Differential geometry and Lie groups A

computational perspective Gallier J. pdf

download

https://textbookfull.com/product/differential-geometry-and-lie-
groups-a-computational-perspective-gallier-j/

Download more ebook instantly today - get yours now at textbookfull.com

We believe these products will be a great fit for you. Click
the link to download now, or visit textbookfull.com
to discover even more!

Basics of algebra topology and differential calculus

Gallier J

https://textbookfull.com/product/basics-of-algebra-topology-and-
differential-calculus-gallier-j/

Theory of Groups and Symmetries Finite groups Lie

groups and Lie Algebras 1st Edition Alexey P. Isaev

https://textbookfull.com/product/theory-of-groups-and-symmetries-
finite-groups-lie-groups-and-lie-algebras-1st-edition-alexey-p-
isaev/

Math 845 Notes on Lie Groups Mark Reeder

https://textbookfull.com/product/math-845-notes-on-lie-groups-
mark-reeder/

From Differential Geometry to Non commutative Geometry

and Topology Neculai S. Teleman

https://textbookfull.com/product/from-differential-geometry-to-
non-commutative-geometry-and-topology-neculai-s-teleman/
Lie Groups 1st Edition Luiz Antonio Barrera San Martin

https://textbookfull.com/product/lie-groups-1st-edition-luiz-
antonio-barrera-san-martin/

Fundamentals of linear algebra and optimization Gallier

J.

https://textbookfull.com/product/fundamentals-of-linear-algebra-
and-optimization-gallier-j/

Theory of Groups and Symmetries: Representations of

Groups and Lie Algebras, Applications 1st Edition
Alexey P. Isaev

https://textbookfull.com/product/theory-of-groups-and-symmetries-
representations-of-groups-and-lie-algebras-applications-1st-
edition-alexey-p-isaev/

Elliptic Quantum Groups Representations and Related

Geometry Hitoshi Konno

https://textbookfull.com/product/elliptic-quantum-groups-
representations-and-related-geometry-hitoshi-konno/

Matrix Inequalities and Their Extensions to Lie Groups

First Edition Liu

https://textbookfull.com/product/matrix-inequalities-and-their-
extensions-to-lie-groups-first-edition-liu/
 Differential Geometry and Lie Groups
A Computational Perspective

Jean Gallier and Jocelyn Quaintance

Department of Computer and Information Science
University of Pennsylvania
Philadelphia, PA 19104, USA
e-mail: [email protected]

c Jean Gallier
Please, do not reproduce without permission of the authors

August 7, 2019
2

To my daughter Mia, my wife Anne,

my son Philippe, and my daughter Sylvie.

To my parents Howard and Jane.

Preface

This book is written for a wide audience ranging from upper undergraduate to advanced
graduate students in mathematics, physics, and more broadly engineering students, especially
in computer science. It covers manifolds, Riemannian geometry, and Lie groups, some central
topics of mathematics. However, computer vision, robotics, and machine learning, to list
just a few “hot” applied areas, are increasingly consumers of differential geometry tools, so
this book is also written for professionals who wish to learn about the concepts and tools
from differential geometry used to solve some of their problems.
Although there are many books covering differential geometry and Lie groups, most of
them assume that the reader is already quite familar with manifold theory, which is a severe
obstacle for a reader who does not possess such a background. In this book, we only assume
some modest background in calculus and linear algebra from the reader, and basically develop
manifold theory from scratch. Additional review chapters covering some basics of analysis,
in particular the notion of derivative of a map between two normed vector spaces, and some
basics of topology, are provided for the reader who needs to firm up her/his background in
these areas. This book is split into two parts.

1. The basic theory of manifolds and Lie groups.

2. Some of the fundamental topics of Riemannian geometry.

The culmination of the concepts and results presented in this book is the theory of nat-
urally reductive homogeneous manifolds and symmetric spaces. It is remarkable that most
familiar spaces are naturally reductive manifolds. Remarkably, they all arise from some suit-
able action of the rotation group SO(n), a Lie group, which emerges as the master player.
The machinery of naturally reductive manifolds, and of symmetric spaces (which are even
nicer!), makes it possible to compute explicitly in terms of matrices all the notions from
differential geometry (Riemannian metrics, geodesics, etc.) that are needed to generalize
optimization methods to Riemannian manifolds. Such methods are presented in Absil, Ma-
hony and Sepulchre [2], and there is even a software package (MANOPT) that implements
some of these procedures.
The interplay between Lie groups, manifolds, and analysis, yields a particularly effective
tool. We tried to explain in some detail how these theories all come together to yield such
a beautiful and useful tool.

3
4

We have also included chapters that present material having significant practical appli-
cations. These include

1. Chapter 8, on constructing manifolds from gluing data, which has applications to

surface reconstruction from 3D meshes.

2. Chapter 21, on the “Log-Euclidean framework,” has applications in medical imaging.

3. Chapter 22, on homogeneous reductive spaces and symmetric spaces, has applications
to robotics, machine learning, and computer vision. For example, Stiefel and Grass-
mannian manifolds come up naturally. Furthermore, in these manifolds, it is possible
to compute explicitly geodesics, Riemannian distances, gradients and Hessians. This
makes it possible to actually extend optimization methods such as gradient descent
and Newton’s method to these manifolds. A very good source on these topics is Absil,
Mahony and Sepulchre [2].

Let us now give motivations for learning the concepts and tools discussed in this book.
The need to generalize concepts and tools used in “flat spaces” such as the real line, the
plane, or more generally Rn , to more general spaces (such as a sphere) arises naturally. Such
concepts and tools include

1. Defining functions.

2. Computing derivatives of functions.

3. Finding minima or maxima of functions.

4. More generally, solving optimization problems.

5. Computing the length of curves.

6. Finding shortest paths between two points.

7. Solving differential equations

8. Defining a notion of average or mean.

9. Computing areas and volumes.

10. Integrating functions.

A way to deal with a space M more complicated than Rn is to cover it with small pieces
Uα , such that each piece Uα “looks” like Rn , which means that there is a bijection ϕα from
Uα to a subset of Rn . Typically, M is a topological space, so the maps ϕα : Uα → Rn , called
charts, are homeomorphisms of Uα onto some open subset of Rn . From an intuitive point of
view, locally, M looks like a piece of Rn .
 5

The mathematical notion formalizing the above idea is the notion of manifold . Having a
“good” notion of what a space M is, the issue of defining real-valued functions f : M → R,
and more generally functions f : M → N between two manifolds M and N , arises. Then it
is natural to wonder what is a function with a certain degree of differentiability, and what
is the derivative of a function between manifolds.

To answer these questions, one needs to add some structure to the charts ϕα : Uα → Rn ,
namely, whenever two charts ϕα : Uα → Rn and ϕβ : Uβ → Rn overlap, which means that
Uα ∩Uβ 6= ∅, then the map ϕβ ◦ϕ−1
α should behave well; technically, this means that it should
k
be C (continuously differentiable up to order k), or smooth.

Another important idea coming from the notion of derivative of a function from Rn to
m
R , is the idea of linear approximation of a function f : M → N between two manifolds. To
accomplish this, we need to define the notion of tangent space Tp M to the manifold M at a
point p ∈ M . Similarly, we have a tangent space Tf (p) N to the manifold N at the point f (p)
(the image of p under f ), and the derivative of f at p is a linear map dfp : Tp M → Tf (p) N
from the tangent space Tp M (with p ∈ M ) to the tangent space Tf (p) N (with f (p) ∈ N ).

Setting up carefully and rigorously the machinery to define manifolds, maps between
them, tangent spaces, and the derivative of a function between manifolds, will occupy the
first third of this volume.

If the manifold M is already naturally a subset of RN for some N large enough, then
matters are simpler, and it is easier to define manifolds, tangent spaces, and derivatives of
functions between manifolds. For pedagogical reasons, we begin with this simpler case in
Chapters 1–3.

