is a set of points M iu Rº such tlrat
1. Tfp E M then p belongs to some CX chart.
2. 4 p E M belongs to two dillerent charts x and y, Wen lhe two charts are CS equivalent.
Intuitively, wc may think of a surface as consisting locally of number of patches 2 sewn? to cach other
so as to form a quilt from a global perspec
The first condition in the definition
the second differentiabi
fatos that cach local pateh looks like a piece of RZ, whercas
joined together smoothly. An-
other way to state this idea is to say that a surface a space that is locally Euclideam and it has a
condition indicates that the patches am
differentiable structure so that the notion of differentiation makes sense. Tf the Euclidean space is
of dimension n, the “surface
is called au r-dimensional manifold
4.4 Example Consider the local coordinate clarl
x(m0) =
in te cos 2, sin sin 2, cos +).
tor equation is equivalent to three scalar functions in two variables
e = sinucose,
y = sinusino,
2 = cos (4.5)
+! 42º =, Uhe chart can
ibly represent the whole sphere because, although a sphere is locally Euclidean, (Uhe earth
Clearly, the surface represented by this chart is part of the sphere
not po:
is locally lat) there is certainly à topological dillerence between a sphere and a plane, Indeed, if
one analyzes the coordinate chart carefully you will note that at the North pole (4 =0. 2= |. the
coordinat 4 = 0 regardles
e number of labels. The fact that it às required to
ation of the 3-dimensional
becom of the
ngular. This happens because 4 = 0 implies that à
value of, so that the North pole has am infin
have two parameters to des
nature of of the surfacs
ibe a patch on a surface in R$ is a manife
constant while varying the other, then
the resulting I-parameter equations describe a curve on the surlace. Thus, [or example, letting
“= constant in equation (4.
s. Fone holds onc of the parameter:
) we gel the equalion of a meridian great circle,
4.5 Notation Given a parametrizatiou of a surface in a local clart x(u,0) = x(u!,u?)
(ue),
we will denote the partial derivatives by any of We following notations:
dx
Ku =X =. Xuu =X
du
Ox
MM = Ka =X =
dy
Ox
Xy = a Kd EE
Que Bud
4.2 The First Fundamental Form
Let 2i(uº) be a local parametrization of a surface. Then, the Tuelidean inner produet in R$ induces
an inner product in the space of tangent vectors at cach point in the surface. This metric on the
surface is obtained as follow
4.2. THE FIRST PUNDAMENTAL FORM 45
2
ds ddr da
dx! Oui
Jus Bu dudu”,
Thus,
where
a natural metrie
We conclude that the surface, by virtuc of being embedded in R$. inheri
sehich awe weill call the induced metric. À pair (44,9). sehero M is amanifold and 9 = gu ade sdu?
is a metrie is called a Riemannian manifold if considered as an entity im itsclf, and Ricmannian
submanifold of Rº if viewed as an object embedded in Tinclidean space. An equivalent version of
the metrie (1.6) can be obtained by using a more traditional caleulus notation
dx Xu du + xudu
ds? = dx-dx
= (xudu + xude) -(xudu + xede)
(x Xu du? 4 Mx Ra )dude(x, ed?
We cam rewrite the last result as
= Edu? + 2Fdudo + Gr”, (4.8)
where
E = =X Xu
Fo = ga=XKuXu
= da = Ko Xu
= g22=K Xu.
“That is
Gug = Xu Xg =€ XX >
4.6 bdf The clement of arclength
ds” = gasdu” o du? (4.9)
is also called the first fundamental form. We must caution the reader that this quantity is not
a form in the sense of diffcrential gcometry since ds? involvos the symimetric tensor product rather
than the wedge product.
The first fundamental form pla: that will find it
convenient to introduce yet a third more modem version. Tollowing the same development as in the
theory of curvos, consider a surface M defined locally by a function (ul, 12) — a(ul, 42). We say
that a quantity X, is a tangent vector at a pointp e M.if X, is a líncar derivation on the space of
€ real-valued functions [/|/: AM — RJ on the surf: The set ol all tangent vectors at a point
p E M is called the tangent space 1,M. As before, a vector field X on the surface is à smooth
such a crucial role in the theory of surfaces
choice ol a tangent vector at eacl point on lhe surface and the union of all tangent spaces is called
the tangent bundle LM.
