The Geometry of Physics –

Problem Solutions
David Luposchainsky, Ernest Yeung
17. 8. 2015

These are my solutions to problems given in Theodore Frankel’s book “The Geometry of Physics” (second
edition). As I could not find any other sources, I do not know whether they are correct or not, so read with
care (especially the index battles). If you have a solution that is not in here already, a better way of showing
something, or just some useful comment, I’d like to hear about it1 .

20110714 - further solutions begun by Ernest Yeung.

Conventions
If not mentioned differently, use the following conventions:
q
• Use Einstein summation. Sometimes I’ll typeset a for clarification though.
• The “” used in the book will be used implicitly, i.e. multiindices are always assumed to be in ascending
order.

1 e-mail: stupid underscore name at gmx dot net

Contents

Manifolds and Vector Fields 6

Submanifolds of Euclidean Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1(3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Tangent Vectors and Mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Tangent or “Contravariant Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Vectors as Differential Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
The Tangent Space to M n at a Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Mappings and Submanifolds of Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Change of Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Tensors and Exterior Forms 7

2.1 Covectors and Riemannian Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1(2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3. The Cotangent Bundle and Phase Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3a. The Cotangent Bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3b. The Pull-Back of a Covector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3c. The Phase Space in Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3d. The Poincaré 1-Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4a. Covariant Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4(2)(i) Contraction invariant under base transformation . . . . . . . . . . . . . . . . 10
2.4(2)(ii) Non-invariant “contraction” . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4(3)(i) Transformation behavior of a contraction . . . . . . . . . . . . . . . . . . . . 10
2.4(3)(ii) Tensor? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4(3)(iii) Tensor? – second attempt . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4(4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 The Graßmann or Exterior Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5a. The Tensor Product of Covariant Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5b. The Grassmann or Exterior Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5c. The Geometric Meaning of Forms in Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5(1) Basis expansion of a form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5(2) Components of –1 · — 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5(3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Exterior Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6(1) Differential of a 3-Form in R4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7 Pull-Backs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7(1) Proof of homomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.7(2) Pull-back onto a surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.8c. Orientability and 2-sided Hypersurfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.9 Interior Products and Vector Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.10(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.10(2) Components of the interior product . . . . . . . . . . . . . . . . . . . . . . . . 15
2.10(3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.10(4) Vector analysis in R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.10(5) Basis expansion of the cross product . . . . . . . . . . . . . . . . . . . . . . . . 16

Integration of Differential Forms 16

3.1 Integration over a Parameterized Subset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1a. Integration of a p-Form in Rp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1b. Integration over Parametrized Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1(3)(i) Higher-dimensional cross product . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3(1) ... in R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3(2) ... in R4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
The Lie derivative 18
4.1 The Lie Derivative of a Vector Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1a. The Lie Bracket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1(1) Coordinate expression for [X, Y] . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 The Lie Derivative of a Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2a. Lie Derivatives of Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2b. Formulas Involving the Lie Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2(1) Coordinate expression for LX –1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2(2) Compositions of derivations and antiderivations . . . . . . . . . . . . . . . . . . . 20
4.2(3) i[X,Y] = LX ¶ iY ≠ iY ¶ LX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2(3) Fugly proof of d–(X, Y) = X(–(Y)) ≠ Y(–(X)) ≠ –([X, Y]) . . . . . . . . . . . . 21
4.3. Differentiation of Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3a. The Autonomous (Time-Independent) Case . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3b. Time-Dependent Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3c. Differentiating Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3(5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.4 A problem set on Hamiltonian mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4(1) Symplectic form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4(1) Symplectic volume form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
P. 147: Derivation of Hamilton’s equations . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4(4) Hamilton in shrt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4(5) Lie derivative of the symplectic Poincaré 2-form . . . . . . . . . . . . . . . . . . . 24
4.4(8) Hmltn n shrtr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4(9) Lie derivative of the pre-symplectic Poincaré 2-form . . . . . . . . . . . . . . . . . 24

The Poincare Lemma and Potentials 25

5.1. A More General Stokes’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.2. Closed Forms and Exact Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.3. Complex Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.5 Finding potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.5(1) Product of a closed and an exact form . . . . . . . . . . . . . . . . . . . . . . . . 26

6 Holonomic and Nonholonomic Constraints 26

6.1. The Robenius Integrability Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.2. Integrability and Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3. Heuristic Thermodynamics via Caratheodory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3a. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3b. The First Law of Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
6.3d. The Second Law of Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

R3 and Minkowski Space 27

7.1 Curvature and Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
7.1.a. Curvature of a Space Curve in R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
7.1(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7.1(2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7.2 Electromagnetism in Minkowski Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
7.2(3) Field strength 2-Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

The Geometry of Surfaces in R3 29

9 Covariant Differentiation and Curvature 29

9.1 Covariant Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
9.3 Cartan’s Exterior Covariant Differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
9.3c. Cartan’s Structural Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
9.3d. The Exterior Covariant Differential of a Vector-Valued Form . . . . . . . . . . . . . . . . 30
9.3(1) Basis expansion of the curvature form . . . . . . . . . . . . . . . . . . . . . . . . 31
9.3(2) Covariant derivative of the identity form . . . . . . . . . . . . . . . . . . . . . . . 31
9.4 Change of Basis and Gauge Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . 31
9.4(1) Transformation of the curvature form . . . . . . . . . . . . . . . . . . . . . . . . . 31
 9.4(2) Transformation of the curvature form . . . . . . . . . . . . . . . . . . . . . . . . . 32
Parallel Displacement and Curvature on a Surface . . . . . . . . . . . . . . . . . . . . . . . . . 33

10 Geodesics 34

11 Relativity, Tensors, and Curvature 34

12 Curvature and Topology: Synge’s Theorem 34

13 Betti Numbers and De Rham’s Theorem 34

14 Harmonic Forms 34

15. Lie groups 34

15.1 Lie Groups, Invariant Vector Fields and Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.1a Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.1b. Invariant Vector Fields and Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.2 One-parameter subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.2(1) Generator of rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.2(2) Generator of A(1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
15.3(1) Maurer-Cartan equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
The Lie Algebra of a Lie Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
15.3a. The Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
15.3b. The Exponential Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

16. Vector Bundles in Geometry and Physics 35

16.3(1) Connection on a tensor product space . . . . . . . . . . . . . . . . . . . . . . . . 35

17. Fiber Bundles, Gauss-Bonnet, and Topological Quantization 35

17.1. Fiber Bundles and Principal Bundles 35

17.1a. Fiber Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
17.1b. Principal Bundles and Frame Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
17.1c. Action of the Structure Group on a Principal Bundle . . . . . . . . . . . . . . . . . . . . . . . 36
17.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
17.3. Chern’s Proof of the Gauss-Bonnet-Poincaré Theorem . . . . . . . . . . . . . . . . . . . . . . . 36
17.3a. A Connection in the Frame Bundle of a Surface . . . . . . . . . . . . . . . . . . . . . . . 36

18. Connections and Associated Bundles 36

18.1. Forms with Values in a Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
18.1.a. The Maurer-Cartan Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

19. The Dirac Equation 37

19.1a. The Rotation Group SO(3) of R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
19.1b. SU (2): The Lie Algebra su(2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
19.1c. SU (2) is Topologically the 3-Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Ad : SU (2) æ SO(3) in More Detail . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

20. Yang-Mills Fields 42

20.1. Noether’s Theorem for Internal Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.1a. The Tensorial Nature of Lagrange’s Equations . . . . . . . . . . . . . . . . . . . . . . . . 42
20.2. Weyl’s Gauge Invariance Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.2a. The Dirac Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.2b. Weyl’s Gauge Invariance Revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.2c. The Electromagnetic Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
The Yang-Mills Nucleon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.3a. The Heisenberg Nucleon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.3b. The Yang-Mills Nucleon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
20.3c. A Remark on Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

21. 43
A. Elasticity 43
A.a. The Classical Cauchy Stress Tensor and Equations of Motion . . . . . . . . . . . . . . . . . . . 43
A.b. Stresses in Terms of Exterior Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
A.f. Hamilton’s Principle in Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
I Manifolds, Tensors and Exterior Forms
Manifolds and Vector Fields
Submanifolds of Euclidean Space
2
1.1(3) Consider for F (A) = det (A), F : Rn æ R, that
! " ! "
det (A + tB) = det A(1 + tA≠1 B) = det (A) det 1 + tA≠1 B
! "
How to deal with det 1 + tA≠1 B ? Recall that
! " ! "
det 1 + tA≠1 B = det (A) (1 + t Tr A≠1 B )
because for det ((1 + tX)),
QQ R Q RR QQ RR
1 tx11 ... tx1n 1 + tx11 ... tx1n
cc .. d c .. .. .. dd = det cc .. .. .. dd
det ((1 + tX)) = det aa . b+a . . . bb aa . . . bb =
1 txn1 ... txnn txn1 ... 1 + txnn
= 1 + t Tr (X) + O(t2 )
q
since recall det (A) = ‡œSn sgn(‡)A1‡1 A2‡2 . . . An‡n , where sum is over all permutations of {1, . . . , n}, and
so only the A11 . . . Ann term would have terms of O(t).
So
d ! "
(DF ) · B = F (x(t))B = det (x) Tr x≠1 B
dt
For x0 œ Sl(n), det (x0 ) = 1. Let B = nr x. Then
1 r 2
(DF ) · B = Tr x≠1 x = r
n
DF = Fú is surjective ’ x œ Sl(n)

Manifolds
Tangent Vectors and Mappings
Tangent or “Contravariant Vectors
Vectors as Differential Operators
The Tangent Space to M n at a Point
Mappings and Submanifolds of Manifolds
Definition 1 M m µ N n (embedded) submanifold of N n . If M locally s.t. F : N n æ Rn≠m
F 1 (x1 . . . xn ) = 0
..
.
F n≠m (x1 . . . xn ) = 0
- i-
- -
n ≠ m diff. F i s.t. - ˆF
ˆxj - has rank n ≠ m

By implicit function thm., submanifold as graph.

(x1 . . . xm , y m+1 . . . y n )
y m+1 = f m=1 (x1 . . . xm )
..
.
y n = f n (x1 . . . xm )
on F (x) = 0
Theorem 2 (1.12) Let F : M m æ N n , q œ N n s.t. F ≠1 (q) µ M m , F ≠1 (q) ”= ÿ
If Fú onto, i.e. Fú rank n, ’ F ≠1 (q),
F ≠1 (q) (n ≠ m)-dim. submanifold of M m
Change of Coordinates

Tensors and Exterior Forms

2.1 Covectors and Riemannian Metrics
q V i q
2.1(1) Want: ai vV = aU j
j vU
A B A B
ˆ ˆ
i dxU (v) = = = (v) = = aVi vVi
j
aU i
aU i
i dxU vU aU i
i vU aVi dxiV aVi dxiV vVj
ˆxjU ˆxjV
or
ˆxiV ˆxiV ˆxjU
j vU = ai = aVi (p0 )vVk = aVi vVi
j j
aU V
vU
ˆxjU ˆxjU ˆxkV
were (1.6) was used.

ˆxiU ˆxiU k ˆxiU ˆxiU j k

v i wi = vU
i i
wU = vVj wV = vV wV
ˆxjV k
ˆxV ˆxjV ˆxkV
transforms as a (0, 2) tensor.

2.1(2)
(i) Recall
ˆxrU ˆxsU U
V
gij = g
ˆxV ˆxjV rs

ˆx ˆx ˆx
= c„s◊ = rc„c◊ = ≠rs„s◊
1
u =r x = r cos („) sin (◊) ˆr ˆ◊ ˆ„
ˆy ˆy ˆy
u2 = ◊ y = r sin („) sin (◊) = s„s◊ = rs„c◊ = rc„s◊
ˆr ˆ◊ ˆ„
u =„3 z = r cos (◊) ˆz ˆz ˆz
= c◊ = ≠rs◊ =0
ˆr ˆ◊ ˆ„

grr = 1
g◊◊ = r2
g„„ = r2 (sin (◊))2

(ii) grad (f ) = Òf is contravariant vector, associated to covector df , df (w) = ÈÒf, wÍ. (Òf )i = g ij ˆx
ˆf
j

g ij is not gij . What is g ij ?