If the manifold M is not embedded in RN for some N , which typically occurs when
M is obtained as a quotient space, such as real projective space RPn (the space of lines
through the origin in Rn+1 ) or the Grassmannian G(k, n) (the space of k-dimensional linear
subspaces of Rn ), then matters are technically more complicated. One needs to introduce
charts and atlases, and the definitions of the tangent space and of the derivative of a map
between manifolds are more technical. One needs to define tangent vectors in terms of various
equivalence relations (on curves, on certain triples, on germs of locally defined functions). We
do this very carefully, even in the case of a C k manifold where 1 ≤ k < ∞ (that is, a manifold
which is not necessarily smooth). We give three equivalent definitions of the tangent space
Tp M to M at p, and prove their equivalence. The first definition involves equivalence classes
of curves through p. The third definition in terms of point derivations applies even to C k -
manifolds, at the price of introducing stationary germs. In the smooth case, this definition
is equivalent to the standard definition found in Tu [112] and Warner [114]. Following J.P.
Serre, the equivalence of the first and of the third definition is elegantly proved by setting
up a bilinear pairing and showing that this pairing is nondegenerate. Chapters 7 and 9 are
devoted to the definitions of tangent spaces, tangent bundles, vector fields, and the related
concepts such as Lie derivatives and Lie brackets, in the framework of general manifolds.
6

Chapter 8 presents a more constructive approach for constructing manifolds using what
we call gluing data. This chapter has applications to surface reconstruction from 3D meshes.
A very important class of manifolds is the class of groups that are also manifolds and
topological groups (which means that multiplication and the inverse operation are smooth).
Such groups are called Lie groups. The prime example is the group SO(3) of rotations in R3 ,
and more generally SO(n). Remarkably, a large subclass of Lie groups turns out to be the
family of closed subgroups of GL(n, R), the group of invertible n × n real matrices. This is a
famous result due to Von Neumann and Cartan, see Theorem 3.8. Such closed subgroups of
GL(n, R) are called linear Lie groups or matrix Lie groups. If G is a linear Lie group, then
its tangent space TI G at the identity, denoted g, has some additional structure besides being
a vector space. It has a noncommutative and nonassociative skew-symmetric multiplication
[X, Y ] (with X, Y ∈ g) called the Lie bracket, which satisfies a strange kind of associativity
axiom called the Jacobi identity. The vector space g with the Lie bracket as multiplication
operation has the algebraic structure of what is called a Lie algebra. In some sense, g is a
linearization of G near I, and the Lie bracket is a measure of the noncommutativity of the
group operation. Remarkably, there is a way of “recovering” G from its Lie algebra g by
making use of the (matrix) exponential exp : g → G. This map is not injective nor surjective
in general. In many cases of interest, such as SO(n) and SE(n), it is surjective. Also, “near”
I, the exponential is bijective. Since we can move from the tangent space TI G = g at I to
to the tangent space Tg G at any other element g ∈ G by left (or right) multiplication, we
obtain a way of parametrizing G using the exponential map.
As a warm-up for the discussion of linear Lie groups and Lie algebras in Chapter 3,
we present some properties of the exponential map of matrices in Chapters 1 and 2. In
particular, we give a formula for the derivative of exp. A discussion of general Lie groups
(not necessarily groups of matrices) is postponed until Chapter 18.
Another important theme of this book is the notion of group action. A manifold, such as
the sphere S n (in Rn+1 ), or projective space RPn , or the Grassmannian G(k, n), may not be
a group, but may have a lot of symmetries given by a group G. For example, the sphere S 2
in R3 has the group of rotations SO(3) as group of symmetries, in the sense that a rotation
in SO(3) moves any point on the sphere to another point on the sphere, so the sphere is
invariant under rotations.
The notion of symmetry of a space under the transformations of a group G is neatly
captured by the notion of action of a group on a set (or a manifold). A (left) action of a
group G on a set X is a binary operation · : G × X → X satisfying the axioms
g1 · (g2 · x) = (g1 g2 ) · x for all g1 , g2 ∈ G and all x ∈ X
1·x=x for all x ∈ X.
Here, g1 g2 denotes the product of the two elements g1 and g2 using the group multiplication
operation on G, and 1 denotes the identity element of G. Intuitively, we can think of g · x,
where g is an element of the group G and x is an element of the set X, as the result of
moving x using the “transformation” g.
 7

A group action is transitive if for any two elements x, y ∈ X, there is some group element
g ∈ G that moves x to y, that is, y = g · x. Many actions that arise in practice are
transitive. For example, the group SO(3) acts transitively on S 2 , and more generally SO(n)
acts transitively on S n−1 . The reason why transitivity is important is that if we consider
any fixed element x ∈ X, we can look at the stabilizer Gx of x, which is the set of elements
of X left fixed by the action of G, namely

Gx = {g ∈ G | g · x = x}.

It can be shown that Gx is a subgroup of G (not necessarily normal), and there is a bijection
between the set G/Gx of left cosets of G and X.
This bijection is very crucial, because it allows us to view X as the set of cosets G/Gx ,
and if the group G is well understood, then this yields a way of inferring information about
X using information about G and Gx . So far, X is a just a set, and G is just a group without
any additional structure, but if X is also a topological space, and G is a topological group,
then we can ask whether the quotient space G/Gx is homeomorphic to X. In general, this
is not the case, but if G is a Lie group and if X is a manifold, then G/Gx is a manifold
diffeomorphic to X.
The above result is very significant because it allows us to study certain manifolds M
that possess a transitive action of a Lie group G in terms of the groups G and Gx . Such
spaces are called homogenous spaces, and it turns out that many familiar manifolds such as
S n , RPn , the Grassmannians G(k, n), the space of symmetric positive definite matrices, the
Lorentz manifolds, etc., are homogenous manifolds.
We begin our study of group actions and homogenous spaces in Chapter 4. We provide
many examples of spaces having a transitive action, and compute explicitly stabilizers for
these actions. The study of homogenous spaces is continued in greater depth, also dealing
with considerations of Riemannian geometry, in Chapter 22,
As a kind of interlude, in Chapter 5, we spend some time investigating the Lorentz groups
O(n, 1), SO(n, 1) and SO0 (n, 1) (and also the groups O(1, n), SO(1, n) and SO0 (1, n)).
When n = 3, these groups arise in the special theory of relativity. It turns out that O(3, 1)
also comes up in computer vision in the study of catadioptric cameras (see Geyer [50],
Chapter 5), and this was one of our original motivations for getting interested in homogeneous
spaces. In Chapter 6, we also investigate the topological structure of the groups O(p, q),
SO(p, q), and SO0 (p, q).
One feature of our exposition worth pointing out is that we give a complete proof of
the surjectivity of the exponential map exp : so(1, 3) → SO0 (1, 3), for the Lorentz group
SO0 (1, 3) (see Section 5.2, Theorem 5.18). Although we searched the literature quite thor-
oughly, we did not find a proof of this specific fact (the physics books we looked at, even the
most reputable ones, seem to take this fact as obvious, and there are also wrong proofs; see
the remark following Theorem 5.5).
8

We are aware of two proofs of the surjectivity of exp : so(1, n) → SO0 (1, n) in the general
case where where n is arbitrary: One due to Nishikawa [90] (1983), and an earlier one
due to Marcel Riesz [97] (1957). In both cases, the proof is quite involved (40 pages or
so). In the case of SO0 (1, 3), a much simpler argument can be made using the fact that
ϕ : SL(2, C) → SO0 (1, 3) is surjective and that its kernel is {I, −I} (see Proposition 5.17).
Actually, a proof of this fact is not easy to find in the literature either (and, beware there are
wrong proofs, again see the Remark following Theorem 5.5). We have made sure to provide
all the steps of the proof of the surjectivity of exp : so(1, 3) → SO0 (1, 3). For more on this
subject, see the discussion in Section 5.2, after Corollary 5.14.
What we have discussed above comprises the basic theory of manifolds, Lie groups, and
homogenous spaces. Chapter 10 gathers some technical tools needed later such as partitions
of unity and covering spaces. For the sake of the reader who feels rusty on some basics
of analysis and topology, we have included two refresher chapters: Chapter 11 on power
series and derivative of functions between normed vector spaces, and Chapter 12 on basics
of topology. These should be consulted as nedeed, but we strongly advise the reader who
has not been exposed to the notion of derivative as a linear map to review Chapter 11.
One of the main gaps in the theory of manifolds that we just sketched is that there is
no way to discuss metric notions such as the notion of length of a curve segment, or the
notion of angle between two curves. We are in a situation similar to the theory of vector
spaces before inner products are introduced. The remedy is to add an inner product to our
manifold M , but since the tangent spaces Tp M (with p ∈ M ) are unrelated, we actually
need to add a family (h−, −ip )p∈M of inner products, one for each tangent space Tp M . We
also need to require that these inner products vary smoothly as p moves in M . A family
of inner products as above is called a Riemannian metric, and a pair (M, h−, −i) where M
is a smooth manifold and (h−, −ip )p∈M is a Riemannian metric is a Riemannian manifold ,
after B. Riemann who was the first to have this idea. If a manifold is too big, then it
may not have a Riemannian metric, but “well-behaved” manifolds, namely second-countable
manifolds, always have a Riemannian metric (this is shown using a partition of unity).
Riemannian metrics are defined in Chapter 13. Having a Riemannian metric allows us to
define the gradient, the Hessian, and the Laplacian, of a function. For functions f : Rn → R,
this is automatic since Rn is equipped with the Euclidean inner product, but for a manifold
M , given a function f : M → R, to convert the linear form dfp into a vector (grad f )p ∈ Tp M
such that dfp (u) = h(grad f )p , ui for all u ∈ Tp M , an inner product is needed on Tp M , and
so a Riemannian metric on M is needed.
The notion of Riemannian metric allows us to discuss metric properties of a manifold,
but there is still a serious gap which has to do with the fact that given a manifold M , in
general, for any two points p, q ∈ M , there is no “natural” isomorphism between the tangent
spaces Tp M and Tq M . Given a curve c : [0, 1] → M on M , as c(t) moves on M , how does
the tangent space Tc(t) M change as c(t) moves?
If M = Rn , then the spaces Tc(t) Rn are canonically isomorphic to Rn , and any vector
v ∈ Tc(0) Rn ∼
= Rn is simply moved along c by parallel transport; that is, at c(t), the tangent
 9

vector v also belongs to Tc(t) Rn . However, if M is curved, for example a sphere, then it is not
obvious how to “parallel transport” a tangent vector at c(0) along a curve c. This problem
is related to the fact that it is not obvious how to define the derivative ∇X Y of a vector field
X with respect to another vector field Y . If X and Y are vector fields on a surface S in R3 ,
then for any point p ∈ S, the derivative (DX Y )p given by

