S.P.

Novikov - MATH 740

Spring 2014

S. Novikov - office 3112 (MATH)

e-mails: [email protected] , [email protected]
office hours: MONDAY, WEDNESDAY, FRIDAY 16-17 pm

Contents
Lecture 1. Introduction: Textbooks and General Remarks.
Local coordinates: What is Cartesian system of coordinates.
Examples. 5

Lecture 2. Manifolds and Atlases. 7

Lecture 3. C ∞ -manifolds, Atlases, Charts. Especially good

Atlases. Implicit functions and Inversion. Imbedding of
Compact Manifolds in RN . Immersions. 11

Lecture 4. Manifolds and Submanifolds: Implicit functions

and Inversion. Vectors and Covectors. 17

Homework 1. 21

Lecture 5. Manifolds and vectors fields. Important Exam-

ples. 22

Lecture 6. Group Manifolds. Lie Algebra. Important Ex-

amples. 25

Lecture 7. Group Manifolds: Compact Lie groups. Most im-

portant Examples. Noncompact Lie groups. Most important
Examples. Lie Algebras. Gradient-like Systems. 29

1
Homework 2. 34

Lecture 8. Riemannian, Pseudo-Riemannian and Symplec-

tic Geometries.Complex Geometry. Restriction of Metric to
submanifolds. Length of curves and Fermát Principle. 34

Lecture 9. Geodesics and Calculus of variations. Length and

Action functionals. Examples. 39

Lecture 10. Variational problem and geodesics on Rieman-

nian manifolds: Action Functional, Lagrangian, Energy, Mo-
mentum. Conservation of Energy and Momentum. 44

Homework 3. 49

Lecture 11. Geodesics and Action Functional. Examples.

Hamiltonian form of Euler - Lagrange equation. Conserva-
tion of Energy and Momentum. Euclidean, Spherical and
Hyperbolic Geometries. 50

Lecture 12. Curvature of curves and surfaces. How to dif-

ferentiate tangent vector fields? 54

Lecture 13. Vector bundles. Connection and Curvature.

Parallel transport. 60

Homework 4. 63

Lecture 14. Vector bundles. Connection and Curvature.

Parallel transport. 65

Lecture 15. Connections in tangent bundle. Curvature.

Ricci curvature. Einstein equation. Spaces of constant cur-
vature. 68
Appendix to Lecture 15. . . . . . . . . . . . . . . . . . . . . . 71

Lecture 16. Tensor fields. Curvature as a tensor field. Gaus-

sian curvature for surfaces in R3 . 71

Homework 5. 75

2
Homeworks 2, 3, 4. Solutions. 76
Homework 2. Solutions. . . . . . . . . . . . . . . . . . . . . . 76
Homework 3. Solutions. . . . . . . . . . . . . . . . . . . . . . 78
Homework 4. Solutions. . . . . . . . . . . . . . . . . . . . . . 80

Lecture 17. Differential forms. 81

Homework 5. Solutions. 84

Lecture 18. Differential forms. De Rham operator. 85

Lecture 19. Differential forms. Cohomology. 90

Homework 6. 94

Lecture 20. Differential forms and tensors. Categorial prop-

erties. Cohomology and Stocks formula. Homotopy invari-
ance of Cohomology. 95

Lecture 21. Homotopy invariance of Cohomology. Examples

of differential forms. Symplectic and Kähler manifolds. 99

Lecture 22. Symplectic Manifolds and Hamiltonian Systems.

Poisson brackets. Preservation of Symplectic form by Hamil-
tonian System. 103

Homework 7. 106

Lecture 23. Volume element in Compact Lie groups. Av-

eraging of differential forms and Riemannian Metric. Coho-
mology of Homogeneous Spaces. 107

Lecture 24. Volume element in Compact Lie groups. Av-

eraging of differential forms and Riemannian Metric. Coho-
mology of Homogeneous Spaces. 111

Lecture 25. Approximation and Transversality. 113

Homework 8. 117

3
Homework 6 and Homework 7. Solutions. 118
Homework 6. Solutions. . . . . . . . . . . . . . . . . . . . . . 118
Homework 7. Solutions. . . . . . . . . . . . . . . . . . . . . . 119

Lecture 26. Transversality. Imbeddings and Immersions of

Manifolds in Euclidean Spaces. 122

Lecture 27. Intersection Index and Degree of Map. 124

Lecture 28. Intersection Index and Degree of Map. 128

Homework 9. 131

Lecture 29. Intersection Index and Degree of Map. 131

Lecture 30. Two applications of differential forms: Degree

of Map and Hopf Invariant. 136

Lecture 31. Comparison of notations of our Lectures with

book of Do Carmo “Riemannian Geometry”. 140

Homework 10. 144

Homeworks 8, 9. Solutions. 145

Homework 8. Solutions. . . . . . . . . . . . . . . . . . . . . . 145
Homework 9. Solutions. . . . . . . . . . . . . . . . . . . . . . 148

Lecture 32. Geodesics. Gauss Lemma. 150

Lecture 33. Local minimality of geodesics. 154

Homework 11. 156

Lecture 34. Riemannian Curvature. Examples. 157

Homeworks 10, 11. Solutions. 160

Homework 10. Solutions. . . . . . . . . . . . . . . . . . . . . . 160
Homework 11. Solutions. . . . . . . . . . . . . . . . . . . . . . 162

4
 Lecture 1. Introduction: Textbooks and Gen-
eral Remarks. Local coordinates: What is
Cartesian system of coordinates. Examples.
Textbooks:
Manfredo P. do Carmo. Riemannian Geometry.
Victor Guillemin, Alan Pollack. Differential Topology.

Additional Literature:
J. Milnor. Morse Theory.
S.P. Novikov, I.A. Taimanov. Modern Geometric Structures And Fields.
B.A. Dubrovin, A.T. Fomenko, S.P. Novikov. Modern Geometry - Methods
and Applications: Parts I, II.

Differential Manifolds, definition, maps, submanifolds.

Language of general topology is necessary: spaces, continuous maps,
homeomorphisms, compactness, metric and Hausdorff spaces.
Basic Tools from multivariable calculus: Implicit Functions, Approxi-
mations and Transversality. Theorems will be stated (but without proof).
Knowledge of Linear Algebra is necessary.
In Riemannian Geometry some theorems from ODE courses will
be needed.
Our theory is C ∞ : - manifolds, maps, . . . . Why?
The “physical” metric in the 4-space-time is NOT RIEMANNIAN.
Why do we need Riemannian metrics?
Concerning Differential Topology: Why do we need Approximations and
Transversality?
What is MANIFOLD?
Definition: manifold is a Hausdorff (or metric) space locally homeomor-
phic to an open domain in Rn .
Rn is an n-manifold.
Open domain U ⊂ Rn is a manifold.
What is a coordinate system?
a) Every coordinate is a continuous function on the space X

f : X →R

5
 b) Collection of functions f1 , . . . , fn

fj : X → R

gives a coordinate system if for every point x ∈ X we have

f⃗(x) = f⃗(y) → x = y

where f⃗(x) = (f1 (x), . . . , fn (x)).

Map x → f⃗(x) gives homeomorphism of X into some open domain U ⊂
R .
n

Examples.
1) X = R2 , f1 = x, f√ 2 = y.
2) X = R , f1 = ρ = x2 + y 2 , f2 = φ.
2
ρ
a) ρ is not a coordinate in R2 ϕ
but ρ is a coordinate in R2 \0 = X ′
b) φ is not a coordinate in R2 \0 because φ is not a function, it is
multivalued.
“Cartesian Coordinates” in Rn

Rn ↔ (x1 , . . . , xn )

points → one-to-one with n-tuples (x1 , . . . , xn ).

“Cartesian Coordinates” in open set U ⊂ Rn .
Manifold = Metric space M n locally homeomorphic to open domains in
R ↔ for every point x ∈ M n there exists an open set x ∈ U ⊂ M n such
n

that local coordinates (Cartesian) are given in U

U → Rn
( )
x → x1 (x), . . . , xn (x) = ⃗x(x)
xj : U → R are continuous functions (one-valued!)

⃗x(x) = ⃗x(y) ↔ x = y

6
 Lecture 2. Manifolds and Atlases.
Manifolds: = Hausdorff (or metric) spaces such that they are “locally
euclidean”: for every x ∈ M there exists an open set x ∈ U with homeo-
morphism

φU : U → Rn = (x1 , . . . , xn ) (so U ⊂ Rn )

The set U represents a “Chart” in the “Atlas” on M . “Local coordinates”

in U (near x)
x → Rn − →i
R
x

are continuous functions in U and

⃗x(x) = ⃗x(y) ↔ x = y

for any two points x, y in U .

For a given “Atlas” {Uα }, covering M , we can introduce “Transition
Maps” in the intersections U ∩ V , where U = Uα , V = Uβ . Thus, for
any x ∈ U ∩ V we can use the local coordinates (x1 , . . . , xn ) (⃗xα in U )
or (y 1 , . . . , y n ) (⃗xβ in V ). The functions xi (y 1 , . . . , y n ) and y k (x1 , . . . , xn )
represent maps of euclidean domains

xi (y), i = 1, . . . , n, y k (x), k = 1, . . . , n

Definition. Manifold M is C ∞ if all the functions xi (y) (given by Atlas for

every pair U , V ) are C ∞ .
Statement. In the C ∞ Atlas all Jacobians det |∂xi /∂y k | are non-zero.
Proof. Since ⃗x(⃗y ) and ⃗y (⃗x) are both C ∞ we have det |∂xi /∂y k | ̸= 0.
∑ ∂xi ∂y k
k ∂xj
= δji
k
∂y
∑
Summation agreement: we do not write , so, in our notations

∂xi ∂y k
= δji
∂y k ∂xj

Definition: Oriented Atlas is such that all det |∂xi /∂y k | > 0
Oriented manifold: = there exists an oriented Atlas.
Examples:

7
 x
n = 2:
S2 is oriented manifold.
RP2 is NOT.
(x, −x) is one point in RP2 .
−x S2
Definitions.
1) C ∞ - function in C ∞ - manifold M with given Atlas of Charts:
M → R is C ∞ in every Chart.
2) C ∞ - map
F
M − → N
(Uα ) (Vβ )
is C ∞ in every Chart. In other words, for every pair Uα , Vβ the corresponding
functions yβk (x1α , . . . , xnα ) are C ∞ in F −1 (Vβ ) ∩ Uα .
Rank of the map F : M → N at the point x ∈ M :
( )
∂yβk Uα −
F
→ Vβ
rkx F = rk where
∂xiα (x) (y)
x

Statement. Rank of the map at the point x does not depend on the choice
of Atlas.
Proof. Let Uα , Vβ be Euclidean domains giving the charts of the manifolds
M and N and the functions ⃗y (⃗x) be represented by the map
F
Uα −→ Vβ
(x) (y)

Consider two other domains Wα and Xβ with coordinates x̃ and ỹ repre-

senting two other charts containing the points x and F (x) respectively. Let
us consider the functions ỹ(x̃) = ỹ(y(x(x̃))) defined by the map
F
Wα → Uα −→ Vβ → Xβ
(x̃) (x) (y) (ỹ)

We have
∂ ỹ i ∂ ỹ i ∂y k ∂xs
= ,
∂ x̃j ∂y k ∂xs ∂ x̃j
where rk |∂ ỹ i /∂y k | ̸= 0, rk |∂xs /∂ x̃j | ̸= 0.

8
 Conclusion
∂ ỹ i ∂y i
rk = rk
∂ x̃j ∂xj
Statement is proved.
F
Special Case: M −
→ R (function).
rkx f = 1 − regular point (∇f |x ̸= 0)
rkx f = 0 − critical point (∇f |x = 0)
Example: f = x +y : (x, y) ̸= (0, 0) - regular point, (0, 0) - critical point.
2 2

Statement. Every local coordinate xiα in the Atlas of Charts for C ∞ -

manifold M = ∪ Uα is such that all the points in the Chart Uα are regular
for xiα : Uα → R.
Proof. In the coordinate system ⃗xα in the Chart Uα we have ∇xiα =
(0, . . . , 1, 0, . . . , 0) ̸= 0.
Statement is proved.
“Good Double Atlas” (GDA):

M = ∪α Uα = ∪α Vα

where Uα ⊂ Vα and there are common coordinates ⃗xα for every Uα , Vα such
that the corresponding domains Uα , Vα in the Euclidean space are defined
by the relations:
V
∑ ∑
Uα : (xjα )2 < 1 , Vα : (xjα )2 < 2 . U
j j

Lemma. For every compact manifold M there exists a GDA.

Proof (for compact M ). For every point x ∈ M we can obviously choose
“small balls” Ux , Vx , x ∈ Ux ⊂ Vx with the required local coordinates.
After that we chose a finite cover of M which gives the required GDA.
Lemma is proved.
∑
Choose a C ∞ - function φ(r), r2 = j |xj |2 , such that:

9
 φ ≡ 0 for r ≤ 0, 1 ϕ(r)
φ ≡ 1 for r ≤ 1,
φ ≡ 0 for r ≥ 2,
φ′ < 0 for 1 < r < 2.
0 1 r 2
Consider C ∞ - functions on the manifold M with GDA:
∑
x̃jα = φ(rα ) xjα , rα2 = (xjα )2
j

We have: x̃jα = xjα for rα ≤ 1, and x̃jα = 0 for rα ≥ 2 (i.e. outside Vα ).

Theorem. Let a compact manifold M n with GDA (finite) be given

M n = U1 ∪ . . . ∪ UQ = V1 ∪ . . . ∪ VQ

Consider the coordinates ⃗xα (rα < 1 in Uα and rα < 2 in Vα ) and the
corresponding functions φ(rα ) and x̃iα on M n .
Then the map F : M n → RN , N = Q(n + 1):
[ ]
F (x) = x̃11 , . . . , x̃n1 , φ(r1 ), . . . , x̃1Q , . . . , x̃nQ , φ(rQ )

is a C ∞ imbedding (nondegenerate).
Terminology:
Imbedding: = rkx F = n at every point x and F (x) = F (y) → x = y.
Immersion: = rkx F = n = dim M at every point x ∈ M .
rkxF=0

M 1 R2
imbedding immersion
Proof of the Theorem.
I. Let x ∈ Uα , then rk F = n because
[ ] ∂ x̃iα
x̃iα = xiα for x ∈ Uα , F = . . . , x̃1α , . . . , x̃nα , . . . , and rk = n
∂xjα
(unit matrix in Uα ).

10
 II. Suppose x ̸= y. Can we still have F (x) = F (y) ?
a) Let x, y ∈ Uα then F (x) ̸= F (y) by the same reason as in I.
b) Let x ∈ Uα , y ∈ Vα (rα > 1). Then φ(rα )|x = 1, φ(rα )|y < 1, so
F (x) ̸= F (y), [ ]
F = . . . , x̃1α , . . . , x̃nα , φ(rα ), . . .
c) Let x ∈ Uα , and y is outside Vα . Then φ(rα )|x = 1, φ(rα )|y = 0, so
F (x) ̸= F (y).
Theorem is proved.

Lecture 3. C ∞-manifolds, Atlases, Charts.

Especially good Atlases. Implicit functions
and Inversion. Imbedding of Compact Man-
ifolds in RN . Immersions.
Partition of Unity.
Let /
∑
Q
ψα (x) = φ(rα ) φ(rβ )
β=1

We can easily see the that all the functions ψα (x) represent C ∞ - functions
on M with the following properties
∑
ψα (x) ≥ 0 , ψα (x) ≡ 1
α

Example of Application.
What is an Integral on a Manifold?
∫
I = f (x) dn σ
M

(dn σ represents some measure on M ).

We can write
∫ ∑ ∑∫
n
I = f (x) ψα (x) d σ = f (⃗xα ) ψα (⃗xα ) dn σ
M α α Vα

11
 We can see now that I is represented as a sum of ordinary integrals
over local domains where we can also put in local coordinates: dn σ =
gα (⃗xα ) dx1α . . . dxnα .
Tangent vector on a C ∞ - manifold M = ∪α (Uα , ⃗xα ):
a) Vector τ is attached to a point x ∈ M .
b) Vector τ is characterized by “components” τ
(τ 1 , . . . , τ n ) in the given system of local coordinates. (x1 ,...,x n)
x
c) Basic vectors ei in the coordinate system ⃗x are identified with the
operators ∂/∂xi :
∂
ei ↔ , τ = τ i ei
∂xi
d) The vector τ is identified with the differential operator τ i ∂/∂xi :
∂
τ ↔ τi ,
∂xi
so τ acts on the functions f (x) at the point x by the formula:
∂f
τ (f ) = τ i
∂xi x

(derivative along the vector τ ).

For a smooth curve ⃗x(t) = (x1 (t), . . . , xn (t)) the vector
d⃗x ( )
τ = = ẋ1 , . . . , ẋn x=x(t0 )
dt
represents the “speed of particle” at the point t0 (nothing to do with “rela-
tivistic speed” in the Special Relativity).
Change of Coordinates:
Let us make a non-degenerate change of coordinates
xi = xi (y 1 , . . . , y n ) , i = 1, . . . , n ,
such that we can write f (⃗x) = f (⃗x(⃗y )) for any smooth function f (⃗x) near
the point x ∈ M . We say that the sets (τ 1 , . . . , τ n ) and (τ ′1 , . . . , τ ′n ) rep-
resent the components of the same vector τ in the coordinate systems {xi }
and {y i } respectively if we have for any f (x) at the point x:
∂f ′j ∂f
τi = τ
∂xi ∂y j

12
 By definition, we can write
i
∂f ′j ∂x ∂f
τi = τ ,
∂xi ∂y j ∂xi
so we come to the conclusion
∂xi
τ i = τ ′j
∂y j
(summation over j is assumed).
In the same way, for the inverse transformation ⃗y = ⃗y (⃗x) we can write

∂y j
τ ′j = τ i
∂xi
where
∂xi ∂y j
= δki
∂y j ∂xk

Every C ∞ - map is given locally by its linear part and the smaller terms
∑
n
i i 1 n
y = F (x , . . . , x ) = Const + Aij xj + O(||x||2 )
j=1

(near ⃗x = 0).
Inversion of Map (C ∞ ).
Let us have a map

F : Rn → Rn , x0 → y0
(x) (y)

given in coordinate form by the functions y i = y i (x1 , . . . , xn ).

If we have the relation
∂y i
det ̸= 0
∂xk x0

then there exist open sets V ∋ x0 , U ∋ y0 and a map

G : Rn → Rn , y0 → x0 ,
(y) (x)

13
defined in the set U , such that for x ∈ V , y ∈ U we have the relations

G (F (x)) ≡ x , F (G(y)) ≡ y

Naturally, we have in this case

( ) ( )
∂xi ∂y j ∂x ∂y
= δki , i.e. = I
∂y j y0 ∂xk x0 ∂y y0 ∂x x0

Implicit Functions.
Let us have a coordinate system (y 1 , . . . , y n+k ) in Rn+k and a system of
k equations z 1 = 0, . . . , z k = 0, z i = z i (y 1 , . . . , y n+k ) with the Condition:
( i)
∂z
rky0 = k
∂y j

(maximal rank).
Statement.
Let us assume that under the above conditions we have the relation
( i)
∂z
det T ≡ det ̸= 0 , j = n + 1, . . . , n + k ,
∂y j y0

Jacobi Matrix
 

 
 R2 z=c
 ∂z i ∂z i ∂z i ∂z i  y1
z=0
, . . . , , , . . . ,

 ∂y 1 ∂y n ∂y n+1 ∂y n+k 

 | {z }  n=k=1
nonzero determinant is here y2
y0

Then:
1) There exists an open set U ∋ y0 ∈ Rn+k near the point y0 , where
the values (y 1 , . . . , y n , z 1 , . . . , z k ) represent a coordinate system;
2) The change
( 1 ) F ( 1 )
y , . . . , y n+k −→ y , . . . , yn, z1, . . . , zk

is C ∞ and nondegenerate.

14
Proof.
Easy to see that the Jacobian Matrix of the transformation F can be
written in the form:
 
1 0 ... 0 ∂z 1 /∂y 1 ... ∂z k /∂y 1
0 1 ... 0 ∂z 1 /∂y 2 ... ∂z k /∂y 2 
. . 
. . ... .. .. ... .. 
. . . . . 
 n 
J = 0 0 ... 1 ∂z 1 /∂y n . . . ∂z /∂y 
k
 
0 0 ... 0 ∂z 1 /∂y n+1 . . . ∂z k /∂y n+1 
. . .. .. .. 
 .. .. ..
. . .
..
. . 
1 n+k
0 0 . . . 0 ∂z /∂y ... ∂z k /∂y n+k
We immediately get then det Jy0 = det T ̸= 0. So, we get our statement
from the Inversion Theorem.
Statement is proved.
Under the above conditions, the “Implicit
y z=c Function Theorem” states that on the
0 z=0 submanifold, given by the relations z 1 = 0,
z=c’
. . . , z k = 0, the values (y n+1 , . . . , y n+k ) can
be locally expressed as explicit functions of
the coordinates (y 1 , . . . , y n ):

y n+1 = φ1 (y 1 , . . . , y n ) , . . . , y n+k = φk (y 1 , . . . , y n )

The Implicit Function Theorem can be considered as a corollary of the

Statement formulated above. Indeed, we have a “local coordinate system”
( 1 )
y , . . . , yn, z1, . . . , zn

near the values ⃗z = 0 in the domain U near the point y0 .

We can then write in this domain for the inverse coordinate transforma-
tion:

y n+1 = φ1 (y 1 , . . . , y n , z 1 , . . . , z k ) , . . . , y n+k = φk (y 1 , . . . , y n , z 1 , . . . , z k )

Putting now ⃗z = 0 we get immediately the statement of the Implicit

Function Theorem.
Another corollary:

15
 Every imbedding of manifold

M n → M n+k
(x) (y)
 
∂y i locally can be given by k nondegenerate
rk ∂xq x0
=n equations z 1 = 0, . . . , z k = 0.

Proof. Consider the imbedding M n → M n+k , given by smooth functions

y i = φi (x1 , . . . , xn ), i = 1, . . . , n + k , such that
( i)
∂y
rk = n
∂xq

Let ( )
∂y i
det ̸= 0 , i = 1, . . . , n
∂xq x0
According to the Inversion Theorem, the transformation
( 1 ) ( )
x , . . . , xn , y n+1 , . . . , y n+k → y 1 , . . . , y n , y n+1 , . . . , y n+k

is locally invertible, so we can introduce the new coordinate system

( )
y ′ = x1 , . . . , xn , y n+1 , . . . , y n+k

near the point y0 = F (x0 ).

The imbedding M n → M n+k can be given now near the point y0 by the
set of equations z 1 = 0, . . . , z k = 0 , where

z 1 = y n+1 − φn+1 (x1 , . . . , xn ) , . . . , z k = y n+k − φn+k (x1 , . . . , xn )

We have also ( )
∂z p
det ̸= 0 ,
∂y n+q
so the values (x1 , . . . , xn , z 1 , . . . , z k ) give also a local coordinate system in
Rn+k .
Corollary is proved.

16
 Lecture 4. Manifolds and Submanifolds: Im-
plicit functions and Inversion. Vectors and
Covectors.
According to the previous lecture, we can formulate here the following
Statement.
Let us have an imbedding

F : M n → N n+k , (rkx F = n , ∀x , F (x) ̸= F (y) , ∀x ̸= y) ,

such that either M n is compact or for every compact set X ⊂ N the

intersection X ∩ F (M ) is compact.
Then:
For every point x ∈ M n ⊂ N n+k there exists local coordinate system
(x , . . . , xn , z 1 , . . . , z k ) in some U ∋ x such that the submanifold M n is
1

given in U by the system

z1 = 0 , . . . , zk = 0

Conclusion.
Every C ∞ - manifold can be given by a set of
local equations in RN : M n → RN (proved
for compact M n ).
Remark. Not all manifolds can be given by global nondegenerate set of
equations in RN

φ1 (y 1 , . . . , y N ) = 0 , . . . , φN −n (y 1 , . . . , y N ) = 0

(Proof later).
RP2 can not be given by nondegenerate global set of equations.
“Tangent Manifold” T ∗ (M n ).
Let us denote again by τ a tangent vector (τ 1 , . . . , τ n ) in M n attached
to a point x ∈ M n .
‘Tangent Manifold” T ∗ (M n ):
( )
M n = ∪α Uα , x1α , . . . , xnα (Atlas, C ∞ )

17
 ( )
T ∗ (M n ) = ∪α Uα × Rn , x1α , . . . , xnα , τ 1 , . . . , τ n
Change of coordinates:
Let us put in the intersection of Charts Uα and Uβ : x = xα , y = xβ .
We have then

xi = xi (y) = xi (y 1 , . . . , y n ) (in Uα ∩ Uβ )

The components of the same vector in the coordinates (x) and (y):

(x, τ ) ↔ (y, τ ′ )

are connected by the relations:

( )
∂xi ∑
τ i = τ ′j is assumed
∂y j j

Covectors (η1 , . . . , ηn ) are attached to the points x ∈ M n .

Change of coordinates

(x, η) ↔ (y, η ′ )

∂xi ∂y j
ηj′ = ηi , ηi = ηj′
∂y j ∂xi
( )
T∗ (M n ) = ∪α Uα × Rn , x1α , . . . , xnα , η1 , . . . , ηn
Scalar product (invariant):

⟨τ , η⟩ = τ i ηi

Conclusion: Spaces of vectors and covectors are dual.

Basis:
∂
vectors : ei ↔
∂xi
covectors : ei ↔ d xi

∂
Vector field : fi (Vector fields)
∂xi
Covector field : gi dxi (Differential 1 − forms)

18
 Vector Fields = Dynamical Systems (Autonomous)
Integration of 1-forms along the path: let

ω = gi (x) dxi
and the path is given by a piecewise
x(t)
smooth one-parametric curve
( ) B
⃗x(t) = x1 (t), . . . , xn (t) A

Definition.
∫ ∫ B ∫ t1 ( )
i dxi
ω = gi (x) dx = gi (x(t)) dt
x(t) A t0 dt
Properties.
Integral does not depend on the choice of coordinates in M n and “time”
t.

t O.K. Nonmonotonic changes of time

t = t (t′ ) are also admissible.

t’
Inner Product of tangent vectors

gij (x) τ i τ̃ j = ⟨τ , τ̃ ⟩

Change of coordinates x = x(y):

∂xi ∂xj ′
gij (x(y)) = gks (y)
∂y k ∂y s
Let Tensor field gij (x) be symmetric:

gij = gji : ⟨τ , τ̃ ⟩ = ⟨τ̃ , τ ⟩

We will require also that tensor gij (x) is nondegenerate.

Under the above conditions we say that gij (x) defines a Pseudoriemannian
Metric on M n .

19
Types of inner products (“Types of Geometry”):
gij dxi dxj > 0 - Riemannian Metric.
Special Types (p, q) of Pseudoriemannian metric:
p = 0 : - Riemannian
p = 1 : - Lorentzian
p = q : - Ultrahyperbolic
Theorem.
In every C ∞ - manifold there exists a C ∞ Riemannian Metric.
Proof. Take GDA (Good Double Atlas)
M n = U1 ∪ . . . ∪ UQ = V1 ∪ . . . ∪ VQ
(|⃗xα | < 1 in Uα and |⃗xα | < 2 in Vα ) and the function φ(r) introduced in
Lecture 2.

1 ϕ(r)

0 1 r 2
Consider quadratic forms
∑
n
( )2
Gα = dx̃jα , x̃jα = xjα · φ(rα )
j=1

Take Riemannian metric

∑
G = Gα
α

(We assume that every point belongs to finite number of domains Vα ).

We claim that it is positive.
Theorem is proved.
Remark.
Lorentzian Nondegenerate Metric does NOT exist in S2 (or RP2 ). [Topol-
ogy claims that Euler characteristics should be 0.]

