Lecture Notes in Lie Groups

arXiv:1104.1106v2 [math.DG] 7 Apr 2011

Vladimir G. Ivancevic∗ Tijana T. Ivancevic†

Abstract
These lecture notes in Lie Groups are designed for a 1–semester third year
or graduate course in mathematics, physics, engineering, chemistry or biology.
This landmark theory of the 20th Century mathematics and physics gives a
rigorous foundation to modern dynamics, as well as field and gauge theories
in physics, engineering and biomechanics. We give both physical and medical
examples of Lie groups. The only necessary background for comprehensive
reading of these notes are advanced calculus and linear algebra.

Contents
1 Preliminaries: Sets, Maps and Diagrams 3
1.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Commutative Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Groups 6

3 Manifolds 8
3.1 Definition of a Manifold . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Formal Definition of a Smooth Manifold . . . . . . . . . . . . . . . . 12
3.3 Smooth Maps Between Smooth Manifolds . . . . . . . . . . . . . . . 13
3.4 Tangent Bundle and Lagrangian Dynamics . . . . . . . . . . . . . . 14
3.5 Cotangent Bundle and Hamiltonian Dynamics . . . . . . . . . . . . . 17
∗
Land Operations Division, Defence Science & Technology Organisation, P.O. Box 1500, Edin-
burgh SA 5111, Australia (e-mail: Vladimir.Ivancevic@dsto.defence.gov.au)
†
Tesla Science Evolution Institute & QLIWW IP Pty Ltd., Adelaide, Australia (e-mail:
tijana.ivancevic@alumni.adelaide.edu.au)

1
4 Lie Groups 18
4.1 Definition of a Lie Group . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 One-Parameter Subgroup . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 Exponential Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.5 Adjoint Representation . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.6 Actions of Lie Groups on Smooth Manifolds . . . . . . . . . . . . . . 26
4.7 Basic Tables of Lie Groups and Their Lie Algebras . . . . . . . . . . 28
4.8 Representations of Lie groups . . . . . . . . . . . . . . . . . . . . . . 31
4.9 Root Systems and Dynkin Diagrams . . . . . . . . . . . . . . . . . . 32
4.9.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.9.2 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.9.3 Dynkin Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.9.4 Irreducible Root Systems . . . . . . . . . . . . . . . . . . . . 36
4.10 Simple and Semisimple Lie Groups and Algebras . . . . . . . . . . . 37

5 Some Classical Examples of Lie Groups 38

5.1 Galilei Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 General Linear Group . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.3 Rotational Lie Groups in Human/Humanoid Biomechanics . . . . . 40
5.3.1 Uniaxial Group of Joint Rotations . . . . . . . . . . . . . . . 41
5.3.2 Three–Axial Group of Joint Rotations . . . . . . . . . . . . . 43
5.3.3 The Heavy Top . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.4 Euclidean Groups of Rigid Body Motion . . . . . . . . . . . . . . . . 45
5.4.1 Special Euclidean Group SE(2) in the Plane . . . . . . . . . 46
5.4.2 Special Euclidean Group SE(3) in the 3D Space . . . . . . . 47
5.5 Basic Mechanical Examples . . . . . . . . . . . . . . . . . . . . . . . 50
5.5.1 SE(2)−Hovercraft . . . . . . . . . . . . . . . . . . . . . . . . 50
5.5.2 SO(3)−Satellite . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.5.3 SE(3)−Submarine . . . . . . . . . . . . . . . . . . . . . . . . 52
5.6 Newton–Euler SE(3)−Dynamics . . . . . . . . . . . . . . . . . . . . 52
5.6.1 SO(3) : Euler Equations of Rigid Rotations . . . . . . . . . . 52
5.6.2 SE(3) : Coupled Newton–Euler Equations . . . . . . . . . . . 53
5.7 Symplectic Group in Hamiltonian Mechanics . . . . . . . . . . . . . 56

6 Medical Applications: Prediction of Injuries 57

6.1 General Theory of Musculo–Skeletal Injury Mechanics . . . . . . . . 57
6.2 Analytical Mechanics of Traumatic Brain Injury (TBI) . . . . . . . . 63
6.2.1 The SE(3)−jolt: the cause of TBI . . . . . . . . . . . . . . . 63
6.2.2 SE(3)−group of brain’s micro–motions within the CSF . . . 64

2
 6.2.3 Brain’s natural SE(3)−dynamics . . . . . . . . . . . . . . . . 65
6.2.4 Brain’s traumatic dynamics: the SE(3)−jolt . . . . . . . . . 68
6.2.5 Brain’s dislocations and disclinations caused by the SE(3)−jolt 69

1 Preliminaries: Sets, Maps and Diagrams

1.1 Sets
Given a map (or, a function) f : A → B, the set A is called the domain of f , and
denoted Dom f . The set B is called the codomain of f , and denoted Cod f. The
codomain is not to be confused with the range of f (A), which is in general only a
subset of B (see [8, 9]).
A map f : X → Y is called injective, or 1–1, or an injection, iff for every y in the
codomain Y there is at most one x in the domain X with f (x) = y. Put another way,
given x and x′ in X, if f (x) = f (x′ ), then it follows that x = x′ . A map f : X → Y
is called surjective, or onto, or a surjection, iff for every y in the codomain Cod f
there is at least one x in the domain X with f (x) = y. Put another way, the range
f (X) is equal to the codomain Y . A map is bijective iff it is both injective and
surjective. Injective functions are called monomorphisms, and surjective functions
are called epimorphisms in the category of sets (see below). Bijective functions are
called isomorphisms.
A relation is any subset of a Cartesian product (see below). By definition, an
equivalence relation α on a set X is a relation which is reflexive, symmetrical and
transitive, i.e., relation that satisfies the following three conditions:

1. Reflexivity: each element x ∈ X is equivalent to itself, i.e., xαx;

2. Symmetry: for any two elements a, b ∈ X, aαb implies bαa; and
3. Transitivity: aαb and bαc implies aαc.

Similarly, a relation ≤ defines a partial order on a set S if it has the following

properties:

1. Reflexivity: a ≤ a for all a ∈ S;

2. Antisymmetry: a ≤ b and b ≤ a implies a = b; and
3. Transitivity: a ≤ b and b ≤ c implies a ≤ c.

A partially ordered set (or poset) is a set taken together with a partial order on
it. Formally, a partially ordered set is defined as an ordered pair P = (X, ≤), where
X is called the ground set of P and ≤ is the partial order of P .

3
1.2 Maps
Let f and g be maps with domains A and B. Then the maps f + g, f − g, f g, and
f /g are defined as follows (see [8, 9]):

(f + g)(x) = f (x) + g(x) domain = A ∩ B,

(f − g)(x) = f (x) − g(x) domain = A ∩ B,
(f g)(x) = f (x) g(x) domain = A ∩ B,
 
f f (x)
(x) = domain = {x ∈ A ∩ B : g(x) 6= 0}.
g g(x)
Given two maps f and g, the composite map f ◦ g, called the composition of f
and g, is defined by
(f ◦ g)(x) = f (g(x)).
The (f ◦g)−machine is composed of the g−machine (first) and then the f −machine,

x → [[g]] → g(x) → [[f ]] → f (g(x)).

√
For example, suppose that y = f (u) = u and u = g(x) = x2 + 1. Since y is a
function of u and u is a function of x, it follows that y is ultimately a function of x.
We calculate this by substitution
p
y = f (u) = f ◦ g = f (g(x)) = f (x2 + 1) = x2 + 1.

If f and g are both differentiable (or smooth, i.e., C ∞ ) maps and h = f ◦ g is the
composite map defined by h(x) = f (g(x)), then h is differentiable and h′ is given by
the product:
h′ (x) = f ′ (g(x)) g ′ (x).
In Leibniz notation, if y = f (u) and u = g(x) are both differentiable maps, then
dy dy du
= .
dx du dx
The reason for the name chain rule becomes clear if we add another link to the chain.
Suppose that we have one more differentiable map x = h(t). Then, to calculate the
derivative of y with respect to t, we use the chain rule twice,
dy dy du dx
= .
dt du dx dt
∂(x,...)
Given a 1–1 continuous (i.e., C 0 ) map F with a nonzero Jacobian ∂(u,...) that
maps a region S onto a region R, we have the following substitution formulas:

4
1. For a single integral,
∂x
Z Z
f (x) dx = f (x(u)) du;
R S ∂u
2. For a double integral,
∂(x, y)
ZZ ZZ
f (x, y) dA = f (x(u, v), y(u, v)) dudv;
R S ∂(u, v)
3. For a triple integral,
∂(x, y, z)
ZZZ ZZZ
f (x, y, z) dV = f (x(u, v, w), y(u, v, w), z(u, v, w)) dudvdw;
R S ∂(u, v, w)
4. Generalization to n−tuple integrals is obvious.

1.3 Commutative Diagrams

Many properties of mathematical systems can be unified and simplified by a presen-
tation with commutative diagrams of arrows. Each arrow f : X → Y represents a
function (i.e., a map, transformation, operator); that is, a source (domain) set X, a
target (codomain) set Y , and a rule x 7→ f (x) which assigns to each element x ∈ X
an element f (x) ∈ Y . A typical diagram of sets and functions is (see [8, 9]):
f ✲ f✲
X Y X f (X)
❅ ❅
h❅ g or h❅ g
❅ ❅
❘ ❄
❅ ❘ ❄
❅
Z g(f (X))
This diagram is commutative iff h = g ◦f , where g ◦f is the usual composite function
g ◦ f : X → Z, defined by x 7→ g(f (x)).
Less formally, composing maps is like following directed paths from one object to
another (e.g., from set to set). In general, a diagram is commutative iff any two paths
along arrows that start at the same point and finish at the same point yield the same
‘homomorphism’ via compositions along successive arrows. Commutativity of the
whole diagram follows from commutativity of its triangular components. Study of
commutative diagrams is popularly called ‘diagram chasing’, and provides a powerful
tool for mathematical thought.
Many properties of mathematical constructions may be represented by universal
properties of diagrams. Consider the Cartesian product X ×Y of two sets, consisting
as usual of all ordered pairs hx, yi of elements x ∈ X and y ∈ Y . The projections

5
hx, yi 7→ x, hx, yi 7→ y of the product on its ‘axes’ X and Y are functions p :
X × Y → X, q : X × Y → Y . Any function h : W → X × Y from a third set W
is uniquely determined by its composites p ◦ h and q ◦ h. Conversely, given W and
two functions f and g as in the diagram below, there is a unique function h which
makes the following diagram commute:
W
❅
f ❅ g
h
❅
✠ ❄ ❘
❅
X ✛p X ×Y q
✲Y

This property describes the Cartesian product X × Y uniquely; the same diagram,
read in the category of topological spaces or of groups, describes uniquely the Carte-
sian product of spaces or of the direct product of groups.

2 Groups
A group is a pointed set (G, e) with a multiplication µ : G × G → G and an inverse
ν : G → G such that the following diagrams commute (see [10, 8, 9]):

1.
(e, 1)
✲ (1, e)
✲
G G×G G
❅ ✒
1❅ µ 1
❅
❘ ❄
❅
G
(e is a two–sided identity)

2.
µ×1✲
G×G×G G×G

1×µ µ
❄ ❄
G×G ✲G
µ
(associativity)

6
 3.
(ν, 1)
✲ (1, ν)
✲
G G×G G
❅ ✒
e❅ µ e
❅
❘ ❄
❅
G
(inverse).

Here e : G → G is the constant map e(g) = e for all g ∈ G. (e, 1) means the map
such that (e, 1)(g) = (e, g), etc. A group G is called commutative or Abelian group
if in addition the following diagram commutes

G×G T ✲G×G
❅
µ❅ µ
❅
❘
❅ ✠
G
where T : G × G → G × G is the switch map T (g1 , g2 ) = (g2 , g1 ), for all (g1 , g2 ) ∈
G × G.
A group G acts (on the left) on a set A if there is a function α : G × A → A
such that the following diagrams commute:

1.
(e, 1)
✲
A G×A
❅
1❅ α
❅
❘ ❄
❅
A
2.
1×α✲
G×G×A G×A

µ×1 α
❄ ❄
G×A ✲A
α
where (e, 1)(x) = (e, x) for all x ∈ A. The orbits of the action are the sets
Gx = {gx : g ∈ G} for all x ∈ A.

7
 Given two groups (G, ∗) and (H, ·), a group homomorphism from (G, ∗) to (H, ·)
is a function h : G → H such that for all x and y in G it holds that

h(x ∗ y) = h(x) · h(y).

From this property, one can deduce that h maps the identity element eG of G to
the identity element eH of H, and it also maps inverses to inverses in the sense that
h(x−1 ) = h(x)−1 . Hence one can say that h is compatible with the group structure.
The kernel Ker h of a group homomorphism h : G → H consists of all those
elements of G which are sent by h to the identity element eH of H, i.e.,

Ker h = {x ∈ G : h(x) = eH }.

The image Im h of a group homomorphism h : G → H consists of all elements

of G which are sent by h to H, i.e.,

Im h = {h(x) : x ∈ G}.

The kernel is a normal subgroup of G and the image is a subgroup of H. The

homomorphism h is injective (and called a group monomorphism) iff Ker h = eG ,
i.e., iff the kernel of h consists of the identity element of G only.

3 Manifolds
A manifold is an abstract mathematical space, which locally (i.e., in a close–up
view) resembles the spaces described by Euclidean geometry, but which globally
(i.e., when viewed as a whole) may have a more complicated structure (see [11]).
For example, the surface of Earth is a manifold; locally it seems to be flat, but viewed
as a whole from the outer space (globally) it is actually round. A manifold can be
constructed by ‘gluing’ separate Euclidean spaces together; for example, a world
map can be made by gluing many maps of local regions together, and accounting
for the resulting distortions.
As main pure–mathematical references for manifolds we recommend popular
graduate textbooks by two ex–Bourbaki members, Serge Lang [13, 12] and Jean
Dieudonne [14, 15]. Besides, the reader might wish to consult some other ‘classics’,
including [11, 16, 17, 18, 19, 20, 3]. Finally, as first–order applications, we recom-
mend three popular textbooks in mechanics, [2, 1, 4], as well as our own geometrical
monographs [8, 9].
Another example of a manifold is a circle S 1 . A small piece of a circle appears to
be like a slightly–bent part of a straight line segment, but overall the circle and the

8
segment are different 1D manifolds. A circle can be formed by bending a straight
line segment and gluing the ends together.1
The surfaces of a sphere 2 and a torus 3 are examples of 2D manifolds. Manifolds
1
Locally, the circle looks like a line. It is 1D, that is, only one coordinate is needed to say
where a point is on the circle locally. Consider, for instance, the top part of the circle, where the
y−coordinate is positive. Any point in this part can be described by the x−coordinate. So, there is a
continuous bijection χtop (a mapping which is 1–1 both ways), which maps the top part of the circle
to the open interval (−1, 1), by simply projecting onto the first coordinate: χtop (x, y) = x. Such
a function is called a chart. Similarly, there are charts for the bottom, left , and right parts of the
circle. Together, these parts cover the whole circle and the four charts form an atlas (see the next
subsection) for the circle. The top and right charts overlap: their intersection lies in the quarter
of the circle where both the x− and the y−coordinates are positive. The two charts χtop and
χright map this part bijectively to the interval (0, 1). Thus a function T from (0, 1) to itself can be
constructed, which first inverts the top chart to reach the circle and then follows the right chart
back to the interval:
 p  p
T (a) = χright χ−1


top (a) = χright a, 1 − a2 = 1 − a2 .

Such a function is called a transition map. The top, bottom, left, and right charts show that the
circle is a manifold, but they do not form the only possible atlas. Charts need not be geometric
projections, and the number of charts is a matter of choice. T and the other transition functions
are differentiable on the interval (0, 1). Therefore, with this atlas the circle is a differentiable, or
smooth manifold.
2
The surface of the sphere S 2 can be treated in almost the same way as the circle S 1 . It can be
viewed as a subset of R3 , defined by: S = {(x, y, z) ∈ R3 |x2 + y 2 + z 2 = 1}. The sphere is 2D, so
each chart will map part of the sphere to an open subset of R2 . Consider the northern hemisphere,
which is the part with positive z coordinate. The function χ defined by χ(x, y, z) = (x, y), maps
the northern hemisphere to the open unit disc by projecting it on the (x, y)−plane. A similar chart
exists for the southern hemisphere. Together with two charts projecting on the (x, z)−plane and two
charts projecting on the (y, z)−plane, an atlas of six charts is obtained which covers the entire sphere.
This can be easily generalized to an nD sphere S n = {(x1 , x2 , ..., xn ) ∈ Rn |x21 + x22 + ... + x2n = 1}.
An n−sphere S n can be also constructed by gluing together two copies of Rn . The transition
map between them is defined as Rn \ {0} → Rn \ {0} : x 7→ x/kxk2 . This function is its own inverse,
so it can be used in both directions. As the transition map is a (C ∞ )−smooth function, this atlas
defines a smooth manifold.
3
A torus (pl. tori), denoted by T 2 , is a doughnut–shaped surface of revolution generated by
revolving a circle about an axis coplanar with the circle. The sphere S 2 is a special case of the
torus obtained when the axis of rotation is a diameter of the circle. If the axis of rotation does not
intersect the circle, the torus has a hole in the middle and resembles a ring doughnut, a hula hoop
or an inflated tire. The other case, when the axis of rotation is a chord of the circle, produces a
sort of squashed sphere resembling a round cushion.
A torus can be defined parametrically by:

x(u, v) = (R + r cos v) cos u, y(u, v) = (R + r cos v) sin u, z(u, v) = r sin v,

where u, v ∈ [0, 2π], R is the distance from the center of the tube to the center of the torus, and r
is the radius of the tube. According to a broader definition, the generator of a torus need not be a
circle but could also be an ellipse or any other conic section.
Topologically, a torus is a closed surface defined as product of two circles: T 2 = S 1 × S 1 . The

9
are important objects in mathematics, physics and control theory, because they
allow more complicated structures to be expressed and understood in terms of the
well–understood properties of simpler Euclidean spaces (see [9]).
The Cartesian product of manifolds is also a manifold (note that not every man-
ifold can be written as a product). The dimension of the product manifold is the
sum of the dimensions of its factors. Its topology is the product topology, and a
Cartesian product of charts is a chart for the product manifold. Thus, an atlas
for the product manifold can be constructed using atlases for its factors. If these
atlases define a differential structure on the factors, the corresponding atlas defines
a differential structure on the product manifold. The same is true for any other
structure defined on the factors. If one of the factors has a boundary, the product
manifold also has a boundary. Cartesian products may be used to construct tori
and cylinders, for example, as S 1 × S 1 and S 1 × [0, 1], respectively.
Manifolds need not be connected (all in ‘one piece’): a pair of separate circles
is also a topological manifold(see below). Manifolds need not be closed: a line
segment without its ends is a manifold. Manifolds need not be finite: a parabola is
a topological manifold.
Manifolds4 can be viewed using either extrinsic or intrinsic view. In the extrinsic
view, usually used in geometry and topology of surfaces, an nD manifold M is seen
as embedded in an (n + 1)D Euclidean space Rn+1 . Such a manifold is called a
‘codimension 1 space’. With this view it is easy to use intuition from Euclidean
spaces to define additional structure. For example, in a Euclidean space it is always
clear whether a vector at some point is tangential or normal to some surface through
that point. On the other hand, the intrinsic view of an nD manifold M is an
abstract way of considering M as a topological space by itself, without any need
surface described above, given the relative topology from R3 , is homeomorphic to a topological torus
as long as it does not intersect its own axis.
One can easily generalize the torus to arbitrary dimensions. An n−torus T n is defined as a
product of n circles: T n = S 1 × S 1 × · · · × S 1 . Equivalently, the n−torus is obtained from the
n−cube (the Rn −generalization of the ordinary cube in R3 ) by gluing the opposite faces together.
An n−torus T n is an example of an nD compact manifold. It is also an important example of a
Lie group (see below).
4

Additional structures are often defined on manifolds. Examples of manifolds with additional
structure include:

• differentiable (or, smooth manifolds, on which one can do calculus;

• Riemannian manifolds, on which distances and angles can be defined; they serve as the
configuration spaces in mechanics;
• symplectic manifolds, which serve as the phase spaces in mechanics and physics;
• 4D pseudo–Riemannian manifolds which model space–time in general relativity.