Y (p + tX(p)) − Y (p)
DX Y (p) = lim .
t→0 t
(if it exists), is a vector in R3 , but there is no reason why it should belong to the tangent
space Tp S to S at p.
Gauss solved this problem by introducing the notion of covariant derivative, which con-
sists in keeping the projection (∇Y X) of (DX Y )p onto the tangent space Tp S, and to discard
the normal component.
However, if M is a general manifold not embedded in RN , then it is not clear how to
perform such a projection. Instead, the notion of covariant derivative is defined in terms of a
connection, which is a bilinear map ∇ : X(M ) × X(M ) → X(M ) defined on vector fields and
satisfying some properties that make it a generalization of the notion of covariant derivative
on a surface. The notion of connection is defined and studied in Chapter 14. Having the
notion of connection, we can define the notion of parallel vector field along a curve, and of
parallel transport, which allows us to relate two tangent spaces Tp M and Tq M .
The notion of covariant derivative is also well-defined for vector fields along a curve. This
is shown in Section 14.2. Given a vector field X along a curve γ, this covariant derivative is
denoted by DX/dt. We then have the crucial notion of a vector field parallel along a curve
γ, which means that DX/dt(s) = 0 for all s (in the domain of γ).
The notion of a connection on a manifold does not assume that the manifold is equipped
with a Riemannian metric. In Section 14.3, we consider connections having additional prop-
erties, such as being compatible with a Riemannian metric or being torsion-free. Then we
have a phenomenon called by some people the “miracle” of Riemannian geometry, namely
that for every Riemannian manifold, there is a unique connection which is torsion-free and
compatible with the metric. Furthermore, this connection is determined by an implicit for-
mula known as the Koszul formula. Such a connection is called the Levi-Civita connection.
If γ is a curve on a smooth Riemannian manifold M , and if X = γ 0 is the vector field of
tangent vectors γ 0 to γ, we can consider the curves γ that satisfy the equation
Dγ 0
= 0. (∗)
dt
0
Intuitively, we can view Dγ
dt
as the tangent component of the acceleration vector γ 00 of the
curve γ, and such curves have an acceleration normal to the manifold. Curves satisfying
equation (∗) are called geodesics. Geodesics are the Riemannian equivalent of straight lines
in Rn . The notion of geodesic is one of the most crucial tools in Riemannian geometry. One
10

of the reasons is that geodesics are locally distance minimizing, and that they provide a way
to parametrize a neighborhood U of any point p on a manifold M by a neighborhood of
the origin in the tangent space Tp M , using the exponential map (not to be confused with
the Lie group exponential) expp : Tp M → M . If the exponential map is surjective, then the
manifold M is said to be complete. A beautiful theorem of Hopf and Rinow states that if a
manifold is complete, then any two points can be joined by a minimal geodesic (a geodesic
of minimal length). This is an important property because the shortest distance between
any two points is achieved by a geodesic. Compact Riemannian manifolds are complete, so
many of the familiar compact manifolds (S n , RPn , G(k, n)) are complete.
R1
Given a curve ω on a Riemannian manifold, the quantity E(ω) = 0 kω 0 (t)k2 dt is called
the energy function. Geodesics between two points p and q turn out to be critical points of
the energy function E on the path space Ω(p, q) of all piecewise smooth curves from p to q.
To define the notion of critical point of the energy function, because the space Ω(p, q) is not
a finite-dimensional manifold, it is necessary to introduce the notion of variation of a curve
and to prove the first variation formula. Here, we make a link with the calculus of variation.
Geodesics are studied throroughly in Chapter 15.
Riemannian metrics, connections, and geodesics, are three of the pilars of differential
geometry. The fourth pilar is curvature.
For surfaces, the notion of curvature can be defined in terms of the curvatures of curves
drawn on the surface. The notion of Gaussian curvature (of course, introduced by Gauss)
gives a satisfactory answer. However, for manifolds of dimension greater than 2, it is not ob-
vious what curvature means. Riemann proposed a definition involving the notion of sectional
curvature, but his seminal paper (1868) did not contain proofs and did not give a general
method to compute such a curvature. It is only fifty years later that the idea emerged that
the curvature of a Riemannian manifold should be viewed as a measure R(X, Y )Z of the
extent to which the operator (X, Y ) 7→ ∇X ∇Y Z is symmetric.
The Riemann curvature operator R turns out to be C ∞ -linear in all of its three arguments,
but it is a rather complicated object. Fortunately, there is a simpler object, the sectional
curvature K(u, v). When ∇ is the Levi-Civita connection, the curvature operator R can
be recovered from the sectional curvature K. There is also an important simpler notion of
curvature Ric(x, y), called the Ricci curvature, which arises as the trace of the linear map
v 7→ R(x, v)y. An even cruder notion of curvature is the scalar curvature. These notions of
curvature are discussed in Chapter 16.
R1
We pointed out earlier that the energy function E(ω) = 0 kω 0 (t)k2 dt determines the
geodesics (between two fixed points p and q) in the sense that its critical points are the
geodesics. A deeper understanding of the energy function is achieved by investigating the
second derivative of E at critical points. To do this we need the notion of 2-parameter
variation and the second variation formula. The curvature operator shows up in this formula.
Another important technical tool is the notion of Jacobi fields, which are induced by geodesic
variations. Jacobi fields can be used to compute the sectional curvature of various manifolds.
 11

Another important theme of differential geometry is the influence of curvature (sectional

or Ricci) on the topology of a Riemannian manifold. This is a vast subject and we only
discuss three results, one of which being the Hadamard and Cartan theorem about complete
manifolds of non-positive curvature.
The goal of Chapter 17 is to understand the behavior of isometries and local isometries,
in particular their action on geodesics. We also intoduce Riemannian covering maps and
Riemannian submersions. If π : M → B is a submersion between two Riemannian manifolds,
then for every b ∈ B and every p ∈ π −1 (b), the tangent space Tp M to M at p splits into two
orthogonal components, its vertical component Vp = Ker dπp , and its horizontal component
Hp (the orthogonal complement of Vp ). If the map dπp is an isometry between Hp and Tb B,
then most of the differential geometry of B can be studied by lifting B to M , and then
projecting down to B again. We also introduce Killing vector fields, which play a technical
role in the study of reductive homogeneous spaces.
In Chapter 18, we return to Lie groups. Not every Lie group is a matrix group, so
in order to study general Lie groups it is necessary to introduce left-invariant (and right-
invariant) vector fields on Lie groups. It turns out that the space of left-invariant vector
fields is isomorphic to the tangent space g = TI G to G at the identity, which is a Lie algebra.
By considering integral curves of left-invariant vector fields, we define the generalization
of the exponential map exp : g → G to an arbitrary Lie group. The notion of immersed
Lie subgroup is introduced, and the correspondence between Lie groups and Lie algebra is
explored. We also consider the special classes of semidirect products of Lie algebras and Lie
groups, the universal covering of a Lie group, and the Lie algebra of Killing vector fields on
a Riemannian manifold.
Chapter 19 deals with two topics:
1. A formula for the derivative of the exponential map for a general Lie group (not
necessarily a matrix group).
2. A formula for the Taylor expansion of µ(X, Y ) = log(exp(X) exp(Y )) near the origin.
The second problem is solved by a formula known as the Campbell-Baker-Hausdorff formula.
An explicit formula was derived by Dynkin (1947), and we present this formula.
Chapter 20 is devoted to the study of metrics, connections, geodesics, and curvature, on
Lie groups. Since a Lie group G is a smooth manifold, we can endow G with a Rieman-
nian metric. Among all the Riemannian metrics on a Lie groups, those for which the left
translations (or the right translations) are isometries are of particular interest because they
take the group structure of G into account. As a consequence, it is possible to find explicit
formulae for the Levi-Civita connection and the various curvatures, especially in the case of
metrics which are both left and right-invariant.
In Section 20.2 we give four characterizations of bi-invariant metrics. The first one refines
the criterion of the existence of a left-invariant metric and states that every bi-invariant
metric on a Lie group G arises from some Ad-invariant inner product on the Lie algebra g.
12