20
 Homework 1.
1. Local coordinates: ρ and φ in R2 .
a) Prove that ρ can not be chosen as a Cartesian coordinate in the
whole domain R2 \0 (whole).
b) Prove that φ can be defined as a one-valued function in any domain
not containing closed paths surrounding 0 .
111111111111111111111111111111
000000000000000000000000000000
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
U 000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
0
000000000000000000000000000000
111111111111111111111111111111
U 000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111
000000000000000000000000000000
111111111111111111111111111111

Define φ in these domains.

2. Spherical coordinates in R3 \0

x = r sin θ cos φ , y = r sin θ sin φ , z = r cos θ

r2 = x2 + y 2 + z 2
Let r = 1 .
N
a) In which domains θ is a good Cartesian
coordinate? r=1
b) In which domains (φ, θ) are good local
coordinates?
S
3. Construct C ∞ functions
1 1
ϕ(r) ψ(r)

0
0 1 r 2 −1 1

using {
0 , x ≤ 0
f (x) =
e−1/|x| , x ≥ 0
2

21
 Lecture 5. Manifolds and vectors fields. Im-
portant Examples.
Classes of manifolds:
a) All C ∞ - manifolds M n ⊂ RN for given N .
b) Manifolds defined by the Global Nondegenerate systems of Equations
in RN
f1 = 0 , . . . , fN −n = 0
(Important cases: n = 2, N = n + 1 = 3).
Classification of manifolds:
n=1: S1 , R
↑ ↑
compact noncompact

n=2: compact orientable manifolds are

S2 , T2 , . . . , S2g . 2
Sg =

nonorientable:
b
y x
2 Klien
RP2 , K2 (Klein Bottle), other. RP = K 2= Bottle
a a
−x
−y
(−x,x) RP 2 b

“Connected sum” of previous manifolds:

RP2 # RP2 # . . . # RP2

M1 M2
K2 = RP2 # RP2 M1 M2

Noncompact 2 - manifolds.
Important manifolds:
RPn - real projective spaces
CPn - complex projective spaces

22
 QPn - “quaternionic” projective spaces
Groups: GLn (R), GLn (C), SLn (R), SLn (C), On ⊃ SOn , Op,q ⊃
SOp,q (case p = 1 - Lorentz Groups), Un ⊃ SUn , Up,q ⊃ SUp,q
Stiefel Manifolds
Grassmann Manifolds
Isometry groups of Rn
Isometry groups of Rp,q
where
Euclidean Metric: gij = δij
Lorentzian Metric: gij = diag (1, −1, . . . , −1) , i.e.
 
1 0 ... 0
0 −1 . . . 0 
 
gij =  .. .. . . . 
. . . .. 
0 0 . . . −1

Vector fields and Dynamical Systems.

Vectors on Manifolds M n
( )
M n = ∪α Uα , x1α , . . . , xnα

τ = (τ 1 , . . . , τ n ).
Covectors on Manifold M n : η = (η1 , . . . , ηn ).
T ∗ (M n ) - space of all vectors on M n .
T∗ (M n ) - space of all covectors on M n .
Vector field
τ (x) = (τ 1 (x), . . . , τ n (x))
locally
generates dynamical system:

ẋi = τ i (x) , i = 1, . . . , n
Solution = curve ⃗x(t): ẋi = τ i (x)

Cauchy Theorem: Let τ (x0 ) ̸= 0 . There exists system of local coordinates

(y 1 , . . . , y n ) such that τ = (1, 0, . . . , 0). In this coordinate system we have

23
 x(t) → y(t) = (t, y02 , . . . , y0n )
solution const

Let τ (x0 ) = 0 . Consider M n × R = M ′ and vector field τ ′ = (τ, 1) in

M ′ near x = x0 .

τ ′ (x0 ) ̸= 0 .
Apply Theorem of Cauchy.
t
One-parametric group generated by vector field.
The equation
ẋ = τ (x)
generates a (local) one parametric group of invertible transformations of (lo-
cal domains) in M n :

St : M n → M n x(t)
x(0) = x0 , x(t) = St (x0 )
x0

S0 = I , S−t = St−1 , St+t′ = St ◦ St′ = St′ ◦ St

Commuting Vector Fields: τ1 (x) , τ2 (x) .

Definition:
[τ1 , τ2 ] = τ1 ◦ τ2 − τ2 ◦ τ1
∂
τ = τ i (x)
∂xi
∂ j ∂ ∂2 i ∂η
j
∂
τη = τi η = τ i ηj − τ
∂xi ∂xj i
∂x ∂x j ∂x ∂xj
i

Corollary
( )
∂η j ∂ i ∂τ
j
∂ i ∂η
j
i ∂τ
j
∂
[τ , η] = τ i
i j
− η = τ − η
∂x ∂x ∂x ∂xj
i ∂x i ∂x i ∂xj

24
 Remark. Sometimes we denote operator τ (f ) acting on scalar functions
by ∇τ f because it coincides in this case with covariant derivative of the scalar
field f along vector field τ .
Statement: Let τ (x0 ) ̸= 0 , η(x0 ) ̸= 0 . Then [τ , η] ≡ 0 ⇒ there exists
local coordinate system (y 1 , . . . , y n ) such that

τ = (1, 0, . . . , 0) , η = (0, 1, 0, . . . , 0)

They generate a (local) commutative group R2 :

St − shifts by (τ )
St′′ − shifts by (η)

St′′ ◦ St = St ◦ St′′

Examples.
1. τ = const in Rn : group of shifts x → x + t · const .
2. τ i = aij xj (linear)

ẋj = τ j (x) , x0 → x(t) , x(0) = x0

- x(t) generate linear maps St .

3. Tr aij = 0 ⇒ St ∈ SLn (R) .
4. aij = −aji ⇒ St ∈ SOn .

Lecture 6. Group Manifolds. Lie Algebra.

Important Examples.
Group Manifolds.
1) GLn (R) , GLn (C) , GLn (Q) .
The notation Q represents here the ”noncommutative field” of quater-
nions:
{a + ib + jc + kd}
where

i2 = j2 = k2 = −1 , ij = −ji = k , jk = −kj = i , ki = −ik = j

25
 dim GLn (R) = n2
dim GLn (C) = 2n2
dim GLn (Q) = 4n2
Question: prove that GLn (R) has 2 components. GLn (C) is connected.
2) SLn (R) , SLn (C) : det = 1
Equation : det A = 1 in Rn or in Cn
2 2

Is this equation nondegenerate ?

3) On , A At = I ,
a) Is this set of equations NONDEGENERATE in GLn (R) ∈ Rn ?
2

b) Is SOn - connected manifold?

c) dim On = n(n − 1)/2
4) Op,q : ⟨Aη , Aζ⟩ = ⟨η , ζ⟩
∑
a) p = 1 , q = n , ⟨η , ζ⟩ = η 0 ζ 0 − nα=1 η α ζ α , (1, n)
∑p ∑p+q
b) ⟨η , ζ⟩ = α=1 η α α
ζ − α α
α=p+1 η ζ , (p, q)
5) Unitary group: ζ, η ∈ Cn
∑ ∑
⟨ζ , η⟩ = ⟨η , ζ⟩ = ζ i η̄ i − ζ i η̄ i
i≤p i>p

Up,q : ⟨Aζ , Aη⟩ = ⟨ζ , η⟩ , (q = 0 : Un )

SUn , SUp,q : det A = 1

How to introduce local coordinates in the group manifold M n = G ?

Let A(t) ∈ G , A(0) = I .

dA
= B ∈ Lie Algebra
dt t=0

Another form
dA −1 dA
Lie Algebra → A or A−1
dt dt
Groups SOn , On :
d
⟨A(t)ζ , A(t)η⟩ = ⟨ζ , η⟩ , ⟨A(t)ζ , A(t)η⟩ = 0 ,
dt
26
 A(t) = A(0) + Ȧ(0) t + O(t2 ) , A(0) = 1
↑
B

Lemma 1: B t = −B (i.e. Ȧt (0) = −Ȧ(0) ).

Proof.

⟨A(t)ζ , A(t)η⟩ = ⟨ζ , η⟩ + t [⟨Bζ , η⟩ + ⟨ζ , Bη⟩] + O(t2 )

d
⟨A(t)ζ , A(t)η⟩ = 0 ⇒ ⟨Bζ , η⟩ + ⟨ζ , Bη⟩ = 0
dt t=0
i.e.
⟨Bζ , η⟩ ≡ − ⟨ζ , Bη⟩
Lemma is proved.
Lemma 2: ∃ ϵ > 0 such that for any A ∈ SOn , ||A − I|| < ϵ we have
A = eB , B t = −B .
Proof.
For small enough ϵ consider the convergent series

B = log (I + A − I) = A − I − (A − I)2 /2 + (A − I)3 /3 − . . .

B t = log (I + At − I) = At − I − (At − I)2 /2 + (At − I)3 /3 − . . .

t
We can write then: A = eB , At = eB . Besides that, from the
commutativity of all the terms of the series we can write also:
t
eB+B = A At = I

which implies B t = −B for small enough B .

Lemma is proved.
Local coordinates in GLn (R) near I can be also taken from a small ball
in the space Rn = space of matrices.
2

Lemma 3: For small enough B we have: Tr B = 0 ⇔ det eB = 1.

Proof.
For the diagonal∏
matrices B = diag (b1 , . . . , bn ) we obviously have the
B
relation det e = ebi = exp (Tr B) . The same property is then also

27
evident for the diagonalizable matrices B = S −1 ◦ diag (b1 , . . . , bn ) ◦ S
from the series representation of the matrix eB . Since the set of diagonaliz-
able matrices is dense in the matrix space and the functions exp (Tr B) and
det eB are analytic functions of the matrix entries we actually have
det eB = eTr B
for any matrix B . The statement of the Lemma for small enough B follows
then from the properties of the function ex .
Lemma is proved.
We can see then that in the local coordinate system in GLn (R) , given
by the entries of the matrix B near I ∈ GLn (R) , the equation det A = 1
has the form Tr B = 0 .
Conclusion. This equation is linear and NONDEGENERATE.
Group SOn : Equations for SOn in the coordinate system, given by the
entries of B near I ∈ GLn (R) , have the form B t = −B .
Unitary group A Āt = I.
Un ⊂ GLn (C) , we can put again A = eB near I = GLn (C) and
introduce a local coordinate system, given by the entries of B in Cn =
2

R2n . In this coordinate system:

2

a) Group SLn (C) is given by equation Tr B = 0 (2 real equations).

b) Group Un is given by equations B t = −B̄ (B̄t = −B) , i.e. n2
linear equations over R .
dim Un = n2 (over R ). U1 = S1 = SO2 dim U2 = 4 , is SU2 = S3 ?
( )
a b
A ∈ SU2 : ⇒ A = ,
c d
where ac̄ + bd¯ = 0 , |a|2 + |b|2 = 1 , |c|2 + |d|2 = 1 , |ad − bc| = 1 .
So, we have
( )
a b
A = , |a|2 + |b|2 = 1 ,
−b̄ ā
i.e. SU2 = S3 .
Quaternion Group: GLn (Q) .

Q : {q = a + ib + jc + kd}

28
 i2 = j2 = k2 = −1 , ij = −ji = k , jk = −kj = i , ki = −ik = j
q̄ ≡ a − ib − jc − kd
GL2 (Q) = {q : q q̄ = 1} = {a2 + b2 + c2 + d2 = 1} = S3 = SU2

Pauli matrices:
( ) ( ) ( )
0 1 0 i i 0
i σx = , i σy = , i σz =
−1 0 i 0 0 −i
Correspondence to the basic quaternions (i, j, k):
i σx ↔ i , i σy ↔ j , i σz ↔ k

SO3 = SU2 /Z2 = S3 /Z2 = RP3

Indeed, consider the norm-conserving transformations
q → q1 q q̄1 , q1 q̄1 = 1
Space Span {1} and the orthogonal space Span {i, j, k} are invariant.
Two transformations coincide iff : q1′ = ±q1 .

SO4 = S3 × S3 /Z2
Consider the transformations
q → q1 q q 2 , q1 q̄1 = 1 , q2 q̄2 = 1
Two transformations coincide iff : (q1′ , q2′ ) = ±(q1 , q2 ) .

Lecture 7. Group Manifolds: Compact Lie

groups. Most important Examples. Non-
compact Lie groups. Most important Exam-
ples. Lie Algebras. Gradient-like Systems.
Group manifolds.

29
 Compact groups: On , SOn , Un , SUn , Tn = Rn /Zn , Spn =
GLn (Q) ∩ SO4n , SO2 = U1 = S1 , SO3 = SU2 /Z2 , SU2 = Sp1 = S3 .
Local coordinates (near I ):
1. SOn : B t = −B , dim SOn = n(n − 1)/2
2. Un : B t = −B̄ , dim Un = n2
Noncompact groups: Rn , Iso (Rn ) −→ Rn (shifts), GLn (R) ,
On
GLn (C) , SLn (R) , SLn (C) , Op,q , Up,q , Sympln (?) .

 
1 ... 0 0 ... 0
 .. . . . .. .. . . . .. 
. . . . 
 
0 ... 1 0 ... 0 
Rp,q : (e1 , . . . , en ) , ⟨ei , ej ⟩ =  
0 . . . 0 −1 . . . 0 
. . . .. .. . . . 
 .. . . . . .. 
0 . . . 0 0 . . . −1

A ∈ Op,q : ⟨Aζ , Aη⟩ = ⟨ζ , η⟩ , ζ, η ∈ Rp,q

A ∈ Up,q : ⟨Aζ , Aη⟩ = ⟨ζ , η⟩ , ζ, η ∈ Cp,q

Symplectic (skew-symmetric) product: ⟨ζ , η⟩ = −⟨η , ζ⟩ :

R2n : (e′1 , e′′1 , . . . , e′n , e′′n ) , ⟨e′i , e′j ⟩ = 0 , ⟨e′′i , e′′j ⟩ = 0 , ⟨e′i , e′′j ⟩ = δij

 
0 1 0 0 ... 0 0
−1 0 0 0 ... 0 0
 
0 0 0 1 ... 0 0
 
 0 −1 0 . . . 0 0
gij = 0 
 .. .. .. .. . . . .. 
 . . . . . .. .
 
0 0 0 0 ... 0 1
0 0 0 0 . . . −1 0

A ∈ Sympn : ⟨Aζ , Aη⟩ = ⟨ζ , η⟩ , ζ, η ∈ R2n (Symplectic Space)

Lie Algebras:

30
 1. SOn : B t = −B , (real entries) .
2. Un : B = −B̄ ,
t
(complex entries) .

SOp,q : ⟨Bζ , η⟩ = −⟨ζ , Bη⟩ , B −?

Spaces Rp,q with pseudoriemannian inner product (“metric”):
 
1 ... 0 0
... 0
 .. .. .. ..
..
. .. 
.
. . . . 
 
0 ... 1 0 ... 0 
gij = ⟨ei , ej ⟩ =  
0 ... 0 −1 . . . 0 
. .. .. . . . 
 .. ..
. . . . .. 
0 . . . 0 0 . . . −1

Consider matrix B = (bij ) . We have

( )
(Bζ)i = bij ζ j , ⟨Bζ , η⟩ = gki bij ζ j η k = b̃kj ζ j η k

Claim. For Lie Algebra we have b̃kj = −b̃jk (Not for B !) .

Proof.
⟨Bζ , η⟩ = gki bij ζ j η k = b̃kj ζ j η k
⟨ζ , Bη⟩ = gki bij ζ k η j = b̃kj ζ k η j = b̃jk ζ j η k
Statement is proved.
Conclusion.
1) Let A(t) ∈ G , A(t) = I + B t + O(t2 ) and ⟨Aζ , Aη⟩ = ⟨ζ , η⟩ .
Then we have B̃ t = −B̃ , B̃kj = gki bij . Here G is any group (G Op,q ,
Sympln , . . . ).
2. Lie Algebra of every group Op,q can be identified with the set of
skew-symmetric matrices

B̃kj = −B̃jk = gki bij

Matrix B is equal to

bij = g ik gks bsj = g ik b̃kj

31
with definition ( )
g ik = (gki )−1

General definition
Let Riemannian (pseudoriemannian) manifold M n be given with Atlas
(α)
of Charts Uα , (x1α , . . . , xnα ) and “metric” gij (x) in the domain Uα . For 2
tangent vectors attached to x we have

⟨τ , η⟩ = gij (x) τ i η j

We define “metric” in the space of covectors ξ = (ξi ), κ = (κj )

⟨ξ , κ⟩ = g ij (x) ξi κj

where (g ij ) = (gkl )−1 , i.e. g ij gjk = δki .

Conclusion. T ∗ (M n ) ∼
= T∗ (M n ) .
Remark. “Skew” metrics

gij (x) = − gij (x)

can be nondegenerate only for n = 2k (n even). We have here ⟨η , η⟩ ≡ 0 .

Gradient vector field
Let function f : M n → R be given. Its gradient is:
( )
∂f
vector : ∇ f (x) =
g ij
g (x) j
∂x
( )
∂f
covector : d f =
∂xi
Gradient system is
∂f
ẋi = g ij = η i (x)
∂xj

Lemma. Let h(x) be any function in M n . We have

dh ∂h
= η i (x) i = ⟨∇g h , ∇g f ⟩
dt ∂x

32
 Proof.
dh ∂h ∂h ∂f
= η i (x) i = g ij (x) i j
= ⟨dh , df ⟩ = ⟨∇g h , ∇g f ⟩
dt ∂x ∂x ∂x
Lemma is proved.
Corollary 1.
df
g ij = − g ji ⇒ ≡ 0
dt
(Symplectic case).
Corollary 2. For Riemannian metric gij η i η j > 0 we have df /dt > 0
along the gradient system
∂f
ẋi = g ij j
∂x
ij −1
(because (g ) = (gij ) ) .
Important example of Symplectic Manifold is T∗ (M n )
dim T∗ (M n ) = 2n , Atlas : {Uα , (x1α , . . . , xnα , pα1 , . . . , pαn )}
Fix α - number of Chart.
∂ ∂
Tangent basis : e′i = , e ′′
i =
∂xi ∂pi
Inner Product is
⟨e′i , e′j ⟩ = 0 , ⟨e′′i , e′′j ⟩ = 0 , ⟨e′i , e′′j ⟩ = − ⟨e′′j , e′i ⟩ = δij
τ = (e′i ) - tangent vector to M n , p = (e′′j ) - tangent covector to M n ,
⟨τ , p⟩ = τ i pi - natural Invariant Inner Product.
Language of 2-forms
∑ ( )
0 I
Ω = dx ∧ dpi ,
i
gij =
−I 0
((x1 , . . . , xn , p1 , . . . , pn ) - local coordinates in T∗ (M n ) ) .
“Gradient systems” in the Symplectic Manifolds are called “Hamiltonian
systems”.

33
 Homework 2.
1. How many projection local coordinate systems for Sn ∈ Rn+1 are needed
to cover all sphere? n = 1, 2, 3, . . . .
2. How many local coordinate systems are needed to cover RPn ? Let
n = 1, 2, 3 .
( )
x ∈ RPn = Rn+1 \0 /x ∼ λx , λ ̸= 0

(x0 , . . . , xn ) ∼ (λx0 , . . . , λxn ) , λ ̸= 0

3. Prove that SO2 = U1 = S1 and SO3 = RP3 = S3 / ± 1 .

4. Which matrices belong to the Lie Algebra of the group SO2,1 ?
5. Find the group O1,1 . How many components does it have?
6. Prove that SO3 = SU2 / ± 1 .
7. Prove that SO4 = S3 × S3 / ± (1, 1) .
8. Prove that GL2 (R) has 2 components. Same for On .

Lecture 8. Riemannian, Pseudo-Riemannian

and Symplectic Geometries.Complex Geom-
etry. Restriction of Metric to submanifolds.
Length of curves and Fermát Principle.
I. Riemannian Geometry = Manifold + Riemannian Metric gij ,
gij η i η j > 0 .
II. Pseudoriemannian Geometry = Manifold + Pseudoriemannian Metric
gij , det gij (x) ̸= 0 .
III. Symplectic Geometry = Manifold + Symplectic Inner Product gij (x) =
− gji (x) , det gij (x) ̸= 0 . Corollary: dim M = 2k .
Main Example (Physics) = T∗ (M n ) .
Flat Geometry:
M n = Rn , gij = const

34
1. Riemannian Case: ||η||2g > 0 , ⟨ei , ej ⟩ = δij .
2. Pseudoriemannian Case: type p, q , ⟨ei , ej ⟩ = ±δij .
3. Symplectic Case: M 2k ,

⟨e′i , e′j ⟩ = 0 , ⟨e′′i , e′′j ⟩ = 0 , ⟨e′i , e′′j ⟩ = − ⟨e′′j , e′i ⟩ = δij

 
0 1 0 0 ... 0 0
−1 0 0 0 ... 0 0
 
0 0 0 1 ... 0 0
 
 0 −1 0 0
gij = 0 ... 0 
 .. .. .. .. .. .... 
 . . . . . . .
 
0 0 0 0 . . . 0 1
0 0 0 0 . . . −1 0

4. Complex Geometry: (z 1 , . . . , z n ) , z l = xl + iy l . Complex Vectors:

η = (η 1 , . . . , η n ) , η j ∈ C .

⟨η , ζ⟩ = gij η i ζ̄ j

⟨ei , ej ⟩ = gij = ḡji

||η||2 > 0 − positive case, ||η||2 = indefinite type → real form of type
(p, q).
Volume element in M n , gij (x):
√ √
dn σ = det gij , dx1 ⊗ · · · ⊗ dxn = g dn x

(Riemannian Case)
√ √
dn σ = (−1)q det gij , dx1 ⊗ · · · ⊗ dxn = ±g dn x
(Pseudoriemannian Case)
Transformation Rule (“Measure”) :

∂xi
x = x(y) , dn x = dx1 . . . dxn = det dy 1 . . . dy n = |J| dn y
∂y j

35
 Important Remark. For manifolds given by Oriented Atlas (J > 0)
we can write dn x = J dn y (differential forms !), J = det ||∂xi /∂y j || .
Lemma 1. Riemannian Metric in the manifold M n defines Riemannian
Metric in every submanifold W k ⊂ M n .
Proof.
Let W k ⊂ M n , locally we have for local coordinates y in M n : y i =
y i (x1 , . . . , xk ) , i = 1, . . . , n, where x represent some local coordinates in
W k.
We define “restriction of metric”
∂y s ∂y k
gij′ (x) = gsk (y(x))
∂xi ∂xj
(restriction of inner product on every linear subspace of tangent space). It
remains positive.
Lemma is proved.
Remark. The analogous lemma is wrong for Pseudoriemannian or Symplec-
tic Geometry because after the restriction to linear subspace metric might
become degenerate.

x0
<η,η>=0

η x2
R1,2 , light cone: ⟨η , η⟩ = 0 .

x1

Riemannian Metric ⇔ length of piecewise smooth curves.

36
 x(t)
B ∫ b√
l(γ) = gij (x(t)) ẋi ẋj dt
γ a

A
“Distance” = minγ l(γ)

a b
Statement. Riemannian metric transforms M n into metric space

l (a, c) ≤ l (a, b) + l(b, c)

Another metric induced by imbedding M n ⊂ RN (Either M n is compact

or M n ∩ DρN is compact for all ρ ≥ 0) .
“Geodesics” = “Locally shortest” paths

S2
b
a

Rn : geodesics = “straight lines” .

b

c1 − speed in the air

P. Fermat (XVII Century)
c 2− speed in the water

a
“Minimal Time Principle”: Light propagates from the point A to
the point B along the path with minimal time among all piecewise
smooth paths joining these 2 points

37
 ∫ ∫ b√
b
|dl| dl2
Time = =
a c(x(t)) a c2 (x)

y
√
dl = dx2 + dy 2
x
{
c1 , y > 0
Let c(x) =
c2 , y < 0

a
I(x)= c1

Minimal time = Minimize this integral:

x minx I(a, b, x) =?
c2

b
So we have a ”Fermat Riemannian Metric”
δij 1 1
gijF = , ||η||2F = ||η||2E , ||η||F = ||η||E
c2 (x) c2 c2

(F = “Fermat”, E = Euclid) .
Speed of light in vacuum c = cvac ∼ 3 · 1010 cm/sec , cmedia < cvacuum
.
How to find geodesics?
Euler - Lagrange (XVIII Century)
Consider more general problem. Let L(x, η) (“Lagrangian”) be a
smooth function in T ∗ (M n ) (tangent manifold). Fix points a, b ∈ M n .
Find “extreme curves” for the action
∫ 1
S(γ) = L (x(t), ẋ(t)) dt
0

38
(η = ẋ) on the piecewise smooth paths: x(t) , x(0) = a , x(1) = b .

b
a x(t)
Examples:
Geometry: action and length functionals
1 1
a) L = gij (x) η i η j = ||η||2
2 2
√
b) L′ = ||η|| = gij η i η j − length
Physics: action functional
1
L = ||η||2 − U (x)
2
(gravity, electric fields)
1 e
L = ||η||2 − e U (x) + Ai (x(t)) ẋi
2 c
- electric field Ei = − ∂U/∂x , magnetic field Bij = ∂Ai /∂xj − ∂Aj /∂xi .
i

Lecture 9. Geodesics and Calculus of vari-

ations. Length and Action functionals. Ex-
amples.
Geodesics: M n , (x1 , . . . , xn ) (local coordinate system), gij (x), gij (x) η i η j > 0
.
Length of path {x(t)} = γ

b ∫ b√
l(γ) = gij (x(t)) ẋi ẋj dt
a γ a

More general:

39
 “Lagrangian” L(x, η) : T ∗ (M n ) → R is given.
“Action functional” is given
∫
S{γ} = L (x(t), ẋ(t)) dt
γ

Examples:
1) L = gij (x) ẋi ẋj /2 = ||ẋ||2 /2
√
2) L′ = gij ẋi ẋj = ||ẋ||
3) L′′ = ||ẋ||2 /2 − U (x) (gravity or electricity)
4) L′′′ = ||ẋ||2 /2 + (e/c) Ai (x) ẋi - magnetic field (its vector-potential).
5) “Relativistic Particle” (?)
Geodesics: Either (1) or (2)
P. Fermat: gij (x) = δij /c2 (x)
“Variation” of path γ

γ + ϵ η(x(t)) = γϵ b
a γ
(locally it makes sense)
η = vector field along (γ) (tangent to M n ).
Variation of Action:
∫ b
S {γ + ϵ η} = L (x(t) + ϵ η(t), ẋ + ϵ η̇) dt
a

Requirement: vector field should satisfy to some

”boundary conditions”. At the firs step we take η(t)
vector fields η(t) is C ∞ and equal to zero near a b
the endpoints (a) and (b)
Extremal curve or critical point: (to find it necessary to solve following
equation for all boundary conditions at the endpoints)
dS
= 0
dϵ ϵ=0

for all η(t) (C ∞ and 0 near the endpoints).

40
Lemma (Euler - Lagrange).
Curve γ is extremal iff
( )
d ∂L ∂L
= ,
dt ∂ ẋi ∂xi

L = L(x, ẋ) , (x, ẋ) ∈ T ∗ (M n ) .

Proof. We have
∫ b
dS d
(γ + ϵ η) = L (x(t) + ϵ η(t), ẋ + ϵ η̇) dt =
dϵ ϵ=0 dϵ a
∫ b( ) ∫ b( )
∂L i ∂L i ∂L d ∂L
= i
η + i
η̇ dt = i
− i
η i dt
a ∂x ∂ ẋ a ∂x dt ∂ ẋ
which is true for all η(t) which are C ∞ and equal to zero near a and b .
Indeed, we have
∫ b ∫ b ( )b ∫ b
∂L i i d ∂L i ∂L d ∂L
i
η̇ dt = − η i
dt + η i
= − ηi dt
a ∂ ẋ a dt ∂ ẋ ∂ ẋ a a dt ∂ ẋi

because η i (a) = η i (b) = 0 .

