Exceptional Lie Groups Explained

Exceptional Lie groups
arXiv:0902.0431v1 [math.DG] 3 Feb 2009
Ichiro Yokota
Preface
In the end of 19 century, W. Killing and E. Cartan classified the complex simple
Lie algebras, called An , Bn , Cn , Dn (classical type) and G2 , F4 , E6 , E7 , E8 (exceptional
type). These simple Lie algebras and the corresponding compact simple Lie groups
have offered many subjects in mathematicians. Especially, exceptional Lie groups are
very wonderful and interesting miracle in Lie group theory.
Now, in the present book, we describe simply connected compact exceptional
simple Lie groups G2 , F4 , E6 , E7 , E8 , in very elementary way. The contents are given
as follows. We first construct all simply connected compact exceptional Lie groups
G concretely. Next, we find all involutive automorphisms of G, and determine the
group structures of the fixed points subgroup G by . Note that they correspond to
classification of all irreducible compact symmetric spaces G/G of exceptional type,
and that they also correspond to classification of all non-compact exceptional simple
Lie groups. Finally, we determined the group structures of the maximal subgroups of
maximal rank. At any rate, we would like this book to be used in mathematics and
physics.
The author thanks K. Abe, K. Mituishi, T. Miyasaka, T. Miyashita, T. Sato, O.
Shukuzawa, K. Takeuchi and O. Yasukura for their advices and encouragements.
Ichiro Yokota
Contents
I.
Exceptional Lie group G2
1.1. Cayley algebra C

1.2. Compact exceptional Lie group G2
1.3. Outer automorphism of the Lie algebra d4
1.4. Lie algebra g2 of G2
1.5. Lie subalgebra su(3) of g2
1.6. Simplicity of g2 C
1.7. Killing form of g2 C
1.8. Roots of g2 C
1.9. Automorphism w of order 3 and subgroup SU (3) of G2
1.10. Involution and subgroup (Sp(1) Sp(1))/Z 2 of G2
1.11. Center z(G2 ) of G2
1.12.
1.13.
1.14.
1.15.
1.16.
Complex exceptional Lie group G2 C

Non-compact exceptional Lie group G2(2) of type G2
Principle of triality in SO(8)
Spinor group Spin(7)
Spinor group Spin(8)
II.
21
22
23
26
28
Exceptional Lie group F4
2.1. Exceptional Jordan algebra J

2.2. Compact exceptional Lie group F4
2.3. Lie algebra f4 of F4
31
33
36
2.4. Simplicity of f4 C
2.5. Killing form of f4
1
2
3
8
9
11
13
14
16
19
21
41
2.6. Roots of f4
2.7. Subgroup Spin(9) of F4
2.8. Connectedness of F4
2.9. Involution and subgroup Spin(9) of F4
2.10. Center z(F4 ) of F4
2.11. Involution and subgroup (Sp(1) Sp(3))/Z 2 of F4
2.12. Automorphism w of order 3 and subgroup (SU (3) SU (3))/Z 3
of F4
C
2.13. Complex exceptional Lie group F4

2.14. Non-compact exceptional Lie groups F4(4) and F4(20) of type F4
45
47
51
55
57
58
59
63
65
66
III.
Exceptional Lie group E6
3.1. Compact exceptional Lie group E6

3.2. Lie algebra e6 of E6
3.3. Simplicity of e6 C
3.4. Element A B of e6 C
3.5. Killing form of e6 C
3.6. Roots of e6 C
3.7. Involution and subgroup F4 of E6
3.8. Connectedness of E6
3.9. Center z(E6 ) of E6
3.10. Involution and subgroup (U (1) Spin(10))/Z 4 of E6
3.11. Involution and subgroup (Sp(1) SU (6))/Z 2 of E6
3.12. Involution and subgroup Sp(4)/Z 2 of E6
3.13. Automorphism w of order 3 and subgroup (SU (3) SU (3)
SU (3))/Z 3 of E6
C
3.14. Complex exceptional Lie group E6

3.15. Non-compact exceptional Lie groups E6(6) , E6(2) , E6(14) and
E6(26) of type E6
IV.
68
68
70
71
73
75
82
82
85
86
90
95
100
105
105
4.1. Freudenthal vector space PC

4.6. Roots of e7 C
4.7. Subgroups E6 and U (1) of E7
4.10. Involution and subgroup (U (1) E6 )/Z 3 of E7
4.11. Involution and subgroup (SU (2) Spin(12))/Z 2 of E7
4.12. Involution and subgroup SU (8)/Z 2 of E7
4.13. Automorphism w of order 3 and subgroup (SU (3) SU (6))/Z 3
of E7
C
4.14. Complex exceptional Lie group E7

4.15. Non-compact exceptional Lie groups E7(7) , E7(5) and E7(25)
of type E7
106
108
108
114
116
118
124
126
129
130
133
142
146
155
155
V.
5.1. Lie algebra e8 C

Complex exceptional Lie group E8 C
Compact exceptional Lie group E8
Roots of e8 C
Involution and subgroup (SU (2) E7 )/Z 2 of E8
e and subgroup Ss(16) of E8
5.8. Involution
5.10. Automorphism w of order 3 and subgroup(SU (3) E6 )/Z 3 of
E8
5.11. Automorphism w3 of order 3 and subgroup SU (9)/Z 3 of E8
5.12. Automorphism z5 of order 5 and subgroup(SU (5) SU (5))/Z 5
of E8
5.13. Non-compact exceptional Lie groups E8(8) and E8(24) of E8
5.4.
5.5.
5.6.
5.7.
157
157
159
160
161
163
173
176
184
184
186
192
199
Notation
R, C = R Re1 , H = R Re1 Re2 Re3 denote the fields of real, complex,
quaternion numbers, respectively.
For R-vector space V , its complexification {u + iv | u, v V } is denoted by V C .
The complex conjugation in V C is denoted by :
(u + iv) = u iv.
RC is briefly denoted by C.
For K-vector space V (K = R, C, C), IsoK (V ) denotes all K-linear isomorphisms
of V . For a K-linear mapping f of V , Vf denotes {v V | f (v) = v}.
For K-vector spaces V, W (K = R, C), HomK (V, W ) denotes all K-homomorphism f : V W . HomK (V, V ) is briefly denoted by HomK (V ).
Let G be a group and an automorphism of G. Then G denotes {g G | (g) =
g}. For s G, Gs denotes {g G | sgs1 = g}.
For topological spaces X, Y , X Y denotes that X and Y are homeomorphic.
For groups G, G , G
= G denotes that G and G are isomorphic as groups. Isomorphic two groups G, G are often identified: G = G .
M (n, K) denotes all n n matrices with entries in K.
E = diag(1, , 1) M (n, K) is the unit matrix.
For A M (n, K), t A denotes the transposed matrix of A and A denotes the
conjugate transposed matrix of A: A = t A.
O(n) = {A M (n, R) | t AA = E} (orthogonal group),
SO(n) = {A O(n) | detA = 1} (spcecial orthogonal group),
U (n) = {A M (n, C) | A A = E} or {A M (n, C) | (t A)A = E} (unitary

group),
SU (n) = {A U (n) | detA = 1} (speccial unitary group),
Sp(n) = {A M (n, H) | A A = E} (symplectic group).
For a Lie group G, its Lie algebra is denoted by the correspoding small Germann
letter g. For example, su(n) is the Lie algebra of the special unitary group SU (n).
Special notations
,
g,
w.
Exceptional Lie group G2

1.1. Cayley algebra C
We denote the division Cayley algebra by C. We now explain this algebra. Consider an 8 dimensional R-vector space with basis {e0 = 1, e1 , e2 , e3 , e4 , e5 , e6 , e7 } and
define a multiplication between them as follows. In the figure below, the multiplication
between e1 , e2 , e3 is defined as
e1 e2 = e3 ,
e2 e3 = e1 ,
e3 e1 = e2 ,
and defined similarly on the other lines. For example, e1 e6 = e7 , e4 e7 = e3 etc. We

regard that e2 , e5 , e7 are also collinear, for example, e5 e7 = e2 . e0 = 1 is the unit of
the multiplication and assume
ei 2 = 1, i 6= 0,
ei ej = ej ei , i 6= j, i 6= 0, j 6= 0,
and the distributive law. Thus C has a multiplication. x1, x R is briefly denoted
by x. In C, the conjugate x, an inner product (x, y), the length |x| and the real part
R(x) are defined respectively by
e1
e7
e2
e4
e3
x0 +
7
X
i=1
xi ei = x0
e6
e5
7
X
i=1
p
|x| = (x, x),
xi ei ,
7
X
i=0
xi ei ,
7
X
i=0
7
X
xi yi ,
yi ei =
i=0
7

X
R x0 +
xi ei = x0 .
i=1
x
by x1 , then we have xx1 = x1 x = 1, and C
|x|2
satisfies all axioms of a field except the associative law x(yz) = (xy)z. Of course
the commutative law xy = yx does not hold. Since the associative law does not
For x C, x 6= 0, we denote
hold, calculations in C are complicated, however the following relations hold. (See
Freudenthal [6] or Yokota [58]). For a, b, x, y C, we have
1
2
3
4
5
(xy, xy) = (x, x)(y, y), |xy| = |x||y|.

(ax, ay) = (a, a)(x, y) = (xa, ya).
(ax, by) + (bx, ay) = 2(a, b)(x, y),
(ax, y) = (x, ay), (xa, y) = (x, ya).
x = x, x + y = x + y, xy = y x.
1
1
6 (x, y) = (y, x) = (xy + yx) = (xy + yx), xx = xx = |x|2 .
2
2
7 a(ax) = (aa)x, a(xa) = (ax)a, x(aa) = (xa)a.
a(ax) = (aa)x, a(xa) = (ax)a, x(aa) = (xa)a,
8 b(ax) + a(bx) = 2(a, b)x = (xa)b + (xb)a.

9 We use a notation {x, y, z} = (xy)z x(yz), called the associator of x, y, z.
Then, we have
{x, y, a} = {y, a, x} = {a, x, y} = {y, x, a} = {x, a, y} = {a, y, x}.
For example, we have
(ax)y + x(ya) = a(xy) + (xy)a,

(xa)y + (xy)a = x(ay) + x(ya),
(ax)y + (xa)y = a(xy) + x(ay).
10 (ax)(ya) = a(xy)a (Moufangs formula).
11 R(xy) = R(yx), R(x(yz)) = R(y(zx)) = R(z(xy)) (= R(xyz)).
For an orthonormal basis {1, a1 , a2 , , a7 } of C with respect to the inner product
(x, y), the following 12.1 12.3 hold.
12.1 ai (aj x) = aj (ai x), in particular, ai aj = aj ai , i 6= j.
12.2 ai (ai x) = x, in particular, ai 2 = 1.
12.3 ai (aj ak ) = aj (ak ai ) = ak (ai aj ), i, j, k are distinct.
1.2. Compact exceptional Lie group G2
Definition. The group G2 is defined to be the automorphism group of the Cayley
algebra C:
G2 = { IsoR (C) | (xy) = (x)(y)}.
Lemma 1.2.1. For G2 , we have
(x, y) = (x, y),
Proof. For G2 and x C, we have
x = x.
x, y C.
To prove this, it is sufficient to show

1 = 1, ei = ei ,
i = 1, 2, , 7.
From (1)(1) = (1 1) = 1 and 1 6= 0, we have 1 = 1, while from (ei )(ei ) =

(ei ei ) = (1) = 1 = 1 for i 6= 0, we get ei = ei . Now,
1
1
((x)(y) + (y)(x)) = ((x)(y) + (y)(x))
2
2

1
= (xy + yx) = ((x, y)) = (x, y).
2
(x, y) =
Theorem 1.2.2. G2 is a compact Lie group.
Proof. G2 is a compact Lie group as a closed subgroup of the orthogonal group

O(8) = O(C) = { IsoR (C) | (x, y) = (x, y)}.
Remark. Since 1 = 1 for G2 , G2 is a subgroup of the orthogonal group
O(7) = { O(C) | 1 = 1}, that is, G2 O(7).
1.3. Outer automorphisms of Lie algebra D4
In order to study the Lie algebra g2 of the group G2 , we consider the Lie algebra
D4 = so(8) = so(C) = {D HomR (C) | (Dx, y) + (x, Dy) = 0}
of the Lie group SO(8).
We define R-linear mappings Gij : C C, i, j = 0, 1, , 7, i 6= j satisfying
Gij ej = ei ,
Gij ei = ej ,
Gij ek = 0, k 6= i, j.
Then Gij D4 and {Gij | 0 i < j 7} forms an R-basis of D4 . Furthermore we

define R-linear mappings Fij : C C, i, j = 0, 1, , 7, i 6= j by
Fij x =
1
ei (ej x),
2
x C.
Lemma 1.3.1. For i, j = 0, 1, , 7, i 6= j, we have Fij D4 , and when i < j,

Fij is expressed in terms of Gij as follows.
2F01
2F
23
2F
45
2F67
=
=
=
=
2F02
2F
13
2F
46
2F57
G01 + G23 + G45 + G67

G01 + G23 G45 G67
G01 G23 + G45 G67
G01 G23 G45 + G67 ,
= G02 G13 G46 + G57

= G02 + G13 G46 + G57
= G02 G13 + G46 + G57
= G02 + G13 + G46 + G57 ,

2F03
2F
12
2F
47
2F56
2F05
2F
14
2F
27
2F36
2F04 = G04 G15 + G26 G37
2F = G + G + G G
15
04
15
26
37
2F
=
G
+
G
+
G
+
G
26
04
15
26
37
2F37 = G04 G15 + G26 + G37 ,
2F06 = G06 G17 G24 + G35

= G05 + G14 G27 G36
2F17 = G06 + G17 G24 + G35

= G05 + G14 + G27 + G36
2F24 = G06 G17 + G24 + G35

= G05 + G14 + G27 G36
2F35 = G06 + G17 + G24 + G35 ,

= G05 + G14 G27 + G36 ,
2F
=
G
+
G
+
G25 + G34
07
07
16
2F = G + G G G
16
07
16
25
34
2F25 = G07 G16 + G25 G34
2F34 = G07 G16 G25 + G34 .
=
=
=
=
G03 + G12 + G47 + G56

G03 + G12 G47 G56
G03 G12 + G47 G56
G03 G12 G47 + G56 ,
In particular, {Fij | 0 i < j 7} forms an R-basis of D4 .
Definition. We define R-linear mappings , , : D4 D4 respectively by

(D)x = Dx,
x C,
(Gij ) = Fij ,
i, j = 0, 1, , 7, i 6= j,
= .
Lemma 1.3.2. The mappings , , are automorphisms of the Lie algebra D4 :

, , Aut(D4 ).
Proof. Since 2 = 1 , is an R-linear isomorphism of D4 . We have
[D1 , D2 ]x = (D1 )(D2 x) (D2 )(D1 x) = D1 (D2 x) D2 (D1 x)
= D1 D2 x D2 D1 x = (D1 D2 D2 D1 )x = [D1 , D2 ]x,
x C.
Hence Aut(D4 ). Since maps the R-basis {Gij | 0 i < j 7} to the R-basis
{Fij | 0 i < j 7} of D4 , is an R-linear isomorphism. To show that is an
automorphism of D4 , it is sufficient to show that
[Gij , Gkl ] = [Gij , Gkl ],
which in turn would follow from the relations
[Fij , Fjk ] = Fik ,
[Fij , Fkl ] = 0,
i, j, k are distinct,
i, j, k, l are distinct.
In the following calculations, let i, j, k, l be all distinct and i, j, k, l 6= 0. For x C,

1
1
[Fi0 , F0k ]x = (Fi0 F0k F0k Fi0 )x = ei (ek x) + ek (ei x)
4
4
1
= ek (ei x) = Fik x,
2
4
[Fi0 , Fkl ]x = (Fi0 Fkl Fkl Fi0 )x =
1
1
ei (el (ek x)) el (ek (ei x))
4
4
1
1
= el (ei (ek x)) + el (ei (ek x)) = 0,
4
4
1
1
[Fij , Fjk ]x = (Fij Fjk Fjk Fij )x = ej (ei (ek (ej x))) ek (ej (ej (ei x)))
4
4
1
1
1
1
= ei (ek (ej (ej x))) + ek (ei x) = ei (ek x) + ek (ei x)
4
4
4
4
1
= ek (ei x) = Fik x,
2
1
1
[Fij , Fkl ]x = (Fij Fkl Fkl Fij )x = ej (ei (el (ek x))) el (ek (ej (ei x)))
4
4
1
1
= ej (ek (ei (el x))) ej (el (ek (ei x))) = = 0.
4
4
Hence Aut(D4 ). Finally, since = , we have Aut(D4 ).
Definition. For a C, we define R-linear mappings La , Ra , Ta : C C respectively by
La x = ax,
Ra x = xa,
Ta x = ax + xa = (La + Ra )x,
x C.
Hereafter, we denote by C0 the subset {a C | a = a} of C.

Lemma 1.3.3. For a C0 , we have
(1)
(2)
La , Ra , Ta D4 .
La = Ra , Ra = La ,
(3)
La = Ta ,
(4)
La = Ra ,
Ra = Ra ,
Ra = Ta ,
Ta = Ta .
Ta = La .
Ta = La .
Proof. (1) (La x, y) = (ax, y) = (x, ay) = (x, ay) = (x, La y), x, y C. Hence
La D4 . Similarly, Ra D4 and Ta = La + Ra D4 .
(2) (La )x = La x = ax = xa = xa = Ra x, x C. Hence La = Ra . The

others can be similarly obtained.
(3) It is sufficient to show that these relations hold for a = ei , i = 1, , 7. We

have
Lei = 2Fi0 , Tei = 2Gi0 .
Indeed,
It follows that
Lei x = ei x = 2Fi0 x, x C,
x=1
2ei ,
2,
x = ei
Tei x = ei x + xei =
0,
x = ej , j 6= 0, i.
Lei = (2Fi0 ) = 2F0i = 2G0i (Lemma 1.3.1) = 2Gi0 = Tei ,

5
Tei = (2Gi0 ) = 2Fi0 = Lei ,

Ra = (Ta La ) = La Ta = Ra .
(4) follows immediately from (2) and (3), since = .
Lemma 1.3.4. The Lie algebra D4 is generated by {La | a C0 } (by taking at
most one time the Lie bracket [ , ]), that is, any D D4 is expressed by
X
[Lbi , Lci ], a, bi , ci C0 .
D = La +
i
Proof. Let D be the Lie subalgebra of D4 generated by {La | a C0 }. Since

Lei = 2Fi0 and [Lei , Lej ] = 4[Fi0 , Fj0 ] = 4Fij , i 6= 0, j 6= 0, i 6= j, we see that D
contains Fij , i, j = 0, 1, , 7, i 6= j. Since {Fij , i < j } is an R-basis of D4 (Lemma
1.3.1), D coincides with D4 .
Theorem 1.3.5. In the automorphism group Aut(D4 ) of D4 , the subgroup S3
generated by and is isomorphic to the symmetric group S3 of degree 3. Furthermore, , , have the following relations.
2 = 1,
2 = 1,
3 = 1,
= .
Proof. Since {La | a C0 } generates D4 (Lemma 1.3.4), it is sufficient to check

= 1, 2 = 1, 3 = 1 etc. for La . However, these relations follow from Lemma
1.3.3.(2),(3),(4). The mapping f : S3 S3 defined by the correspondence

1 2 3
1 2 3
1 2 3
1
,
,
,
1 2 3
2 1 3
3 2 1

1 2 3
1 2 3
1 2 3
, 2
,
2 3 1
3 1 2
1 3 2
2
gives an isomorphism as groups.