Also (r, ◊, „) is a non-coordinate bases.

ds2 = (dr)2 + r2 (d◊)2 + r2 (sin (◊))2 (d„)2

The distance elements are dr, rd◊, r sin (◊) d„ in this non-coordinate basis r‚, ◊,
‚ „.
‚
2 2 2
We’re using ds = |dx| © g(dx, dx) = dx · dx instead of ds = gµ‹ dx dx µ ‹

In the non-coordinate basis {‚ ‚ Coordinate basis {er , e◊ , e„ }

‚ „}.
r, ◊,
Consider
gµÕ ‹ Õ © g(eµÕ , e‹ Õ ) = g–— e– (eµÕ )Â
e— (e‹ Õ ) = g–— –
µÕ
—
‹Õ

gµ‹ = eµ · e‹
grr = 1
g◊◊ = r2
g„„ = r2 (sin (◊))2
So e◊ = ˆ◊ ,
ˆ
e„ = ˆ
ˆ„ are not unit vectors!
In the coordinate basis dx = er dr + e◊ d◊ + e„ d„ = ei dxi
In the noncoordinate basis, dx = r‚dr + ◊rd◊
‚ + „r
‚ sin (◊) d„

r
‚
r
= 1
1
◊
‚ =
◊ r
1
„
=
‚
„ r sin (◊)
So then, for instance
g„„ = r2 (sin (◊))2 = g(ˆ„ , ˆ„ ) = gij eÂi (ˆ„ )Â
ej (ˆ„ ) = gij i
„
j
„ = g„‚„‚r2 (sin (◊))2

g‚
r‚
r
= g‚
◊‚
◊
= g„‚„‚ = 1
g‚
r‚
◊
= g‚‚ = g‚
r„ ◊„‚=0
in non-coordinate basis, we must give up the following two:
dx © dxµ eµ defines eµ coordinate basis
ds2 = gµ‹ dxµ dx‹
Inverse metric components
g rr = 1
1
g ◊◊ = 2
r
1
g „„ =
r2 (sin (◊))2
The isomorphism of V and V ú (e.g. Tp M and Tp M ú ) allows us to introduce notation that replaces one-
forms with vectors and (m, n) tensors with (m + n, 0) tensors.
Replace basis one-forms eµ © –µ with set of vectors defined
eµ (·) © g ≠1 (Â
eµ , ·) = g µ‹ eµ (·)
where eµ basis one form, eµ dual basis vector.
Then
ˆ
er = e r = = r‚
ˆr
1 1
e◊ = 2 e◊ = ◊‚
r r
1 1
e =
„
e„ = „‚
(r sin (◊))2 r sin (◊)
Now
 © eµ ˆµ
Ò in coordinate basis
Â
Òx µ
= eµ
in a coordinate basis
Ò = eµ ˆµ = g µ‹ eµ ˆ‹
So finally
1‚ 1
Ò = r‚ˆr + ◊ˆ ◊ +
‚ „
„ˆ
r r sin (◊)
1‚ 1
Òf = r‚ˆr f + ◊ˆ ◊f +
‚ „f
„ˆ
r r sin (◊)
Also, in this formulation,
1 1
Ò = er ˆr + e◊ ˆ◊ + e„ ˆ„
r 2 (r sin (◊))2
1 1 ˆ 1 ˆ ˆ„ f ˆ
Òf = er ˆr f + 2 ˆ◊ f e◊ + ˆ„ f e„ = (ˆr f ) + ˆ◊ f + =
r (r sin (◊))2 ˆr r2 ˆ◊ (r sin (◊))2 ˆ„
ˆf
= (Òf )i ˆi = g ij j ˆi
ˆx
(cf. MIT Physics 8.962 Spring 1999, Edmund Bertschinger. Introduction to Tensor Calculus for
General Relativity gr1.pdf)
(iii) See above. And as before,
ˆ
= r‚
ˆr
1 ˆ
= ◊‚
r ˆ◊
1 ˆ
= „‚
r sin (◊) ˆ„

2.3. The Cotangent Bundle and Phase Space

2.3a. The Cotangent Bundle

2.3b. The Pull-Back of a Covector

2.3c. The Phase Space in Mechanics Let q 1 . . . q m local generalized coordinates, M m configuration space
of a dynamical system.

L : TMm æ R

Consider (U, q), U V ”= ÿ, q œ U V

(V, r)
r = r(q)

ˆrj i ˆ ṙj ˆrj

ṙj = q̇ (2.27) =
ˆq i ˆ q̇ i ˆq i
ˆL ˆL ˆq j ˆL ˆ q̇ j ˆL ˆ q̇ j ˆL ˆq j
fii © = + = =
ˆ ṙi ˆq j ˆ ṙi ˆ q̇ j ˆ ṙi ˆ q̇ j ˆ ṙi ˆ q̇ j ˆri
j
ˆq
=∆ fii = pj i (2.29)
ˆr
p’s are covector.

=∆ p : T M m æ T ú M m
cotangent bundle. T ú M m of covectors to configuration space is phase space.
1ÿ
T (q, q̇) = gjk (q)q̇ j q̇ k (2.31)
2
jk
ˆL ˆT ÿ
pi = i = i = gij (q)q̇ j (2.32)
ˆ q̇ ˆ q̇ j

think of 2T as Riemannian metric on M m .

ÿ
Èq̇, q̇Í = gij (q)q̇ i q̇ j
ij

2.3d. The Poincaré 1-Form

2.4 Tensors
2.4a. Covariant Tensors
Definition 3 covariant tensor of rank r

Q : E ◊ ··· ◊ E æ R
Q(v1 . . . vr )
vector space of covariant rth rank tensors E ú ¢ · · · ¢ E ú = ¢r E ú

2nd. rank covariant tensor – ¢ — : E ◊ E æ R

– ¢ —(v, w) © –(v)—(w)
2.4(1) For any v, w tangent vectors,
ˆ
v = vi
ˆxi
ˆ
w = wi i
ˆx
ˆ ˆ
(– ¢ —)(v, w) = –(v)—(w) = ai dxi (v j j )bk dxk (wl l ) = ai bk v j wl ” ij ” kl = aj v j bk wk =
ˆx ˆx
= aj bk dxj (v)dxk (w) = aj bk dxj ¢ dxk (v, w)
For –, — in components,
– = ai dxi
— = bj dxj

ai bj dxi ¢ dxj (v, w) = ai bj v i wj = aj v j bk wk =∆ – ¢ — = ai bj dxi ¢ dxj

2.4(2)(i) Contraction invariant under base transformation

3 Õi 4
ˆx j ˆx
k
ˆxÕi ˆxk ! j "
A i = A(dx , ˆ i ) = A
Õi Õi Õ
dx , ˆ k = A dx , ˆ k = Aj j
ˆxj ˆxÕi ˆxj ˆxÕi
¸ ˚˙ ˝ ¸ ˚˙ ˝
ˆxk =Aj k
=”jk
ˆxj

This is the transformation law of a scalar.

2.4(2)(ii) Non-invariant “contraction”

ÿ ÿ ÿ 3 4 ÿ ˆxj ˆxk
ˆxj ˆxk
AÕii = A(ˆ Õi , ˆ Õi ) = A ˆ j , ˆk = A(ˆ j , ˆ k )
i i i
ˆxÕi ˆxÕi i
ˆxÕi ˆxÕi ¸ ˚˙ ˝
=Ajk
ÿ ˆxj ˆxk
= Ajk ”= Aii
i
ˆxÕi ˆxÕi
q
Since the differential quotients do not cancel out, the value of i Aii is dependant on coordinates; a coordinate-
dependant number is neither a scalar nor any other sort of tensor.

2.4(3)(i) Transformation behavior of a contraction

ˆxk ˆx¸ ˆxÕi m ˆxk ˆx¸ ˆxÕi ˆxk

gji v =
Õ Õi
gk¸ v = gk¸ v m
= gk¸ v ¸
ˆxÕj ˆxÕi ˆxm ˆxÕj ¸ˆxÕi˚˙ˆxm˝ ˆxÕj
=”m
¸

Thus, gji v i transforms like a vector.

2.4(3)(ii) Tensor?
3 4
ˆ ˆxÕi k ˆ 2 xÕi ˆx¸ k ˆxÕi ˆv k ˆ 2 xÕi ˆx¸ k ˆxÕi ˆv k ˆx¸
ˆjÕ v Õi = v = v + = v +
ˆxÕj ˆxk ¸
ˆx ˆx ˆxk Õj k
ˆx ˆxÕj ˆx¸ ˆxk ˆxÕj ˆx¸ ˆxÕj
ˆxk ¸˚˙˝
=ˆ¸ v k
2 Õi ¸ ¸ Õi
ˆ x ˆx k ˆx ˆx
= ¸ ˆxk ˆxÕj
v + Õj ˆ vk
k ¸
ˆx
¸ ˚˙ ˝ ˆx ˆx
”=0

Although the second term is the correct tensor transformation law, the first term prevents ˆj v i from forming
a tensor.
2.4(3)(iii) Tensor? – second attempt
Using the result of (ii), one gets

ˆ 2 xÕi ˆx¸ k ˆx¸ ˆxÕi ˆ 2 xÕj ˆx¸ k ˆx¸ ˆxÕj

ˆjÕ v Õi ≠ ˆiÕ v Õj = v + ˆ¸ v k
≠ v ≠ ˆ¸ v k
ˆx¸ ˆxk ˆxÕj ˆxÕj ˆxk ˆx¸ ˆxk ˆxÕi ˆxÕi ˆxk
3 2 Õi 4 3 ¸ 4
ˆ x ˆx¸ k ˆ 2 xÕj ˆx¸ k ˆx ˆxÕi ˆx¸ ˆxÕj
= v ≠ v + ˆ¸ v k
≠ ˆ¸ v k
ˆx¸ ˆxk ˆxÕj ˆx¸ ˆxk ˆxÕi ˆxÕj ˆxk ˆxÕi ˆxk
ˆx¸ ˆxÕi ! k "
”= 0 + Õj k
ˆ¸ v ≠ ˆk v ¸
ˆx ˆx

2.4(4)
(i)
1
L = L(q, q̇) = gij (q)q̇ i q̇ j ≠ V
2
3 4
d ˆL ˆL
k
= k
dt ˆ q̇ ˆq

ˆV i 1 ˆ 2 V i j
V = V (q) = V (0) + q + qq
ˆq i 2 ˆq i ˆq j
Assume g symmetric in indices.
ˆV
ˆq k
= 0 i.e. q = 0 nondegenerate minimum for V .
3 4
d ˆL ˆ2V j
= gij (0)q̇ j = ≠ q = ≠Qij q j
dt ˆ q̇ i ˆq i ˆq j
(ii)
(iii)

2.5 The Graßmann or Exterior Algebra

2.5a. The Tensor Product of Covariant Tensors
2.5b. The Grassmann or Exterior Algebra
1
– = –J dxJ = –J dxJ
p!
–J antisymmetric in J and –J antisymmetric in J

–J = p!–J

Lemma 1 (2.46)
”M ”JKL = ”M
IJ IKL

Proof

I = (i1 . . . ip )
J = (j1 . . . jq+r )
J = (j1 < · · · < jq+r )
K = (k1 . . . kq )
L = (l1 . . . lr )
M = (m1 . . . mp+q+r )
KL fixed. put KL into (unique) increasing order, by as many transpositions as total number of inversions
(cf. Tu, L.W., Introduction to Manifolds, Springer, 2008), Proposition 3.6)
so ”JKL ”= 0 for only 1 J
Suppose ”JKL = 1, KL even permutation of J (permutation is bijective)
M fixed so suppose IJ even permutation of M
I‡(KL) = f (M )
 ‡ even permutation of KL, so put ‡(KL) into KL by even number of transpositions
This defines even permutation g that’s bijective on I‡(KL)

g(I‡(KL)) = IKL = gf (M )

so IKL even permutation of M .

If IJ odd permutation of M , gf odd permutation, ”M
IKL
= ≠1

2.5c. The Geometric Meaning of Forms in Rn

2.5(1) Basis expansion of a form
! "
aJ dxJ (ˆ K ) = aJ dxJ (ˆ K ) = aJ ”K
J
= aK = – (ˆ K )

Since this is true for all ˆ K , – = aJ dxJ .

2.5(2) Components of –1 · — 2
ÿ
(–1 · — 2 )i<j<k = lmn
”ijk –l —mn
l,m<n

All summands where ijk is not a permutation of lmn vanish, so there are 6 possible permutations left:

(A) (B) (C) (D) (E) (F)

i = l i = m i = n i = l i = n i = m
j = m j = n j = l j = n j = m j = l
k = n k = l k = m k = m k = l k = n

Of these 6, (C), (D) and (E) contradict i < j < k (given by the problem) with respect to m < n (from the
definition of the wedge product), leaving only 3 summands. Thus,
ÿ
(–1 · — 2 )i<j<k = lmn
”ijk –l —mn = ”ijk
ijk
–i —jk + ”ijk
kij
–k —ij + ”ijk
jik
–j —ik
¸˚˙˝ ¸˚˙˝ ¸˚˙˝ ¸˚˙˝
l,m<n
(A)æ+1 (B)æ+1 (F )æ≠1 ≠—ki

= –i —jk + –j —ki + –k —ij .

2.5(3) In R3 ,
Given
–1 = a1 dx1 + · · · + a3 dx3
— 1 = b1 dx1 + b2 dx2 + b3 dx3
fl1 = r1 dx1 + r2 dx2 + r3 dx3
“ 2 = c1 dx2 · dx3 + c2 dx3 · dx1 + c3 dx1 · dx2
–1 · “ 2 = (a1 c1 + a2 c2 + a3 c3 )dx1 · dx2 · dx3 = a · cdx1 · dx2 · dx3 = a · cvol(dx)
–1 · — 1 = (a1 b2 ≠ a2 b1 )dx1 · dx2 + (a1 b3 ≠ a3 b1 )dx1 · dx3 + (a2 b3 ≠ a3 b2 )dx2 · dx3
–1 · — 1 · fl1 = ((a1 b2 ≠ a2 b1 )r3 + (≠r2 )(a1 b3 ≠ a3 b1 ) + r1 (a2 b3 ≠ a3 b2 ))dx1 · dx2 · dx3 = r · (a ◊ b)vol(dx)
2.6 Exterior Differentiation
2.6(1) Differential of a 3-Form in R4
ÿ
— 3 = —J dxJ = —ijk dxi · dxj · dxk
i<j<k

= —123 dx · dx2 · dx3 + —124 dx1 · dx2 · dx4

1

+ —134 dx1 · dx3 · dx4 + —234 dx2 · dx3 · dx4

∆ d— 3 = d—123 · dx1 · dx2 · dx3 + d—124 · dx1 · dx2 · dx4
+ d—134 · dx1 · dx3 · dx4 + d—234 · dx2 · dx3 · dx4
ˆ—123 ˆ—124
= dxi · dx1 · dx2 · dx3 + dxi · dx1 · dx2 · dx4
ˆxi ˆxi
ˆ—134 ˆ—234
+ dxi · dx1 · dx3 · dx4 + dxi · dx2 · dx3 · dx4
ˆxi ˆxi
ˆ—123 ˆ—124
= dx4 · dx1 · dx2 · dx3 + dx3 · dx1 · dx2 · dx4
ˆx4 ˆx3
ˆ—134 ˆ—234
+ dx2 · dx1 · dx3 · dx4 + dx1 · dx2 · dx3 · dx4
ˆx2 ˆx1
ˆ(≠—123 ) ˆ—124
= 4
dx1 · dx2 · dx3 · dx4 + dx1 · dx2 · dx3 · dx4
ˆx ˆx3
ˆ(≠—134 ) ˆ—234
+ dx1 · dx2 · dx3 · dx4 + dx1 · dx2 · dx3 · dx4
ˆx2 ˆx1
æ rename components: —234 æ —1 , ≠—134 æ —2 , —124 æ —3 , ≠—123 æ —4
3 4
ˆ—1 ˆ—2 ˆ—3 ˆ—4
= + + + dx1 · dx2 · dx3 · dx4
ˆx1 ˆx2 ˆx3 ˆx4

In cartesian coordinates, this says something like d(B · dV) = div (B) dH (“H: Hyperspace volume”).