Take now η = (0, . . . , 0, η i , 0, . . . , 0)

ηi =
Conclusion: We have the Euler - Lagrange System of ODE
( )
∂L d ∂L
=
∂xi dt ∂ ẋi

near t = t0 for all t0 ∈ (a, b) .

Lemma is proved.
Terminology:

∂L/∂ ẋi = pi = “Momentum”

∂L/∂xi = fi = “force”

ṗi = fi

41
 1
L = gij (x) ẋi ẋj ⇒ pi = gij ẋj
2

T ∗ (M n ) → T∗ (M n )
ẋ → p
velocity momentum
vector (covector)

Equation of Geodesics:
( )
∂L 1 ∂gij
ṗk = k
= k
ẋi ẋj
∂x 2 ∂x
√
Another form (for length) L′ = gij ẋi ẋj :
( )
d pk ∂L′ 1 ∂gij i j
√ = k
= √ ẋ ẋ
dt i
gij (x) ẋ ẋj ∂x 2 gij (x) ẋ ẋ ∂xk
i j

′
Conclusion. Let parameter
√ t for the length functional L is “natural”, i.e.
t = l(γ) (length), dt = gij (x(t)) ẋi ẋj . Then we have same equations
√
for both L and L′ because gij (x(t)) ẋi ẋj = 1 .
Corollaries.
1) Geodesics for Rn , gij = δij , are the b
straight lines. a
x0 =ct timelike
lightlike
2) Relativistic Particles: (x0 , x1 , x2 , x3 ),
gij = diag (1, −1, −1, −1) .
Let x0 = ct, spacelike
η 2 < 0 - “spacelike” vector,
x=x1
η 2 = 0 - “lightlike” vector,
η 2 > 0 - “timelike” vector.

42
 x(t)
b
Requirement:
For every real material object we have:
⟨ẋ , ẋ⟩ ≥ 0.
If the mass of the object > 0 then we have:
⟨ẋ , ẋ⟩ > 0 .

“Time which you lived” :

∫
1 b√ 1
τ = ⟨ẋ , ẋ⟩ dt = length (γ)
c a c
Let
⃗v = (ẋ1 , ẋ2 , ẋ3 ) , ⃗ = (ẋ1 /c, ẋ2 /c, ẋ3 /c)
w
Then √ √
⟨ẋ , ẋ⟩ = c 1 − w2

For a particle of mass m we put :

√
L = −mc ⟨ẋ , ẋ⟩

“Momentum”:
∂L m ẋi
pi = = √ , i = 1, 2, 3
∂ ẋi 1 − w2

Energy :
∂L
E = ẋi − L
∂ ẋi

Examples :
1) Geodesics
1
L = gij (x) ẋi ẋj ⇒ E = L
2
√
L′ = gij (x) ẋi ẋj ⇒ E ≡ 0

43
 2) Gravity or electric field
1 1
L′′ = gij (x) ẋi ẋj − U (x) ⇒ E = gij (x) ẋi ẋj + U (x)
2 2
3) Magnetic field
1 e 1
L′′′ = gij (x) ẋi ẋj + Ai (x) ẋi ⇒ E = gij (x) ẋi ẋj
2 c 2

Equations :
1) Geodesics
( )
1 ∂gij
ṗk = ẋi ẋj , pk = gkj ẋj
2 ∂xk
2) Gravity or electric field
( )
1 ∂gij ∂U
ṗk − ẋi j
ẋ = − , pk = gkj ẋj
2 ∂xk ∂xk
3) Magnetic field
( )
1 ∂gij e ∂Ai i e
ṗk − k
ẋi ẋj = ẋ , pk = gkj ẋj + Ak (x)
2 ∂x c ∂xk c
i.e. ( ) ( )
d ( ) 1 ∂gij e ∂Ai ∂Ak
gkj ẋj − i
ẋ ẋ j
= − ẋi
dt 2 ∂xk c ∂x k ∂xi

Magnetic field
∂Ai ∂Ak
Bik = k
−
∂x ∂xi

Lecture 10. Variational problem and geodesics

on Riemannian manifolds: Action Functional,
Lagrangian, Energy, Momentum. Conserva-
tion of Energy and Momentum.
M n , (x1 , . . . , xn ) , gij (x) .

44
“Lagrangian”: L(x, η) : T ∗ (M n ) → R .
∫
S{γ} = L (x(t), ẋ(t)) dt (Action)
γ

“Momentum”: pi = ∂L/∂ ẋi = ∂L/∂v i

∫ b( )
dS ∂L d ∂L
δ S (γ, η) → (γ + ϵ η) = i
− η i dt
dϵ ϵ=0 a ∂x dt ∂ ẋi

where η(t) represents “variation” of the path γ .

a η(t) b
Euler - Lagrange equation:
( )
d ∂L ∂L
δS = 0 ↔ =
dt ∂ ẋi ∂xi

Examples
1) L = gij (x) ẋi ẋj /2
√
2) L = gij ẋi ẋj
3) Physics

L = gij (x) ẋi ẋj /2 − U (x) + (e/c) Ai (x(t)) ẋi

↑ ↑
gravity, magnetic
electric field
field

“Energy”
∂L ∂L
E = ẋi i
− L = vi i − L
∂ ẋ ∂v

45
 Energy Conservation Law

Theorem 10.1. For the Euler - Lagrange System we have the following
conservation law:
dE
= 0
dt
(i.e. E is constant along the trajectories of the Euler - Lagrange System).
Proof.
( ) ( )
dE d i ∂L i ∂L i d ∂L ∂L i ∂L i
= ẋ − L = ẍ + ẋ − ẋ − ẍ =
dt dt ∂ ẋi ∂ ẋi dt ∂ ẋi ∂xi ∂ ẋi
[ ( ) ]
d ∂L ∂L
= − ẋi = 0
dt ∂ ẋi ∂xi
Theorem is proved.

Momentum Conservation Law

Theorem 10.2. Let L(x, v) does not depend on x1 :

L = L (x2 , . . . , xn , v 1 , . . . , v n )

Then we have: ṗ1 = 0 on the trajectories of the Euler - Lagrange System.

Proof. ( )
d ∂L ∂L
ṗ1 = = = 0
dt ∂ ẋ1 ∂x1
Theorem is proved.
Definition. Vector field ζ = (ζ i (y)) is called “Symmetry” of Lagrangian
if L(x, v) does not depend on x1 in the local coordinate system (x) where
ζ = (1, 0, . . . , 0) .
Corollary. In the original system (y 1 , . . . , y n ) we have ζ = (ζ 1 , . . . , ζ n )
The component pζ = pi ζ i is a conservative quantity ṗζ = 0 because pζ
is exactly the first component of p in the system (x1 , . . . , xn ) where ζ =
(1, 0, . . . , 0) .

46
Examples.
1) L = gij (x) ẋi ẋj /2
Energy
∂L
E = ẋi − L = L
∂ ẋi
Ė = 0 ⇒ - parameter along geodesics is NATURAL because
1
E = ||ẋ||2g = const
2
along trajectory.

2) Let a surface M 2 ⊂ R3 be invariant under

rotations around z - axis:

L = gij (x) ẋi ẋj /2 , n = 2 , x = (ρ, φ) , surface Φ(ρ, z) = 0 ,

∂L/∂φ = 0 .

“Angular Momentum”

∂L
pφ = = const − along geodesics
∂ φ̇
Geometrical meaning - later.

Consider now
1 [ 0 2 ] 1
L′ = a1 · (ẋ ) − (ẋ1 )2 − (ẋ2 )2 − (ẋ3 )2 = a1 · ⟨ẋ , ẋ⟩
2 2

47
 √ √
L′′ = a2 · (ẋ0 )2 − (ẋ1 )2 − (ẋ2 )2 − (ẋ3 )2 = a2 · ⟨ẋ , ẋ⟩ , t = x0 /c

Relativity
∂L′
Use L′ : E ′ = L′ , pi
(4)
=
∂ ẋi
time τ ∼ length by theorem above so
( 0 )
(4) ∂L′ dx dt dx1 dt dx3 dt
pi = = a1 · ,− , ..., −
∂ ẋi dt dτ dt dτ dt dτ
dt dτ 1 1
= 1/ = √ ∑3 = √
dτ dt
c2 − (v α )2 c2 − v 2
α=1

where ( )
dx1 dx2 dx3
v = , ,
dt dt dt

4-D Formalism
So we have for the 4-momentum
( )
1 dx1 /dt dx2 /dt dx3 /dt
p (4)
= a1 · √ ,− √ ,− √ ,− √
1 − w2 c 1 − w2 c 1 − w2 c 1 − w2
where w = v/c .
Clearly we have to choose: a1 = mc , where m is the mass of a
particle.
3-D Formalism
Use L′′ . We use time t = x0 /c , x0 = ct .
√
L′′ = a2 · (c2 − (ẋ1 )2 − (ẋ2 )2 − (ẋ3 )2

∂L′′ vα a2 vα
p(3)
α = = − a 2 · √ = − √
∂ ẋα c2 − v 2 c 1 − v 2 /c2
v α = dxα /dt , α = 1, . . . , 3 . We choose a2 = − mc .
So we finally have

∂L′′ mc2
E ′′ = ẋα − L′′
= √ , w = v/c
∂ ẋα 1 − w2

48
E ′′ = “physical energy” E . Let p0 = E/c . We have

p20 − p21 − p22 − p30 = m2 c2

- “MASS SURFACE” in the space R1,3

(Hyperbolic = Lobachevsky - Bolyai Space).

“Velocity” of particle

vi
v = (v 1 , v 2 , v 3 ) , √ = pi
1 − v 2 /c2

111111111111111111111
000000000000000000000
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
We have “Poincare” Model for 3D Hyperbolic Space 000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
(ball) 000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111

Homework 3.
1. Prove that group O(1, 1) consists of transformations:
( )
2 2 ch φ sh φ
P , P = 1 , T , T = 1 ,
sh φ ch φ
( )
1 0
U (1, 1) = Z2 × Z2 × R (topologically) , gij =
0 −1

O(2) ∼
= Z2 × S1 (topologically)

2. Prove equality SU (1, 1) = SL2 (R) .

( )
a b
SU (1, 1) = , |a|2 − |b|2 = 1 , SL2 (R) : {A, det A = 1}
b̄ ā

Use change of basis (e1 , e2 ) ↔ (e = e1 + ie2 , ē = e1 − ie2 ) .

49
3. Prove that SL2 (R) ∼
= R2 × S1 (topologically).
4. Calculate Euclidean Riemannian metric of R2 in polar coordinates.
5. Calculate Riemannian Metric of sphere S2 :

S2 : x2 + y 2 + z 2 = 1 , R3 : dl2 = dx2 + dy 2 + dz 2

in the spherical coordinates θ, φ :

z = r cos φ , x = r sin θ cos φ , y = r sin θ sin φ , (r = 1)

6. Calculate Riemannian Metric of “pseudosphere” (hyperbolic Lobachevsky

plane)

x0
H2 in R1,2

H2 : ρ2 = (x0 )2 − (x1 )2 − (x2 )2 = 1 , x0 > 0

x2
R1.2 : dl2 = (dx0 )2 − d(x1 )2 − d(x2 )2

x1

Lecture 11. Geodesics and Action Func-

tional. Examples. Hamiltonian form of Eu-
ler - Lagrange equation. Conservation of
Energy and Momentum. Euclidean, Spher-
ical and Hyperbolic Geometries.
Geodesics:
b ∫
S {γ} = L(x, v) dt , v = ẋ
a γ γ

50
 1
L = gij (x) ẋi ẋj , L(x, η) : T ∗ (M n ) → R
2
Euler - Lagrange Equations:
∂L
ṗi = , pi = gij v j
∂v i
∂gij k j 1 ∂gjk k j
gij (x) ẍj + k
ẋ ẋ = ẋ ẋ
∂x 2 ∂xi
( )
1 ∂gjk ∂gij
ẍj gij = i
− k
ẋk ẋj
2 ∂x ∂x
( )
1 ∂gjk ∂gij
p
ẍ = g pi
i
− k
ẋk ẋj
2 ∂x ∂x
Let us introduce the “Christoffel Symbols”
( )
1 pi ∂gij ∂gik ∂gjk
p p
Γkj = Γjk = g + −
2 ∂xk ∂xj ∂xi
We have then
ẍp + Γpkj ẋk ẋj = 0

Euclidean (Pseudoriemannian) Metric: Γpkj = 0 .

Examples:

1. R2 : dx2 + dy 2 ⇒ dz dz̄ = dr2 + r2 (dφ)2

dz dz̄
2. S2 : dθ2 + sin2 θ dφ2 ⇒
(1 + |z|2 )2

dz dz̄ dz dz̄
3. H2 : dχ2 + sh2 χ dφ2 ⇒ ⇒
(1 − |z| )
2 2 y2

Lemma 1. For the general metric of the form dl2 = dρ2 + Φ2 (ρ) (dφ)2
every line φ = const is geodesics.
Proof. We have
1 ( 2 )
L = ρ̇ + Φ2 (ρ) φ̇2
2
51
 Euler - Lagrange equations:
d ( 2 )
ρ̈ = ΦΦ′ (ρ) φ̇2 , Φ (ρ) φ̇ = 0
dt
So we have: φ̇ = 0 always satisfy to the Euler - Lagrange equations.
Lemma is proved.
Remark. Integrate Euler - Lagrange equations now:

Φ2 φ̇ = a = const ⇒ φ̇ = a/Φ2 (ρ)

and therefore: ρ̈ = ΦΦ′ (ρ) a2 /Φ4 (ρ) = a2 Φ′ (ρ)/Φ3 (ρ) ,

∂V (ρ) 1
so : ρ̈ = − a2 , V (ρ) = ,
∂ρ 2Φ2 (ρ)

∂V (ρ) ρ̇2 a2
so : ρ̈ ρ̇ + a2 ρ̇ or + = b = const
∂ρ 2 Φ2 (ρ)
We have then √
ρ̇ = 2(a2 /Φ2 (ρ) − b)
and
∫
dρ
t = √ ⇒ ρ = ρ(t) , φ̇ = a/Φ2 (ρ)
2(a2 /Φ2 (ρ) − b)

Conclusion. Every geodesics for the M 2 = {R2 , S2 , H2 } can be obtained

from the straight line φ̇ = 0 (or φ = const) by the action of the following
groups:
1) R2 : Iso (R2 ) − straight lines (Euclidean space).
2) S2 : SO3 − big circles (round sphere).
3) H : SO2,1 (= SL2 (R), SU (1, 1)?)
2
Isometry
- circles orthogonal to the boundary in group
the Poincare model: SU1,1
dz dz̄
|z| < 1 , dl2 =
(1 − |z|2 )2

52
 Isometry
group
Klein Model:
SL 2(R)
y = Im z > 0 , dl2 = dz dz̄/y 2 .
y=0
How to prove that they represent the same metric?
Geodesic Flow = Dynamical system in T ∗ (M n ) ∼
= T∗ (M n ) .

T ∗ (M n ) : ẋp = v p , v̇ p = Γpij xi xj

∂L
T ∗ (M n ) : ṗi = , ẋi = g ik (x) pk , (pi = gij (x) ẋj )
∂xi

Theorem. Let
∂L
E(x, p) = H(x, p) , E = ẋi − L
∂ ẋi
Then we have
∂H ∂H
ṗi = − , ẋi =
∂xi ∂pi

Proof. We change (x, v) → (x, p) . By the Euler - Lagrange equations we

have for E = v i pi − L(x, v(x, p)) :
( ) ( i) ( ) ( i)
∂E ∂v ∂L ∂v
1) = pi + v −j
i
= v j = ẋj
∂pj x ∂pj x ∂v x ∂pj x
( ) ( i) ( ) ( ) ( i)
∂E ∂v ∂L ∂L ∂v
2) = p i − − = − ṗj
∂xj p ∂xj p ∂xj v ∂v i x ∂xj p
Theorem is proved.
Definition. H(x, p) = “Hamiltonian” (function) (generator of time
dynamics).
Remark. Hamiltonian system is a “skew-gradient” flow in the inner product
of (x, p) - space ( )
0 I
gij = = − gji
−I 0

53
 2-form
∑
n
Ω = dxi ∧ dpi
i=1

is invariant under the flow (later).

v
H=const

Example. n = 1 , H = p2 /2 + V (x) ,
L = ẋ2 /2 − V (x) , p = v . x

dH
≡ 0 , (energy conservation)
dt

For geodesics:
1 ij 1
H = g (x) pi pj , L = gij (x) ẋi ẋj ,
2 2
(g ij ) = (gij )−1 , g ik gkj = δji , gij = gji .

Lecture 12. Curvature of curves and sur-

faces. How to differentiate tangent vector
fields?
Curvature.
1. Curvature of curves γ(t) ⊂ Rn . Let γ̇ ̸= 0 . Choose normal parameter
t = length s , γ → γ(s) , |γ| = 1 , γ̇ = τ .
Definition. |γ̈| = curvature

54
 n
τ
Special case n = 2 (oriented plane) :

(τ, n) - orthonormal oriented frame.

Statement.
dτ
= k ·n
ds
(k = curvature ).
Proof.
dτ
⟨τ , τ ⟩ = 1 ⇒ 2 ⟨τ , ⟩ = 0 ⇒ τ̇ ⊥ n ⇒ τ̇ = k · n
ds
Statement is proved.
Frené:
(τ (s) , n(s)) = A(s) (τ0 , n0 ) , A(s) ∈ SO2 ,
so
dA −1
A = B(s) = − B t (s)
ds
So we have:
( )
dA dA 0 k
= B(s) A(s) or = A(s)
ds ds −k 0
or
τ̇ = k n , ṅ = − k τ

2. Curvature of Hypersurfaces

M n ⊂ Rn (x0 , x1 , . . . , xn )

z
τ
Let M n is given by the equation

x0 = z = F (x1 , . . . , xn )

0
x1 ... xn

55
 Coordinate system is chosen in such a way that z ⊥ M n at the point
P : (x1 , . . . , xn ) = (0, . . . , 0) , so we have near P :
∂z ∂F
= = 0 , j = 1, . . . , n
∂xj P ∂xj (0,...,0)

Curvature form (“Second quadratic form”) kij dxi dxj :

∂2F
kij dxi dxj = dxi dxj
∂xi ∂xj
(at P only!)
It describes curvature of Normal Sections (curves) in Rn+1 , n ≥ 2 .
′′
Principal curvatures′′ = eigenvalues kj
( 2 )
′′ ′′ ∂ F
Gaussian Curvature = det = K
∂xi ∂xj
( 2 )
′′ ′′ ∂ F
Mean Curvature = Tr
∂xi ∂xj

K<0

convex K>0

Gauss Theorem. Gauss Curvature depends only on the Riemannian Metric

in M n . (Later).
Riemannian Metric for M 2 ⊂ R3 . Let z = F (x, y) , where (z, x, y) is
an orthonormal coordinate system in R3 .
dl2 = dz 2 + dx2 + dy 2 = (Fx dx + Fy dy)2 + dx2 + dy 2 =

56
 ( ) ( )
= 1 + Fx2 dx2 + 1 + Fy2 dy 2 + 2 Fx Fy dx dy
For the point P ∈ M n where z ⊥ M n we have gij (x) = δij .

Invariant form of characteristic Polynomial

det (kij − λ gij )

∑
Here (λ1 , . . . , λn ) - principal curvatures, λj - mean curvature,
λ1 . . . λn - Gaussian curvature, kij transforms as a quadratic form in
the tangent space.
Our Goal : Riemannian Curvature and Curvature in the vector
bundles.
Derivative of function
∑
d : f → df = fxi dxi

covector field:
(ηi ) = (fxi )
Vector field: ( i) d ( i)
ζ −
→ ∂i ζ
Transformation rule x = x(y)

∂xi
(x) ζ i = ζ ′j (y)
∂y j

Consider now
( i
)
i ∂ ′j ∂x ∂ ∂y p ∂
∂k ζ = ζ , where =
∂xk ∂y j ∂xk ∂xk ∂y p

We have
( )
∂ i ∂y p ∂ ′j ∂x
i
∂y p ∂xi ∂ζ ′j ∂y p ∂ 2 xi ′j
ζ = ζ = + ζ
∂xk ∂xk ∂y p ∂y j ∂xk ∂y j ∂y p ∂xk ∂y p ∂y j

57
Conclusion: Second derivative enters in the change of coordinates rule
( ) ( )
∂ i ∂ ′j ∂y p ∂xi ∂ 2 xi ∂y p ′j
ζ = ζ + ζ
∂xk ∂y p ∂xk ∂y j ∂y p ∂y j ∂xk

TENSOR LAW NONTENSOR LAW

For covectors:
( ) ( )
∂ ∂y p ∂ ′ ∂y
j
∂ ′ ∂y p ∂y j ∂ 2yj
(η i ) = ηj i = η + η′
∂xk ∂xk ∂y p ∂x ∂y p j ∂xk ∂xi ∂xk ∂xi j

Conclusion 1. Skew symmetric expression ∂k ηi − ∂i ηk transforms as a

tensor (inner product)
( ) ∂y p ∂y q
(x) (∂k ηi − ∂i ηk ) = ∂p′ ηq′ − ∂q′ ηp′ (y) , x = x(y)
∂xk ∂xi
or
∂xk ∂xi
(x) p q
(∂k ηi − ∂i ηk ) = ∂p′ ηq′ − ∂q′ ηp′
∂y ∂y
which makes sense even if number of (x)′ s (n) is not equal to number of (y)′ s
(m) !
(y) → (x)
f :
Nm → Mn
We call the corresponding tensor “differential 2-form”

hij (x) = − hji (x)

∂xi ∂xq
h′pq = hij , h → f ∗ (h)
∂y p ∂y q

h′ ← h
∗
f :
2 − form in N m 2 − form in M n

Conclusion 2. Quantities ∂i ηj and ∂k ζ p are not tensors. How to make

tensors?

58
“Covariant derivatives”

∇i η j = ∂i η j + Γjik η k

∇i ηj = ∂i ηj − Γkij ηk
where Γijk are “Christoffel Symbols”.
Requirements: ∇i η j and ∇i ηj are tensors.

Difficulty: ∇i ∇j ̸= ∇j ∇i , ∇i ∇j − ∇j ∇i = R̂ij (“Curvature”).

More general (local picture): consider Ψ(x) ∈ RLx , where RLx is a linear
space with
∑ basis (e1 (x), . . . , eL (x)) depending on the point (x1 , . . . , xn ) ,
i
Ψ(x) = Ψ ei (x) . Consider the set of linear equations

∂Ψi /∂xj = Aijk Ψk (x1 , . . . , xn )

Can we solve this system?

Let us define Operators of “Covariant Derivatives”
∂
∇j = j
− Aijk (x)
∂x

Lemma. The set of linear equations (above) is solvable for all “initial
data” Ψ(x0 ) = Ψ0 if and only if

∇i ∇j − ∇j ∇i = Rij = 0

for all x ∈ Rn .
Proof. For every solution Ψ(x) we have

∂i Ψ = Ai Ψ (Ai = matrix)

and
∂i ∂j Ψ = ∂i (Aj Ψ) = ∂j (Ai Ψ)
or
(∂i Aj − ∂j Ai ) Ψ + Aj (∂i Ψ) − Ai (∂j Ψ) = 0
or
(∂i Aj − ∂j Ai − Ai Aj + Aj Ai ) Ψ = 0

59
where
(∂i Aj − ∂j Ai − Ai Aj + Aj Ai ) = Rij (matrix)
So we have Rij Ψ = 0 for all x ∈ Rn . Therefore Rij ≡ 0 .
Lemma is proved.

Rij = [∇i , ∇j ]

Examples.
a) L = n , Ψ - vector field, Aijk = − Γijk .
b) L = n , Ψ - covector field, Aijk = Γijk .
c) L = 1 , Ai - scalar values, Ai Aj − Aj Ai = 0 ,

∂i Aj − ∂j Ai = Rij

Lecture 13. Vector bundles. Connection

and Curvature. Parallel transport.
Vector bundles and Curvature.
Vector bundle = Family of Linear Spaces RN x depending on parameter
x ∈ X and “locally trivial”:
For every point x ∈ X there exists an open set U ∋ x such that we can
choose a basis [eU1 (x), . . . , eUN (x)] ∈ RN
x continuously depending on x ∈ X .
For x ∈ U ∩ V we have

eUi (x) = aji U V (x) eVj (x)

Real case: aij ∈ GLn (R)

Complex case: aij ∈ GLn (C)
Orthogonal case: aij ∈ On
Unitary case: aij ∈ Un
- For all pairs U , V (!).

60
Other groups are also possible and define the “structural group” of the
bundle.
Our case: X is a C ∞ -manifold and aij (x) are C ∞ -functions.

Example. Let X = M k ⊂ RL . Consider

family of tangent k-spaces to M k .

Map G : M k → Gk,L (Grassmann Manifold).

Curvature is local quantity.
Locally vector bundle is given by the product U × Rn according to the
choice of the basis e1 (x), . . . , en (x) in Rn .
Consider set of linear ODE’s
∂Ψj j
∑
p
= A ip (x) Ψ (x) , Ψ = Ψj (x) ej
∂xi
i = 1, . . . , n , j, p = 1, . . . , N .
or:
∇i Ψj = ∂i Ψj − Ajip Ψp = 0

Is this system solvable?

For n = 1 it is true.
For n > 1 it may be wrong! “Curvature” Rij of this “Differen-
tial Geometric Connection” Ajip (x) should be equal to zero:
[ ]
∂ ∂
[∇i , ∇q ] = ∇i ∇q − ∇q ∇i = − Aip (x) ,
j
− Aql (x) = 0
j
∂xi ∂xq

As we saw in the previous lecture, we can formulate the following Lemma:

Lemma. The set of linear equations (above) is solvable for all “initial
data” Ψ(x0 ) = Ψ0 if and only if
∇i ∇j − ∇j ∇i = Rij = 0

61
for all x ∈ Rn .

Rij = [∇i , ∇j ]

Examples.
a) L = n , Ψ - vector field, Aijk = − Γijk .
b) L = n , Ψ - covector field, Aijk = Γijk .
c) L = 1 , Ai - scalar values, Ai Aj − Aj Ai = 0 ,

∂i Aj − ∂j Ai = Rij

“Gauge Transformations”, Curvature

We have ∂i Ψ = Ai Ψ (Matrix Form), Ai = Akij (x) , so

∇i = ∂i − Ai , ∇q = ∂q − Aq ,

∂Aq ∂Ai
[∇i , ∇q ] = Riq = − + + Ai Aq − Aq Ai
∂xi ∂xq
(Matrices in RN ) .
1) For tangent (cotangent) case N = n .
2) For scalar case N = 1 , n is any.
3) For euclidean case ⟨ei , ej ⟩ = δij we should have (Ai )kj = − (Ai )jk
(skew symmetry).
Change of Basis:
ei = bij (x) e′j , [e = B e′ ]
or
Ψi ei = Ψi bji e′j , Ψ′j = Ψi bji

⃗ ′ = B(x) Ψ
Ψ ⃗

Lemma 2. Let
∂Ψj
= Ajik Ψk (x)
∂xi

62
 Then for the new basis (e′j ) we have

∂Ψ′j
i
= A′jik Ψ′k (x)
∂x
where
∂B −1
A′i = B Ai B −1 + B
∂xi

Proof. We have Ψ′ = B(x) Ψ , so

∂Ψ′j ∂B ∂B −1 ′
= B Ai Ψ + Ψ = B Ai B −1 Ψ′ + B Ψ
∂xi ∂xi ∂xi
Lemma is proved.
Conclusion.
∂B −1
A′i = B Ai B −1 + B
∂xi

Let B −1 = G(x) , we have then

∂G
A′i = G−1 Ai G − G−1
∂xi
- Gauge Transformation.