Theorem 1.3.6. (Principle of infinitesimal triality in D4 ). For any D1
D4 , there exist D2 , D3 D4 such that
(D1 x)y + x(D2 y) = D3 (xy),
x, y C.
Also such D2 , D3 are uniquely determined for D1 and we have

D2 = D1 ,
D3 = D1 .
Proof. If a C0 , then La , Ra , Ta D4 (Lemma 1.3.3.(1)) and the equality

(ax)y + x(ya) = a(xy) + (xy)a implies that
(La x)y + x(Ra y) = Ta (xy),
x, y C.
(i)
Similarly, for b C0 , we have

(Lb x)y + x(Rb y) = Tb (xy),
x, y C.
(ii)
Applying Ta on (ii) and using (i), we have

(La Lb x)y + (Lb x)(Ra y) + (La x)(Rb y) + x(Ra Rb y) = Ta Tb (xy).
Exchanging a for b, and subtracting from the above, we get
([La , Lb ]x)y + x([Ra , Rb ]y) = [Ta , Tb ](xy).
Since D1 D4 is expressed as
D1 = L a +
[Lb , Lc ],
(iii)
a, b, c C0
P
P
(Lemma 1.3.4), putting D2 = Ra + [Rb , Rc ], D3 = Ta + [Tb , Tc ] and using (i),
(iii), we obtain
(D1 x)y + x(D2 y) = D3 (xy), x, y C.
Next, we shall show that D2 and D3 are determined uniquely for D1 . To prove this,
it is sufficient to show for D1 = 0 that D2 = D3 = 0. Now, suppose
x(D2 y) = D3 (xy),
x, y C.
Putting x = 1, we have D2 y = D3 y, so that D2 = D3 (= D). Therefore

x(Dy) = D(xy),
x, y C.
(iv)
Putting D1 = p, we have 2(p, 1) = (p, 1)+(1, p) = (D1, 1)+(1, D1) = 0, which implies
p C0 . Furthermore, putting y = 1 in (iv), we have xp = Dx, so (iv) becomes
x(yp) = (xy)p,
for all x, y C.
We therefore see that p R, so that p = 0 since p C0 . Hence Dx = xp = 0,

and so D = 0. This proves the uniqueness. Finally, if we express D1 D4 as
P
P
P
D1 = La + [Lb , Lc ], a, b, c C0 , then D2 = Ra + [Rb , Rc ], D3 = Ta + [Tb , Tc ]
from the arguments above and the uniqueness. Hence we have D2 = D1 , D3 = D1
(Lemma 1.3.3).
Lemma 1.3.7. For D1 , D2 , D3 D4 , the relation
(D1 x)y + x(D2 y) = (D3 )(xy),
x, y C
(D2 x)y + x(D3 y) = (D1 )(xy),

(D3 x)y + x(D1 y) = (D2 )(xy),
x, y C,
x, y C.
implies
Proof. For D2 D4 , there exist D3 , D1 D4 such that

(D2 x)y + x(D3 y) = (D1 )(xy),
x, y C,
and D3 = D2 , D1 = D2 (Theorem 1.3.6). The assumption of the lemma is

D2 = D1 , D3 = D1 (Theorem 1.3.6). Hence, using Theorem 1.3.5, we have
D3 = D2 = D1 = 1 D1 = D1 = D3 = D3 ,
D1 = D2 = 1 D2 = D1 .
1.4. Lie algebra g2 of G2

Theorem 1.4.1. The Lie algebra g2 of the Lie group G2 is given by
g2 = {D HomR (C) | D(xy) = (Dx)y + x(Dy)}.
Proof. If D HomR (C) satisfies (exp tD)(xy) = ((exp tD)x)((exp tD)y), t R,
then by differentiating with respect to t and putting t = 0, we get D(xy) = (Dx)y +
x(Dy). Conversely, if D HomR (C) satisfies D(xy) = (Dx)y + x(Dy), then it is not
difficult to verify that = exp tD satisfies (xy) = (x)(y).
In order to study the Lie algebra g2 , we need the following Lie algebra b3 = so(7)
of SO(7):
b3 = {D D4 | D1 = 0}
= {D D4 | D = D} = {D D4 | D = D}.
Lemma 1.4.2. g2 is a Lie subalgebra of b3 . Moreover we have
g2 = {D D4 | D = D, D = D}
= {D D4 | D = D, S3 }
= {D b3 | D = D}.
Proof. Let D g2 . Putting x = y = 1 in D(xy) = (Dx)y + x(Dy), we get

D1 = 0. We next show that
Dx C0 , x C.
If i 6= 0, then (Dei )ei + ei (Dei ) = D(ei ei ) = D(1) = 0, and so Dei C0 . Together
with D1 = 0, we have Dx C0 , x C. Now, note that
xy + yx = (xy + yx) = 2(x, y),
x, y C0 .
Applying D on the relation above, we have (Dx)y + x(Dy) + (Dy)x + y(Dx) = 0.

Since Dx, Dy C0 , we get
(Dx, y) + (x, Dy) = 0,
8
x, y C.
Hence g2 b3 . The relation in the lemma follows easily from Theorem 1.3.5 and
Theorem 1.3.6.
Theorem 1.4.3. Any element of g2 is expressed by the sum of elements of the
following seven types.
G23 + G45 + G67 ,
G12 + G47 + G56 ,
G14 G27 G36 ,
G13 G46 + G57 ,
G15 + G26 G37 ,
G17 G24 + G35 ,
, , R
++ =0
G16 + G25 + G34 .
Conversely, the above seven elements belong to g2 . In particular, the dimension of g2

is 14:
dim g2 = 14.
P
Proof. Let D g2 . Since D b3 (Lemma 1.4.2), D = 0<i<j ij Gij , ij R.
The condition D = D (Lemma 1.4.2) implies that
X
X
ij Gij .
ij Fij =
0<i<j
0<i<j
Applying the above element on 1 C, we have

X
ij ej ei = 0.
0<i<j
Replacing ej ei by ek and comparing the coefficient of each e1 , e2 , , e7 , we have

23 + 45 + 67 = 0,
13 46 + 57 = 0,
12 + 47 + 56 = 0,
15 + 26 37 = 0,
14 27 36 = 0,
17 24 + 35 = 0,
16 + 25 + 34 = 0,
from which the first result follows. Conversely, any of these seven elements D obviously belongs to b3 and the condition D = D (Lemma 1.4.2) is verified from the
table of Lemma 1.3.1.
1.5. Lie subalgebra su(3) of g2
The Cayley algebra C naturally contains the field C of complex numbers as
C = {x0 + x1 e1 | xi R}.
Any element x C is expressed by
x = x0 + x1 e1 + x2 e2 + x3 e3 + x4 e4 + x5 e5 + x6 e6 + x7 e7
(xi R)
= (x0 + x1 e1 ) + (x2 + x3 e1 )e2 + (x4 + x5 e1 )e4 + (x6 + x7 e1 )e6 ,

9
that is,
x = a + m1 e 2 + m2 e 4 + m3 e 6 ,
a, mi C.
We associate such element x of C with the element of C C 3
m1
a + m2 .
m3
In C C 3 , we define a multiplication, an inner product ( , ) and a conjugation

respectively by
(a + m)(b + n) = (ab hm, ni) + (an + bm m n),
(a + m, b + n) = (a, b) + (m, n),
a + m = a m,
where the real valued symmetric inner product (m, n), the Hermitian inner product
hm, ni and the exterior product m n are usually defined respectively by
3
X
X
1
mi n i ,
(mi , ni ), hm, ni =
(m n + n m) =
2
i=1
i=1
m2 n 3 n 2 m3
m n = m3 n 1 n 3 m1
m1 n 2 n 1 m2

m1
n1
for m = m2 , n = n2 C 3 . Since these operations correspond to their
m3
n3
respective operations in C, hereafter, we identify C C 3 with C , that is,
(m, n) =
C C 3 = C.
We shall study the following subalgebra (g2 )e1 of g2 :
(g2 )e1 = {D g2 | De1 = 0}.
Theorem 1.5.1.
(g2 )e1
= su(3).
Proof. We define a mapping : su(3) = {D M (3, C) | D = D, tr(D) =

0} (g2 )e1 by
a + m C C 3 = C.
(D)(a + m) = Dm,
We first prove that (D) (g2 )e1 . For elements

e1 (E11 E22 ),
E13 E31 ,
e1 (E22 E33 ), E12 E21 , e1 (E12 + E21 ),

e1 (E13 + E31 ), E23 E32 , e1 (E23 + E32 )
10
of an R-basis of su(3) (where Ekl M (3, R) is the matrix with the (k, l)-entry is 1
and otherwise are 0), we have
(e1 (E11 E22 )) = G23 + G45 ,
(E12 E21 ) = G24 + G35 ,
(E13 E31 ) = G26 + G37 ,
(E23 E32 ) = G46 + G57 ,
(e1 (E22 E33 )) = G45 + G67 ,

(e1 (E12 + E21 )) = G25 + G34 ,
(e1 (E13 + E31 )) = G27 + G36 ,
(e1 (E23 + E32 )) = G47 + G56 .
Hence (D) g2 (Theorem 1.4.3). Clearly (D)e1 = 0, so that (D) (g2 )e1 .
Obviously : su(3) g2 is a homomorphism as Lie algebras and is injective, so
that we identify su(3) and (su(3). We set
S1 = 2G12 G47 G56 , S2 = 2G13 G46 + G57 ,
S3 = 2G14 + G27 + G36 , S4 = 2G15 + G26 G37 ,
S5 = 2G16 G25 G34 , S6 = 2G17 G24 + G35 ,
and let S be the R-vector subspace of g2 spanned by S1 , , S6 . Then we have the
following decomposition of g2 .
g2 = su(3) S.
Now, we shall show that : su(3) (g2 )e1 is onto. Let B (g2 )e1 . Denote
B=D+
6
X
xi Si ,
D su(3), xi R.
i=1
From the condition Be1 = 0, we have

2x1 e2 2x2 e3 2x3 e4 2x4 e5 2x5 e6 2x6 e7 = 0.
Hence x1 = = x6 = 0, so that B = D su(3). Thus the proof of Theorem 1.5.1 is
completed.
1.6. Simplicity of g2 C
Let CC = {x1 + ix2 | x1 , x2 C} be the complexification of the Cayley algebra C.
In the same manner as in C, we can also define in CC the multiplication xy, the inner
product (x, y) such that they satisfy properties 1 12.3 of Section 1.1 (except some
formulas about the length |x|). CC is called the complex Cayley algebra. CC has two
complex conjugations, namely,
x1 + ix2 = x1 + ix2 , (x1 + ix2 ) = x1 ix2 ,
xi C.
The complex conjugation is a complex-conjugate linear transformation of CC and

satisfies
(xy) = ( x)( y), x, y CC .
11
For a while, in the Lie algebra su(3) g2 , we use the following notations.
H1 = G23 + G45 , H2 = G45 + G67
L12 = G24 + G35 , L21 = G25 + G34 , L13 = G26 + G37 ,
L21 = G27 + G36 , L23 = G46 + G57 , L32 = G47 + G56 .
Lemma 1.6.1. The Lie brackets [D, S], D su(3), S S are given as follows.
H1
H2
L12
L21
L13
L31
L23
L32
S1
S2
0
S3
S4
S5
S6
0
0
S2
S3
S1 S4
0
S4
S4
S1
S3
S2
S6
0
S5
0
0 S5
0
S6
S4
S3
S3
S2
S1
0
0
S6
S5
S5
S6
0
0
S6
S5
0
0
0
0
S1
S2
S2 S1
S3
S4
S4 S3
Lemma 1.6.2. S is a su(3)-irreducible R-module, and hence we have

[su(3), S] = S.
Proof. Evidently S is a su(3)-R-module. Let W be a non-zero su(3)-invariant
R-snbmodule of S. If W contains some Sk , then, from the table of Lemma 1.6.1,
we can see that W contains all Sk , k = 1, 2, , 6, and hence W = S. Now, let
P6
S = k=1 xk Sk , xk R be a non-zero element of W and assume that x1 6= 0 (without
the loss of generality). Applying H1 on S, we have x1 S2 x2 S1 x3 S4 + x4 S3 W .
Next, applying L13 and L31 on it, we have
x1 S6 + x2 S5 W
(i) and
x1 S5 x2 S6 W
(ii)
Taking (ii)x2 (i)x1 , we have (x1 2 + x2 2 )S5 W . Since x1 2 + x2 2 6= 0, we have

S5 W and so W = S. Consequently the irreducibility of S is proved. Finally, since
[su(3), S] is a su(3)-invariant R-submodle of S, from the irreducibility of S, we have
[su(3), S] = S.
Theorem 1.6.3. The Lie algebra g2 C is simple and so g2 is also simple.
Proof. We shall prove that g2 is simple, because the proof of that g2 C is simple
is the same as the case of g2 . We use the decomposition of g2
g2 = su(3) S
(Theorem 1.5.1).
Let p : g2 su(3) and q : g2 S be projections of g2 = su(3) S. Now, let a be

a non-zero ideal of g2 . Then p(a) is an ideal of g2 . Indeed, if D p(a), then there
exists S S such that D + S a. For any D su(3), we have
a [D , D + S] = [D , D] + [D , S],
12
[D , S] S
(Lemma 1.6.1), hence [D , D] p(a).
We show that either su(3) a 6= {0} or S a 6= {0}. Assume that su(3) a = {0}
and S a = {0}. The mapping p|a : a su(3) is injective because S a = {0}.
Since p(a) is a non-zero ideal of su(3) and su(3) is simple, we have p(a) = su(3).
Hence dim a = dim p(a) = dim su(3) = 8. On the other hand, since su(3) a = {0},
q|a : a S is also injective, we have dim a dim S = 6. This leads to a contradiction.
We now consider the following two cases.
(1) Case su(3) a 6= {0}. From the simplicity of su(3), we have su(3) a = su(3),
hence a su(3). On the other hand, we have
a [a, S] [su(3), S] = S (Lemma 1.6.1).
Hence a su(3) S = g2 .
(2) Case Sa 6= {0}. Choose a non-zero element S Sa a. Under the actions

of su(3), we can see that S1 a (Lemma 1.6.1). Hence 0 6= 4H1 + 2H2 = [S1 , S2 ] a.
So this case can be reduced to the case (1). Thus we have a = g2 , which proves the
simplicity of g2 .
1.7. Killing form of g2 C
Lemma 1.7.1. In su(3) g2 , the Lie brackets between H1 and Lij , Lji of Section
1.6 are given by
[H1 , L12 ] = 2L21 , [H1 , L21 ] = 2L12 , [H1 , L13 ] = L31 ,
[H1 , L31 ] = L13 , [H1 , L23 ] = L32 ,
[H1 , L32 ] = L23 .
Theorem 1.7.2. The killing form B2 of the Lie algebra g2 C is given by
B2 (D1 , D2 ) = 4tr(D1 D2 ),
Di g2 C .
Proof. Since tr(D1 D2 ) is a g2 C -adjoint invariant bilinear form of g2 C and g2 C is

simple (Theorem 1.6.3), there exists k C such that
B2 (D1 , D2 ) = ktr(D1 D2 ).
To determine this k, let D1 = D2 = H1 . Then from
[H1 , [H1 , L12 ] ] = [H1 , 2L21 ] = 4L12 , [H1 , [H1 , L21 ] ] = [H1 , 2L12] = 4L21 ,
[H1 , [H1 , L13 ] ] = [H1 , L31 ] = L13 ,
[H1 , [H1 , L31 ] ] = [H1 , L13 ] = L31 ,
[H1 , [H1 , L23 ] ] = [H1 , L32 ] = L23 , [H1 , [H1 , L32 ] ] = [H1 , L23 ] = L32 ,
[H1 , [H1 , S1 ] ] = [H1 , S2 ] = S1 ,
[H1 , [H1 , S3 ] ] = [H1 , S4 ] = S3 ,
[H1 , [H1 , S5 ] ] = 0, [H1 , [H1 , S6 ] ] = 0
13
[H1 , [H1 , S2 ] ] = [H1 , S1 ] = S2 ,

[H1 , [H1 , S4 ] ] = [H1 , S3 ] = S4 ,
(Lemma 1.6.1), we have

B2 (H1 , H1 ) = tr((adH1 )2 ) = (4) 2 + (1) 8 = 16.
On the other hand,
H1 H1 e2 = H1 e3 = e2 ,
H1 H1 e3 = H1 e2 = e3 ,
H1 H1 e4 = H1 e5 = e4 , H1 H1 e5 = H1 e4 = e,
H1 H1 ei = 0 otherwise.
Hence tr(H1 H1 ) = (1) 4 = 4. Therefore k = 4.
1.8. Roots of g2 C
We recall the C-Lie isomorphism f : sl(3, C) su(3)C and the embedding :
su(3)C g2 C ,
f (A) = A tA,
(D)(a + m) = Dm,
1
(1 + ie1 ),
2
a + m C C (C 3 )C = CC ,
and we regard sl(3, C) as a subalgebra of g2 C under the composition of f and .

Further we know that the Lie algebra sl(3, C) has roots (k l ), 1 k < l 3
relative to the Cartan subalgebra
0
n 1 0
o

h = 0 2 0 i C, 1 + 2 + 3 = 0
0
0 3
of sl(3, C), and Ekl is a root vector associated with the root k l .
Theorem 1.8.1. The rank of the Lie algebra g2 C is 2. The roots of g2 C relative
to some Cartan subalgebra of g2 C are given by
(1 2 ),
(1 3 ),
(2 3 ),
1 ,
with 1 + 2 + 3 = 0.
Proof.
h = {i1 G23 i2 G45 i3 G67 g2 C | k C} sl(3, C) g2 C
is an abelian subalgebra of g2 C (it will be a Cartan subalgebra of g2 C ). The roots of

sl(3, C) is also roots of g2 C , so we have the table of roots and associated root vectors
as follows.
(1 2 ) : (G24 + G35 ) + i(G25 + G34 ),
(1 3 ) : (G26 + G37 ) + i(G27 + G36 ),
(2 3 ) : (G46 + G57 ) + i(G47 + G56 ).

14
The remainder roots and associated root vectors are found as follows.
1
: (2G12 G47 G56 ) i(2G13 G46 + G57 ),
: (2G14 + G27 + G36 ) i(2G15 + G26 G37 ),
: (2G16 G25 G34 ) i(2G17 G24 + G35 ).
Theorem 1.8.2. In the root system of g2 C of Theorem 1.8.1,

1 = 1 2 ,
2 = 2
is a fundamental root system of the Lie algebra g2 C and

= 21 + 33
is the highest root. The Dynkin diagram and the extended Dynkin diagram of g2 C are
respectively given by
i
2
3
i
2
Proof. All positive roots of g2 C are expressed by

1 2 = 1 ,
1 = 1 + 2 ,
1 3 = 21 + 32 , 2 3 = 1 + 32 ,
2 = 2 ,
3 = 1 + 22 .
Hence = {1 , 2 } is a fundamental root system of g2 C . The real part of hR of h is

hR = {i1 G23 i2 G45 i3 G67 | i R, 1 + 2 + 3 = 0}
and the Killing form B2 on hR is given by
B2 (H, H ) = 8
3
X
k k ,
k=1
for H = i1 G23 i2 G45 i3 G67 , H = i1 G23 i2 G45 i3 G67 hR (Theorem 1.7.2). Now, the canonical element Hi hR associated with i (B2 (H , H) =
(H), H hR ) are determined as follows.
1
1
H1 = iG23 + iG45 ,
8
8
Hence we have
H2 =
1
1
1
iG23 iG45 + iG67 .
24
12
24
11
1
(1 + 1) = ,
88
4
1
1 1
(1 + 4 + 1) =
,
(2 , 2 ) = B2 (H2 , H2 ) = 8
24 24
12
1 1
1
(1 , 2 ) = B2 (H1 , H2 ) = 8
(1 2) = ,
8 24
8
1
1
(, ) = , (, 1 ) = , (, 2 ) = 0.
4
8
(1 , 1 ) = B2 (H1 , H1 ) = 8
15
Using them, we can draw the Dynkin diagram and the extended Dynkin diagram of
g2 C .
According to Borel-Siebenthal theory (Borel and Siebenthal [4]), the Lie algebra
g2 has two subalgebras as maximal subalgebras with the maximal rank 2.
(1) One is a subalgebra of type C1 C1 which is obtained as the fixed points by
an involution of g2 .
(2) The other is a subalgebra of type A2 which is obtained as the fixed points by
an automorphism w of order 3 of g2 .
These subalgebras will be realized in the group G2 in Theorem 1.10.1 and Theorem
1.9.4, respectively.
1.9. Automorphism w of order 3 and subgroup SU (3) of G2
We shall study the following subgroup (G2 )e1 of G2 :
(G2 )e1 = { G2 | e1 = e1 }.
Theorem 1.9.1.
(G2 )e1
= SU (3).
Proof (cf. Theorem 1.5.1). We define a mapping : SU (3) (G2 )e1 by

a + m C C 3 = C.
(A)(a + m) = a + Am,
We first prove that (A) (G2 )e1 . For = (A), A SU (3) and x = a + m, y =
e (which is the adjoint matrix of
b + n C C 3 = C, using that if A SU (3) then A
1
A) = A = A , we have
(x)(y) = (a + Am)(b + An)
= (ab hAm, Ani) + (aAn + bAm Am An)
e
= (ab hm, A Ani) + (aAn + bAm t A(m
n))
= (ab hm, ni) + A(an + bm m n)
= (A)((a + m)(b + n)) = (xy).
Hence (A) G2 . Clearly (A)e1 = e1 , so that (A) (G2 )e1 . Evidently is a

homomorphism. We show that is onto. Let (G2 )e1 . Note that induces a
C-linear transformation of C 3 . Now let
e2 = a1 ,
e4 = a2 ,
e6 = a3
and construct a matrix A = (a1 , a2 , a3 ) M (3, C). From (e2 )(e4 ) = (e2 e4 ) =
e6 , we have a1 a2 = a3 , namely, ha1 , a2 i a1 a2 = a3 , hence we have
ha1 , a2 i = 0,
a3 = a1 a2 .
16
Similarly we have ha2 , a3 i = ha3 , a1 i = 0. Moerover from (ek )(ek ) = (ek ek ) =