2.7 Pull-Backs
2.7(1) Proof of homomorphism
Notation: Let (Fú vI ) = (Fú vi1 , Fú vi2 , . . .) .
ÿ
F ú (– · —) (vI ) = (– · —) (Fú vI ) = ”IJK –(Fú vJ ) —(Fú vK ) = –(Fú vJ ) · —(Fú vK )
J,K

= (F ú –(vJ )) · (F ú —(vK )) ’ vI (= vJK )

∆ F (– · —) = (F –) · (F —)
ú ú ú
2.7(2) Pull-back onto a surface
Let (u, v) = (y 1 , y 2 ).

— 2 = —12 dx1 · dx2 + —13 dx1 · dx3 + —23 dx2 · dx3

ˆx1 i ˆx2 j ˆx1 i ˆx3 j ˆx2 i ˆx3 j
∆ iú — = —12 dy · dy + — 13 dy · dy + —23 dy · j dy
ˆy i ˆy j ˆy i ˆy j ˆy i ˆy
3 1 2⇠⇠ 1 2
ˆx 1 ⇠ˆx ⇠ ˆx ˆx
= —12 1⇠
dy⇠ · 1 dy 1 + 1 dy 1 · 2 dy 2
⇠ˆy⇠ ˆy ˆy ˆy
1 2 1 2⇠⇠
4
ˆx 2 ˆx 1 ˆx 2 ⇠ˆx ⇠ 2
+ dy · 1 dy + ⇠ dy⇠· 2 dy
ˆy 2 ˆy ⇠ˆy 2 ⇠ ˆy
3 1 3⇠⇠ 1
ˆx 1 ⇠ˆx ⇠ 1 ˆx 1 ˆx3 2
+ —13 dy
⇠ ⇠ · dy + dy · dy
⇠ˆy⇠ 1 ˆy 1 ˆy 1 ˆy 2
4
ˆx1 2 ˆx3 1 ˆx1 2 ⇠ˆx ⇠3⇠⇠2
+ dy · dy + dy
⇠ ⇠ · dy
ˆy 2 ˆy 1 ⇠ˆy⇠2 ˆy 2
3 2
ˆx 1 ⇠ˆx ⇠3⇠⇠1 ˆx2 1 ˆx3 2
+ —23 dy
⇠ ⇠ · dy + 1 dy · 2 dy
⇠ˆy⇠ 1 ˆy 1 ˆy ˆy
3⇠⇠
4
ˆx2 2 ˆx3 1 ˆx2 2 ⇠ˆx ⇠ 2
+ dy · 1 dy + ⇠ dy⇠· 2 dy
ˆy 2 ˆy ⇠ˆy 2 ⇠ ˆy
3 3 1 2 4 3 1 3 4
ˆx ˆx ˆx1 ˆx2 ˆx ˆx ˆx1 ˆx3
= —12 ≠ + —13 ≠
ˆy 1 ˆy 2 ˆy 2 ˆy 1 ˆy 1 ˆy 2 ˆy 2 ˆy 1
3 2 3 44
ˆx ˆx ˆx2 ˆx3
+ —23 1 2
≠ 2 1 dy 1 · dy 2
ˆy ˆy ˆy ˆy

If one now defines, by renaming the components of — again (—23 æ —1 , ≠—13 = —31 æ —2 , —12 æ —3 ),
b = (—1 , —2 , —3 ), the last term can be identified as b · n dy 1 · dy 2 , and one gets the desired expression

iú — = (b, n) du · dv .

2.8
2.8c. Orientability and 2-sided Hypersurfaces
If M orientable if ÷ orientation ’ T Mxn to M n , cont., or cover M by (U, Ï), |J| > 0 ’ overlap.
Converse: cont. orientation ’ T Mxn , M orientable.

If M orientable, ’ p, q œ M , curve C, p = C(0), C(t), t ‘æ ei (t) cont.,

q = C(1)
Contrapositive! (Mö bius strip)

2.8c. Orientability and 2-Sided Hypersurfaces

M submanifold of W r
N transverse to M if N never tangent to M , N ”= 0 on M
hypersurface M n in W n+1 2-sided in W if ÷ (cont.) transverse vector field N along M
Möbius band “1-sided”, @ cont. unit N
if M n 2-sided hypersurface of orientable W n+1 , then M n orientable

2.9 Interior Products and Vector Analysis

2.9a. Interior Products and Contractions
Definition 4 interior product
iv –1 = –(v) if – 1-form
iv –p (w2 . . . wp ) = –p (v, w2 . . . wp ) if – p-form

Clearly iA+B = iA + iB
iaA = aA
Theorem 5 (2.75) iv : p
æ p≠1
antiderivation

iv (–p · — q ) = [iv –p ] · — q + (≠1)p –p · [iv — q ]

Theorem 6 (2.76) in components
ÿ ÿ
iv – = v j aji2 <···<ip dxi2 · · · · · dxip
i2 <···<ip j
q
i.e. (iv –)i2 <···<ip = j v j aji2 <···<ip
or

[iv –]k = v j –jk

2.9b. Interior Product in R3

v ≈∆ pseudo-2-form v 2 © iv vol3
Ô Ô
iv gdu1 · · · · · dun = gv i iˆi du1 · · · · · dun
ÿ ÿ
iˆi du1 · · · · · dun = 1...n
” ji ”jI duI = 1...n
”iI duI =??? = ‘iI 1 „i
1...n du · · · · · du · . . . du
n

I,j I

cf. Nakahara 5.4.3. Interior product and Lie derivative of forms

ˆ
X = Xµ
ˆxµ
1
Ê=
ʵ ...µ dxµ1 · · · · · dxµp
p! 1 r
p
1 1 ÿ is i
iX Ê = X ‹ Ê‹i2 ...ip dxi2 · · · · · dxip = X Êi1 ...is ...ip (≠1)s≠1 dxi1 · · · · · dx
„ s · · · · · dxip
(p ≠ 1)! p! s=1
Ê 1 = È , wÍ
iv Ê 1 = Ê 1 (v) = Èv, wÍ

v 1 · Ê 2 = Èv, wÍvol3 (2.82) (1)

v 1 · Ê 2 = v 1 · iw vol3 = [iw v 1 ] · vol3 + ≠iw (v 1 · vol3 ) = (iw v 1 )vol3 = Èv, wÍvol3

v◊w 2 form v 1 · Ê 2 iv◊w vol3 = v 1 · w1
1 form ≠iv Ê 2

2.10(1) Given T...j...

...i...
components of a mixed tensor, p times contravariant and q times covariant, then it
transforms as such, by definition,
ˆy k1 ˆy ki ˆy kq ˆxj1 ˆxjj ˆxjp ...ii ...
...ki ...
T...l j ...
= i 1
. . . ii . . . iq l 1
. . . lj . . . lp T...jj ...
ˆx ˆx ˆx ˆy ˆy ˆy

ˆy k1 ˆy k ˆy kq ˆxj1 ˆxjj ˆxjp ...ii ... ˆy k1 ‰k

ˆy ˆy kq ˆxj1 ‰
ˆxjj ˆxjp ...i...
...k...
T...k... = . . . . . . . . . . . . T = . . . . . . . . . . . . T
i
ˆx 1 i
ˆx 1 i
ˆx q ˆy 1 l ˆy k l
ˆy p ...jj ... i
ˆx 1 i
ˆx i i l
ˆx q ˆy 1 l
ˆy j ˆy lp ...i...
A3 4 Bjj
≠1
ˆy k ˆxjj ˆxjj ˆy k ˆy ˆy k
= = = ” iji
j
ˆxii ˆy k ˆy k ˆxii ˆx ˆxii
k

2.10(2) Components of the interior product

Let – = –J dxJ . In general we have the expansion
! "
iv – = ivj ˆ j – (ˆ k , ˆ L ) dxk · dxL = v j – (ˆ j , ˆ L ) dxL = v j –jL dxL
For a single component this yields
! "
(iv –)K = v j –jL dxL (ˆ K ) = v j –jL dxL (ˆ K ) = v j –jL ”K
L
= v j –jK
2.10(3) Recall
5 3 46
2 1 ˆ Ô ij ˆf
Ò f= f © div(gradf ) = Ô gg
g ˆui ˆuj
Note that g ij is the inverse of gij .
Note that g = r4 (sin (◊))2

ˆr (r2 sin (◊) ˆr f ) + ˆ◊ (r2 sin (◊) (1/r2 )ˆ◊ f ) + ˆ„ (r2 sin (◊) (1/(r2 (sin (◊))2 ))ˆ„ f )
=∆ (1/r)ˆr (r2 ˆr f ) + (1/r2 )ˆ◊ (sin (◊) ˆ◊ f ) + (1/(r2 (sin (◊))2 ))ˆ„ (ˆ„ f )

2.10(4) Vector analysis in R3

…grad(f ) …grad(g)
˙˝¸˚ ˙˝¸˚
0 0
grad (f g) … d(f · g ) = df · g + f · dg = (df ) g + f (dg) … f grad (g) + g grad (f )
…grad(f )
˙˝¸˚
2
div (f B) … d(f · — ) = df · — 2 + f · d— 2 … f div (B) + Ègrad (f ) , BÍ
¸ ˚˙ ˝ ¸˚˙˝
…grad(f )·B …div(B)

2.10(5) Basis expansion of the cross product

v ◊ B … ≠iv — 2
= ≠iv iB vol3
= ≠v k B l iˆ k iˆ l vol3
= ≠v k B l vol3 (ˆ l , ˆ k , ˆ m )dxm
Ô
= g v k B l Áklm dxm

If you’re wondering how the “identification stuff” works, read the chapter about the Hodge star operator, it’s
around page 360. I have no idea why Frankel placed it that late. You might also be interested in the definition
of the cross product in 3.1(3)(i).

Integration of Differential Forms

one does not integrate vectors; one integrates forms.
If there is extra structure available, for example, a Riemannian metric, then it is possible to rephrase an
integration, say of exterior 1-forms or 2-forms, in terms of a vector interations involving “arc lengths” or
“surface areas,” but we shall see that even in this case we are complicating a basically simple situation.
If a line integral of a vector occurs in a problem, then usually a deeper look at the situation will show that the
vector in question was in fact a covector, that is, a 1-form!
For example, the strength of the electric field can be determined by the work done in moving a unit charge
very slowly along a small path, that is, by a line integral. The electric field strength is a 1-form.
-Theodore Frankel.

3.1 Integration over a Parameterized Subset

How does one integrate the Poincaré 2-form Ê over a surface in phase space? -Theodore Frankel.

3.1a. Integration of a p-Form in Rp

define integral of a p-form over region (U, o) µ Rp ,
orientation o; o(u) = ±1
⁄ ⁄ ⁄
– = a(u)du1 · · · · · dup ©= o(u)a(u)du1 . . . dup (3.1) (2)
(U,o) U
! "
(e1 . . . ep ) = ˆ
ˆu1 . . . ˆuˆ p has same orientation as o(u)
3.1b. Integration over Parametrized Subsets
oriented parameterized p-subset of manifold M to be pair (U, o; F ),
oriented region (U, o) in Rp and diff. F : U æ M

define ⁄ ⁄
– = p
F ú –p (3.3) (3)
(U,o;F ) (U,o)

we make no requirements on the rank of DF .