Homework 4.
1. Prove that the group O(1, 1) (connected component) can be written in
the form ( ) ( )
√ 1 √ w
ch ψ sh ψ 1−w2 1−w2
= √ w √ 1
sh ψ ch ψ 1−w2 1−w2

2. a) Prove that every isometry of R2 , preserving orientation, is either shift

or rotation around some point.
b) Prove that every isometry of R2 , inverting orientation, is product of
reflection and shift along this line (line of reflection).
3. How many local coordinate systems are needed to cover RP2 ? Find cover
by 3 domains.

63
4. Find cover of sphere S2g with g handles by 2 systems of local coordinates.

2
Sg =

5. Introduce “pseudospherical” coordinates

x0 = ρ ch θ , x1 = ρ sh θ cos φ , x2 = ρ sh θ sin φ
Restrict metric
(dx0 )2 − (dx1 )2 − (dx2 )2
on the “pseudoshere” ρ = 1 . Calculate it.
6. Prove that the following metrics are equivalent:
dz dz̄
a) |z| < 1 , 2
dl(1) =
(1 − |z|2 )
2 dw dw̄
b) Im w > 0 , dl(2) =
(Im w)2
1111111111111111111111
0000000000000000000000
0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
111111111111
000000000000 0000000000000000000000
1111111111111111111111
000000000000
111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
000000000000
111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
000000000000
111111111111 (P) 0000000000000000000000
1111111111111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
0000000000000000000000
1111111111111111111111
000000000000
111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
000000000000
111111111111 0000000000000000000000
1111111111111111111111
000000000000
111111111111
(K)
Use transformation
az + b
w =
cz + d
Find a, b, c, d such that ball maps into upper plane.
7. Find isometry groups for P and K in the form
az + b
z →
cz + d
Prove that
( ) ( )
a b a b
∈ SU (1, 1) for P and ∈ SL2 (R) for K
c d c d

64
 Lecture 14. Vector bundles. Connection
and Curvature. Parallel transport.
Vector bundle=“locally trivial” family of vector spaces RN x , x ∈ X with
“local bases” [e1 (x), . . . , eN (x)], x ∈ U ⊂ X, ∀x ∃U ∋ x.
U U

For x ∈ U ∩ V
eUi (x) = aj,U
i
V
(x)eN
j (x).

“Group” aj,U
i
V
(x) ∈ G ⊂ GLN (R)
Differential geometrical connection (on vector bundle).
Locally matrix-valued functions Ai (x) = (Ai )jk (x) are given for all x ∈
U ⊂ X, and operators ∇Ui = ∂i − AUi acting on functions ΨU : U → RN ,
ΨU = (Ψj,U ),
∇Ui ΨU = ∂i ΨU − AUi ΨU .
In intersection x ∈ U ∩ V we have

ΨU = g U V ΨV , G(x) = g U V (x),

such that:
AVi = G−1 (x)AUi G(x) − G−1 ∂i G
(gauge equivalence).
“Curvature”
U
Rij = ∇Ui ∇Uj − ∇Uj ∇Ui (matrix functions).

Theorem (Later):
V
Rij = G−1 Rij
U
G (!)
ξ-tangent vector to X = M n in the point x ∈ M n , x1 , . . . , xn .
( U )
∂Ψ
∇ξ =
U
i
− Ai Ψ ξ i
U
∂x

– covariant derivative along ξ ∈ TX∗ .

Curve γ = {xi (t)}. Covariant derivative along γ:

65
 ∇Uγ = ∇Uẋ(t) = ẋi ∇Ui ΨU .
xi (t) γ (t)
x0
Definition. “Parallel” vector field along the line γ(t):
∇Uẋ(t) ΨU (x(t)) ≡ 0 for all t.
“Parallel transport” •+ vector ΨU (x0 ) = Ψ0 along the line γ(t), γ(0) =
x0 .
γ Results for γ1 and γ2 may be
different!
x0 γ’

Geodesics: N = n, vector bundle is tangent vector bundle to X = M n ,

Rnx = Tx∗ , e1 , . . . , en – standard basis in tangent spaces in local coordinates
ei ≡ ∂/∂xi for U ⊂ X, X = M n .
Definition. γ(t) = {xj (t)} is “geodesic curve” if ∇Uẋ(t) ẋ(t) ≡ 0.

Lemma 1. Curve γ(t) = {xj (t)} is geodesic iff the following equation is

true:
ẍj + Γ̃jik ẋi ẋk = 0,
where Γ̃jik = −Ajik (x) (Connection).
Proof. By definition, we have
∇Uẋ(t) ẋ(t) = ẋk ∇Uk ẋ(t) = ẋk ∂k ẋj (t) − ẋk Ajik (x)ẋi (t) = ẍj (t) + Γ̃jik ẋi ẋk = 0.
| {z }
Geodesic equation

“Variational geodesics.”
ẍj + Γjik ẋi ẋk = 0
we need
1
Γkij = − (Akij + Akji ) − symmetric part
2
for these 2 equations coincide. Calculating variation, we obtain:
[ [ ]] [ ]
1 ∂gij ∂giq 1 lq ∂gij ∂giq ∂gqj
Γij = Symmetrization of g −
l lq
+ = g − q + + .
2 ∂xq ∂xj 2 ∂x ∂xj ∂xi

66
Calculation of Connection and Curvature: (it is linear operation)

1. Covariant derivative of scalars is trivial:

∇ξ f (x) = ξ i ∂i f.

2. Covariant derivatives of vectors, covectors and tensors like inner prod-

ucts are defined:

∇i ξ j (x) = ∂i ξ j + Γjik ξ k vectors,

∇i ηj (x) = ∂i ηj + Γ̃ρij ηρ covectors,

∇i tjk (x) = ∂i tjk + Gpl
ijk tpl inner products of vectors,

such that

a) ∇i (ξ j ηj ) = ∇i (ξ j )ηj + ξ j ∇i (ηj ) = ∂i (ξ j ηj ) (scalars).

b) ∇i (ξk ηl ) = ∇i (ξk )ηl + ξk ∇i (ηl ) (product of two covectors).

c) ∇i gkl ≡ 0, gkl =Riemannian Metric.

Theorem. There exist unique symmetric connection such that

− Γ̃ijk = Γijk (1)

∇i tkl = ∂i tkl − −
Γpik tpl Γpil tkp (2)
[ ]
1 is ∂gjs ∂gsk ∂gjk
i i
Γjk = Γkj = g + − (3)
2 ∂xk ∂xj ∂xs

Proof of (1): we have (a):

∇i (η k ξk ) = ∂i i(η k ξk ) = (∂i η k + Γkis η s )ξk + η k (∂i ξk + Γ̃sik ξs ), ⇒ Γ = −Γ̃.

Proof of (2): we have “Leibnitz”(b):

∇i (ηl ξk ) = (∂i ηl + Γ̃sik ηs )ξk + ηl (∂i ξk + Γ̃sik ξs ),

so we see that (2) is true for the “products” (ηl ξl ), so it is true for linear
combinations of products and so it is true fro all tensors of the type (tkl ).
Proof of the formula for Γkij – next lecture.

67
 Lecture 15. Connections in tangent bundle.
Curvature. Ricci curvature. Einstein equa-
tion. Spaces of constant curvature.
Differential-geometrical connections in tangent bundle T ∗ (M n ), x1 , . . . , xn .

Aijk (x) = Γijk (x)

∇i η j (x) = ∂i η j + Γjik η k , (vectors)
∇i ηk (x) = ∂i ηk + Γ̃jik ηj , (covectors)
∇i tjk (x) = ∂i tjk + Gplijk tpl , (inner products - tensors)
Axioms:
1. ∇i f (x) = ∂i f (x), scalars.
2. ∇i (ξ j ηj ) = ∇i (ξ j )ηj + ξ j ∇i (ηj ).
3. ∇i (ξk ηl ) = ∇i (ξk )ηl + ξk ∇i (ηl ).
4. ∇i gkl ≡ 0, (gkl =Riemannian Metric).
Items 2-3 are Leibnitz rule.
Definition. Connection in T ∗ (M ) is “symmetric” if Γkij (x) = Γkji (x) (“torsion”=0).
“Torsion tensor” = Tijk (x) = Γkij (x) − Γkji (x).
Theorem. There exists a unique differential-geometrical connection on
T ∗ (M ) extended to covectors and tensors as above, symmetric and com-
patible with Riemannian metric ∇i gkj = 0.
Proof.
Step 1: Prove that Γ̃kij = −Γkij (follows from the Axioms 1 and 2).
Step 2: Prove that ∇i tjk = ∂i tjk − Γsij tsk − Γsik tjs .
Proof. From the ∑ Step ∑= −Γ.
1 we have Γ̃
(p) (p)
From the Axiom 3 we have
(p) (p)
our result for ∇i ( P ξk ηj ), tkj = P ξk ηj . Every tensor tkj can be
presented in that form (may be as a series).
Step 3: From the condition ∇i gkj = 0 we have
∂i gkj = Γsij gks + Γsik gsj

68
Let us solve the system of linear equations for the triple (ijk) using condition
Γkij = Γkji (3 equations for 3 unknown quantities Γkij = Γkji ) or Γij,k = gks Γsij .
Solution:
1
Γij,k = (∂j gik + ∂i gkj − ∂k gij ) .
2
Our equations are:
∂i gkj = Γij,k + Γik,j
u = ∂i gkj , v = ∂k gij , w = ∂j gki ,
u−v+w
a = Γij,k , ⇒ a = ,
2
1
Γij;k = (∂i gkj − ∂k gij + ∂j gki ) .
2
1
Γkij = g ks Γij;s = g ks (∂i gsj + ∂j gis − ∂s gij ) .
2
For the case Γkij = Γkji , gij = gji we obtain the same formula as for geodesics
(for Calculation of Variations)

ẍk + Γkij ẋi ẋj = 0.

Curvature.

(Rkp )ij = ∇i ∇j − ∇j ∇i = R̂ij , (where) R̂ij are matrices.

∂Γj ∂Γi
∇i ∇j − ∇j ∇i = [∂i + Γi , ∂j + Γj ] = − + Γi Γj − Γj Γi .
∂xi ∂xj | {z }
product of matrices

Example. Consider special coordinates for M 2 ⊂ R3

(z ⊥ M 2 , x, y – local coordinates near x0 = (0, 0)),
z
x,y
δij = gij (0), z = F (x, y), gxx = 1 + Fx2 ,

gxy = Fx Fy , gyy = 1 + Fy2

We have finally:

gij (0) = δij , ∂i gkl (0) = 0, ⇒ Γijk (0) = 0,

69
 ∂Γkij
=? Calculation is required.
∂xl
Conclusions from calculations (later):
s
1. Rij;kl = gis Rj;kl = Rkl;ij .

2. Rji;kl = −Rij;kl = Rij;lk .

For n = 2 we have: i, j, k, l = 1, 2. The whole curvature tensor Rij,kl can
be defined by 1 scalar function R (why?)
k
Rij = Ri,kj Ricci Curvature

R = Rii , Rji = g is Rsj , Scalar Curvature

Gauss Theorem. R/2 = Gaussian Curvature of the surface M 2 ⊂ R3

(Calculation later).
For n = 2 and M 2 = S 2 , R2 , H 2 we have R = const.
For n = 3 we have: Ricci Curvature Rij completely determines the whole
tensor Rij,kl . Why? They have the same number of components.
Curvature of Rn , S n , H n .

 0,
R= > 0, ?

< 0.

Curvature of conformally Euclidean metric gij = ϕ2 (x)δij .

What is “Curvature of 2-directions”?
Einstein equation: n = 4, gij -indefinite.
1
Rij − Rgij = λgij , no matter, only gravity.
2
“cosmological constant” λ ̸= 0.
Curvature of metrics in the compact groups like SOn , Un , . . . ?
What does it mean – “Constant curvature”?

∇i Rij,kl ≡ 0 “locally symmetric spaces”.

Compact groups, in particular.

70
 Appendix to Lecture 15.

1
Γkij = g ks Γij;s = g ks (gsj,i + ∂j gis,j − gij,s ) .
2
Our notation:
∂f /∂xk ≡ f, k
∂ Γ̂q ∂ Γ̂i
R̂iq = − + Γ̂i Γ̂q − Γ̂q Γ̂i , where Γ̂i = Γkij .
∂xi ∂xq
Our system:

gij (0) = δij , gij,p (0) = 0, g ij (0) = δij = gij (0), g ij ,p (0) = 0

R̂iq = Rjk;iq = Rj k;iq , (x, y = 0, 0).

∂Γjqk 1
i
= (gjk,iq + gqj,ik − gqk,ij )
∂x 2
∂Γjik 1
q
= (gjk,iq + gij,qk − gik,qj )
∂x 2
2Rjk;iq = gjk,iq + gqj,ik − gqk,ij − gjk,iq − gij,qk + gik,qj =
= gqj,ik − gqk,ij − gij,qk + gik,qj
(All formulas are valid at the point 0,0)

Lecture 16. Tensor fields. Curvature as a

tensor field. Gaussian curvature for surfaces
in R3 .
• What is a tensor field?

• Calculations: Curvature is a tensor field.

• Calculation of Curvature in the special coordinates.

• Algebraic properties of Curvature Tensor.

• Gauss Theorem for M 2 ⊂ R3 .

71
 • Curvature Tensor at the most symmetric spaces Rn , S n , H n .

What is a Tensor? (Scalars, vectors, covectors, inner products, Rieman-

nian Curvature.)
M n , x1 , . . . , xn -local coordinates. Tensor field of the type (k, l) is defined
by components:

Tji11,...,j
,...,ik
l
(x), dim = nk+l , Vector bundle Ψ : M n → Tensors,

such that for x = x(y) we have change

( )( )
∂xj1 ∂y p1
T́qp11,...,q
,...,pk
(y) = Tji11,...,j
,...,ik
(x(y)) · ... · ...
l l
∂y q1 ∂xi1

Examples: T i (x) – vector field.

∂y p
T́ p (y) = T i (x(y)) ,
∂xi
∂xj
T́q (y) = Tj (x(y)) ,
∂y q
Inner products:
∂xi ∂xj
ǵpq = gij .
∂y p ∂y q
Linear operators:
∂y p ∂xj
ápq = aij .
∂xi ∂y q
Inner product of covectors:
∂y p ∂y q
ǵ pq = g ij .
∂xi ∂xj

(1) (2)
Operations: linear (sum), product of tensors T IJ · T PQ = TJQ
IP
, permutation
of indexes: gij → gji .
...i...
∑ ...i...
“Trace”: T...i... = i T...i... .
Calculations.

Ψ = GΨ′ , ∇i = ∂i + Γi , ∇′i = ∂i + Γ′i , Γ′i = G−1 Γi G + G−1 ∂i G.

72
 Rij = ∇i ∇j − ∇j ∇i = ∂i Γj − ∂j Γi + Γi Γj − Γj Γi .

′
Theorem: Rij = ∇′i ∇′j − ∇′j ∇′i = G−1 Rij G.
Proof: We have ∂i G−1 = −G−1 ∂i G G−1 , because ∂i (G−1 G) = 0 = ∂i G−1 G+
G−1 ∂i G = 0. O.K.

∂i Γ′j = ∂i (G−1 Γj G + G−1 ∂j G) = −(G−1 (∂i G)G−1 Γj G)+

+G−1 ∂i Γj G + G−1 Γj ∂i G − G−1 ∂i G G−1 ∂j G + G−1 ∂i ∂j G,

Γ′i Γ′j = (G−1 Γi G + G−1 ∂i G)(G−1 Γj G + G−1 ∂j G) =
= G−1 Γi Γj G + G−1 (∂i G)G−1 Γj G + G−1 Γi ∂j G + (G−1 ∂i G)(G−1 ∂j G)
′
Conclusion: R̂ij = G−1 R̂ij G.
i
Corollary. Rj;kl is a tensor for tangent bundle T ∗ (M n ), R̂kl = Rp;kl
s
.
Gauss Curvature:
z
z = F (x, y),
g11 = 1 + Fx2 , z ⊥ x, y,
g12 = 1 + Fx Fy , x,y Fx = Fy = 0 at 0, 0.
g22 = 1 + Fy2 ,
x=x1 ,y=x2
“Curvature form” (x = 0, y = 0), matrix gij (0, 0) = δij = g ij (0, 0).

( )
Fxx Fxy
K = det = Fxx Fyy − Fxy
2
, x = 0, y = 0.
Fxy Fyy
Calculate Riemannian Curvature: R12;12 – only this component is non-
trivial.
2 2 2
g12;12 = Fxx Fyy + Fxy , g11;22 = 2Fxy , g22;11 = 2Fxy , x, y = 0, 0.

Using formulas for Rij;kl (0, 0) in these coordinates (See Appendix to the
previous Lecture), we obtain the following results:
p
R = 2R12;12 = 2K, Rij = Ri;pj , R = Rpp .

Results are valid in all coordinates.

73
Consider a pair of vector fields ξ, η. Define ∇η = η i ∇i .
Curvature:
R̂ηξ = ∇η ∇ξ − ∇ξ ∇η − ∇[η,ξ] ,
[η, ξ] = [η i ∂i , ξ j ∂j ] = (η i ∂ξ j /∂xi − ξ i ∂η j /∂xi )∂j .
We have
R̂η,ξ = R̂ij η i ξ j (Matrix)
for any vector bundle.
Quadratic Form (tangent bundle, symmetric connection)

Rij;kl = Rkl;ij = −Rji;kl − Rij;lk ,

< ij >= − < ji > – basic vectors in ∧2 Rn .

Curvature along 2-direction: let η, ξ be unit orthogonal vectors (in the Rie-
mannian Metric)
|η| = 1, |ξ| = 1, < η, ξ >= 0.
Then the sectional curvature is:

Rij;kl η k ξ l η i ξ j = R < η ∧ ξ, η ∧ ξ > .

| {z }
Sum by k,l

Metrics: S n , Rn , H n .
Quadratic form Rij;kl is determined by one constant R. All curvatures if
all 2-directions are the same at all points:

• Positive for S n .

• Negative for H n .

• 0 for Rn .

Symmetry Groups are On+1 for S n , Iso(Rn ) and O1,n for H n . Dimension
of these groups is n(n + 1)/1. Any points can be mapped to any point
(homogeneous space). Any unit tangent vector cam be mapped to any unit
tangent vector.

74
 Every pair (x, v) can be mapped to every pair (y, w), where x, y are
points, v, w are tangent vectors at the points x, y respectively, |v| = |w| = 1.

Remark. Group On acts homogeneously on the space Vn,k of orthonormal

k-frames (τ1 , . . . , τk ) in Rn , τi ⊥ τj , |τi | = 1.
Vn,k =“Stiefel Manifold”, Vn,1 = S n−1 , Vn,n = On .

Homework 5.
1. Prove that RP2 \point is diffeomorphic to MOBIUS BAND.
b

2. Prove that KLEIN BOTTLE a a is the same as

RP2 # RP2 , where # is “connected sum”

3. Prove that metric of Sn , Rn , Hn can be written in the form

a) dl2 = dρ2 + sin2 ρ (dΩ)2 , Sn ,

where (dΩ)2 is the metric of Sn−1 ,
b) dl2 = dr2 + r2 (dΩ)2 , Rn ,
c) dl2 = dχ2 + sh2 ρ (dΩ)2 , Hn .

4. Introduce “conformal coordinates” in Sn

x0 N

x 1... x n
0

75
 Prove that
∑ ∑
n
2 (dxi )2 2
dl = , r = (xi )2
(1 + r2 )2 i=1

5. Prove that straight lines passing through center are geodesics for the
metric of S2 :

dz dz̄
dl2 = 4
(1 + |z|2 )2

Homeworks 2, 3, 4. Solutions.
Homework 2. Solutions.
1. Projection Coordinates for Sn ⊂ Rn+1 . We need 2(n + 1) coordinate
domains

n=1

2. RPn : (x0 , . . . , xn ) ∼ (λx0 , . . . , λxn ) , λ ̸= 0 ,

{ }
Uj = xj ̸= 0 ⇒ (x0 , . . . , 1, . . . , xn )

- local coordinates, Uj = Rn .
3. SO3 = RP3 : axis + angle ≤ π .

76
 11111111111111
00000000000000
π
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111 (φ = π) ∼
= (φ = − π) .
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
−π11111111111111
00000000000000
00000000000000
11111111111111
4. SO1,2 : B̃ij = − B̃ji , Bik = g ks B̃si ,
 
1 0 0
g ks = 0 −1 0 
0 0 −1

So we have Bik = product of (g ks ) by skew symmetric matrix B̃si .

5. O1,1 ∋ T, P, P T, 1 (in different connected components). We have T 2 =
P 2 = 1 , P T = T P − Z2 × Z2 . Component of 1 is
( )
ch ψ sh ψ
sh ψ ch ψ

6 - 7. Quaternions R4 ∋ q

SO3 : q → q1 q q1−1 , |qj | = 1 , j = 1, 2

SO3 : q1 and q2 − any |qj | = 1

8.
a) GLn (R) has 2 components only.
b) GLn (C) is connected.
(b) Proof.
Set of matrices with distinct eigen-values is dense. Write linear operators
in basis of eigenvectors for GLn (C):
 
λ1 . . . 0
 .. . . .
. . ..  , λj ̸= 0 , λj ∈ C\0
0 . . . λn

it is connected space.
(b) is proved.

77
(a) Proof.
Let our field is R . Choose the same type basis consisting of 1-dim blocks
if λj is real or 2-dim blocks (λ, λ̄) for complex λ. We have for the forms of
blocks
λj ∈ R − (λj )
( )( )
cos bj sin bj ±1 0
λj ∈ C − a j
sin bj sin bj 0 1
Deform λj → ±1 , aj → ±1 , bj → 0 .
Now remember that ( )
−1 0
0 −1
is connected with 1

A
ξ−
→ −ξ
ξ ∈ R2

Multiplying our matrices by

( )
−1 0
A =
0 −1

many times we see that there are 2 components ( det > 0 , det < 0) .
(a) is proved.

Homework 3. Solutions.

1. Already was solved in HW2.

2. ( )
a b
SU (1, 1) , , ⟨A ψ , A φ⟩ = ⟨ψ , φ⟩
c d
vectors (a, b) = ζ , (c, d) = η

⟨ζ , ζ⟩ = 1 , ⟨η , η⟩ = − 1 , ⟨ζ , η⟩ = 0 ⇒

78
 |a|2 − |b|2 = 1 , |c|2 − |d|2 = − 1 , a c̄ − b d¯ = 0

solution: ( )
a b
, a ā − b b̄ = 1
b̄ ā

SL2 (R) : e1 , e2 → a e 1 + b e2 , c e 1 + d e 2
ad − bc = 1 , a, b, c, d ∈ R
Take basis
e = e1 + i e2 , ē = e1 − i e2
e1 = (e + ē)/2 , e2 = (e − ē)/2i
In the new basis we have
1 [( ) ( ) ]
e → a + d − i (b − c) e + a − d + i (b + c) ē = q e + p ē
2
1 [( ) ( ) ]
ē → a − d − i (b + c) e + a + d + i (b − c) ē = p̄ e + q̄ ē
2
where
1( )
q q̄ − p p̄ = (a+d) + (b−c) − (a−d) − (b+c)
2 2 2 2
= ad − bc = 1
4

4. SL2 (R) ∼
= S1 × R2 (?)
( ) ( )
cos φ sin φ λ µ
A ∈ SL2 (R) ⇒ A = ×
− sin φ cos φ 0 1/λ

where λ ∈ R+ \0 ∼
= R, µ ∈ R.
Remark. Every “semisimple” Lie Group is topologically a product:
G ∼= K × RN , where K is a “compact group” and RN is a “Borel
subgroup” (upper triangle).
4.
dl2 = dθ2 + sin2 θ (dφ)2 , S2 ⊂ R3
For polar coordinates in R2 : x = ρ cos φ , y = ρ sin φ we have

dl2 = dρ2 + ρ2 (dφ)2

79
5.
H2 : z 2 − x2 − y 2 = 1
z = ch θ , x = sh θ cos φ , y = sh θ sin φ
− dl2 = dθ2 + sh2 θ (dφ)2

Homework 4. Solutions.
1. We have parametrization (physics)
1 w
ch ψ = √ , sh ψ = √
1 − w2 1 − w2
because
(ch ψ)2 − (sh ψ)2 = 1

Remark. Physics: let w = v/c .

Lorentz transformation:
1 v/c
x0 = c t = √ x′0 + √ x′1
1 − v /c
2 2 1 − v 2 /c2

v/c 1
x1 = √ x′0 + √ x′1
1 − v /c
2 2 1 − v /c
2 2

So we have
1 v/c2
t = √ t′ + √ x′
1 − v /c
2 2 1 − v /c
2 2

v 1
x = √ t′ + √ x′
1 − v /c
2 2 1 − v /c
2 2

Note, that we get the Galilean Transformation

t ≃ t′ , x ≃ x′ + v t′

in the case v/c ≪ 1 .

2. a) Iso (R2 )+ ∼
= shifts + rotations SO2 .

1 → SO2 → Iso (R2 )+ → shifts = R2

80
 b) Iso (R2 )− ∼
= reflection + shift along the reflexion line .
B’
A

A’
B

3. Already was solved for RPn : ⃗x ̸= 0 , (x0 , . . . , xn ) ∼ λ (x0 , . . . , xn ) ,

(n + 1) systems :
Ui : (x0 , . . . , 1, . . . , xn )

4. S2g ⊂ R3 Two such domains for T2 , g = 1 .

Two such domains for any g

(glue them along the boundary
strips).

Lecture 17. Differential forms.

Last chapter.

1. Geodesics, Calculus of Variations, Fermat Principle, Lagrangian, Ac-

tion Functional (Length and “Kinetic Energy”=Natural Parameter),
Euler-Lagrange equations, Momentum, Energy, Conservation Laws,
Examples.

2. Curvature of curves, Curvature of Hypersurfaces in Rn+1 , Quadratic

(2nd) Form, Principal Curvatures and Gaussian Curvature: definition
via the special coordinates.

81
 3. Vector Bundles and Differential-Geometrical (Linear) Connection, Cur-
vature, Gauge Transformation for Connection and Curvature.
Parallel Transport.

4. Tangent Bundle and “Christoffel Symbols”, Cotangent Bundle, Ten-

sor Bundles (Inner Products), Geodesics (new definition), compatibil-
ity with Riemannian Metric. Symmetric Connections. Formulas for
Christoffel Symbols and Riemannian Curvature. Special Coordinates.

5. Symmetries of the Riemannian Curvature Tensor, Ricci Tensor and

Scalar Curvature. Examples. Einstein Equations (n = 4). Curvature
for n = 2. Gauss Theorem. Curvature for n = 3. Curvature of Rn , S n ,
H n.

Next chapter. Differential forms.

• m = 0: Differential 0-form is a scalar function

f (x) : M n → R.

• m = 1: Differential 1-form is a covector field ω written in the form

∑
ω= ωi (x)dxi .

• m = n Differential n-form = object of integration:

Ω = f (x)dx1 ∧ . . . ∧ dxn , (locally)

such that for x = x(y) we have

∂xi j
Ω = f (x(y))dx1 ∧ . . . ∧ dxn , dxi = dy , (summation in j)
∂y j

Definition. Differential k-from in M n is a smooth quantity Ω which locally

in every Chart of Atlas (x1 , . . . , xn ) can be written in the form
∑
Ω= fI (x) |dxi1 ∧ .{z . . ∧ dxi}k = fI dxI ,
I dxI

82
where I = (i1 < i2 . . . < ik ), and for x = x(y) we have

Ω = fI (x(y))dxi = gJ (y)dy J , dy J = dy j1 ∧ . . . ∧ dy jk , j1 < j2 < . . . < jk .

The symbols dxi form an Associative Algebra under multiplication ∧

(exterior product), such that

dxi ∧ dxj = −dxj ∧ dxi ,

(bilinear, associative.)
Examples:
1.
∑ ∑ ∑ ∑
( ai dxi ) ∧ ( bj dxj ) = ai bj dxi ∧ dxj = (ai bj − aj bi )dxi ∧ dxj .
i j ij i<j

2. Change of coordinates:
∂xi j
x = x(y), dxi = dy , (summation in j)
∂y j

3. Corollary:

(a) k ≤ n.
(b) Ωn = f (x)dx1 ∧ . . . ∧ dxn .