Lhe coordinate chart map
a:RÉ— MER?
46 CHAPTYER 4. VHEORY OF SUREAG
induces à push-forward map
aiIRS UM
defined by
0) Inquey= Va o 7)
Just as in the case of curvos, when we revert back to cla
xi (4º). sehat we really mean is (r/c 0) (4º), where 2!
examples of tangent vectors on M are given by the pusl-forward of the standard basis of TR?
us
ical notation to describe a surface as
are the coordinate functions in Rê. Particular
These tangent vectors which earlier we called xo are defined by
à à
NO luis az
(ao)
“fe
us
Tn this formalism, the first fundamental form 7 is just the symmetrie bilincar tensor defined by
induced metric
MAES UN,V)=<AYS, (4.10)
where X and Y are any pair of vector ficlds in TM.
Orthogonal Parametric Curves
Lei Y and W be vectors tangent to a surface AL defined locally by a chart x(uº). Since the vectors
Xalpha Spam the tangent space of M at cach point, the vectors V and F7 can be written as lincar
combinations
Voxa
Wo = Wºxa.
The functions Vº and Fº are called the curvilincar coordinates of the vectors. We cam caleulate
the length and the inner product of the vectors using the induced Riemannian metrie,
[HIP = <yyD>=< Vez vêxa >= CV? Coxa xs >
VIP = gastov?
IP = gesWowo,
and
<vW> = <VixaWêx >= VW? < xx >
= ganvtWo,
The angle 8 subtended by the the vectors « and His the given by the equation
<LW>
AA
cos) =
(4.11)
Let uº = 6º(t) and 4º = “S(t) be two curves on the surface. Then the total differentials
dy
»
dt
tó .
due = CC gr amd duto
a
THE SECOND FUNDAMENTAL FORM 49
Tigure 4.2: Normal Curvature
by the constrainl ol the curve to lie on a surface. “The geodesic curvature vector , measures Lhe
“sideward” component ol the curvature in the tangent plane to the surface. Thus, il one draws a
straight line on a flat piece of paper and then smoolhly bend the paper into à surface, ten the
traight line would now acquire some curvature. Since the line vas originally straight, there is no
sideward component of curvature so A, = Qin this case. This means that the entire contribution
to the curvature comes from the normal component, reficeting the fact that the only reason there is
eurvature here is due to the bend im the surface itself.
s a point p € M and a direction vector X, E TM, one can geometrically
ss of all unit speed curvos in M
vlúch contain the poiat p and whose tangent vectors line up with the direction of X. Of course,
but al au infinitesimal level, all Lhese curves can be obtained
plane containing the vector X and the normal to M.
Similarly. if one specifi
envision the normal curvature by considering the equivalence cla
Were are infinitely many such curve
by intersecting (he surface will a ” verti
AI curves in Uhis equivalence class lave lhe same normal curvature and their geodesic curvatures
vamish. 1u Uis sense, the normal curvature is more of à property pertuining to a direction on the
surface at a point, whercas, the geodesic enrvature really depends on the curve itself. Tt might be
impossible for a hiker vealking on the ondulating hills of the Ozarkts to find a straight line trail, since
the rolling hills of the terrain extend in all directions. TIowever, it might be possible to walk on a
path with zero geodesie curvature as long as the hiker can maintain the samc compass direction.
To find an explicit formula for the normal eurvature wc first differentiate equation (4.15)
dt
ds
d du
= ea
r =
)
dus
“da
x)
duê
du” du?
*escgs ds
Taking the inner product of the last equation with the normal and noticing that < x,,n >= 0, we
get
au du” du?
Ka <f,uD>=<X,0>—>———
ds ds
basdutdu?
+“ 417
Goadurdu? (417)
sehere
bu =< Xu. > (4.18)
4.9 Definition “Lhe expression
HH = bogdutS du” (4.19)
is called the second fundamental form .