10
for surrounding (n + 1)D Euclidean space. This view is more flexible and thus it
is usually used in high–dimensional mechanics and physics (where manifolds used
represent configuration and phase spaces of dynamical systems), can make it harder
to imagine what a tangent vector might be.

3.1 Definition of a Manifold

Consider a set M (see Figure 1) which is a candidate for a manifold. Any point
x ∈ M has its Euclidean chart, given by a 1–1 and onto map ϕi : M → Rn , with its
Euclidean image Vi = ϕi (Ui ). More precisely, a chart ϕi is defined by (see [8, 9])

ϕi : M ⊃ Ui ∋ x 7→ ϕi (x) ∈ Vi ⊂ Rn ,

where Ui ⊂ M and Vi ⊂ Rn are open sets.

Figure 1: Geometric picture of the manifold concept.

Clearly, any point x ∈ M can have several different charts (see Figure 1). Con-
sider a case of two charts, ϕi , ϕj : M → Rn , having in their images two open sets,
Vij = ϕi (Ui ∩ Uj ) and Vji = ϕj (Ui ∩ Uj ). Then we have transition functions ϕij
between them,

ϕij = ϕj ◦ ϕ−1
i : Vij → Vji , locally given by ϕij (x) = ϕj (ϕ−1
i (x)).

If transition functions ϕij exist, then we say that two charts, ϕi and ϕj are compat-
ible. Transition functions represent a general (nonlinear) transformations of coordi-
nates, which are the core of classical tensor calculus.

11
 A set of compatible charts ϕi : M → Rn , such that each point x ∈ M has its
Euclidean image in at least one chart, is called an atlas. Two atlases are equivalent
iff all their charts are compatible (i.e., transition functions exist between them), so
their union is also an atlas. A manifold structure is a class of equivalent atlases.
Finally, as charts ϕi : M → Rn were supposed to be 1-1 and onto maps, they
can be either homeomorphisms, in which case we have a topological (C 0 ) manifold,
or diffeomorphisms, in which case we have a smooth (C k ) manifold.

3.2 Formal Definition of a Smooth Manifold

Given a chart (U, ϕ), we call the set U a coordinate domain, or a coordinate neigh-
borhood of each of its points. If in addition ϕ(U ) is an open ball in Rn , then U
is called a coordinate ball. The map ϕ is called a (local) coordinate map, and the
component functions (x1 , ..., xn ) of ϕ, defined by ϕ(m) = (x1 (m), ..., xn (m)), are
called local coordinates on U [8, 9].
Two charts (U1 , ϕ1 ) and (U2 , ϕ2 ) such that U1 ∩ U2 6= ∅ are called compatible
if ϕ1 (U1 ∩ U2 ) and ϕ2 (U2 ∩ U1 ) are open subsets of Rn . A family (Uα , ϕα )α∈A of
compatible charts on M such that the Uα form a covering of M is called an atlas.
The maps ϕαβ = ϕβ ◦ ϕ−1 α : ϕα (Uαβ ) → ϕβ (Uαβ ) are called the transition maps,
for the atlas (Uα , ϕα )α∈A , where Uαβ = Uα ∩ Uβ , so that we have a commutative
triangle:
Uαβ ⊆ M
❅
ϕα ❅ ϕβ
❅
✠ ❘
❅
ϕα (Uαβ ) ✲ ϕβ (Uαβ )
ϕαβ
An atlas (Uα , ϕα )α∈A for a manifold M is said to be a C k −atlas, if all transi-
tion maps ϕαβ : ϕα (Uαβ ) → ϕβ (Uαβ ) are of class C k . Two C k atlases are called
C k −equivalent, if their union is again a C k −atlas for M . An equivalence class of
C k −atlases is called a C k −structure on M . In other words, a smooth structure on
M is a maximal smooth atlas on M , i.e., such an atlas that is not contained in any
strictly larger smooth atlas. By a C k −manifold M , we mean a topological manifold
together with a C k −structure and a chart on M will be a chart belonging to some
atlas of the C k −structure. Smooth manifold means C ∞ −manifold, and the word
‘smooth’ is used synonymously with C ∞ .
Sometimes the terms ‘local coordinate system’ or ‘parametrization’ are used
instead of charts. That M is not defined with any particular atlas, but with an
equivalence class of atlases, is a mathematical formulation of the general covariance
principle. Every suitable coordinate system is equally good. A Euclidean chart

12
may well suffice for an open subset of Rn , but this coordinate system is not to be
preferred to the others, which may require many charts (as with polar coordinates),
but are more convenient in other respects.
For example, the atlas of an n−sphere S n has two charts. If N = (1, 0, ..., 0)
and S = (−1, ..., 0, 0) are the north and south poles of S n respectively, then the two
charts are given by the stereographic projections from N and S:

ϕ1 : S n \{N } → Rn , ϕ1 (x1 , ..., xn+1 ) = (x2 /(1 − x1 ), . . . , xn+1 /(1 − x1 )), and
ϕ2 : S n \{S} → Rn , ϕ2 (x1 , ..., xn+1 ) = (x2 /(1 + x1 ), . . . , xn+1 /(1 + x1 )),

while the overlap map ϕ2 ◦ ϕ−1 n n

1 : R \{0} → R \{0} is given by the diffeomorphism
−1 2
(ϕ2 ◦ ϕ1 )(z) = z/||z|| , for z in R \{0}, from Rn \{0} to itself.
n

Various additional structures can be imposed on Rn , and the corresponding man-

ifold M will inherit them through its covering by charts. For example, if a covering
by charts takes their values in a Banach space E, then E is called the model space
and M is referred to as a C k −Banach manifold modelled on E. Similarly, if a cov-
ering by charts takes their values in a Hilbert space H, then H is called the model
space and M is referred to as a C k −Hilbert manifold modelled on H. If not other-
wise specified, we will consider M to be an Euclidean manifold, with its covering by
charts taking their values in Rn .
For a Hausdorff C k −manifold the following properties are equivalent: (i) it is
paracompact; (ii) it is metrizable; (iii) it admits a Riemannian metric;5 (iv) each
connected component is separable.

3.3 Smooth Maps Between Smooth Manifolds

A map ϕ : M → N between two manifolds M and N , with M ∋ m 7→ ϕ(m) ∈ N , is
called a smooth map, or C k −map, if we have the following charting [8, 9]:

5
Recall the corresponding properties of a Euclidean metric d. For any three points x, y, z ∈ Rn ,
the following axioms are valid:

M1 : d(x, y) > 0, for x 6= y; and d(x, y) = 0, for x = y;

M2 : d(x, y) = d(y, x); M3 : d(x, y) ≤ d(x, z) + d(z, y).

13
 ✬ ✩ ✬ ✩
✬ ✩ ✬ ✩
U M N V
ϕ ✲⊙
m⊙
✫ ✪ ✫ ✪
ϕ(m)
✫ ✪ ✫ ✪

φ ψ

✬ ✩ ✬ ✩
✻ ✻

φ(U ) ❄ ❄ψ(V )
⊙ ✲⊙
φ(m) ψ ◦ ϕ ◦ φ−1 ψ(ϕ(m))
✫ ✪ ✫ ✪

Rm ✲ ✛ Rn
This means that for each m ∈ M and each chart (V, ψ) on N with ϕ (m) ∈ V there
is a chart (U, φ) on M with m ∈ U, ϕ (U ) ⊆ V , and Φ = ψ ◦ ϕ ◦ φ−1 is C k , that is,
the following diagram commutes:
ϕ ✲V ⊆N
M ⊇U

φ ψ
❄ ❄
φ(U ) ✲ ψ(V )
Φ
Let M and N be smooth manifolds and let ϕ : M → N be a smooth map. The
map ϕ is called a covering, or equivalently, M is said to cover N , if ϕ is surjective
and each point n ∈ N admits an open neighborhood V such that ϕ−1 (V ) is a union
of disjoint open sets, each diffeomorphic via ϕ to V .
A C k −map ϕ : M → N is called a C k −diffeomorphism if ϕ is a bijection,
ϕ−1 : N → M exists and is also C k . Two manifolds are called diffeomorphic if there
exists a diffeomorphism between them. All smooth manifolds and smooth maps
between them form the category M.

3.4 Tangent Bundle and Lagrangian Dynamics

The tangent bundle of a smooth n−manifold is the place where tangent vectors live,
and is itself a smooth 2n−manifold. Vector–fields are cross-sections of the tangent
bundle. The Lagrangian is a natural energy function on the tangent bundle (see
[8, 9]).

14
 In mechanics, to each nD configuration manifold M there is associated its 2nD
velocity phase–space manifold, denoted by T M and called the tangent bundle of
M (see Figure 2). The original smooth manifold M is called the base of T M .
There is an onto map π : T M → M , called the projection. Above each point
x ∈ M there is a tangent space Tx M = π −1 (x) to M at x, which is called a fibre.
The fibreGTx M ⊂ T M is the subset of T M , such that the total tangent bundle,
TM = Tx M , is a disjoint union of tangent spaces Tx M to M for all points
m∈M
x ∈ M . From dynamical perspective, the most important quantity in the tangent
bundle concept is the smooth map v : M → T M , which is an inverse to the projection
π, i.e, π ◦ v = IdM , π(v(x)) = x. It is called the velocity vector–field. Its graph
(x, v(x)) represents the cross–section of the tangent bundle T M . This explains
the dynamical term velocity phase–space, given to the tangent bundle T M of the
manifold M .

Figure 2: A sketch of a tangent bundle T M of a smooth manifold M .

If [a, b] is a closed interval, a C 0 −map γ : [a, b] → M is said to be differentiable

at the endpoint a if there is a chart (U, φ) at γ(a) such that the following limit exists
and is finite:
d (φ ◦ γ)(t) − (φ ◦ γ)(a)
(φ ◦ γ)(a) ≡ (φ ◦ γ)′ (a) = lim . (1)
dt t→a t−a
Generalizing (1), we get the notion of the curve on a manifold. For a smooth
manifold M and a point m ∈ M a curve at m is a C 0 −map γ : I → M from an
interval I ⊂ R into M with 0 ∈ I and γ(0) = m.
Two curves γ 1 and γ 2 passing though a point m ∈ U are tangent at m with
respect to the chart (U, φ) if (φ ◦ γ 1 )′ (0) = (φ ◦ γ 2 )′ (0). Thus, two curves are tangent
if they have identical tangent vectors (same direction and speed) in a local chart on
a manifold.
For a smooth manifold M and a point m ∈ M, the tangent space Tm M to M at
m is the set of equivalence classes of curves at m:
Tm M = {[γ]m : γ is a curve at a point m ∈ M }.

15
 A C k −map ϕ : M ∋ m 7→ ϕ(m) ∈ N between two manifolds M and N induces
a linear map Tm ϕ : Tm M → Tϕ(m) N for each point m ∈ M , called a tangent map,
if we have:
Tm (M ) T M T (N ) Tϕ(m) (N )

T (ϕ) ✲ ⊙
⊙

πM πN

✬ ✩ ✬ ✩
❄ ❄
M ⊙ ✲ ⊙ N
ϕ
✫ ✪ ✫ ✪
m ϕ(m)

i.e., the following diagram commutes:

Tm ϕ ✲ Tϕ(m) N
Tm M

πM πN
❄ ❄
M ∋m ✲ ϕ(m) ∈ N
ϕ

with the natural projection π M : T M → M, given by π M (Tm M ) = m, that takes a

tangent vector v to the point m ∈ M at which the vector v is attached i.e., v ∈ Tm M .
For an nD smooth manifold M , its nD tangent bundle T M G is the disjoint union
of all its tangent spaces Tm M at all points m ∈ M , T M = Tm M .
m∈M
To define the smooth structure on T M , we need to specify how to construct
local coordinates on T M . To do this, let (x1 (m), ..., xn (m)) be local coordinates of a
point m on M and let (v 1 (m), ..., v n (m)) be components of a tangent vector in this
coordinate system. Then the 2n numbers (x1 (m), ..., xn (m), v 1 (m), ..., v n (m)) give
a local coordinate
G system on T M .
TM = Tm M defines a family of vector spaces parameterized by M . The
m∈M
inverse image π −1
M (m) of a point m ∈ M under the natural projection π M is the
tangent space Tm M . This space is called the fibre of the tangent bundle over the
point m ∈ M .

16
 A C k −map ϕ : M → N between two manifolds M and N induces a linear
tangent map T ϕ : T M → T N between their tangent bundles, i.e., the following
diagram commutes:
Tϕ ✲ TN
TM

πM πN
❄ ❄
M ✲N
ϕ
All tangent bundles and their tangent maps form the category T B. The category
T B is the natural framework for Lagrangian dynamics.
Now, we can formulate the global version of the chain rule. If ϕ : M → N and
ψ : N → P are two smooth maps, then we have T (ψ ◦ ϕ) = T ψ ◦ T ϕ. In other
words, we have a functor T : M ⇒ T B from the category M of smooth manifolds
to the category T B of their tangent bundles:
M TM
❅ ❅
ϕ ❅ (ψ ◦ ϕ) T
=⇒ Tϕ ❅ T (ψ ◦ ϕ)
❅ ❅
✠ ❘
❅ ✠ ❘
❅
N ✲P TN ✲ TP
ψ Tψ

3.5 Cotangent Bundle and Hamiltonian Dynamics

The cotangent bundle of a smooth n−manifold is the place is where 1–forms live, and
is itself a smooth 2n−manifold. Covector–fields (1–forms) are cross-sections of the
cotangent bundle. The Hamiltonian is a natural energy function on the cotangent
bundle (see [8, 9]).
A dual notion to the tangent space Tm M to a smooth manifold M at a point m
is its cotangent space Tm∗ M at the same point m. Similarly to the tangent bundle,

for a smooth manifold M of dimension n, its cotangent bundle T ∗ M is the G disjoint

∗ M at all points m ∈ M , i.e., T ∗ M =
union of all its cotangent spaces Tm Tm∗ M.

m∈M
Therefore, the cotangent bundle of an n−manifold M is the vector bundle T ∗ M =
(T M )∗ , the (real) dual of the tangent bundle T M .
If M is an n−manifold, then T ∗ M is a 2n−manifold. To define the smooth
structure on T ∗ M , we need to specify how to construct local coordinates on T ∗ M .
To do this, let (x1 (m), ..., xn (m)) be local coordinates of a point m on M and let
(p1 (m), ..., pn (m)) be components of a covector in this coordinate system. Then the
2n numbers (x1 (m), ..., xn (m), p1 (m), ..., pn (m)) give a local coordinate system on
T ∗ M . This is the basic idea one uses to prove that indeed T ∗ M is a 2n−manifold.

17
 G
T ∗M = ∗ M defines a family of vector spaces parameterized by M , with
Tm
m∈M
the conatural projection, π ∗M : T ∗ M → M, given by π ∗M (Tm ∗ M ) = m, that takes a

covector p to the point m ∈ M at which the covector p is attached i.e., p ∈ Tm ∗ M.

−1
The inverse image π M (m) of a point m ∈ M under the conatural projection π ∗M is
the cotangent space Tm ∗ M . This space is called the fibre of the cotangent bundle

over the point m ∈ M .

In a similar way, a C k −map ϕ : M → N between two manifolds M and N induces
a linear cotangent map T ∗ ϕ : T ∗ M → T ∗ N between their cotangent bundles, i.e.,
the following diagram commutes:
T ∗ϕ ✲ T ∗N
T ∗M

π ∗M π ∗N
❄ ❄
M ✲N
ϕ
All cotangent bundles and their cotangent maps form the category T ∗ B. The
category T ∗ B is the natural stage for Hamiltonian dynamics.
Now, we can formulate the dual version of the global chain rule. If ϕ : M → N
and ψ : N → P are two smooth maps, then we have T ∗ (ψ ◦ ϕ) = T ∗ ψ ◦ T ∗ ϕ. In
other words, we have a cofunctor T ∗ : M ⇒ T ∗ B from the category M of smooth
manifolds to the category T ∗ B of their cotangent bundles:
M T ∗M
❅ ✒ ❅
■
ϕ ❅ (ψ ◦ ϕ) T∗ T ∗ϕ ❅ T ∗ (ψ ◦ ϕ)
❅
=⇒ ❅
✠ ❘
❅ ❅
N ✲P T ∗N ✛ T ∗P
ψ T ∗ψ

4 Lie Groups
In this section we present the basics of classical theory of Lie groups and their Lie
algebras, as developed mainly by Sophus Lie, Elie Cartan, Felix Klein, Wilhelm
Killing and Hermann Weyl. For more comprehensive treatment see e.g., [21, 22, 23,
24, 25].
In the middle of the 19th Century S. Lie made a far reaching discovery that
techniques designed to solve particular unrelated types of ODEs, such as separable,
homogeneous and exact equations, were in fact all special cases of a general form of
integration procedure based on the invariance of the differential equation under a

18
continuous group of symmetries. Roughly speaking a symmetry group of a system
of differential equations is a group that transforms solutions of the system to other
solutions. Once the symmetry group has been identified a number of techniques
to solve and classify these differential equations becomes possible. In the classical
framework of Lie, these groups were local groups and arose locally as groups of
transformations on some Euclidean space. The passage from the local Lie group to
the present day definition using manifolds was accomplished by E. Cartan at the
end of the 19th Century, whose work is a striking synthesis of Lie theory, classical
geometry, differential geometry and topology.
These continuous groups, which originally appeared as symmetry groups of dif-
ferential equations, have over the years had a profound impact on diverse areas such
as algebraic topology, differential geometry, numerical analysis, control theory, clas-
sical mechanics, quantum mechanics etc. They are now universally known as Lie
groups.
A Lie group is smooth manifold which also carries a group structure whose
product and inversion operations are smooth as maps of manifolds. These objects
arise naturally in describing physical symmetries.6
6
Here are a few examples of Lie groups and their relations to other areas of mathematics and
physics:

1. Euclidean space Rn is an Abelian Lie group (with ordinary vector addition as the group
operation).
2. The group GLn (R) of invertible matrices (under matrix multiplication) is a Lie group of
dimension n2 . It has a subgroup SLn (R) of matrices of determinant 1 which is also a Lie
group.
3. The group On (R) generated by all rotations and reflections of an nD vector space is a Lie
group called the orthogonal group. It has a subgroup of elements of determinant 1, called
the special orthogonal group SO(n), which is the group of rotations in Rn .
4. Spin groups are double covers of the special orthogonal groups (used e.g., for studying
fermions in quantum field theory).
5. The group Sp2n (R) of all matrices preserving a symplectic form is a Lie group called the
symplectic group.
6. The Lorentz group and the Poincaré group of isometries of space–time are Lie groups of
dimensions 6 and 10 that are used in special relativity.
7. The Heisenberg group is a Lie group of dimension 3, used in quantum mechanics.
8. The unitary group U (n) is a compact group of dimension n2 consisting of unitary matrices.
It has a subgroup of elements of determinant 1, called the special unitary group SU (n).
9. The group U (1) × SU (2) × SU (3) is a Lie group of dimension 1 + 3 + 8 = 12 that is the
gauge group of the Standard Model of elementary particles, whose dimension corresponds to:
1 photon + 3 vector bosons + 8 gluons.