Lemma. C ∞ differential forms is a ring ∧∗ (M n , R) such that every C ∞ -map

f : M n → N m there is a natural “induced map” of rings:

f ∗ : ∧∗ (N m ) → ∧∗ (M n ),

commuting with all algebraic operations (addition, multiplication ∧).

Proof. “Change of coordinates”
[ ]
f ∂xi j f∗
x = x(y), y → x, ⇒ dx ⇒ i
dy (sum)
∂y j
commutes with change of coordinates in N and M and with all algebraic
operations like ∧.

83
 For scalar functions we have:
ϕ : N → R, f ∗ ϕ(y) = ϕ(x(y)) by definition.
For forms
∂xi
f ∗ dxi =
(x(y))dy j by definition,
∂y j
products are mapped into products.
The proof is finished.

Homework 5. Solutions.
1.
η η
a
RP 2= RP 2\ Pt= = a a
a
−η −η

2.
b
a a
RP 2 ♯ RP 2 =?.
K2 =
K 2 \ MOBEUS ∼
= MOBEUS
Mobeus b
band (dashed)

3. Already was done

4. “Conformal coordinates” in Sn

1 ψ
ϕ S n \N ∼
= 1Rn n
x ,...,x
1 x l

0 x
84
 ∑
(dxi )2
dl 2
= const · ?
(1 + |x|2 )2
a) n = 1 (line): x = 2 tg ψ
b) n = 2 preservation of angles
5.

dz dz̄
|z| < 1 , dl2 = 4
(1 + |z|2 )2

Central line = geodesics? Metric does not depend on the angle φ !

Lecture 18. Differential forms. De Rham

operator.
Differential forms.
Consider a C ∞ -manifold M n with Atlas of Charts. Locally k-forms are:
∑
Ω= fi1 ...ik (x)dxi1 ∧ . . . ∧ dxik .
i1 <i2 <...<ik

Multiplication: Associative and

dxi ∧ dxj = −dxj ∧ dxi .

Ω=0 for k > n

Ω = f (x)dx1 ∧ . . . ∧ dxn for k = n
Ω = f (x) for k = 0
f (x) · Ω = Ω · f (x)

Change of coordinates: x = x(y)

f∗ ∂xi j
dxi ⇒ dy (sum).
∂y j

85
Functorial properties: g : M → N , x = x(y).
y x

∑ ∂xi
G∗ Ω = fI (x(y))dxi1 ∧ . . . ∧ dxik , dxi ⇒ (x(y))dy j .
∂y j
Algebra of forms.
∑
n
∗
∧ (M ) = ∧k (M ),
k=0

g (Ω1 ∧ Ω2 ) = g ∗ Ω1 ∧ g ∗ Ω2 .
∗

Integration of forms.

1. ∫
f (x) = ±f (P ), k = 0, ±P = “point with orientation”.
±P

2. ∫
f (x)dx1 ∧ . . . ∧ dxn = ordinary integral
D

D-domain (local). To integrate over large domains, one can write

∫ ∑
Ω= ϕq (x)Ω,
D q

where ∑
ϕq (x) ≡ 1, ϕq is a C ∞ function,
q

ϕq ≡ 0 outside of small domain Dqn ⊂ M n . The family ϕq is called

“Partition of Unity”.

3. Let Ω be a k-form and g : Dk → M n , g is C ∞ . By definition:

∫ ∫
Ω = g ∗ Ω − integral of k-form in k-ball Dk .
g(Dk ) Dk

86
Properties: Integration of forms does not depend on local coordinates both
in M n and Dk (restrict the k-form to the k-dimensional body and integrate).
Differentiation of forms:

d : ∧k (M n ) → ∧k+1 (M n )

1. df is usual differential.

2. d(dxi1 ∧ . . . ∧ dxik ) = 0.

3. d(f dxi1 ∧ . . . ∧ dxik ) = df ∧ dxi1 ∧ . . . ∧ dxik .

Example:
∑ ∂fi ∑ ( ∂fj ∂fi
)
d(fi dx ) = dfi ∧ dx =
i i
j
dx ∧ dx =
j i
i
− j dxi ∧ dxj .
i,j
∂x i<j
∂x ∂x

Examples of forms:
Momentum
∑ ∂f i (pi ) in Calculus of Variations. Differential of function
df = ∂xi
dx .
Electric field (Ei )-covector field.
2-forms: Magnetic field in R3 . B = Bαβ dxα dxβ .
Electromagnetic field in R4 = (x0 , x1 , x2 , x3 ).

F = Fij dxi ∧ dxj = cFoj dt ∧ dxα + Bαβ dxα ∧ dxβ ,

i, j = 0, 1, 2, 3, α, β = 1, 2, 3, x0 = cdt, F0α = Eα – electric field, Bαβ = Fαβ

– magnetic field.
Symplectic 2-form in the cotangent space T∗ M n with local coordinates
i
(x , pi )
∑n
Ω= dxi ∧ dpi
i=1

| ∧ .{z
Ω . . ∧ Ω} − volume forms in T∗ (M n ).
n times

It defines nondegenerate inner product given by matrix

( )
0 1n
gij = skew symmetric.
−1n 0

87
 Kähler Riemannian Metric in Complex Space (Manifolds) with complex
linear coordinates (z 1 , . . . , z n ).
∑
dl2 = gα,β̄ dz α dz̄ β > 0.
α,β

Coordinates: (z 1 , . . . , z n , z̄ 1 , . . . , z̄ n ). Associated 2-form is

i∑
Ω= gα,β̄ dz α ∧ dz̄ β , z = x + iy.
2 α,β

Examples: n=1.

 dzdz̄ : R2
dl2 = gαβ dz α dz̄ β = dzdz̄ S2
 :
(1 ± |z|2 )2 H2

i dz ∧ dz̄ dx ∧ dy
Ω= = − Area Form.
2 (1 ± |z| )
2 2 (1 ± (x2 + y 2 ))2

Theorem: The operator d is well-defined in ∧k (M n ) by the definition above.

It commutes with C ∞ -maps g : M → N and has the following properties:

1. d(Ωk ∧ Ωl ) = (dΩk ) ∧ Ωl + (−1)k Ωk ∧ (dΩl ).

2. d ◦ d = 0.

Proof:

a)
( ) ∂f ∂xi j
d g ∗ f (x(y)) = g ∗ df : d(g ∗ f ) = d(f (x(y))) = dy = g ∗ df.
∂xi ∂y j

b)
d(dxi ) = d(0) ∧ dxi = 0,
( i )
∗ dx j ∂ 2 xi
i
d(g dx ) = d dy = dy q ∧ dy j ≡ 0.
dy j ∂y j ∂y q

88
So d commutes with the map g ∗ for k = 1.
For k > 1 our result is obvious by the following reason: d(dxi1 ∧ . . . ∧
dxik ) = 0. Let us prove now that the Statement 1 of the Theorem. Let
Ωk = f · dxi1 ∧ . . . ∧ dxik , Ωl = g · dxj1 ∧ . . . ∧ dxjl .
[ ]
d(Ωk ∧ Ωl ) = d (f · dxi1 ∧ . . . ∧ dxik ) ∧ (g · dxj1 ∧ . . . ∧ dxjl ) =

= (df · g + f · dg) ∧ dxi1 ∧ . . . ∧ dxik ∧ dxj1 ∧ . . . ∧ dxjl =

= df ∧ dxi1 ∧ . . . ∧ dxik ∧ gdxj1 ∧ . . . ∧ dxjl +
+(−1)k f dxi1 ∧ . . . ∧ dxik ∧ dg ∧ dxj1 ∧ . . . ∧ dxjl =
= (dΩk ) ∧ Ωl + (−1)k Ωk ∧ (dΩl ).
Let us prove now that d ◦ d = 0.
∂f j
d(f · dxi1 ∧ . . . ∧ dxik ) = df ∧ dxi1 ∧ . . . ∧ dxik = dx ∧ dxi1 ∧ . . . ∧ dxik .
∂xj
d◦d(f ·dxi1 ∧. . .∧dxik ) = d(df )∧dxi1 ∧. . .∧dxik −df ∧d(dxi1 ∧. . .∧dxik ) ≡ 0.
Theorem is proved.
We have:
∑ ( ∂Aj ∂Ai
)
d (Aj dx )j
= − j dxi ∧ dxj .
| {z } ∂xi ∂x
i<j
“covector field” | {z }
“curl”=2-form

Let n = 3, B12 → B̃ 3 , B13 → −B̃ 2 , B23 → B̃ 1 .

( )
1 2 3
∂ B̃ ∂ B̃ ∂ B̃
d(B12 dx1 ∧dx2 +B13 dx1 ∧dx3 +B23 dx2 ∧dx3 ) = + + dx1 ∧dx2 ∧dx3
∂x1 ∂x2 ∂x3
| {z }
divB̃

Faraday Laws:

n = 3 : a) dB = 0, Magnetic field
1 ∂B α
b) dE = c ∂t , E = Eα dx ,
n = 4 : d(Fij dxi ∧ dxj ) = 0. Electromagnetic field

89
 Lecture 19. Differential forms. Cohomol-
ogy.
Consider the algebra of differential forms on a manifold N m . Let Ω ∈
∧k (N m ). Then dΩ ∈ ∧k+1 (N m ). Consider a map f : M → N . Let us
check, that f ∗ dΩ = df ∗ Ω.
Proof. Let ϕ : N → R, f ∗ ϕ(y) = ϕ(x(y)).
( )
∗ ∗ ∂ϕ i ∂ϕ ∗ i ∂ϕ ∂xi j
f dϕ(y) = f dx = (x(y))f (dx ) = (x(y)) dy =
∂xi ∂xi ∂xi ∂y j

∂(f ∗ ϕ) j
j
dy . O.K. f ∗ d = df ∗ ,
∂y
for scalars and dxi . But the product ∧ commutes with f ∗ . Every form is a
combination of ϕ(x) · dxi1 ∧ . . . ∧ dxik . So we have

f ∗ d = df ∗ for all k ≥ 0.

Homology (Cohomology).

H k (M n , R) = Ker d/ Im d in ∧k (M n ).

Ker d ⊂ ∧k : {Ω | dΩ = 0} closed forms

′
Im d ⊂ ∧ : {Ω | Ω = dΩ }
k
exact forms

Examples: a) k = 0 ⇒ Ω = ϕ is locally constant.

Conclusion: H 0 (M n , R) = Rp , where p is the number of components.
Closed Forms.
∑
1. Forms with constant coefficients in Rn or in T n = Rn /Zn : Ω = aI dxI ,
aI = const.

2. Closed 1-form
∑ ∑ ( ∂ψj ∂ψi
)
ω= i
ψi (x)dx , dω = i
− j dxi ∧ dxj .
i<j
∂x ∂x

90
 ∂ψj ∂ψi
dω = 0 ⇔ i
− j ≡ 0,
∂x ∂x

∫x γ1
ϕ(x) = ω ⇒ dϕ = ω if ϕ is well-defined. x1
x0 γ2
x0

Necessary condition: dω = 0.

τ
Stokes Formula. Consider an oriented n
manifold M n with boundary
∂M n = W n−1 with “induced” orientation.
What is it? M
W
Consider “external” normal vector n to W ⊂ M and tangent frame τ ton

W . Let (n, τ ) form an oriented frame in M n at the point P . We say that τ

is an oriented (n − 1)-frame to W in the induced orientation.
Theorem. For every (n − 1)-form Ω in M n we have

1. ∫ ∫ ∫ ∫
··· Ω= ··· dΩ,
W n−1 Mn

W n−1 is a closed C ∞ manifold, M n is compact “manifold with bound-

ary W ”.

2. For every manifold N , n−1-form Ω in ∧n−1 (N ) and mapping f : M N →

N we have ∫ ∫ ∫ ∫
· · · f Ω = · · · f ∗ dΩ,
∗

W n−1 Mn

Proof for the case M n = I n (cube) with coordinates (x1 , . . . , xn ), 0 ≤ xi ≤ 1,

91
 (0,1) (1,1)

x2 W = ∂I n (I n =“ball”).

(0,0) x1 (1,0)

a) Case n = 1. We have

∫b
Ψ′ dx = Ψ(b) − Ψ(a), Ω = Ψ(x). b
a
a

b) Case n > 1. We have

∑
n ∑
n
Ωn−1 = ci ∧ . . . ∧ dxn =
Ψi dx1 ∧ . . . ∧ dx Ω(i)
n ,
i=1 i=1

ci means that this multiplier is omitted.

where dx
(i)
Consider every summand Ωn separately.

xi =1 ∫ ∫ ∫ ∫
··· Ω(i)
n = ··· Ω(i)
n =
∂I n [xi =0,1]

xi =0
[∫ ∫
= (−1)i−1 ci ∧ . . . ∧ dxn −
· · · Ψi (x1 , . . . , xi = 1, . . . , xn )dx1 ∧ . . . ∧ dx

∫ ∫ ]
− ··· i c
Ψi (x , . . . , x = 0, . . . , x )dx ∧ . . . ∧ dx ∧ . . . ∧ dx =
1 n i 1 n

∫ ∫
= (−1) i−1
··· ∂i Ψi (x1 , . . . , xi , . . . , xn )dx1 ∧ . . . ∧ dxi ∧ . . . ∧ dxn =
In
∫ ∫
= ··· dΩ(i)
n .
In

92
 Let n = 2. Orientation for n = 2:
d c

x2 Here a = (0, 0), b = (1, 0), c = (1, 1),

d = (0, 1), (n, τ ) = (x1 , x2 ).

a x1 b
Ω = Ψ1 (x1 , x2 )dx1 + Ψ2 (x1 , x2 )dx2 ,
( )
∂Ψ2 1 2 ∂Ψ1 1 2
dΩ = (x , x ) − (x , x ) dx1 ∧ dx2 ,
∂x1 ∂x2
∫ ∫ 1 ∫ 1 ∫ 0 ∫ 0
1 1 2 2 1 1
Ω= Ψ1 (x , 0)dx + Ψ1 (1, x )dx + Ψ1 (x , 1)dx + Ψ1 (0, x2 )dx2
0 0 1 1
∂I 2

Let Ω = Ψ1 dx1 for n = 2. We have

∂Ψ1 2 ∂Ψ1
2
dx ∧ dx1 = − 2 dx1 ∧ dx2
dΩ =
∂x ∂x
 
∫∫ ∫1 ∫1 ∫1
∂Ψ ( )
dΩ = −  dx  dx = − dx1 Ψ1 (x1 , 1) − (Ψ1 (x1 , 0) =
1 2 1
∂x 2
I2 0 0 0

∫1
( )
= dx1 Ψ(x1 , 0) − (Ψ(x1 , 1) .
0

We have ∫ ∫ 1 ∫ 1
Ω= Ψ1 (x , 0)dx −
1 1
Ψ1 (x1 , 1)dx1 .
0 0
∂I 2

These two expressions coincide.

“Ordinary homology” for the C ∞ - manifolds:

0
• ϕ : I k → M n – “singular cube”.
0 1

93
 0 2
ψ : σ → M – “singular simplex”.
k n
•
σ k =< 0, 1, . . . , k >. 0 1
0 1

ϕ, ψ are assumed to be C ∞ maps.

“Boundary”:
∂I k = ∂(I k−1 × I 1 ) = ∂(I k−1 ) × I 1 + (−1)k−1 I k−1 ∂I 1 .
∑n ∑
n
∂σ k = (−1)i < 0, 1, . . . , bi, . . . , k >= (−1)i σik−1 .
i=0 i=0
k−1 ϕ
“Degenerate cube” I −−−−−→ I
k
−
→ M n.
projection
Every form Ωk defines a linear form

C ∞ − singular cubes ⇒R
C ∞ − singular simplices ⇒ R,

(integration of Ω along singular cubes, singular simplices, degenerate cubes

→ 0).

Homework 6.
1. Find all geodesics for Sn , Rn , Hn (for Hn in P - model and in K -
model.
2. Find all closed geodesics in torus T2 with euclidean metric.
3. Find all closed geodesics in RP2 (metric of constant positive curvature).
4. Prove that parallel transport along any path preserves inner product
γ(t) : [xi (t)] ,
ξ (t)
⟨η(t), ζ(t)⟩ = const if ∇ẋ η = 0, ∇ẋ ζ = 0
η(t) γ (t)
(∇s gij ≡ 0 ).

94
 Lecture 20. Differential forms and tensors.
Categorial properties. Cohomology and Stocks
formula. Homotopy invariance of Cohomol-
ogy.
Category of C ∞ -manifolds and C ∞ -maps. Functors like spaces of tensors
with low indices. Differential k-forms ∧k (M n ): exterior multiplication ∧,
operator d and integration, their properties. Stokes formula fro integration.
Singular cubes and simplices, boundary operator ∂. Cohomology, forms and
singular complexes.

a) H∧k (M n , R) (forms).

b) H k (M n , R) (simplices and cubes).

c) Hk (M n , R) (simplices and cubes).

a) Ker
| {z d} / Im d
|{z} = H∧k (M n , R).
closed forms exact forms

∗ ∗
| {z∂} /
b) Ker Im
| {z∂} = H k (M n , R) (cubes, simplices).
cocycles coboundaries

c) |Ker
{z ∂} / Im ∂ = Hk (M n , R) (cubes, simplices).
|{z}
cycles boundaries

∂I k = ∂(I k−1 × I 1 ) = ∂(I k−1 ) × I 1 + (−1)k−1 I k−1 ∂I 1 .

∑n
∂σ k = (−1)i σik−1 , σik−1 =< 0, 1, . . . , bi, . . . , k >
i=0
“simplicial chains”:
∑
λs (σ k , ϕs ), ϕs : σ k → M n (C ∞ ).
s

“Simplicial cochains” = functional ψ on simplicial chains:

[ ]
∑ ∑
ψ(chain) = ψ λs (σ k , ϕs ) = ψ(σ k , ϕs ).
s s

95
 “Boundary operator:” ∂ : k-chains→ k − 1-chains (linear).
∑
n
k
∂(σ , ϕ) = (−1)i (< 0, 1, . . . , bi, . . . , k >, ϕ),
i=0

where ϕ is naturally defined at the boundary simplices.

“Coboundary operator:” ∂ : k-cochains→ k + 1-cochains.

< ∂ ∗ a, b >=< a, ∂b >, a is a cochain, b is a chain.

“Special cochains”= C ∞ -form Ω

∫ ∑ ∫
< Ω, b >= Ω = λs Ω, b is a chain.
s
b (σ k ,ϕs )

“Stokes formula”
∫ ∫
dΩ =< dΩ, b >=< Ω, ∂b >= Ω,
b ∂b

Conclusion: Stokes formula defines a correct homomorphism

H∧k (M n , R) → Hsimplices
k
(M n , R).

In Algebraic topology:
k
Hsimplices (M n , R) → Hom(Hk (M n ), R).

Isomorphism
H∧k (M n , R) ∼
= Hom(Hk (M n ), R)
was claimed by Poincaré in 1895 and proved by De-Rham in 1930s.
Lemma 1. Cohomology H∧k (M n , R) form a ring (operation ∧) such that

a ∧ b = (−1)kl b ∧ a, a ∈ H∧k , a ∈ H∧l .

Proof. Let Ω, Ω′ represent a, b, i.e. Ω + Im d) ∼

= a, Ω′ + Im d) ∼
= b. Define

a∧b∼
= (Ω ∧ Ω′ + Im d) in H k+l (M n , R).

96
We have

(Ω + du) ∧ (Ω′ + dv) = Ω ∧ Ω′ + du ∧ Ω′ + Ω ∧ dv + du ∧ dv,

but

du ∧ Ω′ = d(u ∧ Ω′ ), Ω ∧ dv = ±d(Ω ∧ v), du ∧ dv = d(u ∧ dv).

The Lemma is true.

Instead of R maybe any ring, for noncommutative ring may take place
a ∧ b ̸= ±b ∧ a. For associative rings multiplication of forms and homology
is associative.
Definition. C ∞ maps g, h : N → M are homotopic if there exists a C ∞
map F : N × I → M such that F t=1 = g, F t=0 = h.
Theorem. For homotopic maps g, h induced maps of cohomology coincide.

g ∗ = h∗ : H∧k (M , R) → H∧k (N , R).

(x) (y)

Proof. Let Ω ∈ ∧k (M ) be a k-form and F ∗ Ω ∈ ∧k (N × R) be a k-form.

Coordinates in N × R are y 1 , . . . , y n , t. Every form can be written u =
a + dt ∧ b, where a, b do not contain dt.
Define operator:
D : ∧k (N × R) → ∧k−1 (N )
by formula
∫1
Du = b(t)dt, u = a + dt ∧ b, a → 0.
0

We have:
1

t N I Lemma. Ddu + dDu = u t=1

−u t=0
.
0

Proof.
du = dy a + dt ∧ ȧ − dt ∧ dy b,

97
 ∫ 1 ∫ 1 ∫ 1
Ddu = ȧdt − dy bdt = a t=1
−a t=0
− dy bdt,
0 0 0
but
∫ 1 ∫ 1
a t=1
=u t=1
, a t=0
=u t=0
, bdt = Du, dy bdt = dDu.
0 0

O.K.
Finally we have for every closed k-form Ω in M :

Ω → F ∗ Ω → DF ∗ Ω in ∧k−1 (N ).

t N I
dΩ = 0, dF ∗ Ω = 0,
0

so we have
X  X 
dD(F ∗ Ω) + X
Dd(F ∗
XΩ)
∗
X=F Ω t=1
− F ∗Ω t=0
= g ∗ Ω − h∗ Ω.

So g ∗ Ω = h∗ Ω + Im d,

g ∗ ≡ h∗ : H k (M ) → H k (N ).

Theorem is proved.
Corollary. For every contractible manifold M like point, Ball, space Rn and
so on cohomology are the same:
{
∗ R, k = 0,
H∧ (point) =
0, k ̸= 0.

Poincaré Lemma. For every manifold M n every closed k-form Ω is “locally

exact” (k > 0):

Ω = dω in contractible open domain U ⊂ M n .

98
 Lecture 21. Homotopy invariance of Co-
homology. Examples of differential forms.
Symplectic and Kähler manifolds.
Homotopy invariance of cohomology.

g:N →M
C ∞ − maps:
h:N →M

are homotopic if exists a map F such that:

F :→ N × I → M, F t=1
= g, F t=0
= h.

Theorem. If g and h are homotopic, then

g ∗ = h∗ : H k (M, R) → H k (N, R).

Proof. Let u ∈ ∧∗ (N × I). Then u can be uniquely written as:

u = a + dt ∧ b = aI (y, t)dy I + bI (y, t)dt ∧ dy I .

Define the following operator D:

 1 
∫
Du =  bI (y, t)dt dy I .
0

Lemma.
Ddu + dDu = u t=1
−u t=0
.

Proof.
du = dy a + dt ∧ ȧ − dt ∧ dy b,

1
∫ 1 ∫ 1
t dy Du = dy bdt = (dy b)dt
0 0
0 y
N

99
 Ddu = D(d(a + dt ∧ b)) = D(dy a + dt ∧ ȧ − dt ∧ dy b) =
∫ 1
= a t=1 − a t=0 − dy bdt = a t=1 − a t=0 − dDu.
0
O.K.
Proof of the Theorem.

F :→ N × I → M ,
Ω

F ∗ dΩ = df ∗ Ω = 0,
XX 
dD(F ∗ Ω) + X
Dd(F ∗
XΩ)
∗
X=F Ω t=1
− F ∗Ω t=0
= h∗ Ω − g ∗ Ω.
g ∗ ≡ h∗ : H k (M ) → H k (N ).
O.K.
Homotopy equivalent manifolds N , M :

∃ N−
→M −
→N
ϕ ψ

such that

ψ · ϕ is homotopic to 1N : N → N
ϕ · ψ is homotopic to 1M : M → M.

Corollary: M and N are homotopy equivalent implies:

ϕ∗ : H ∗ (M ) → H ∗ (N )
ψ ∗ : H ∗ (N ) → H ∗ (M ).

are isomorphisms of rings.

In particular, H ∗ (Rn ) = H ∗ (Dn ) = H ∗ (M ) = H ∗ (point), where M is
any contractible manifold (map 1 : M → M is homotopic to const:
M → point).
Poincaré Lemma. Every closed k-form Ω is “locally exact” (k > 0).
Examples of forms

100
1. k = 0, df = 0 ⇒ f = const (in every component).
H 0 (M ) = Rp , where p = number of components.

2. M n is an oriented manifold. Every n-form Ω is closed: dΩ = 0 (obvi-

ous). Riemannian metric in M n generates a “volume form”
√
Ω = dn σ = det gij dx1 ∧ . . . ∧ dxn locally in oriented atlas.
Form Ω = dn σ is closed and not exact for closed (compact) manifolds
because ∫
Ω = volume M > 0.
Mn

So H (M ) ̸= 0 for closed oriented manifolds. For connected nonori-

n n

ented manifolds and and manifolds with boundary we have H n (M n ) =

0 (not proved yet).
3. “Nondegenerate 2-forms”:
∑
Ω= Ωijdxi ∧ dxj (locally)
i<j

det Ωij ̸= 0, Ωij = −Ωji we have n = 2k.

Theorem. For every 2-form in 2k-dimensional manifold we have:

1 √
Ω ∧ . . . ∧ Ω = det Ωij dx1 ∧ . . . ∧ dx2k .
k! | {z }
k times

Proof. Let M k = Rn and Ωij = const.

Step 1. Choose such basis that
( ) ∑
0 1
k
Ωij = , Ω= dxi ∧ dpi .
−1 0
i=1

Step 2. Proof Theorem in this basis:

( k )k
∑
dxi ∧ dpi = k! dx1 ∧ dp1 ∧ . . . ∧ dxk ∧ dpk .
i=1

101
 We have:
(dxi ∧ dpi ) ∧ (dxj ∧ dpj ) = (dxj ∧ dpj ) ∧ (dxi ∧ dpi ).

Step 3. Return to the original basis. Both sides are n-forms and
transform in the same way.
So Theorem is proved.

Lemma. Every skew-symmetric inner product can be reduced to the

form  
0 1 0
Ωij =  −1 0 0  ,
0 0 0

Proof (for nondegenerate forms). For every vector e there exists e′

such that < e, e′ >= 1. Find orthogonal compliment to the subspace
(e, e′ ). Iterate this process. Our Lemma follows.
√
Remark. det Ωij is a polynomial of the matrix entries if Ωij = −Ωji .
It is called “Pfaffian”.

Definition. Manifold M 2k is called “simplistic” if there exists a 2-form

Ωij dxi ∧ dxj such that det Ωij ̸= 0 and dΩ = 0.
Examples.
a) T∗ (M k ). We have
∑
k
Ω= dxi ∧ dpi .
i=1

b) Kähler (complex manifolds). Complex coordinates (z 1 , . . . , z n ) (locally).

o < gαβ dz α dz̄ β − Riemannian Metric.
i∑
Ω= gαβ dz α ∧ dz̄ β , dΩ = 0, det gαβ =
̸ 0.
2 α,β

Corollary. Every simplectic manifold is oriented.

√
Proof. Ωk = det Ωij dx1 ∧ . . . ∧ dx2k is a “volume form”.

102
 Lecture 22. Symplectic Manifolds and Hamil-
tonian Systems. Poisson brackets. Preser-
vation of Symplectic form by Hamiltonian
System.

Symplectic form. M n , Ω, dΩ = 0, n = 2k, Ωk = k!dn σ – volume.