(1) = 1, we have hak , ak i = 1, hence A U (3). Finally det A = (a3 , a1
a2 ) = (a3 , a3 ) = ha3 , a3 i = 1 (where (a, b) is the ordinary inner product in C 3 :
(a, b) = t ab). Hence we have A SU (3) and (A) = , which shows that is onto.
Ker = {E} is easily obtained. Thus we have the isomorphism SU (3)
= (G2 )e1 .
Theorem 1.9.2.
G2 /SU (3) S 6 .
Proof. S 6 = {a C | a
= a, |a| = 1} is a 6 dimensional sphere. Since the group
G2 is a subgroup of O(7) = { O(C) | 1 = 1} (Remark of Theorem 1.2.2), G2 acts
on S 6 . We shall show that this action is transitive. To prove this, it is sufficient to
show that any element a S 6 can be transformed to e1 S 6 by some G2 . Now,
for a1 S 6 , choose any element a2 S 6 such that (a1 , a2 ) = 0. Let
a3 = a1 a2 .
6
Then a3 S and a3 satisfies (a1 , a3 ) = (a2 , a3 ) = 0. Choose any element a4 S 6

such that (a1 , a4 ) = (a2 , a4 ) = (a3 , a4 ) = 0. Let
a5 = a1 a4 ,
a6 = a4 a2 ,
a7 = a3 a4 .
Then the set {a0 = 1, a1 , a2 , , a7 } is an orthonormal R-basis of C. Indeed, |ai | =

1, 0 i 7 are trivial, and we need to verify that (ai , aj ) = 0, i 6= j. However this
can be checked by direct calculations such as
(a4 , a7 ) = (a4 , a3 a4 ) = (1, a3 )(a4 , a4 ) = 0,
(a1 , a6 ) = (a1 , a4 a2 ) = (a1 a2 , a4 ) = (a3 , a4 ) = 0,
(a3 , a6 ) = (a3 , a4 a2 ) = (a3 a2 , a4 ) = (a1 , a4 ) = 0, etc.
Now, since {e0 = 1, e1 , e2 , , e7 } and {a0 = 1, a1 , a2 , , a7 } are both orthonormal
R-bases, the R-linear isomorphism : C C satisfying
ei = ai ,
i = 0, 1, , 7
belongs to O(7): O(7). Moreover we claim that G2 , that is, satisfies

(xy) = (x)(y),
x, y C.
To show this, it is sufficient to note that

(ei ej ) = (ei )(ej ),
i, j = 0, 1, , 7
which can be also checked by direct calculations such as

(e4 )(e7 ) = a4 a7 = a4 (a3 a4 ) = a4 (a4 a3 ) = a3 = e3 = (e4 e7 ),
(e1 )(e6 ) = a1 a6 = a1 (a4 a2 ) = a4 (a1 a2 ) = a4 a3 = a3 a4 = a7

= e7 = (e1 e6 ),
(e3 )(e6 ) = a3 a6 = (a1 a2 )(a4 a2 ) = (a2 a1 )(a4 a2 ) = a2 (a1 a4 )a2

= a2 a5 a2 = a2 a2 a5 = a5 = e5 = (e3 e6 ), etc.
17
Hence G2 and e1 = a1 , and so 1 a1 = e1 . This shows the transitivity. The

isotropy subgroup (G2 )e1 of G2 at e1 is SU (3) (Theorem 1.9.1). Thus we have the
homeomorphism G2 /SU (3) S 6 .
Since S 6 and SU (3) are both simply connected, from G2 /SU (3) S 6 (Theorem
1.9.2), we see that G2 is also simply connected. Hence we have the following theorem.
Theorem 1.9.3. G2 = { IsoR (C) | (xy) = (x)(y)} is a simply connected
compact simple Lie group.
Remark 1. Since G2 is connected, G2 is contained in SO(7) = { O(7) | det =
1}: G2 SO(7).
Remark 2. Since we know that the dimension of the group G2 as dim G2 =

dim g2 = 14 (Theorem 1.4.3), G2 /SU (3) S 6 is proved as follows. The group G2
acts on S 6 . The isotropy subgroup (G2 )e1 of G2 at e1 is SU (3) (Theorem 1.9.1) and
dim(G2 /(G2 )e1 ) = dim G2 dim SU (3) = 14 8 = 6 = dim S 6 . Therefore we have
G2 /SU (3) S 6 .
Using the mapping : SU (3) G2 , we define a mapping w : C C by
w = (diag(1 , 1 , 1 ))
1
3
where 1 = +
e1 C C. This w is defined as
2
2
a + m C C 3 = C.
w(a + m) = a + 1 m,
Then w G2 and w3 = 1.
We shall study the following subgroup (G2 )w of G2 :

(G2 )w = { G2 | w = w}.
Theorem 1.9.4.
(G2 )w = (G2 )e1

= SU (3).
Proof. Recall the mapping : SU (3) G2 of Theorem 1.9.1. We first show that
(SU (3)) (G2 )w . Indeed, for A SU (3) and a + m C C 3 = C, we have
w(A)(a + m) = w(a + Am) = a + 1 Am
= a + A1 m = (A)w(a + m).
Hence w(A) = (A)w, so that (A) (G2 )w . Conversely, let (G2 )w . We
consider the R-vector subspace Cw = {x C | wx = x} of C, then Cw = C. Since
satisfies w = w, Cw is invariant under . Since the restriction of to Cw induces
an automorphism of C, we have
e1 = e1
or e1 = e1 .
18
In the latter case, consider the mapping 1 : C C defined by 1 (a + m) = a + m.

Then we have 1 G2 and 1 e1 = e1 . Let = 1 . Since e1 = e1 , we have
SU (3) (G2 )e1 (Theorem 1.9.1) (G2 )w . Therefore, 1 = 1 (G2 )w ,
which is a contradiction. Indeed,
1 m = w(1 m) = 1 (wm) = 1 m = 1 m
for all m C 3 ,
which is false. Therefore e1 = e1 , and so SU (3) (Theorem 1.9.1). Thus we have

(G2 )w = (G2 )e1 .
1.10. Involution and subgroup (Sp(1) Sp(1))/Z 2 of G2
The Cayley algebra C naturally contains the field H of quaternions as
H = {x0 + x1 e1 + x2 e2 + x3 e3 | xi R}.
Any element x C is expressed by
x = x0 + x1 e1 + x2 e2 + x3 e3 + x4 e4 + x5 e5 + x6 e6 + x7 e7
(xi R)
= (x0 + x1 e1 + x2 e2 + x3 e3 ) + (x4 + x5 e1 x6 e2 + x7 e3 )e4 ,

that is,
x = m + ae4 ,
m, a H.
In H He4 , we define a multiplication, an inner product ( ,

and an R-linear transformation respectively by
), a conjugation
(m + ae4 )(n + be4 ) = (mn ba) + (an + bm)e4 ,

(m + ae4 , n + be4 ) = (m, n) + (a, b),
m + ae4 = m ae4 ,
(m + ae4 ) = m ae4 .
Since these operations correspond to their respective operations in C, hereafter, we

identify H He4 with C, that is,
H He4 = C.
We shall study the following subgroup (G2 ) of G2 :
(G2 ) = { G2 | = }.
Theorem 1.10.1.
(G2 )
= (Sp(1) Sp(1))/Z 2 , Z 2 = {(1, 1), (1, 1)}.
Proof. We define a mapping : Sp(1) Sp(1) (G2 ) by

(p, q)(m + ae4 ) = qmq + (paq)e4 ,
19
m + ae4 H He4 = C.
We first show that (p, q) (G2 ) . For = (p, q), p, q Sp(1) and x = m + ae4 ,
y = n + be4 H He4 = C, we have
(x)(y) = (qmq + (paq)e4 )(qnq + (pbq)e4 )
= ((qmq)(qnq) (pbq)(paq)) + ((paq)(qnq) + (pbq)(qmq))e4
= q(mn ba)q + (p(an + bm)q)e4
= (p, q)((m + ae4 )(n + be4 )) = (xy).

Hence (p, q) G2 . Clearly (p, q) = (p, q), so that (p, q) (G2 ) . Evidently
is a homomorphism. We shall show that is onto. Let (G2 ) . Since satisfies
= , C = {x C | x = x} = H is invariant under , so that the restriction of
to C induces an automorphism of H. Hence there exists q Sp(1) satisfying
m = qmq,
mH
(Proposition 108 of Yokota [58]). By putting = (1, q)1 , we have (G2 ) and
|H = 1. Therfore induces a transformation of C = {x C | x = x} = He4 .
By putting e4 = pe4 , where p H, we have |p| = |pe4 | = |e4 | = |e4 | = 1, which
implies p Sp(1). From
(m + ae4 ) = m + (a)(e4 ) = m + a(pe4 ) = m + (pa)e4 = (p, 1)(m + ae4 ),
we have = (p, 1). Therefore we have
= (1, q) = (1, q)(p, 1) = (p, q),
(p, q) Sp(1) Sp(1),
which shows that is onto. Ker = {(1, 1), (1, 1)} = Z 2 is easily obtained. Thus
we have the isomorphism (Sp(1) Sp(1))/Z 2
= (G2 ) .
Remark 1.
(Sp(1) Sp(1))/Z 2
= SO(4).
Indeed, a mapping f : Sp(1) Sp(1) SO(4) = SO(H) defined by

f (p, q)x = pxq,
xH
induces the isomorphism (Sp(1) Sp(1))/Z 2

= SO(4) as groups.
Remark 2. Since (G2 ) is connected as the fixed points subgroup by an automorphism of the simply connected group G2 , the fact that : Sp(1) Sp(1) (G2 )
is onto can be proved as follows. The elements
2G12 G47 G56 , G47 + G56 ,
2G13 G46 G57 ,
G46 + G57 ,
2G23 G45 G67 , G45 + G67
forms an R-basis of (g2 ) . So dim(g2 ) = 6 = 3 + 3 = dim(sp(1) sp(1)). Hence
is onto.
20
1.11. Center z(G2 ) of G2

Theorem 1.11.1. The center z(G2 ) of the group G2 is trivial :
z(G2 ) = {1}.
Proof. Let z(G2 ). From the commutativity with : = , we have
SO(4) (Remark 1 of Theorem 1.10.1) and so z(SO(4)). Since the center
z(SO(4)) of SO(4) is the group of order 2, we have
=1
or = .
However
/ z(G2 ) (Theorem 1.10.1). Hence = 1.
1.12. Complex exceptional Lie group G2 C
Definition. The group G2 C is defined to be the automorphism group of the
complex Cayley algebra CC :
G2 C = { IsoC (CC ) | (xy) = (x)(y)}.
Lemma 1.12.1. For G2 C , we have
(x, y) = (x, y),
x, y CC .
Proof. The equality x = (1)(x) holds for all x CC , and so we have 1 = 1.

We shall show that
ek = ek , k = 1, 2, , 7.
Note that (ek )(ek ) = (ek ek ) = (1) = 1. Let x = ek and N (x) = xx C.
Then xx = 1 and N (x)N (x) = 1, which shows that N (x) = 1. If N (x) = 1,
then x = xxx = xN (x) = x, so x C. From xx = 1, we have x = i. Then
(ek ) = (i), and so ek = i, which is a contradiction. Hence N (x) = 1, that is,
xx = 1 and x = xxx = xN (x) = x. Thus we have
x = x,
x CC .
Now we have
1
1
((x)(y) + (y)(x)) = ((x)(y) + (y)(x))
2
2
1

= ((xy + yx) = ((x, y)) = (x, y).
2
(x, y) =
We define a positive definite Hermitian inner product hx, yi in CC by

hx, yi = ( x, y),
21
For HomC (CC ), we denote the complex conjugate transpose of with respect to
the inner product hx, yi by : h x, yi = hx, yi.
Lemma 1.12.2. (1) For G2 C , we have = 1 G2 C .
(2) For any G2 , its complexificated mapping C : CC CC belongs to G2 C :

C G2 C . Identifying with C , we regard G2 as a subgroup of G2 C : G2 G2 C .
Now, for G2 C , we have, G2 if and only if = , that is,
G2 = { G2 C | = }.
Proof. (1) h x, yi = hx, yi = ( x, y) = (1 x, y) = h 1 x, yi for all
x, y CC . Hence = 1 G2 C .
(2) Let G2 C satisfy = . Then since x = x = x, we have x C

for x C. Hence induces an R-transformation of C and G2 , further we
have = ( )C .
Theorem 1.12.3. The polar decomposition of the group G2 C is given by
G2 C G2 R14 .
In particular, G2 C is a simply connected complex Lie group of type G2 .
Proof. Evidently G2 C is an algebraic subgroup of IsoC (CC ) = GL(8, C). If
G2 C , then G2 C (Lemma 1.12.2.(1)). Hence, from Chevalleys lemma
(Chevalley [5]), we have
G2 C (G2 C U (CC )) Rd = G2 Rd ,
where U (CC ) = { IsoC (CC ) | hX, Y i = hX, Y i} and d = dim G2 C dim G2 =
2 14 14 = 14. Since G2 is simply connected (Theorem 1.9.3), G2 C is also simply
connected. The Lie algebra of the group G2 C is g2 C , so that G2 C is a complex simple
Lie group of type G2 .
1.13. Non-compact exceptional Lie group G2(2) of type G2
In C = H He4 , we define a multiplication by
(m + ae4 )(n + be4 ) = (mn + ba) + (an + bm)e4 .
This algebra C is called the split Cayley algebra, and is isomorphic to (CC ) = {x
CC | x = x} as algebras. Now, the group G2(2) is defined to be the automorphism
group of the split Cayley algebras C :
G2(2) = { IsoR (C ) | (xy) = (x)(y)}.
and which can also be defined by
22
G2(2) = (G2 C ) = { G2 C | = }.
Theorem 1.13.1. The polar decomposition of the Lie groups G2(2) is given by
G2(2) (Sp(1) Sp(1))/Z 2 R8 .
Proof. In the split Cayley algebra C , the inner product (x, y) is defined as
1
(x, y) = (xy + yx). If we define an inner product (x, y) of C by (x, y) , then
2
(x, y) is a positive definite inner product. For G2(2) , the transpose t of with
respect to the inner product (x, y) is t = 1 G2(2) . Since G2(2) is an algebraic
subgroup of IsoR (C ) = GL(8, R), from Chevalleys lemma, we have
G2(2) (G2(2) O(8)) Rd
= ((G2 C ) ) Rd = ((G2 C ) ) Rd = (G2 ) Rd
= (Sp(1) Sp(1))/Z 2 Rd (Theorem 1.10.1), d = 8.
where O(8) = { IsoR (C ) | (x, y) = (x, y)}.

Theorem 1.13.2. The center z(G2(2) ) of the group G2(2) is trivial:
z(G2(2) ) = {1}.
1.14. Principle of triality in SO(8)
For a S n1 = {a Rn | (a, a) = 1}, we define an element Da O(n) =
O(Rn ) = {A IsoR (Rn ) | (Ax, Ay) = (x, y)} by
Da x = x 2(x, a)a,
x Rn .
Da is called the reflection with respect to the hyperplane orthogonal to a. Its determinant is 1: det(Da ) = 1.
Lemma 1.14.1. The group O(n) is generated by reflections, that is, any A O(n)
can be expressed by the product of finite number of reflections:
A = Dam Da2 Da1 ,
ai S n1 .
In particular, A SO(n) can be expressed by the product of even number of reflections.

A = Da2m Da2 Da1 ,
ai S n1 .
Proof. (See Theorem 24 of Yokota [58).

From now on, the group SO(8) is identified with the group
SO(C) = { IsoR (C) | (x, y) = (x, y), det = 1}.
23
Theorem 1.14.2. (Principle of triality in SO(8)). For any 3 SO(8), there

exist 1 , 2 SO(8) such that
(1 x)(2 y) = 3 (xy),
x, y C.
Moreover, 1 , 2 are determined uniquely for 3 up to the sign, that is, for 3 , such
1 , 2 have to be 1 , 2 or 1 , 2 .
Proof. Since 3 SO(8) is expressed by the product of even number of reflections (Lemma 1.14.1), it is sufficient to show the existence of 1 , 2 only for
3 = Db Da , a, b C, |a| = |b| = 1. Now, since
Da x = x 2(x, a)a = x (xa + ax)a = axa,
x C,
we have 3 x = Db Da x = b(axa)b. Define mappings 1 , 2 : C C respectively by

1 x = b(ax), 2 x = (xa)b. We then see that 1 , 2 SO(8) and
(1 x)(2 y) = (b(ax))((ya)b) = b(a(xy)a)b = 3 (xy),
x, y C.
Next, we shall show the uniqueness of 1 , 2 up to the sign. To prove this, it is

sufficient to show in the case 3 = 1. Now, let
(1 x)(2 y) = xy,
x, y C.
Let 1 1 = p, then |p| = 1 and p(2 y) = y, so 2 y = py. Similarly we have 1 x = x

q,
where q = 2 1. Hence (x
q )(
py) = xy. If we let x = y = 1, then q p = 1, so q = p.
Therefore we have
(xp)(
py) = xy, x, y C.
Putting py instead of y, we have
(xp)y = x(py),
x, y C.
From this, we see that p is a real number. Hence p = 1 because |p| = 1. Thus we
have 1 = 2 = 1 or 1 = 2 = 1.
Lemma 1.14.3. For 1 , 2 , 3 O(8), the relation
(1 x)(2 y) = 3 (xy),
x, y C
(2 x)(3 y) = 1 (xy),
x, y C,
implies
(3 x)(1 y) = 2 (xy),
x, y C.
Proof. If x = 0 or y = 0, then the statement is trivially valid, so we may

assume x, y 6= 0. Now, multiply 1 x from left and 3 (xy) from right on the relation
24
(1 x)(2 y) = 3 (xy)). Then we get |x|2 (2 y)(3 (xy)) = 1 x|xy|2 . Therefore

(2 y)(3 (xy)) = 1 x|y|2 .
Putting yz instead of x, we have (2 y)(3 (yyz)) = 1 (yz)|y|2 , that is,
(2 y)(3 z) = 1 (yz).
The other relation is similarly obtained.
Lemma 1.14.4. If 1 , 2 , 3 O(8) satisfy
(1 x)(2 y) = 3 (xy),
x, y C,
then 1 , 2 , 3 SO(8).
Proof. Suppose 1
/ SO(8). Since the mapping : C C defined by x = x
belongs to O(8), and since det = 1, we have 1 = 1
SO(8). Using the
1
principle of triality (Theorem 1.14.2) on the element 1 (cf. Lemma 1.14.3), there
exist 2 , 3 SO(8) such that
(1 (1 x))(2 (2 y)) = 3 ((1 x)(2 y)) = 3 (3 (xy)),
x, y C.
Setting 2 2 = 2 and 3 3 = 3 , the above relation becomes

x(2 y) = 3 (xy),
x, y C.
Put x = 1, then 2 y = 3 y, so 2 = 3 . Hence we have

x(2 y) = 2 (xy),
x, y C.
(i)
Put 2 1 = p, then |p| = 1. Lett y = 1 in (i), then xp = 2 x. Hence (i) becomes

x(yp) = (xy)p,
x, y C.
(ii)
Again let y = p, then x = (xp)p, and so x p = xp. Then

px = xp
for all x C.
Therefore p R, hence p = 1 since |p| = 1. Thus (ii) becomes x y = xy, and so

xy = yx for all x, y C.
But this is a contradiction. Therefore 1 SO(8). As for 2 , 3 , use Lemma 1.14.3
and the argument above, to show 2 , 3 SO(8).
As a corollary of the principle of triality (Theorem 1.14.2), we have the following
proposition.
25
Proposition 1.14.5. For a C such that |a| = 1, the mapping a : C C

defined by
a x = axa1 , x C
belongs to the group G2 if and only if a3 = 1.
Proof. If we apply Lemma 1.14.3 to the Moufang formula (ax)(ya) = a(xy)a,
then
(xa)(aya) = (xy)a, x, y C.
If we replace x by ax and y by ya respectively, then
(axa)(aya2 ) = a(xy)a2 ,
x, y C.
Now, the mapping a belongs to G2 if and only if