19
 A Lie group is a group whose elements can be continuously parametrized by
real numbers, such as the rotation group SO(3), which can be parametrized by the
Euler angles. More formally, a Lie group is an analytic real or complex manifold
that is also a group, such that the group operations multiplication and inversion
are analytic maps. Lie groups are important in mathematical analysis, physics and
geometry because they serve to describe the symmetry of analytical structures. They
were introduced by Sophus Lie in 1870 in order to study symmetries of differential
equations.
While the Euclidean space Rn is a real Lie group (with ordinary vector addition
as the group operation), more typical examples are given by matrix Lie groups, i.e.,
groups of invertible matrices (under matrix multiplication). For instance, the group
SO(3) of all rotations in R3 is a matrix Lie group.
One classifies Lie groups regarding their algebraic properties 7 (simple, semisim-
ple, solvable, nilpotent, Abelian), their connectedness (connected or simply con-
nected) and their compactness.8
To every Lie group, we can associate a Lie algebra which completely captures
the local structure of the group (at least if the Lie group is connected).9
7
If G and H are Lie groups (both real or both complex), then a Lie–group–homomorphism
f : G → H is a group homomorphism which is also an analytic map (one can show that it is
equivalent to require only that f be continuous). The composition of two such homomorphisms
is again a homomorphism, and the class of all (real or complex) Lie groups, together with these
morphisms, forms a category. The two Lie groups are called isomorphic iff there exists a bijective
homomorphism between them whose inverse is also a homomorphism. Isomorphic Lie groups do
not need to be distinguished for all practical purposes; they only differ in the notation of their
elements.
8
An n−torus T n = S 1 × S 1 × · · · × S 1 (as defined above) is an example of a compact Abelian
Lie group. This follows from the fact that the unit circle S 1 is a compact Abelian Lie group (when
identified with the unit complex numbers with multiplication). Group multiplication on T n is then
defined by coordinate–wise multiplication.
Toroidal groups play an important part in the theory of compact Lie groups. This is due in part
to the fact that in any compact Lie group one can always find a maximal torus; that is, a closed
subgroup which is a torus of the largest possible dimension.
9
Conventionally, one can regard any field X of tangent vectors on a Lie group as a partial
differential operator, denoting by Xf the Lie derivative (the directional derivative) of the scalar
field f in the direction of X. Then a vector–field on a Lie group G is said to be left–invariant if it
commutes with left translation, which means the following. Define Lg [f ](x) = f (gx) for any analytic
function f : G → R and all g, x ∈ G. Then the vector–field X is left–invariant iff XLg = Lg X for
all g ∈ G. Similarly, instead of R, we can use C. The set of all vector–fields on an analytic manifold
is a Lie algebra over R (or C).
On a Lie group G, the left–invariant vector–fields form a subalgebra, the Lie algebra g associated
with G. This Lie algebra is finite–dimensional (it has the same dimension as the manifold G) which
makes it susceptible to classification attempts. By classifying g, one can also get a handle on the
group G. The representation theory of simple Lie groups is the best and most important example.
Every element v of the tangent space Te at the identity element e of G determines a unique

20
4.1 Definition of a Lie Group
A Lie group is a smooth (Banach) manifold M that has at the same time a group
G−structure consistent with its manifold M −structure in the sense that group mul-
tiplication µ : G × G → G, (g, h) 7→ gh and the group inversion ν : G → G, g 7→
g−1 are C k −maps. A point e ∈ G is called the group identity element (see e.g.,
[21, 22, 1, 3]).
For example, any nD Banach vector space V is an Abelian Lie group with group
operations µ : V × V → V , µ(x, y) = x + y, and ν : V → V , ν(x) = −x. The identity
is just the zero vector. We call such a Lie group a vector group.
Let G and H be two Lie groups. A map G → H is said to be a morphism of
Lie groups (or their smooth homomorphism) if it is their homomorphism as abstract
groups and their smooth map as manifolds.
Similarly, a group G which is at the same time a topological space is said to be
a topological group if both maps (µ, ν) are continuous, i.e., C 0 −maps for it. The
homomorphism G → H of topological groups is said to be continuous if it is a
continuous map.
A topological group (as well as a smooth manifold) is not necessarily Hausdorff.
left–invariant vector–field whose value at the element g of G is denoted by gv; the vector space
underlying the Lie algebra g may therefore be identified with Te .
Every vector–field v in the Lie algebra g determines a function c : R → G whose derivative
everywhere is given by the corresponding left–invariant vector–field: c′ (t) = T Lc(t) v and which
has the property: c(s + t) = c(s)c(t), (for all s and t) (the operation on the r.h.s. is the group
multiplication in G). The formal similarity of this formula with the one valid for the elementary
exponential function justifies the definition: exp(v) = c(1). This is called the exponential map,
and it maps the Lie algebra g into the Lie group G. It provides a diffeomorphism between a
neighborhood of 0 in g and a neighborhood of e in G. This exponential map is a generalization of
the exponential function for real numbers (since R is the Lie algebra of the Lie group of positive
real numbers with multiplication), for complex numbers (since C is the Lie algebra of the Lie
group of non–zero complex numbers with multiplication) and for matrices (since M (n, R) with the
regular commutator is the Lie algebra of the Lie group GL(n, R) of all invertible matrices). As the
exponential map is surjective on some neighborhood N of e, it is common to call elements of the
Lie algebra infinitesimal generators of the group G.
The exponential map and the Lie algebra determine the local group structure of every connected
Lie group, because of the Baker–Campbell–Hausdorff formula: there exists a neighborhood U of
the zero element of the Lie algebra g, such that for u, v ∈ U we have

exp(u)exp(v) = exp(u + v + 1/2[u, v] + 1/12[[u, v], v] − 1/12[[u, v], u] − ...),

where the omitted terms are known and involve Lie brackets of four or more elements. In case u
and v commute, this formula reduces to the familiar exponential law :

exp(u)exp(v) = exp(u + v).

Every homomorphism f : G → H of Lie groups induces a homomorphism between the corre-

sponding Lie algebras g and h. The association G =⇒ g is called the Lie Functor.

21
A topological group G is Hausdorff iff its identity is closed. As a corollary we have
that every Lie group is a Hausdorff topological group.
For every g in a Lie group G, the two maps,

Lg : G → G, h 7→ gh,
Rh : G → G, g 7→ gh,

are called left and right translation maps. Since Lg ◦ Lh = Lgh , and Rg ◦ Rh = Rgh ,
it follows that (Lg )−1 = Lg−1 and (Rg )−1 = Rg−1 , so both Lg and Rg are diffeomor-
phisms. Moreover Lg ◦ Rh = Rh ◦ Lg , i.e., left and right translation commute.
A vector–field X on G is called left–invariant vector–field if for every g ∈ G,
L∗g X = X, that is, if (Th Lg )X(h) = X(gh) for all h ∈ G, i.e., the following diagram
commutes:
T Lg ✲ TG
TG
✻ ✻
X X

G ✲G
Lg
A Riemannian metric on a Lie group G is called left-invariant if it is preserved by
all left translations Lg , i.e., if the derivative of left translation carries every vector to
a vector of the same length. Similarly, a vector field X on G is called left–invariant
if (for every g ∈ G) L∗g X = X.

4.2 Lie Algebra

An algebra A is a vector space with a product. The product must have the property
that
a(uv) = (au)v = u(av),
for every a ∈ R and u, v ∈ A. A map φ : A → A′ between algebras is called an
algebra homomorphism if φ(u · v) = φ(u) · φ(v). A vector subspace I of an algebra
A is called a left ideal (resp. right ideal) if it is closed under algebra multiplication
and if u ∈ A and i ∈ I implies that ui ∈ I (resp. iu ∈ I). A subspace I is said
to be a two–sided ideal if it is both a left and right ideal. An ideal may not be an
algebra itself, but the quotient of an algebra by a two–sided ideal inherits an algebra
structure from A.
A Lie algebra is an algebra A where the multiplication, i.e., the Lie bracket
(u, v) 7→ [u, v], has the following properties:
LA 1. [u, u] = 0 for every u ∈ A, and
LA 2. [u, [v, w]] + [w, [u, v]] + [v, w, u]] = 0 for all u, v, w ∈ A.

22
 The condition LA 2 is usually called Jacobi identity. A subspace E ⊂ A of a Lie
algebra is called a Lie subalgebra if [u, v] ∈ E for every u, v ∈ E. A map φ : A → A′
between Lie algebras is called a Lie algebra homomorphism if φ([u, v]) = [φ(u), φ(v)]
for each u, v ∈ A.
All Lie algebras (over a given field K) and all smooth homomorphisms between
them form the category LAL, which is itself a complete subcategory of the category
AL of all algebras and their homomorphisms.
Let XL (G) denote the set of left–invariant vector–fields on G; it is a Lie subalge-
bra of X (G), the set of all vector–fields on G, since L∗g [X, Y ] = [L∗g X, L∗g Y ] = [X, Y ],
so the Lie bracket [X, Y ] ∈ XL (G).
Let e be the identity element of G. Then for each ξ on the tangent space Te G we
define a vector–field Xξ on G by Xξ (g) = Te Lg (ξ). XL (G) and Te G are isomorphic
as vector spaces. Define the Lie bracket on Te G by [ξ, η] = [Xξ , Xη ] (e) for all
ξ, η ∈ Te G. This makes Te G into a Lie algebra. Also, by construction, we have
[Xξ , Xη ] = X[ξ,η] ; this defines a bracket in Te G via left extension. The vector space
Te G with the above algebra structure is called the Lie algebra of the Lie group G
and is denoted g.
For example, let V be a nD vector space. Then Te V ≃ V and the left–invariant
vector–field defined by ξ ∈ Te V is the constant vector–field Xξ (η) = ξ, for all η ∈ V .
The Lie algebra of V is V itself.
Since any two elements of an Abelian Lie group G commute, it follows that all
adjoint operators Adg , g ∈ G, equal the identity. Therefore, the Lie algebra g is
Abelian; that is, [ξ, η] = 0 for all ξ, η ∈ g.
For example, G = SO(3) is the group of rotations of 3D Euclidean space, i.e.
the configuration space of a rigid body fixed at a point. A motion of the body is
then described by a curve g = g(t) in the group SO(3). Its Lie algebra g = so(3) is
the 3D vector space of angular velocities of all possible rotations. The commutator
in this algebra is the usual vector (cross) product (see, e.g. [1, 3, 9]).
A rotation velocity ġ of the rigid body (fixed at a point) is a tangent vector to
the Lie group G = SO(3) at the point g ∈ G. To get the angular velocity, we must
carry this vector to the tangent space T Ge of the group at the identity, i.e. to its Lie
algebra g = so(3). This can be done in two ways: by left and right translation, Lg
and Rg . As a result, we obtain two different vector fields in the Lie algebra so(3) :

ω c = Lg−1 ∗ ġ ∈ so(3) and ω x = Rg−1 ∗ ġ ∈ so(3),

which are called the ‘angular velocity in the body’ and the ‘angular velocity in space,’
respectively.
The dual space g∗ to the Lie algebra g = so(3) is the space of angular momenta
π. The kinetic energy T of a body is determined by the vector field of angular

23
velocity in the body and does not depend on the position of the body in space.
Therefore, kinetic energy gives a left-invariant Riemannian metric on the rotation
group G = SO(3).

4.3 One-Parameter Subgroup

Let Xξ be a left–invariant vector–field on G corresponding to ξ in g. Then there is
a unique integral curve γ ξ : R → G of Xξ starting at e, i.e., (see, e.g. [8, 9])


γ̇ ξ (t) = Xξ γ ξ (t) , γ ξ (0) = e

γ ξ (t) is a smooth one–parameter subgroup of G, i.e., γ ξ (t + s) = γ ξ (t) · γ ξ (s), since,

as functions of t
both sides equal γ ξ (s) at t = 0 and both satisfy differential equation
γ̇(t) = Xξ γ ξ (t) by left invariance of Xξ , so they are equal. Left invariance can be
also used to show that γ ξ (t) is defined for all t ∈ R. Moreover, if φ : R → G is a
one–parameter subgroup of G, i.e., a smooth homomorphism of the additive group
R into G, then φ = γ ξ with ξ = φ̇(0), since taking derivative at s = 0 in the relation

φ(t + s) = φ(t) · φ(s) gives φ̇(t) = Xφ̇(0) (φ(t)) ,

so φ = γ ξ since both equal e at t = 0. Therefore, all one–parameter subgroups of G

are of the form γ ξ (t) for some ξ ∈ g.

4.4 Exponential Map

The map exp : g → G, given by (see, e.g. [3, 8, 9]):

exp(ξ) = γ ξ (1), exp(0) = e

is called the exponential map of the Lie algebra g of G into G. exp is a C k −-

map, similar to the projection π of tangent and cotangent bundles; exp is locally a
diffeomorphism from a neighborhood of zero in g onto a neighborhood of e in G; if
f : G → H is a smooth homomorphism of Lie groups, then

f ◦ expG = expH ◦Te f .

Also, in this case

exp(sξ) = γ ξ (s).
Indeed, for fixed s ∈ R, the curve t 7→ γ ξ (ts), which at t = 0 passes through e,
satisfies the differential equation
d


γ ξ (ts) = sXξ γ ξ (ts) = Xsξ γ ξ (ts) .
dt

24
Since γ sξ (t) satisfies the same differential equation and passes through e at t = 0, it
follows that γ sξ (t) = γ ξ (st). Putting t = 1 induces exp(sξ) = γ ξ (s).
Hence exp maps the line sξ in g onto the one–parameter subgroup γ ξ (s) of G,
which is tangent to ξ at e. It follows from left invariance that the flow Ftξ of X
satisfies Ftξ (g) = g exp(sξ).
Globally, the exponential map exp is a natural operation, i.e., for any morphism
ϕ : G → H of Lie groups G and H and a Lie functor F, the following diagram
commutes:
F(ϕ) ✲
F(G) F(H)

exp exp
❄ ❄
G ✲H
ϕ
Let G1 and G2 be Lie groups with Lie algebras g1 and g2 . Then G1 × G2 is a
Lie group with Lie algebra g1 × g2 , and the exponential map is given by:

exp : g1 × g2 → G1 × G2 , (ξ 1 , ξ 2 ) 7→ (exp1 (ξ 1 ), exp2 (ξ 2 )) .

For example, in case of a nD vector space, or infinite–dimensional Banach space,

the exponential map is the identity.
The unit circle in the complex plane S 1 = {z ∈ C : |z| = 1} is an Abelian Lie
group under multiplication. The tangent space Te S 1 is the imaginary axis, and we
identify R with Te S 1 by t 7→ 2πit. With this identification, the exponential map
exp : R → S 1 is given by exp(t) = e2πit .
The nD torus T n = S 1 ×···×S 1 (n times) is an Abelian Lie group. The exponen-
tial map exp : Rn → T n is given by

exp(t1 , ..., tn ) = (e2πit1 , ..., e2πitn ).

Since S 1 = R/Z, it follows that T n = Rn /Zn , the projection Rn → T n being given

by the exp map.

4.5 Adjoint Representation

For every g ∈ G, the map (see, e.g. [1, 3, 8, 9]):


Adg = Te Rg−1 ◦ Lg : g → g

is called the adjoint map, or adjoint operator associated with g.

For each ξ ∈ g and g ∈ G we have

exp (Adg ξ) = g (exp ξ) g−1 .

25
 The relation between the adjoint map and the Lie bracket is the following: For
all ξ, η ∈ g we have
d
Ad η = [ξ, η].
dt t=0 exp(tξ)
Left and right translations induce operators on the cotangent space T ∗ Gg dual
to Lg∗ and Rg∗ , denoted by (for every h ∈ G):

L∗g : T ∗ Ggh → T ∗ Gh , Rg∗ : T ∗ Ghg → T ∗ Gh .

The transpose operators Ad∗g : g → g satisfy the relations Ad∗gh = Ad∗h Ad∗g (for every
g, h ∈ G) and constitute the co-adjoint representation of the Lie group G. The
co-adjoint representation plays an important role in all questions related to (left)
invariant metrics on the Lie group. According to A. Kirillov, the orbit of any vector
field X in a Lie algebra g in a co-adjoint representation Ad∗g is itself a symplectic
manifold and therefore a phase space for a Hamiltonian mechanical system.
A Lie subgroup H of G is a subgroup H of G which is also a submanifold of
G. Then h is a Lie subalgebra of g and moreover h = {ξ ∈ g| exp(tξ) ∈ H, for all
t ∈ R}.
One can characterize Lebesgue measure up to a multiplicative constant on Rn by
its invariance under translations. Similarly, on a locally compact group there is a
unique (up to a nonzero multiplicative constant) left–invariant measure, called Haar
measure. For Lie groups the existence of such measures is especially simple: Let G be
a Lie group. Then there is a volume form U b5, unique up to nonzero multiplicative
constants, that is left–invariant. If G is compact, U b5 is right invariant as well.