Property (without proof): locally there exists a system of coordinates
x1 , . . . , xk , p1 , . . . , pk such that
∑
k
Ω= dxi ∧ dpi .
i=1

Example. Consider cotangent bundle of a C ∞ manifold M n = T∗ N k , with

coordinates (x, p). Change of coordinates
∂xi
x = x(y), p´j = pi
∂y j

Lemma 1. 2-form Ω is well-defined (independent of the choice of coordinates

in N ).
∑ ∑
k
Proof. Ω = dxi ∧ dpi , Ώ = dy i ∧ dṕi . Why Ω = Ώ.
i i=1

∂xi ∂ 2 xi
dṕj = dpi j
+ pi j k
dy k
∂y ∂y ∂y
∑ ∑ ∂xi XXX∂X xXj
2 i 
dy j ∧ dṕj = dy j ∧ dpi + dpi  X
 j ∂y k X
dy ∧
XXX =
dy k

j ij
∂y j ∂y X
∑ ∂xi ∑
= j
dy j ∧ dpi = dxi ∧ dpi .
ij
∂y i

O.K.
More generic simplectic form in T∗ N is
∑ e ∑
dxi ∧ dpi + B = ΩB , B = bij (x)dxi ∧ dxj , dB = 0.
c i<j

103
For k = 3 it is “Magnetic Field” (correction of simplectic form).
Hamiltonian system in: T∗ N = M is defined by “Hamiltonian” H(x, p),

∂H
ẋi =
∂pi
∂H
ṗi = −
∂xi
Another form: Take
∑ covector field (dH). Take inner product given by
simplistic form Ω = dx ∧ dpi .
i

( )
0 1
Ωij = .
−1 0

Construct vector field ηH = Ω−1 (dH).

Dynamical System.
ż = η(z), z = (x, p),
ẋ = Hp , ṗ = −Hx .

“Poisson Bracket”.

{f, g} =< df, dg >Ω = − < dg, df >Ω .

It provides the structure of Lie algebra.

For any functions f (x, p) we have:

f˙ = {H, f }.

We have {H, H} = 0 – conservation of energy.

Geodesics. M = T∗ N , H = 12 g ij pi pj . Lagrangian L = 12 gij ẋi ẋj ,

∂L ∂L
H = ẋi i
− L, pi = .
∂ ẋ ∂ ẋi
Euler-Lagrange equations:
∂L
ṗi = ⇔ ẍi + Γijk ẋj ẋk = 0.
∂xi

104
Theorem 1. In every compact closed simplectic manifold M n (n = 2k)
with form Ω (nondegenerate, we have

H 2j (M ) ̸= 0, j = 1, . . . , k.

∫
Proof. We know that Ωk = k! volume, so Ωk ̸= 0. Therefore H 2j ̸= 0,
because Ω2j can not be exact (if Ω2j is exact then Ω2k is exact.)
Theorem 2. For any simplectic manifold M n and Hamiltonian system with
Hamiltonian H : M n → R simplectic form is preserved by mapping St :
M n → M n (time shift by our system).
Proof. Locally we choose such coordinates (x1 , . . . , xk , p1 , . . . , pk ) that Ω =
∑
k
dxi ∧ dpi (Darboux).
i=1
Our system is:
∂H ∂H
ṗi = − , ẋi =
∂xi ∂pi
Small time shift is:

St : pi → pi + ṗi t + O(t2 )
xi → xi + ẋi t + O(t2 )

St∗ pi = pi + ṗi t + O(t2 )

St∗ xi = xi + ẋi t + O(t2 )

We have
∑ ∑[ ∂H
] [
∂H
]
St∗ (dxi ∧dpi ) = i i
(dx +ẋ t)∧(dpi +ṗi t) = i
dx + td ∧ dpi − td i =
i i
∂pi ∂x

∑ [ ]
∂H ∂H
= dx ∧ dpi + t d
i
∧ dpi − dx ∧ d i =
i

i
∂pi ∂x
∑ [ ∂ 2H  
 ∂ 2
H ∂ 2 
H

 ∂ 2
H
]

j ∧ dpi + 
dp dx ∧ dpi − dx∧
j i
dx − dx ∧ i
j i
  ∂xi ∂xj
= Ω+ t ∂p dpk =

∂p i j ∂p i ∂x j ∂x ∂pk
ij

=Ω

105
 Theorem is proved.
Remark 1. “Multivalued Hamiltonian”= closed 1-form dH.
Remark 2. “Poisson structure”: Ω may be degenerate, Ω∗ is well-defined
but degenerate (Ω∗ = Ωij ).
Remark 3. The Poisson Structure is a symplectic inner product of
covectors which defines a Lie Algebra of functions (the Poissoon Bracket):
{f, g} =< df, dg >Ω
{{f, g}, h} + {{g, h}, f } + {{h, f }, g} = 0.
Here we have Ωij = (Ω)−1 if it is nondegenerate. This definition implies
in that case dΩ = 0, Ω = Ωij dxi ∧ dxj .
Remark 4. According to the Darboux Theorem, every nondegenerate
Symplecttic Geometry with dΩ = 0 is always ”flat”, i.e. there exists system of
local coordinates such that Ω is constant (and reduced to standard canonical
form.)

Homework 7.
1. Let action functional is
∫ ∫
S {γ} = i j
gij ẋ ẋ dt + Ai (x) dt
γ γ

where γ = {xi (t)} .

Prove that the Euler - Lagrange equation depends on the form
( )
d Ai dxi = Ω
(only).
2. Let γ(t) be a plain curve, closed.
γ (t) Prove that:
I
k(s) ds = 2π × n , n = integer

where s is natural parameter and k(s) is curva-

ture.
106
3. Calculate area form in R2 , S2 , H2 in the spherical (pseudospherical)
coordinates (polar for R2 ).
4. Prove that
xdy − ydx
dφ = c ·
x2 + y 2
in the plane R2 .
5. Let a complex line bundle Cx1 be given over M n , (x1 , . . . , xn ) with
an imaginary Differential - Geometric Connection {AUj dxj = iAUjReal dxj } ,
U ⊂ M n , AUjReal ∈ R . Let the bundle is “unitary” (i.e. in the domains
U ∩ V change of basis eU (x) = g U V (x) eV (x) , g U V ∈ U1 = S1 (eiψ )) .
Prove that 2-form
( ) ( ) ( )
H = H U = dAUReal = dAVReal
in U ∩ V is well defined as a closed 2-form in Λ2 (M n ) .
dH = 0
(first Chern class). Its integrals along the closed 2-submanifolds are always
2πn, n ∈ Z.
Remark. This fact (well-known in Topology of characteristic classes)
was also discovered by Dirac in 1930-50s in process of Heisenberg-Schrodinger
quantization of ”Magnetic Monopole”: the Hilbert space of state is in fact
space of sections of complex line bundle where vector-potential of magnetic
field is a covariant derivative of sections. (The Feinman quantization through
the path integral leads to different topology: the Action functional for mag-
netic monopole is in fact a closed 1-form on the space of paths whose periods
along 1-cycles in this space should be quantized, i.e. they should be equal to
2πn, n ∈ Z assuming that the Plank constant is equal to one in our units.)

Lecture 23. Volume element in Compact Lie

groups. Averaging of differential forms and
Riemannian Metric. Cohomology of Homo-
geneous Spaces.

Homogeneous spaces. M n = G/H, G – Lie group (compact).

107
Theorem. Let group G be compact. There exists Riemannian Metric in
M = G/H s.t. g ∗ gij (x) = gij (x), g ∈ G, g : M → M , g(x)- action of the
group G.
Remark. We assume that there exists a G-invariant volume form Ω in
the group G. We call it dσ(g).
Proof of the Theorem. Take any Riemannian Metric gij (x) in M . Con-
sider family of metrics g ∗ gij (x) = gijh (x) in M , depending on g ∈ G as
parameter. Integrate:
∫
gijg (x)dσ(g) = Gij (x).
G

1. This is a Riemannian (positive!) metric.

2. This metric is G-invariant because for ψ ∈ G we have
∫ ∫ ∫
∗ ∗ g
ψ Gij = ψ gij (x)dσ(g) = gij (x)dσ(g) = gijψg (x)dσ(ψg)
ψg

G G G

The last identity is based on the following:

Lemma 1. In every compact Lie Group there exists a double-invariant
volume form.
Proof. Consider left-invariant metric in G : gij and its volume form Ω :
h∗ gij = gij , h∗ Ω = Ω. But this n-form is also right-invariant because for
g → hgh−1 this form maps into itself (space of right-invariant n-forms is
1-dimensional and group is compact).
So the Theorem is proved.
Theorem 2. For homogeneous manifolds with compact group G: M =
G/H every closed differential form is homologous to an invariant closed form.
Proof. Let dω = 0 in ∧k (M ). Consider the integral
∫
1
ω= h∗ ω dσ(h),
|G|
G
∫
|G| = dσ(h), h ∈ G, h : M → M.
G

108
 1. This is an invariant form.

2. This is a closed form:

∫ ∫
1 ∗ 1
dω = h ω dσ(h) = h∗ (dω) dσ(h) = 0.
|G| |G|
G G

3. This from represents the same cohomology class as ω: all forms h∗ ω

belongs to the same cohomology class if the group is connected. So
all “integral sums” belong to the same cohomology class.

So our Theorem follows.

Examples.

1. Spheres S n , G = SOn+1 . Only invariant forms are k = 0 (scalars ) and

k = n. So we have:

H 0 (S n ) = H n (S n ) = R.

2. Tori T n = Rn /Zn , G = T n (abelian group). Spaces of invariant forms

are ∧k Rn . All of them are closed: dω = 0 for invariant forms, so
H ∗ (T n ) = ∧(v1 , . . . , vn ).

3. S n1 × . . . × S nk . Forms v1 , . . . , vk , dim vk = k. We have vj2 = 0 only

relations.

H ∗ (S n1 × . . . × S nk ) = {v1 , . . . , vk , vj2 = 0}.

4. Cohomology of RP 2n+1 are the same as for S 2n+1 ( H ∗ ( , R) ).

Cohomology of RP 2n are 0 (k > 0).
Cohomology of CP n are generated by 2-form Ω and its powers:

1, Ω, Ω2 , . . . , Ωn .

Action of SOn and Un in Rn and Cn . Invariant differential forms:

109
 1. Rn , dx1 , . . . , dxn , A ∈ SOn .
∑
A(dxi ) = aij dxj , AAt = 1.
j

Invariant differential forms are:

k = 0, scalar, k = n, dx1 ∧ . . . ∧ dxn .

No other invariant forms!

2. Cn , dz 1 , . . . , dz n , dz̄ 1 , . . . , dz̄ n , A ∈ Un .
∑ i j 
A(dz i ) = aj dz , 
∑ i j
j
A(dz̄ i ) = āj dz 
j

Invariant forms are:

k = 0 – scalars.
∑
k = 2 (2-forms) i
2
dz j ∧ dz̄ j = Ω.
j
l
k = 2l, Ω .
For l = n we have the “volume element”
( )n
i
Ω=± dz 1 ∧ . . . dz n ∧ dz̄ 1 ∧ . . . ∧ dz̄ n .
2

For n = 1.
R2 = C1 : dz, dz̄, A = (eiϕ ) = U1 .

dz → eiϕ dz
dz ∧ dz̄ → dz ∧ dz̄.
dz̄ → e−iϕ dz̄

No other invariant differential forms for Cn and G = Un (with con-

stant coefficients).

Important homogeneous spaces G/H with compact group H.

110
 1. S n = SOn+1 /SOn .
S 2n−1 = Un /Un−1 .
S 4n−1 = Spn /Spn−1 .

2. CP n = Un+1 /U1 × Un = SUn+1 /Un .

3. Stiefel Manifolds.
Vn,k = SOn /SOn−k (orthogonal k-frames in Rn ).
C
Vn,k = Un /Un−k (complex version).

4. Grassmann Manifolds.
Gn,k = SOn /SOk × SOn−k .
GCn,k = Un /Uk × Un−k

5. Lie Groups (compact, simple) G

G = {G × G/G}, g → h1 gh−1
2 , h1 ∈ G, h2 ∈ G.

6. “Principal” homogeneous spaces

G = G/(1), H = 1.

Lecture 24. Volume element in Compact Lie

groups. Averaging of differential forms and
Riemannian Metric. Cohomology of Homo-
geneous Spaces.
Theorem. Let G be a compact Lie group with invariant volume element
dσ(h):
For h1 , h2 : G → G, ϕ : h → h1 hh−1
2 we have
ϕ∗ dσ(h) = dσ(h).
Let G act (from the left) in the manifold M h : M → M , h ∈ G. Group
G is connected. Then

1. Every closed differential form is homologous to an invariant form.

111
 2. Operator of “Averaging” maps H k (M ) as a unit homomorphism:
∫
1
ω= h∗ ω dσ(h), h ∈ G,
|G|
G

0 = dω ⇒ dω̄ = 0, ω − ω̄ = dv, ω̄ = dv ⇒ ω̄ = dv̄,

i.e.

H ∗ (M ) ∼ ∗
= Hinv (M ) ← calculated with G − invariant forms only.

Proof was given at the previous lecture.

Examples.
1. S n , G = SOn+1 , H = SOn . Invariant forms k = 0, n only. H 0 = R,
H n = R.

2. G = S 3 = SU2 , H = (1). Invariant forms k = 0, 1, 2, 3

∧0 = R, ∧1 = R3 , ∧2 = R3 , ∧0 = R,

d : ∧1 → ∧2 is an isomorphism.

3. Un+1 /U1 × Un = CP n = SUn+1 /Un . G = SUn+1 , H = Un . H acts on

Cn (tangent to 1). Constant forms in Cn (all closed).
Basis: dz i1 ∧ . . . ∧ dz ik ∧ dz̄ j1 ∧ . . . ∧ dz̄ jl , type (k, l).
Action: a ∈ Un , ∑
A(dz i ) = aij dz j ,
∑
j
A(dz̄ i ) = āij dz j
j

For n = 1.
dz → eiϕ dz
⇒ invariant form dz ∧ dz̄.
dz̄ → e−iϕ dz̄

∑
n
k ≥ 1: invariant form are dz i ∧ dz̄ i and all Ωk ,k = 1, . . . , n.
i=1

No other invariant forms (!).

112
Cohomology rings.

1. Tori T n = S 1 × . . . × S 1 = Rn /Zn

H ∗ (T n ) = ∧(v1 , . . . , vn ), dim Vj = 1.

2. Spheres S n , G = SOn+1 . Only invariant forms are k = 0 (scalars ) and

k = n. So we have:

H 0 (S n ) = H n (S n ) = R.

3. S n1 × . . . × S nk , G = SOn1 +1 × . . . × SOnk +1 .

H ∗ (S n ) = {1, vn |vn2 = 0}.

H ∗ (S n1 × . . . × S nk ) = ∧∗ (1, vn1 , . . . , vnk |vj2 = 0), no other relations.

4. CP n .

H ∗ (CP n ) = {1, u, u2 , . . . , un }, dim u = 2, un+1 = 0, G = SUn+1 , H = Un .

5. Homology of Vn,k , Gn,k can be computed using forms

H ∗ ( , R) =?

Vn,k = SOn /SOn−k ; Gn,k = SOn /SOk × SOn−k ,

C
Vn,k = Un /Un−k ; GCn,k = Un /Uk × Un−k .

6. Lie Groups G (compact): G = G × G/G.

Lecture 25. Approximation and Transver-

sality.

Differential topology. C ∞ -manifolds, Atlases, Charts, Orientations, C k -

maps, rank of map, C ∞ -submanifolds: (existence), M n ⊂ RN (M n compact).

113
Approximation. Every continuous map of compact manifolds f : M n →
x
N m , y(x) can be approximated by homotopic C ∞ -map g : M n → N m :
y
max |f (x), g(x)| < ϵ.
x

More exactly. Let f : M → N be C 0 -map which is C ∞ in open domain

U ⊂ M n . For every domain V , V ⊂ U ⊂ M n approximation g : M n → N m
can be chosen such that f ≡ g in V .
000000000000000
111111111111111
111111111111111
000000000000000
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111 U ⊂ Mn
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111
(“Locality Property”.)
000000000000000
111111111111111
000000000000000
111111111111111
000000000000000
111111111111111

V
Lemma. For ϵ > 0 small enough maps f and g are homotopic.
Proof. Take Riemannian Metric gij (y) in N m .

ε For ϵ > 0 small enough geodesics joining

g(x)
f (x), g(x) is unique in the ball of radius
f(x) 2ϵ.

Our homotopy it such that every point moves from f (x) to g(x) along
this short geodesics homogeneously and reaches its end at t = 1. O.K.
Corollary. Homotopy class (C ∞ ) for c∞ maps M → N are the same as C 0
homotopy classes.
Transversality. Let W m−k ⊂ N m be a C ∞ -submanifold. A map f : M n →
N m is called “transversal along W ” iff for every x ∈ f −1 (W ) map of
tangent spaces

(π ◦ df ) : Txn
df
/ Tm π / T m /T m−k ∼
= Rk
f (x) normal projection 6N W
normal plane

has rank k.
Examples.

1. n < k ⇒ f 1 (Ω) = ∅ empty.

2. n = k:

114
 (a) f 1 (W ) are isolated points x1 , . . . , xl , (df )xk has rank k.
(b) Special case: m = k, dim W = 0 (point).

a) 2
N 1 b) m=k=n
1 W
f(M )
...
x1 x2 xl W
k=1, m=1, n=2 Txi ∼ = R = Rm
= TN ∼ n

Theorem (without proof).

1. Every map can be approximated by a map transversal along W ⊂ N
(approximation is “local”).
2. For every map f submanifold W ⊂ N can be approximated (“locally”)
by W̃ ∈ N such that f is transversal along W̃ .
“Locality” = “remains unchanged in the almost entire area where it al-
ready was transversal”.
Theorem 1. Let map f : M n → N n ⊃ W n−k be transversal along W ∈ N .
Then f −1 (W ) is a C ∞ -submanifold in M of codimension (and map df normal
is isomorphic).
1
1 W
f(M ) Proof. Let y 1 , . . . , y m be local coordinates in
N m and W m−k is given locally by equations:
m=2 W : y 1 = . . . = y k = 0.
2
N
For the transversal map f : M → N along W ⊂ N we have functions
y 1 (f (x)) = f ∗ y 1 , . . . , f ∗ y k = y k (f (x)),
and the inverse image f −1 (W ) is given by equations:
y 1 (f (x)) = 0, . . . , y k (f (x)) = 0
∥ ∥
1
z (x) ... k
z (x) = V n−k ⊂ M

115
 These equations are NONDEGENERATE because of TRANSVER-
SALITY long W (i.e. at the points f (x) ∩ W ).
So we have nondegenerate submanifold

V n−k ⊂ M

and map
V n−k → W n−k
is isomorphic along normal k-planes by definition of TRANSVERSAL-
ITY.
f
Examples. Function Rn − → R ⊃ W = point. Level f (x) = y0 is transversal
iff df ̸= 0 for all f (y). Topological invariants.
1

1. Let n = 1.

N +−N−= 0 N +−N−= 1

(a) local = numbers N+ and N− of “positive” and “negative” points.

(b) global = N+ − N− .

2. Degree of map. f : M n → N n – closed (oriented).

f −1 (point) = N+ ∪ N− ,

• degree (mod 2) is N+ − N− (mod 2).

• degree (over Z) is N+ − N− (both M n , N n are oriented).

3. Intersection index.

f
f(M 1) Mn −
→ N n+p ⊃ W p .
1 2
W N
(M n , f ) ◦ W p ∈ Z
if oriented, otherwise modulo 2.
n=p=1 n+p=2

116
IMPORTANT 1ST CASE: “Empty set transversality”
f : M n → N m ⊃ W m−k , n < k, f 1 (Ω) = ∅ (transversal map).

Homework 8.
∑
1. Prove that the expression i pi dxi (locally in T∗ (N k )) defines correctly
1-form in T∗ (N k ) .
2. Prove that geodesic
∑ flow (Hamiltonian H(x, p) = g ij (x) pi pj / 2 ) pre-
i n
serves the form i pi dx at the level H = const in the manifold T∗ (M ) .

3. Prove that every closed k-form Ω in M n is homologous to the form

Ω′ − Ω = dω , such that Ω′ ≡ 0 in the disc near the point x0 :

111111111111
000000000000
000000000000
111111111111
000000000000
111111111111 x0
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111 Disk
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111

4. Prove that the Euler - Lagrange equation in T∗ (N k ) for the functional

∫ (∑ )
S {γ} = pi dxi − H dt ,
γ i

γ = {x(t), p(t)} , H = H(x, p) , is exactly a hamiltonian system

∂H ∂H
ṗi = − , ẋi =
∂xi ∂pi

5. Prove that the flow ẋi = ξ i (x) preserves the volume form
Ω = dx1 ∧ · · · ∧ dxn if
∑n
∂ξ i (x)
= 0 ( ⇔ St∗ Ω = Ω )
i=1
∂xi

117
 Homework 6 and Homework 7. Solutions.
Homework 6. Solutions.
1. Geodesics for Sn , Rn , Hn . Metric

a) dl2 = dρ2 + sin2 ρ (dΩ)2 , Sn ,

where (dΩ)2 is the metric of Sn−1 ,

b) dl2 = dr2 + r2 (dΩ)2 , Rn ,

c) dl2 = dχ2 + sh2 ρ (dΩ)2 , Hn .

Straight lines through zero η0 ∈ Sn−1 is geodesics (equation does not
depend on angles on Sn−1 ).

Other geodesics will be found by the action of groups SOn+1 (for Sn ),

SOn ∗ Rn (for Rn ), and SOn,1 (for Hn ).
2. Closed geodesic in T2 : straight lines with rational parameters
(0,1) (1,1)

m
y y = x + const
n

(0,0) x (1,0)

RP 2
η
ξ
3. Closed geodesic in RP2 (all nongomotopic to zero)
−ξ
−η

118
4. Parallel transport along the path γ(t) , x1 (t), . . . , xn (t) , connection
∇ gij ≡ 0 .

ξ (t)

η(t) γ (t)

d d ( )
⟨η(t) , ζ(t)⟩ = gij η i (t) ζ j (t) =
dt dt
( ) ( ) ( )
= ∇ẋ gij η (t) ζ (t) + gij ∇ẋ η (t) ζ j (t) + gij η i (t) ∇ẋ ζ j (t) = 0
i j i

(Leibnitz Identity).

Homework 7. Solutions.
1. ∫ ∫
S {γ} = γ
gij ẋi ẋj dt + γ
Ai (x) dt

↑ ↑

S0 (γ) S1 (γ)
A = Ai (x) dxi is one-form in M n .
( )
d ∂L1 ∂L1
EL(i) = EL0(i) + i
− = 0
dt ∂ ẋ ∂xi
( )
∂L1 d ∂L1 ∂Ai k ∂L1 ∂Aj j
i
= Ai (x) , i
= k
ẋ , i
= ẋ
∂ ẋ dt ∂ ẋ ∂x ∂x ∂xi
∑ ∂Ai ∑ ∂Aj
EL(i) = EL0(i) + k
ẋk
− i
ẋj =
k
∂x j
∂x
( )
∂Ai ∂Ak
= EL0(i) + − ẋk = EL0(i) + Bik ẋk
∂xk ∂xi
(dA = B).

119
2.
dφ
∫ z }| {
γ
k(s) ds = 2π n ?
γ - closed
φ - rotation of normal (and tangent)
γ (t) vector.
dẋ(s)
k(s) =
ds
s - natural parameter.
Another form:

→ S1 : Gauss Map : G∗ (dφ) = k(s) ds

G
γ −

(check!)
3. Area form
√
R2 : dx ∧ dy = r dr ∧ dφ , r = det gij
( )
1 0 √
dl2 2 2 2
= dr + r dφ , gij = , g = r
0 r2
4 dz ∧ dz̄
S2 : = sin θ dθ ∧ dφ
(1 + |z|2 )2
( )
2 4 dz dz̄ 1 0
dl = = dθ2 + sin2 θ dφ2 , gij =
(1 + |z|2 )2 0 sin2 θ
4 dz ∧ dz̄
H2 : = sh χ dχ ∧ dφ
(1 − |z|2 )2
( )
4 dz dz̄ 1 0
dl 2
= = dχ2 + sh2 χ dφ2 , gij =
(1 − |z| )
2 2 0 sh2 χ

4.
xdy − ydx y
dφ = const · , φ = arctg
x2 + y 2 x

120
 y
ϕ
x

5. Complex Line bundle Cx1 over M n , (x1 , . . . , xn ) :

M n = ∪α Uα , x1α , . . . , xnα

Let Uα = U , Uβ = V , we have in U ∩ V for the bases eU (x) , eV (x) :

UV
eU (x) = g U V (x) eV (x) , g U V (x) = eiφ (x)

- U1 - connection, group G = U (1) . We have for the connection AUi (x) dx

in U ∩ V : ( U V )−1 V U V ( )−1
AU = g A g − gU V dg
where ( )−1 ( )−1
gU V AV g U V = AV , gU V dg = i dφ(x)
Finally, we have for AUReal :

AUReal = AVReal − dφ(x)

Conclusion:
d AUReal = AVReal = H
(same for all U α ).
Obviously d H = 0 .
Remark 1.
a) H ∈ H 2 (M n , R) is the first Chern class of a line bundle.
b) It does NOT depend on the choice of connection.
Proof. Let us have ÃU = AU + iΨU , ÃV = AV + iΨV . We have then
in U ∩ V : ΨU = ΨV = Ψ . So we get, that Ψ is a globally defined 1-form
in M n and
H̃ = H + dΨ
where dΨ is an exact form in M n .

121
 c) ∫
1
H
2π 2−cycle

is integer ( ∈ Z).

Lecture 26. Transversality. Imbeddings and

Immersions of Manifolds in Euclidean Spaces.

Transversality.
f : M n → N m ⊃ W m−k
(along W )
df π m−k ∼ k
→ Tfm(x) −
Txn − → TNm /TW =R ,
rk(π ◦ df ) = k for all x ∈ f −1 (W ).

Example. n < k ⇒ f −1 (W ) = ∅.
Applications (imbeddings and immersions).
Theorem 1. Every map f : M n → RN can be approximated by NONDE-
GENERATE imbedding if N ≥ 2n + 1 (M n is compact C ∞ manifold).
Proof.
ϕ ψ
Step 1. Consider any C ∞ -map M n − → RN and C ∞ imbedding M n −
→ RQ .
Their product ϕ × ψ : M n → RN +Q = RP gives us C ∞ imbedding Φ :
M ⊂ RN +Q = RP , N + Q > 2n + 1.
n

Step 2. Project the imbedding M n ∈ RP into the space RP −1 along the

vector (direction) l (unit vector ±l in S N +Q−1 ).
n
M
l

Consider the projection πl ◦ Φ : M n → RP −1 .

πl

122
Lemma 1. For all vectors ±l outside of the image R : M n × M n \∆ → S P −1
x−y
R(x, y) = ∈ S P −1 , ±l ̸∈ R(M n × M n \∆).
|x − y|
the projection map πl ◦ Φ is imbedding.
Proof. πl (x) = πl (y) ⇒ ±l ∈ T (M × M ).
Lemma 2.
τ Consider the map T1∗ (M n ) −→
R 1
S P −1 , where
x (x, τ ) ∈ T1∗ (M n ), |τ | = 1, τ is tangent to M n at x.
R1 (x, τ ) = τ .
Let ±l ̸∈ Im R1 . Then the projection πl ◦ Φ : M n → RP −1 is an nonde-
generate immersion.
Proof. No one tangent vector to M n ∈ RP belongs to the kernel of the
projection if l not parallel to τ for all (x, τ ).
Lemma is true. O.K.
Proof of the Theorem 1. Let Φ be any imbedding M n ⊂ RN where
N > 2n + 1. We project it along the direction l, ±l ̸∈ Image(M × M \∆).
Almost all vectors have this property because dim Image(M ×M \∆) = 2n,
less then dim S N −1 ⊂ RN , N − 1 > 2n.
Here Transversality works. We apply transversality to the pair

M × M \∆ → S N −1 ⊃ (point).