In Section 20.3 we show that if G is a Lie group equipped with a left-invariant metric,
then it is possible to express the Levi-Civita connection and the sectional curvature in terms
of quantities defined over the Lie algebra of G, at least for left-invariant vector fields. When
the metric is bi-invariant, much nicer formulae are be obtained. In particular the geodesics
coincide with the one-parameter groups induced by left-invariant vector fields.
Section 20.5 introduces simple and semisimple Lie algebras. They play a major role in
the structure theory of Lie groups
Section 20.6 is devoted to the Killing form. It is an important concept, and we establish
some of its main properties. Remarkably, the Killing form yields a simple criterion due to
Élie Cartan for testing whether a Lie algebra is semisimple.
We conclude this chapter with a section on Cartan connections (Section 20.7). Un-
fortunately, if a Lie group G does not admit a bi-invariant metric, under the Levi-Civita
connection, geodesics are generally not given by the exponential map exp : g → G. If we
are willing to consider connections not induced by a metric, then it turns out that there
is a fairly natural connection for which the geodesics coincide with integral curves of left-
invariant vector fields. These connections are called Cartan connections. This chapter makes
extensive use of results from a beautiful paper of Milnor [84].
In Chapter 21 we present an application of Lie groups and Riemannian geometry. We
describe an approach due to Arsigny, Fillard, Pennec and Ayache, to define a Lie group
structure and a class of metrics on symmetric, positive-definite matrices (SPD matrices)
which yield a new notion of mean on SPD matrices generalizing the standard notion of
geometric mean.
SPD matrices are used in diffusion tensor magnetic resonance imaging (for short, DTI),
and they are also a basic tool in numerical analysis, for example, in the generation of meshes
to solve partial differential equations more efficiently. As a consequence, there is a growing
need to interpolate or to perform statistics on SPD matrices, such as computing the mean
of a finite number of SPD matrices.
Chapter 22 provides the culmination of the theory presented in the book, the concept of
a homogeneous naturally reductive space.
The goal is to study the differential geometry of a manifold M presented as the quotient
G/H of a Lie group G by a closed subgroup H. We would like to endow G/H with a
metric that arises from an inner product on the Lie algebra g of G. To do this, we consider
G-invariant metrics, which are metrics on G/H such that the left multiplication operations
τg : G/H → G/H given by
τg (h2 H) = gg2 H
are isometries. The existence of G-invariant metrics on G/H depends on properties of a
certain representation of H called the isotropy representation (see Proposition 22.21). The
isotropy representation is equivalent to another representation AdG/H : H → GL(g/h) of H
involving the quotient algebra g/h.
 13

This representation is too complicated to deal with, so we consider the more tractable
situation where the Lie algebra g of G factors as a direct sum
g = h ⊕ m,
for some subspace m of g such that Adh (m) ⊆ m for all h ∈ H, where h is the Lie algebra
of H. Then g/h is isomorphic to m, and the representation AdG/H : H → GL(g/h) becomes
the representation Ad : H → GL(m), where Adh is the restriction of Adh to m for every
h ∈ H. In this situation there is an isomorphism between To (G/H) and m (where o denotes
the point in G/H corresponding to the coset H). It is also the case that if H is “nice” (for
example, compact), then M = G/H will carry G-invariant metrics, and that under such
metrics, the projection π : G → G/H is a Riemannian submersion.
It is remarkable that a simple condition on m, namely Ad(H) invariance, yields a one-to-
one correspondence between G-invariant metrics on G/H and Ad(H)-invariant inner prod-
ucts on m (see Proposition 22.22). This is a generalization of the situation of Proposition
20.3 characterizing the existence of bi-invariant metrics on Lie groups. All this is built into
the definition of a reductive homogeneous space given by Definition 22.8.
It is possible to express the Levi-Civita connection on a reductive homogeneous space in
terms of the Lie bracket on g, but in general this formula is not very useful. A simplification
of this formula is obtained if a certain condition holds. The corresponding spaces are said
to be naturally reductive; see Definition 22.9. A naturally reductive space has the “nice”
property that its geodesics at o are given by applying the coset exponential map to m;
see Proposition 22.27. As we will see from the explicit examples provided in Section 22.7,
naturally reductive spaces “behave” just as nicely as their Lie group counterpart G, and the
coset exponential of m will provide all the necessary geometric information.
A large supply of naturally reductive homogeneous spaces are the symmetric spaces. Such
spaces arise from a Lie group G equipped with an involutive automorphism σ : G → G (with
σ 6= id and σ 2 = id). Let Gσ be the set of fixed points of σ, the subgroup of G given by
Gσ = {g ∈ G | σ(g) = g},
and let Gσ0 be the identity component of Gσ (the connected component of Gσ containing 1).
Consider the +1 and −1 eigenspaces of the derivative dσ1 : g → g of σ, given by
k = {X ∈ g | dσ1 (X) = X}
m = {X ∈ g | dσ1 (X) = −X}.
Pick a closed subgroup K of G such that Gσ0 ⊆ K ⊆ Gσ . Then it can be shown that G/K is
a reductive homogenous space and that g factors as a direct sum k ⊕ m, which makes G/K
a reductive space. Furthermore, if G is connected and if both Gσ0 and K are compact, then
G/K is naturally reductive.
There is an extensive theory of symmetric spaces and our goal is simply to show that
the additional structure afforded by an involutive automorphism of G yields spaces that are
14

naturally reductive. The theory of symmetric spaces was entirely created by one person,
Élie Cartan, who accomplished the tour de force of giving a complete classification of these
spaces using the classification of semisimple Lie algebras that he had obtained earlier. In
Sections 22.8, 22.9, and 22.10, we provide an introduction to symmetric spaces.
In the past five years, we have also come to realize that Lie groups and homogeneous
manifolds, especially naturally reductive ones, are two of the most important topics for their
role in applications. It is remarkable that most familiar spaces, spheres, projective spaces,
Grassmannian and Stiefel manifolds, symmetric positive definite matrices, are naturally re-
ductive manifolds. Remarkably, they all arise from some suitable action of the rotation group
SO(n), a Lie group, who emerges as the master player. The machinery of naturally reductive
manifolds, and of symmetric spaces (which are even nicer!), makes it possible to compute
explicitly in terms of matrices all the notions from differential geometry (Riemannian met-
rics, geodesics, etc.) that are needed to generalize optimization methods to Riemannian
manifolds.
Since we discuss many topics ranging from manifolds to Lie groups, this book is already
quite big, so we resolved ourselves, not without regrets, to omit many proofs. The purist
may be chagrined, but we feel that it is more important to motivate, demystify, and explain,
the reasons for introducing various concepts and to clarify the relationship between these
notions rather than spelling out every proof in full detail. Whenever we omit a proof, we
provide precise pointers to the literature. In some cases (such as the theorem of Hopf and
Rinow), the proof is just too beautiful to be skipped, so we include it.
The motivations for writing these notes arose while the first author was coteaching a
seminar on Special Topics in Machine Perception with Kostas Daniilidis in the Spring of
2004. In the Spring of 2005, the first author gave a version of his course Advanced Geo-
metric Methods in Computer Science (CIS610), with the main goal of discussing statistics
on diffusion tensors and shape statistics in medical imaging. This is when he realized that
it was necessary to cover some material on Riemannian geometry but he ran out of time
after presenting Lie groups and never got around to doing it! Then, in the Fall of 2006 the
first author went on a wonderful and very productive sabbatical year in Nicholas Ayache’s
group (ACSEPIOS) at INRIA Sophia Antipolis, where he learned about the beautiful and
exciting work of Vincent Arsigny, Olivier Clatz, Hervé Delingette, Pierre Fillard, Grégoire
Malandin, Xavier Pennec, Maxime Sermesant, and, of course, Nicholas Ayache, on statistics
on manifolds and Lie groups applied to medical imaging. This inspired him to write chapters
on differential geometry, and after a few additions made during Fall 2007 and Spring 2008,
notably on left-invariant metrics on Lie groups, the little set of notes from 2004 had grown
into a preliminary version of this manuscript. The first author then joined forces with the
second author in 2015, and with her invaluable assistance, produced the present book, as
well, as a second volume dealing with more advanced topics.
We must acknowledge our debt to two of our main sources of inspiration: Berger’s
Panoramic View of Riemannian Geometry [14] and Milnor’s Morse Theory [81]. In our
opinion, Milnor’s book is still one of the best references on basic differential geometry. His
 15

exposition is remarkably clear and insightful, and his treatment of the variational approach
to geodesics is unsurpassed. We borrowed heavily from Milnor [81]. Since Milnor’s book
is typeset in “ancient” typewritten format (1973!), readers might enjoy reading parts of it
typeset in LATEX. We hope that the readers of these notes will be well prepared to read
standard differential geometry texts such as do Carmo [39], Gallot, Hulin, Lafontaine [49]
and O’Neill [91], but also more advanced sources such as Sakai [100], Petersen [93], Jost [64],
Knapp [68], and of course Milnor [81].
The chapters or sections marked with the symbol ~ contain material that is typically
more specialized or more advanced, and they can be omitted upon first (or second) reading.
Acknowledgement: We would like to thank Eugenio Calabi, Ching-Li Chai, Ted Chinburg,
Chris Croke, Ron Donagi, Harry Gingold, H.W. Gould, Herman Gluck, David Harbater, Julia
Hartmann, Jerry Kazdan, Alexander Kirillov, Florian Pop, Steve Shatz, Jim Stasheff, George
Sparling, Doran Zeilberger, and Wolfgand Ziller for their encouragement, advice, inspiration
and for what they taught me. We also thank Christine Allen-Blanchette, Arthur Azevedo
de Amorim, Kostas Daniilidis, Carlos Esteves, Spyridon Leonardos, Stephen Phillips, João
Sedoc, Marcelo Siqueira, and Roberto Tron for reporting typos and for helpful comments.
16
Contents