⁄ ⁄ 5 6 ⁄ 5 6
ˆ ˆ 1 ˆ ˆ
– ©p
(F – ) ú p
. . . p du · · · · · du = o(u) (F – )
p ú p
. . . p du1 . . . dup (3.4)
(U,o;F ) (U,o) ˆu1 ˆu U ˆu1 ˆu
(4)
⁄ 5 6
ˆ ˆ
= o(u) –p Fú 1 . . . Fú p du1 . . . dup (3.5) (5)
U ˆu ˆu
⁄ ⁄ ⁄ b ⁄ b
dxj
–1 = ai dxi = F ú [ai dxi ] = aj dt (3.6) (6)
C C a a dt

3.1(3)(i) Higher-dimensional cross product

Ai · (A1 ◊ · · · ◊ Ai ◊ · · · ◊ An≠1 ) := voln (Ai , A1 , . . . , Ai , . . . An≠1 ) = 0

3.3 Stokes’ Theorem

3.3(1) ... in R3
• p=2

Ê 1 = w1 dx1 + w2 dx2 + w3 dx3

3 4 3 4
1 ˆw3 ˆw2 2 3 ˆw3 ˆw1
∆ dÊ = + dx · dx + + dx1 · dx3
ˆx2 ˆx3 ˆx1 ˆx3
3 4
ˆw2 ˆw1
+ + dx1 · dx2
ˆx1 ˆx2
This corresponds to the classical Stokes’ Theorem
⁄ ⁄
rot (W) dA = Wds
A ˆA

• p=3

Ê 2 = w12 dx1 · dx2 + w13 dx1 · dx3 + w23 dx2 · dx3

3 4
ˆw23 ˆw31 ˆw12
∆ dÊ 2 = + + dx1 · dx2 · dx3
ˆx1 ˆx2 ˆx3
This corresponds to Gauß’s Law
⁄ ⁄
div (W) dV = WdA
V ˆV

3.3(2) ... in R4
• p=2

Ê 1 = w1 dx1 + w2 dx2 + w3 dx3 + w4 dx4

3 4 3 4
ˆw1 ˆw2 ˆw1 ˆw3
∆ dÊ 1 = ≠ dx1
· dx 2
+ ≠ dx1 · dx3
ˆx2 ˆx1 ˆx3 ˆx1
3 4 3 4
ˆw1 ˆw4 1 4 ˆw2 ˆw3
+ ≠ dx · dx + ≠ dx2 · dx3
ˆx4 ˆx1 ˆx3 ˆx2
3 4 3 4
ˆw2 ˆw4 2 4 ˆw3 ˆw4
+ ≠ dx · dx + ≠ dx3 · dx4
ˆx4 ˆx2 ˆx4 ˆx3
 It could be said to be some analogon to the classical Stokes’ Theorem in R4
⁄ ⁄
curl (W) dA = Wds
A ˆA

• p=3

Ê 2 = w12 dx1 · dx2 + w13 dx1 · dx3 + w14 dx1 · dx4

+ w23 dx2 · dx3 + w24 dx2 · dx4 + w34 dx3 · dx4
3 4
ˆw12 ˆw13 ˆw23
∆ dÊ 2 = ≠ + dx1 · dx2 · dx3
ˆx3 ˆx2 ˆx1
3 4
ˆw12 ˆw14 ˆw24
+ ≠ + dx1 · dx2 · dx4
ˆx4 ˆx2 ˆx1
3 4
ˆw13 ˆw14 ˆw34
+ ≠ + dx1 · dx3 · dx4
ˆx4 ˆx3 ˆx1
3 4
ˆw23 ˆw24 ˆw34
+ ≠ + dx2 · dx3 · dx4
ˆx4 ˆx3 ˆx2

The classical analogon is of obviously

⁄ ⁄
wtf(W) dV = WdA
V ˆV

• p=4
Ê 3 and dÊ 3 have already been calculated in 2.6(1). Using these forms, one gets a 4-dimensional analogon
to Gauß’s Theorem
⁄ ⁄
div (W) dH = WdV
H ˆH

The Lie derivative

4.1 The Lie Derivative of a Vector Field
4.1a. The Lie Bracket
X, Y - vector fields on M
„(t) = „t be local flow generated by field X
„t x - pt. t seconds along integral curve at X, the “orbit” of x, starts at time 0 at pt. x.
Compare Y„t x at that pt. with results of pushing Yx to pt. „t x by differential „tú
Figure 4.1.
Lie derivative of Y with respect to X.

[Y„t x ≠ „tú Yx ]
[LX Y ]x © lim = (4.1) (7)
tæ0 t
[„≠tú Y„t x ≠ Yx ] [„≠tú Y„t x ≠ Yx ]
= lim „tú = lim (4.2) (8)
tæ0 t tæ0 t
Hadamard’s Lemma. (4.3)
Let f be cont. diff. in neighborhood U of x0
Then for sufficiently small t, ÷ g = g(t, x) = gt (x) cont. diff. in t, pt. x œ U s.t.

g0 (x) = Xx (f )
f („t x) = f (x) + tgt (x)

i.e.
f ¶ „t = f + tgt
If we accept this for the moment, we many proceed with ÷ of limit.
 At x
[Y„t x ≠ „tú Yx ] (2.60) Y„ x (f ) ≠ Yx (f ¶ „t ) Y„ x (f ) ≠ Yx (f + tgt )
(f ) = lim (f ) = lim t = lim t =
t¶0 t tæ0 t tæ0 t
[Y„t x (f ) ≠ Yx (f )] tangent vector def. on integral curve
= lim ≠ lim Yx (gt ) = Xx [Y (f )] ≠ Yx (g0 ) = Xx {Y (f )} ≠ Yx {X(f )}
tæ0 t tæ0

remark (4.2)
d
LX Yx = {
(„≠t )ú Y„t x }t=0 (4.7) (9)
dt
Proof of Hadamard’s Lemma: Define F (t, x) = (f ¶ „t )(x)
fix t, x, put F(s) = F (st, x)
s1 s1 d s1
Then (f ¶ „t )(x) ≠ f (x) = F(1) ≠ F(0) = 0 F Õ (s)ds = 0 ds F (st, x)ds = 0 tF1 (st, x)ds
F1 denotes derivativeswith respect to 1st. variable.
1
Thus, define gt (x) © 0 F1 (st, x)ds
then (f ¶ „t )(x) ≠ f (x) = tgt (x)

4.1b. Jacobi’s Variational Equation

Use the fact that X j = dxj
dt along the orbit
3 4
dY i ÿ ˆX i
[LX Y ]i = ≠ Yj (10)
dt j
ˆxj
3 4
ˆW i ˆV j ˆ
LV W = [V, W ] = Vi i
≠ Wi i
ˆx ˆx ˆxj
W j = W j (◊(t))
If ◊˙i = V i
ˆW j dW j
Vi i
=
ˆx dt
j
dW ˆV j
(LV W )j = ≠ Wi i (4.8) (11)
dt ˆx
-
d --
(LV W )p = d(◊≠t )◊t (x) (W◊t (x) )
dt -t=0
-
ˆ◊i ˆ --
d(◊≠t )◊t (x) (W◊t (x) ) = (≠t, ◊(t, x))W j
(◊(t, x))
ˆxj ˆxi -x
For 5 6
ˆ 1
W = =
ˆx 0
5 6 5 6
x(t) x cos (t) ≠ y sin (t)
◊= =
y(t) x sin (t) + y cos (t)
so that
5 65 6 5 6 d 5 6 5 6
ˆ◊i cos ((≠t)) ≠ sin ((≠t)) 1 cos (t) ≠ sin (t) t=0 0
(≠t, ◊(t, x))W (◊(t, x)) =
j
= dt
≠æ ≠≠æ
ˆxj sin ((≠t)) cos ((≠t)) 0 ≠ sin (t) ≠ cos (t) ≠1

ˆ ˆ
=∆ (LV W )p = ≠ = LV
ˆy ˆx

4.1(1) Coordinate expression for [X, Y]

3 4 3 4
ˆ j ˆ j ˆ i ˆ
[X, Y] = X(Y) ≠ Y(X) = X Yi
≠Y X
ˆui ˆuj ˆuj ˆui
2 ⇠ 2 ⇠
j ⇠ˆ⇠ j ⇠ˆ⇠
j i
ˆY ˆ i ⇠ j ˆX ˆ i ⇠
= Xi i + X
⇠ ⇠Y ≠ Y ≠ Y ⇠X
3 ˆu ˆu
i i
4ˆu ˆu
j ˆuj ˆui ⇠ ˆuj ˆuj
j j
ˆY ˆX ˆ
= Xi i ≠ Y i
ˆu ˆui ˆuj
4.2 The Lie Derivative of a Form
If a flow deforms some attribute, say volume, how does one measure the deformation? -Theodore Frankel

4.2a. Lie Derivatives of Forms

4.2b. Formulas Involving the Lie Derivative
Theorem 7 (4.24)
LX iY ≠ iY LX = i[X,Y ] (12)

Theorem 8 (4.25)
d–(X, Y ) = X(–(Y )) ≠ Y (–(X)) ≠ –([X, Y ]) (13)

Proof :
d–(X, Y ) = (iX d–)(Y ) = (LX – ≠ diX –)(Y ) = iY LX – ≠ Y (–(X)) = LX iY – ≠ i[X,Y ] – ≠ Y (–(X)) =
= LX –(Y ) ≠ –([X, Y ]) ≠ Y (–(X)) = X(–(Y )) ≠ Y (–(X)) ≠ –([X, Y ])

Note that
diX – = d(–(X))
ˆ(–(X)) i
d(–(X))(Y ) = Y = Y (–(X))
ˆxi
Done.

4.2(1) Coordinate expression for LX –1

LX –1 = iX d– + diX – = iX d–i dui + diX –i dui = iX d–i · dui + d–i dui (X)
ˆ–i j ˆ–i j i ˆX i j
= iX du · du i
+ du X + –i du
ˆuj ˆuj ˆuj
ˆ–i ! " ˆ–i j ! " ˆ–i ˆX i
= j
iX duj dui ≠ j
du iX dui + j duj X i + –i j duj
ˆu ¸ ˚˙ ˝ ˆu ¸ ˚˙ ˝ ˆu ˆu
=X j =X i
ˆ–i j i ˆ–j j i ˆ–j j i ˆX j i
= X du ≠ X du + X du +–j du
ˆuj ¸ ˆu
i
˚˙ ˆu
i
˝ ˆui
=0
3 4
j ˆ–i ˆX j
= X + –j dui
ˆuj ˆui

4.2(2) Compositions of derivations and antiderivations

(◊A ≠ A◊)(–p · — q ) = ◊(A– · — + (≠1)p – · A—) ≠ A(◊– · — + – · ◊—)

= ◊A– · — + A– · ◊— + (≠1)p ◊– · A— + (≠1)p – · ◊A—
≠ A◊– · — ≠ (≠1)deg(◊–) ◊– · A— ≠ A– · ◊— ≠ (≠1)p – · A◊—
(Notice that derivations alter their argument’s degree by
an even number; thus the 2nd and 7th, and the 3rd and 6th
summand cancel each other out)
= (◊A ≠ A◊)– · — + (≠1)p – · (◊A ≠ A◊)—
 (AB ≠ BA)(–p · — q ) = A(B– · — + (≠1)p – · B—) ≠ B(A– · — + (≠1)p – · A—)
= AB– · — + (≠1)deg(B–) B– · A— + (≠1)p A– · B—
+ (≠1)p (≠1)p – · AB— + BA– · — + (≠1)deg(A–) A– · B—
¸ ˚˙ ˝
=1
+ (≠1)p B– · A— + (≠1)p (≠1)p – · BA—
¸ ˚˙ ˝
=1
(Again, the 2nd and 7th, 2nd and 6th summands cancel out
as an antiderivation alters its argument’s degree by an
uneven number)
= (AB + BA)– · — + – · (AB + BA)—

4.2(3) i[X,Y] = LX ¶ iY ≠ iY ¶ LX
As stated in the corresponding chapter, it’s enough to verify the formula for functions and differentials of
functions.
Functions:
i[X,Y] f = 0
LX iY f ≠ iY LX f = 0
¸˚˙˝ ¸ ˚˙ ˝
=0 =0

Differentials:
i[X,Y] df = df ([X, Y])
= [X, Y](f )
LX iY df ≠ iY LX df = iX diY df + d iX iY df ≠iY iX ¸˚˙˝
dd f ≠ iY diX df
¸ ˚˙ ˝
=0 =0

= iX dY(f ) ≠ iY dX(f ) = X(Y(f )) ≠ Y(X(f ))

= [X, Y](f )

4.2(3) Fugly proof of d–(X, Y) = X(–(Y)) ≠ Y(–(X)) ≠ –([X, Y])

Step 1: Calculate single terms.
3 4 1 2
k l ˆ–i ˆ ˆ k l ˆ–i
d–(X, Y) = X Y du j
· du i
, = X Y ” j i
” l ≠ ” j i
” k
ˆuj ˆuk ˆul ˆuj k l

ˆ–i ˆ–i
= X j Y i j ≠ X iY j j
ˆu ˆu
ˆ ! " ˆ–j ˆY j
X(–(Y)) = X i i –j Y j = X i Y j i + –j X i i
ˆu ˆu ˆu
i ˆ
! " j i ˆ–j i ˆX
j
Y(–(X)) = Y i
–j X = X Y
j
i
+ –j Y i
ˆu 33 ˆu ˆu4 4
i i i i
ˆY ˆX j ˆY j ˆX
–([X, Y]) = –([X, Y]i ˆ i ) = – X j
≠ Y j
ˆ i = – i X ≠ –i Y
ˆuj ˆuj ˆuj ˆuj
Step 2: Smash them together.
X(–(Y)) ≠ Y(–(X)) ≠ –([X, Y])
j j i i
ˆ–j i ˆY j i ˆ–j i ˆX j ˆY j ˆX
= X iY j + –j X ≠ X Y ≠ –j Y ≠ –i X + –i Y
ˆui ˆui ˆui ˆui ˆuj ˆuj
j i i j
ˆ– ˆ– ˆY ˆY ˆX ˆX
= X i Y j i ≠ X j Y i i +–j X i i ≠ –i X j j +–i Y j j ≠ –j Y i
j j
ˆu ˚˙ ˆu ˝ ¸ ˆu ˚˙ ˆu ˝ ¸ ˆu ˚˙ i
ˆu ˝
¸
=d–(X,Y) =0 =0

= d–(X, Y)

4.3. Differentiation of Integrals

How does one compute the rate of change of an integral when the domain of integration is also
changing?
4.3a. The Autonomous (Time-Independent) Case
Let – p-form

V oriented, compact submanifold of M , dimV = p

flow „t : M æ M , i.e. 1-parameter “group” of diffeomorphisms „t

defined V (t) © „t V
Fig. 4.5. EY : 20141031
- I don’t get
- this
Let X = d „t (x)- . X = „˙ t (x)-
dt t=0 t=0
⁄ ⁄
I(t) = –= „út –
V (t) V
s s 5⁄ 6
[I(t + h) ≠ I(t)] [ „ú – ≠ V „út –] „út {„úh – ≠ –}
I Õ (t) = lim = lim V t+h = lim =
hæ0 h hæ0 h hæ0 V h
C⁄ D ⁄
{„úh – ≠ –} {„úh – ≠ –}
= lim = lim
hæ0 V (t) h V (t) hæ0 h
EY: 20141031 I don’t understand the steps in between the lines, equality 4; what happened to the „út out in
front?