4.6 Actions of Lie Groups on Smooth Manifolds

Let M be a smooth manifold. An action of a Lie group G (with the unit element
e) on M is a smooth map φ : G × M → M, such that for all x ∈ M and g, h ∈ G,
(i) φ(e, x) = x and (ii) φ (g, φ(h, x)) = φ(gh, x). In other words, letting φg : x ∈
M 7→ φg (x) = φ(g, x) ∈ M , we have (i’) φe = idM and (ii’) φg ◦ φh = φgh . φg is a
diffeomorphism, since (φg )−1 = φg−1 . We say that the map g ∈ G 7→ φg ∈ Dif f (M )
is a homomorphism of G into the group of diffeomorphisms of M . In case that M
is a vector space and each φg is a linear operator, the function of G on M is called
a representation of G on M (see, e.g. [1, 3, 8, 9]).
An action φ of G on M is said to be transitive group action, if for every x, y ∈ M ,
there is g ∈ G such that φ(g, x) = y; effective group action, if φg = idM implies g = e,
that is g 7→ φg is 1–1; and free group action, if for each x ∈ M , g 7→ φg (x) is 1–1.
For example,

26
 1. G = R acts on M = R by translations; explicitly,

φ : G × M → M, φ(s, x) = x + s.

Then for x ∈ R, Ox = R. Hence M/G is a single point, and the action is

transitive and free.

2. A complete flow φt of a vector–field X on M gives an action of R on M , namely

(t, x) ∈ R × M 7→ φt (x) ∈ M.

3. Left translation Lg : G → G defines an effective action of G on itself. It is also

transitive.

4. The coadjoint action of G on g∗ is given by

∗
Ad∗ : (g, α) ∈ G × g∗ 7→ Ad∗g−1 (α) = Te (Rg−1 ◦ Lg ) α ∈ g∗ .

Let φ be an action of G on M . For x ∈ M the orbit of x is defined by

Ox = {φg (x)|g ∈ G} ⊂ M

and the isotropy group of φ at x is given by

Gx = {g ∈ G|φ(g, x) = x} ⊂ G.

An action φ of G on a manifold M defines an equivalence relation on M by the

relation belonging to the same orbit; explicitly, for x, y ∈ M , we write x ∼ y if there
exists a g ∈ G such that φ(g, x) = y, that is, if y ∈ Ox . The set of all orbits M/G is
called the group orbit space (see, e.g. [1, 3, 8, 9]).
For example, let M = R2 \{0}, G = SO(2), the group of rotations in plane, and
the action of G on M given by
  
cos θ − sin θ
, (x, y) 7−→ (x cos θ − y sin θ, x sin θ + y cos θ).
sin θ cos θ

The action is always free and effective, and the orbits are concentric circles, thus the
orbit space is M/G ≃ R∗+ .
A crucial concept in mechanics is the infinitesimal description of an action. Let
φ : G × M → M be an action of a Lie group G on a smooth manifold M . For each
ξ ∈ g,
φξ : R × M → M, φξ (t, x) = φ (exp(tξ), x)

27
is an R−-action on M . Therefore, φexp(tξ) : M → M is a flow on M ; the correspond-
ing vector–field on M , given by

d
ξ M (x) = φexp(tξ) (x)
dt t=0

is called the infinitesimal generator of the action, corresponding to ξ in g.

The tangent space at x to an orbit Ox is given by

Tx Ox = {ξ M (x)|ξ ∈ g}.

Let φ : G × M → M be a smooth G − −action. For all g ∈ G, all ξ, η ∈ g and

all α, β ∈ R, we have:
(Adg ξ)M = φ∗g−1 ξ M , [ξ M , η M ] = − [ξ, η]M , and (αξ + βη)M = αξ M + βη M .
Let M be a smooth manifold, G a Lie group and φ : G × M → M a G−action
on M . We say that a smooth map f : M → M is with respect to this action if for
all g ∈ G,
f ◦ φg = φg ◦ f .
Let f : M → M be an equivariant smooth map. Then for any ξ ∈ g we have

T f ◦ ξ M = ξ M ◦ f.

4.7 Basic Tables of Lie Groups and Their Lie Algebras

One classifies Lie groups regarding their algebraic properties (simple, semisimple,
solvable, nilpotent, Abelian), their connectedness (connected or simply connected)
and their compactness (see Tables A.1–A.3). This is the content of the Hilbert 5th
problem.

28
Some real Lie groups and their Lie algebras:

Lie Description Remarks Lie Description dim

group algb. /R
Rn Euclidean space Abelian, simply Rn the Lie bracket n
with addition connected, not is zero
compact
R× nonzero real Abelian, not R the Lie bracket 1
numbers with connected, not is zero
multiplication compact
R>0 positive real Abelian, simply R the Lie bracket 1
numbers with connected, not is zero
multiplication compact
S1 = complex num- Abelian, con- R the Lie bracket 1
R/Z bers of absolute nected, not sim- is zero
value 1, with ply connected,
multiplication compact
H× non–zero simply con- H quaternions, 4
quaternions nected, not with Lie bracket
with multiplica- compact the commutator
tion
S3 quaternions of simply con- R3 real 3−vectors, 3
absolute value nected, com- with Lie bracket
1, with mul- pact, simple the cross prod-
tiplication; a and semi– uct; isomorphic
3−sphere simple, isomor- to su(2) and to
phic to SU (2), so(3)
SO(3) and to
Spin(3)
GL(n, R) general linear not connected, M(n, R) n−by-n matri- n2
group: invert- not compact ces, with Lie
ible n−by-n bracket the
real matrices commutator
GL+ (n, R) n−by-n real simply con- M(n, R) n−by-n matri- n2
matrices with nected, not ces, with Lie
positive deter- compact bracket the
minant commutator

29
Classical real Lie groups and their Lie algebras:

Lie Description Remarks Lie Description dim

group algb. /R
SL(n, R) special linear simply con- sl(n, R) square matrices n2 −
group: real nected, not with trace 0, 1
matrices with compact if with Lie bracket
determinant 1 n>1 the commutator
O(n, R) orthogonal not connected, so(n, R) skew– n(n−
group: real compact symmetric 1)/2
orthogonal square real
matrices matrices, with
Lie bracket
the commuta-
tor; so(3, R) is
isomorphic to
su(2) and to R3
with the cross
product
SO(n, R) special orthogo- connected, com- so(n, R) skew– n(n−
nal group: real pact, for n ≥ 2: symmetric 1)/2
orthogonal ma- not simply con- square real
trices with de- nected, for n = matrices, with
terminant 1 3 and n ≥ Lie bracket the
5: simple and commutator
semisimple
Spin(n) spinor group simply con- so(n, R) skew– n(n−
nected, com- symmetric 1)/2
pact, for n = 3 square real
and n ≥ 5: matrices, with
simple and Lie bracket the
semisimple commutator
U (n) unitary group: isomorphic to u(n) square com- n2
complex unitary S 1 for n = 1, plex matrices
n−by-n matri- not simply A satisfying
ces connected, A = −A∗ , with
compact Lie bracket the
commutator
SU (n) special unitary simply con- su(n) square complex n2 −
group: complex nected, com- matrices A with 1
unitary n−by-n pact, for n ≥ 2: trace 0 satisfy-
matrices with simple and ing A = −A∗ ,
determinant 1 semisimple with Lie bracket
the commutator

30
Basic complex Lie groups and their Lie algebras:10

Lie Description Remarks Lie Description dim

group algb. /C
Cn group operation Abelian, simply Cn the Lie bracket n
is addition connected, not is zero
compact
C× nonzero com- Abelian, not C the Lie bracket 1
plex numbers simply con- is zero
with multiplica- nected, not
tion compact
GL(n, C) general linear simply con- M (n, C) n−by-n matri- n2
group: invert- nected, not ces, with Lie
ible n−by-n compact, for bracket the
complex matri- n = 1: iso- commutator
ces morphic to
C×
SL(n, C) special linear simple, sl(n, C) square matrices n2 −
group: complex semisimple, with trace 0, 1
matrices with simply con- with Lie bracket
determinant 1 nected, for the commutator
n ≥ 2: not
compact
O(n, C) orthogonal not connected, so(n, C) skew– n(n−
group: com- for n ≥ 2: not symmetric 1)/2
plex orthogonal compact square complex
matrices matrices, with
Lie bracket the
commutator
SO(n, C) special orthogo- for n ≥ 2: so(n, C) skew– n(n−
nal group: com- not compact, symmetric 1)/2
plex orthogonal not simply square complex
matrices with connected, for matrices, with
determinant 1 n = 3 and Lie bracket the
n ≥ 5: simple commutator
and semisimple

4.8 Representations of Lie groups

The idea of a representation of a Lie group plays an important role in the study of
continuous symmetry (see, e.g., [22]). A great deal is known about such representa-
tions, a basic tool in their study being the use of the corresponding ’infinitesimal’
representations of Lie algebras.
Formally, a representation of a Lie group G on a vector space V (over a field K)
10
The dimensions given are dimensions over C. Note that every complex Lie group/algebra can
also be viewed as a real Lie group/algebra of twice the dimension.

31
is a group homomorphism G → Aut(V ) from G to the automorphism group of V .
If a basis for the vector space V is chosen, the representation can be expressed as a
homomorphism into GL(n, K). This is known as a matrix representation.
On the Lie algebra level, there is a corresponding linear map from the Lie algebra
of G to End(V ) preserving the Lie bracket [·, ·].
If the homomorphism is in fact an monomorphism, the representation is said to
be faithful.
A unitary representation is defined in the same way, except that G maps to
unitary matrices; the Lie algebra will then map to skew–Hermitian matrices.
Now, if G is a semisimple group, its finite–dimensional representations can be
decomposed as direct sums of irreducible representations. The irreducibles are in-
dexed by highest weight; the allowable (dominant) highest weights satisfy a suitable
positivity condition. In particular, there exists a set of fundamental weights, indexed
by the vertices of the Dynkin diagram of G (see below), such that dominant weights
are simply non–negative integer linear combinations of the fundamental weights.
If G is a commutative compact Lie group, then its irreducible representations
are simply the continuous characters of G. A quotient representation is a quotient
module of the group ring.

4.9 Root Systems and Dynkin Diagrams

A root system is a special configuration in Euclidean space that has turned out to
be fundamental in Lie theory as well as in its applications. Also, the classification
scheme for root systems, by Dynkin diagrams, occurs in parts of mathematics with
no overt connection to Lie groups (such as singularity theory, see e.g., [22]).

4.9.1 Definitions
Formally, a root system is a finite set Φ of non–zero vectors (roots) spanning a
finite–dimensional Euclidean space V and satisfying the following properties:

1. The only scalar multiples of a root α in V which belong to Φ are α itself and
-α.

2. For every root α in V , the set Φ is symmetric under reflection through the
hyperplane of vectors perpendicular to α.

3. If α and β are vectors in Φ, the projection of 2β onto the line through α is an

integer multiple of α.

The rank of a root system Φ is the dimension of V . Two root systems may
be combined by regarding the Euclidean spaces they span as mutually orthogonal

32
subspaces of a common Euclidean space. A root system which does not arise from
such a combination, such as the systems A2 , B2 , and G2 in Figure 3, is said to be
irreducible.
Two irreducible root systems (V1 , Φ1 ) and (V2 , Φ2 ) are considered to be the same
if there is an invertible linear transformation V1 → V2 which preserves distance up
to a scale factor and which sends Φ1 to Φ2 .
The group of isometries of V generated by reflections through hyperplanes as-
sociated to the roots of Φ is called the Weyl group of Φ as it acts faithfully on the
finite set Φ, the Weyl group is always finite.

4.9.2 Classification
It is not too difficult to classify the root systems of rank 2 (see Figure 3).

Figure 3: Classification of root systems of rank 2.

Whenever Φ is a root system in V and W is a subspace of V spanned by Ψ =

Φ ∩ W , then Ψ is a root system in W . Thus, our exhaustive list of root systems
of rank 2 shows the geometric possibilities for any two roots in a root system. In
particular, two such roots meet at an angle of 0, 30, 45, 60, 90, 120, 135, 150, or 180
degrees.
In general, irreducible root systems are specified by a family (indicated by a
letter A to G) and the rank (indicated by a subscript n). There are four infinite
families:

• An (n ≥ 1), which corresponds to the special unitary group, SU (n + 1);

• Bn (n ≥ 2), which corresponds to the special orthogonal group, SO(2n + 1);

• Cn (n ≥ 3), which corresponds to the symplectic group, Sp(2n);

• Dn (n ≥ 4), which corresponds to the special orthogonal group, SO(2n),

as well as five exceptional cases: E6 , E7 , E8 , F4 , G2 .

33
4.9.3 Dynkin Diagrams
A Dynkin diagram is a graph with a few different kinds of possible edges (see Figure
4). The connected components of the graph correspond to the irreducible subalge-
bras of g. So a simple Lie algebra’s Dynkin diagram has only one component. The
rules are restrictive. In fact, there are only certain possibilities for each component,
corresponding to the classification of semi–simple Lie algebras (see, e.g., [26]).

Figure 4: The problem of classifying irreducible root systems reduces to the problem
of classifying connected Dynkin diagrams.

The roots of a complex Lie algebra form a lattice of rank k in a Cartan subalgebra
h ⊂ g, where k is the Lie algebra rank of g. Hence, the root lattice can be considered
a lattice in Rk . A vertex, or node, in the Dynkin diagram is drawn for each Lie
algebra simple root, which corresponds to a generator of the root lattice. Between
two nodes α and β, an edge is drawn if the simple roots are not perpendicular. One
line is drawn if the angle between them is 2π/3, two lines if the angle is 3π/4, and
three lines are drawn if the angle is 5π/6. There are no other possible angles between
Lie algebra simple roots. Alternatively, the number of lines N between the simple
roots α and β is given by
2 hα, βi 2 hβ, αi
N = Aαβ Aβα = = 4 cos2 θ,
|α|2 |β|2

where Aαβ = 2hα,βi

|α|2
is an entry in the Cartan matrix (Aαβ ) (for details on Cartan
matrix see, e.g., [22]). In a Dynkin diagram, an arrow is drawn from the longer root
to the shorter root (when the angle is 3π/4 or 5π/6).
Here are some properties of admissible Dynkin diagrams:

1. A diagram obtained by removing a node from an admissible diagram is admis-

sible.
2. An admissible diagram has no loops.
3. No node has more than three lines attached to it.

34
 4. A sequence of nodes with only two single lines can be collapsed to give an
admissible diagram.

5. The only connected diagram with a triple line has two nodes.

A Coxeter–Dynkin diagram, also called a Coxeter graph, is the same as a Dynkin

diagram, but without the arrows. The Coxeter diagram is sufficient to characterize
the algebra, as can be seen by enumerating connected diagrams.
The simplest way to recover a simple Lie algebra from its Dynkin diagram is to
first reconstruct its Cartan matrix (Aij ). The ith node and jth node are connected
by Aij Aji lines. Since Aij = 0 iff Aji = 0, and otherwise Aij ∈ {−3, −2, −1}, it is
easy to find Aij and Aji , up to order, from their product. The arrow in the diagram
indicates which is larger. For example, if node 1 and node 2 have two lines between
them, from node 1 to node 2, then A12 = −1 and A21 = −2.
However, it is worth pointing out that each simple Lie algebra can be constructed
concretely. For instance, the infinite families An , Bn , Cn , and Dn correspond to
the special linear Lie algebra gl(n + 1, C), the odd orthogonal Lie algebra so(2n +
1, C), the symplectic Lie algebra sp(2n, C), and the even orthogonal Lie algebra
so(2n, C). The other simple Lie algebras are called exceptional Lie algebras, and
have constructions related to the octonions.
To prove this classification Theorem, one uses the angles between pairs of roots to
encode the root system in a much simpler combinatorial object, the Dynkin diagram.
The Dynkin diagrams can then be classified according to the scheme given above.
To every root system is associated a corresponding Dynkin diagram. Otherwise,
the Dynkin diagram can be extracted from the root system by choosing a base, that
is a subset ∆ of Φ which is a basis of V with the special property that every vector
in Φ when written in the basis ∆ has either all coefficients ≥ 0 or else all ≤ 0.
The vertices of the Dynkin diagram correspond to vectors in ∆. An edge is
drawn between each non–orthogonal pair of vectors; it is a double edge if they make
an angle of 135 degrees, and a triple edge if they make an angle of 150 degrees. In
addition, double and triple edges are marked with an angle sign pointing toward the
shorter vector.
Although a given root system has more than one base, the Weyl group acts
transitively on the set of bases. Therefore, the root system determines the Dynkin
diagram. Given two root systems with the same Dynkin diagram, we can match up
roots, starting with the roots in the base, and show that the systems are in fact the
same.
Thus the problem of classifying root systems reduces to the problem of classifying
possible Dynkin diagrams, and the problem of classifying irreducible root systems
reduces to the problem of classifying connected Dynkin diagrams. Dynkin diagrams

35
encode the inner product on E in terms of the basis ∆, and the condition that this
inner product must be positive definite turns out to be all that is needed to get the
desired classification (see Figure 4).
In detail, the individual root systems can be realized case–by–case, as in the
following paragraphs:
An . Let V be the subspace of Rn+1 √ for which the coordinates sum to 0, and
let Φ be the set of vectors in V of length 2 and with integer coordinates in Rn+1 .
Such a vector must have all but two coordinates equal to 0, one coordinate equal to
1, and one equal to -1, so there are n2 + n roots in all. √
Bn . Let V = Rn , and let Φ consist of all integer vectors in V of length 1 or 2.
The total number of roots is 2n2 . √
Cn : Let V = Rn , and let Φ consist of all integer vectors in V of 2 together
with all vectors of the form 2λ, where λ is an integer vector of length 1. The total
number of roots is 2n2 . The total number of roots is 2n2 . √
Dn . Let V = Rn , and let Φ consist of all integer vectors in V of length 2. The
total number of roots is 2n(n − 1). √
En . For V8 , let V = R8 , and let E8 denote the set of vectors α of length 2
such that the coordinates of 2α are all integers and are either all even or all odd.
Then E7 can be constructed as the intersection of E8 with the hyperplane of vectors
perpendicular to a fixed root α in E8 , and E6 can be constructed as the intersection
of E8 with two such hyperplanes corresponding to roots α and β which are neither
orthogonal to one another nor scalar multiples of one another. The root systems
E6 , E7 , and E8 have 72, 126, and 240 roots respectively. √
F4 . For F4 , let V = R4 , and let Φ denote the set of vectors α of length 1 or 2
such that the coordinates of 2α are all integers and are either all even or all odd.
There are 48 roots in this system.
G2 . There are 12 roots in G2 , which form the vertices of a hexagram.