Out Theorem follows (nondegeneracy see below).

Theorem 2. Every map M n → RN can be approximated by immersion, if
N ≥ 2n.
Proof is similar. Image of the map T ∗ (M ) → S N −1 , (x, τ ) → τ , τ ∈ Rnx ,
|τ | = 1 does not touch “typical” point (±l) ∈ S N −1 , because dim T ∗ (M ) is
equal to 2n − 1 < N − 1, if N > 2n.
So for N > 2n we can project imbeddings along πl and image in RN −1
remains an immersion M n → RN −1 .
So for N − 1 ≥ 2n we can project preserving immersion.
Theorem follows.

123
Example.

• N = 2n. Projection to RN −1 may be singular (not immersion).

n=1
N = 2n + 1. Projection to RN −1 may
•
be not imbedding. N=2

Singularities of projection:

R3

π: R1 R1 (projection from R2 )

R2
knot diagrams
πl

R1

Remark 1. Our results are true for all manifolds (C ∞ ):

a) M n → N N , N > 2n, imbeddings are dense.

b) M n → N N , N ≥ 2n, immersions are dense.

Remark 2. Our results are true for noncompact manifolds.

Lecture 27. Intersection Index and Degree

of Map.
Intersection Index:

f : M n → N m ⊃ W m−k , f transversal along W, k = n.

Intersection f (M n ) ∩ W m−k is “transversal” in the point x ∈ M if linear
spaces (df )x · Rnx ∈ Rmf (x) and Rf (x) (tangent to W ) jointly generate R
m−k m
m
(tangent to N in f (x)).

124
 N τW
W • τN - oriented frame to N .
f(M)
• τM - oriented frame to M .
τM
• τW - oriented frame to W .

“Sign” of intersection point xi = sign [τN /τM τW ].

xj

1. Intersection number:
∑
M ◦ W in N = (−1)sign(xj )
xj ∈f −1 (W )

2. Nonoriented Case: intersection number is from Z2 .

Degree of map.

f : M n → N n ⊃ W = (point), k = n = m.

f -transversal along W :

f −1 (W ) = x1 ∪ . . . ∪ xp ∈ M n ,

M n – oriented (closed). Degree of f [deg f ] is equal to

∑
deg f = (−1)sign(xj )
xj ∈f −1 (W )

( )
sign(xj ) : τN /τM = sign det df xj
,

df xj : Rntangent to M → Rntangent to N.
| {z }
linear map

deg f is a particular case of intersection number for the case dim W = 0

(W =point). If M and/or N are nonoriented,

degW f ∈ Z2 .

125
Theorem 1. Degree of map and Intersection Index are the same for homo-
topic maps.
Proof. Consider homotopy F such that

F : M × I → N ⊃ W, F t=0
= f, F t=1
= g, F ∈ C ∞ ,

and F is transversal along W .

Consider F −1 W . We have picture with 4 possibilities:
+ + −
2)
4) F
3)
1) W
+ − +

1. sign(x1 ) = − sign(x2 ).
x1 x2

x1 x2
2. sign(x1 ) = − sign(x2 ).

3. no point for t = 0, 1.

4. sign remains unchanged.

126
Conclusion. Total sum remains unchanged.
O.K.
Corollary (N = S n ): degree of map does not depend on the point W ⊂ N n .
Proof. Let N = S n . Rotate sphere S n ϕt , such that ϕ0 = 1, ϕ1 (W1 ) = W2 ,

Consider homotopy process

W2
ϕt ◦ f = F (x, t), F t=o
(x) = W1 , F t=1
(x) = W2 .
W1
O.K.
Examples.
1. Jordan Theorem.

A
γ1
S 1 ⊂ R2 (C ∞ -imbeddings). Two paths γ1 ,
B γ2
γ2 connecting a pair of points A, B.

γ (γ1 ∪ γ2 ) ◦ S 1 = 0 (Homework 9).

1 γ2
If γ1 ◦ S 1 = 1 ⇒ γ2 ◦ S 1 = ±1.

Conclusion. If γ1 ◦ S 1 = 1, then every path γ2 connecting A and B

crosses S 1 .

Remark. Intersection of 2 closed submanifolds M n ◦ W k ∈ Rm , m =

n + k is equal to 0.
2. Gauss Theorem: Let w = fn (z) be a polynomial in R2 = C. It defines
a map f : S 2 → S 2 , deg f = n, plus every point f −1 (w) is positive. So
we have exactly n points in f −1 (w) : z 1 , . . . , z n in transversal case.

127
 3. F : M n → N n – map of closed oriented manifolds, m = n = k, W -
point.

Theorem.
∫ ∫
∗
f (Ω) = (deg f ) Ω, (Why?)
| {z }
M integer N

Application in Geometry will be presented.

Lecture 28. Intersection Index and Degree

of Map.
Intersection index and degree of map are homotopy invariant.
Theorem. Let M n , N n be oriented (closed) manifolds and f : M n → N n
be a C ∞ -map. Then for every n-form Ω in N we have
∫ ∫
∗
f (Ω) = (deg f ) · Ω.
| {z }
M integer N

Proof. By transversality theorem, almost all points W ∈ N are “transver-

sal” along W ∈ N . The ∫set of transversal points in N is open and has mea-
sure=1. So the integral Ω depends only on the set of transversal points in
N
w ∈ N . Let U ∋ w be a small open set containing w such that all points in
U are transversal. Let U be connected set.
f
We have f −1 (U ) = U1 ∪ . . . ∪ Ul where Uj → − U is a diffeomorphism
preserving (+) or reversing (−) orientation with sign = (−1)sj , j = 1, . . . , l.
We have ∫ ∫
∗ sj
f Ω = (−1) Ω,
Uj U

and   ( )
∫ ∫ ∑
f ∗Ω =  Ω · s
(−1) j .
j
f −1 (U ) U

128
We have
∑
deg f = (−1)sj , (deg f calculated at the point w ∈ U ),
j

by definition.
We know that deg f does not depend on w: it is same for all transversal
points w ∈ N , and their measure is 1.
So we calculated ∫ ∫
∗
f (Ω) = (deg f ) · Ω.
M N

Theorem is proved.
Corollaries.

1. This Theorem is true also for manifolds with boundary

f : (M, ∂M ) → (N, ∂N ) f
N
M
assuming that f (∂M ) ⊂ ∂N .
M N
In this case deg fM = deg f∂M (Why?) (Homework 10).

2. Degree of map f : M → N is well-defined for “proper” maps f such

that f −1 (compact) is compact.
∫
If Ω is such that Ω < ∞, we have
N
∫ ∫
∗
f (Ω) = (deg f ) · Ω.
M N

(See examples of polynomial mappings)

f = Pn (x) : R → R , y = a0 xn + . . . + an .
(x) (x)

(See Homework 10)

What is “Linking number” for 2 submanifolds {M n , W k } in Rn+k+1 ?

129
a) {M n , W k } = M n ◦ B k+1 , ∂B k+1 = W k in Rn+k+1 .
b) {M n , W k } = Z n+1 ◦ W k , ∂Z k+ = M n in Rn+k+1 .
c) {M n , W k } = Z n+1 ◦ B k+1 in R+
n+k+2
.

Z B
M
n+k+2
W R+
M W M W
(here n−k=0)

Intersection Number (in manifolds).

1.
M n ◦ W k = (−1)kn W k ◦ M n (oriented case).

2. Self-intersection (M n = W n ).
M n ◦ M n =? in N 2n .
Perturb M = (M̃ ).

M
M ◦ M = M̃ ◦ M .
M
Corollary. M n ◦ M n = 0 = −M n ◦ M n , n = 2k + 1.
Modulo 2 (nonorientable case).
Example.

RP 2 ξ
−η

M 1 ◦ M 1 = 1.

η
−ξ

130
 Fixpoint (Lefschetz) problem:
f : M n → M n (orientable).
Solutions: f (x) = x.
Algebraic number of solutions: Let ∆ ⊂ M ×M , (x, x) ∈ ∆. “Lefschetz
Number” (x, f (x)) ∈ ∆f .
L(f ) = ∆f ◦ ∆.

Homework 9.
1. Prove that for all open 2-manifolds we have H 2 (M 2 , R) = 0 .
2. Prove that degree of map f : S1 → S1 can be calculated by
f (x + 2π) = f (x) + 2πm , m ∈ Z , m = deg f
(2π - circles).
3. Prove that intersection index of 2 closed curves in any domain U ⊂ R2
is equal to 0.
4. Prove that every map M n → Sm \W m−k is homotopic to zero for n <
k−1 .
5. Prove that every map Sn → Sm is homotopic to zero for n < m .
6. Calculate cohomology ring H ∗ (SO4 , R) = ?
7. Calculate H 1 (M 2 , R) = ? for M 2 = R2 \(∗1 ∪ ∗2 ) :
R2
*1 *2

Lecture 29. Intersection Index and Degree

of Map.
Intersection Index, Degree of Map.
For M n ◦ W k ⊂ Rn+k we have Intersection Index.

131
 For f : M n → N n we have deg f.
They have the following properties:

1. They are homotopy invariant.

2. M n ◦ W k = (−1)kn W k ◦ M n .

3. M ◦ M = 0 for n = k = 2l + 1 (oriented case, M ◦ M ∈ Z).

4. In nonoriented case M n ◦ M n may be ̸= 0.

5. Linking number {M n , W k } in Rn+k+1 .

M = ∂Z, W = ∂B, M ∩ W = ∅, {M n , W k } = M ◦ B = Z ◦ W.

Let M, W ∈ Rn+k+1 = ∂Rn+k+2

+
Z B
We have B ◦ Z = {M, W }.

M W M W
Gauss formula:
Let M = S 1 ∈ R3 , W = S 1 ∈ R3 .
M
dγ 1 dγ 2 dγ 3
W det dx 1
dx 2
dx3
γ 1 − x1 γ 2 − x2 γ 3 − x3
∥
γ (t) I I
x(s) 1 (d⃗γ (t) × d⃗x(s), ⃗γ (t) − ⃗x(s))
{M, W } =
M = γ(t), W = x(s), 4π |⃗γ (t) − ⃗x(s)|3
M W
γ(t) ̸= x(s),
ϕ = {γ(t), x(s)},
Is it closed 2-form in R3 × R3 \∆?
ϕ : T 2 → R3 × R3 \∆.

Fixed Point Problem.

f : M n → M n , f (x) = x (?)

132
Algebraic (Lefschetz) number of fixpoints:

L(f ) = ∆ ◦ ∆f in M n × M n , (x, x) ∈ ∆, (x, f (x)) ∈ ∆f ,

∆ = ∆f ⇒ f (x) = x.

1. Special case f ∼ 0 : M n → const.

∆ ◦ δf = 1.

Example. M n = S n , ϕ : S n → Dn (south hemisphere).

ϕ f ψ
Map: S n −
→ D−
n
−
→ D−
n
−
→ Sn

Sn D n− g = ψ ◦ f ◦ ϕ ∼ 0 (obvious).

S’

Conclusion: ∆ ◦ ∆f = 1. There exists fix point (at least one).

2. Special case f ∼ 1 : M → M (identical). We have

{
0, n = 2k + 1,
∆ ◦ ∆f = ∆ ◦ ∆ =
χ(M n ), n = 2k.

Example: “Euler Characteristics” χ(M n ).

X
St (X) : M n → M n , X − vector field,

St (X) ∼ 1, M n ⊂ T ∗ (M n ) M

133
 n
M =(M ,tX)
χ(M n ) = M̃ ◦ M =?
(M n,0)

Consider “gradient” vector (covector) field df for f : M → R.

T∗ (M ) ∼
= T ∗ (M ), we use T∗ (M ).

M̃ ◦ M =?
Critical point: (df ) xk
= 0. Critical point xk is nondegenerate iff the
quadratic form ( )
2 ∂ 2f
df= dxi dxj
∂xj ∂xj xk

is non-degenerate.
“Sing” of xk is equal to the number of negative squares in the form
(d2 f )xk , which is the “Morse index” n(xk ).

signf xk = (−1)Morse Index(xk )

= (−1)m(xk ) .

Lemma. Intersection Index M ◦ M in T∗ (M ) is equal to:

∑
n
M ◦M = (−1)j mj (f ) = χ(M n ) for orientable manifolds,
j=0

mj (f ) = ♯points xk with Morse Index = j.

Example, n = 2. χ(M 2 ) = m0 (f ) − m1 (f ) + m2 (f ), m0 – minima, m1 –

saddles, m2 – maxima.
Proof of Lemma. Variation of manifold M given by vector field X is
SX (t) : x → x + tX(x) + O(t2 ) where t ↘ 0. For the fix point X(x) = 0 we
have:
( )
∂X i
SX (t) : ⃗x → ⃗x+tA⃗x+O(t ), where xk = (0, . . . , 0) and A = (Aj ) =
2 i
.
∂xj xk

Now calculate intersection index M̃ ◦ M in T∗ (M n ) = T ∗ (M n ).

134
 T*(M) τ’
M
M
M M
xk
τ

Local coordinates in T ∗ (M ) are (x1 , . . . , xk , η 1 , . . . , η k ), ⃗η – tangent vector.

Basis is: τ1 = ∂
∂x1
, . . . , τn = ∂
∂xn
, τ1′ = ∂
∂η 1
,. . . , τn′ = ∂
∂η n
.

Tangent space to M ⊂ T ∗ (M ) is span(τ1 , . . . , τn ) at the point xk (fixpoint

for X).
Tangent space to M̃ = SX (t)M for t ↘ 0 is span(τ1 + tAτ1′ , . . . , τn + tAτn′ )
(neglecting term O(t2 )).
(τ1 + tAτ1′ , . . . , τn + tAτn′ , τ1 , . . . , τn ) − basis M̃ ◦ M.
sing is equal to sign det A.
O.K.
Let gij X j = df and gij (xk ) = δij . We have
( ) ( 2 )
∂X i ∂ f
sign det A x = sign det = sign = (−1)♯ negative squares .
k ∂xj xk ∂xi ∂xj xk
Lemma is proved.
So Euler Characteristics is:
1. Total algebraic singularity of vector field.
2.
∑
n
χ(M n ) = (−1)j mj (f ), for f : M → R.
j=0

mj (f ) = ♯points xk with Morse Index = j.

Example. Maxwell (XIX century).

Floating island, no beaches.

under m0 − m1 + m2 = 1.
water

135
 Lecture 30. Two applications of differential
forms: Degree of Map and Hopf Invariant.
Two applications of differential forms.
f
I) Homotopy theory: Degree of map S n −
→ S n (spheres may be replaced
by any closed orientable manifolds:)
∫
A) Take n-form Ω ∈ ∧n (S n ), Ω = 1.
Sn

Definition. ∫
deg f = f ∗ (Ω)
Sn

Proof.
Let F be a homotopy from f to g: F : S n ×I → S n , (0 ≤ t ≤ 1), F (x, 0) =
f , F (x, 1) = g. Consider F ∗ (Ω)). We have dF ∗ (Ω) = F ∗ (dΩ) = 0.
∫ ∫ ∫
∗ ∗
F (Ω) − F (Ω) = dF ∗ (Ω) = 0.
x x
(S n ,1)
 (S n ,0)
 (S n ×I)
∗ ∗
g Ω f Ω
∫ ∫
Ω → Ω + dω ⇒ Ω + dω = Ω.
(S n ,t) (S n ,t)

Theorem is proved.
∫
We calculated integral f ∗ Ω geometrically and proved geometrically
( ) S n

∫ ∗ ∫
that f Ω = Ω · deg f .
Sn Sn
∫
B) Consider S 3 → S 2 and fix 2-form Ω in ∧2 (S 2 ) such that Ω = 1.
S2

Let f ∗ Ω = dω, ω ∈ ∧1 (S 3 ).

Define “Hopf Invariant”

∫
H(f ) = ω ∧ f ∗ Ω.
S3

136
 Theorem. H(f ) is homotopy invariant.
∫
Proof. Consider F : S 3 × I → S 2 , Ω = 1. Take F ∗ (Ω) ∈ ∧2 (S 3 × I).
S2
Take

F ∗ (Ω) = dω, ω ∈ ∧2 (S 3 × I), ω ′ = ω t=0

, ω ′′ = ω t=1
.

We have

d(ω ∧ F ∗ (Ω)) = dω ∧ F ∗ (Ω) = F ∗ (Ω) ∧ F ∗ (Ω) = F ∗ (Ω ∧ Ω) = 0.

So ∫ ∫ ∫
∗ ∗
ω ∧ F (Ω) − ω ∧ F (Ω) = (F ∗ Ω)2 = 0.
t=1
x t=0
x S 3 ×I
 
∫ ∫
′′ ∗
ω ∧g Ω ω′ ∧ f ∗Ω
S3 S3

Here
F t=1
= g, F t=0
= f.

How to calculate H(f ) geometrically?

Let Ω → Ω′ = Ω + dv: H(t) is the same.
Let ω → ω ′ = ω + u, du = 0: H(t) is the same.
Theorem is proved.

II) Riemannian Geometry: Gaussian Curvature.

Let M n ⊂ Rn+1 (x0 , x1 , . . . , xn ) and M n is locally (near P ∈ M n ) defined
by equation:

z
x0 = z(x1 , . . . , xn ), P = (0, 0, . . . , 0), (dz) P
= 0.
n(P)

Consider map (Gauss).

P
G
x Q−
→ n(Q) ∈ S n ⊂ Rn−1 , Q ∈ M n .

137
Definition dσ – volume element at S n .

K ∗ = G∗ dσ.

n-form K ∗ is the “Gauss form” or “Curvature Form”. Let dσM is a volume

element of M n induces from M n ⊂ Rn Riemannian Metric. We have by
definition
K ∗ = KdσM .
In our system of coordinates z = x0 = z(x1 , . . . , xn ), z ⊥ M n (at P ), we have
gij (P ) = gij (0) = δij and (∂gij /∂xq ) = 0. So we have

(dσM )P = dx1 ∧ . . . ∧ dxn , K ∗ = Kdx1 ∧ . . . ∧ dxn ,

and K is ordinary Gauss Curvature:

( 2 )
∂ z
KP = det , P = (0, 0, . . . , 0).
∂xi ∂xJ P

Theorem 1. K ∗ = KdσM = G∗ (dσ), where dσ is invariant volume element

in S n ⊂ Rn+1 .
Proof. Let us describe volume element dσN in S n . Let S n be given by
equations (N =“north pole”)
w2 +√ u21 + . . . + u2n = 1,
N
n w = 1 − (u1 )2 − . . . − (un )2 ,
S N = (0, 0, . . . , 0),
w u1 , . . . , un – local coordinates.
u
Construct Gauss map:
G : Q → nQ ∈ S n , Q = (x1 , . . . , xn )
z M n : −z(x1 , . . . , xn ) + x0 = 0,
nQ – normal vector.
n P Q
M (1, −zx1 , . . . , −zxn )
nQ = √
1 + (zx1 )2 + . . . + (zxn )2
x
We see that P → N , where P : ⃗x = 0.

138
 The volume element of sphere at N is:

dσ = du1 ∧ . . . ∧ dun , u − local coordinates in S n , u0 = w.

Map:
(1, −∇z )
⃗u = nQ = √ ,
1 + (∇z )2
1 −∂z/∂xj
w=√ , uj = √ , j = 1, . . . , n.
1 + (∇z )2 1 + (∇z )2

Calculate Jacobian:
( )
∂uj
J0 = det , x = 0, 0, . . . , 0 .
∂xk T

J0 =?. Remember that (∇Z)P ≡ 0. So we have:

( ) ( 2 )
∂⃗u ∂ z
J0 = det = (−1) · det
n
,
∂⃗x (0,...,0) ∂xi ∂xi (0,...,0)

So we have
( )
∗ ∂ 2z
KdσM = G (dσ) = (−1) · det n
dx1 ∧ . . . ∧ dxn .
∂xi ∂xi (0,...,0)

For n = 2k we have (−1)n = 1.

Let n = 2. By definition
( 2 )
∂ z
K = det , where z = x0 = z(x, y), ∇Z T
= 0.
∂xi ∂xi T

Theorem is proved.
Conclusion: ∫ ∫∫
Kdσ = (deg G) · dσ.
M2 S2

How to calculate deg G? Let n = 1, 2. For n = 2 we expressed Gauss

Curvature through the Curvature Tensor (Riemannian).

139
 How to calculate deg G in R3 (the same consideration can be done for
any compact M n ⊂ Rn+1 ).
M 2 ⊂ R3 , G : M 2 → S 2 , h : M 2 → R.
z
N
Lemma 1. Let S, N ∈ S 2 (the
South and the North poles) be
transversal points for G. Then
∇h = 0 iff Pj ∈ G−1 (N ) or Pj ∈
G−1 (S).
S

Lemma 2. For the function z on M n we have1 :

G−1 (N )
signPj = (−1)Morse Index in

G−1 (N )
signSj = (−1)n+Morse Index in

Calculation:
∑
n {
n j n 0, n = 2k + 1,
χ(M ) = (−1) mj (z) = degN G+(−1) degS G =
2 deg G, n = 2k.
j=0

Lecture 31. Comparison of notations of our

Lectures with book of Do Carmo “Rieman-
nian Geometry”.
We recommend book Do Carmo (D.C.) “Riemannian Geometry”. Let us
make some comparison of notations and prove some theorems (f.i. “local
minimality” of geodesics).
Manifolds:
1
Morse Index of z(⃗x) at P , (∇z)P = 0 in equal to the ♯ of negative squares in the form
z
(d )P .

140
 D.C. “Differential Structure”, Hausdorff.
Our Course: “Atlas of Charts”, C ∞ manifolds, metric spaces, “Dou-
ble good Atlas”.
Local coordinates - same.
Imbedding and Immersions M n = RN (proof of existence is missing in
D.C.).
Partition of unity (proof is missing in D.C.) - we proved for compact
manifolds.
Tangent vectors, basis of Tτn as ∂i = ∂/∂xi in local coordinates
(x1 , . . . , xn ) .
∑ i
Vector field X = a (x) ∂i (locally).
Our notation X = (ai ) (index is “upper”). Action of vector field on
functions f (x) :
∑ ∂f ∂f
D.C. : X (f ) = ai (x) i
= ai i
∂x ∂x
Commutator
[X , Y ] = X (Y (t)) − Y (X(t))
( i i
)
j ∂b j ∂a
i i i
X = (a ) , Y = (b ) , [X , Y ] = a −b
∂xj ∂xj
f
C ∞ -maps : M n −
→ N m , differential of map, rank :

d fτ : Tτn → Tfm(τ ) (linear)

Implicit function theorem / local inversion theorem of the map (no proof
in D.C. and in our course).
Approximation and Transversality are missing in D.C. or presented in the
highly reduced form - see C.P.
Examples of manifolds - similar.
Orientation.
i
Covectors∑(basis (dxi ) in local coordinates) are missing in D.C. Covector
fields ω = i ui (x) dx (1-forms).
Manifolds T ∗ (M n ) (vectors), Atlas, Charts.
Manifolds T∗ (M n ) (covectors, missing in D.C.)

141
 Riemannian and Pseudoriemannian Metric = inner product (symmetric,
nondegenerate) is ⟨X , Y ⟩ , gij = ⟨∂i , ∂j ⟩ .
Symplectic inner products and differential forms are missing in
D.C.
√
Volume element dn σ = det gij dx1 ∧ · · · ∧ dxn (locally).
Induced Metric for submanifolds M n ⊂ RN or M n ⊂ N k (good only
for positive Riemannian Metrics).
Lie Groups, right and left invariant vector fields, right and left invariant
Riemannian Metric on Compact Lie Groups, bi-invariant metric and inner
product of right-invariant vector fields (D.C.).
Riemannian Metric and length of curves. Existence of Riemannian metric
in every manifold (proved by partition of unity in D.C.; another proof -
imbedding in RN ). Volume provided by Riemannian metric.
Existence of biinvariant volume on a compact group G. Existence of
invariant metric in the spaces with action of compact group G (“averaging
procedure” - both in D.C. and our course).
Connections (“affine”) - considered in D.C. for tangent bundles only:
vector field X, X = ai (x) ∂i ,
∑
∇X Y = ai ∇ i Y , Y = bj (x) ∂j
i
( )
∂bj ∂bj
∇i Y = b (∇i ∂j ) +
j
∂j = + bj Γkij ∂j
∂xi ∂xi
for components, ∇i (∂j ) = Γkij ∂k .
Symmetric Connection” Γkij = Γkji {∇i ∂j = ∇j ∂i } .
Curve
c(t) = {xj (t)} , X = (ẋ)
Let V be a vector fields, V = (v j ) , X(t) - tangent to c.
( j )
d dv
(V ) = ∇X V = q k j
+ v ẋ Γqk ∂k
dt dt

D.C. : Xj ↔ ∂j , ∇Xi Xj = Γkij ∂k , X = (ẋj (t)) along c(t) .

Tensors. Examples:

142
 1. Scalars (type (0, 0)) - f (x) .
∑ i
2. Vectors (type (1, 0)) - (ui ) u ∂i .
∑
3. Covectors (type (0, 1)) - (vi ) vi dxi .
4. Inner products of vectors gij (type (0, 2)) .
5. Inner products of covectors g ij (type (2, 0)) .
6. Operators (aij ) (type (1, 1)) .
General Tensors (type (k, l)) .
Components Tji11...j ...ik
l
(x)
e1 , . . . , en - basis of tangent vectors
e1 , . . . , en - basis of covectors
ei1 ⊗ ei2 ⊗ . . . ⊗ eik ⊗ ej1 ⊗ ej2 ⊗ . . . ⊗ ejl - basis in tensor space
(k, l).
Operations : Linear, Tensor product

(k, l) ⊗ (p, q) ⊂ (k + p, l + q)
′ ′′ ′ ′′
Tj′i′ ⊗ Tj′i′′ = Tji′ ji ′′ (x)
Permutation of indices (lower or upper)
Trace
′ ′′ Trace ′ ′′
∑
n
′ ′′
Tji′ jji i′′ −−−→ Tji′ ji ′′ = Tji′ ii ji ′′
i=1

Examples: A = (aij ) , aij → aij = Tr A (trace), gij → gji - permuta-

tion.
Extension of connection to all tensor fields. Axioms.
a) Trivial for scalar fields ∇i f = ∂i f .
b) Satisfies to Leibnitz formula for tensor products.
c) Annihilates Riemannian metric ∇i gkl ≡ 0 .
d) Commutes with trace

∂
∇X ⟨V , W ⟩ = ui ⟨V , W ⟩ = ⟨∇X V , W ⟩ + ⟨V , ∇X W ⟩
∂xi

143
where X = (ui ) ,
( )
∇k ⟨V , W ⟩ = ∇k gij v i wj = ⟨∇k V , W ⟩ + ⟨V , ∇k W ⟩

∇k = ∇Xk , Xk = ∂k .
For parallel Transform
d dV dW
⟨V , W ⟩ = ⟨ , W ⟩ + ⟨V , ⟩
dt dt dt
dV dW d⟨V , W ⟩
= 0, = 0 ⇒ = 0
dt dt dt

Levi - Civita Theorem. Symmetric connection compatible with metric

gij is given by the formula
( )
1 pi ∂gij ∂gik ∂gjk
p
Γjk = g + −
2 ∂xk ∂xj ∂xi

“symmetric”: ∇k ∂j = ∇j ∂k ↔ ∇X Y − ∇Y X = [X , Y ] .

Homework 10.
1. Prove that for complex polynomial

w = Pn (z) : S2 → S2

we have deg f = n .
2. Calculate degree of rational map

Pn (z)
w = , (irreducible fraction)
Qk (z)

3. Calculate degree of real polynomial

y = a0 xn + a1 xn−1 + . . . + an : R → R , aj ∈ R

4. Let γ(t) = γ(t + 2π) : S1 → R2 and γ̇ ̸= 0 .

144
 γ (t) f(x)
f

Consider “Gauss Map” :

f
S1 →
− S1 , f (γ̇) = γ̇/|γ̇| ∈ S1

Prove that f ∗ (dφ) = k(s) ds , where s - natural parameter, k - curva-

ture. Calculate deg f = ?
5. Consider 2-form in R3 \0 :

Ω = x dy ∧ dz − y dx ∧ dz + z dx ∧ dy

Prove that ( )
Ω
d = 0
(x + y + z 2 )3/2
2 2

Prove that
Ω|unit sphere = area element in S2
which is SO3 - invariant.