4.3b. Time-Dependent Fields

A time-dependen vector field on a manifold M does not generate a flow!
’ time-dependent tensor field A(t, x) on M , should be considered a tensor field on product manifold R ◊ M
R ◊ M has local coordinates (t = x0 , x1 . . . xn )
solve the system of ODE

dxi
= v i (t, x) xi (s = 0) = xi0 , i = 1...n
ds (4.39)
dt
=1 t(s = 0) = t0
ds
get a flow „s : R ◊ M æ R ◊ M

4.3c. Differentiating Integrals

Let „t : M æ M 1-parameter family of diffeomorphisms of M
don’t assume they form a flow, assume „0 = 1 and (t, x) æ „t x smooth as a function of (t, x) œ R ◊ M
Let Êt (x) = Ê(t, x) œ p (M ) be 1 parameter family of forms on M
Let V ™ M submanifold, dimV = p

Problems
4.3(1) A, B time dependent vector fields on R3
fl(t, x) function
Using ⁄ ⁄ ⁄ ⁄
d d
–= –= LX – = Lv+ ˆ –
dt V (t) dt W (t) W (t) W (t)
ˆt

if p = 1,
⁄ ⁄ ⁄ ⁄ ⁄
d d ˆ– ˆ–
–= –= Lv+ ˆ – = Lv – + = + iv d– + div –
dt V (t) dt W (t) W (t)
ˆt
W (t) ˆt W (t) ˆt

Additional Problems on Fluid Flow

4.3(5)
(i)
 i
(ii) For circulation C(t)
u˜ ,
j ⁄ ⁄ 3 ⁄ 4
d ˆu˜ 1 dp
u =
˜
+ Lu u˜ = d ÎuÎ2 ≠ „ ≠ =0
dt C(t) C(t) ˆt C(t) 2 fl

since C(t) is a closed curve. Then the circulation is constant in time.

2
(iii) vorticity Ê œ (M )
Ê := du˜
For some compact submanifold S µ M , dimS = 2,
⁄ ⁄ ⁄ ⁄ ⁄ ⁄ 3 ˜ 4
d ˆÊ ˆÊ ˆdu˜ ˆu
Ê= L ˆ +u Ê = + Lu Ê = + diu Ê + iu dÊ = + diu Ê = d + iu Ê =
dt S S
ˆt
S ˆt S ˆt S ˆt S ˆt
⁄ ⁄ ⁄
ˆu˜ ˆu˜
= ( + iu Ê) = ( + Lu u˜ ≠ diu u˜ ) = 0 ≠ du2 = 0
ˆS ˆt ˆS ˆt ˆS

since ˆS is a closed curve.

(iv)

4.4 A problem set on Hamiltonian mechanics

4.4(1) Symplectic form
Ê is obviously closed as Ê = d⁄. In order to show non-degeneracy, let
ˆ ˆ
X = Qi i
+ Pi
ˆq ˆpi
X ”= 0

Then
! "
iX Ê = iX dpi · dq i = (iX dpi ) dq i ≠ dpi iX dq i = Pi dq i ≠ Qi dpi ”= 0

i.e. there is no Y ”= 0 so that iY iX Ê = Ê(X, Y) = 0 for all X ”= 0, so Ê is a non-degenerate bilinearform.

4.4(1) Symplectic volume form

n
fi n
fi
Ê n := Ê= dpik · dq ik = dpi1 · dq i1 · dpi2 · dq i2 · . . . · dpin · dq in
k=1 k=1

All summands with equal indices vanish, only distinct ik indices yield a term, thus there are (n ≠ k + 1) choices
for ik . Combine them all to get a total of
n
Ÿ
(n ≠ k + 1) = (n ≠ 1 + 1)(n ≠ 2 + 1) · · · (n ≠ n + 1) = n(n ≠ 1) · · · 1 = n!
k=1

So n! choices exist. Next, rearrange the “wedge factors” so the indices are in ascending order, yielding a factor
of ±1. Now

Ê n = ±n! dp1 · dq 1 · dp2 · dq 2 · . . . · dpn · dq n

P. 147: Derivation of Hamilton’s equations The paragraph below (4.49) says “comparing these two expres-
sions” and doesn’t explain it any further. This is what’s happening.
Let

H = H(q, p, t) = pi q̇ i ≠ L(q, q̇, t)

Then
ˆH i ˆH ˆH
dH = dH(q, p, t) = dq + dpi + dt
ˆq i ˆpi ˆt
but also
Z
dH = d(pi q̇ i ≠ L(q, q̇, t)) = dpi q̇ i + H
ˆL ˆL i ˆL
pi dH
q̇ i ≠
H dq i ≠ Z dq̇ ≠ dt
ˆq i ˆ q̇ i Z ˆt
¸˚˙˝ ¸ ˚˙ Z ˝
d
= dt ˆL
=ṗi =≠pi dq̇ i
ˆ q̇ i

ˆL
= ≠ṗi dq i + q̇ i dpi ≠ dt
ˆt
Comparing these two results for dH yields Hamilton’s equations
ˆH ˆH ˆL ˆH
q̇ i = ṗi = ≠ =≠
ˆpi ˆq i ˆt ˆt

4.4(4) Hamilton in shrt

3 4
ˆ ˆ
iX Ê = dpi · dq i X j j + X n+j
ˆq ˆpj
3 4 3 4
ˆ ˆ
= dpi · dq i X j j + dpi · dq i X n+j
ˆq ˆpj
3 4 3 4 3 4 3 4
ˆ ˆ ˆ ˆ
= dpi X j dq i ≠ dq i X j j dpi + dpi X n+j dq i ≠ dq i X n+j dpi
ˆqj ˆq ˆpj ˆpj
¸ ˚˙ ˝ ¸ ˚˙ ˝ ¸ ˚˙ ˝ ¸ ˚˙ ˝
=0 =X i =X n+i =0
ÿ dqi
dpi i ˆH i ˆH
= ≠X i dpi + X n+i dq i = ≠ dpi + dq = ≠ dq ≠ i dpi = ≠dH(q, p)
i
dt dt ˆpi ˆq

4.4(5) Lie derivative of the symplectic Poincaré 2-form

LX Ê = iX dÊ + diX Ê = iX d2 ⁄ ≠ d2 H = 0

Since L is a derivation on the exterior algebra, LX Ê n vanishes as well.

4.4(8) Hmltn n shrtr

This is basically the same procedure as in 4.4(4).

ˆq i ˆ ˆpi ˆ ˆ
X= + +
ˆt ˆq i ˆt ˆpi ˆt

0 = iX = iX (dpi · dq i ≠ dH · t) = (iX dpi )dq i ≠ dpi (iX dq i ) ≠ (iX dH)dt + dH (iX dt)
¸ ˚˙ ˝
=1
3 4
ˆpi i ˆq i ˆH i ˆH ˆH ˆH ˆH ˆH
= dq ≠ pi ≠ iX i
dq + dp i + dt dt + i dq i + dpi + dt
ˆt ˆd ˆq ˆpi ˆt ˆq ˆpi ˆt
ˆpi i ˆq i ˆH ˆq i ˆH ˆpi ˆH ˆH ˆH ˆH
= dq ≠ pi ≠ i dt ≠ dt ≠ dt + i dq i + dpi + dt
ˆt ˆd ˆq ˆt ˆpi ˆt ˆt ˆq ˆpi ˆt
¸ ˚˙ ˝
=≠ dH
dt dt=≠ ˆt dt
ˆH

3 4 3 4 3 4
ˆpi ˆH ˆq i ˆH ˆH ˆH
= + i dq i + ≠ + dpi + ≠ + dt
ˆt ˆq ˆt ˆpi ˆt ˆt
ˆH ˆH
∆ q̇ i = ; ṗi = ≠ i
ˆpi ˆq
I don’t think it can still become any shorter.

4.4(9) Lie derivative of the pre-symplectic Poincaré 2-form

LX = iX d + d iX = iX d2 =0
¸˚˙˝
=0
The Poincare Lemma and Potentials
5.1. A More General Stokes’s Theorem
Let V compact oriented submanifold of M n
smooth F : M n æ W m
F (V ) µ W need not be a submanifold, might have self-interactions, pathologies.
⁄ ⁄
— =
p
F ú—p (5.1) (14)
F (V ) V

generalizes (3.17), def.

⁄ ⁄ ⁄ ⁄ ⁄
d— p≠1
= ú
F d— p≠1
= dF — ú p≠1
= ú p≠1
F — = — p≠1
F (V ) V V ˆV F (ˆV )

(question, 2nd., 3rd. equality)

Answer: recall Naturality (cf. wikipedia exterior derivative)

contravariant smooth functor M ‘æ (M )
k
k k

Fú
k
(N ) k
(M )

d d

Fú
k+1
(N ) k+1
(M )
So that
dF ú = F ú d

define ˆF (V ) = F (ˆV )

generalized Stoke’s thm. ⁄ ⁄

d— p≠1
= — p≠1 (5.2) (15)
F (V ) ˆF (V )

manifold needs only “piecewise smooth” boundaries.

5.2. Closed Forms and Exact Forms

— p closed if d— = 0
— p exact if — p = d–p≠1 , some –p≠1

Theorem 9 (5.3) Let M n with 1st Betti number 0, b1 = 0, i.e. ’ closed oriented piecewise smooth curve C
is the boundary of some compact oriented “surface”. Then ’ closed 1-form — 1 on M n is exact.

Let x, y œ M , y fixed.
oriented C(y, x) starts at y, ends at x
define ⁄
f (x) © —1
C(y,x)

If ÷ another C 1 (y, x), then C ≠ C Õ closed oriented curve.

By given, ÷ oriented compact surface F (V ) s.t. ˆF (V ) = C.
⁄ ⁄ ⁄ j ⁄
—≠ —= —= —= d— = 0
C CÕ C≠C Õ ˆF (V ) F (V )
s s
— = C Õ —. f independent of curve.
C
Let vx vector at x.
Let vector field v coincide with vx at x, defined in neighborhood of curve C(y, x), v = 0 at y.
 „t flow generated by v, „t C(y, x) curve joining y to „i x
5 6
d„i x
= vf
dt
C ⁄t=0 D ⁄
d d
df (v) = f {„t x}t=0 = — = Lv — =
dt dt „t C(y,x) C(y,x)
t=0

5.3. Complex Analysis

5.5 Finding potentials
5.5(1) Product of a closed and an exact form
Let Ÿ be a closed k-form and Á be an exact form with dÁ̃ = Á. Then
! "
Ÿ · Á = Ÿ · dÁ̃ = (≠1)k (d(Ÿ · Á̃) ≠ ¸˚˙˝
dŸ · Á̃) = d (≠1)k Ÿ · Á̃
=0

6 Holonomic and Nonholonomic Constraints

6.1. The Robenius Integrability Condition
Can one always find a surface orthogonal to a family of curves in R3 ?

6.2. Integrability and Constraints

6.3. Heuristic Thermodynamics via Caratheodory
Can one go adiabatically from some state to any nearby state?

6.3a. Introduction
6.3b. The First Law of Thermodynamics
Consider system of regions of fluids separated by “diathermous” membranes
allow only passage of heat, not fluids

assume system connected

assume each state in thermal equilibrium
Let pi , vi (uniform) pressure and volume of ith region
at thermal equilibrium, “equations of state” pi vi = ni RTi
pi vi = ni RTi eliminate all but 1 pressure

=∆ p1 , v1 , v2 , . . . vn

assume globally defined energy function U

path in M n+1 represents sequence of states each in equilibrium, i.e. assume very slow changes in time,
quasi-static irreversible processes, e.g. “stirring”
on M , dimM = n + 1, assume ÷ work 1-form W , work done by system

W = pi dvi = pi (U, v1 . . . vn )dvi i = 1...n

heat 1-form, heat added or removed from system, assume Q ”= 0

n
ÿ
Q= Qi (U, v1 . . . vn )dvi (v0 = U )
i=0

1st. law of thermodynamics

dU = Q ≠ W
energy conservation
6.3c. Some Elementary Changes of State
1. Heating at constant volume.

path “I œ M , dimM = n + 1 s.t. dv1 = · · · = dvn = 0. W = 0 dU = Q0 dU . “˙ I = c0 ˆU

ˆ

2. Quasi-static adiabatic process. No heat exchanged,

Q(“˙ II ) = 0 so dU = ≠W
3. Stirring at constant volume adiabatic, but not quasistatic
Q, W makes no sense but
work is being done by (or on) system, U (y Õ ) ≠ U (x), difference of internal energy
assume connected mechanical manifold V , dimV = n
diff. fi : M æ V
fi onto
fiú onto
fi submersion
By main thm. on submanifolds of Sec. 1.3d,
if v œ V , then fi ≠1 (v) 1-dim. embedded submanifold of M
assume ’ fi ≠1 (v) connected, we’re assuming given any pair of states
lying on fi ≠1 (v), 1 of them can be obtained by other by “heating at constant volume”
assume W on M is 0 when W |fi≠1 (v) = 0
on the other hand, Q ”= 0 on fi ≠1 (v); dU = Q ”= 0 (first law)
(U, v 1 . . . v n ) local coordinate system for M (U global coordinate)

6.3d. The Second Law of Thermodynamics

cyclic process starts and ends at the same state
Kelvin 2nd. law of thermodynamics
@ quasistatic cyclic process can Q converted entirely into W