4.9.4 Irreducible Root Systems

Irreducible root systems classify a number of related objects in Lie theory, notably:

1. Simple complex Lie algebras;

2. Simple complex Lie groups;

3. Simply connected complex Lie groups which are simple modulo centers; and

4. Simple compact Lie groups.

In each case, the roots are non–zero weights of the adjoint representation.
A root system can also be said to describe a plant’s root and associated systems.

36
4.10 Simple and Semisimple Lie Groups and Algebras
A simple Lie group is a Lie group which is also a simple group. These groups, and
groups closely related to them, include many of the so–called classical groups of
geometry, which lie behind projective geometry and other geometries derived from
it by the Erlangen programme of Felix Klein. They also include some exceptional
groups, that were first discovered by those pursuing the classification of simple Lie
groups. The exceptional groups account for many special examples and configu-
rations in other branches of mathematics. In particular the classification of finite
simple groups depended on a thorough prior knowledge of the ‘exceptional’ possi-
bilities.
The complete listing of the simple Lie groups is the basis for the theory of the
semisimple Lie groups and reductive groups, and their representation theory. This
has turned out not only to be a major extension of the theory of compact Lie groups
(and their representation theory), but to be of basic significance in mathematical
physics.
Such groups are classified using the prior classification of the complex simple
Lie algebras. It has been shown that a simple Lie group has a simple Lie algebra
that will occur on the list given there, once it is complexified (that is, made into a
complex vector space rather than a real one). This reduces the classification to two
further matters.
The groups SO(p, q, R) and SO(p + q, R), for example, give rise to different real
Lie algebras, but having the same Dynkin diagram. In general there may be different
real forms of the same complex Lie algebra.
Secondly, the Lie algebra only determines uniquely the simply connected (uni-
versal) cover G∗ of the component containing the identity of a Lie group G. It may
well happen that G∗ is not actually a simple group, for example having a non–trivial
center. We have therefore to worry about the global topology, by computing the
fundamental group of G (an Abelian group: a Lie group is an H−space). This was
done by Elie Cartan.
For an example, take the special orthogonal groups in even dimension. With
−I a scalar matrix in the center, these are not actually simple groups; and having
a two–fold spin cover, they aren’t simply–connected either. They lie ‘between’ G∗
and G, in the notation above.
Recall that a semisimple module is a module in which each submodule is a direct
summand. In particular, a semisimple representation is completely reducible, i.e.,
is a direct sum of irreducible representations (under a descending chain condition).
Similarly, one speaks of an Abelian category as being semisimple when every object
has the corresponding property. Also, a semisimple ring is one that is semisimple as
a module over itself.

37
 A semisimple matrix is diagonalizable over any algebraically closed field contain-
ing its entries. In practice this means that it has a diagonal matrix as its Jordan
normal form.
A Lie algebra g is called semisimple when it is a direct sum of simple Lie algebras,
i.e., non–trivial Lie algebras L whose only ideals are {0} and L itself. An equivalent
condition is that the Killing form
B(X, Y ) = Tr(Ad(X) Ad(Y ))
is non–degenerate [27]. The following properties can be proved equivalent for a
finite–dimensional algebra L over a field of characteristic 0:
1. L is semisimple.
2. L has no nonzero Abelian ideal.
3. L has zero radical (the radical is the biggest solvable ideal).
4. Every representation of L is fully reducible, i.e., is a sum of irreducible repre-
sentations.
5. L is a (finite) direct product of simple Lie algebras (a Lie algebra is called
simple if it is not Abelian and has no nonzero ideal ).
A connected Lie group is called semisimple when its Lie algebra is semisimple;
and the same holds for algebraic groups. Every finite dimensional representation
of a semisimple Lie algebra, Lie group, or algebraic group in characteristic 0 is
semisimple, i.e., completely reducible, but the converse is not true. Moreover, in
characteristic p > 0, semisimple Lie groups and Lie algebras have finite dimensional
representations which are not semisimple. An element of a semisimple Lie group or
Lie algebra is itself semisimple if its image in every finite–dimensional representation
is semisimple in the sense of matrices.
Every semisimple Lie algebra g can be classified by its Dynkin diagram [22].

5 Some Classical Examples of Lie Groups

5.1 Galilei Group
The Galilei group is the group of transformations in space and time that connect
those Cartesian systems that are termed ‘inertial frames’ in Newtonian mechanics.
The most general relationship between two such frames is the following. The origin
of the time scale in the inertial frame S ′ may be shifted compared with that in S;
the orientation of the Cartesian axes in S ′ may be different from that in S; the
origin O of the Cartesian frame in S ′ may be moving relative to the origin O in S
at a uniform velocity. The transition from S to S ′ involves ten parameters; thus the
Galilei group is a ten parameter group. The basic assumption inherent in Galilei–
Newtonian relativity is that there is an absolute time scale, so that the only way

38
in which the time variables used by two different ‘inertial observers’ could possibly
differ is that the zero of time for one of them may be shifted relative to the zero of
time for the other (see, e.g. [1, 8, 9]).
Galilei space–time structure involves the following three elements:

1. World, as a 4D affine space A4 . The points of A4 are called world points or

events. The parallel transitions of the world A4 form a linear (i.e., Euclidean)
space R4 .

2. Time, as a linear map t : R4 → R of the linear space of the world parallel

transitions onto the real ‘time axes’. Time interval from the event a ∈ A4 to
b ∈ A4 is called the number t(b − a); if t(b − a) = 0 then the events a and b are
called synchronous. The set of all mutually synchronous events consists a 3D
affine space A3 , being a subspace of the world A4 . The kernel of the mapping
t consists of the parallel transitions of A4 translating arbitrary (and every)
event to the synchronous one; it is a linear 3D subspace R3 of the space R4 .

3. Distance (metric) between the synchronous events,

ρ(a, b) =k a − b k, for all a, b ∈ A3 ,

given by the scalar product in R3 . The distance transforms arbitrary space of

synchronous events into the well known 3D Euclidean space E 3 .

The space A4 , with the Galilei space–time structure on it, is called Galilei space.
Galilei group is the group of all possible transformations of the Galilei space, pre-
serving its structure. The elements of the Galilei group are called Galilei transfor-
mations. Therefore, Galilei transformations are affine transformations of the world
A4 preserving the time intervals and distances between the synchronous events.
The direct product R × R3 , of the time axes with the 3D linear space R3 with
a fixed Euclidean structure, has a natural Galilei structure. It is called Galilei
coordinate system.

5.2 General Linear Group

The group of linear isomorphisms of Rn to Rn is a Lie group of dimension n2 , called
the general linear group and denoted Gl(n, R). It is a smooth manifold, since it is
a subset of the vector space L(Rn , Rn ) of all linear maps of Rn to Rn , as Gl(n, R)
is the inverse image of R\{0} under the continuous map A 7→ det A of L(Rn , Rn ) to
R. The group operation is composition (see, e.g. [1, 3, 8, 9]).

(A, B) ∈ Gl(n, R) × Gl(n, R) 7→ A ◦ B ∈ Gl(n, R)

39
and the inverse map is

A ∈ Gl(n, R) 7→ A−1 ∈ Gl(n, R).

If we choose a basis in Rn , we can represent each element A ∈ Gl(n, R) by an

invertible (n × n)−-matrix. The group operation is then matrix multiplication and
the inversion is matrix inversion. The identity is the identity matrix In . The group
operations are smooth since the formulas for the product and inverse of matrices are
smooth in the matrix components.
The Lie algebra of Gl(n, R) is gl(n), the vector space L(Rn , Rn ) of all linear
transformations of Rn , with the commutator bracket

[A, B] = AB − BA.

For every A ∈ L(Rn , Rn ),

∞ i
X t
γ A : t ∈ R 7→γ A (t) = Ai ∈ Gl(n, R)
i!
i=0

is a one–parameter subgroup of Gl(n, R), because

∞
X ti−1
γ A (0) = I and γ̇ A (t) = Ai = γ A (t) A
(i − 1)!
i=0

Hence γ A is an integral curve of the left–invariant vector–field XA . Therefore, the

exponential map is given by
∞
n n A
X Ai
exp : A ∈ L(R , R ) 7→ exp(A) ≡ e = γ A (1) = ∈ Gl(n, R).
i!
i=0

For each A ∈ Gl(n, R) the corresponding adjoint map

AdA : L(Rn , Rn ) → L(Rn , Rn )

is given by
AdA B = A · B · A−1 .

5.3 Rotational Lie Groups in Human/Humanoid Biomechanics

Local kinematics at each rotational robot or (synovial) human joint, is defined as
a group action of an nD constrained rotational Lie group SO(n) on the Euclidean
space Rn . In particular, there is an action of SO(2)−-group in uniaxial human joints

40
(cylindrical, or hinge joints, like knee and elbow) and an action of SO(3)−-group
in three–axial human joints (spherical, or ball–and–socket joints, like hip, shoulder,
neck, wrist and ankle). In both cases, SO(n) acts, with its operators of rotation,
on the vector x = {xµ }, (i = 1, 2, 3) of external, Cartesian coordinates of the parent
body–segment, depending, at the same time, on the vector q = {q s }, (s = 1, · · · , n)
on n group–parameters, i.e., joint angles (see [5, 6, 8, 9]).
Each joint rotation R ∈ SO(n) defines a map

R : xµ 7→ ẋµ , R(xµ , q s ) = Rqs xµ ,

where Rqs ∈ SO(n) are joint group operators. The vector v = {vs }, (s = 1, · · · , n)
of n infinitesimal generators of these rotations, i.e., joint angular velocities, given by
∂R(xµ , q s )
 
∂
vs = −
∂q s q=0 ∂x µ

constitute an nD Lie algebra so(n) corresponding to the joint rotation group SO(n).
Conversely, each joint group operator Rqs , representing a one–parameter subgroup of
SO(n), is defined as the exponential map of the corresponding joint group generator
vs
Rqs = exp(q s vs )
This exponential map represents a solution of the joint operator differential equation
in the joint group–parameter space {q s }
dRqs
= vs Rq s .
dq s

5.3.1 Uniaxial Group of Joint Rotations

The uniaxial joint rotation in a single Cartesian plane around a perpendicular axis,
e.g., xy−plane about the z axis, by an internal joint angle θ, leads to the following
transformation of the joint coordinates:

x′ = x cos θ − y sin θ, y ′ = x sin θ + y cos θ.

In this way, the joint SO(2)−group, given by

   
cos θ − sin θ
SO(2) = Rθ = |θ ∈ [0, 2π] ,
sin θ cos θ

acts in a canonical way on the Euclidean plane R2 by

     
cos θ − sin θ x x cos θ −y sin θ
SO(2) = , 7 →
− .
sin θ cos θ y x sin θ y cos θ

41
Its associated Lie algebra so(2) is given by
  
0 −t
so(2) = |t ∈ R ,
t 0

since the curve γ θ ∈ SO(2) given by

 
cos tθ − sin tθ
γ θ : t ∈ R 7−→ γ θ (t) = ∈ SO(2),
sin tθ cos tθ
 
1 0
passes through the identity I2 = and then
0 1
 
d 0 −θ
γ θ (t) = ,
dt t=0 θ 0

so that I2 is a basis of so(2), since dim (SO(2)) = 1.

The exponential map exp : so(2) → SO(2) is given by
   
0 −θ cos tθ − sin tθ
exp = γ θ (1) = .
θ 0 sin tθ cos tθ

The infinitesimal generator of the action of SO(2) on R2 , i.e., joint angular

velocity v, is given by
∂ ∂
v = −y +x ,
∂x ∂y
since
  
d d cos tv − sin tv x
vR2 (x, y) = exp(tv) (x, y) = .
dt t=0 dt t=0 sin tv cos tv y

The momentum map J : T ∗ R2 → R associated to the lifted action of SO(2) on

T ∗ R2 ≃ R4 is given by

J (x, y, p1 , p2 ) = xpy − ypx , since

J (x, y, px , py ) (ξ) = (px dx + py dy)(vR2 ) = −vpx y + −vpy x.

The Lie group SO(2) acts on the symplectic manifold (R4 , ω = dpx ∧dx+dpy ∧dx)
by
  
cos θ − sin θ
φ , (x, y, px , py )
sin θ cos θ
= (x cos θ − y sin θ, x sin θ + y cos θ, px cos θ − py sin θ, px sin θ + py cos θ) .

42
5.3.2 Three–Axial Group of Joint Rotations
The three–axial SO(3)−group of human–like joint rotations depends on three pa-
rameters, Euler joint angles q i = (ϕ, ψ, θ), defining the rotations about the Cartesian
coordinate triedar (x, y, z) placed at the joint pivot point. Each of the Euler angles
are defined in the constrained range (−π, π), so the joint group space is a constrained
sphere of radius π (see [5, 6, 8, 9]).
Let G = SO(3) = {A ∈ M3×3 (R) : At A = I3 , det(A) = 1} be the group
of rotations in R3 . It is a Lie group and dim(G) = 3. Let us isolate its one–
parameter joint subgroups, i.e., consider the three operators of the finite joint rota-
tions Rϕ , Rψ , Rθ ∈ SO(3), given by
     
1 0 0 cos ψ 0 sin ψ cos θ − sin θ 0
Rϕ =  0 cos ϕ − sin ϕ  , Rψ =  0 1 0  , Rθ =  sin θ cos θ 0 
0 sin ϕ cos ϕ − sin ψ 0 cos ψ 0 0 1

corresponding respectively to rotations about x−axis by an angle ϕ, about y−axis

by an angle ψ, and about z−axis by an angle θ.
The total three–axial joint rotation A is defined as the product of above one–
parameter rotations Rϕ , Rψ , Rθ , i.e., A = Rϕ · Rψ · Rθ is equal11
 
cos ψ cos ϕ − cos θ sin ϕ sin ψ cos ψ cos ϕ + cos θ cos ϕ sin ψ sin θ sin ψ
A =  − sin ψ cos ϕ − cos θ sin ϕ sin ψ − sin ψ sin ϕ + cos θ cos ϕ cos ψ sin θ cos ψ  .
sin θ sin ϕ − sin θ cos ϕ cos θ

However, the order of these matrix products matters: different order products give
different results, as the matrix product is noncommutative product. This is the
reason why Hamilton’s quaternions 12 are today commonly used to parameterize the
SO(3)−group, especially in the field of 3D computer graphics.
The one–parameter rotations Rϕ , Rψ , Rθ define curves in SO(3) starting from
1 0 0
!
I3 = 0 1 0 . Their derivatives in ϕ = 0, ψ = 0 and θ = 0 belong to the asso-
0 0 1
ciated tangent Lie algebra so(3). That is the corresponding infinitesimal generators
of joint rotations – joint angular velocities vϕ , vψ , vθ ∈ so(3) – are respectively given
11
Note that this product is noncommutative, so it really depends on the order of multiplications.
12
Recall that the set of Hamilton’s quaternions H represents an extension of the set of complex
numbers C. We can compute a rotation about the unit vector, u by an angle θ. The quaternion q
that computes this rotation is  
θ θ
q = cos , u sin .
2 2

43
by

0 0 0 0 0 1
" # " #
∂ ∂ ∂ ∂
vϕ = 0 0 −1 = −y +z , vψ = 0 0 0 = −z +x ,
0 1 0 ∂z ∂y −1 0 0 ∂x ∂z
0 −1 0
" #
∂ ∂
vθ = 1 1 0 = −x +y .
0 0 0 ∂y ∂x

Moreover, the elements are linearly independent and so

  
 0 −a b 
so(3) =  a 0 −γ  |a, b, γ ∈ R .
−b γ 0
 

The Lie algebra so(3) is identified with R3 by associating to each v = (vϕ , vψ , vθ ) ∈

0 −a b
" #
R3 the matrix v ∈ so(3) given by v = a 0 −γ . Then we have the following
−b γ 0
identities:

1. u[
× v = [û, v]; and

2. u · v = − 12 Tr(û · v).

The exponential map exp : so(3) → SO(3) is given by Rodrigues relation

!2
sin kvk 1 sin kvk
exp(v) = I + v+ 2
kvk
v2
kvk 2
2

where the norm kvk is given by

p
kvk = (v 1 )2 + (v 2 )2 + (v 3 )2 .

The the dual, cotangent Lie algebra so(3)∗ , includes the three joint angular mo-
menta pϕ , pψ , pθ ∈ so(3)∗ , derived from the joint velocities v by multiplying them
with corresponding moments of inertia.
Note that the parameterization of SO(3)−rotations is the subject of continuous
research and development in many theoretical and applied fields of mechanics, such
as rigid body, structural, and multibody dynamics, robotics, spacecraft attitude
dynamics, navigation, image processing, etc.

44
5.3.3 The Heavy Top
Consider a rigid body moving with a fixed point but under the influence of gravity.
This problem still has a configuration space SO(3), but the symmetry group is only
the circle group S 1 , consisting of rotations about the direction of gravity. One says
that gravity has broken the symmetry from SO(3) to S 1 . This time, eliminating the
S 1 symmetry mysteriously leads one to the larger Euclidean group SE(3) of rigid
motion of R3 . Conversely, we can start with SE(3) as the configuration space for
the rigid–body and ‘reduce out’ translations to arrive at SO(3) as the configuration
space. The equations of motion for a rigid body with a fixed point in a gravitational
field give an interesting example of a system that is Hamiltonian. The underlying
Lie algebra consists of the algebra of infinitesimal Euclidean motions in R3 (see
[1, 3, 8, 9]).
The basic phase–space we start with is again T ∗ SO(3), parameterized by Euler
angles and their conjugate momenta. In these variables, the equations are in canon-
ical Hamiltonian form. However, the presence of gravity breaks the symmetry, and
the system is no longer SO(3) invariant, so it cannot be written entirely in terms of
the body angular momentum p. One also needs to keep track of Γ, the ‘direction of
gravity’ as seen from the body. This is defined by Γ = A−1 k, where k points upward
and A is the element of SO(3) describing the current configuration of the body. The
equations of motion are
I2 − I3
ṗ1 = p2 p3 + M gl(Γ2 χ3 − Γ3 χ2 ),
I2 I3
I3 − I1
ṗ2 = p3 p1 + M gl(Γ3 χ1 − Γ1 χ3 ),
I3 I1
I1 − I2
ṗ3 = p1 p2 + M gl(Γ1 χ2 − Γ2 χ1 ),
I1 I2
and Γ̇ = Γ × Ω,

where Ω is the body’s angular velocity vector, I1 , I2 , I3 are the body’s principal
moments of inertia, M is the body’s mass, g is the acceleration of gravity, χ is the
body fixed unit vector on the line segment connecting the fixed point with the body’s
center of mass, and l is the length of this segment.

5.4 Euclidean Groups of Rigid Body Motion

In this subsection we give description of two most important Lie groups in classical
mechanics in 2D and 3D, SE(2) and SE(3), respectively (see [4, 8, 9]).