(axa)(aya) = a(xy)a,
x, y C.
From the uniqueness of the principle of triality up to sign, we have

aya2 = aya.
Therefore a2 = a, so that a3 = 1. The converse also holds.
1
3
e1 C C. Then 1 3 = 1, so 1 G2 by Proposition 1.14.5.
Let 1 = +
2
2
This 1 is nothing but w of Section 1.9: 1 = w, because
1 (a + m) = 1 (a + m)1 = 1 a1 + 1 2 m = a + 1 m = w(a + m),
for a + m C C 3 = C.
1.15. Spinor group Spin(7)
If we use the principle of triality for SO(7), then there exist
e, SO(8)
satisfying
(x)(e
y) = (xy),
x, y C.
Putting x = 1, we have
ey = y, and so
e = . Then (i) becomes
(x)(e
y) =
e(xy),
x, y C.
(i)
(ii)
Conversely, suppose that ,

e SO(8) satisfy (ii). Putting x = 1, we have (1)(e
y) =
ey, and so we have 1 = 1. Hence SO(7).

e3 of SO(8) by
Definitioin. We define a subgroup B
e3 = {e
B
SO(8) | (x)(e
y) =
e(xy), x, y C for some SO(7)}.
26
e3 is a compact group.
B
Theorem 1.15.1.
e3 /G2 S 7 ,
B
e3 is connected.
In particular, the group B
e3 SO(7) = G2 .
B
e3 is a subgroup
Proof. S 7 = {a C | |a| = 1} is a 7 dimensional sphere. Since B
e3 acts on S 7 . We shall show that B
e3 acts transitively on S 7 .
of SO(8), the group B
7
To prove this, it is sufficient to show that any b0 S can be transformed to 1 S 7
e3 . Now, we choose any element a1 S 7 such that (1, a1 ) = 0. and
by some B
choose any element a2 S 7 such that (1, a2 ) = (a1 , a2 ) = 0. Let a3 S 7 be the
element determined by a3 b0 = a2 (a1 b0 ). More precisely, if we let a3 = (a2 (a1 b0 ))b0 ,
then a3 S 7 and satisfies (1, a3 ) = (a1 , a3 ) = (a2 , a3 ) = 0. Choose any element
a4 S 7 such that (1, a4 ) = (a1 , a4 ) = (a2 , a4 ) = (a3 , a4 ) = 0. Let a5 , a6 , a7 S 7 be
elements determined by
a5 b0 = a1 (a4 b0 ),
a6 b0 = a2 (a4 b0 ),
a7 b0 = a6 (a1 b0 ).
Then, {a0 = 1, a1 , a2 , , a7 } forms an orthonormal R-basis of C. To prove this, we

need to show (ai , aj ) = ij , i, j = 0, 1, , 7. We will only show the following two
since the others can be proved in a similar manner.
(a2 , a6 ) = (a2 b0 , a6 b0 ) = (a2 b0 , a2 (a4 b0 )) = (b0 , a4 b0 ) = (1, a4 ) = 0,
(a3 , a7 ) = (a3 b0 , a7 b0 ) = (a2 (a1 b0 ), a6 (a1 b0 )) = (a1 (a2 b0 ), a1 (a6 b0 ))
= (a2 b0 , a6 b0 ) = (a2 b0 , a2 (a4 b0 )) = (b0 , a4 b0 ) = (1, a4 ) = 0.
Now, since {e0 , e1 , , e7 } and {a0 = 1, a1 , , a7 } are both orthonormal R-bases of
C, the R-linear isomorphism : C C satisfying
ei = ai ,
i = 0, 1, , 7
belongs to O(7). Moreover, this satisfies

(x)((y)b0 ) = ((xy))b0 ,
x, y C.
(i)
To prove this, it is sufficient to show that

(ei )((ej )b0 ) = ((ei ej ))b0 ,
i, j = 0, 1, , 7.
Again we need to verify many cases, but here we will only show the following two
examples.
(e1 )((e3 )b0 ) = a1 (a3 b0 ) = a1 (a2 (a1 b0 )) = a1 (a1 (a2 b0 )) = a2 b0
= (e2 )b0 = (e1 e3 )b0 ,
(e2 )((e5 )b0 ) = a2 (a5 b0 ) = a2 (a1 (a4 b0 )) = a1 (a2 (a4 b0 )) = a1 (a6 b0 )

= a6 (a1 b0 ) = a7 b0 = (e7 )b0 = (e2 e5 )b0 .
27
Now, if we put
b i = ai b 0 ,
i = 0, 1, , 7,
then {b0 , b1 , , b7 } is an orthonormal R-basis of C. The R-linear isomorphism

e:
C C satisfying
eei = bi , i = 0, 1, , 7
belongs to O(8). Since
ex = (x)b0 , it follows from (i) that
(x)(e
y) =
e(xy),
x, y C.
(ii)
Since O(7),
e O(8) satisfy (ii), SO(7) ,
e SO(8) (Lemma 1.14.4). Hence
1
e
e B3 and
e1 = b0 , and so
e b0 = 1. This shows the transitivity. The isotropy
e3 at 1 S 7 is G2 . Indeed, if
e3 satisfies
subgroup of B
e B
e1 = 1, then we have
e
=
e, so
e G2 . Conversely, G2 satisfies B3 and 1 = 1. Thus we have
e3 /G2 S 7 .
the homeomorphism B
e3
B
= Spin(7).
Theorem 1.15.2.
(From now on, we identify these groups).
e3 satisfy the principle of triality

Proof. Suppose SO(7) and
eB
(x)(e
y) =
e(xy),
x, y C.
(xy) = (e
x)(e
y),
x, y C.
f3 SO(7) by p(e
We define a mapping p : B
) = . It is not difficult to see that p is a
homomorphism. The principle of triality implies that p is onto and Ker p = {1, 1}.
Next, we shall prove that p is continuous. From Lemma 1.14.3, we have
Consider the matrices of and

e with respect to the R-basis {e0 , e1 , , e7 }. Then
we can see that each component of matrix is a polynominal of components of matrix
e3 )). Therefore p is continuous. Hence we have the
e (for example, e1 = (e
e2 )(e
isomorphism
e3 /{1, 1}
B
= SO(7).
e3 is isomorphic to Spin(7) as the universal covering group of SO(7).

Therefore B
1.16. Spinor group Spin(8)
e 4 of SO(8) SO(8) SO(8) by

Definition. We define a subgroup D
e 4 = {(1 , 2 , 3 ) SO(8) SO(8) SO(8) | (1 x)(2 y) = 3 (xy), x, y C}.

D
e 4 is a compact group.
D
e 4 satisfies (x)(e
Since an element (,
e, e
) of D
y) =
e(xy), x, y C, we see that
e 4 contains Spin(7) as a subgroup under the identification
D
28
e 4.
Spin(7)
e (,
e, e
) D
e 4 /Spin(7) S 7 .
Proposition 1.16.1.
D
e 4 is connected.
In particular, the group D
Proof. S 7 = {a C | |a| = 1} is a 7 dimensional sphere. We define an action of

e 4 on S 7 by
D
(1 , 2 , 3 )a = 1 a, a S 7 .
This action is transitive. Let a S 7 . Since SO(8) acts transitively on S 7 , there

exists 1 SO(8) such that 1 1 = a. For 1 , choose 2 , 3 SO(8) satisfying the
principle of triality
(1 x)(2 y) = 3 (xy). x, y C.
e 4 and (1 , 2 , 3 )1 = a, which shows the transitivity. The

Then (1 , 2 , 3 ) D
e 4 at 1 S 7 is Spin(7). Indeed, if (1 , 2 , 3 ) D
e 4 satisotropy subgroup of D
isfies (1 , 2 , 3 )1 = 1, then 1 1 = 1. Therefore 1 SO(7), which shows that
e 4 /Spin(7)
(1 , 2 , 3 ) Spin(7) and vice versa. Thus we have the homeomorphism D
7
S .
Theorem 1.16.2.
e4
D
= Spin(8).
e 4 SO(8) by
Proof. We define a mapping p : D
p(1 , 2 , 3 ) = 1 .
Evidently, p is a homomorphism. The principle of triality implies that p is onto and

Ker p = {(1, 1, 1), (1, 1, 1)}. Thus we obtain the isomorphism
e 4 /{(1, 1, 1), (1, 1, 1)}
D
= SO(8).
e 4 is isomorphic to Spin(8) as the universal covering group of SO(8).

Therefore D
Theorem 1.16.3. The center z(Spin(8)) of the group Spin(8) is isomorphic to

the group Z 2 Z 2 :
z(Spin(8)) = {(1, 1, 1), (1, 1, 1), (1, 1, 1), (1, 1, 1)}
= {(1, 1, 1), (1, 1, 1)} {(1, 1, 1), (1, 1, 1)}

= Z2 Z2.
Proof. The proof follows easily from z(SO(8)) = {1, 1} and the principle of
triality.
29
Theorem 1.16.4. We define automorphisms , , : Spin(8) Spin(8) respectively by

(1 , 2 , 3 ) = (1 , 3 , 2 ),
(1 , 2 , 3 ) = (3 , 2 , 1 ),
(1 , 2 , 3 ) = (2 , 3 , 1 ),
where : SO(8) SO(8) in the right side is defined by ()x = x, x C. Then
we have relations
2 = 1, 2 = 1, 3 = 1, = .
The subgroup S3 generated by , in the automorphism group Aut(Spin(8)) of Spin(8)
is isomorphic to the symmetric group S3 . Also, we have
Spin(7) = { Spin(8) | = },
G2 = { Spin(8) | = , S3 }
= { Spin(8) | = , = }
= { Spin(8) | = }
= { Spin(7) | = }.
Moreover we have the isomorphism
Spin(8)/{(1, 1, 1), (1, 1, 1)}
= Spin(8)/{(1, 1, 1), (1, 1, 1)},
that is,
SO(8)
= Ss(8).
Proof. The group multiplication between , , is the same as that of Theorem
1.3.5. Next, we shall show
G2 = { Spin(8) | = }.
If (1 , 2 , 3 ) Spin(8) satisfies (1 , 2 , 3 ) = (1 , 2 , 3 ), then we have 1 =
2 = 3 (= ), that is,
(x)(y) = (xy),
x, y C,
(i)
Put a = 1, then |a| = 1, and put x = 1 and y = 1 in (i), then we have a(y) = (y)
and (x)a = (x), respectively. Hence, we get
a(x) = (x)a
for all x C.
hence a R, and so a = 1 from |a| = 1. In the case a = 1, let x = y = 1 in

(i), then (1)(1) = 1 which is a contradiction. Hence a = 1, so that = .
Therefore we have G2 . Finally, the automorphism 2 : Spin(8) Spin(8)
satisfies
2 (1, 1, 1) = (1, 1, 1),
hence 2 induces the isomorphism SO(8)
= Ss(8).
30
Exceptional Lie group F4

2.1. Exceptional Jordan algebra J
Let J = J(3, C) denote all 3 3 Hermitian matrices with entries in the Cayley
algebra C:
J = {X M (3, C) | X = X},
where X = tX. Any element X J is
1
X = X(, x) = x3
x2
of the form
x3 x2
2 x1 , i R, xi C.
x1 3
J is a 27 dimensional R-vector space. In J, the multiplication X Y , called the Jordan

multiplication, is defined by
X Y =
1
(XY + Y X).
2
In J, we define the trace tr(X), an inner product (X, Y ) and a trilinear form tr(X, Y, Z)
respectively by
tr(X) = 1 + 2 + 3 , X = X(, x),
(X, Y ) = tr(X Y ), tr(X, Y, Z) = (X, Y Z).
Moreover, in J, we define a multiplication X Y , called the Freudenthal multiplication, by
X Y =
1
(2X Y tr(X)Y tr(Y )X + (tr(X)tr(Y ) (X, Y ))E),
2
(where E is the 3 3 unit matrix) and a trilinear form (X, Y, Z) and the determinant
detX respectively by
(X, Y, Z) = (X, Y Z),
detX =
1
(X, X, X).
3
For X = X(, x), Y = Y (, y) and Z = Z(, z) J, the explicit forms in the terms
of their entries are as follows.
(X, Y ) =
tr(X, Y, Z) =
3
X
i=1
3
X
(i i + 2(xi , yi )),
(i i i + R(xi yi+1 zi+2 + xi zi+1 yi+2 )
i=1
+ i ((yi+1 , zi+1 ) + (yi+2 , zi+2 )) + i ((zi+1 , xi+1 ) + (zi+2 , xi+2 ))

+ i ((xi+1 , yi+1 ) + (xi+2 , yi+2 ))),
31
(X, Y, Z) =
3
X
1
i=1
(i i+1 i+2 + i i+2 i+1 ) + R(xi yi+1 zi+2 + xi zi+1 yi+2 )

(i (yi , zi ) + i (zi , xi ) + i (xi , yi )) ,
detX = 1 2 3 + 2R(x1 x2 x3 ) 1 x1 x1 2 x2 x2 3 x3 x3 .
Lemma 2.1.1. The followings hold in J.
(1) (i)
(ii)
(2) (i)
(ii)
X Y = Y X,
X Y = Y X.
1
E X = X, E X = (tr(X)E X), E E = E.
2
The inner product (X, Y ) is symmetric and positive definite.
tr(X, Y, Z) = tr(Y, Z, X) = tr(Z, X, Y ) = tr(X, Z, Y ) = tr(Y, X, Z)
= tr(Z, Y, X).
The similar statement is also valid for (X, Y, Z).
(iii)
(iv)
(3) (i)
(ii)
(X, E) = (X, E, E) = tr(X, E, E) = tr(X), tr(X, Y, E) = (X, Y ).

1
tr(X Y ) = (tr(X)tr(Y ) (X, Y )).
2
(X X) X = (detX)E (Hamilton-Cayley).
(X X) (X X) = (detX)X.
Proof. (1) is evident.

(2) is clear from the explicit forms of (X, Y ), tr(X, Y, Z), (X, Y, Z) etc.
(3) Using the following explicit form
2 3 x1 x1 x1 x2 3 x3
X X = x1 x2 3 x3 3 1 x2 x2
x3 x1 2 x2 x2 x3 1 x1
x3 x1 2 x2
x2 x3 1 x1 ,
1 2 x3 x3
X = X(, x),
each formula is obtained by direct calculations.
In J, we adopt the following notations:
1 0 0
0 0 0
0 0 0
E1 = 0 0 0 , E2 = 0 1 0 , E3 = 0 0 0 ,
0 0 0
0 0 0
0 0 1
0 x 0
0 0 0
0 0 x
F1 (x) = 0 0 x , F2 (x) = 0 0 0 , F3 (x) = x 0 0 .
x 0 0
0 0 0
0 x 0
The tables of the Jordan multiplication and the Freudenthal multiplication among
elements above are given as follows.
Ei Ej = 0, i 6= j
Ei Ei = Ei
1
Ei Fj (x) = Fj (x), i 6= j
Ei Fi (x) = 0
2
Fi (x) Fi (y) = (x, y)(Ei+1 + Ei+2 ), Fi (x) Fi+1 (y) = 1 Fi+2 (xy),
2
32
Ei Ei+1 = Ei+2
2
Ei Fj (x) = 0, i 6= j
Fi (x) Fi+1 (y) = Fi+2 (xy),

2
where the indexes are considered as mod 3.
Ei Ei = 0
1
Ei Fi (x) = Fi (x)
Fi (x) Fi (y) = (x, y)Ei ,
2.2. Compact exceptional Lie group F4

Definition. The group F4 is defined to be the automorphism group of the Jordan
algebra J:
F4 = { IsoR (J) | (X Y ) = X Y }.
Lemma 2.2.1. (1) For F4 , we have E = E.
(2) For F4 , we have tr(X) = tr(X), X J.
Proof. (1) Applying on E X = X, we have E X = X. Let X = 1 E,

then E E = E, that is, E = E.
(2) We use the Hamilton-Cayley identity X (X X) = (detX)E (Lemma

2.1.1.(3)), that is,
1
X (X X) tr(X)X 2 + (tr(X)2 tr(X 2 ))X = (detX)E.
2
(i)
We put X in the place of X of (i) and then apply 1 F4 on the obtained

expression. Then
1
X (X X) tr(X)X 2 + (tr(X)2 tr((X 2 ))X = (detX)E.
2
By subtracting (i)(ii), we get
1
(tr(X) tr(X))X 2 + (tr(X)2 tr(X)2 + tr((X)2 ) tr(X 2 ))X
2
= (detX det(X))E.
Let X = Fi (ej ), i = 1, 2, 3, j = 0, 1, , 7, then
1
tr(Fi (ej ))(Ei+1 + Ei+2 ) + (tr(Fi (ej ))2 + tr((Fi (ej ))2 ) 2)Fi (ej )
2
= det(Fi (ej ))E.
Comparing the entries of both sides of the equation above, we have
tr(Fi (ej )) = 0 (= tr(Fi (ej )))
and tr((Fi (ej ))2 ) = 2, hence
tr(Ei ) = tr((E Fi (1)2 )) = tr(E) tr((Fi (1))2 ) = 3 2 = 1 = tr(Ei ),
33
(ii)
for i = 1, 2, 3. Consequently, we have tr(X) = tr(X) for every X = Ei , Fi (ej ) of the

R-basis of J. Thus the lemma is proved.
For HomR (J), we denote by t the transpose of with respect to the inner
product (X, Y ): (t X, Y ) = (X, Y ).
Lemma 2.2.2. For IsoR (J), the following four conditions are equivalent.
(1) det (X) = det X,
for all X J.
(2) (X, Y, Z) = (X, Y, Z),

t 1
(3) X Y =
(X Y ),
(4) X X = t 1 (X X),
for all X, Y, Z J.
for all X, Y J.
for all X J.
Proof. (1) (2) det (X) = det X implies that (X, X, X) = (X, X, X).
Putting X + Y + Z in place of X and comparing the coefficient of , we obtain
(2).
(2) (1) is evident.
(4) (3) Putting X + Y in place of X and comparing the coefficient of , we

obtain (3).
(3) (4) is evident.
(2) (3) is easily obtained.

Lemma 2.2.3. If IsoR (J) satisfies det (X) = det X for all X J, then we
have
det (t 1 X) = det (t X) = det X, for all X J.
Proof. We have
t 1
(Y Y ) t 1 (Y Y ) = (Y Y ) (Y Y ) (Lemma 2.2.2.(4))
= (det Y )Y (Lemma 2.1.1.(3)) = (det Y )Y = ((det Y )Y )
= ((Y Y ) (Y Y )), Y J.
Let Y = X X, X J in the above, then

t 1
((det X)X) t 1 ((det X)X) = ((det X)X (det X)X).