Caratheodory (1909) 2nd. law of thermodynamics

’ neighborhood N – state x, ÷ y not accessible from x via quasistatic adiabatic paths, i.e. paths s.t. Q = 0

II Geometry and Topology

R3 and Minkowski Space
7.1 Curvature and Special Relativity
7.1.a. Curvature of a Space Curve in R3
! ds "2 st
x = x(t) dt = v2 s(t) = 0 Îx(u)Îdu
ÎvÎ = v
dx dx ds
ẋ = =v= = Tv
dt ds dt
d2 s dT ds dT
a = ẍ = v = v̇T + v Ṫ = 2 T + v = v̇T + v 2
dt ds dt ds
dT
v ◊ a = v3 T ◊ = v 3 Ÿ(s)T ◊ n
ds
so 3 4
dx dt v
unit tangent vector T = = ẋ =
ds ds v
Note that
dT 1 d 1 d
·T= (T · T) = (1) = 0
ds 2 ds 2 ds
Now
dT
= Ÿ(s)n(s) (7.1)
ds
where n principal normal, Ÿ(s) Ø 0 curvature of C.
Îv ◊ a
=∆ Ÿ =
v3
7.1(1)

x = cos (Êt)
y = sin (Êt)
z = kt
- - Q R
- ex
- ey ez -- kÊ 2 s(Êt)
v ◊ a = -- ≠ÊsÊt ÊcÊt k -- = a≠kÊ 2 c(Êt)b
-≠Ê 2 cÊt ≠Ê 2 sÊt 0 - Ê3
Ô
Îv ◊ aÎ k2 Ê4 + Ê6 Ê2
=∆ Ÿ = =  =
v3 (Ê 2 + k 2 )3 Ê2 + k2

7.1(2)

Given B = T ◊ n,
dB dT dn dn
= ◊n+T◊ =T◊
ds ds ds ds
so 3 4
dB dn dn dn
n◊ = n ◊ (T ◊ )= n◊ T ≠ (n ◊ T) =0
ds ds ds ds
Indeed
dB
T· =0
ds
dB d d
B· = (B · B) = (1) = 0
ds ds ds
Then dB
ds Î n.

Define torsion dB
ds = · (s)n

Using CAB ≠ BAC,

n ◊ B = n ◊ (T ◊ n) = T
T ◊ B = T ◊ (T ◊ n) = (n · T)T ≠ (T · T)n = ≠n
=∆ n = B ◊ T
So
dn dB dT dn
= ◊T+B◊ = · n ◊ T + B ◊ Ÿn = ≠· B + ≠ŸT =
ds ds ds ds

7.2 Electromagnetism in Minkowski Space

7.2(3) Field strength 2-Form
Notation: dx0 = dt; dxij··· = dxi · dxj · · · ·. The expansion for úF was taken from (14.20) combined with
(3.41).

F · F = (Ei dxi0 + BJ={1,2,3} dxJ ) · (Ek dxk0 + BL={1,2,3} dxL )

X i0k0
X + BJ BLX
= ≠Ei Ek dxX dxX X + Ei BL dxi0L + BJ Ek dxJk0
JL

= ≠2Ei BJ dx0iJ
= ≠2(E1 B23 dx0123 + E2 B13 dx0213 + E3 B12 dx0312 )
0123
= ≠2 (E1 B23 + E2 B31 + E3 B12 ) dx
¸ ˚˙ ˝ ¸ ˚˙ ˝
=ÈE,BÍ =vol4

= ≠2 ÈE, BÍ vol4
F · úF = (Ei dxi0 + BJ={1,2,3} dxJ ) · (≠(úB)k dxk0 + (úE)L={1,2,3} dxL )
X i0k0
X + BJ E úX
= ≠Ei B ú dxX
j dxX
L
X + Ei E ú dx0iL ≠ BJ B ú dx0Jk
JL
L k
0kJ
= Bk BJ dx
ú
≠ Ei EL dx0iL
ú

= Bkú BJ dx0kJ ≠ Ek EJú dx0kJ

= (Bkú BJ ≠ Ek EJú )dx0kJ
(permute k, J so their combination is in increasing order;
permuting the double indices of B, E cancels out the minus)
= (ÎBÎ2 ≠ ÎEÎ2 ) vol4

The Geometry of Surfaces in R3

9 Covariant Differentiation and Curvature
9.1 Covariant Differentiation

Definition 10 affine connection or covariant differentiation is operator Ò(X, v) ‘æ ÒX v vector X at p

vector field v at p
ÒX (av + bw) = aÒX v + bÒX w
(9.2) (16)
ÒaX+bY v = aÒX v + bÒY v

ÒX (f v) = X(f )v + f ÒX v (“Leibniz rule”)

demand if X smooth, ÒX v smooth vector field.

in our work up until now, we have always used local coordinates x to yield a basis ˆ
ˆxi for tangent vectors in
a patch U .
For many purposes, however, it is advantageous to use a more general basis.
frame of vector fields in U - n linearly independent smooth vector fields

e = (e1 . . . en )

coordinate frame = special case, ei = ˆx

ˆ
i for some coordinate system x in U

frame e usually not coordinate frame, since [ei , ej ] usually not 0 while [ˆi , ˆj ] = 0

Theorem 11 (9.3) frame e is locally a coordinate frame iff

[ei , ej ] = 0 ’ i, j

Proof:
We need only show that [ei , ej ] = 0 implies ÷ functions (xi ) such taht

ˆ
ei =
ˆxi
Let ‡ be the dual form basis. From (4.25)

d‡ i (ej , ek ) = ≠‡ i ([ej , ek ]) (9.4) (17)

 Let e = (e1 . . . en ) frame in U . Then X = ej X j
(9.2)
≠≠≠æ ÒX (ek v k ) = X(v k )ek + v k ÒX ek = X j ej v k ek + v k X j Òej ek = X j ej (v k )ek + X j ei Êjk
i
vk =
(9.5)
= X j ei Êjk
i
v k + X j ej (v k )ek
(18)
where Êjk
i
defined

Òej ek = ei Êjk
i
(9.6) (19)
when ej = ˆj coordinate frame, Êjk
i
= i
jk
since X(v ) = dv (X)
k k

ÒX v = ei {dv i (X) + X j Êjk

i
vk }
i
Êjk coefficients of affine connection.

using dual basis ‡ of 1-forms,

ÒX v = ei {dv i (X) + Êjk

i
‡ j (X)v k } = ei {dv i + Êjk
i
‡ j v k }(X) (9.7) (20)
i.e.

ÒX v = dv(X) + Êjk
i
+ Êjk
i
‡ j (X)v k ei (21)

when frame e is coordinate frame ei = ˆi = ˆxi ,

ˆ
‡ i = dxi

ˆv i
ÒX v = ˆi { + Êjk
i
v k }dxj (X)
ˆxj
i.e.
5 6
ˆv i
(ÒX v)i = + Ê i
jk v k
Xj (9.8) (22)
ˆxj
since ÒX v assumed to be vector, conclude
- ˆv i
Òj v i = v i -j © + Êjk
i
vk (9.9) (23)
ˆxj
form the components of a mixed tensor, covariant derivative of vector v.

9.3 Cartan’s Exterior Covariant Differential

9.3c. Cartan’s Structural Equations
denote
row matrix e © (e1 . . . en )
column ‡ © (‡ 1 . . . ‡ n )T

n ◊ n matrix of connection 1-forms Ê = (Ê ij )

column vector of torsion 2-forms · = (· 1 . . . · n )T

9.3d. The Exterior Covariant Differential of a Vector-Valued Form

– vector-valued p-form
locally, – = ei ¢ –i , ’ –i = aiJ (x)‡ J locally defined p-form

exterior covariant differential, vector-valued (p + 1) form Ò–

defined by Leibniz rule
Ò– = Ò(ei ¢ –i ) = (Òei ) ¢· –i + ei ¢ d–i
where
(Òei ) ¢· –i = (ek ¢ Ê ki ) ¢· –i © ek ¢ (Ê ki · –i )
column of p forms – = (–1 . . . –n )T

Ò– = e ¢ (d– + Ê · –) (9.31) (24)

9.3(1) Basis expansion of the curvature form
! " ! " ! "
◊ij = dÊ ij + Ê ir · Ê rj = d Ê i¸j du¸ + Ê ikr duk · Ê r¸j du¸
! " 1! "
= ˆk Ê i¸j + Ê ikr Ê r¸j ¸duk ˚˙
· du˝¸ = (k¸ )duk¸ + (k¸ )duk¸
¸ ˚˙ ˝ 2
© du
k¸
© “(k¸ )”

(In the second summand, commute the wedge product,

afterwards rename k ¡ ¸)
1 ! "
= (k¸ )duk¸ ≠ (¸k )duk¸
2
1! "
= ˆk Ê i¸j ≠ ˆ¸ Ê ikj + Ê ikr Ê r¸j ≠ Ê i¸r Ê rkj duk · du¸
2¸ ˚˙ ˝
=Rijk¸

1 i
= R duk · du¸
2 jk¸

9.3(2) Covariant derivative of the identity form

! " ! "
Ò“dr” = Ò ei ¢ ‡ i = ei ¢ d‡ i + Ê ij · ‡ j = ei ¢ · i
¸ ˚˙ ˝
=· i

Remark: The reason for calling ei ¢ ‡ i the identity form is because

ei ¢ ‡ i (v) = ei ¢ ‡ i (v j ej ) = ei v j ‡ i (ej ) = ei v i = v
¸ ˚˙ ˝
=”ji

9.4 Change of Basis and Gauge Transformations

9.4(1) Transformation of the curvature form
For readability, let P̄ © P ≠1 .

◊Õ = dÊ Õ + Ê Õ · Ê Õ
= d(P̄ ÊP + P̄ dP )
+ (P̄ ÊP + P̄ dP ) · (P̄ ÊP + P̄ dP )
= d(P̄ ÊP ) + d(P̄ dP )
+ P̄ ÊP · P̄ ÊP + P̄ ÊP · P̄ dP + P̄ dP · P̄ ÊP + P̄ dP · P̄ dP
2⇠
= dP̄ · ÊP + P̄ dÊP ≠ P̄ Ê · dP + dP̄ · dP + ⇠ P̄ d⇠P
+ P̄ ÊP · P̄ ÊP + P̄ ÊP · P̄ dP + P̄ dP · P̄ ÊP + P̄ dP · P̄ dP
(Use 0 = d1 = d(P̄ P ) = dP̄ P + P̄ dP … dP̄ = ≠P̄ dP P̄ ;
Also, the matrices “commute” with the wedge product, i.e. “A · B = AB · ”)
= ≠P̄ dP · P̄ ÊP + P̄ dÊP ≠ P̄ Ê · dP ≠ P̄ dP · P̄ dP
+ P̄ Ê · ÊP + P̄ Ê · dP + P̄ dP · P̄ ÊP + P̄ dP · P̄ dP
= P̄ dÊP + P̄ Ê · ÊP
= P̄ (dÊ + Ê · Ê)P
= P̄ ◊P

And this dear children is why indices should be left away. (Yes, it’s the same exercise.)
 ◊Õi j = dÊ Õi j + Ê Õi k · Ê Õk j
= d(P̄ i l Ê l m P m j + P̄ i l dP l j )
+ (P̄ i l Ê l m P m k + P̄ i l dP l k ) · (P̄ k n Ê n o P o j + P̄ k n dP n j )
= d(P̄ i l Ê l m P m j ) + d(P̄ i l dP l j )
+ P̄ i l Ê l m P m k · P̄ k n Ê n o P o j + P̄ i l Ê l m P m k · P̄ k n dP n j
+ P̄ i l dP l k · P̄ k n Ê n o P o j + P̄ i l dP l k · P̄ k n dP n j
⇠
2 ⇠
⇠
= dP̄ i l · Ê l m P m j + P̄ i l dÊ l m P m j ≠ P̄ i l Ê l m · dP m j + dP̄ i l · dP l j + ⇠ ld P j
P̄ i⇠ l

+ P̄ i l Ê l m P m k · P̄ k n Ê n o P o j + P̄ i l Ê l m P m k · P̄ k n dP n j
+ P̄ i l dP l k · P̄ k n Ê n o P o j + P̄ i l dP l k · P̄ k n dP n j
(Use 0 = d” i j = d(P̄ i k P k j ) = dP̄ i k P k j + P̄ i k dP k j … dP̄ i j = ≠P̄ i k dP k l P̄ l j ;
Also, the matrices “commute” with the wedge product, i.e. “Ai k · B k j = Ai k B k j · ”)
= ≠P̄ i r dP r s · P̄ s l Ê l m P m j + P̄ i l dÊ l m P m j ≠ P̄ i l Ê l m · dP m j ≠ P̄ i r dP r s · P̄ s l dP l j
+ P̄ i l Ê l m · Ê m o P o j + P̄ i l Ê l m · dP m j
+ P̄ i l dP l m · P̄ m n Ê n o P o j + P̄ i l dP l k · P̄ k n dP n j
= P̄ i l dÊ l m P m j + P̄ i l Ê l m · Ê m n P n j
= P̄ i l (dÊ l m + Ê l n · Ê n m )P m j
= P̄ i l ◊l m P m j

9.4(2) Transformation of the curvature form

The Transformation rule for basis vectors is
eÕ = eP … eÕi = ej P ji
The transformation from cartesian to polar coordinates is given by
3 4 3 4 3 4
x x(r, Ï) r cos (Ï)
= =
y y(r, Ï) r sin (Ï)
A B 3 4
cos (Ï) ≠r sin (Ï)
ˆx ˆx
ˆ(x, y)
P = = ˆy ˆy =
ˆr ˆÏ
ˆ(r, Ï) ˆr ˆÏ
sin (Ï) r cos (Ï)