Contents 17

I Introduction to Differential Manifolds and Lie Groups 23

1 The Matrix Exponential; Some Matrix Lie Groups 25
1.1 The Exponential Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2 Some Classical Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.3 Symmetric and Other Special Matrices . . . . . . . . . . . . . . . . . . . . . 41
1.4 Exponential of Some Complex Matrices . . . . . . . . . . . . . . . . . . . . 43
1.5 Hermitian and Other Special Matrices . . . . . . . . . . . . . . . . . . . . . 46
1.6 The Lie Group SE(n) and the Lie Algebra se(n) . . . . . . . . . . . . . . . 47
1.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

2 Adjoint Representations and the Derivative of exp 61

2.1 Adjoint Representations Ad and ad . . . . . . . . . . . . . . . . . . . . . . . 61
2.2 The Derivative of exp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

3 Introduction to Manifolds and Lie Groups 71

3.1 Introduction to Embedded Manifolds . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Linear Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.3 Homomorphisms of Linear Lie groups and Lie Algebras . . . . . . . . . . . . 102
3.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

4 Groups and Group Actions 121

4.1 Basic Concepts of Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2 Group Actions: Part I, Definition and Examples . . . . . . . . . . . . . . . 127
4.3 Group Actions: Part II, Stabilizers and Homogeneous Spaces . . . . . . . . 139
4.4 The Grassmann and Stiefel Manifolds . . . . . . . . . . . . . . . . . . . . . 147
4.5 Topological Groups ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

5 The Lorentz Groups ~ 163

17
18 CONTENTS

5.1 The Lorentz Groups O(n, 1), SO(n, 1) and SO0 (n, 1) . . . . . . . . . . . . . 163
5.2 The Lie Algebra of the Lorentz Group SO0 (n, 1) . . . . . . . . . . . . . . . 178
5.3 The Surjectivity of exp : so(1, 3) → SO0 (1, 3) . . . . . . . . . . . . . . . . . 184
5.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

6 The Structure of O(p, q) and SO(p, q) 197

6.1 Polar Forms for Matrices in O(p, q) . . . . . . . . . . . . . . . . . . . . . . . 197
6.2 Pseudo-Algebraic Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.3 More on the Topology of O(p, q) and SO(p, q) . . . . . . . . . . . . . . . . . 206
6.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

7 Manifolds, Tangent Spaces, Cotangent Spaces 211

7.1 Charts and Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
7.2 Tangent Vectors, Tangent Spaces . . . . . . . . . . . . . . . . . . . . . . . . 229
7.3 Tangent Vectors as Derivations . . . . . . . . . . . . . . . . . . . . . . . . . 234
7.4 Tangent and Cotangent Spaces Revisited ~ . . . . . . . . . . . . . . . . . . 243
7.5 Tangent Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
7.6 Submanifolds, Immersions, Embeddings . . . . . . . . . . . . . . . . . . . . 254
7.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

8 Construction of Manifolds From Gluing Data ~ 263

8.1 Sets of Gluing Data for Manifolds . . . . . . . . . . . . . . . . . . . . . . . 263
8.2 Parametric Pseudo-Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . 279

9 Vector Fields, Integral Curves, Flows 283

9.1 Tangent and Cotangent Bundles . . . . . . . . . . . . . . . . . . . . . . . . 283
9.2 Vector Fields, Lie Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . 288
9.3 Integral Curves, Flows, One-Parameter Groups . . . . . . . . . . . . . . . . 296
9.4 Log-Euclidean Polyaffine Transformations . . . . . . . . . . . . . . . . . . . 306
9.5 Fast Polyaffine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
9.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310

10 Partitions of Unity, Covering Maps ~ 313

10.1 Partitions of Unity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
10.2 Covering Maps and Universal Covering Manifolds . . . . . . . . . . . . . . . 322
10.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330

11 Basic Analysis: Review of Series and Derivatives 333

11.1 Series and Power Series of Matrices . . . . . . . . . . . . . . . . . . . . . . . 333
11.2 The Derivative of a Function Between Normed Spaces . . . . . . . . . . . . 343
11.3 Linear Vector Fields and the Exponential . . . . . . . . . . . . . . . . . . . 359
11.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
CONTENTS 19

12 A Review of Point Set Topology 365

12.1 Topological Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
12.2 Continuous Functions, Limits . . . . . . . . . . . . . . . . . . . . . . . . . . 372
12.3 Connected Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
12.4 Compact Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
12.5 Quotient Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
12.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396

II Riemannian Geometry, Lie Groups, Homogeneous Spaces 399

13 Riemannian Metrics, Riemannian Manifolds 401
13.1 Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
13.2 Riemannian Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
13.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410

14 Connections on Manifolds 413

14.1 Connections on Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414
14.2 Parallel Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
14.3 Connections Compatible with a Metric . . . . . . . . . . . . . . . . . . . . . 423
14.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430

15 Geodesics on Riemannian Manifolds 433

15.1 Geodesics, Local Existence and Uniqueness . . . . . . . . . . . . . . . . . . 434
15.2 The Exponential Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
15.3 Complete Riemannian Manifolds, Hopf-Rinow, Cut Locus . . . . . . . . . . 449
15.4 Convexity, Convexity Radius . . . . . . . . . . . . . . . . . . . . . . . . . . 456
15.5 Hessian of a Function on a Riemannian Manifold . . . . . . . . . . . . . . . 458
15.6 The Calculus of Variations Applied to Geodesics . . . . . . . . . . . . . . . 460
15.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467

16 Curvature in Riemannian Manifolds 471

16.1 The Curvature Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
16.2 Sectional Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
16.3 Ricci Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
16.4 The Second Variation Formula and the Index Form . . . . . . . . . . . . . . 488
16.5 Jacobi Fields and Conjugate Points . . . . . . . . . . . . . . . . . . . . . . . 493
16.6 Jacobi Fields and Geodesic Variations . . . . . . . . . . . . . . . . . . . . . 500
16.7 Topology and Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
16.8 Cut Locus and Injectivity Radius: Some Properties . . . . . . . . . . . . . . 513
16.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516

17 Isometries, Submersions, Killing Vector Fields 519

20 CONTENTS

17.1 Isometries and Local Isometries . . . . . . . . . . . . . . . . . . . . . . . . . 520

17.2 Riemannian Covering Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
17.3 Riemannian Submersions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
17.4 Isometries and Killing Vector Fields . . . . . . . . . . . . . . . . . . . . . . 532
17.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534

18 Lie Groups, Lie Algebra, Exponential Map 537

18.1 Lie Groups and Lie Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . 539
18.2 Left and Right Invariant Vector Fields, Exponential Map . . . . . . . . . . . 545
18.3 Homomorphisms, Lie Subgroups . . . . . . . . . . . . . . . . . . . . . . . . 552
18.4 The Correspondence Lie Groups–Lie Algebras . . . . . . . . . . . . . . . . . 556
18.5 Semidirect Products of Lie Algebras and Lie Groups . . . . . . . . . . . . . 557
18.6 Universal Covering Groups ~ . . . . . . . . . . . . . . . . . . . . . . . . . . 567
18.7 The Lie Algebra of Killing Fields ~ . . . . . . . . . . . . . . . . . . . . . . 568
18.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 570

19 The Derivative of exp and Dynkin’s Formula ~ 585

19.1 The Derivative of the Exponential Map . . . . . . . . . . . . . . . . . . . . 585
19.2 The Product in Logarithmic Coordinates . . . . . . . . . . . . . . . . . . . 587
19.3 Dynkin’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
19.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589

20 Metrics, Connections, and Curvature on Lie Groups 591

20.1 Left (resp. Right) Invariant Metrics . . . . . . . . . . . . . . . . . . . . . . 592
20.2 Bi-Invariant Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
20.3 Connections and Curvature of Left-Invariant Metrics . . . . . . . . . . . . . 601
20.4 Connections and Curvature of Bi-Invariant Metrics . . . . . . . . . . . . . . 607
20.5 Simple and Semisimple Lie Algebras and Lie Groups . . . . . . . . . . . . . 612
20.6 The Killing Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615
20.7 Left-Invariant Connections and Cartan Connections . . . . . . . . . . . . . 621
20.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

21 The Log-Euclidean Framework 635

21.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
21.2 A Lie Group Structure on SPD(n) . . . . . . . . . . . . . . . . . . . . . . . 636
21.3 Log-Euclidean Metrics on SPD(n) . . . . . . . . . . . . . . . . . . . . . . . 637
21.4 A Vector Space Structure on SPD(n) . . . . . . . . . . . . . . . . . . . . . 641
21.5 Log-Euclidean Means . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
21.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643

22 Manifolds Arising from Group Actions 645

22.1 Proper Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647
22.2 Proper and Free Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
CONTENTS 21