(1) Case det X 6= 0. In this case, we have t 1 X t 1 X = (X X). Hence,
3det (t 1 X) = (t 1 X, t 1 X t 1 X)
= (t 1 X, (X X)) = (X, X X) = 3det X,
hence we have det (t 1 X) = det X. Next, if we use 1 instead of , we can see
also that det (t X) = det X.
34
(2) Case det X = 0. If det (t 1 X) 6= 0, we can use the result of (1). If we put t X
instead of X in det (t 1 X) = detX of (1), then det t 1 (t X)) = det (t X). Hence
0 = det X = det t X) 6= 0 which is a contradiction. Thus we have det (t 1 X) =
det (t X) = 0, so det (t 1 X) = det (t X) = det X is also valid.
Lemma 2.2.4. For IsoR (J), the following five conditions are equivalent.
(1)
(2)
(X Y ) = X Y .
tr(X, Y, Z) = tr(X, Y, Z), (X, Y ) = (X, Y ).
(3)
det (X) = det X, (X, Y ) = (X, Y ).
(4)
det (X) = det X, E = E.
(5)
(X Y ) = X Y .
Proof. (1) (2) (X, Y ) = tr(X Y ) = tr((X Y )) = tr(X Y ) (Lemma

2.2.1) = (X, Y ). Also tr(X, Y, Z) = (X, Y Z) = (X, (Y Z)) = (X, Y
Z) = tr(X, Y, Z).
(2) (1) (X Y, Z) = tr(X, Y, Z) = tr(X, Y, Z) = (X Y, Z) = ((X
Y ), Z) holds for all Z, so we have X Y = (X Y ).
(2) (3) Since we have already shown (2) (1), we can use tr(X) = tr(X)
(Lemma 2.2.1.(2)). Now,
3
1
3det(X) = tr(X, X, X) tr(X)(X, X) + tr(X)3
2
2
1
3
3
= tr(X, X, X) tr(X)(X, X) + tr(X) = 3detX.
2
2
(3) (5) ((X Y ), Z) = (X Y, Z) = (X, Y, Z) = (X, Y, Z) (Lemma
2.2.2) = (X Y, Z) holds for all Z J, so we have X Y = (X Y ).
(5) (4)
(det (X))X = (X X) (X X) (Lemma 2.1.1.(3))
= ((X X) (X X)) = (det X)X (Lemma 2.1.1.(3))

and so we have det (X) = det X. In the relation
X E = (X E) =
1
(tr(X)E X),
2
if we denote E = P , then
X P =
1
1
tr(X)P X,
2
2
X J.
Let X = 1 E1 and P = 1 E1 + 2 E2 + 3 E3 + F1 (p1 ) + F2 (p2 ) + F3 (p3 ), then

1
(2 E3 + 3 E2 F1 (p1 ))
2
1
= ((1 E1 + 2 E2 + 3 E3 + F1 (p1 ) + F2 (p2 ) + F3 (p3 )) E1 ),
2
where = tr(1 E1 ). By comparing entries of both sides, we have
35
0 = 1 1,
3 = 2 ,
2 = 3 ,
p1 = p1 ,
0 = p2 ,
0 = p3 .
Consequently we have p2 = p3 = 0. Similarly, by letting X = 1 E2 , we also have

p1 = 0. Again put X = 1 F1 (1), then
1
1
1 F1 (1) = ((1 E1 + 2 E2 + 3 E3 ) F1 (1)),
2
2
where = tr(1 F1 (1)). By Comparing entries of F1 -parts, we see that 1 = 1.
Similarly 2 = 3 = 1. Therefore we have E = E.
(4) (2) tr(X) = (X, E, E) = (X, E, E) = (X, E, E) (Lemma 2.2.2.(2))
1
= tr(X). Hence (tr(X)tr(Y ) (X, Y )) = (X, Y, E) (Lemma 2.1.1) = (X, Y, E)
2
1
1
= (X, Y, E) = (tr(X) tr(Y )(X, Y )) = (tr(X)tr(Y )(X, Y )). There2
2
fore we obtain
(X, Y ) = (X, Y ).
Next, using the
1
tr(Z)(X, Y ) +
2
1
1
relation (X, Y, Z) = tr(X, Y, Z) tr(X)(Y, Z) tr(Y )(Z, X)
2
2
1
tr(X)tr(Y )tr(Z) and (X, Y, Z) = (X, Y, Z), we obtain
2
tr(X, Y, Z) = tr(X, Y, Z).
Theorem 2.2.5. F4 is a compact Lie group.

Proof. F4 is a compact Lie group as a closed subgroup of the orthogonal group
O(27) = O(J) = { IsoR (J) | (X, Y ) = (X, Y )} (Lemma 2.2.4.(2)).
The group F4 contains G2 as
consider a mapping
e : J J,
1 x3
e x3 2
x2 x1
a subgroup in the following way. For G2 , we

x2
1
x1 = x3
3
x2
x3
2
x1
x2
x1 .
3
Then
e F4 . So we identify G2 with
e F4 : G2 F4 .
2.3. Lie algebra f4 of F4
In order to investigate the Lie algebra f4 of the group F4 , it will be helpful to

study the Lie algebra e6(26) of the group
E6(26) = { IsoR (J) | det (X) = det X}
= { IsoR (J) | t 1 (X Y ) = X Y }.
36
Lemma 2.3.1. The Lie algebra e6(26) of the group E6(26) is given by
e6(26) = { HomR (J) | (X, X, X) = 0}
= { HomR (J) | (X, Y, Z) + (X, Y, Z) + (X, Y, Z) = 0}

= { HomR (J) | t (X Y ) = X Y + X Y }.
Proof. It is easy to verify that these conditions in e6(26) are equivalent (see
Lemma 2.2.2). Now, if HomR (J) satisfies ((exp t)X, (exp t)X, (exp t)X) =
(X, X, X) for all t R, then we have (X, X, X) = 0 by putting t = 0 after differentiating with respect to t. Conversely, if HomR (J) satisfies t (X Y ) = X
Y +X Y , then it is easy to verify that = exp t satisfies t 1 (X Y ) = X Y .
Theorem 2.3.2. The Lie algebra f4 of the group F4 is given by
f4 = { HomR (J) | (X Y ) = X Y + X Y }

n
tr(X, Y, Z) + tr(X, Y, Z) + tr(X, Y, Z) = 0 o
= HomR (J)
(X, Y ) + (X, Y ) = 0
= { HomR (J) | (X, X, X) = 0, (X, Y ) + (X, Y ) = 0}
= { HomR (J) | (X, X, X) = 0, E = 0}
= { HomR (J) | (X Y ) = X Y + X Y }.
Proof. The theorem follows easily from Lemma 2.2.4.

We define an R-vector space M by
M = {A M (3, C) | A = A}.
For X, Y M (3, C), we define [X, Y ] M (3, C) by
[X, Y ] = XY Y X.
[M , J] J,
Lemma 2.3.3.
[J, J] M .
e:JJ
Since [M , J] J, any element A M induces an R-linear mapping A
defined by
e = 1 [A, X], X J.
AX
2
Lemma 2.3.4. For X J, there exists a C0 such that
[X, XX] = aE.

Proof. Let X = xij , xij C, xij = xji . The (i, j)-entry aij of [X, XX] =
X(XX) (XX)X is given by
X
X
aij =
(xik (xkl xlj ) (xik xkl )xlj ) =
{xik , xkl , xlj }, i, j = 1, 2, 3.

k,l
k,l
37
Since xii is real, if the bracket { , , } contains xii , then { , , } is 0. If i 6= j, then

aij is a sum of a(xa) (ax)a, a(ax) (aa)x and so on, so aij = 0. If i = j, then
a11 = {x12 , x23 , x31 } {x13 , x32 , x21 },
a22 = {x21 , x13 , x32 } {x23 , x31 , x12 },
a33 = {x31 , x12 , x23 } {x32 , x21 , x13 },
however they are equal, that is, a11 = a22 = a33 ( = a). Hence we have [X, XX] = aE.
Since X, XX J, we have [X, XX] M (Lemma 2.3.3). Therefore, (aE) = aE
and so aE = aE, which imply that a = a.
To prove the following Proposition 2.3.6, in M (3, C), we define a real valued symmetric inner product (X, Y ) by
(X, Y ) =
1
tr(XY + Y X ).
2
Lemma 2.3.5. The inner product (X, Y ) of M (3, C) satisfies

(XY, Z) = (Y Z, X) = (ZX, Y ) = (X, Y Z) = (Y, ZX) = (Z, XY ).

Proof. Let X = xij , Y = yij , Z = zij . Then we have
1
(XY, Z) = R(XY, Z) = R(tr((XY )Z + Z (Y X )))
2

1 X
= R
((xij yjk )zki + z ji (y kj xik ))
2
i,j,k

1 X
= R
((yjk zki )xij + xik (z ji y kj ))
2
i,j,k
1
= R(tr((Y Z)X + X (Z Y ))) = (Y Z, X).
2
e f4 .
Proposition 2.3.6. For A M , tr(A) = 0, we have A
Proof. From the equivalent conditions in Theorem 2.3.2, it is sufficient to show

the following two formulas:
(
([A, X], Y ) + (X, [A, Y ]) = 0, X, Y J,
tr([A, X], Y, Z) + tr(X, [A, Y ], Z) + tr(X, Y, [A, Z]) = 0, X, Y, Z J.
Now, the left side of the first formula = (AX, Y ) (XA, Y ) + (X, AY ) (X, Y A) = 0
(Lemma 2.3.5). Next, we show that
([A, X], XX) = 0,
X J.
Certainly, if [X, XX] = aE, a C0 (Lemma 2.3.4), then

38
([A, X], XX) = (AX, XX) (XA, XX)
= (A, X(XX)) (A, (XX)X) (Lemma 2.3.5)

1
= (A, [X, XX]) = (A, aE) = tr(Aa + aA )
2
1
1
= tr(Aa aA) = (tr(A)a atr(A)) = 0.
2
2
Now, putting X + X + Z in the place of X, and comparing the coefficient of ,
we obtain
([A, X], Y Z + ZY ) + ([A, Y ], XZ + ZX) + ([A, Z], XY + Y X) = 0,
which is the required second formula.
In M , we adopt
0 0
A1 (a) = 0 0
0 a
the following notation:
0 0
0
a , A2 (a) = 0 0
0
a 0
a
0 ,
0
0
A3 (a) = a
0
a
0
0
0
0.
0
ei (a) f4 (Proposition 2.3.6) and the operation of A

ei (a) on J is given by
Then A
ei (a)Ei = 0
ei (a)Fi (x) = (a, x)(Ei+1 Ei+2 )
A
A
ei (a)Ei+1 = 1 Fi (a)
ei (a)Fi+1 (x) = 1 Fi+2 (ax)
A
A
2
2
ei (a)Fi+2 (x) = 1 Fi+1 (xa).

ei (a)Ei+2 = Fi (a),
A
A
2
2
Proposition 2.3.7. The Lie subalgebra d4 of f4 :
d4 = { f4 | Ei = 0, i = 1, 2, 3}
is isomorphic to the Lie algebra D4 = so(8):
D4 = {D HomR (C) | (Dx, y) + (x, Dy) = 0}
under the correspondence
D4 D1 d4
given by
1
x3
x2
x3
2
x1

0
x2
x1 = D3 x3
3
D2 x2
D3 x3
0
D1 x1
D2 x2
D1 x1 ,
0
where D2 , D3 are elements of D4 which are determined by D1 from the principle of

triality :
(D1 x)y + x(D2 y) = D3 (xy), x, y C.
(From now on, we identify these Lie algebras d4 and D4 ).
39
Proof. We define a mapping : D4 d4 by (D1 ) = . We first prove that

d4 f4 . Indeed,
(X, X, X) = (X, X X)

0
D3 x3 D2 x2
2 3 x1 x1

= D3 x3
0
D1 x1 , x1 x2 3 x3
0
x3 x1 2 x2
D2 x2 D1 x1
x1 x2 3 x3
3 1 x2 x2
x2 x3 1 x1
x3 x1 2 x2
x2 x3 1 x1
1 2 x3 x3
= 2(D1 x1 , x2 x3 1 x1 ) + 2(D2 x2 , x3 x1 2 x2 ) + 2(D3 x3 , x1 x2 3 x3 )

= 2((D1 x1 , x2 x3 ) + (D2 x2 , x3 x1 ) + (D3 x3 , x1 x2 ))
= 2((D1 x1 , x2 x3 ) + (D2 x2 , x3 x1 ) + (D3 x3 , x1 x2 ))
which is equal to 0, if we use the relation

(D1 x1 , x2 x3 ) = (x1 , D1 (x2 x3 )) = (x1 , (D2 x2 )x3 + x2 (D3 x3 ))
= (x1 , x3 (D2 x2 ) + (D3 x3 )x2 ) = (x3 x1 , D2 x2 ) (x1 x2 , D3 x3 ).
Hence f4 . Ei = 0, i = 1, 2, 3 are evident, so that d4 . Clearly is injective.
We shall show that is onto. Let d4 . We put
Ji = {Fi (x) | x C} = {X J | 2Ei+1 X = 2Ei+2 X = X}.
Since X Ji for X Ji , induces R-linear mappings : Ji Ji and Di : C C
satisfying
Fi (x) = Fi (Di x), x C,
for i = 1, 2, 3. Applying on Fi (x) Fi (y) = (x, y)(Ei+1 + Ei+2 ), we have Fi (Di x)
Fi (y) + Fi (x) Fi (Di y) = 0, and hence we have
(Di x, y) + (x, Di y) = 0,
x, y C.
1
Hence Di D4 , i = 1, 2, 3. Moreover, by applying on F1 (x) F2 (y) = F3 (xy), we
2
see that
(D1 x)y + x(D2 y) = D3 (xy), x, y C.
This shows that is onto. Thus Proposition 2.3.7 is proved.
Theorem 2.3.8. Any element f4 is uniquely expressed by
e
= D + A,
D d4 , A M , diagA = 0,
where diagA = 0 means that all diagonal elements aii of A are 0. In particular, the
dimension of f4 is
dim f4 = 28 + 24 = 52.
Proof. Applying f4 on Ei Ei = Ei and Ei Ej = 0, i 6= j, we have
2Ei Ei = Ei ,
Ei Ej + Ei Ej = 0.
40
From these relations, we see that each Ei is of the form
0 a3 a2
0 a3
0
0
0 a2
E1 = a3 0
0 , E2 = a3 0 a1 , E3 = 0
0 a1 ,
a2
0
0
0 a1 0
a2 a1 0
0
a3 a2
ai C. If we construct a matrix A = 2 a3 0
a1 using these elements ai ,
a2 a1 0
then A M with diagA = 0, and we have
e i,
Ei = AE
i = 1, 2, 3.
e then DEi = 0, i = 1, 2, 3, hence D d4 . Thus, can be

If we put D = A,
e where D d4 , A M , diagA = 0. To show the uniqueness
expressed by = D + A,
of the expression, it is sufficient to prove that
e = 0, D d4 , A M , diagA = 0, then D = 0, A = 0.
D+A
e i = 0, i = 1, 2, 3, hence A = 0 and so D = 0.
Certainly, if we apply it on Ei , then AE
Finally, we have dim f4 = 28 + 24 = 52 from the expression above. Thus the theorem
is proved.
2.4. Simplicity of f4 C
Let JC = {X1 + iX2 | X1 , X2 J} be the complexification of the Jordan algebra
J. In the same manner as in J, in JC we can also define the multiplications X Y ,
X Y , the inner product (X, Y ), the trilinear forms tr(X, Y, Z), (X, Y, Z) and the
determinant detX. They have the same properties as those of J. JC is called the
complex exceptional Jordan algebra. JC has two complex conjugations, namely,
X1 + iX2 = X 1 + iX 2 , (X1 + iX2 ) = X1 iX2 ,
Xi J.
The complex conjugation of JC satisfies

(X Y ) = X Y, (X Y ) = X Y,
X, Y JC .
We define Lie algebras e6 C and f4 C respectively by

e6 C = { HomC (JC ) | (X, X, X) = 0}
= { HomC (JC ) | t (X Y ) = X Y + X Y },
f4 C = { HomC (JC ) | (X Y ) = X Y + X Y }
= { HomC (JC ) | (X Y ) = X Y + X Y }.
These are the complexification of the Lie algebras e6(26) and f4 respectively. Then the
properties of e6(26) and f4 stated in Section 2.3 also hold for e6 C and f4 C . Hereafter
we will describe f4 C , but the statements are also valid for f4 using J instead of JC .
41
e : JC JC by
For A JC , we define a C-linear mapping A
e = A X,
AX
X JC .
e e6 C .
Proposition 2.4.1. (1) For A JC , tr(A) = 0, we have A
e B]
e f4 C .
(2) For A, B JC , we have [A,
e X, X) = (A X, X X) = (A, X (X X))
Proof. (1) (AX,
= (A, (det X)E) = (det X)(A, E) = (det X)tr(A) = 0,
e e6 C .
Hence A
X JC .
h

i
1
1
A tr(A)E , B tr(B)E
e6 C ,
3
3
e B]E
e = A(
e BE)
e B(
e AE)
e = A B B A = 0.
[A,
C
e B]
e f4 (Theorem 2.3.2).
Hence [A,
(2)
e B]
e =
[A,
Lemma 2.4.2. In f4 C , we have

ei , E
ej ] = 0,
[E
ei+1 (a),
ei , Fei+1 (a)] = 1 A
[E
2
[Fei (a), Fei (b)] d4 C ,
ei (a)] = A
ei (Di a),
[D, A
ei (a), A
ei (b)] d4 C ,
[A
[Fe1 (ei ), Fe1 (ej )] = Gij .
ei , Fei (a)] = 0,
[E
ei , Fei+2 (a)] = 1 A
ei+2 (a),
[E
2
1e
[Fei (a), Fei+1 (b)] = A
i+2 (ab),
2
where D = (D1 , D2 , D3 ) d4 C ,
ei+2 (ab),
ei (a), A
ei+1 (b)] = 1 A
[A
2
ei+1 (ba).
ei (a), A
ei+2 (b)] = 1 A
[A
2
Lemma 2.4.3. (1) Any non-zero element x CC can be transformed to 1 by

successive actions of d4 C .
(2) CC is a d4 C -irreducible C-module.

X
Di ai | Di d4 C , ai CC = CC .
(3) d4 C CC =
i
P7
Proof. (1) Let 0 6= x = i=0 xi ei , xi C. Suppose xi 6= 0 (moreover we may
assume i 6= 0). Then we have
G
i0
xi e0 x0 ei
x
Gj0 (j6=i)
xi ej
xj 1 G0j
e0 = 1.
The case x0 6= 0 is reduced to the above by the action of Gi0 .
(2) Let W 6= {0} be a d4 C -invariant C-submodule of CC . If e0 = 1 W , then

Gi0 e0 = ei . Therefore W = CC . Now, since a non-zero element x of W can be
transformed to 1 by d4 C from (1) above, W contains 1, so that W = CC , because
Gi0 e0 = ei .
42
(3) Since d4 C CC is a d4 C -invariant C-submodule of CC , from the irreducibility of

CC of (2), we have d4 C CC = CC .
Recall that any element f4 C is uniquely expressed by
e1 (a1 ) + A
e2 (a2 ) + A
e3 (a3 ),
=D+A
D d4 C , ai CC ,
where Ai (ai ) (M )C (Theorem 2.3.8).
Theorem 2.4.4. The Lie algebra f4 C is simple and so f4 is also simple.

e C = {A
ei (a) | a CC } and A
eC = A
e C A
e C A
e C , then
Proof. Let denote A
i
1
2
3
C
e A
e A
e = d4 C A
e .
f4 C = d4 C A
1
2
3
e C . Now, let a be
e C be projections of f C = d4 C A
Let p : f4 C d4 C and q : f4 C A
4
a non-zero ideal of f4 C . Then p(a) is an ideal of d4 C . Indeed, if D p(a), then there
3
X
ei (ai ) a. For any D d4 C , we have
A
exist ai CC , i = 1, 2, 3, such that D +
i=1
3
3
i
h
X
X
ei (Di ai ) (Lemma 2.4.2),
ei (ai ) = [D , D] +
A
A
a D , D +
i=1
i=1
hence [D , D] p(a).
e C a 6= {0}. Assume that d4 C a = {0}

We show that either d4 C a 6= {0} or A
e C a = {0}. The mapping p|a : a d4 C is injective because AC a = {0}.
and A
Since p(a) is a non-zero ideal of d4 C and d4 C is simple, we have p(a) = d4 C . Hence
dimC a = dimC p(a) = dimC d4 C = 28. On the other hand, since d4 C a = {0},
e C = 8 3 = 24. This leads
e C is also injective, we have dimC a dimC A
q|a : a A
to a contradiction.
(1) Case d4 C a 6= {0}. From the simplicity of d4 C , we have d4 C a = d4 C , hence

a d4 C . On the other hand, we have
e C] = A
e C,
a [a, f4 C ] [d4 C , A
i
i
i = 1, 2, 3.
ei (ai )] = A
ei (Dai ) (Lemma 2.4.2) and d4 C CC = CC
The last equation follows from [D, A
C
e A
e C A
e C = f C.
(Lemma 2.4.3.(3)). Hence a d4 C A
1
e C a 6= {0}. Choose a non-zero element