Using Mathematica to skip the annoying 2nd semester homework assignment parts of finding the inverse and
calculating derivatives,
3 4
≠ sin (Ï) dÏ ≠ sin (Ï) dr ≠ r cos (Ï) dÏ
dP =
cos (Ï) dÏ cos (Ï) dr ≠ r sin (Ï) dÏ
3 4
cos (Ï) sin (Ï)
P =
≠1
≠ 1r sin (Ï) 1r cos (Ï)
Multiplying these two expressions yields, as desired,
3 4
⇠ÊP⇠
⇠ 0 ≠rdÏ
ÊÕ = ⇠
P ≠1 + P ≠1 dP = 1 1
r dÏ r dr

Since ◊ = 0, ◊Õ vanishes as well. This is obvious from the transformation behavior of ◊; direct computation
confirms this, as
◊Õ = dÊ Õ + Ê Õ · Ê Õ
3 4 3 4 3 4
0 ≠rdÏ 0 ≠rdÏ 0 ≠rdÏ
=d 1 1 + 1 1 · 1 1
dÏ r dr dÏ dr r dÏ r dr
3 r 4r 3 r ⇠ ⇠ ⇠⇠⇠ ≠ rdÏ · 1r dr
⇠ 4
d0 " d(≠rdÏ)
⇢ ⇠
0 · 0 ≠⇠
⇠ ⇠
⇠· r dÏ1
0 ·⇠(≠rdÏ)
= !⇢ ! ⇠ " +
rdÏ ⇠ (
( ⇠
d 1 dÏ d⇠
⇠
1⇠
r dr
1
⇠⇠
⇠dÏ ·⇠ 0 + 1r dr · 1r dÏ ( 1
r dÏ
(( ·( (
(≠rdÏ) 1
+⇠
r dr
⇠· ⇠⇠1
r dr
3 r 4r 3 4
0 ≠dr · dÏ 0 dr · dÏ
= + 1
≠ r12 dr · dÏ 0 r 2 dr · dÏ 0
3 4
0 0
=
0 0
This exercise made the advantage of the matrix notation clear: use the connection coefficients like normal
matrices, only that you put a wedge in between their components’ differential form “factors”.

Parallel Displacement and Curvature on a Surface

When is parallel displacement independent of path?
We saw in Section 8.7 that parallel displacement of a vector between 2 pts. of a surface is path-dependent;
This phenomenon is referred to as holonomy.

Theorem 12 (9.61) Let U µ M 2 compact in Riemannian surface with piecewise smooth boundary ˆU
Assume U covered by single orthonormal frame field e (e.g. U contained in coordinate patch)
Let unit vector v parallel translated around ˆU
e defined orientation.
Then angle – between v0 , vf is
⁄⁄ ⁄⁄
–= KdS = K‡ 1 · ‡ 2
U

Proof. parametrize ˆU , let T tangent, let – = arccos

! Èe1", vÍ
Although – (like v) is not single-valued on ˆU , d– = d–ds ds well-defined.
j
– = arccos Èv0 , vf Í = d–
ˆU

For
v = e1 cos (–) + e2 sin (–)
then

Òv = e(dv + Êv) = e1 (dv 1 + Ê12 v 2 ) + e2 (dv 2 + Ê21 v 1 ) = e1 (≠ sin (–) d– + Ê12 sin (–)) + e2 (cos (–) d– + Ê21 cos (–)) =
= (≠e1 sin (–) + e2 cos (–))(d– ≠ Ê12 )

To say that v parallel displaced around ˆU is to say

Òv(T ) = 0, i.e. d– ≠ Ê12 = 0 along ˆU (9.62)

d–(T ) = Ê12 (T )
Then j j ⁄⁄
–= d– = Ê12 = dÊ12 =
ˆU ˆU U
⁄⁄ ⁄⁄
= ◊12 = K‡ 1 · ‡ 2
U U
10 Geodesics
11 Relativity, Tensors, and Curvature
12 Curvature and Topology: Synge’s Theorem
13 Betti Numbers and De Rham’s Theorem
14 Harmonic Forms
III Lie Groups, Bundles, and Chern Forms
15. Lie groups
15.1 Lie Groups, Invariant Vector Fields and Forms
15.1a Lie Groups
2
Topological GL(n, R) is an open subset of Rn and as such is a n2 -dim. manifold.
(cf. pp. 392)

Examples
4. G = Sl(n, R). Sl(n, R) subgroup of Gl(n, R), detx = 1. Prob. 1.1(3). Submanifold of dimSl(n, R) = n2 ≠ 1
5. G = O(n), Sec. 1.1. Submanifold of dim n(n≠1) 2 .
6. G = U (n). Sec. 1. submanifold of complex n2 space or real 2n2 space.
8. G = T n abelian group of diagonal matrices of form z = diag[ei◊1 . . . ei◊n ] (15.2)
This group is topologically S 1 ◊ · · · ◊ S 1 , n-torus. Since circle connected, T n connected. From this, U (n)
also connected!

15.1b. Invariant Vector Fields and Forms

15.2 One-parameter subgroups

15.2(1) Generator of rotations
Ë2k J 2k Ë2k+1 J 2k+1 Ë2k (J 2 )k Ë2k+1 J(J 2 )k
Œ
ÿ Œ
ÿ Œ
ÿ Œ
ÿ Œ
ÿ
Ëk J k
eËJ = = + = +
k! (2k)! (2k + 1)! (2k)! (2k + 1)!
k=0 k=0 k=0 k=0 k=0

Ë2k (≠1)k Ë2k+1 (≠1)k

ÿŒ Œ
ÿ
=I +J = I cos (Ë) + J sin (Ë)
(2k)! (2k + 1)!
k=0 k=0

15.2(2) Generator of A(1)

3 4
a b
X=
0 0
433 4 3 4 3 4
a b
2 a b aa ab a b
X = = =a = aX
0 0 0 0 0 0 0 0
I
I n=0
∆ Xn =
an≠1 X n > 0

1 k k 1 k k≠1 1 ÿ 1 k k
ÿŒ ÿŒ Œ
∆ etX = t X =I+ t a X=I+ X t a
k! k! a k!
k=0 k=1 k=1
AΠB 3 4 3 4 3 4
1 ÿ 1 t0 a0 1 0 1 ab 1 b
=I+ X k k
t a ≠ = + eta ≠ a
a k! 0! 0 1 0 0 0 0
k=0
3 ta b ta b 4
=
e ae ≠ a
0 0
15.3(1) Maurer-Cartan equations

d‡ U (XR , XS ) = (
XR (U((( (X(
(U(
XS((‡
(
((‡ (X S )) ≠ ( R )) ≠ ‡ ([XR , XS ])
U

= ≠‡ U (CRS
T
XT ) = ≠CRS
U
(25)

1 U (25) 1 U R
∆ d‡ U = d‡ (XR , XS )‡ R · ‡ S = ≠ CRS ‡ · ‡S
2 2

∆ 0 = d(d‡ U )(XL , XM , XS )
(4.27)
= XL (d‡ U (XM , XS )) ≠ XM (d‡ U (XL , XS )) + XS (d‡ U (XL , XM ))
≠ d‡ U ([XL , XM ], XS ) + d‡ U ([XL , XS ], XM ) ≠ d‡ U ([XM , XS ], XL )
U⇠ ⇠ ⇠⇠ U⇠⇠
XL
=⇠ ⇠⇠
⇠(≠C MS) ≠ ⇠ XM
⇠(≠C⇠⇠LSU
) +⇠ ⇠⇠
XS⇠(≠C LM )

≠ d‡ U (CLM
R
XR , XS ) + d‡ U (CLS
R
XR , XM ) ≠ d‡ U (CM R
S XR , XL )

= CRS
U R
CLM + CRM
U R
CSL + CRL
U R
CM S

The Lie Algebra of a Lie Group

15.3a. The Lie Algebra
15.3b. The Exponential Map
Theorem 13 (15.27) map exp : g æ G sending A ‘æ eA diffeomorphism of some neighborhood of 0 œ g onto
neighborhood of e œ G.

Pf.
vector X œ g
- 3 4-
d - d 1 -
expú (X) = (exp tX)-- = 1 + tX + t2 X 2 + . . . -- =X
dt t=0 dt 2 t=0
exp ú : g æ g is the identity, exp local diffeomorphism by inverse function thm. (The Jacobian is nonsingular).

If G not a matrix group,

given X at e, etX = exp (tX) curve through e whose tangent vector at t = 0 is vector X (recall etX is the
integral curve through e of left invariant vector field X).
Thus expú (X) = X

16. Vector Bundles in Geometry and Physics

16.3(1) Connection on a tensor product space

ÒÕÕX = ÒÕÕX (ea ¢ eÕR ⁄aR ) = ÒX ea ¢ eÕR ⁄aR + ÒX (ea ¢ eÕR ⁄aR )
= X i eb Êia
b
¢ eÕ ⁄aR + ea ¢ eÕS ÊiR ⁄ X +ea ¢ eÕR d⁄
S aR i aR i
¸ ˚˙ X˝
¸ ˚˙ R ˝ ¸ ˚˙ ˝
b¡a S¡R =X i ˆi ⁄aR

=X i
ea ¢ eÕR (ˆi ⁄aR + a bR
Êib ⁄ + ÊiS ⁄ )
R aS

17. Fiber Bundles, Gauss-Bonnet, and Topological Quantization

A vector bundle is a family of vector spaces parameterized by points in the base space. How do we
parameterize a family of manifolds, say Lie groups?

17.1. Fiber Bundles and Principal Bundles

17.1a. Fiber Bundles
17.1b. Principal Bundles and Frame Bundles
frame e at p chosen
 f (p) = e– (p)g– (p) (17.4) (26)

– : U– ◊ G æ fi ≠1 (U– )
(27)
– (p, g) = e– (p)g = (e– )i g ij = fj
in an overlap, the same frame (17.4) will have another representation

f (p) = e— (p)g— (p) (17.5) (28)

e— (p) = e– (p)·–— (p)

·–— (p) © ·–—
g– (p) = ·–— (p)g— (p)
diffeomorphism
·–— (p) : G æ G
left translation of G by (transition) matrix ·–— (p)

17.1c. Action of the Structure Group on a Principal Bundle

Let f = (f1 . . . fn ) frame at p, f œ P

Theorem 14 (17.8)
(f œ P, g œ G) æ (f g) œ P
freely when g ”= e and
fi(f g) = fi(f )
i.e. preserves fibers

Proof: fi(f ) = p

– : U– ◊ G æ fi ≠1 (U– ) local trivialization

– (p, g– ) = f =∆ – (f ) = (p, g– ) ÷ ! g– for f
≠1

Let g œ G,
right action of g on fi ≠1 (U– ) is (locally action)

– (p, g– g) = fg
u
if p œ U– U—

fg = — (p, g— g) = — (p, ·—– (p)g– g) = – (p, g– g)

·—– = ≠1
— –

We see in this proof that the essential point is that left translations in G (say by ·—– ) commute with right
translations (say by g).

17.2.
17.3. Chern’s Proof of the Gauss-Bonnet-Poincaré Theorem
17.3a. A Connection in the Frame Bundle of a Surface
3 4
dx
Ê œ g = u(1) (17.14) (29)
dt

18. Connections and Associated Bundles

18.1. Forms with Values in a Lie Algebra
What do we mean by g ≠1 dg?
18.1.a. The Maurer-Cartan Form
If we think of Ê as being a form that takes its values in the fixed vector space g, rather than as a matrix of
1-forms, we shall have an equivalent picture that is in many ways more closely related to the terminology used
in physics.
exterior form is differential form
Maurer-Cartan 1-form on G
Let {ER } basis for g
{XR } left invariant fields on G obtained by left translating E’s
{‡ R } left invariant 1-forms on G forming, ’ g œ G, basis dual to {XR }

‡ R (XS ) = ” RS
Then
© ER ¢ ‡ R (18.1) (30)
(Yg ) = ER ‡ (Yg ) = ER Y
R R

Y = XR Y at g œ G, left translates back to 1

R

: Tg G æ Te G
cf. Nakahara
: Y ‘æ (Lg≠1 )ú Y = (Lg )≠1
ú Y, Y œ Tg G
Classically, Cartan wrote ’ p œ M , vector valued 1 form taking each Y vector at p into itself
dp = ˆi ¢ dxi = ˆi ¢ ” ij dxj
= g ≠1 dg (18.2) (31)
dg takes Y at g into Y , g ≠1
left transltates Y back to e

19. The Dirac Equation

Spin is what makes the world go ’round. -Theodore Frankel

19.1a. The Rotation Group SO(3) of R3

S T

E1 = U ≠1V
1
S T
1
E2 = U V
≠1
S T
≠1
E3 = U1 V

(Ei )’s are basis for so(3)

[Ei , Ej ] = ‘ijk Ek
ckij = ‘ijk
Consider 1-parameter group of rotations with angular velocity Ê, Ê = d◊
dt
-
dr --
= Ê ◊ r(0)
dt -t=0
On the other hand, 1-parameter subgroup is of form R(t) = etS , S skew-symmetric matrix
r(t) = R(t)r(0) = etS r(0)
-
dr --
= Sr(0) so S(r) = Ê ◊ r
dt -t=0
Ej (r) = ej ◊ r
R(t) = exp (Ej Ê j t) © exp (E · Êt) (19.3) (32)

R(◊) = exp (◊E · n) (19.4) (33)