22.3 Riemannian Submersions and Coverings ~ . . . . . . . . . . . . . . . . . . . 652

22.4 Reductive Homogeneous Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 658
22.5 Examples of Reductive Homogeneous Spaces . . . . . . . . . . . . . . . . . 668
22.6 Naturally Reductive Homogeneous Spaces . . . . . . . . . . . . . . . . . . . 672
22.7 Examples of Naturally Reductive Homogeneous Spaces . . . . . . . . . . . . 678
22.8 A Glimpse at Symmetric Spaces . . . . . . . . . . . . . . . . . . . . . . . . 685
22.9 Examples of Symmetric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . 693
22.10 Types of Symmetric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 707
22.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712

Bibliography 717

Symbol Index 725

Index 733
22 CONTENTS
 Part I

Introduction to Differential Manifolds

and Lie Groups

23
Chapter 1

The Matrix Exponential; Some

Matrix Lie Groups

Le rôle prépondérant de la théorie des groupes en mathématiques a été longtemps

insoupçonné; il y a quatre-vingts ans, le nom même de groupe était ignoré. C’est Galois
qui, le premier, en a eu une notion claire, mais c’est seulement depuis les travaux de
Klein et surtout de Lie que l’on a commencé à voir qu’il n’y a presque aucune théorie
mathématique où cette notion ne tienne une place importante.
—Henri Poincaré

The purpose of this chapter and the next two chapters is to give a “gentle” and fairly
concrete introduction to manifolds, Lie groups and Lie algebras, our main objects of study.
Most texts on Lie groups and Lie algebras begin with prerequisites in differential geometry
that are often formidable to average computer scientists (or average scientists, whatever that
means!). We also struggled for a long time, trying to figure out what Lie groups and Lie
algebras are all about, but this can be done! A good way to sneak into the wonderful world
of Lie groups and Lie algebras is to play with explicit matrix groups such as the group
of rotations in R2 (or R3 ) and with the exponential map. After actually computing the
exponential A = eB of a 2 × 2 skew symmetric matrix B and observing that it is a rotation
matrix, and similarly for a 3 × 3 skew symmetric matrix B, one begins to suspect that there
is something deep going on. Similarly, after the discovery that every real invertible n × n
matrix A can be written as A = RP , where R is an orthogonal matrix and P is a positive
definite symmetric matrix, and that P can be written as P = eS for some symmetric matrix
S, one begins to appreciate the exponential map.
Our goal in this chapter is to give an elementary and concrete introduction to Lie groups
and Lie algebras by studying a number of the so-called classical groups, such as the general
linear group GL(n, R), the special linear group SL(n, R), the orthogonal group O(n), the
special orthogonal group SO(n), and the group of affine rigid motions SE(n), and their Lie
algebras gl(n, R) (all matrices), sl(n, R) (matrices with null trace), o(n), and so(n) (skew

25
26 CHAPTER 1. THE MATRIX EXPONENTIAL; SOME MATRIX LIE GROUPS

symmetric matrices). Lie groups are at the same time, groups, topological spaces, and
manifolds, so we will also have to introduce the crucial notion of a manifold .
The inventors of Lie groups and Lie algebras (starting with Lie!) regarded Lie groups as
groups of symmetries of various topological or geometric objects. Lie algebras were viewed
as the “infinitesimal transformations” associated with the symmetries in the Lie group. For
example, the group SO(n) of rotations is the group of orientation-preserving isometries of
the Euclidean space En . The Lie algebra so(n, R) consisting of real skew symmetric n × n
matrices is the corresponding set of infinitesimal rotations. The geometric link between a Lie
group and its Lie algebra is the fact that the Lie algebra can be viewed as the tangent space
to the Lie group at the identity. There is a map from the tangent space to the Lie group,
called the exponential map. The Lie algebra can be considered as a linearization of the Lie
group (near the identity element), and the exponential map provides the “delinearization,”
i.e., it takes us back to the Lie group. These concepts have a concrete realization in the
case of groups of matrices and, for this reason, we begin by studying the behavior of the
exponential maps on matrices.
We begin by defining the exponential map on matrices and proving some of its properties.
The exponential map allows us to “linearize” certain algebraic properties of matrices. It also
plays a crucial role in the theory of linear differential equations with constant coefficients.
But most of all, as we mentioned earlier, it is a stepping stone to Lie groups and Lie algebras.
On the way to Lie algebras, we derive the classical “Rodrigues-like” formulae for rotations
and for rigid motions in R2 and R3 . We give an elementary proof that the exponential map
is surjective for both SO(n) and SE(n), not using any topology, just certain normal forms
for matrices (see Gallier [48], Chapters 12 and 13).
In Chapter 2, in preparation for defining the Lie bracket on the Lie algebra of a Lie
group, we introduce the adjoint representations of the group GL(n, R) and of the Lie algebra
gl(n, R). The map Ad : GL(n, R) → GL(gl(n, R)) is defined such that AdA is the derivative
of the conjugation map AdA : GL(n, R) → GL(n, R) at the identity. The map ad is the
derivative of Ad at the identity, and it turns out that adA (B) = [A, B], the Lie bracket of A
and B, and in this case, [A, B] = AB − BA. We also find a formula for the derivative of the
matrix exponential exp.
Chapter 3 gives an introduction to manifolds, Lie groups and Lie algebras. Rather than
defining abstract manifolds in terms of charts, atlases, etc., we consider the special case of
embedded submanifolds of RN . This approach has the pedagogical advantage of being more
concrete since it uses parametrizations of subsets of RN , which should be familiar to the
reader in the case of curves and surfaces. The general definition of a manifold will be given
in Chapter 7.
Also, rather than defining Lie groups in full generality, we define linear Lie groups us-
ing the famous result of Cartan (apparently actually due to Von Neumann) that a closed
subgroup of GL(n, R) is a manifold, and thus a Lie group. This way, Lie algebras can be
“computed” using tangent vectors to curves of the form t 7→ A(t), where A(t) is a matrix.
Another Random Scribd Document
with Unrelated Content
The text on this page is estimated to be only 20.14%
accurate

248 THE HOME MECHANIC. WALL PANEL DECORATIONS.

The text on this page is estimated to be only 2.70%
accurate

THE HOME MECHANIC. 249 WALL DECOHATIONS. »fl >^ f

K O k^ »v s K fi % J2! » w O E4 S 0 f M O c w » ^ H d tt 53 t^ hj a
M ^H ^ • O £S Hj M « »^ 3
The text on this page is estimated to be only 6.17%
accurate

250 THE HOME MECHANIC. td >— I ?? O ^ H W g O O o I

— I o l-H o l-H fa to
THE HOME MECHANIC. 251
The text on this page is estimated to be only 28.25%
accurate