(2) Case A
e C a a.
e1 (a1 ) + A
e2 (a2 ) + A
f3 (a3 ) A
A
If a1 6= 0, under the actions of d4 C of Lemma 2.4.3.(1), we have

e1 (1) + A
e2 (b) + A
e3 (c) a.
A
43
e1 (1) a, hence from Lemma 2.4.2,

If b = c = 0, then A
e1 (1), A
e1 (e1 )] d4 C a
0 6= [A
e1 (1), A
e1 (e1 )]F2 (1) = 2F2 (e1 )), so the case is
(as for the first inequality, note that [A
e1 (1), then
reduced to the case (1). If b 6= 0, then take the Lie bracket with A
e3 (b) + A
e2 (c) a
A
and then
e3 (1) + A
e2 (c ) a
A
under some actions of d4 C . If c = 0, then this can be reduced to the case (1) as in
the above. The case c 6= 0 can also be reduced to the case (1) by the same argument
as above. Thus we have a = f4 C , which proves the simplicity of f4 C .
Lemma 2.4.5. (1) For f4 C and A, B JC , we have
f B]
f
e B]
e ] = [A,
e + [A,
e B].
[, [A,
X
ei , B
ei ], Ai , Bi JC .
[A
(2) Any element f4 C is expressed as =
i
e B]
e ]X = [A,
e B]X
e [A,
e B]X
e
Proof. (1) [, [A,
= (A (B X) B (A X)) (A (B X) B (A X))
= A (B X) + A (B X) + A (B X) B (A X)
B (A X) B (A X) A (B X) + B (A X)
= A (B X) B (A X) + A (B X) B (A X)
f B]X
f
e + [A,
e B]X,
= [A,
X JC .
P e e
(2) By (1) we see that a = { i [Ai , Bi ] | Ai , Bi JC } is an ideal of f4 C . From the
simplicity of f4 C , we have a = f4 C .
Proposition 2.4.6. (1) J0 C = {X JC | tr(X) = 0} is an f4 C -irreducible Cmodule.

o
nX
i Bi | i f4 C , Bi J0 C = J0 C .
(2) f4 C J0 C =
i
Proof. (1) Evidently J0 C is an f4 C -C-module (Lemma 2.2.1.(2)). Now, let W be

a non-zero f4 C -invariant C-submodule of J0 C .
Case (i) If W contains some Fi (x) 6= 0 (i = 1, 2, 3), then W = J0 C . Indeed, under

some actions of d4 C , we can assume Fi (1) W (Lemma 2.4.3.(1)). Next, from
ei+2 (1)
ei (1)
A
A
Fi (1) Ei+1 Ei+2 Fi+2 (1) Fi+1 (1),
ei (a)Fi+2 (1) = Fi+1 (a), etc., we have Ei Ei+1 , Fi (a) W and hence we have
and A
W = J0 C . Now, in the general case, choose a non-zero element X from W:
0 6= X = (E1 E2 ) + (E2 E3 ) + F1 (x1 ) + F2 (x2 ) + F3 (x3 ) W.
44
Case x1 6= 0. Under the action of d4 C of Lemma 2.4.3.(1), we have

F1 (1) + F2 (y2 ) + F3 (y3 ) W.
If y2 = y3 = 0, then F1 (1) W , hence we can reduce to the case (i). If y2 6=
e2 (a), then we have
0, choose a CC , a 6= 0, such that (y2 , a) = 0, and apply A
C
F3 (a) + F1 (ay3 ) W . Again under the action of d4 , we have
F3 (1) + F1 (z1 ) W.
If z1 = 0, we can reduce to the case (i). If z1 6= 0, choose b CC , b 6= 0, such that
e1 (b), then F2 (b) W , which is again reduced to the case (i).
(z1 , b) = 0 and apply A
Similarly for x2 6= 0 or x3 6= 0 the arguments above hold.
e1 (1) on the non-zero element
Case x1 = x2 = x3 = 0. Applying A
X = (E1 E2 ) + (E2 E3 ) W,
we have F1 ( 2) W . If 2 6= 0, we can reduce to the case (i). If 2 = 0,

e3 (1), we have F3 (3) W which is again the case (i). We have just proved
applying A
that W = J0 C .
(2) Since f4 C J0 C is an f4 C -invariant C-submodule of J0 C , we have f4 C J0 C = J0 C
from (1).
2.5. Killing form of f4 C
Lemma 2.5.1. For A, B, C, D JC , we have
Proof.
e B]C,
e D) = ([C,
e D]A,
e
([A,
B).
e B]C,
e D) = (A (B C), D) (B (A C), D)
([A,
= (B C, A D) (A C, B D)
e D]A,
e
= (C (D A), B) (D (C A), B) = ([C,
B).
Definition. In f4 C , we define an inner product (1 , 2 )4 by

e B])
e 4 = (A, B),
(, [A,
f4 C , A, B JC .
More precisely, we define
(1 , 2 )4 =
X
i,j
ei , B
ei ]Cj , Dj ) =
([A
X
i,j
ej , D
e j ]Ai , Bi ),
([C
P e e
P e e
C
for 1 = i [A
i , Bi ], 2 =
j [Cj , Dj ], Ai , Bi , Cj , Dj J (Lemma 2.4.5.(2)). Then,
Lemma 2.5.1 shows that the definition of the inner product (1 , 2 )4 is independent
of the choice of the expressions of 1 , 2 , and that (1 , 2 )4 is symmetric.
45
Lemma 2.5.2. The inner product (1 , 2 )4 of f4 C is f4 C -adjoint invariant:

([, 1 ], 2 )4 + (1 , [, 2 ])4 = 0,
J.
, i f4 C .
e B],
e 2 = [C,
e D],
e A, B, C, D
Proof. It is sufficient to show the lemma for 1 = [A,
e B]
e ], [C,
e D])
e 4 + ([A,
e B],
e [, [C,
e D]
e ])4
([, [A,
f B]
f [C,
f D]
f 4 (Lemma 2.4.5.(1))
e + [A,
e B],
e D])
e 4 + ([A,
e B],
e [C,
e + [C,
e D])
= ([A,
f B]C,
f
e D) + ([A,
e B]C,
e B]C,
e
e B]C,
e D)
= ([A,
D) + ([A,
D) + ([A,
= (A (B C), D) (B (A C), D) + (A (B C), D) (B (A C), D)

+(A (B C), D) (B (A C), D) + (A (B C), D) (B (A C), D),
which is equal to 0, if we use the relation (X Y, Z) = (X Y, Z) (X Y, Z)

for the above last two terms.
Theorem 2.5.3. The Killing form B4 of the Lie algebra f4 C is given by
1 , 2 f4 C .
B4 (1 , 2 ) = 9(1 , 2 )4 = 3tr(1 2 ),
Proof. Since f4 C is simple (Theorem 2.4.4), there exist k, k C such that

B4 (1 , 2 ) = k(1 , 2 )4 = k tr(1 2 ).
e1 (1). Since A
e1 (1) = 2[E
e3 , Fe1 (1)]
To determine these k, k , we put = 1 = 2 = A
(Lemma 2.4.2), we have
e1 (1), A
e1 (1))4 = 2(A
e1 (1), [E
e3 , Fe1 (1)])4
(, )4 = (A
e1 (1)E3 , F1 (1)) = (F1 (1), F1 (1)) = 2.
= 2(A
On the other hand, (ad)2 is calculated as follows.
e1 (1), [A
e1 (1), Gi0 ] ] = [A
e1 (1), A
e1 (ei )] = Gi0 , i 6= 0
[A
e1 (1), [A
e1 (1), A
e1 (ei )] ] = [A
e1 (1), Gi0 ] = A
e1 (ei ), i 6= 0
[A
e2 (ei ),
e1 (1), A
e3 (ei )] = 1 A
e1 (1), [A
e1 (1), A
e2 (ei )] ] = 1 [A
[A
2
4
e1 (1), [A
e1 (1), A
e3 (ei )] ] = 1 [A
e3 (ei ),
e1 (1), A
e2 (ei )] = 1 A
[A
2
4
the others = 0.
Hence we have
e1 (1))2 ) = (1) 7 2 +
B4 (, ) = tr((adA
Therefore k = 9. Next, we will calculate tr().
46
1
8 2 = 18.
4
e1 (1)F1 (1) = 1 (E2 E3 ),

e1 (1)A
e1 (1)E2 = 1 A
A
2
2
1
1
e1 (1)F1 (1) = (E2 E3 ),
e1 (1)A
e1 (1)E3 = A
A
2
2
e1 (1)A
e1 (1)F1 (1) = A
e1 (1)(E2 E3 ) = F1 (1),
A
e1 (1)A
e1 (1)F2 (ei ) = 1 A
e1 (1)F3 (ei ) = 1 F2 (ei ),
A
2
4
1e
1
e
e
A1 (1)A1 (1)F3 (ei ) = A1 (1)F3 (ei ) = F3 (ei ),
2
4
the others = 0.
Hence we have
e1 (1))2 ) =
tr() = tr((A
Therefore k = 3.
1
1
21+
8 2 = 6.
2
4
Lemma 2.5.4. For a non-zero element A J0 C , there exists B J0 C such that

e B]
e 6= 0.
[A,
e B]
e = 0 for all B J0 C . Then 0 = (, [A,
e B])
e 4 =
Proof. Assume that [A,
C
C
C
C
e A])
e 4 = (B, A) for any f4 . Since f4 J0 = J0 (Proposition 2.4.6.(2)),
(, [B,
we have (J0 C , A) = 0, so that A = 0.
2.6. Roots of f4 C
Before we obtain the roots of the Lie algebra f4 C , we recall the roots of the Lie
algebra D4 C :
D4 C = {D HomC (CC ) | (Dx, y) + (x, Dy) = 0}.
If we let Hk = iGk4+k for k = 0, 1, 2, 3, then
h = {H =
3
X
k=0
i Hk | k C}
is a Cartan subalgebra of D4 C , and roots of D4 C relative to h are given by

(k l ), (k + l ),
0 k < l 3.
The root vectors associated with these roots are respectively given by
k l
: (Gkl + G4+k4+l ) i(Gk4+l + Gl4+k ),
k + l
: (Gkl G4+k4+l ) + i(Gk4+l Gl4+k ),
k + l
k l
: i(Gkl + G4+k4+l ) (Gk4+l + Gl4+k ),
: i(Gkl G4+k4+l ) + (Gk4+l Gl4+k ).

47
The Lie algebra D4 C is contained in f4 C as

D4 C D (D, D, D) d4 C f4 C
(Theorem 1.3.6, Proposition 2.3.7). Outer automorphisms , of D4 C induce Clinear transformations of h and the matrices of , with respect to the C-basis
H0 , H1 , H2 , H3 of h are respectively given by
1 1
1 1
1
= =
1
2 1
1 1
1 1
1 1
,
1
1
1
1
1 1 1
1 1
1 1
=
1
1 1
2
1 1 1
1
1
1
1
(Lemma 1.3.1). Note that they are orthogonal matrices.

Theorem 2.6.1. The rank of the Lie algebra f4 C is 4. The roots of f4 C relative
to some Cartan subalgebra of f4 C are given by
(k l ), (k + l ) , 0 k < l 3,
0 , 1 , 2 , 3 ,
1
1
( 0 + 1 + 2 3 ),
(0 1 + 2 3 ),
2
2
1
1
(0 + 1 + 2 + 3 ),
( 0 1 + 2 + 3 ),
2
2
1
1
(0 + 1 + 2 3 ),
( 0 1 + 2 3 ),
2
2
1
1
( 0 + 1 + 2 + 3 ),
(0 1 + 2 + 3 ).
2
2
Proof. We use the decomposition in Theorem 2.4.4:
Let h =
e C A
e C A
e C.
f4 C = d4 C A
1
2
3
3
n
o
X
H =
k Hk | k C d4 C f4 C . Since h is a Cartan subalgebra of
k=0
d4 C (it will be also a Cartan subalgebra of f4 C ), the roots of d4 C :

(k l ), (k + l ),
0k<l3
e1 (a)] = A
e1 (Ha) (Lemma 2.4.2), where
are also roots of f4 C . Furthermore, from [H, A
H(ek + ie4+k ) =
3
X
k iGj4+j (ek + ie4+k )
j=0
= ik (e4+k + iek ) = k (ek + ie4+k ),

e1 (ek + ie4+k ) is an associated root vector for
we see that k is a root of f4 C and A
e1 (ek ie4+k ) is
0 k 3. Similarly, k for 0 k 3 is root of f4 C and A
48
e2 (a)] = A
e2 ((H)a),
an associated root vector. Next, by Lemma 2.4.2 we have [H, A
where
H =
3
X
k Hk
k=0
1
= (0 (H0 + H1 H2 + H3 ) + 1 (H0 + H1 + H2 H3 )
2
+2 (H0 + H1 + H2 + H3 ) + 3 (H0 H1 + H2 + H3 ))
1
1
= (0 1 + 2 3 )H0 + (0 + 1 + 2 3 )H1
2
2
1
1
+ (0 + 1 + 2 + 3 )H2 + (0 1 + 2 + 3 )H3 ,
2
2
e2 (ek +
and so we see that coefficients of H0 , H1 , H2 , H3 are roots of f4 C and that A
i + ie4+k ) is associated root vector for 0 k 3. The roots above with negative sign
e2 (ek ie4+k ) are associated root vectors. Finally, from
are also roots of f4 C and A
e3 (a)] = A
e3 ((H)a) (Lemma 2.4.2), where
[H, A
1
1
(0 + 1 2 + 3 )H0 + (0 + 1 + 2 3 )H1
2
2
1
1
+ (0 + 1 + 2 + 3 )H2 + (0 1 + 2 + 3 )H3 ,
2
2
H =
we obtain the remainders of roots.

Theorem 2.6.2. In the root system of Theorem 2.6.1,
1 = 0 1 , 2 = 1 2 , 3 = 2 , 4 =
1
(0 1 2 + 3 )
2
is a fundamental root system of the Lie algebra f4 C and

= 21 + 32 + 43 + 24
is the highest root. The Dynkin diagram and the extended Dynkin diagram of f4 C are
i
3
4
i
3
2
4
Proof. All positive roots of f4 C are expressed by 1 , 2 , 3 , 4 as follows.

0
1
3
3
= 1 + 2 +
=
2 +
=
= 1 + 22 +
49
3
3
3
33 + 24 ,
0 1
0 2
0 + 3
1 2
1 + 3
2 + 3
=
=
=
=
=
=
0 + 1
0 + 2
0 + 3
1 + 2
1 + 3
2 + 3
=
=
=
=
=
=
1
1 +
1
1
2
2 + 23 + 24
2
+ 2 + 23 + 24
+ 22 + 23 + 24
1 + 22
1 + 2
21 + 32
2
1 + 32
1 + 22
1
( 0 + 1 + 2 + 3 )
2
1
(0 1 2 + 3 )
2
1
( 0 + 1 2 + 3 )
2
1
( 0 1 + 2 + 3 )
2
1
(0 1 + 2 + 3 )
2
1
(0 + 1 2 + 3 )
2
1
( 0 1 2 + 3 )
2
1
(0 + 1 + 2 + 3 )
2
+
+
+
+
+
+
23
23
43 + 24
23
43 + 24
43 + 44 ,
= 1 + 22 + 33 + 4
=
= 1 + 22 + 23 + 4
= 1 +
2 + 23 + 4
3 + 4
2 + 3 + 4
1 + 2 + 3 + 4
2 + 23 + 4 .
Hence = {1 , 2 , 3 , 4 } is a fundamental root system of f4 C . The real part hR of

h is
3
X
k Hk | k R}.
hR = {H =
k=0
The Killing form B4 of f4 C is B4 (1 , 2 ) = 3tr(1 2 ) (Theorem 2.5.3), so that

B4 (H, H ) = 18
3
X
k=0
k k ,
H=
3
X
k=0
k Hk , H =
3
X
k=0
k Hk hR .
Indeed, since
HEi = 0, i = 1, 2, 3,
HF1 (x) = F1 (Hx), HF2 (x) = F2 ((H)x), HF3 (x) = F3 ((H)x),
we have
50
B4 (H, H ) =
3
X
k l B4 (Hk , Hl )
k,l=0
X
= 3 (k l (tr(Hk Hl ) + tr((Hk )(Hl )) + tr((Hk )(Hl ))))
k,l
3
X
X
= 3 k l (2 + 2 + 2)kl = 18
k k .
k=0
k,l
Now, the canonical elements Hi hR corresponding to i (B4 (H , H) = (H), H

hR ) are determined by
1
(H0 H1 ),
18
1
=
H2 ,
18
1
(H1 H2 ),
18
1
=
(H0 H1 H2 + H3 ).
36
H1 =
H2 =
H3
H4
Hence we have
1 1
1
(1 + 1) =
18 18
9
and the other inner products are similarly calculated. Hence, the inner products
induced by the Killing form B4 between 1 , 2 , 3 , 4 and are given by
(1 , 1 ) = B4 (H1 , H1 ) = 18
(1 , 1 ) = (2 , 2 ) =
1
,
9
1
,
18
1
(3 , 4 ) = ,
36
(3 , 3 ) = (4 , 4 ) =
1
1
, (2 , 3 ) = ,
18
18
(1 , 3 ) = (1 , 4 ) = (2 , 4 ) = 0,
1
1
(, ) = , (, 1 ) = , (, i ) = 0, i = 2, 3, 4,
9
18
using them, we can draw the Dynkin diagram and the extended Dynkin diagram of
f4 C .
(1 , 2 ) =
According Borel-Siebenthal theory, the Lie algebra f4 has three subalgebras as

maximal subalgebras with the maximal rank 4.
(1) The first one is a subalgebra of type B4 obtained as the fixed points of an
involution of f4 .
(2) The second one is a subalgebra of type C1 C3 obtained as the fixed points
of an involution of f4 .
(3) The third one is a subalgebra of type A2 A2 obtained as the fixed points of
an automorphism w of order 3 of f4 .
These subalgebras will be realized as subgroups of the group F4 in Theorems 2.9.1,
2.11.2 and 2.12.2, respectively.
2.7. Subgroup Spin(9) of F4
51
Theorem 2.7.1.
{ F4 | Ei = Ei , i = 1, 2, 3}
= Spin(8).
Proof. We recall the group Spin(8) of Theorem 1.16.2 and define a mapping
: Spin(8) D4 = { F4 | Ei = Ei , i = 1, 2, 3} by

1
3 x3 2 x2
1 x3 x2
(1 , 2 , 3 ) x3 2 x1 = 3 x3
2
1 x1 .
x2 x1 3
3
2 x2 1 x1
We first prove that = (1 , 2 , 3 ) D4 . Indeed, the fact that det (X) = det X
can be seen by observing that
R((1 x1 )(2 x2 )(3 x3 )) = R((3 (x1 x2 ))3 x3 )
= (3 (x1 x2 ), 3 x3 ) = (x1 x2 , x3 ) = R(x1 x2 x3 ),
which together with E = E, shows that F4 and Ei = Ei , i = 1, 2, 3. Therefore
D4 . Certainly is a homomorphism. We shall show that is onto. Let D4 .
We put
Ji = {Fi (x) | x C} = {X J | 2Ei+1 X = 2Ei+2 X = X},
i = 1, 2, 3.
Since X Ji , X Ji , induces R-isomorphisms : Ji Ji and i : C C

satisfying
Fi (x) = Fi (i x), x C,
for i = 1, 2, 3. Applying on Fi (x) Fi (y) = (x, y)(Ei+1 + Ei+2 ), since the left side
becomes Fi (i x) Fi (i y) = (i x, i y)(Ei+1 + Ei+2 ), we have
(i x, i y) = (x, y),
x, y C.
1
Hence i O(8), i = 1, 2, 3. Moreover, by applying on F1 (x) F2 (y) = F3 (xy),
2
we see that
(1 x)(2 y) = 3 (xy),
x, y C.
From Lemma 1.14.4, we have 1 , 2 , 3 SO(8), so that (1 , 2 , 3 ) Spin(8) and

(1 , 2 , 3 ) = . Therefore is onto. Evidently Ker = {(1, 1, 1)}. Consequently
is an isomorphism.
We shall study the following subgroup (F4 )E1 of F4 :
(F4 )E1 = { F4 | E1 = E1 }.
We define R-vector subspaces J01 , J23 of J respectively by

o
n 0 0 0

J01 = {X J | E1 X = 0, tr(X) = 0} = 0 x R, x C ,
0 x
52
J23
n 0
= {Y J | 2E1 Y = Y } = y 3
y2
y3
0
0
y2
o

0 y2 , y3 C .
0
Lemma 2.7.2. Suppose that we are given an element A J01 such that (A, A) =
2. Choose any element X0 J01 such that (X0 , X0 ) = 2, (A, X0 ) = 0. Choose any
element Y0 J23 such that (Y0 , Y0 ) = 2, 2A Y0 = Y0 . Let
Z0 = 2X0 Y0 .
Choose any element X1 J01 such that (X1 , X1 ) = 2, (A, X1 ) = (X0 , X1 ) = 0.
Choose any element X2 J01 such that (X2 , X2 ) = 2, (A, X2 ) = (X0 , X2 ) =
(X1 , X2 ) = 0. Let
Y1 = 2Z0 X1 ,
Z2 = 2X2 Y0 ,
X3 = 2Y1 Z2 .
Finally, choose any element X4 J01 such that (X4 , X4 ) = 2, (A, X4 ) = (X0 , X4 ) =
(X1 , X4 ) = (X2 , X4 ) = (X3 , X4 ) = 0. Let
Z4 = 2X4 Y0 , Y2 = 2Z0 X2 , Y3 = 2Z0 X3 ,
X5 = 2Y1 Z4 , X6 = 2Y2 Z4 ,
X7 = 2Y3 Z4
and moreover let
Yi = 2Z0 Xi ,
Zi = 2Xi Y0 ,
i = 4, 5, 6, 7,
i = 1, 3, 5, 6, 7.
Then, an R-linear mapping : J J satisfying

E = E,
E1 = E1 ,
F1 (ei ) = Xi , F2 (ei ) = Yi ,
(E2 E3 ) = A,
F3 (ei ) = Zi , i = 0, 1, , 7
belongs to (F4 )E1 .