45
5.4.1 Special Euclidean Group SE(2) in the Plane
The motion in uniaxial human joints is naturally modelled by the special Euclidean
group in the plane, SE(2). It consists of all transformations of R2 of the form Az +a,
where z, a ∈ R2 , and
  
cos θ − sin θ
A ∈ SO(2) = matrices of the form .
sin θ cos θ

In other words,
 group
 SE(2) consists of matrices of the form:
Rθ a
(Rθ , a) = , where a ∈ R2 and Rθ is the rotation matrix:
0 I
 
cos θ − sin θ
Rθ = , while I is the 3 × 3 identity matrix. The inverse (Rθ , a)−1
sin θ cos θ
is given by
 −1  
−1 Rθ a R−θ −R−θ a
(Rθ , a) = = .
0 I 0 I
The Lie algebra se(2) of SE(2) consists of 3 × 3 block matrices of the form
   
−ξJ v 0 1
, where J= , (J T = J −1 = −J),
0 0 −1 0

with the usual commutator bracket. If we identify se(2) with R3 by the isomorphism
 
−ξJ v
∈ se(2) 7−→ (ξ, v) ∈ R3 ,
0 0

then the expression for the Lie algebra bracket becomes

[(ξ, v1 , v2 ), (ζ, w1 , w2 )] = (0, ζv2 − ξw2 , ξw1 − ζv1) = (0, ξJ T w − ζJ T v),

where v = (v1 , v2 ) and w = (w1 , w2 ).

The adjoint group action of
   
Rθ a −ξJ v
(Rθ , a) on (ξ, v) =
0 I 0 0

is given by the group conjugation,

     
Rθ a −ξJ v R−θ −R−θ a −ξJ ξJa + Rθ v
= ,
0 I 0 0 0 I 0 0

46
or, in coordinates,
Ad(Rθ ,a) (ξ, v) = (ξ, ξJa + Rθ v). (2)
In proving (2) we used the ∗
 µidentityRθ J = JRθ . Identify the dual algebra, se(2) ,
with matrices of the form 2J 0 , via the nondegenerate pairing given by the
α 0
trace of the product. Thus, se(2)∗ is isomorphic to R3 via
 µ 
2J 0 ∈ se(2)∗ 7−→ (µ, α) ∈ R3 ,
α 0

so that in these coordinates, the pairing between se(2)∗ and se(2) becomes

h(µ, α), (ξ, v)i = µξ + α · v,

that is, the usual dot product in R3 . The coadjoint group action is thus given by

Ad∗(R −1 (µ, α) = (µ − Rθ α · Ja + Rθ α). (3)

θ ,a)

Formula (3) shows that the coadjoint orbits are the cylinders T ∗ Sα1 = {(µ, α)| kαk =
const} if α 6= 0 together with the points are on the µ−axis. The canonical cotangent
bundle projection π : T ∗ Sα1 → Sα1 is defined as π(µ, α) = α.

5.4.2 Special Euclidean Group SE(3) in the 3D Space

The most common group structure in human biodynamics is the special Euclidean
group in 3D space, SE(3). Briefly, the Euclidean SE(3)−group is defined as a
semidirect (noncommutative) product of 3D rotations and 3D translations, SE(3) :=
SO(3) ✄ R3 (see [4, 8, 9]). Its most important subgroups are the following:

Subgroup Definition
SO(3), group of rotations Set of all proper orthogonal
in 3D (a spherical joint) 3 × 3 − rotational matrices
Set of all 3 × 3 − matrices:

SE(2), special Euclidean group cos θ sin θ rx
in 2D (all planar motions)  − sin θ cos θ ry 
0 0 1
SO(2), group of rotations in 2D Set of all proper orthogonal
subgroup of SE(2)–group 2 × 2 − rotational matrices
(a revolute joint) included in SE(2) − group
3
R , group of translations in 3D
Euclidean 3D vector space
(all spatial displacements)

47
Lie Group SE(3) and Its Lie Algebra An element of SE(3) is a pair (A, a)
where A ∈ SO(3) and a ∈ R3 . The action of SE(3) on R3 is the rotation A followed
by translation by the vector a and has the expression

(A, a) · x = Ax + a.

Using this formula, one sees that multiplication and inversion in SE(3) are given by

(A, a)(B, b) = (AB, Ab + a) and (A, a)−1 = (A−1 , −A−1 a),

for A, B ∈ SO(3) and a, b ∈ R3 . The identity element is (l, 0).

The Lie algebra of the Euclidean group SE(3) is se(3) = R3 × R3 with the Lie
bracket
[(ξ, u), (η, v)] = (ξ × η, ξ × v − η × u). (4)
The Lie algebra of the Euclidean group has a structure that is a special case of
what is called a semidirect product. Here it is the product of the group of rotations
with the corresponding group of translations. It turns out that semidirect products
occur under rather general circumstances when the symmetry in T ∗ G is broken (see
[4, 8, 9]).
The dual Lie algebra of the Euclidean group SE(3) is se(3)∗ = R3 × R3 with the
same Lie bracket (5).

Representation of SE(3) In other words, SE(3) := SO(3) ✄ R3 is the Lie group

consisting of isometries of R3 .
Using homogeneous coordinates, we can represent SE(3) as follows,
  
R p 3
SE(3) = ∈ GL(4, R) : R ∈ SO(3), p ∈ R ,
0 1

with the action on R3 given by the usual matrix–vector product when we identify
R3 with the section R3 × {1} ⊂ R4 . In particular, given
 
R p
g= ∈ SE(3),
0 1

and q ∈ R3 , we have
g · q = Rq + p,
or as a matrix–vector product,
    
R p q Rq + p
= .
0 1 1 1

48
Lie algebra of SE(3) The Lie algebra of SE(3) is given by
  
ω v
se(3) = ∈ M4 (R) : ω ∈ so(3), v ∈ R3 ,
0 0
where the attitude matrix ω : R3 → so(3) is given by
 
0 −ω z ω y
ω =  ωz 0 −ωx  .
−ω y ωx 0

The exponential map of SE(3) The exponential map, exp : se(3) → SE(3), is
given by    
ω v exp(ω) Av
exp = ,
0 0 0 1
where

1 − cos kωk kωk − sin kωk 2

A=I+ 2 ω+ ω ,
kωk kωk3
and exp(ω) is given by the Rodriguez’ formula,
sin kωk 1 − cos kωk 2
exp(ω) = I + ω+ ω .
kωk kωk2
In other words, the special Euclidean group SE(3) := SO(3) ✄ R3 is the Lie
group consisting of isometries of the Euclidean 3D space R3 . An element of SE(3)
is a pair (A, a) where A ∈ SO(3) and a ∈ R3 . The action of SE(3) on R3 is the
rotation A followed by translation by the vector a and has the expression
(A, a) · x = Ax + a.
The Lie algebra of the Euclidean group SE(3) is se(3) = R3 × R3 with the Lie
bracket
[(ξ, u), (η, v)] = (ξ × η, ξ × v − η × u). (5)
Using homogeneous coordinates, we can represent SE(3) as follows,
  
R p 3
SE(3) = ∈ GL(4, R) : R ∈ SO(3), p ∈ R ,
0 1
with the action on R3 given by the usual matrix–vector product when we identify
R3 with the section R3 × {1} ⊂ R4 . In particular, given
 
R p
g= ∈ SE(3),
0 1

49
and q ∈ R3 , we have
g · q = Rq + p,
or as a matrix–vector product,
    
R p q Rq + p
= .
0 1 1 1

The Lie algebra of SE(3), denoted se(3), is given by

  
ω v 3
se(3) = ∈ M4 (R) : ω ∈ so(3), v ∈ R ,
0 0

where the attitude (or, angular velocity) matrix ω : R3 → so(3) is given by

 
0 −ωz ω y
ω =  ωz 0 −ω x  .
−ωy ω x 0

The exponential map, exp : se(3) → SE(3), is given by

   
ω v exp(ω) Av
exp = ,
0 0 0 1

where

1 − cos kωk kωk − sin kωk

A=I+ 2 ω+ ω2 ,
kωk kωk3
and exp(ω) is given by the Rodriguez’ formula,

sin kωk 1 − cos kωk 2

exp(ω) = I + ω+ ω .
kωk kωk2

5.5 Basic Mechanical Examples

5.5.1 SE(2)−Hovercraft
Configuration manifold is (θ, x, y) ∈ SE(2), given by matrix
 
cos θ sin θ x
P =  − sin θ cos θ y .
0 0 1

50
 Kinematic equations of motion in Lie algebra se(2):
 
0 ω vx
Ṗ = P  −ω 0 vy  , (ω = θ̇, vx = ẋ, vy = ẏ).
0 0 0
Kinetic energy:
1 1
Ek = m(vx2 + vy2 ) + Iω 2 ,
2 2
where m, I are mass and inertia moment of the hovercraft.
Dynamical equations of motion:
mv̇x = mωvy + u1 ,
mv̇y = −mωvx + u2 ,
I ω̇ = τ u2 ,
where τ = −h is the torque applied at distance h from the center–of–mass, while
u1 , u2 are control inputs.

5.5.2 SO(3)−Satellite
Configuration manifold is rotation matrix R ∈ SO(3), with associated angular–
velocity (attitude) matrix ω = (ω 1 , ω 2 , ω 3 ) ∈ so(3) ≈ R3 given by
 
0 −ω 3 ω 2
ω ∈ so(3) 7−→  ω 3 0 −ω 1  .
−ω 2 ω 1 0
Kinematic equation of motion in so(3):
Ṙ = Rω,
Kinetic energy:
1
Ek = ω T Iω,
2
where inertia tensor I is given by diagonal matrix,
I = diag{I1 , I2 , I3 }.
Dynamical Euler equations of motion:
Iω̇ = Iω × ω + τ i ui ,
where × is the cross–product in 3D, τ i are three external torques and ui = ui (t) are
control inputs.

51
5.5.3 SE(3)−Submarine
The motion of a rigid body in incompressible, irrotational and inviscid fluid is de-
fined by the configuration manifold SE(3), given by a pair of rotation matrix and
translation vector, (R, p) ∈ SE(3), such that angular velocity (attitude) matrix and
linear velocity vector, (ω, v) ∈ se(3) ≈ R6 .

Kinematic equations of motion in se(3):

ṗ = Rv, Ṙ = Rω.

Kinetic energy (symmetrical):

1 1
Ek = vT Mv + ω T Iω,
2 2
where mass and inertia matrices are diagonal (for a neutrally buoyant ellipsoidal
body with uniformly distributed mass),

M = diag{m1 , m2 , m3 },
I = diag{I1 , I2 , I3 }.

Dynamical Kirchhoff equations of motion read:

Mv̇ = Mv × ω, Iω̇ = Iω × ω + Mv × v.

By including the body–fixed external forces and torques, fi , τ i , with input controls
ui = u( t), the dynamical equations become:

Mv̇ = Mv × ω + fi ui ,
Iω̇ = Iω × ω + Mv × v + τ i ui .

5.6 Newton–Euler SE(3)−Dynamics

5.6.1 SO(3) : Euler Equations of Rigid Rotations
Unforced Euler equations read in vector form

π̇ ≡ Iω̇ = π × ω, with I = diag{I1 , I2 , I3 }

and in scalar form

I1 ω̇ 1 = (I2 − I3 )ω 2 ω3
I2 ω̇ 2 = (I3 − I1 )ω 3 ω1 .
I3 ω̇ 3 = (I1 − I2 )ω 1 ω2

52
Using rotational kinetic–energy Lagrangian
1 1
L(ω) = Ekrot = ω t Iω = (I1 ω 21 + I2 ω 22 + I3 ω 23 ) (t = ‘transpose’)
2 2
Regarding the angular momentum π = ∂ω L = Iω = (I1 ω 1 , I2 ω 2 , I3 ω 3 ) as a vector,
we can derive unforced Euler equations: π̇ = π × ω as a system of Euler–Lagrange–
Kirchhoff equations
d
∂ω L = ∂ω L × ω.
dt
Forced Euler equations read in vector form

π̇ + ω × π = T

and in scalar form

I1 ω̇1 + (I3 − I2 )ω 2 ω 3 = T1
I2 ω̇2 + (I1 − I3 )ω 3 ω 1 = T2
I3 ω̇3 + (I2 − I1 )ω 1 ω 2 = T3

5.6.2 SE(3) : Coupled Newton–Euler Equations

Forced coupled Newton–Euler equations read in vector form

ṗ ≡ Mv̇ = F + p × ω, with M = diag{m1 , m2 , m3 }

π̇ ≡ Iω̇ = T + π × ω + p × v, I = diag{I1 , I2 , I3 },

with principal inertia moments given in Cartesian coordinates (x, y, z) by density

ρ−dependent volume integrals
ZZZ ZZZ ZZZ
2 2 2 2
I1 = ρ(z +y )dxdydz, I2 = ρ(x +y )dxdydz, I3 = ρ(x2 +y 2 )dxdydz,

In tensor form, the forced–coupled Newton–Euler equations read

ṗi ≡ Mij v̇ j = Fi + εjik pj ω k ,

π̇ i ≡ Iij ω̇ j = Ti + εjik π j ω k + εjik pj v k ,

where the permutation symbol εjik is defined as


+1 if (i, j, k) is (1, 2, 3), (3, 1, 2) or (2, 3, 1),

εjik = −1 if (i, j, k) is (3, 2, 1), (1, 3, 2) or (2, 1, 3),

0 otherwise: i = j or j = k or k = i.


53
 In scalar form these equations read

ṗ1 = F1 − m3 v3 ω 2 + m2 v2 ω 3
ṗ2 = F2 + m3 v3 ω 1 − m1 v1 ω 3
ṗ3 = F3 − m2 v2 ω 1 + m1 v1 ω 2
.
π̇ 1 = T1 + (m2 − m3 )v2 v3 + (I2 − I3 )ω 2 ω 3
π̇ 2 = T2 + (m3 − m1 )v1 v3 + (I3 − I1 )ω 1 ω 3
π̇ 3 = T3 + (m1 − m2 )v1 v2 + (I1 − I2 )ω 1 ω 2

These coupled rigid–body equations can be derived from the Newton–Euler kinetic
energy
1 1
Ek = vt Mv + ω t Iω
2 2
or, in tensor form
1 1
E = Mij v̇ i v̇ j + Iij ω̇ i ω̇ j .
2 2
Using the Kirchhoff–Lagrangian equations
d
∂v Ek = ∂v Ek × ω + F
dt
d
∂ω Ek = ∂ω Ek × ω + ∂v Ek × v + T,
dt
or, in tensor form
d
∂ i E = εjik (∂vj E) ω k + Fi ,
dt v
d
∂ωi E = εjik (∂ωj E) ω k + εjik (∂vj E) v k + Ti
dt
we can derive linear and angular momentum covectors

p = ∂v Ek , π = ∂ω Ek

or, in tensor form

pi = ∂vi E, π i = ∂ωi E,
and in scalar form

p = [p1 , p2 , p3 ] = [m1 v1 , m2 v2 , m2 v3 ]
π = [π 1 , π 2 , π 3 ] = [I1 ω 1 , I2 ω 2 , I3 ω 3 ],

54
with their respective time derivatives, in vector form
d d d d
ṗ = p = ∂v Ek , π̇ = π = ∂ω Ek
dt dt dt dt
or, in tensor form
d d d d
ṗi = pi = ∂vi E, π̇ i = π i = ∂ωi E,
dt dt dt dt
and in scalar form
ṗ = [ṗ1 , ṗ2 , ṗ3 ] = [m1 v̇1 , m2 v̇2 , m3 v̇3 ]
π̇ = [π̇ 1 , π̇ 2 , π̇ 3 ] = [I1 ω̇ 1 , I2 ω̇ 2 , I3 ω̇ 3 ].
In addition, for the purpose of biomechanical injury prediction/prevention, we
have linear and angular jolts, respectively given in vector form by
Ḟ = p̈ − ṗ × ω − p × ω̇
Ṫ = π̈ −π̇ × ω − π × ω̇ − ṗ × v − p × v̇,
or, in tensor form13
Ḟi = p̈i − εjik ṗj ω k − εjik pj ω̇ k ,
Ṫi = π̈ i − εjik π̇ j ω k − εjik π j ω̇ k − εjik ṗj v k − εjik pj v̇ k ,
where the linear and angular jolt covectors are
Ḟ ≡ Ḟi = Mv̈ ≡ Mij v̈ j = [Ḟ1 , Ḟ2 , Ḟ3 ],
Ṫ ≡ Ṫi = Iω̈ ≡ Iij ω̈ j = [Ṫ1 , Ṫ2 , Ṫ3 ],
where
v̈ = v̈ i = [v̈ 1 , v̈ 2 , v̈ 3 ]t , ω̈ = ω̈ i = [ω̈ 1 , ω̈ 2 , ω̈ 3 ]t ,
are linear and angular jerk vectors.
In scalar form, the SE(3)−jolt expands as

 Ḟ1 = p̈1 − m2 ω3 v̇2 + m3 (ω 2 v̇3 + v3 ω̇2 ) − m2 v2 ω̇ 3 ,
Ḟ = p̈2 + m1 ω3 v̇1 − m3 ω 1 v̇3 − m3 v3 ω̇ 1 + m1 v1 ω̇ 3 ,
 2
Ḟ3 = p̈3 − m1 ω2 v̇1 + m2 ω 1 v̇2 − v2 ω̇ 1 − m1 v1 ω̇ 2 ,

 Ṫ1 = π̈ 1 − (m2 − m3 ) (v3 v̇2 + v2 v̇3 ) − (I2 − I3 ) (ω 3 ω̇ 2 + ω 2 ω̇ 3 ) ,
Ṫ = π̈ 2 + (m1 − m3 ) (v3 v̇1 + v1 v̇3 ) + (I1 − I3 ) (ω 3 ω̇ 1 + ω 1 ω̇ 3 ) ,
 2
Ṫ3 = π̈ 3 − (m1 − m2 ) (v2 v̇1 + v1 v̇2 ) − (I1 − I2 ) (ω 2 ω̇ 1 + ω 1 ω̇ 2 ) .
13
In this paragraph the overdots actually denote the absolute Bianchi (covariant) derivatives, so
that the jolts retain the proper covector character, which would be lost if ordinary time derivatives
are used. However, for simplicity, we stick to the same notation.