252 THE HOME MECHANIC.

The text on this page is estimated to be only 6.17%
accurate

THE HOME MECHANIC. 253 I— I O "^ tI— I o W b3 W O O

c
 254 THE HOME MECHANIC. A COMPLETE SYSTEM OP
COACH PAINTING. This is a brancli of the art of great importance,
yet one but very imperfectly known to the masses. One reason for
this statement is that coach painters generally keep a very close
surveillance over visitors, few, if any, are allowed to enter their
workshops, and no class of painters are so unwilling to communicate
any of the principals or theories of their business. House painting,
sign painting, etc., are branches of the art that cannot be well
concealed from the public gaze, for the operations are so common,
and of every day occurrence, that any person of moderate ability can
soon obtain sufficient information to enable them to do a passable
piece of work, especially in house painting. Graining requires not
only good judgment, a true sense of the various woods and marbles
he wishes to imitate ; but also the organ of imitation well developed.
Sign painting being in part mechanical, has been acquired with more
ease than many other branches of the art, yet there are some points
got only by long practice and through the help of others in the same
line. But, as stated above, coach painting differs so much from any
of the above in the method, mixtures, etc., that even a good clear
knowledge of any of the others is of no use when applied to coach
painting. In starting with giving the pupil a correct idea of coach
painting, it will be necessary first to mention the kind of materials,
tools, etc., required. In no branch of painting should more care and
knowledge be employed than in selecting the most suitable stock
before commencing a piece of work ; for without paying strict
attention to this fact, good work cannot be done. A first-rate coach
painter cannot execute a piece of work to
 THE HOME MECHANIC. 255 advantage with poor materials
; the paints must be of the best quality, varnish, japan, oil and
turpentine dryers, etc. ; let those be good and the chances are all in
your favor. MATERIALS REQUIRED. White lead, B. B. brand, ground
in oil. Same quality, dry. Japan dryer. No. 1. Black Japan. Body copal
varnish (flowing). Body (hard), for bringing up the work, for rubbing
down and using in the previous coats, then finish with the body
flowing. Carriage varnish, Nos. 1 and 2. Drop, or Frankfort black.
Chrome yellows, of all shades to orange. Indian red. Vermilion.
Venetian red. Orange mineral. Red lead. Scarlet lake. Crimson lake. )
ri i, • j. ? • i_ Purple lake. ] ^^^^^ P^'^*^^' ''^'^ P'^^' Yellow
lake. Carmine Purple Brown. Burned Umber. Raw Umber. Burned and
Raw Sienna. Chrome greens, assorted shades. Olive, or quaker's
green. Ultramarine blue. Prussian blue. I Cobalt blue. French yellow
(ocher). Whiting, best quality. Ground pumice. Pumice in lump.
Rotten stone, very fine.
 256 THE HOME MECHANIC. Tripoli powder. Granulated zinc
(putty powder). Fine wheat flower. Olive oil for polishing. Brushes of
suitable sizes, short for painting. Oval bound varnish brushes. Sash
tools, assorted sizes. Pencils, large and small (sable). Long liners for
striping (camel hair or sable). For fine lines, long black sable. Sand
paper, assorted numbers. Emery paper, for rubbing old work, viz. :
iron, springs, bolts, shackles, etc. Dusters. A few tressels to stand
bodies upon of a suitable hight. A wheel stand, which is an upright
mortised into a square block of wood, about two inches above the
half diameter of the largest wheel ; have a pin of iron put through
the upright, a little less than the size of the axle, and long enough to
hold the wheel while painting or when striping. And another, made
with a block or stool, with an upright bolt or pin from the center, so
as to place a wheel upon a horizontal position. A bench for the flag
and muller. A small paint mill attached to the bench for grinding
colors. A square putty knife. A diamond pointed putty knife. A pallet
knife. A hal/round file. } ^•"' "^"^'''''S «« work. Flat broad chisel.
Flat scraper, or old plane iron, used when firing off old paint. Varnish
pots (tin), with a wire across the mouth to wipe the brush over. Paint
cans, assorted sizes. Oil cans, etc., marked for boiled and raw.
Turpentine can. Japan can. Cloth, double milled, for rubbing the
varnish coats. Sponge and chamois skin, etc.
 THE HOME MECHANIC. 257 Having mentioned almost all
the necessaries required to furnish a coach painter, I will now
proceed with the general instruction for preparing a carriage for
painting. First, take and dust all the machine carefully before taking
it into the paint room. Have two boards like small table tops
connected by a bolt in the center, so as to re. volve wben wanted.
Place these upon the tressels, then take the body off the under
carriage and set it upon the smaK tressels before mentioned, then
take off the wheels and let the carriage rest upon the axle on a
tressel of a proper hight, same as the wbeels ; take the wheel stand
and place upon it one of the wheels, then the work will be all easy of
access and ready to commence painting. In coach painting very little
oil paint is required, and here is the difference that occurs between
house and coach painting. Only two or three coats of oil paint is
sufficient in bringing up the work. This we call priming. Just give as
many coats as will carry out a slight gloss, no more. The best oil for
this purpose is good drying oil, say twothirds oil and one-third
turpentine, and have a little of the fine litharge ground into it, about
two ounces to twenty pounds of paint. The litharge causes the paint
to become hard, and rub down better than if only drying oil alone
was used. We mention litharge because it is the best dryer for this
purpose. Patent dryer is, no doubt, convenient, being already
ground, but it never dries paint so hard as litharge, nor rubs down
with the sand paper so well, being more apt to hang to or clog the
paper, and roll or draw upon the work. After all is first coated and
dry, take your sand paper and rub down very smooth, making sure
to cut down all inequalities, ridges, or deficiencies in finish by the
coach builder, otherwise by neglecting to do so on the first coat, you
would have the mortification to find that those places would cut
through in your second coat, thus losing your time to a considerable
amount. You now dust your work carefully, and mix up some putty
(better make your own) with whiting and oil (raw oil) and about one-
fourth part white lead, and work into it while l^neading it, a little
pulverized litharge to make it
 258 THE HOME MECHANIC. dry good and hard. This is very
important, as the putty should be dry, otherwise it would rub up in
the process of the work, and spoil the entire job. Observe that the
putty must be stitf, and well worked up, before using it. You now
take your putty -knife and fill up every crack, crevice, knot hole, etc.,
etc ; level every inequality smooth and fine j study rather to have
your filling-up rather higher than otherwise, for you will put that all
right when rubbing down the next coat, but if below the level, you
have to commence puttying again. That is bad work, so be sure to
putty carefully. All this done, then dust off the work again, previous
to second coating. Should your color be too stout or thick, thin with
turpentine, but be sure not to use thin color, for it not only covers
badly, but rubs down very tacky. For all dark colors, use a dark lead
color for your oil coats— '^ merely black and white," but, for
preparing for any such color as light green, let the color be light lead
color. If for a yellow, begin with white, or slightly tinted with a little
chrome yellow. In laying your second coat, be sure to lay it regular,
not fat in some places and spare in others, but as far as possible an
equal fair coat, and we would wish to guard you against one of the
very worst faults a painter can commit, viz. : to leave full or fat
edges upon any part of the work, such as the springs, shackles, bolt-
heads, or nuts, or any part of the can-iage, for if left by neglect,
they will tell upon you at some stage of the work, and at a time,
possibly, when you cannot well repair the evil done ; therefore be
careful. When this second coat is dry, examine minutely, that you
have missed none of the places requiring putty. The second coat
being thoroughly dry, proceed to rub down with your sand-paper,
using a finer quality this time, so as not to cut so deep, hoping that
all projections have been cut down upon the previous coats. Now
dust off your work, and commerce to give the third coat, putting on
a very tight coat (that is, rubbing it out well, not too much paint).
After the work is all dry and hard, you next proceed to
 THE HOME MECHANIC. 259 giving the filling up coats.
Filling is a mixture prepared as follows : Take dry French j^ellow,
litharge, a little white lead, whiting — say about one-sixteenth
litharge — a small quantity of white lead, a small amount of whiting
; a little red lead \\ill also improve it ; take your drying Japan and
pour in as much as will nearly mix it,- put in a very little drying oil,
very careful not to have too much, otherwise the filling will be soft
and tough ; turpentine to thin to a suitable thickness, to make it
spread like a stiff coat of paint. This filling must be laid full, a coat to
flow on freely. A little practice is necessary to find out the exact
quantities of thinners to be used in mixing good ''filling up stuff."
After all is right, commence to coat your body, laying it fullest where
you see any hollow or want that you cannot putty very well, as this
is really its use. Then, if a good job, coat your shafts, springs,
wheels, etc., etc.. etc., and when all is done over, let it stand until all
is perfectly hard (try with your thumb nail) j if it feels hard and don't
peel up, it will stand another coat. In the most of cases, two coats of
filling will be sufficient. The last coat must be allowed to dry hard, as
it has to be all rubbed down with lump pumice stone. Rub the
pumice flat upon a stone before you commence rubbing down
carriage work. When the student begins to rub he is to be very
cautious to avoid cutting through, and here he will find a " proof"
whether or not he has paid attention to rubbing down carefully with
his sand-paper, for if he has not, of course his pumice will cut in all
places neglected, as stated above. In rubbing down with lump
pumice, you must use plenty water ; keep your sponge in your left
hand, and supply water whenever required ; it will cut quicker, and
not be liable to roll up, as it would undoubtedly do if not kept
perfectry wet. Draw your finger or side of the hand over the parts
you are rubbing and you will see when you have it done enough,
that is. all perfectly firm and smooth. Then you have done all that is
required. Take your sponge and wash off all the work ,• wring out
your wash leather (chamois) and dry off the job ; then if all is
perfect, all filled,
 260 THE HOME MECHANIC. the work will feel like a piece of
ivory, quite smooth and level. A panel or piece of coach work cannot
be got level or fit for finishing without being filled up by this method,
for the grain of the wood will always be more or less seen unless
done so. We now come to the next stage of operations, viz. : color.
Whatever color the carriage is to be painted, the color must be fresh
ground, as it is a mixture that drys very quick. If you want a dark
green of the olive shades, you take deep chrome yellow and drop
black ; have your black powdered ; mix the two together in a pot
with the drying japan, and a little turpentine — not too much spirits
at first, as it would not grind well if thin ; put your mixture into the