Proof. We have to show that the setting of the lemma is appropriate and prove
that satisfies
(X Y ) = X Y, X, Y J.
For this purpose, it is sufficient to show that this holds for generators X, Y = Ei ,
Fj (ek ) of J. We have to check 272 = 729 times if honestly doing so we reduce the
number of times after preparing some lemma. However we omit its proof. In details,
see Yokota [40].
Proposition 2.7.3.
(F4 )E1 /Spin(8) S 8 .
In particular, the group (F4 )E1 is connected.
Proof. S 8 = {X J01 | (X, X) = 2} is an 8 dimentional sphere. For (F4 )E1

and X S 8 , we have X S 8 . Hence the group (F4 )E1 acts on S 8 . This action is
transitive. Indeed, for a given A S 8 , by constructing (F4 )E1 of Lemma 2.7.2,
53
we have (E2 E3 ) = A, and which shows the transitivity. We determine the isotropy
subgroup of (F4 )E1 at E2 E3 S 8 . Let (F4 )E1 satisfy (E2 E3 ) = E2 E3 .
Since (F4 )E1 satisfies E1 = E1 and E = E, it also satisfies (E2 + E3 ) = E2 +
E3 . Therefore we have E2 = E2 and E3 = E3 , so that Spin(8). Conversely,
Spin(8) satisfies (E2 E3 ) = E2 E3 . Thus we have the homeomorphism
(F4 )E1 /Spin(8) S 8 .
Remark. If we know the dimension of the group (F4 )E1 , without using Lemma
2.7.2, Proposition 2.7.3 can be simply proved as follows. The group (F4 )E1 acts on S 8 .
The isotropy subgroup of (F4 )E1 at E2 E3 is Spin(8) and dim((F4 )E1 /Spin(8)) =
dim(F4 )E1 dim Spin(8) = 36 28 = dim S 8 . Therefore, we have (F4 )E1 /Spin(8)
S8.
Theorem 2.7.4.
(F4 )E1
= Spin(9).

Proof. Let O(9) = O(J01 ) = { IsoR (J01 ) | ( X, Y ) = (X, Y )}. Consider
the restriction = |J01 of (F4 )E1 to J01 , then O(9). Hence we can define
a homomorphism p : (F4 )E1 O(9) by p() = . Since p is continuous and (F4 )E1
is connected (Proposition 2.7.3), the mapping p induces a homomorphism
p : (F4 )E1 SO(9).
We shall show p is onto. Let SO(8) = { SO(9) | (E2 E3 ) = E2 E3 }. The
restriction p of p : (F4 )E1 SO(9) to Spin(8) = { (F4 )E1 | (E2 E3 ) = E2 E3 }
coincides with the homomorphism p : Spin(9) SO(8) in Theorem 1.16.2. In
particular, p : Spin(8) SO(8) is onto. Hence, from the following commutative
diagram
1
Spin(8)
p
SO(8)
(F4 )E1
p
SO(9)
S8
=
S8
we see that p : (F4 )E1 SO(9) is onto by the five lemma. Ker p = {1, }, where =
(1, 1, 1) (which corresponds to defined in the following Section 2.9). Indeed,
let Ker p, then satisfies X = X for all X J01 . From (E2 E3 ) = E2 E3
follows that Spin(8). Let = (1 , 2 , 3 ) Spin(8). Since F1 (x) = F1 (x),
that is, F1 (1 x) = F1 (x), so 1 x = x for all x C, hence we have 1 = 1. From the
principle of triality, we have = (1, 1, 1) = 1 or = (1, 1, 1) = . Hence we have
Ker p = {1, }, and so we have the isomorphism
(F4 )E1 /{1, }
= SO(9).
Therefore (F4 )E1 is isomorphic to the group Spin(9) as the universal covering group
of SO(9).
54
Spin(9)/Spin(7) S 15 .
Theorem 2.7.5.
where Spin (7) = {e

SO(8) | (e
x)(y) =
e(xy), x, y C for some SO(7)}.
Proof. Let S 15 = {Y J23 | (Y, Y ) = 2}. For Spin(9) and Y S 15 , we have

Y S 15 . Hence Spin(9) acts on S 15 . We shall show that this action is transitive.
Let Y0 S 15 . Choose any element A J01 such that (A, A) = 2, 2A Y0 = Y0 .
Using these A and Y0 , construct Xi , Yi , Zi and of Lemma 2.7.2, then Spin(9)
and satisfies F2 (1) = Y0 , which shows the transitivity. We determine the isotropy
subgroup Spin(9)F2 (1) of Spin(9) at F2 (1) S 15 . Let Spin(9) satisfies F2 (1) =
F2 (1). Applying on F2 (1)F2 (1) = E1 +E3 we have F2 (1)F2 (1) = E1 +E3 , so we
get E3 = E3 . Since always satisfies (E2 +E3 ) = E2 +E3 , we have E2 = E2 , and
so Spin(8). Denote = (1 , 2 , 3 ), then F2 (2 (1)) = F2 (1) implies 2 1 = 1,
so that 2 SO(7). Hence = (1 , 2 , 3 ) Spin (7). Conversely Spin(7)
satisfies F2 (1) = F2 (1). Therefore Spin(9)F2 (1) = Spin (7) is proved. Thus we have
the homeomorphism Spin(9)/Spin(7) S 15 .
(Spin (7) and Spin(7) are conjugate in the group O(7). Indeed, a mapping f :
Spin (7) Spin(7),
f ( ) = 1 , (x = x, x C)
gives the conjugation. In particular, Spin (7)
= Spin(7))
2.8. Connectedness of F4
We denote by (F4 )0 the connected component of F4 containing the identity 1.
Lemma 2.8.1. For
Y (, y), where
y1
y2
y3
a C, we define a mapping 1 (a) : J J by 1 (a)X(, x) =
= 1
2 3
(a, x1 )
2 + 3
+
cos 2|a| +
sin 2|a|
=
2
2
|a|
2 3
(a, x1 )
2 + 3
cos 2|a|
sin 2|a|,
=
2
2
|a|
(2 3 )a
2(a, x1 )a
= x1
sin2 |a|
sin 2|a|
2|a|
|a|2
x3 a
sin |a|
= x2 cos |a|
|a|
ax2
sin |a|
= x3 cos |a| +
|a|

sin |a|
if a = 0, then
means 1 , then 1 (a) (F4 )0 .
|a|
0 0 0
e1 (a) f4 (Proposition 2.3.6) and
Proof. For A1 (a) = 0 0 a , we have A
0 a 0
55
e1 (a). Hence 1 (a) (F4 )0 .

1 (a) = exp A
Proposition 2.8.2. Any element

by some element (F4 )0 :
1
X = 0
0
X J can be transformed to a diagonal form

0
2
0
0
0 ,
3
i R.
Moreover, 1 , 2 , 3 are uniquely determined (up to their permutation) independent of

the choice of (F4 )0 .
Proof. For a given X J, consider a space X = {X | (F4 )0 }. Since (F4 )0 is

compact (Theorem 2.2.5), X is also compact. Let 1 2 + 2 2 + 3 2 be the maximal value
of all 1 2 + 2 2 + 3 2 for Y = Y (, y) X and let X0 = X(, x) be an element of X
which attains its maximal value. Then X0 is of diagonal form. Certainly, suppose X0
is not of diagonal form, for example, the 2 3 entry x1 of X0 is non-zero: x1 6= 0. Let
x1
a(t) =
t, t > 0, and construct 1 (a(t)) (F4 )0 of Lemma 2.8.1. Since |a(t)| = t
|x1 |
(a(t), x1 )
and
= |x1 | for Y ((t), y(t)) = 1 (a(t))X0 X, we have
|a(t)|
1 (t)2 + 2 (t)2 + 3 (t)2
+
2
2 3
2
3
= 1 2 +
+
cos 2t + |x1 | sin 2t
2
2
2
+
2 3
2
3
cos 2t |x1 | sin 2t

+
2
2

2
+ 2
2
3
2
3
+2
cos 2t + |x1 | sin 2t
= 1 2 + 2
2
2
+ 2

2
2
3
2
3
= 1 2 + 2
+2
+ |x1 |2 sin2 (2t + t0 ) (for some t0 R)
2
2

2
+ 2
2
3
2
3
2
+2
+ |x1 |2
1 + 2
2
2
= 1 2 + 2 2 + 3 2 + 2|x1 |2
which is the maximal value and attains at some t > 0. This contradicts the maximum
of 1 2 + 2 2 + 3 2 . Hence x1 = 0. x2 = x3 = 0 can be similarly proved by constructing
2 (a), 3 (a) (F4 )0 analogous to 1 (a) of Lemma 2.8.1. Hence X0 is of diagonal form.
We now give the proof of
the latter halfof the proposition. If X J is transformed
1 0 0
to a diagonal form X = 0 2 0 by (F4 )0 , then
0 0 3
P3
tr(X) = tr(X) (Lemma 2.2.1.(2)) = i=1 i ,
P3
(X, X) = (X, X) (Lemma 2.2.4) = i=1 i 2 ,
P3
tr(X, X, X) = tr(X, X, X) (Lemma 2.2.4) = i=1 i 3 .
56
Hence, 1 , 2 , 3 are uniquely determined (up to order) as the solutions of the following
simultaneous equation:
1 + 2 + 3 = tr(X)
1 2 + 2 2 + 3 2 = (X, X)
3
1 + 2 3 + 3 3 = tr(X, X, X).
The space CP2 , called the Cayley projective plane, is defined as

CP2 = {X J | X 2 = X, tr(X) = 1}.
Theorem 2.8.3.
F4 /Spin(9) CP2 .
In particular, the group F4 is connected.
Proof. For F4 and X CP2 , we have X CP2 . Hence F4 acts on CP2 .

We shall prove that the group (F4 )0 acts transitively on CP2 . To prove this, it is
sufficient to show that any element X CP2 can be transformed to E1 CP2 by
some (F4 )0 . Now, X CP2 J can be transformed to a diagonal form by
(F4 )0 :
1 0 0
X = 0 2 0 , i R
0 0 3
(Proposition 2.8.2). From the
is,
2
1
0
0
condition X X = X, we have X X = X, that

0
2 2
0

0
1
0 = 0
3 2
0
0
2
0
0
0 .
3
Hence i 2 = i , so that i = 1 or i = 0, i = 1, 2, 3. Next, from tr(X) = tr(X) = 1,

we have i = 1, i+1 = i+2 = 0 for some i, that is, X = Ei . Moreover, E2 ,
E3 are transformed to E1 respectively
by (F4 )0 . Certainly,
if we define a mapping
0 1 0
: J J, X = T XT 1, where T = 1 0 0 SO(3), then (F4 )0 and
0 0 1
E2 = E1 . Hence X = E1 . In the case X = E3 , the situation is similar to the
above. Therefore the transitivity is proved. Since we have CP2 = (F4 )0 E1 , CP2 is
connected. Now, the group F4 acts transitively on CP2 and the isotropy subgroup
of F4 at E1 CP2 is Spin(9) (Theorem 2.7.4). Thus we have the homeomorphism
F4 /Spin(9) CP2 . Finally, the connectedness of F4 follows from the connectedness
of CP2 and Spin(9).
2.9. Involution and subgroup Spin(9) of F4
57
Definition. We define an R-linear transformation of J by
1 x3 x2
1
x3 x2
x3 2 x1 = x3
2
x1 .
x2 x1 3
x2 x1
3
Then F4 and 2 = 1. Observe that this is the same as = (1, 1, 1)

Spin(8) Spin(9) F4 of Theorem 2.7.4.
We shall study the following subgroup (F4 ) of F4 :
(F4 ) = { F4 | = }.
For this end, we consider R-vector subspaces J and J
of , respectively by
0
n 1 0
J = {X J | X = X} = 0 2 x1
0 x1 3
of J, which are eigenspaces

o

i R, x1 C
= {X J | E1 X = 0} E1 (where E1 = {E1 | R})
= J(2, C) E1 ,
J = {X J | X = X} =
= {X J | 2E1 X = X}.
0
x3
x2
x3
0
0
x2
o

0 xi C
0
Then, J = J J and J , J are invariant under the action of (F4 ) .

Theorem 2.9.1.
(F4 ) = (F4 )E1 = Spin(9).
Proof. We shall show that for (F4 ) we have E1 = E1 . Let (F4 ) . We

first show that
E2 , E3 J(2, C).
Certainly we have
E2 = (F2 (1) F2 (1)) = F2 (1) F2 (1)
= (F2 (x2 ) + F3 (x3 )) (F2 (x2 ) + F3 (x3 )) (for some x2 , x3 C)

= (x2 , x2 )E2 + (x3 , x3 )E3 F1 (x2 x3 ) J(2, C).
Similarly E3 J(2, C). So E1 = (E E2 E3 ) = E E2 E3 is of the form

E1 = E1 + 2 E2 + 3 E3 + F1 (x1 ).
Then we have
1 = (E1 , E1 ) = (E1 , E1 ) = 1 + 2 2 + 3 2 + 2|x1 |2 ,
58
so that 2 = 3 = x1 = 0, therefore E1 = E1 , that is, (F4 )E1 . Conversely

let (F4 )E1 , that is, F4 satisfies E1 = E1 . Noting that J = J J and
J = {X J | E1 X = 0} E1 , J = {X J | 2E1 X = X} are invariant under
, we have
X = (X1 + X2 )
X1 J , X2 J
= (X1 X2 ) = X1 X2 = (X1 ) + (X2 )

= (X1 + X2 ) = X,
X J.
Thus = , and so (F4 ) . Therefore we have shown that (F4 ) = (F4 )E1
=
Spin(9) (Theorem 2.7.4).
2.10. Center z(F4 ) of F4
Theorem 2.10.1. The center z(F4 ) of the group F4 is trivial:
z(F4 ) = {1}.
Proof. Let z(F4 ). From the commutativity with : = , we have
Spin(9) (Theorem 2.9.1) and so z(Spin(9)). Since the center z(Spin(9)) of
Spin(9) is the group of order 2, we have
=1
or
= .
However,
/ z(F4 ) (Theorem 2.9.1), which implies that = 1.
According to a general theory of compact Lie groups, it is known that the center
of the simply connected simple Lie group of type F4 is trivial. Hence F4 has to be
simply connected. Thus we have the following Theorem.
Theorem 2.10.2. F4 = { IsoR (J) | (XY ) = XY } is a simply connected
compact Lie group of type F4 .
Remark. If we know that the space CP2 is simply connected, the simply connectedness of the group F4 follows from F4 /Spin(9) CP2 of Theorem 2.8.3.
2.11. Involution and subgroup (Sp(1) Sp(3))/Z 2 of F4
We define an involutive R-linear transformation of J by
1 x3 x2
1 x3 x2
x3 2 x1 = x3 2 x1 .
x2 x1 3
x2 x1 3
This is the same as G2 F4 .
59
We shall study the following subgroup (F4 ) of F4 :

(F4 ) = { F4 | = }.
Any element
1
X = x3
x2
X J is expressed by

x3 x2
1 m3
2 x1 = m3 2
x1 3
m2 m1

0
m2
m1 + a3 e4
3
a2 e 4
a3 e 4
0
a1 e4
a2 e4
a1 e 4 ,
0
where xi = mi + ai e4 H He4 = C. We associate such element X J with the

element of J(3, H) H 3 ,
1 m3 m2
m3 2 m1 + (a1 , a2 , a3 ).
m2 m1 3
In J(3, H)H 3 , we define a multiplication , an inner product ( , ) and an R-linear

transformation respectively by

1
1
(M + a) (N + b) = M N (a b + b a) (aN + bM ),
2
2
(M + a, N + b) = (M, N ) + 2(a, b),
(M + a) = M a.
(In J(3, H) the multiplication M N and the inner product (M, N ) are analogously
1
defined as in J, and in H 3 the inner product (a, b) is defined naturally by (a b +
2
b a)). Since these operations correspond to their respective operations in J, hereafter,
we identify J(3, H) H 3 with J, that is,
J(3, H) H 3 = J.
The group F4,H is defined to be the automorphism group of the Jordan algebra
JH = J(3, H) (in which multiplications and are analogously defined as in J):
F4,H = { IsoR (JH ) | (M N ) = M N }
= { IsoR (JH ) | (M N ) = M N }.
Proposition 2.11.1.
F4,H
= Sp(3)/Z 2 , Z 2 = {E, E}.
Proof. We define a mapping : Sp(3) F4,H by

(A)M = AM A ,
M JH .
It is not difficult to see that is well-defined and is a homomorphism. We shall show

that is onto. For a given F4,H , we consider Ei , i = 1, 2, 3. Since Ei satisfies
Ei Ei = Ei and tr(Ei ) = 1 (that is, Ei HP2 = {M JH | M M =
60
M, tr(M ) = 1} (the quaternion

plane)), there exists Ai Sp(3) such that
projective
ai1
Ei = Ai Ei Ai . Let ai = ai2 be the i-th column vector of Ai , then we have
ai3
ai1 ai1 ai1 ai2 ai1 ai3

Ei = ai2 ai1 ai2 ai2 ai2 ai3 . Construct a matrix A = (a1 , a2 , a3 ), then we
ai3 ai1 ai3 ai2 ai3 ai3
have
Ei = AEi A , i = 1, 2, 3.
Since AA = A(E1 + E2 + E3 )A = E1 + E2 + E3 = E = E, we have A Sp(3).
If we let = (A)1 , then F4,H and satisfies
Ei = Ei ,
i = 1, 2, 3.
Analogously as in Theorem 2.7.1, induces orthogonal transformations 1 , 2 , 3

: H H satisfying
(1 m)(2 n) = 3 (mn),
m, n H.
(i)
Put p = 1 1 and q = 2 1, then we have |p| = |q| = 1. Letting m = 1 and n = 1 in

(i), we have p(2 n) = 3 n and (1 m)q = 3 m, and hence
2 m = p(1 m)q,
3 m = (1 m)q.
Substitute m = p(1 m) in (i), then we see that satisfies

(m)(n) = (mn),
m, n H,
that is, is an automorphism of H: Aut(H). Hence, is expressed by m = rmr,

using r Sp(1) (Proposition 108 of Yokota [58]). Consequently
1 m = prmr, 2 m = rmrq, 3 m = qrmr p, m H.
qr 0 0
Construct a matrix B = 0 pr 0 , then B Sp(3) and we have
0 0 r
M = BM B ,
M JH ,
that is, = (B). Hence

= (A) = (A)(B) = (AB),
AB Sp(3).
Therefore is onto. Ker = {E, E} can be easily obtained. Thus we have the
isomorphism Sp(3)/Z 2
= F4,H .
Theorem 2.11.2.
(F4 )
= (Sp(1) Sp(3))/Z 2 , Z 2 = {(1, E), (1, E)}.
61
Proof. We define a mapping : Sp(1) Sp(3) (F4 ) by

(p, A)(M + a) = AM A + paA ,
M + a JH H 3 = J.
We first show that (p, A) (F4 ) . For p Sp(1), A Sp(3), M, N JH and

a, b H 3 , we have
AM A AN A = A(M N )A ,
(paA ) (pbA ) = Aa ppbA = A(a b)A ,
(paA )(AN A ) = p(aN )A , etc.