55
5.7 Symplectic Group in Hamiltonian Mechanics
Here we givea brief description
 of symplectic group (see [4, 8, 9]).
0 I
Let J = , with I the n × n identity matrix. Now, A ∈ L(R2n , R2n )
−I 0
is called a symplectic matrix if AT J A = J. Let Sp(2n, R) be the set of 2n × 2n
symplectic matrices. Taking determinants of the condition AT J A = J gives det A =
±1, and so A ∈ GL(2n, R). Furthermore, if A, B ∈ Sp(2n, R), then (AB)T J(AB) =
B T AT JAB = J. Hence, AB ∈ Sp(2n, R), and if AT J A = J, then JA = (AT )−1 J =
(A−1 )T J, so J = (A − 1)T JA−1 , or A−1 ∈ Sp(2n, R). Thus, Sp(2n, R) is a group.
The symplectic Lie group
Sp(2n, R) = A ∈ GL(2n, R) : AT J A = J


is a noncompact, connected Lie group of dimension 2n2 + n. Its Lie algebra

sp(2n, R) = A ∈ L(R2n , R2n ) : AT J A = J = 0 ,


called the symplectic Lie algebra, consists of the 2n × 2n matrices A satisfying

AT J A = 0.
Consider a particle of mass m moving in a potential V (q), where q i = (q 1 , q 2 , q 3 ) ∈
R . Newtonian second law states that the particle moves along a curve q(t) in R3 in
3

such a way that mq̈ i = − grad V (q i ). Introduce the 3D–momentum pi = mq̇ i , and
the energy (Hamiltonian)
3
1 X 2
H(q, p) = pi + V (q).
2m
i=1

Then
∂H ∂V
= = −mq̈ i = −ṗi , and
∂q i ∂q i
∂H 1
= pi = q̇ i , (i = 1, 2, 3),
∂pi m

and hence Newtonian law F = mq̈ i is equivalent to Hamiltonian equations

∂H ∂H
q̇ i = , ṗi = − .
∂pi ∂q i
Now, writing z = (q i , pi ),
!
  ∂H
0 I ∂q i
= q̇ i , ṗi = ż,


J grad H(z) = ∂H
−I 0 ∂pi

56
so the complex Hamiltonian equations read

ż = J grad H(z).

Now let f : R3 × R3 → R3 × R3 and write w = f (z). If z(t) satisfies the complex

Hamiltonian equations then w(t) = f (z(t)) satisfies ẇ = AT ż, where AT = [∂wi /∂z j ]
is the Jacobian matrix of f . By the chain rule,

ẇ = AT J grad H(z) = AT J A grad H(z(w)).

z w

Thus, the equations for w(t) have the form of Hamiltonian equations with energy
K(w) = H(z(w)) iff AT J A = J, that is, iff A is symplectic. A nonlinear transfor-
mation f is canonical iff its Jacobian matrix is symplectic. Sp(2n, R) is the linear
invariance group of classical mechanics.

6 Medical Applications: Prediction of Injuries

6.1 General Theory of Musculo–Skeletal Injury Mechanics
The prediction and prevention of traumatic brain injury, spinal injury and musculo-
skeletal injury is a very important aspect of preventive medical science. Recently,
in a series of papers [28, 29, 30], we have proposed a new coupled loading-rate
hypothesis as a unique cause of all above injuries. This new hypothesis states that
the unique cause of brain, spinal and musculo-skeletal injuries is a Euclidean Jolt,
which is an impulsive loading that strikes any part of the human body (head, spine
or any bone/joint) – in several coupled degrees-of-freedom simultaneously. It never
goes in a single direction only. Also, it is never a static force. It is always an
impulsive translational and/or rotational force coupled to some mass eccentricity.
This is, in a nutshell, our universal Jolt theory of all mechanical injuries.
To show this, based on the previously defined covariant force law, we have firstly
formulated the fully coupled Newton–Euler dynamics of:
1. Brain’s micro-motions within the cerebrospinal fluid inside the cranial cavity;
2. Any local inter-vertebral motions along the spine; and
3. Any local joint motions in the human musculo-skeletal system.

Then, from it, we have defined the essential concept of Euclidean Jolt, which
is the main cause of all mechanical injuries. The Euclidean Jolt has two main
components:
1. Sudden motion, caused either by an accidental impact or slightly distorted
human movement; and

57
Figure 5: Human brain and its SE(3)−group of microscopic three-dimensional mo-
tions within the cerebrospinal fluid inside the cranial cavity.

2. Unnatural mass distribution of the human body (possibly with some added
masses), which causes some mass eccentricity from the natural physiological body
state.
What does this all mean? We will try to explain it in “plain English”. As we
live in a 3D space, one could think that motion of any part of the human body, ei-
ther caused by an accidental impact or by voluntary human movement, “just obeys
classical mechanics in 6 degrees-of-freedom: three translations and three rotations”.
However, these 6 degrees-of-freedom are not independent motions as it is suggested
by the standard term “degrees-of-freedom”. In reality, these six motions of any
body in space are coupled. Firstly, three rotations are coupled in the so-called ro-
tation group (or matrix, or quaternion). Secondly, three translations are coupled
with the rotation group to give the full Euclidean group of rigid body motions in
space. A simple way to see this is to observe someone throwing an object in the
air or hitting a tennis ball: how far and where it will fly depends not only on the
standard “projectile” mechanics, but also on its local “spin” around all three axes
simultaneously. Every golf and tennis player knows this simple fact. Once the spin
is properly defined we have a “fully coupled Newton–Euler dynamics” – to start with.

58
Figure 6: Human body representation in terms of SE(3)/SE(2)-groups of rigid-body
motion, with the vertebral column represented as a chain of 26 flexibly-coupled
SE(3)-groups.

The covariant force law for any biodynamical system (which we introduced earlier
in our biodynamics books and papers, see our references in the cited papers above)
goes one step beyond the Newton–Euler dynamics. It states:

Euclidean Force covector field =

Body mass distribution × Euclidean Acceleration vector field

This is a nontrivial biomechanical generalization of the fundamental Newton’s

definition of the force acting on a single particle. Unlike classical engineering me-
chanics of multi-body systems, this fundamental law of biomechanics proposes that
forces acting on a multi-body system and causing its motions are fundamentally dif-
ferent physical quantities from the resulting accelerations. In simple words, forces
are massive quantities while accelerations are massless quantities. More precisely,
the acceleration vector field includes all linear and angular accelerations of individual
body segments. When we couple them all with the total body’s mass-distribution
matrix of all body segments (including all masses and inertia moments), we get the
force co-vector field, comprising all the forces and torques acting on the individual
body segments. In this way, we have defined the 6-dimensional Euclidean force for
an arbitrary biomechanical system.
Now, for prediction of injuries, we need to take the rate-of-change (or derivative,

59
Figure 7: Schematic latero-frontal view of the left knee joint. Although designed
to perform mainly flexion/extension (strictly in the sagittal plane) with some re-
stricted medial/lateral rotation in the semi-flexed position, it is clear that the knee
joint really has at least six-degrees-of-freedom, including three micro-translations.
The injury actually occurs when some of these microscopic translations become
macroscopic, which normally happens only after an external jolt.

with respect to time) of the Euclidean biomechanical force defined above. In this way,
we get the Euclidean Jolt, which is the sudden change (in time) of the 6-dimensional
Euclidean force:

Euclidean Jolt covector field =

Body mass distribution × Euclidean Jerk vector field

And again, it consists of two components: (i) massless linear and angular jerks
(of all included body segments), and (ii) their mass distribution. For the sake of sim-
plicity, we can say that the mass distribution matrix includes all involved segmental
masses and inertia moments, as well as “eccentricities” or “pathological leverages”
from the normal physiological state.

Therefore, the unique cause of all brain, spine and musculo-skeletal injuries has
two components:
1. Coupled linear and angular jerks; and
2. Mass distribution with “eccentricities”.

60
In other words, there are no injuries in static conditions without any mass
eccentricities; all injuries are caused by mutually coupled linear and an-
gular jerks, which are also coupled with the involved human mass distri-
bution.

The Euclidean Jolt causes two forms of discontinuous brain, spine or musculo-
skeletal injury:
1. Mild rotational disclinations; and
2. Severe translational dislocations (or, fractures).

In the cited papers above, we have developed the soft-body dynamics of biome-
chanical disclinations and dislocations, caused by the Euclidean Jolt, using the
Cosserat multipolar viscoelastic continuum model.

Implications of the new universal theory are various, as follows.

A. The research in traumatic brain injury (TBI, see Figure 5) has so far iden-
tified the rotation of the brain-stem as the main cause of the TBI due to various
crashes/impacts. The contribution of our universal Jolt theory to the TBI research
is the following:

1. Rigorously defined this brain rotation as a mechanical disclination of the

brain-stem tissue modelled by the Cosserat multipolar soft-body model;
2. Showing that brain rotation is never uni-axial but always three-axial;
3. Showing that brain rotation is always coupled with translational disloca-
tions. This is a straightforward consequence of our universal Jolt theory.

These apparently ‘obvious’ facts are actually radically new: we cannot separately
analyze rapid brain’s rotations from translations, because they are in reality always
coupled.

One practical application of the brain Jolt theory is in design of helmets. Briefly,
a ‘hard’ helmet saves the skull but not the brain; alternatively, a ‘soft’ helmet pro-
tects the brain from the collision jolt but does not protect the skull. A good helmet
is both ‘hard’ and ‘soft’. A proper helmet would need to have both a hard external
shell (to protect the skull) and a soft internal part (that will dissipate the energy
from the collision jolt by its own destruction, in the same way as a car saves its
passengers from the collision jolt by its own destruction).

Similarly, in designing safer car air-bags, the two critical points will be (i) their

61
placement within the car, and (ii) their “soft-hard characteristics”, similar to the
helmet characteristics described above.

B. In case of spinal injury (see Figure 6), the contribution of our universal Jolt
theory is the following:

1. The spinal injury is always localized at the certain vertebral or inter-vertebral

point;
2. In case of severe translational injuries (vertebral fractures or discus herniae)
they can be identified using X-ray or other medical imaging scans; in case of micro-
scopic rotational injuries (causing the back-pain syndrome) they cannot be identified
using current medical imaging scans;
3. There is no spinal injury without one of the following two causes:
a. Impulsive rotational + translational loading caused by either fast human
movements or various crashes/impacts; and/or
b. Static eccentricity from the normal physiological spinal form, caused by
external loading;
c. Any spinal injury is caused by a combination of the two points above:
impulsive rotational + translational loading and static eccentricity.

This is a straightforward consequence of our universal Jolt theory. We cannot

separately analyze translational and rotational spinal injuries. Also, there are no
“static injuries” without eccentricity. Indian women have for centuries carried bulky
loads on their heads without any spinal injuries; they just prevented any load ec-
centricities and any jerks in their motion.

The currently used “Principal loading hypothesis” that describes spinal injuries
in terms of spinal tension, compression, bending, and shear, covers only a small
subset of all spinal injuries covered by our universal Jolt theory. To prevent spinal
injuries we need to develop spinal jolt awareness: ability to control all possible im-
pulsive spinal loadings as well as static eccentricities.

C. In case of general musculo-skeletal injury (see Figure 7 for the particular case
of knee injury), the contribution of our universal Jolt theory is the following:

1. The injury is always localized at the certain joint or bone and caused by an
impulsive loading, which hits this particular joint/bone in several coupled degrees-
of-freedom simultaneously;
2. Injury happens when most of the body mass is hanging on that joint; for
example, in case of a knee injury, when most of the body mass is on one leg with

62
a semi-flexed knee — and then, caused by some external shock, the knee suddenly
“jerks” (this can happen in running, skiing, and ball games, as well as various
crashes/impacts); or, in case of shoulder injury, when most of the body mass is
hanging on one arm and then it suddenly jerks.

To prevent these injuries we need to develop musculo-skeletal jolt awareness. For

example, never overload a flexed knee and avoid any kind of uncontrolled motions
(like slipping) or collisions with external objects.

6.2 Analytical Mechanics of Traumatic Brain Injury (TBI)

6.2.1 The SE(3)−jolt: the cause of TBI
In this subsection we give a brief on TBI mechanics. For more details and references,
see [28].
In the language of modern dynamics, the microscopic motion of human brain
within the skull is governed by the Euclidean SE(3)–group of 3D motions (see next
subsection). Within brain’s SE(3)–group we have both SE(3)–kinematics (consisting
of SE(3)–velocity and its two time derivatives: SE(3)–acceleration and SE(3)–jerk)
and SE(3)–dynamics (consisting of SE(3)–momentum and its two time derivatives:
SE(3)–force and SE(3)–jolt), which is brain’s kinematics × brain’s mass–inertia
distribution.
Informally, the external SE(3)–jolt14 is a sharp and sudden change in the SE(3)–
force acting on brain’s mass–inertia distribution (given by brain’s mass and inertia
matrices). That is, a ‘delta’–change in a 3D force–vector coupled to a 3D torque–
vector, striking the head–shell with the brain immersed into the cerebrospinal fluid.
In other words, the SE(3)–jolt is a sudden, sharp and discontinues shock in all
6 coupled dimensions of brain’s continuous micro–motion within the cerebrospinal
fluid (Figure 5), namely within the three Cartesian (x, y, z)–translations and the
three corresponding Euler angles around the Cartesian axes: roll, pitch and yaw.
If the SE(3)–jolt produces a mild shock to the brain (e.g., strong head shake), it
causes mild TBI, with temporary disabled associated sensory-motor and/or cognitive
functions and affecting respiration and movement. If the SE(3)–jolt produces a hard
14
The mechanical SE(3)–jolt concept is based on the mathematical concept of higher–order tan-
gency (rigorously defined in terms of jet bundles of the head’s configuration manifold), as follows:
When something hits the human head, or the head hits some external body, we have a collision.
This is naturally described by the SE(3)–momentum, which is a nonlinear coupling of 3 linear New-
tonian momenta with 3 angular Eulerian momenta. The tangent to the SE(3)–momentum, defined
by the (absolute) time derivative, is the SE(3)–force. The second-order tangency is given by the
SE(3)–jolt, which is the tangent to the SE(3)–force, also defined by the time derivative.

63
shock (hitting the head with external mass), it causes severe TBI, with the total
loss of gesture, speech and movement.
The SE(3)–jolt is the absolute time–derivative of the covariant force 1–form (or,
co-vector field). The fundamental law of biomechanics is the covariant force law :

Force co-vector field = Mass distribution × Acceleration vector–field,

which is formally written (using the Einstein summation convention, with indices
labelling the three Cartesian translations and the three corresponding Euler angles):

Fµ = mµν aν , (µ, ν = 1, ..., 6)

where Fµ denotes the 6 covariant components of the external “pushing” SE(3)–force

co-vector field, mµν represents the 6×6 covariant components of brain’s inertia–
metric tensor, while aν corresponds to the 6 contravariant components of brain’s
internal SE(3)–acceleration vector-field.
D
Now, the covariant (absolute, Bianchi) time-derivative dt (·) of the covariant
SE(3)–force Fµ defines the corresponding external “striking” SE(3)–jolt co-vector
field:
D D  
(Fµ ) = mµν (aν ) = mµν ȧν + Γνµλ aµ aλ , (6)
dt dt
D ν
where dt (a ) denotes the 6 contravariant components of brain’s internal SE(3)–jerk
vector-field and overdot (˙) denotes the time derivative. Γνµλ are the Christoffel’s
symbols of the Levi–Civita connection for the SE(3)–group, which are zero in case
of pure Cartesian translations and nonzero in case of rotations as well as in the
full–coupling of translations and rotations.
In the following, we elaborate on the SE(3)–jolt concept (using vector and tensor
methods) and its biophysical TBI consequences in the form of brain’s dislocations
and disclinations.

6.2.2 SE(3)−group of brain’s micro–motions within the CSF

The brain and the CSF together exhibit periodic microscopic translational and ro-
tational motion in a pulsatile fashion to and from the cranial cavity, in the fre-
quency range of normal heart rate (with associated periodic squeezing of brain’s
ventricles). This micro–motion is mathematically defined by the Euclidean (gauge)
SE(3)−group.
In other words, the gauge SE(3)−group of Euclidean micro-motions of the brain
immersedin the cerebrospinal
 fluid within the cranial cavity, contains matrices of
R b
the form , where b is brain’s 3D micro-translation vector and R is brain’s
0 1
3D rotation matrix, given by the product R = Rϕ · Rψ · Rθ of brain’s three Eulerian

64
micro-rotations, roll = Rϕ , pitch = Rψ , yaw = Rθ , performed respectively about
the x−axis by an angle ϕ, about the y−axis by an angle ψ, and about the z−axis
by an angle θ,
     
1 0 0 cos ψ 0 sin ψ cos θ − sin θ 0
Rϕ =  0 cos ϕ − sin ϕ  , Rψ =  0 1 0  , Rθ =  sin θ cos θ 0 .
0 sin ϕ cos ϕ − sin ψ 0 cos ψ 0 0 1

Therefore, brain’s natural SE(3)−dynamics within the cerebrospinal fluid is

given by the coupling of Newtonian (translational) and Eulerian (rotational) equa-
tions of micro-motion.

6.2.3 Brain’s natural SE(3)−dynamics

To support our coupled loading–rate hypothesis, we formulate the coupled Newton–
Euler dynamics of brain’s micro-motions within the scull’s SE(3)−group of motions.
The forced Newton–Euler equations read in vector (boldface) form

Newton : ṗ ≡ Mv̇ = F + p × ω, (7)

Euler : π̇ ≡ Iω̇ = T + π × ω + p × v,

where × denotes the vector cross product,15

M ≡ Mij = diag{m1 , m2 , m3 } and I ≡ Iij = diag{I1 , I2 , I3 }, (i, j = 1, 2, 3)

are brain’s (diagonal) mass and inertia matrices,16 defining brain’s mass–inertia
distribution, with principal inertia moments given in Cartesian coordinates (x, y, z)
by volume integrals
ZZZ ZZZ ZZZ
2 2 2 2
I1 = ρ(z +y )dxdydz, I2 = ρ(x +z )dxdydz, I3 = ρ(x2 +y 2 )dxdydz,

15
Recall that the cross product u × v of two vectors u and v equals u × v = uv sin θn, where θ
is the angle between u and v, while n is a unit vector perpendicular to the plane of u and v such
that u and v form a right-handed system.
16
In reality, mass and inertia matrices (M, I) are not diagonal but rather full 3 × 3 positive–
definite symmetric matrices with coupled mass– and inertia–products. Even more realistic, fully–
coupled mass–inertial properties of a brain immersed in (incompressible, irrotational and inviscid)
cerebrospinal fluid are defined by the single non-diagonal 6 × 6 positive–definite symmetric mass–
inertia matrix MSE(3) , the so-called material metric tensor of the SE(3)−group, which has all
nonzero mass–inertia coupling products. In other words, the 6 × 6 matrix MSE(3) contains: (i)
brain’s own mass plus the added mass matrix associated with the fluid, (ii) brain’s own inertia plus
the added inertia matrix associated with the potential flow of the fluid, and (iii) all the coupling
terms between linear and angular momenta. However, for simplicity, in this paper we shall consider
only the simple case of two separate diagonal 3 × 3 matrices (M, I).