50 CHAPTYER 4. VPHEORY OF SURA
4.10 Proposition “The second fundamental form is syummetric.
Proof: In the classical formulation of the second fundamental form the proof is lrivial. We have
dos = ba since for a (º patch x(u”), we have xag = X5a because the partial derivalives conmute,
We will denote the coeflicients of lhe second fundamental form by
= bn =<X%en>
f = bo=<xen>
= by=<myn>
9 = by=<xn>
+ so that equation (4.19) can be written as
TI = edu? + 9fdudo + gor? (4.20)
and equation (4.17) as
CTT Pdul+2Fdudo+
MOTO Ceni +afdado + gd
We would also like to point out that just as the first fundamental form can be represented as
1=< dx dx >,
so cam we represent We second fundamental form as
H=-<dedu>
To see this it suffices to note that differentiation of the identity < x. >= O implies that
<Xyg.n >> — <XDy >.
Yherefore,
<dxdn> = <xodu, ndo” >
< xadut,ngdu” >
< xong > dudu”
—<xgn > dutd?
=IT
From a computational point a view, a more useful formula for the coelficients of We second
fundamental formula cam be derived by first applying We classical vector identity
4 JACA. 1
(4x 8) (C «D=| BG 55] (4.22)
to compute
xs xxo PO = (uu xx) (ru x)
Xu Xu Xu Xu
des | Eu Xu Xu Xu
Xu Xu Xu Ko
= EG-Fê
Consequentiy, the normal vector can be written as
Xu X Xy Xu X Xy
n=> DD =
“xxx VECSF?
4.4 CURVAPURE
“Thus, we can wrile We coeficients bos direclly as Lriple products involving derivatives of (
expressions for Uese coelficients are
(XX Xau)
(ru xu ue )
(x xuxa)
s =
4 VEG— +
Jhe first fundamental form on à surface measures the (square) of the distance between tivo
infinitesimally separated points, There is à similar interpretation of We second fundamental form
as we show below. “The second fundamental form measures the distance from a point on the surface
Lo the tangent plane at à second infinitesimally separated point. To see Uis simple geometrical
interpretation, consider a point xy = x(ug) E M and a nearby point x(us + dus). Txpanding on a
Taylor serics, sec get
x(ui + du") = xp + (xpJadu” + Jogdutdu 4...
We recall that the distance formula [tom a point x to a plane which contains xq is just Lhe scalar
projection of (x — xy) onto the normal. Since the normal to the plane at xy is the same as the unit,
normal to the surface and < x,,n >= 0, we find that the distance D is
D = <x-xn>
1 ;
=3< (xoJas nm > dutdu?
1
Sta
sia
he first fundamental form (or rather, its determinant) also appears in caleulus in the context
of caleulating Ve area of a parametrized surface. the reason is that if one considers an infinitesimal
parallelogram subtended by the vectors xudu ad x,do, then the dillerential of surface area is given
by the length of the cross product of these two infinitesimal tangent vectors. That is
ds
bre x xe] dede
ff VEG- P? dude
information about the shape of the surface at a point.
Tor example, the discus basl=e9— $? > 0 then all the neighboring
points lic on the same side of the tangent plane, and hence, the surface is concave im one direction.
Hat a point on a surface é > 0, the poimt is called am elliptic poimt, if 6 < O. the point is called
hyperbolic or a saddle point, and if 6 = 0, the point is called parabolic.
The second fundamental form contain
ion above indicates that if b
4.4 Curvature
Curvature and all related questions which surround curvature, constitute the central object of study
1 geometry. One would like to be able to answer questions such as, what quantilies
remain invariant as one surface is smoothly changed into another? There is certainly something
in differenti
intrinsically different from a cone, which wc can construct from a flat piece of paper and a sphere
vehich we can not. What is it that makes these two surfa
the shortest path between tsvo objects sehen the path is constrained to be om a surfac
o different? ow does one caleulate
?