bench mill and grind it into a clean pot (always washing out the mill,
to keep it clean), then see if the shade is of the kind desired; if too
dark, you want more yellow, or vice versa-, if the color is wanted
warm olive or quakeFs green, put in a little India or Venetian red.
Now take a clean brush, and make up the color to the proper
thickness and commence work at once ; as the color drys pr sets
very quick, be expeditious in laying it on. The same care is not now
so necessary as regards fat edges, as the color will all dry hard. Still,
avoid all careless or slovenly work j always work clean, 'if you want
the work to look well. A small drop of drying oil can be used in very
warm weather, if it should set too fast to allow time to lay the color,
yet an expert never uses any oil, and it is better not to use any,
unless you wish to let the work stand for a longer period of time
than usual. By working the color free of oil, two coats per day can
be given easily. Two coats of this color will be sufficient. I have often
done a good piece of work with one flowing coat. All plain colors are
done in the same manner } all opaque colors that cover well require
no more. Greens, browns, olives, yellows, etc., etc., but for an
ultramarine blue, the work has to be got up very differently. You
must ground up after your filling has been rubbed down, washed,
dried, etc, with a coat in " oil," that is not
 THEJ HOME MECHANIC. iS^l in "japan," using just so much
oil with the color to keep it from setting too quick ; for such colors
use sugar of lead for the dryer, in preference to japan dryers, as the
color will be purer. Grind some good Prussian blue in oil, then add to
a mixture of white lead as much blue as will make a ground dark
enough for to bear the ultramarine blue, that is about as dark as you
wish the carriage to be when finished. You now proceed to give a
nice careful coat over all Ih? work that is to be finished blue. When
dry, if solid enough, it will require no more Prussian blue, but
generally two coats are required. When satisfied that your ground is
good, take some of the best ultramarine blue, grind it on the stone
with a little varnish, pour out some of your body flowing varnish into
a clean pot, put in your ultramarine already ground, and with a clean
brush mix well the blue with your varnish 5 try if there is blue
enough in it, and, if all right, commence to give a very regular
flowing coat all over, and the colored varnish coat will flow on very
evenly and give a tone and transparency, a depth of color which
cannot be got by using full ultramarine alone. We always
recommend two coats for a coach wanted in this peculiar color,
second coat the same as the first, only before giving the second coat
rub down all the work with ground pumice and water, with a cloth.
You thus remove the gloss, and the next coat will lay and flow
better. Let the work stand for a few days, then rub down again with
ground pumice and water, wash, and dry with your skin, after which
the work is all ready for picking out and striping, of which I will lay
down general rules hereafter. CLARET OR LAKE COLORS. Your first
coat is composed of vermilion and rose pink, in oil, as for blue.
Filling all rubbed down, washed, dried, etc., dust clean, then proceed
to give a coat ; this you will find to cover very well. When dry, give
another light coat J as soon as hard, rub with ground pumice and
water, same process as for blue. Dust clean and nice. If you wish a
very rich light claret, you must calculate not to have too much rose
pink in your ground color, for
 2Q2 THE HOME MECHANIC. if you do, you will not get the
color you auticipate. But for darli clarets, use considerable rose pink.
The student must pay attention to these important hints. For the
light shade, have some scarlet lake 5 grind in a similar way as
directed for ultramarine ; take the bodyflowing varnish and put in
the lake ; mix well, and proceed to coat very carefully. This color
sometimes takes three coats, but with due care, and a good brush
hand, it can be attained with two coats. DARKER SHADES. Use more
rose pink in the ground. Then, instead of scarlet lake, use the best
crimson lake, and in the same manner as for the light claret ; two
coats will be enough if carefully laid. In all those colors got, the
principal idea is to have a solid, true body, not shady but fair. STILL
A DARKER SHADE. Vermilion, rose pink, a little ultramarine blue, for
the ground. This is for a purple shade of claret. Ground good and
solid. Then grind some purple lake, same as above j put into your '^
body -flowing varnish;'^ two coats will be enough, if managed well.
Should any imperfections appear, give another coat and all will be
sure. TO PAINT A CARRIAGE PURE CARMINE. This is sometimes
required for fancy machines, or fireengines, hose carriages, etc.
Many painters fail in painting with this very valuable color, and also
through ignorance of how to use it. They very generally attempt to
get the color by the use of the carmine alone. This color, if pure, is
worth $3 per ounce. Ground your work with the best English
vermilion ; see that the color is complete and solid. Then take pure
carmine, grind in a little drying oil, put it into the body-flowing
varnish, and coat very carefully. It will take two coats to produ
 THE HOME MECHANIC. ^63. By this method, one ounce
bottle will do a whole machme, thereby saving a large amount of
cash for carmine, and time, which is capital, and at the same time
produce a very superior color to anything that can be realized from
pure carmine. All the above colors are produced by what is called in
the trade, glaizing. ' A green can be done in the same way : laying a
light I green for a ground, then use the green lake to glaize, as in
the clarets, carmines, ultramarines, etc. ^' JAP AX BROWN." Grind
drop black in japan ; add to it a little vermilion, just enough to see it.
This makes a very rich color, and looks very well with vermilion lines,
or an orange line ; either looks well. OXFORD BROWNS. Take a little
chrome yellow, white lead, India red, best ocher, burned umber, just
white enough to be seen ; yellow is the principal color ; red to warm
it, umber to give it the brown tone. Many varieties of Oxford brown
can be made by the adding more or less of the leading colors of the
compositions. CHROME GREENS OF ALL SHADES, Or greens
composed of chrome yellow and Prussian blue. Let all the greens be
ground in japan. Being opaque, they all work in that way remarkably
well, and by far superior to any attempt in oil. FAWN COLROS.
Yellow, red, a little black, or a little burned umber may be added,
even burned terra de sienna is sometimes put in the mixture. Coach
colors vary from the shades in general use by house and sign
paintf^rs. They are strictly a class of colors peculiar to ttiat branch
of the art., and arf
 264 THE HOME MECHANIC. tied to no cornmon established
rule. I have seen a carriage painted with the cleaning of several
pots, forming a color that would puzzle almost the best colorist to
imitate, yet it looked first rate, although a nondescript ! We advise
the adoption of the principle, as it not only produces varieties of
color but will be a point of economy of great importance, as much
valuable color is often thrown away through this want of judgment,
mto the common receptacle for all left colors called the " smudge
barrel." DRABS. Any variety can be made to suit the most fastidious.
Composition drabs, red black, and yellow, umber ; also some mineral
colors, from which many fine tints can be made. Raw umber and
white alone constitute a good cool drab, and can be toned either
with a little " chrome/^ or " red/' so as to be either rich, or warm.
Any fancy shade of color wanted can be mixed with perfect ease, for
very soon the student will acquire sufficient knowledge of color to
mix any shade he desires. A GOOD, DURABLE, CEHAP COLOR.
Vermilion and Prussian blue. This is an easy way of getting at a
cheap, rich purple, with a small quantity of white lead. ANOTHER.
Vermilion and drop black makes a cheap plum brown, or claret ;
covers well, and looks and stands well, although far deficient to the
claret got up by glaizing with crimson lake, already mentioned 5 but
for low priced work looks very well indeed, and will not cost one-half
the price j as much time is saved, and the materials moderate.
Having given a general outline of colors, we propose to next point
out to the student the manner of " breaking out," or ^' picking out,"
as the trade term has it. That is, laying ui the lines or stripes, '^ blue
ground." If a large carriage, with heavy wheels^ draw lines (with
 THE HOME MECHANIC. S65 japan mixed color) with ground
Frankfort black, from three-quarter inch to one inch broad, on every
part of the carriage, spokes and springs, filloes, hub, etc., etc., etc.,
shafts, etc. Then, with a light primrose, or light orange color, draw
fine lines about three-eighths or a quarter of an inch from the broad
black line, showing that separating distance clear and distinct of the
blue between. And great care is required in drawing the fine lines, so
as to have them true and oqui-distant Wherever the broad black line
is drawn, carry your two fine lines all through the work, bolt heads
and nuts black with one fine line round the edge. For an extra finish
a fine line can be drawn down the center of the black, a deep
orange, or pure white, or gold. This style looks very beautiful. In
drawing fine lines, as a general rule, the color is ground in drying oil,
as it combines more closely, and gives out from the pencil much
better, and makes the lines more regular, although for work in a
hurry, where the lines are not required to be so fine, japan color will
do. GREENS. Generally pick out with black, and if a light green, black
lines will be enough. If a little more expense is to be gone to, run up
the center of the black lines with a white, and not too fine. This
makes a good neat finish. If a dark green, pick out with black, and
run very fine line on each edge of the black, three-eighths of an inch
off the black. With a very bright green, that looks very nice indeed,
or any other color to suit the taste of the artist. CLARETS. Almost all
colors of this class are picked out with black. Fine side lines are
either vermilion or a rich orange -, or side lines of orange, not too
dark, and a vermilion line run up the center of the black ; or one
large black line, with a gold line up the center, about one-sixteenth
of an inch strong.
 266 THE HOME MECHANIC. FAWN COLORS. Pick out with
black, fine line with white on each edge, or brown drab shade — any
color that will show well and be in harmony. OXFORD BROWN. Pick
out with black, fine line vermilion or medium tint of chrome yellow
with a perceptible touch of red in it. Sometimes part the black line
with white down the center. JAPAN OR PLUM BROWN. Vermilion line
looks best of any color, and for cheap work requires nothing more.
BLACK BOLTS AND NUTS. Some dark colors (very dark) look well to
pick out with drab, or fawn color, edge lines vermilion, center line
black This looks very clean and showy. OLIVE, OR QUAKER'S
GREENS. Pick out generally with black, fine line with white orange,
or light green. DRABS OF ALL SHADES. Pick out with black, fine line
with vermilion, or orange high colored. Extra finish, center line
white. PURPLE. Pick out with black, fine line with a bright line of
orange or lilac, or with vermilion. These remarks apply to most of
the colors in use by coach painters, although there can still be a
greater variety of work done than we can enumerate, which can be
put in practice at any time. The next operation is to varnish. Now, it
appears a
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com