65
dependent on brain’s density ρ = ρ(x, y, z),

v ≡ v i = [v1 , v2 , v3 ]t and ω ≡ ω i = [ω 1 , ω 2 , ω 3 ]t

(where [ ]t denotes the vector transpose) are brain’s linear and angular velocity
vectors17 (that is, column vectors),

F ≡ Fi = [F1 , F2 , F3 ] and T ≡ Ti = [T1 , T2 , T3 ]

are gravitational and other external force and torque co-vectors (that is, row vectors)
acting on the brain within the scull,

p ≡ pi ≡ Mv = [p1 , p2 , p3 ] = [m1 v1 , m2 v2 , m2 v2 ] and

π ≡ π i ≡ Iω = [π 1 , π 2 , π 3 ] = [I1 ω1 , I2 ω 2 , I3 ω 3 ]

are brain’s linear and angular momentum co-vectors.

In tensor form, the forced Newton–Euler equations (7) read

ṗi ≡ Mij v̇ j = Fi + εjik pj ω k , (i, j, k = 1, 2, 3)

j
π̇ i ≡ Iij ω̇ = Ti + εjik π j ω k + εjik pj v k ,

where the permutation symbol εjik is defined as


+1 if (i, j, k) is (1, 2, 3), (3, 1, 2) or (2, 3, 1),

εjik = −1 if (i, j, k) is (3, 2, 1), (1, 3, 2) or (2, 1, 3),

0 otherwise: i = j or j = k or k = i.


In scalar form, the forced Newton–Euler equations (7) expand as


 ṗ1 = F1 − m3 v3 ω 2 + m2 v2 ω 3
Newton : ṗ = F2 + m3 v3 ω 1 − m1 v1 ω 3 , (8)
 2
ṗ3 = F3 − m2 v2 ω 1 + m1 v1 ω 2

 π̇ 1 = T1 + (m2 − m3 )v2 v3 + (I2 − I3 )ω 2 ω 3
Euler : π̇ = T2 + (m3 − m1 )v1 v3 + (I3 − I1 )ω 1 ω 3 ,
 2
π̇ 3 = T3 + (m1 − m2 )v1 v2 + (I1 − I2 )ω 1 ω 2

showing brain’s individual mass and inertia couplings.

17
In reality, ω is a 3 × 3 attitude matrix. However, for simplicity, we will stick to the (mostly)
symmetrical translation–rotation vector form.

66
 Equations (7)–(8) can be derived from the translational + rotational kinetic
energy of the brain18
1 1
Ek = vt Mv + ω t Iω, (9)
2 2
or, in tensor form
1 1
E = Mij v i v j + Iij ω i ω j .
2 2
For this we use the Kirchhoff–Lagrangian equations:
d
∂v Ek = ∂v Ek × ω + F, (10)
dt
d
∂ω Ek = ∂ω Ek × ω + ∂v Ek × v + T,
dt
∂Ek ∂Ek
where ∂v Ek = ∂v , ∂ω Ek = ∂ω ; in tensor form these equations read

d
∂ i E = εjik (∂vj E) ωk + Fi ,
dt v
d
∂ω i E = εjik (∂ω j E) ω k + εjik (∂vj E) v k + Ti .
dt
Using (9)–(10), brain’s linear and angular momentum co-vectors are defined as

p = ∂v Ek , π = ∂ω Ek ,

or, in tensor form

pi = ∂vi E, π i = ∂ωi E,
with their corresponding time derivatives, in vector form
d d d d
ṗ = p = ∂v E, π̇ = π = ∂ω E,
dt dt dt dt
or, in tensor form
d d d d
ṗi = pi = ∂vi E, π̇ i = π i = ∂ωi E,
dt dt dt dt
18
In a fully–coupled Newton–Euler brain dynamics, instead of equation (9) we would have brain’s
kinetic energy defined by the inner product:
" ! !#
1 p p
Ek = MSE(3) .
2 π π

67
or, in scalar form

ṗ = [ṗ1 , ṗ2 , ṗ3 ] = [m1 v̇1 , m2 v̇2 , m3 v̇3 ], π̇ = [π̇ 1 , π̇ 2 , π̇ 3 ] = [I1 ω̇ 1 , I2 ω̇ 2 , I3 ω̇ 3 ].

While brain’s healthy SE(3)−dynamics within the cerebrospinal fluid is given

by the coupled Newton–Euler micro–dynamics, the TBI is actually caused by the
sharp and discontinuous change in this natural SE(3) micro-dynamics, in the form
of the SE(3)−jolt, causing brain’s discontinuous deformations.

6.2.4 Brain’s traumatic dynamics: the SE(3)−jolt

The SE(3)−jolt, the actual cause of the TBI (in the form of the brain’s plastic
deformations), is defined as a coupled Newton+Euler jolt; in (co)vector form the
SE(3)−jolt reads19

Newton jolt : Ḟ = p̈ − ṗ × ω − p × ω̇ ,
SE(3) − jolt :
Euler jolt : Ṫ = π̈ −π̇ × ω − π × ω̇ − ṗ × v − p × v̇,

where the linear and angular jolt co-vectors are

Ḟ ≡ Mv̈ = [Ḟ1 , Ḟ2 , Ḟ3 ], Ṫ ≡ Iω̈ = [Ṫ1 , Ṫ2 , Ṫ3 ],

where
v̈ = [v̈1 , v̈2 , v̈3 ]t , ω̈ = [ω̈ 1 , ω̈ 2 , ω̈ 3 ]t ,
are linear and angular jerk vectors.
In tensor form, the SE(3)−jolt reads20

Ḟi = p̈i − εjik ṗj ω k − εjik pj ω̇ k , (i, j, k = 1, 2, 3)

Ṫi = π̈ i − εjik π̇ j ω k − εjik π j ω̇ k − εjik ṗj v k − εjik pj v̇ k ,

in which the linear and angular jolt covectors are defined as

Ḟ ≡ Ḟi = Mv̈ ≡ Mij v̈ j = [Ḟ1 , Ḟ2 , Ḟ3 ],

Ṫ ≡ Ṫi = Iω̈ ≡ Iij ω̈ j = [Ṫ1 , Ṫ2 , Ṫ3 ],

where v̈ = v̈ i , and ω̈ = ω̈ i are linear and angular jerk vectors.

19
Note that the derivative of the cross–product of two vectors follows the standard calculus
d
product–rule: dt (u × v) = u̇ × v + u × v̇.
20
In this paragraph the overdots actually denote the absolute Bianchi (covariant) time-derivative
(6), so that the jolts retain the proper covector character, which would be lost if ordinary time
derivatives are used. However, for the sake of simplicity and wider readability, we stick to the same
overdot notation.

68
 In scalar form, the SE(3)−jolt expands as

 Ḟ1 = p̈1 − m2 ω 3 v̇2 + m3 (ω2 v̇3 + v3 ω̇ 2 ) − m2 v2 ω̇3 ,
Newton jolt : Ḟ = p̈2 + m1 ω 3 v̇1 − m3 ω 1 v̇3 − m3 v3 ω̇ 1 + m1 v1 ω̇ 3 ,
 2
Ḟ3 = p̈3 − m1 ω 2 v̇1 + m2 ω 1 v̇2 − v2 ω̇1 − m1 v1 ω̇ 2 ,

 Ṫ1 = π̈ 1 − (m2 − m3 ) (v3 v̇2 + v2 v̇3 ) − (I2 − I3 ) (ω 3 ω̇2 + ω 2 ω̇ 3 ) ,
Euler jolt : Ṫ = π̈ 2 + (m1 − m3 ) (v3 v̇1 + v1 v̇3 ) + (I1 − I3 ) (ω 3 ω̇1 + ω 1 ω̇ 3 ) ,
 2
Ṫ3 = π̈ 3 − (m1 − m2 ) (v2 v̇1 + v1 v̇2 ) − (I1 − I2 ) (ω 2 ω̇1 + ω 1 ω̇ 2 ) .

We remark here that the linear and angular momenta (p, π), forces (F, T) and
jolts (Ḟ, Ṫ) are co-vectors (row vectors), while the linear and angular velocities
(v, ω), accelerations (v̇, ω̇) and jerks (v̈, ω̈) are vectors (column vectors). This
bio-physically means that the ‘jerk’ vector should not be confused with the ‘jolt’ co-
vector. For example, the ‘jerk’ means shaking the head’s own mass–inertia matrices
(mainly in the atlanto–occipital and atlanto–axial joints), while the ‘jolt’means ac-
tually hitting the head with some external mass–inertia matrices included in the
‘hitting’ SE(3)–jolt, or hitting some external static/massive body with the head
(e.g., the ground – gravitational effect, or the wall – inertial effect). Consequently,
the mass-less ‘jerk’ vector represents a (translational+rotational) non-collision effect
that can cause only weaker brain injuries, while the inertial ‘jolt’ co-vector represents
a (translational+rotational) collision effect that can cause hard brain injuries.

6.2.5 Brain’s dislocations and disclinations caused by the SE(3)−jolt

Recall from introduction that for mild TBI, the best injury predictor is considered
to be the product of brain’s strain and strain rate, which is the standard isotropic
viscoelastic continuum concept. To improve this standard concept, in this subsec-
tion, we consider human brain as a 3D anisotropic multipolar Cosserat viscoelastic
continuum, exhibiting coupled–stress–strain elastic properties. This non-standard
continuum model is suitable for analyzing plastic (irreversible) deformations and
fracture mechanics in multi-layered materials with microstructure (in which slips
and bending of layers introduces additional degrees of freedom, non-existent in the
standard continuum models.
The SE(3)−jolt (Ḟ, Ṫ) causes two types of brain’s rapid discontinuous deforma-
tions:

1. The Newton jolt Ḟ can cause micro-translational dislocations, or discontinuities

in the Cosserat translations;
2. The Euler jolt Ṫ can cause micro-rotational disclinations, or discontinuities in
the Cosserat rotations.

69
 To precisely define brain’s dislocations and disclinations, caused by the SE(3)−jolt
(Ḟ, Ṫ), we first define the coordinate co-frame, i.e., the set of basis 1–forms {dxi },
given in local coordinates xi = (x1 , x2 , x3 ) = (x, y, z), attached to brain’s center-
of-mass. Then, in the coordinate co-frame {dxi } we introduce the following set of
brain’s plastic–deformation–related SE(3)−based differential p−forms21 :

the dislocation current 1–form, J = Ji dxi ;

the dislocation density 2–form, α = 21 αij dxi ∧ dxj ;
the disclination current 2–form, S = 21 Sij dxi ∧ dxj ; and
the disclination density 3–form, Q = 3!1 Qijk dxi ∧ dxj ∧ dxk ,
where ∧ denotes the exterior wedge–product. These four SE(3)−based differen-
tial forms satisfy the following set of continuity equations:

α̇ = −dJ − S, (11)
Q̇ = −dS, (12)
dα = Q, (13)
dQ = 0, (14)

where d denotes the exterior derivative.

In components, the simplest, fourth equation (14), representing the Bianchi iden-
21
Differential p−forms are totally skew-symmetric covariant tensors, defined using the exterior
wedge–product and exterior derivative. The proper definition of exterior derivative d for a p−form
β on a smooth manifold M , includes the Poincaré lemma: d(dβ) = 0, and validates the general
Stokes formula Z Z
β= dβ,
∂M M
where M is a p−dimensional manifold with a boundary and ∂M is its (p−1)−dimensional boundary,
while the integrals have appropriate dimensions.
A p−form β is called closed if its exterior derivative is equal to zero,

dβ = 0.

From this condition one can see that the closed form (the kernel of the exterior derivative operator
d) is conserved quantity. Therefore, closed p−forms possess certain invariant properties, physically
corresponding to the conservation laws.
A p−form β that is an exterior derivative of some (p − 1)−form α,

β = dα,

is called exact (the image of the exterior derivative operator d). By Poincaré lemma, exact forms
prove to be closed automatically,
dβ = d(dα) = 0.
This lemma is the foundation of the de Rham cohomology theory

70
tity, can be rewritten as

dQ = ∂l Q[ijk] dxl ∧ dxi ∧ dxj ∧ dxk = 0,

where ∂i ≡ ∂/∂xi , while θ[ij...] denotes the skew-symmetric part of θij....

Similarly, the third equation (13) in components reads
1
Qijk dxi ∧ dxj ∧ dxk = ∂k α[ij] dxk ∧ dxi ∧ dxj , or
3!
Qijk = −6∂k α[ij] .

The second equation (12) in components reads

1
Q̇ijk dxi ∧ dxj ∧ dxk = −∂k S[ij] dxk ∧ dxi ∧ dxj , or
3!
Q̇ijk = 6∂k S[ij] .

Finally, the first equation (11) in components reads

1 1
α̇ij dxi ∧ dxj = (∂j Ji − Sij ) dxi ∧ dxj , or
2 2
α̇ij = 2∂j Ji − Sij .

In words, we have:

• The 2–form equation (11) defines the time derivative α̇ = 12 α̇ij dxi ∧ dxj of the
dislocation density α as the (negative) sum of the disclination current S and
the curl of the dislocation current J.

• The 3–form equation (12) states that the time derivative Q̇ = 3!1 Q̇ijk dxi ∧dxj ∧
dxk of the disclination density Q is the (negative) divergence of the disclination
current S.

• The 3–form equation (13) defines the disclination density Q as the divergence
of the dislocation density α, that is, Q is the exact 3–form.

• The Bianchi identity (14) follows from equation (13) by Poincaré lemma and
states that the disclination density Q is conserved quantity, that is, Q is the
closed 3–form. Also, every 4–form in 3D space is zero.

From these equations, we can derive two important conclusions:

71
 1. Being the derivatives of the dislocations, brain’s disclinations are higher–order
tensors, and thus more complex quantities, which means that they present a
higher risk for the severe TBI than dislocations — a fact which is supported
by the literature (see review of existing TBI–models given in Introduction of
[28]).

2. Brain’s dislocations and disclinations are mutually coupled by the underlaying

SE(3)−group, which means that we cannot separately analyze translational
and rotational TBIs — a fact which is not supported by the literature.

For more medical details and references, see [28].

References
[1] Arnold, V.I., Mathematical Methods of Classical Mechanics. Springer, New
York, (1978)

[2] Abraham, R., Marsden, J., Foundations of Mechanics. Benjamin, Reading, MA,
(1978)

[3] Abraham, R., Marsden, J., Ratiu, T., Manifolds, Tensor Analysis and Applica-
tions. Springer, New York, (1988)

[4] Marsden, J.E., Ratiu, T.S., Introduction to Mechanics and Symmetry, A Basic
Exposition of Classical Mechanical Systems. (2nd ed), Springer, New York,
(1999)

[5] Ivancevic, V. Symplectic Rotational Geometry in Human Biomechanics. SIAM

Rev. 46(3), 455–474, (2004)

[6] Ivancevic, V., Ivancevic, T., Human-Like Biomechanics, A Unified Mathe-

matical Approach to Human Biomechanics and Humanoid Robotics. Springer,
Berlin, (2005)

[7] Ivancevic, V., Ivancevic, T., Natural Biodynamics. World Scientific, Singapore
(2006)

[8] Ivancevic, V., Ivancevic, T., Geometrical Dynamics of Complex Systems, A

Unified Modelling Approach to Physics, Control, Biomechanics, Neurodynamics
and Psycho-Socio-Economical Dynamics. Springer, Dordrecht, (2006)

[9] Ivancevic, V., Ivancevic, T., Applied Differential Geometry, A Modern Intro-
duction. World Scientific, Singapore, (2007)

72
[10] Switzer, R.K., Algebraic Topology – Homology and Homotopy. (in Classics in
Mathematics), Springer, New York, (1975)

[11] de Rham, G., Differentiable Manifolds. Springer, Berlin, (1984)

[12] Lang, S., Fundamentals of Differential Geometry. Graduate Texts in Mathe-

matics, Springer, New York, (1999)

[13] Lang, S., Introduction to Differentiable Manifolds (2nd ed.). Graduate Texts in
Mathematics, Springer, New York, (2002)

[14] Dieudonne, J.A., Foundations of Modern Analysis (in four volumes). Academic
Press, New York, (1969)

[15] Dieudonne, J.A., A History of Algebraic and Differential Topology 1900-1960.

Birkh´’auser, Basel, (1988)

[16] Spivak, M., Calculus on Manifolds, A Modern Approach to Classical Theorems

of Advanced Calculus. HarperCollins Publishers, (1965)

[17] Spivak, M., A comprehensive introduction to differential geometry, Vol.I-V,

Publish or Perish Inc., Berkeley, (1970-75)

[18] Choquet-Bruhat, Y., DeWitt-Morete, C., Analysis, Manifolds and Physics (2nd
ed). North-Holland, Amsterdam, (1982)

[19] Choquet-Bruhat, Y., DeWitt-Morete, C., Analysis, Manifolds and Physics, Part
II, 92 Applications (rev. ed). North-Holland, Amsterdam, (2000)

[20] Bott, R., Tu, L.W., Differential Forms in Algebraic Topology. Graduate Texts
in Mathematics, Springer, New York, (1982)

[21] Chevalley, C., Theory of Lie groups, Princeton Univ. Press, Princeton, (1946)

[22] Helgason, S., Differential Geometry, Lie Groups and Symmetric Spaces. (2nd
ed.) American Mathematical Society, Providence, RI, (2001)

[23] Gilmore, R., Lie Groups, Lie Algebras and Some of their Applications (2nd
ed.), Dover, (2002)

[24] Fulton, W., Harris, J., Representation theory. A first course, Graduate Texts
in Mathematics, Springer, New York, (1991)

[25] Bourbaki, N., Elements of Mathematics, Lie Groups and Lie Algebras, Springer,
(2002)

73
[26] Conway, J.H., Curtis, R.T., Norton, S.P., Parker, R.A., Wilson, R.A., Atlas
of Finite Groups: Maximal Subgroups and Ordinary Characters for Simple
Groups. Clarendon Press, Oxford, (1985)

[27] Schafer, R.D., An Introduction to Nonassociative Algebras. Dover, New York,

(1996)

[28] V.G. Ivancevic, New mechanics of traumatic brain injury, Cogn. Neurodyn.
3:281-293, (2009)
http://www.springerlink.com/content/p27023577564202h/?p=4351a9d0d76a4fd4b45d6720dad056f3&pi=

[29] V.G. Ivancevic, New mechanics of spinal injury, IJAM, 1(2): 387–401, (2009)
http://www.worldscinet.com/ijam/01/0102/S1758825109000174.html

[30] V.G. Ivancevic, New mechanics of generic musculo-skeletal injury, BRL,

4(3):273–287, (2009)
http://www.worldscinet.com/brl/04/0403/S1793048009001022.html

74