54 CHAPTYER 4. VPHEORY OF SURA
4.18 Thcorem “Lhe Gauss map is à self adjoint operator on EM.
Proof: We have already shown that L: TM — TM is a lincar map. Recall that am operator
L on a linear space is sell adjoint il < LX,Y >=< X,LY >, so that the theorem is equivalent to
proving Uat Wal the second fundamental for is symmetric (LLX, Y] = H[Y, X). Computing the
diflerence of lhese two quantities, we gel
MXOT-H[Y,X] = <LXYD>-<LYX>
<VyNYD>-<FyrN,
X>.
Since < X,N >=< Y,N >= 0, and lhe connection is compatible with We metric, we know that
<FxNV> 0 NNxY>
<FENXD — -<NSçX>,
hence,
XY] TI(y.X] <NVyXD-<NVxy >,
<NVEX -VxY >
< NX, Y]>
=0 (UT [XY] E (MO)
One of the most important topies in an introductory course linear algebra deals with the spectrum
of self adjoint operators. “The main result in tis area states Uial if one considers the eigenvalue
equation
EX = 8X (4.30)
then the cigenvalues arc always real and cigenvectors corresponding to different cigenvalues arc or
thogonal. In the current situation, the vector spas
in question are the tangent spaces at cach point
of a surface in R$, so the dimension is 2. Tlence. we expeet two cigenvalues and two cigenvectors
LX = mN (4.31)
LX» = NX. (4.32)
4.19 Definition The cigenvalues x, and xy of the Gauss map £ arc called the principal cur-
vatures and the cigenvectors Xy and Xy are called the principal di
uations may occur depending on the classification of the cigenvalues at cach
point p on the surface:
tions.
Several possible
1. Tf a, É xo and both cigenvalues are positive, then p is called an clliptic poimt
2. Al kyio < 0, then p is called à hyperbolic point.
3. Tay = 40 £ 0, then p is called an umbilic point.
4. if 81x = 0. then p is called a parabolie point
Tt is also well known from lincar algebra, that the the determinant and the trace of a self adjoint.
operator are the only invariants under a adjoint (simila invariants
ansformation. Clearly thes
are important in the case of the operator L.. and they deserve to be given special names.
4.20 Definition The determinant A = det(7) is called the Gaussian curvature of M and
HF = (1/2)Tr(£) is callod the mean curvature.
4.4 CURVAPURE
Since any self-adjoint operator is diagonalizable and in a diagonal bas
Lis diag(ur, 12), E follows inmediately Uiab
, lhe matrix representing
Ko = ans
= dl tr) (4.33)
4.21 Proposition Lei X and Y be any linearly independent vectors in 1(M). Then
LXxXLY = EX xY)
(EXXDA(X x 17) = 2X xy) (4.34)
Proof: Since EX,LY € T(M). they can be expresses as lincar combinations of the basis vectors
X and Y,
EX = 0X+hY
LO = aX+Hboy.
computing the cross product, sec get
ixxty = [O blxxy
aa by
= deMIX xY).
Similary
EXXIDAM XE) = (ntbX XY)
= IUL(X x 7)
(HA x 4).
4.22 Proposition
K =
(4.35)
Proof: Starting with equations (4.34) take the dot product of both sides with X x Y and use the
vector identity (1.22). We immediately ger
| <LX.X> <EXY > |
K=
<INX> <EX,X>
<XND> <XY>
<YX> <yr>
<LX> o <FY> <IVX> <IrY>
| <XX> <XY> |
[RS SP | | SNÃS <A Y> |
2H =
<YÃ> <YY>
56 CHAPTYER 4, VHEORY OF SURFACES
Yhe result follows by taking X = xu and Y = xy
4.23 Thcorem (Euler) Let X; and X be unit eigenvectors of L andei X = (cos) X1+(sin 6) X.
Yhen
1HA,X) = nicos?6 + nosin” 6
Proof: Easy. Just compute LH(X,X) =< LX,X >, using the fact the LXG = 11X, LXo = 9X,
aud noting that the eigenveclors are orthogonal. We gel,
<LXXND> = <(cos0)m Xi + (sinO)aa Ko», (cos0)X, + (sin 0) Xa >
ricos? 0 < XX > +rgsin? 6 < No, Xa >
2a.
= «cos? 0 + nysin