Homeworks 8, 9. Solutions.
Homework 8. Solutions.
∑
1. Prove that i pi dxi is 1-form in T∗ (N k ) (local coordinates (xi , pj )).
Change of coordinates x = x(y) ,

∂xi
(system x) pi = p′j (system y)
∂y j
So we have
∑ ∑ ∂xi ∑
p′j dy j = pi dy j
= pi dxi
j i,j
∂y j i

∑
Remark. dpi ∧ dxi is a well-defined 2-form also (done before).

145
 ij
2. Prove that geodesic
∑ flowi with Hamiltonian H(x, p) = g (x) pi pj / 2
preserves
∑ the form i pi dx [correction due to student: it preserves the
form i pi dxi restricted to the (2n−1) - dimensional submanifold T1 (M n ) ,
|ẋ| = 1 (unit tangent vectors)].
Proof. Calculation shows that for the shift St along the flow we have for
geodesic flow ( )
d ∗ ∑
St pi dxi = dH
dt i t=0
For geodesic flow have
1 ij 1
H(x, p) = g (x) pi pj = gij (x) ẋi ẋj
2 2
so T1 (M n ) is given by the condition H = 1 and dH = 0 on T1 (M n ) .
Calculation: We have

St∗ (pi ) = pi + t ṗi + O(t2 ) , St∗ (xi ) = xi + t ẋi + O(t2 )

For ẋi = Hpi , ṗi = − Hxi we have then

d ∗ d ∗ i
S (pi ) = − Hxi , S (x ) = Hpi
dt t t=0 dt t t=0

d ∗ i
S (dx ) = dHpi = Hpi pj dpj + Hpi xj dxj
dt t t=0
Finally:
( )
d ∗ ∑ [ ]
St pi dxi = − Hxi dxi + pi Hpi pj dpj + Hpi xj dxj =
dt i t=0
( )
∂H
= d pi − H = dL
∂pi
where L is a “Lagrangian”.
For
1 ij 1
H(x, p) = g (x) pi pj = gij (x) ẋi ẋj
2 2
we have H = L and also dH = dL .

146
 2 is proved.
3. Locally every closed form is exact: ω = dν in the domain U ⊂ M n .
Let U is the ball |x|2 < ϵ . Multiply ν by function φ (C ∞ , φ ≡ 1 in U
and φ ≡ 0 outside of the ball |x|2 > 2ϵ ).
Our form is ω − d(φν) .
4. Let ∫ (∑ )
S {γ} = pi (t) ẋ (t) dt − H(p, x) dt
i
,
γ i
∑
γ = {x(t), p(t)} , L = pi ẋi − H(x, p) . We have from the Euler
Lagrange equations
( )
d ∂L
∂L ∂H
ṗi = = − i =
dt ∂ ẋi
∂x i ∂x
( )
∂H ∂L d ∂L
ẋi − = = = 0
∂pi ∂pi dt ∂ ṗi
Remark. We can write
∫
( )
S {γ} = d−1 (Ω) − H dt
γ

where Ω is the simplectic form.

In our case
∑ (∑ )
Ω = dpi ∧ dxi = d pi dxi

If cohomology class [Ω] ∈ H 2 (M 2n ) is nonzero (Dirac Monopole, Com-

pact simplectic Manifolds) the expression S{γ} defines correctly a “Mul-
tivaled Functional” or closed 1-form on the space of paths γ (f.i. closed
paths).
5. Consider volume element dx1 ∧ · · · ∧ dxn = Ω and the flow ẋi = ξ i (x) .
We have
d ∗
(S Ω) =
dt t t=0
1[ ]
= lim d(x1 + tξ 1 ) ∧ · · · ∧ d(xn + tξ n ) − dx1 ∧ · · · ∧ dxn =
t→0 t
( 1 )
∂ξ ∂ξ n ⃗
= 1
+ ... + n
dx1 ∧ · · · ∧ dxn = div ξ(x) ·Ω
∂x ∂x

147
 Homework 9. Solutions.
1. Let M 2 = R2 \(point) . It is homotopy equivalent to S1 , so H 2 (M 2 ) =
H 2 (S1 ) = 0 .
Every open 2-manifold is homotopy equivalent to “graph” like
111111111
000000000
000000000
111111111
000000000
111111111
000000000
111111111
000000000
111111111
000000000
111111111
00000
11111
000000000
111111111
00000000
11111111
00000
11111
000000000
111111111
00000000
11111111
00000
11111
000000000
111111111
00000000
11111111
00000
11111
B2 D2
00000000
11111111
00000000
11111111
00000000
11111111
00000000
11111111
00000000
11111111 D2
00000000
11111111
00000000
11111111

For every 2-form in D2 we have ω = dν (in D2 ).

Construct function φ ≡ 1 in D2 and φ ≡ 0 outside of domain B 2 ⊃
D . So ω ∼ ω − d(φν) = ω ′ . And ω ′ is nonzero only in strips
2

1111111111
0000000000
0000000000
1111111111 strip
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0000000000
1111111111
0 1111111111
0000000000

Every such strip is homotopy equivalent to circle. Our result follows.

2. Degree of map S1 → S1 , m = deg f ?
f (x + 2π) = f (x) + 2πm , m ∈ Z

Proof.

4π 4π
Inverse
a image
deg=2
2π 2π of point (a)
a appears twice

0 2π 0 2π

2 is proved.
3. Deform one circle far: S1 (t) , 0 ≤ t ≤ 1 , S2 (0) = S2 . We have
S1 ◦ S2 (0) = S1 ◦ S2 (1) = Ø

148
 S1(0) S1(1)
S2(1)
S2(0)

4. Every map M n → Sn+k is homotopic to zero for k > 0 because it is

homotopic to C ∞ (approximation), and image of C ∞ -map does not cover
Sn+k = Sm . Homotopy process
F
Mn × I −
→ Sm

can be approximated by smooth (C ∞ ) map F transversal along W m−k .

Here m = n + k . For n < k − 1 we have

F (M n × I) ∩ W n−k = Ø

So our statement follows.

5. Same argument for f : Sn → Sm , n < m : C ∞ - map can not cover
Sm if m > n .
6. H ∗ (SO4 , R) = ? We have SO4 = S3 × S3 / ± (1, 1) . Group G =
SO4 ×SO4 . G - invariant torus in SO4 appear (as well as for S3 ×S3 ) only
in dimensions 0, 3, 6 . So we have for H ∗ (SO4 , R) in different dimensions:
 
 1 , ω1 , ω2 , ω1 ∧ ω 2 
H ∗ (SO4 , R) :
 
D=1 D=3 D=3 D=6

H0 = R , H3 = R ⊕ R , H6 = R .
7. We have
1111111111
0000000000 000000000
111111111
000000000
111111111
111111111111
000000000000 0000000000
1111111111
00000
11111
000000000
111111111
000000000000
111111111111 0000000000
1111111111
00000
11111
0000000000
1111111111 000000000
111111111
000000000000
111111111111
000000000000
111111111111 00000
11111
0000000000
1111111111 000000000
111111111
R2 \(∗1 ∪ ∗2 ) ∼ 000000000000
111111111111 ∼ 00000
11111
=1111111111
0000000000 000000000
111111111
00000
11111
= 000000000000
111111111111 0000000000
1111111111 000000000
111111111
000000000000
111111111111
000000000000
111111111111 0000000000 111111111
1111111111 000000000
2
D D 2 B2
Use arguments same as in Problem 3 above for calculation of H 2
( )
H 1 R2 \(∗1 ∪ ∗2 ) = R ⊕ R

149
 Lecture 32. Geodesics. Gauss Lemma.
1. Geodesics: smooth curves γ(t) = {xi (t)} such that
D d⃗x
≡ 0 ( or ∇ẋ ẋ = 0 )
dt dt
Equation
ẍi + Γikj ẋk ẋj = 0

Calculus of variation:
a) Length
∫ b ∫ b
l(γ) = |ẋ(t)| dt = L0 (x, ẋ) dt
a a

where √
L0 (x, ẋ) = gij (x) ẋi ẋj
is a “Lagrangian”.
b) “Action”
∫ b ∫ b
1
S(γ) = gij (x) ẋi ẋj dt = L1 (x, ẋ) dt
a 2 a

Euler - Lagrange equation (EL)

d ∂L ∂L
=
dt ∂ ẋi ∂xi
Momentum:
∂L
pi =
∂ ẋi
Energy:
E = pi xi − L
Conservation of energy:
dE
≡ 0
dt
For L = L1 we have
1 1
E = L1 = gij (x) ẋi ẋj = ∥ẋ∥2
2 2
150
 The condition E = const means that the parameter t along geodesics
is NATURAL (∼ length).
For L = L0 with natural parameter and L1 Euler - Lagrange equa-
tions coincide.
Exponential Map:
V - domain in TP (M n ) = Rn , P = M n .
v
0
expP : V → M n , v ∈ TP , expP (V ) : exp P(v)
P geodesics
“Geodesic ball”: Bϵ (0) - ball in TP ,

expP (Bϵ (0)) − geodesic ball in M n

It exists for ϵ , - small enough (depends on P ∈ M n ).

For compact M n it exists for any radius.
For symmetric connection we have ∇i ∂j = ∇j ∂i (or ∇Xi Xj = ∇Xj Xi
for Xi = ∂i ). Let s (u, v) be a parametrized surface

s = {x1 (u, v), . . . , xn (u, v)}

Lemma 1.
D ∂s D ∂s
=
∂v ∂u ∂u ∂v
where
∂s ∂xi ∂s ∂xj
= ∂i = X , = ∂j = Y
∂u ∂u ∂v ∂v
or ∇X Y = ∇Y X at s.
Proof. Vector fields X, Y commute at the surface s, so [X, Y ] = 0 at s.
But we have
∇X Y − ∇Y X − [X , Y ] = 0
at s.
Lemma is proved.

151
Gauss Lemma. Let expP (v) be defined for v ∈ TR = Rn , and w ∈ TP .
Then we have

⟨(d expP )v (v) , (d expP )v (w)⟩ = ⟨v , w⟩

Proof. For v ∥ w lemma is obvious because γ(t) = expP (tv) is

geodesics for 0 ≤ t ≤ 1 .
Let now w ⊥ v , so ⟨v, w⟩ = 0 . Put

w
dv
w = , v(0) = v, |v(s)| = const
ds s=0
v
v(s)

For small ϵ > 0 all u = t v(s) , |s| < ϵ , have well-defined expP (u) ,
0 ≤ t ≤ 1.
Consider surface f : A → M n , parametrized by (t, s) :

A : f (t, s) = expP (tv(s)) , 0 ≤ t ≤ 1 , |s| < ϵ

where f (t, s0 ) - geodesics through P .

By definition

∂f ∂f
⟨ , ⟩ = ⟨(d expP )v (w) , (d expP )v (v)⟩
∂s ∂t t=1,s=0

We have also
∂ ∂f ∂f D ∂f ∂f ∂f D ∂f D ∂f ∂f
⟨ , ⟩ = ⟨ , ⟩ + ⟨ , ⟩ = ⟨ , ⟩ =
∂t ∂s ∂t dt ∂s ∂t ∂s dt ∂t dt ∂s ∂t
D ∂f ∂f 1 ∂ ∂f ∂f
= ⟨ , ⟩ = ⟨ , ⟩ = 0 ,
ds ∂t ∂t 2 ∂s ∂t ∂t
∂f ∂f
so ⟨ , ⟩ is t − independent.
∂s ∂t
Calculate it for t → 0 :
We have
∂f
(t, 0) = (d expP )tv (tw)
∂t

152
and
∂f
(t, 0) → 0 , t → 0 .
∂s
∂f ∂f ∂f ∂f
So ⟨ , ⟩ ≡ 0 for all t , and ⟨ , ⟩ = 0
∂s ∂t ∂s ∂t t=1,s=0

Lemma is proved.
Theorem. Let ϵ > 0 is small enough and expP (Bϵ (0)) ⊂ M n is a geodesic
ball. Let c(t) be a smooth curve joining γ(0) and γ(1) where γ is a
geodesics started in P .
γ (t)

P
c(t)
Then length of c(t) is: l (c) ≥ l (γ) , and l (c) = l (γ) implies c = γ .
Proof. Let c ⊂ Bϵ (round geodesic ball). Curve c(t) can be written as

c = expP (r(t) · v(t)) , |v(t)| = 1 ,

where v(t) is a curve in tangent space TP , and r(t) is such that 0 < r ≤ 1 .
Let c(t1 ) ̸= P for all t1 ∈ (0, 1] and r(t) > 0 for all t1 ∈ (0, 1] .
We have for

f (r(t), t) = expP (r(t) · v(t)) , |v(t)| = 1 :

dc ∂f ′ ∂f
= r (t) +
dt ∂r ∂t
By Gauss Lemma (above) we have

∂f ∂f
⟨ , ⟩ = 0
∂r ∂t
We have also |∂f /∂r| = 1 , so
2 2
dc ′ 2 ∂f
= |r (t)| +
dt ∂t

153
and ∫ 1 ∫ 1
dc
dt ≥ |r′ (t)| dt ≥ r(1) − r(0)
ϵ dt ϵ
For ϵ → 0 we have r(0) = 0 . So we get
l (c) ≥ r(1) = l (γ)
At the same time the condition ∂f /∂t ≡ 0 means v(t) = v(0) = const ,
i.e. c = γ (geometrically for r′ (t) ≥ 0 , otherwise: l (c) > l (γ) ) .
Theorem is proved.

Lecture 33. Local minimality of geodesics.

Last lecture: For every point P ∈ M n there exists an ϵ-ball (geodesic) B
containing P as a center such that geodesics is a shortest line between P
and any point Q ∈ B .
Corollary 1. Let X be compact space and f, g : X → M n such 2 maps
that distance ρ (f (x), g(x)) is smaller than the minimum of all ϵ (P ) for all
P ∈ f (X) . Then f and g are homotopic.

γ
Proof. Move f (x) to g(x) monotonically along g(x)
unique minimal geodesics joining these points
f(x)
Corollary is proved.
Corollary 2. Let M n be compact (or metrically complete) Riemannian
manifold. For every 2 points P, Q with distance ρ (P, Q) = d there exists
geodesics γ , l(γ) = d , joining P and Q .
Proof. Let us remind that d(P, Q) = minγ̃ l (γ̃) , γ̃ joins P and Q and
γ̃ is smooth or piecewise smooth curve:
γ
Q
P

Let us remind you the “Arcellát Principle” claiming that the set of piece-
wise smooth curves of length ≤ Γ is “precompact”. It means in particular
that every “Cauchy Sequence” of curves for which length is well-defined.

154
 Consider sequence of piecewise smooth curves γn such that l (γn ) → d ,
d = ρ (P, Q) . Limit curve γ∞ is such that l (γ∞ ) = d , but it can be not
smooth (even not piecewise smooth).
Take 2 points P1 , P2 at γ∞ such that ρ (P1 , P2 ) < ϵ (small enough).
Q
P P1 P2 γ (t)

Replace γ∞ from P1 to P2 by small geodesics joining them. By con-

struction new curve is shorter than γ∞ or γ∞ is geodesics between P1 and
P2 . So, our statement follows.
Corollary is proved.
Remark. In the book D.C. the “Hopf - Rinow” theorem is proved for this
result. In particular, it claims that expτ is defined for all vectors in Tτ (M n )
in complete Riemannian manifold. For compact manifold it simply follows
from the fact that manifold T1 (M n ) is compact and geodesic flow is well
defined for all time.
However, we presented here the “Arcellát” argument which was used in
many other variational problems since the great work of Hilbert who was
first to prove existence of minimizing geodesics in XIX Century.
See Hopf - Rinow theorem in D.C.
Now we return to Curvature

R (X, Y ) Z − curvature tensor

X = (ui ) , Y = (v j ) , Z = (z k ) , W = (wl ) .
( )
R (X, Y ) = ∇Y ∇X − ∇X ∇Y − ∇[X,Y ]
- curvature operator.
Index notations

( )
Rkl ji
= (∇j ∇i − ∇i ∇j ) = R̂ji = − R̂ij = (Rji )jk

(Here [∂i , ∂j ] = 0 , X = ∂i , Y = ∂j , Z = ∂k , W = ∂l ).
R (X, Y ) − linear 0-order operator on tangent space (matrix).

155
operator
k
R (X, Y ) = Rl,ij ui v j = R̂X,Y

vector
k
R (X, Y ) Z = Rl,ij ui v j z l

number
⟨R (X, Y ) Z , W ⟩ = Rl,ij
k
ui v j z l wp gkp

l
Formulas for symmetric connection were given, and for Rk,ij through Γkij .
(Especially for coordinate system such that gij (P ) = δij , ∂k Γij |P = 0).
p

Homework 11.
1. Compact Lie groups, bi-invariant Riemannian metric g → h1 g h−1
2 .
At ∼ 3 ∼
Prove that e are geodesics (for SO3 = RP and SU2 = S ). A is
3

from Lie Algebra: At = − A (G = SOn ) , At = − Ā , tr A = 0 for

G = SUn .
2. Prove that the form

⟨AR , AR ⟩ = L(g, ġ)

gives right-invariant Riemannian metric where AR (t) = ġ(t) g −1 (t) , ⟨A , A⟩

- any inner product in Rn = Lie Algebra ( n = k(k − 1)/2 for SOk ,
n = k 2 − 1 for SUk ). It is bi-invariant if ⟨A , A⟩ = Tr A2 , At = − A
and ⟨A , A⟩ = Tr A Ā , tr A = 0 , At = − Ā (groups G = SOk and
SUk ). For left-invariant metric take AL = g −1 (t) ġ(t) .
3. Find all geodesics in SU2n joining I and − I .
4. Calculate intersection matrices for all compact 2-manifolds (closed, mod-
ulo 2).
5. Prove that for any open domain in compact 2-manifold U ⊂ M 2 inter-
section matrix of 1-cycles modulo 2 has finite rank (i.e. factor-space by the
annihilator is finite-dimensional).

156
 Lecture 34. Riemannian Curvature. Exam-
ples.
Consider M n , ψ ∈ Rm
x .

Curvature tensor.
Connection: ∇X , X = ∂i , ∇X ψ = ∂i ψ + Γ̂i ψ
Curvature:
∇X ∇Y − ∇Y ∇X − ∇[X,Y ] = R̂
X = ∂i , Y = ∂j , R̂ij = ∇i ∇j − ∇j ∇i : Rm
x → Rx
m

1-form (locally) - connection

A = Γ̂i dxi

2-form (locally) - curvature

R = R̂ij dxi ∧ dxj = dA + A ∧ A

(Matrix Multiplication).
1
R̂ = dA + [A ∧ A]
2
- same; written through the commutator - Lie Algebra multiplica-
tion.
Gauge Transformation: ψ = G(x) φ(x)

Γ̂i → G−1 Γ̂i G − G−1 ∂i G :

∇i (ψ) = ∇i (Gφ) = ∂i (Gφ) + Γ̂i (Gφ)

( )
or : G ∂i φ + G−1 ∂i G φ + G−1 Γ̂i G φ

R̂ij → G−1 R̂ij G

(calculation).
l
Conclusion: for tangent bundle R̂ij = (Rk,ij ) is tensor.

157
Symmetric connection compatible with Riemannian metric (tangent bun-
dle)
∇i ⟨X , Y ⟩ = ⟨∇i X , Y ⟩ + ⟨X , ∇i Y ⟩

Corollary: ( )
1 ks ∂gij ∂gis ∂gsj
Γkij = g − s + j
+
2 ∂x ∂x ∂xi

Geodesic Equation

ẍi + Γijk (x) ẋi ẋk = 0

1) Euler - Lagrange equation for action

∫
1
S = ||ẋ||2 dt or
γ 2

2) ∇ẋ ẋ = 0 .
Calculation of Curvature: Choose coordinate system such that

gij (P ) = δij , ∂gij /∂xk |P = 0

in such point
∂ s ∂ s
s
Rijk = Γik − Γ
∂xj ∂xi jk
s
Symmetries of Curvature: Let Rij,kl = gis Rj,kl then

Rij,kl = − Rij,lk = − Rji,kl , Rkl,ij = Rij,kl

Bianchi:
Rij,ks + Rjk,is + Rki,js = 0

So R̂ is quadratic form on the space Λ2 TP = Λ2 Rn satisfying to Bianchi

identity (above) (check D.C. pages 91 - 93).
Sectional Curvature

R (X, Y, X, Y )
K (X, Y ) =
Area (X, Y )

158
X, Y - vectors,
√
Area (X, Y ) = |X ∧ Y | = |X|2 |Y |2 − ⟨X, Y ⟩2 =

Y
√ ( ) 111111111111111111111
000000000000000000000
000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
⟨X, X⟩ ⟨X, Y ⟩ 000000000000000000000
111111111111111111111
= det 000000000000000000000
111111111111111111111
⟨X, Y ⟩ ⟨Y, Y ⟩ 000000000000000000000
111111111111111111111
000000000000000000000
111111111111111111111
X
X Y
X ∧ Y ∈ Λ2 TP = Λ2 Rn
R (X ∧ Y , X ∧ Y ) - quadratic form on Λ2 Rn .
K > 0 - “Positive Curvature” (like Sn )
K < 0 - “Negative Curvature” (like Hn )
Ricci Curvature: (see D.C.)

s
Rji = Rij = Ri,sj

Scalar Curvature (R → K = const · R ) .

Rji = g is Rsj , R = Rii

1
Ric = Rij
n−1
Notation (D.C.)
1 ∑
RicP (X) = ⟨R (X, Zi ) X , Zi ⟩
n−1 i

Here |X| = 1 and Z1 , . . . , Zn−1 are orthonormal basis in Rn = TP ,

Zi ⊥ X . We have
1 ∑
K = RicP (Zj )
n j

2 - dimensional Manifolds

159
 R = R12,12 - the only nonzero component.

K<0
K>0

R = Gaussian Curvature M 2 ⊂ R3 .
3 - dimensional Manifolds
Rij determines all Curvature Tensor.
4 - dimensional Manifolds
Einstein Equation:
1
Rij − R gij = λ Tij + µ gij
2
where the term λ Tij represents “Matter” and µ gij is the “cosmological
term” (or “dark energy”).
Lie Groups and “symmetric” spaces (“constant curvature tensor”)
( i )
∇s Rj,kl ≡ 0

Homeworks 10, 11. Solutions.

Homework 10. Solutions.
1. f : S2 → S2 , f (z) = z n + a1 z n−1 + . . . + a0 .
We have homotopy ft (z) = z n + t (a1 z n−1 + . . . + a0 ) . For t = 0 we
have f0 = z n . For t = 1 we have f1 = f (z) .
For f0 = z n we have exactly n roots of equation

zn = 1

The point 1 = W ∈ S2 is transversal. So we have deg f0 = deg f1 = n .

2. Rational function f = P/Q = w .
Solve equation
P (z) = w Q(z)

160
(Fraction is irreducible).
Take point w0 such that roots are distinct. We have deg f = max (deg P, deg Q).
3. Real polynomial f = a0 xn + . . . + an , a0 ̸= 0 . Let n = 2k + 1 .
Then:
a0 >0 a0 <0
a0 > 0 : deg f = 1 ,

a0 − 0 : deg f = −1 .

Consider the case n = 2k .

4. Gauss Map:
n(P)

γ(t) = γ(t + 2π), τ (P)

s - natural parameter.
γ̇ ̸= 0 . P P → n(P ) ∈ S1 .
t = s · const ,
γ (t)

n
dτ = k(s) ds = dφ - obvious.
τ

5. 2-form
Ω = x dy ∧ dz − y dx ∧ dz + z dx ∧ dy
1. Rotate in (y, z) - plane:
x → x , dy ∧ dz → dy ∧ dz
2. Rotate in (x, y) - plane:
z → z , dx ∧ dy → dx ∧ dy
3. Rotate in (x, z) - plane:
y → y , dx ∧ dz → dx ∧ dz

For (y, z) - plane we have: rotation of Ω is equal to rotation of x dy ∧ dz

plus rotation of dx ∧ (z dy − y dz) where
x → x , z dy − y dz → z dy − y dz

161
 Similar arguments we have for other relations.
So our form is invariant under all 3 rotations. Therefore it is invariant
under the whole SO3 .

Homework 11. Solutions.

1. Compact Lie group, Bi-Invariant Metric (SO3 = RP3 and SU2 = S3 ) .
Let At = − A be real 3 × 3 matrix. eAt ∈ SO3 . In some basis we
have  
0 1 0
A = −1 0 0
0 0 0
and eAt are rotations around Z - axis. It is a straight line in RP3 = SO3
−π

presenting it as a ball B 3 (radius π). So all these geodesics

0
π
are straight lines from 0 to π . So we proved it for SO3 . For SU2 it is
the same geodesics because SU2 / ± 1 = SO3 .
2. We have for g → gg0 :
d
ġ(t) g −1 (t) → (g(t)g0 ) (g(t)g0 )−1 → ġ(t) g −1 (t)
dt
Same proof for g → g0 g and g −1 ġ for left shifts. So ⟨A(t), A(t)⟩ is right
(left) invariant inner product for A = ġ g −1 (A = g −1 ġ).
Right and Left shifts of the group commute with each other

Rh1 Lh2 = Lh2 Rh1 : g → h2 g h 1

Right shift maps right invariant metric into itself, and left shift maps
right invariant metric into another right invariant metric g̃ij .

Rh∗ 1 (gij ) = gij , L∗h2 (gij ) = g̃ij

For h2 = h−11 metric g̃ij can be obtained from gij at the Lie Alge-
bra by conjugation. So g̃ij = gij for inner product ⟨A, A⟩ invariant under
conjugation. For SOn it is ⟨A, A⟩ = − Tr A2 , At = − A .

162
3. Geodesics in SUn are eAt (if it starts at I for t = 0). So we need:
eAT = −I for some T . We have the following basis for su2 :
( ) ( ) ( )
i 0 0 −1 0 i
, ,
0 −i 1 0 i 0

Let us put ( )
ia 0
A = , a ∈ R
0 −ia
Fix time moment T such that eiaT = − 1 . We have then eAT = − I .
All other solutions can be obtained by the change of basis in C2 (same T ) :
( ) ( )
i 0 i 0
→ g g −1 ,
0 −i 0 −i

g ∈ SU2 . We have g ∈ SU2 /U1 .

4. All compact 2-manifolds M 2 :

( 2 )  Z2 ⊕ · · · ⊕ Z2 (2g times , M 2 is orientable) ,
H1 M , Z2 = Z2 ⊕ · · · ⊕ Z2 (k times , for M 2 = RP2 # . . . #RP2 ) ,

Z2 ⊕ Z2 , M 2 = K 2 .

 
0 1 ... 0 0
1 0 ... 0 0
 
 .. .. . . .... 
M 2 = S2g : . . . . . over Z2
 
0 0 ... 0 1
0 0 ... 1 0
 
1 0 ... 0
0 1 ... 0
 
M 2 = RP2 # . . . #RP2 :  .. .. . . ..  k times
. . . .
0 0 ... 1
( )
2 2 0 1
M = K :
1 1

163
 K 2 # RP2 = T2 # RP2 = RP2 # RP2 # RP2 (?)

K 2 = RP2 # RP2 (?)

5. Let U be an open domain in any compact 2-manifold M 2 . Then inter-

section matrix ( )
H1 (U, Z2 ) → H1 M 2 , Z2
depends only on image of cycles in M 2 where it has finite rank. So inner
product in H1 (U, Z2 ) also has finite rank.

164