19.1b. SU (2): The Lie Algebra su(2)
su(2) = g = {X|X = ≠X † ; tr(X) = 0}
ig = {X|X = X † ; tr(X) = 0},
basis 5 6
1
‡1 =
1
5 6
≠i
‡2 = (19.5)
i
5 6
1
‡3 =
≠1

[‡j , ‡k ] = 2i‘ijk ‡i (19.6)

We shall see that SU (2) simply connected.
Lie group theory states that ÷ homomorphism from SU (2) onto SO(3) Pf. : Frobenius thm.
Ad : SU (2) æ SO(3)
Claim: adjoint representation Ad(g) = gY g ≠1 of SU (2) on 3-dim. Lie algebra su(2) yields (Thm. 19.2)
standard representation of SO(3) on R3
5 6
z x ≠ iy
x = x · ‡ = xR ‡R = = xú
x + iy ≠z
inverse
1
x= tr(xú ‡1 )
2
1
y = tr(xú ‡2 ) (19.8)
2
1
z = tr(xú ‡3 )
2
S T
1
e1 = U0V ‘æ ‡1
0
S T
0
e2 = U1V ‘æ ‡2
0
S T
0
e3 = U0V ‘æ ‡3
1
ei · ‡ = ‡ i
real scalar product in ig
Èh, hÕ Í = tr(hhÕ )
as tr(‡j ‡k ) = 2”jk
Recall, ’ Lie group G acts on g by adjoint action

Ad : G æ Gl(g)
Ad(g)(X) = gXg ≠1 ’X œ g
In this case, SU (2) = {u|u† u = 1}, su(2) = {X|X † = ≠X, trX}, ‡i , i = 1, 2, 3 basis for su(2)

Ad : G æ Gl(g)
Ad(g)(X) = gXg ≠1 ’X œ g
Ad : SU (2) æ su(2)
Ad(u)(X) = uXu≠1 ’ X œ su(2)
Consider action of SU (2) on isu(2) = ig hermitian traceless matrices X. We’ll still call this Ad
’ u œ SU (2)
Ad(u) : ig æ ig
Ad(u) : isu(2) æ isu(2) xú ‘æ uxú u≠1 ’ xú œ isu(2)
 ’ u œ SU (2), we’re associated a 3 ◊ 3 matrix

R3 æ ig
Ad(u) : R3 æ R3 using (19.7)
x ‘æ x · ‡ = xR ‡R = xú

Note Ad is a representation of SU (2) by 3 ◊ 3 matrices

Ad(uuÕ )(xú ) = uuÕ xú (uuÕ )≠1 = Ad(u)Ad(uÕ )xú

Note
ÈAd(u)xú , Ad(u)xú Í = tr(uxú u≠1 uxú u≠1 )tr(xú xú ) = Èxú xú Í
so Ad(u) œ O(3), Ad representation of SU (2) by orthogonal 3 ◊ 3 matrices.

19.1c. SU (2) is Topologically the 3-Sphere

fundamental
5 6representation of SU (2) by 2 ◊ 2 complex unitary matrices uu = 1
†

u11 u12
u21 u22
Recall that the general form of SU (2) matrices is the following: (cf. wikipedia)
3 4
– ≠—
SU (2) = { |–, — œ C, |–|2 + |—|2 = 1}
— –

S 3 µ C2 ¥ R4

S 3 = {(z1 , z2 )T ||z1 |2 + |z2 |2 = 1}

Note SU (2) : S 3 æ S 3 as U U † = 1 ’ U œ SU (2) i.e. (unitary)
Note SU (2) acts transitively on S 3

Pf: 5 6
z ≠z 2
(1, 0)T œ S 3 , u= 1 œ SU (2)
z2 z 1
3 4 3 4 3 4
1 z z1
u = 1 ’ œ S 3 i.e. (arbitrary)
0 z2 z2
3 4 3 4
z1 3 1
So any œ S can be “reached” from œ S 3 by some u œ SU (2)
z2 0
From (17.10), topologically
SU (2)
S3 ¥
H
3 4
1
where H is stability subgroup of pt.
0
But (19.11), H = {1}

=∆ SU (2) ¥ S 3
In fact,

SU (2) æ S 3
3 4 3 4
z1 ≠z 2 z
u= ‘æ 1
z2 z 1 z2
In particular SU (2) = S 3 connected.
Since Ad(u) œ O(3) orthogonal matrix, detAd(u) = ±1
since u cont., and connected S 3 , detAd(u) = +1. Thus Ad(u) œ SO(3)

Ad : SU (2) æ SO(3)
Ad : SU (2) æ SO(3) in More Detail
Theorem 15 (19.12) representation Ad : SU (2) æ SO(3) given in (19.10)

u œ SU (2)
xú œ isu(2) (19.10) (34)
≠1
xú ‘æ uxú u

is onto, i.e. ’ rotation in R3 , of form (19.10)

Furthermore, this representation is 2 : 1, i.e. ’ rotation R, ÷ exactly 2 ±u œ SU (2), s.t. Ad(±u) = R

EY : 20150217 What we have is this:

R3 =
ú
su(2) 1
ú (X) =
x≠1 (tr(X‡1 ), tr(X‡2 ), tr(X‡3 ))T
(x, y, z) ‘æ x ‡R
ú R 2
Ad
SU (2) ≠≠æ Gl(su(2))
u ‘æ Ad(u) µ SO(3)
Ad(u)
isu(2) ≠≠≠≠æ isu(2)
xú ‘æ uxú u≠1
Pf: Let u(t) 1-parameter
! " subgroup of SU (2)
u(t) = exp ti h , h 2 ◊ 2 hermitian matrix (i.e. h = h† ), u(t) œ SU (2)

3
u(t) æ 1-parameter subgroup of SO(3) under isu(2) æ R
≠1
‘ú x
xú æ
Adu(t)x ≥ Adu(t)xú = e≠ith xú eith

- -
d -- d -- ≠ith
Ad(u(t))x = e xú eith = ≠i[h, xú ] = ≠i[hj ‡j , xk ‡k ] = ≠ihj xk [‡j , ‡k ] = ≠ihj xk ‘jki ‡ i (2i) =
dt -0 dt -0
ú

= 2‘jki hj xk ‡ i = 2(h ◊ x)i ‡i

EY : 20150217 Keep in mind

Ad(u)
R3 R3
ƒ

Ad(u)
isu(2) isu(2)

angular velocity vector of 1-parameter group Adu(t)x œ R3

Ê = 2h

From (19.3)
R(t) = exp (Ej Ê j t) © exp (E · Êt) (19.3) (35)
1‡ 2
Ad exp · ht xú ≥ R(t)x = exp (E · 2ht)x (19.13) (36)
i
or

Ad exp ( ‡i ht)
isu(2) ≠≠≠≠≠≠≠≠æ isu(2)

isu(2) æ R3
1‡ 2
Ad exp ht xú ‘æ R(t)x = exp (E · 2ht)x
i
1‡ 2
Adú = E– (19.14) (37)
–
2i
 e.g. h œ isu(2) Q R
0
h = ‡3 , h = a0b
1
t=◊
u(t) œ SU (2) 3 4 3 4 5 ≠i◊ 6 3 4
t t e ◊
u(t) = exp h = exp ‡3 = = exp ‡3
i i ei◊ i
5 6
1
with ‡3 =
≠1
exp (E · 2ht) œ SO(3)
S T S T
Adú ( ‡2i
–
) ≠2◊ cos (2◊) ≠ sin (2◊)
‘æ exp (E · 2h◊) = exp (2◊E3 ) = exp U2◊ V = U sin (2◊) cos (2◊) V
1
S T
≠1
with ·h = E · ‡3 = E3 = U 1 V

For SU (2), for 0 Æ ◊2fi, 5 ≠i◊ 6

! " e
exp ◊‡i 3 is a simple closed curve,
ei◊
exp (2◊E3 ) yields 2 full rotations

’ rotation of R3 is a rotation about some size, i.e. R = exp (E · Ê◊) œ SO(3)

By (19.13), 1‡ 2
Ad exp · ht xú ‘æ R(t)x = exp (E · 2ht) = exp (E · Ê◊)
i
So that 1‡ 2
Ad exp · Ê◊ = R
2i
for
Ê = 2h
‡
E=
2i
So Ad onto. Ad : SU (2) æ SO(3) ! "
If Ad(u) = R, u = u(t) = exp ti h ‘æ R = exp (E · Êt) Ad(u)xú = uxú u≠1 ‘æ Rx
Ad(≠u)xú = uxú u≠1 ‘æ Rx
So Ad representation is at least 2 : 1 i.e. not faithful

It’s an elementary result of group theory that

if „ : G æ GÕ homomorphism of G onto GÕ , then GÕ isomorphic to coset G/H, where H = „≠1 (eÕ ) is kernel

(17.10) fundamental principle, ’ G that acts on GÕ by

(g, g Õ ) ‘æ „(g)g Õ

and stability subgroup of eÕ œ GÕ is kernel H = „≠1 (eÕ ) ker„ = H = „≠1 (eÕ )

Ad : SU (2) æ SO(3) kerAd = {±1}

(17.11) æ SU (2) is fiber bundle over SO(3);

p≠1 : SO(3) æ SU (2)

p≠1 (R) ‘æ {±u} exactly 2 pts.

ƒ
S3 SU (2)
p p
ƒ
RP 3 SO(3)
 S 3 /p = RP 3
x œ S3
[x] = {x, ≠x}
x ≥ ≠x

20. Yang-Mills Fields

20.1. Noether’s Theorem for Internal Symmetries
How do symmetries yield conservation laws?
„ N -tuple „a (t, x) = „a (x), local representation of a section of some vector bundle E,
E
fi
M
In the case of a Dirac electorn, we have seen that E is the bundle of complex 4-component Dirac spinors
over a perhaps curved spacetime. If E is not a trivial bundle (or if we insist on using curvilinear coordinates)
we shall have to deal with the fact that ˆj „a do not form a tensor.

20.1a. The Tensorial Nature of Lagrange’s Equations

Let M n+1 (pseudo-) Riemannian manifold, let E vector bundle over M ; for definiteness, let fiber be RN .
section of this bundle over U µ M is described by N real-valued functions {„aU },
where „V = cV U „U and
cV U (x) is N ◊ N transition matrix function, caV U b .

a
{ }
notation a
{ –}
a
– = ·–— —

Lagrangian L0 (x, „, „x ) © L0 (x, , ˆj a

)

20.2. Weyl’s Gauge Invariance Revisited

20.2a. The Dirac Lagrangian
20.2b. Weyl’s Gauge Invariance Revisited
20.2c. The Electromagnetic Lagrangian
Instead of considering a change of (spacetime) coordinates x, we shall look at a change of the field (fiber)
coordinate Â, i.e. a gauge transformation.
Since the phase of  is not measurable, we should be able to have invariance under a local gauge transfor-
mation, where – = –(x) varies with the spacetime point x!
Clearly the Dirac equation and Lagrangian are not invariant under such a substitution because of the appear-
ance of terms involving d–.
It must be that there is some background field that is interacting with the electron. This background field will
manifest itself through the appearance of the connection.

The Yang-Mills Nucleon

How did the groups SU (2) and SU (3) appear in particle physics?

20.3a. The Heisenberg Nucleon

20.3b. The Yang-Mills Nucleon
20.3c. A Remark on Terminology
We have related the connection matrices Ê to the gauge potentials A by

Ê = ≠iqA
q is called a generalized charge.

21.
A. Elasticity
A.a. The Classical Cauchy Stress Tensor and Equations of Motion
B(t) compact body, might be portion of larger body in motion
mass 3 form
m3 © fl vol
mass conservation ⁄ ⁄
d 3
m = Lv+ ˆ m3 = 0
dt B(t) B(t)
ˆt

b external force density (per unit mass)

⁄ ⁄ ⁄
d
v i m3 = b i m3 + tij nj da
dt B(t) B(t) ˆB(t)

A.b. Stresses in Terms of Exterior Forms

t pseudo (n ≠ 1) form on M n with values in tangent bundle T M (vector bundle language)

t = er ¢ tr © er ¢ trJ ‡ J (38)
⁄ ⁄ ⁄ ⁄
t= e r ¢ tr = e r tr = er trJ dxJ
ˆB ˆB ˆB ˆB
as total traction that part of body outside ˆB exerts on B
(2.73) ’ Riemannian M , write (n ≠ 1) form tr in terms of vector t(r)
Ô Ô
tr = i(tr )vol = i(t(r) ) g‘I dxI = gt(r)i ‘iJ dxJ

Ô
trJ = gt(r)i ‘iJ (A.6) (39)
1
tri © t(r)i = Ô trJ ‘iJ
g
relation between stress form tr and Cauchy’s stress tensor tri
assuming t = er ¢ tr is (n ≠ 1) form section of the tangent bundle, thus from (9.31), we have
20141102 EY recall
Ò– = e ¢ (d– + Ê · –) (9.31)
and
Ò– = Ò(ei ¢ –i ) = (Òei ) ¢· –i + ei ¢ d–i
where
(Òei ) ¢· – = (ek ¢ Ê ki ¢· –i © ek ¢ (Ê ki · –i )
i

Òt = Ò(er ¢ tr ) = Òer ¢· tr + er ¢ dtr = er ¢ (dtr + Ê rs · ts ) = er ¢ Òtr (A.7) (40)

A.f. Hamilton’s Principle in Elasticity

⁄
”U = ”ERC dX R · S C (A.24) (41)
B
⁄
”U = S CR ”ERC VOLn (A.25) (42)
B
Hamilton’s principle ⁄ ⁄ ⁄
” T dt ≠ ”U dt + ”W dt = 0 (A.26) (43)
s
We don’t write ” U dt because we don’t assume
s
We don’t assume that ÷ stored energy function U so we don’t write ” U dt, ÷ only differential ”U