From which, we see that (p, A) satisfies
(p, A)((M + a) (N + b)) = (p, A)(M + a) (p, A)(N + b),
hence (p, A) F4 . Clearly (p, A) = (p, A), so that (p, A) (F4 ) . Certainly
is a homomorphism. We shall show that is onto. Let (F4 ) . Since the
restriction = |JH of to JH = {X J | X = X} belongs to F4,H , there exists
A Sp(3) such that
M = AM A , M JH
(Proposition 2.11.1). Let = (1, A)1 , then |JH = 1. Therefore G2 .
Certainly, since Spin(7), by letting = (1 , , ), we see that (1 x)(y) =
(xy). Now let y = 1, then we see 1 = . Since (G2 ) and |H = 1, there
exists p Sp(1) such that
(m + ae4 ) = m + (pa)e4 ,
m + ae4 H He4 = C,
(Theorem 1.10.1). Hence we have

(M + a) = M + pa = (p, E)(M + a),
so that = (p, E). Hence we have = (1, A) = (1, A)(p, E) = (p, A).
Therefore is onto. Ker = {(1, E), (1, E)} = Z 2 can be easily obtained. Thus
we have the isomorphism (Sp(1) Sp(3))/Z 2
= (F4 ) .
Remark. Since (F4 ) is connected as fixed points subgroup of F4 by the involution
of the simply connected Lie group F4 , the fact that : Sp(1) Sp(3) F4 is onto
can be proved as follows. The elements
Gij , 0 i < j 3, 4 i < j 7,
ek (ej ), 0 j 3, k = 1, 2, 3
A
forms an R-basis of (f4 ) . So dim((f4 ) ) = 6 2 + 4 3 = 24 = 3 + 21 = dim(sp(1)

sp(3)). Hence is onto.
62
F4
2.12. Automorphism w of order 3 and subgroup (SU (3) SU (3))/Z 3 of

We define an R-linear transformation w of J by

1 x3 x2
1 x3
w x3 2 x1 = x3 2
x2 x1 3
x2 x1
This w is the same as w G2 F4 .
x2
x1 .
3
We shall study the following subgroup (F4 )w of F4 :

(F4 )w = { F4 | w = w}.
We associate an
1
x3
x2
element
x3
2
x1
of J with the element
x2
x1 ,
3
1
a3
a2
i R, xi = ai + mi C C 3 = C
a3
2
a1
a2

a1 + m1 , m2 , m3
3
of J(3, C) M (3, C). In M (3, C), we define the exterior product M N by
m2 n3 m3 n1 m1 n2
M (3, C),
+
+
+
M N =
n2 m3 n3 m1 n1 m2
where M = (m1 , m2 , m3 ), N = (n1 , n2 , n3 ) M (3, C). Then, for P, A M (3, C)

and M, N M (3, C), we have
P M P N = t Pe(M N ),
e
M A N A = (M N ) tA,
e are the adjoint matrices of P and A, respectively. Further, in M (3, C),

where Pe and A
we define a real valued symmetric product (M, N ) by
X
1
(M, N ) = tr(M N + N M ) =
(mij , nij ),
2
i,j

where M = mij , N = nij M (3, C).
In J(3, C) M (3, C), we define the multiplication , the inner product ( , ) and
the R-linear transformation w respectively by
1
1
(X + M ) (Y + N ) = (X Y (M N + N M )) (M Y + N X + M N ),
2
2
(X + M, Y + N ) = (X, Y ) + 2(M, N ),

1
3
e1 ,
w(X + M ) = X + 1 M,
1 = +
2
2
63
(in J(3, C), multiplications and are defined as in J). Since these operations
correspond to their respective operations in J, we identify J(3, C) M (3, C) with J,
that is,
J(3, C) M (3, C) = J.
The group F4,C is defined to be the automorphism group of the Jordan algebra
JC = J(3, C) :
F4,C = { IsoR (JC ) | (X Y ) = X Y }
= { IsoR (JC ) | (X Y ) = X Y }.
The group Z 2 = {1, } acts on the group SU (3) by

A = A,
A SU (3),
and the group SU (3) Z 2 be the semi-direct product of groups SU (3) and Z 2 under
this action.
Proposition
2.12.1. F4,C
= (SU (3)/Z 3 ) Z 2 , Z 3 = {E, 1 E, 1 2 E}, 1 =
1
3
+
e1 .
2
2
Proof. We define a mapping : SU (3) Z 2 F4,C by
(A, 1)X = AXA ,
(A, )X = AXA ,
X JC .
It is not difficult to see that is well-defined and that it is a homomorphism. The

proof that is onto is the same as that of Proposition 2.11.1 (use C, SU (3), instead
of H, Sp(3)), and we only need to modify a little the last part. Since Aut(C) can
be either = 1 or = (where x = x, x C), we have
1 x = px, 2 x = xq, 3 x = qxp
Now, choose r C such that r3
Then, B SU (3) and we have
X = BXB
or 1 x = px, 2 x = xq, 3 x = q x p.
qr 0
= qp and construct a matrix B = 0 pr
0 0
or X = BXB ,
0
0 .
r
X JC .
The remaining proof is again analogous to that of Proposition 2.11.1.

w
Theorem 2.12.2. (F4 )
= (SU (3) SU (3))/Z 3 , Z 3 = {(E, E), (1 E, 1 E),
3
1
e1 .
(1 2 E, 1 2 E)}, 1 = +
2
2
Proof. We define a mapping : SU (3) SU (3) (F4 )w by
(P, A)(X + M ) = AXA + P M A ,

64
X + M J(3, C) M (3, C) = J.
We first show that (P, A) (F4 )w . For P, A SU (3) and X + M, Y + N

J(3, C) M (3, C), we have
AXA AY A = A(X Y )A ,
(P M A ) (P N A ) = AM P P N A = A(M N )A,
(P M A )(AY A ) = P (M Y )A ,
e = P M N A , etc.
P M A P N A = t Pe (M N )t A
Using them, we see that (P, A) satisfies
(P, A)((X + M ) (Y + N )) = (P, A)(X + M ) (P, A)(Y + N ),

hence (P, A) F4 . Clearly w(P, A) = (P, A)w, so that (P, A) (F4 )w . Certainly is a homomorphism. We shall show that is onto. Let (F4 )w . Since the
restriction = |JC of to JC = {X J | wX = X} belongs to F4,C , there exists
A SU (3) such that
X = AXA
or X = AXA ,
X JC
(Proposition 2.12.1). In the former case, let = (E, A)1 , then |JC = 1, and so
G2 . Moreover (G2 )e1 = (G2 )w = SU (3) (Theorem 1.9.4), and hence, there
exists P SU (3) such that
(X + M ) = X + P M = (P, E)(X + M ),
X + M JC M (3, C) = J.
Therefore we have
= (E, A) = (E, A)(P, E) = (P, A).
In the latter case, consider the mapping 1 : J J given by 1 (X + M ) = X + M ,
X + M J and remember that 1 G2 F4 . Let = 1 (E, A)1 . Then F4
and |JC = 1, which shows that (G2 )e1 = (G2 )w (Theorem 1.9.4) (F4 )w . Since
, (E, A) (F4 )w , we have 1 (F4 )w , so that 1 (G2 )w which is a contradiction
(Theorem 1.9.4). Consequently the proof of is onto is completed. The fact that
Ker = {(E, E), (1 E, 1 E), (1 2 E, 1 2 E)} = Z 3 can be easily obtained. Thus we
have the isomorphism (SU (3) SU (3))/Z 3
= (F4 )w .
Remark 1. Since (F4 )w is connected as the fixed points subgroup of F4 by an
automorphism of order 3 of the simply connected group F4 (Rasevskii [32]), that
: SU (3) SU (3) (F4 )w is onto is proved as follows. The elements
G01 , G23 , G45 , G67 , G26 + G37 , G27 + G36 ,
G24 + G35 , G25 + G34 , G46 + G57 , G47 + G56 ,
e1 (1), A
e2 (1), A
e3 (1), A
e1 (e1 ), A
e2 (e1 ), A
e3 (e1 )
A
65
forms an R-basis of (f4 )w . So dim(f4 )w = 10 + 6 = 16 = 8 + 8 = dim(su(3) + su(3)).

Hence is onto.
Remark 2. The group F4 has a subgroup which is isomorphic to the group
(SU (3) SU (3))/Z 3 ) Z 2 , which is the semi-direct product of the groups (SU (3)
SU (3))/Z 3 and Z 2 (the action of Z 2 = {1, 1 } on the group (SU (3) SU (3)) is
1 (P, A) = (P , A)).
2.13. Complex exceptional Lie group F4 C
Definition. The group F4 C is defined to be the automorphism group of the
complex Jordan algebra JC :
F4 C = { IsoC (JC ) | (X Y ) = X Y }
= { IsoC (JC ) | det (X) = det X, (X, Y ) = (X, Y )}
= { IsoC (JC ) | det (X) = det X, E = E}
= { IsoC (JC ) | (X Y ) = X Y }.
We define a positive definite Hermitian inner product hX, Y i in JC by

hX, Y i = ( X, Y ).
For HomC (JC ), we denote the complex conjugate transpose of with respect to
hX, Y i by : h X, Y i = hX, Y i.
Lemma 2.13.1. (1) For F4 C , we have = 1 F4 C .
(2) For any F4 , its complexificated mapping C : JC JC belongs to F4 C :

C F4 C . Identifying with C , we regard F4 as a subgroup of F4 C : F4 F4 C .
For F4 C , we have F4 if and only if = , that is,
F4 = { F4 C | = }.
Proof. (1) h X, Y i = hX, Y i = ( X, Y ) = (1 X, Y ) = h 1 X, Y i for
all X, Y JC . Hence = 1 F4 C .
(2) Let F4 C satisfy = . Then, since X = X = X, we have

X J for X J. Hence induces an R-transformation of J and F4 , further
we have = ( )C .
Theorem 2.13.2. The polar decomposition of the Lie group F4 C is given by
F4 C F4 R52 .
In particular, F4 C is a simply connected complex Lie group of type F4 .
66
Proof. Evidently F4 C is an algebraic subgroup of IsoC (JC ) = GL(27, C). If

F4 C , then F4 C (Lemma 2.13.1.(1)). Hence from Chevalleys lemma, we
have
F4 C (F4 C U (JC )) Rd = F4 Rd ,
where U (JC ) = { IsoC (JC ) | hX, Y i = hX, Y i} and d = dim F4 C dim F4 =

2 52 52 = 52. Since F4 is simply connected (Theorem 2.10.2), F4 C is also simply
connected. The Lie algebra of the group F4 C is f4 C , so that F4 C is a complex simple
Lie group of type F4 .
2.14. Non-compact exceptional Lie groups F4(4) and F4(20) of type F4
Consider the following two R-vector spaces
J(3, C ) = {X M (3, C ) | X = X},
J(1, 2, C) = {X M (3, C) | I1 X I1 = X},

where I1 = diag(1, 1, 1) M (3, R). We define the Jordan multiplication X Y in
J(3, C ) and J(1, 2, C) respectively by
X Y =
1
(XY + Y X).
2
Then we have the isomorphisms

J(3, C )
= {X J(3, CC ) | X = X} = (J(3, CC )) ,
J(1, 2, C)
= {X J(3, CC ) | X = X} = (J(3, CC ))
as Jordan algebras, therefore we identify them, respectively. We denote by F4(4) and
F4(20) respectively the automorphism groups of the Jordan algebras J(3, C ) and J(1,
2, C):
F4(4) = { IsoR (J(3, C )) | (X Y ) = X Y },
F4(20) = { IsoR (J(1, 2, C)) | (X Y ) = X Y }.

They can also be defined by
F4(4) = (F4 C ) ,
F4(20) = (F4 C ) .
Theorem 2.14.1. The polar decompositions of the Lie groups F4(4) and F4(20)
are respectively given by
F4(4) (Sp(1) Sp(3))/Z 2 R28 ,
F4(20) Spin(9) R16 .
Proof. These are facts corresponding to Theorems 2.11.2 and 2.9.1.

Theorem 2.14.2. The centers z(F4(4) ) and z(F4(20) ) are trivial :
z(F4(4) ) = {1},
z(F4(20) ) = {1}.
67

Let JC be the complex exceptional Jordan algebra.
Definition. We define the groups E6 C and E6 respectively by
E6 C = { IsoC (JC ) | det (X) = det X}
= { IsoC (JC ) | (X, Y, Z) = (X, Y, Z)},
E6 = { IsoC (JC ) | det (X) = det X, hX, Y i = hX, Y i}
= { IsoC (JC ) | (X, Y, Z) = (X, Y, Z), hX, Y i = hX, Y i}
= { IsoC (JC ) | X Y = t 1 (X Y ), hX, Y i = hX, Y i}
= { IsoC (JC ) | X Y = (X Y ), hX, Y i = hX, Y i}.

If we define an involutive automorphism of the group E6 C by
() = t 1 ,
E6 C
(Lemma 2.2.3), then the definition of the group E6 can be also given by
E6 = { E6 C | () = } = (E6 C ) .
Theorem 3.1.1. E6 is a compact Lie group.
Proof. E6 is a compact Lie group as a closed subgroup of the unitary group
U (27) = U (JC ) = { IsoC (JC ) | hX, Y i = hX, Y i}.
Before investigating the Lie algebra e6 of the group E6 , we will study the Lie
algebra e6 C of the group E6 C .
Theorem 3.2.1. (1) The Lie algebra e6 C of the Lie group E6 C is given by
e6 C = { HomC (JC ) | (X, X, X) = 0}.
(2) Any element e6 C is uniquely expressed by
= + Te,
f 4 C , T J0 C .
In particular, the dimension of e6 C is
dimC (e6 C ) = 52 + 26 = 78.

68
Proof. (1) is proved as similar way to Lemma 2.3.1.

(2) For e6 C , by letting T = E, we obtain T JC , and tr(T ) = 0. Certainly
tr(T ) = (T, E, E) = (E, E, E) = 0. If we put = Te, then e6 C . Moreover
f4 C , because E = E TeE = T T = 0. Hence we have = + Te. To prove
the uniqueness of the expression, it is sufficient to show that
+ Te = 0, f4 C , T J0 C
implies
= 0, T = 0.
Certainly, let apply it on E, then we have T = 0, so that = 0. Finally, we have

dimC (e6 C ) = 52 + 26 = 78 from the expression above. Therefore the theorem is
proved.
Theorem 3.2.2. The Lie bracket [1 , 2 ] in e6 C is given by
g
[1 + Te1 , 2 + Te2 ] = ([1 , 2 ] + [Te1 , Te2 ]) + (g
1 T2 2 T1 ),
where i = i + Tei , i f4 C , Ti J0 C .
f for f C , T J0 C . Now,
Proof. It is sufficient to show that [, Te] = T
4
[, Te]X = (T X) T X = T X + T X T X
f X, X JC .
= T X = T
We shall investigate the Lie algebra e6 of the Lie group E6 .
Lemma 3.2.3. For = + Te e6 C , f4 C , T J0 C , we have

In particular, t e6 C .
() = t = t ( + Te) = Te.
Proof. (t X, Y ) = (X, Y ) = (X, Y + TeY ) = (X, Y ) (X, T Y )

= (X, Y ) (TeX, Y ) = (( Te)X, Y ) X, Y JC .
Therefore t = Te e6 C .
Theorem 3.2.4. (1) The Lie algebra e6 of the Lie group E6 is given by
e6 = { HomC (JC ) | (X, X, X) = 0, hX, Y i + hX, Y i = 0}.
(2) Any element e6 is uniquely expressed by
= + iTe,
f4 , T J0 .
Proof. (1) The proof is evident (cf. Lemma 2.3.1).

(2) For e6 C , the condition e6 is equivalent to () = . Now, if is
of the form = + f
T , f4 C , T J0 C (Theorem 3.2.1.(2)), then () = is
69
g
T = + f
T (Lemma 3.2.3), that is, = and T = T . Hence f4 ,
and T is of the form T = iT, T J0 .
Proposition 3.2.5. The complexification of the Lie algebra e6 is e6 C .
Proof. For e6 C , the conjugate transposed mapping of with respect to
the inner product hX, Y i of JC is = t e6 C , and for e6 C , belongs to e6
if and only if = . Now, any element e6 C can be uniquely expressed as
=

+
+i
,
2
2i
+
,
e6 .
2
2i
Hence e6 C is the complexification of e6 .

Theorem 3.3.1. The Lie algebra e6 C is simple and so e6 is also simple.
Proof. We use the decomposition of e6 C of Theorem 3.2.1.(2):
C
e6 C = f4 C e
J0 .
C
C
J0 . Now, let a be
J0 be projections of e6 C = f4 C e
Let p : e6 C f4 C and q : e6 C e
a non-zero ideal of e6 C . Then p(a) is an ideal of f4 C . Indeed, if p(a), then there
exists T J0 C such that + Te a. For any 1 f4 C , we have
hence [1 , ] a.
a [1 , + Te] = [1 , ] + g
1 T (Theorem 3.2.2),
C
We shall show that either f4 C a 6= {0} or e
J0 a 6= {0}. Assume that f4 C a = {0}
C
C
J a = {0}.
and e
J a = {0}. Then the mapping p|a : a f4 C is injective because e
0
Since p(a) is a non-zero ideal of f4 C and f4 C is simple, we have p(a) = f4 C . Hence

dimC a = dimC p(a) = dimC f4 C = 52. On the other hand, since f4 C a = {0},
C
C
J0 = dimC J0 C = 26. This
q|a : a e
J0 is also injective, we have dimC a dimC e
leads to a contradiction.
(1) Case f4 C a 6= {0}. From the simplicity of f4 C , we have f4 C a = f4 C , hence

a f4 C . On the other hand, we have
C
C
g
a [a, e6 C ] [f4 C , e
J0 ] = f4 C
J0 C (Lemma 3.2.2) = e
J0 (Proposition 2.4.6.(2)).
C
J0 = e 6 C .
Consequently a f4 C e
C
e C a a.
e (A J0 C ) be a non-zero element of J
(2) Case e
J0 a 6= {0}. Let A
0
e B]
e 6= 0 (Lemma 2.5.4), then 0 6= [A,
e B]
e f4 C a.
Choose B J0 C such that [A,
Hence this case is reduced to the case (1).
70
Therefore we have a = e6 C .
Proposition 3.3.2. (1) JC is a simple Jordan algebra.
(2) JC is an e6 C -irreducible C-module.
o
nX
i Ai | i e6 C , Ai JC = JC .
(3) e6 C JC =
i
Proof. (1) Let a be a non-zero ideal of JC and X = X(, x) a non-zero element

of a.
(i) Case 1 6= 0. From 1 E1 = (2X E1 X) E1 a, we have E1 a.
Next, from F2 (1) = 2E1 F2 (1) a and E1 + E3 = F2 (1) F2 (1) a, we have
E3 = (E1 + E3 ) E1 a. Similarly E2 a, so that E = E1 + E2 + E3 a. Now,
for any X JC , X = E X a, and so a = JC . In the case 2 6= 0 or 3 6= 0, the
statement is also valid.
(ii) Case 1 = 2 = 3 = 0, x1 6= 0. We have F1 (x1 ) = 4(X E2 ) E3 a. Choose
a CC such that (x1 , a) = 1, then F1 (x1 ) F1 (a) = (x1 , a)(E2 + E3 ) = E2 + E3 a.
Hence this case is reduced to the case (i) and so a = JC .
(2) Let W be a non-zero e6 C -invariant C-submodule of JC . For any A JC and
X W , we have

1
1
A X = A tr(A)E X + tr(A)X W.
3
3
Hence, W is an ideal of JC . From the simplicity of JC ((1) above), we have W = JC .
(3) e6 C JC is an e6 C -invariant C-submodule of JC . Hence from the irreducibility
of JC ((2) above), we have e6 C JC = JC .
3.4. Element A B of e6 C
Definition. For A, B JC , we define an element A B e6 C by

e B]
e + A B 1 (A, B)E
A B = [A,
3
(Proposition 2.4.1, Theorem 3.2.1).
Lemma 3.4.1. For A, B JC , we have

(A B)X =
1
1
(B, X)A + (A, B)X 2B (A X),
2
6
X JC .
Proof. Consider Hamilton-Cayley formula X (X X) = (det X)E, X JC ,

that is,
1
1
X (X X) tr(X)X X + (tr(X)2 (X, X))X = (X, X, X)E.
2
3
If we put A + B + X in place of X, then taking the coefficient of , we have
71
A (B X) + B (X A) + X (A B) tr(A)B X tr(B)A X
1
tr(X)A B + (tr(A)tr(B)X + tr(B)tr(X)A + tr(X)tr(A)B)
2
1
((A, B)X + (B, X)A + (X, A)B) = (A, B, X)E.
2
Therefore, using the above, we have
1
1
1
(B, X)A + (A, B)X 2B (A X) + (A, B)X
2
6
3
1
1
= (B, X)A + (A, B)X 2B (A X) + tr(B)A X + tr(A X)B
2
2
(tr(B)tr(A X) (B, A X))E
1
1
= (B, X)A + (A, B)X 2B (A X) + tr(A)B X + tr(X)B A
2
2
1
tr(A)tr(X)B + (A, X)B + tr(B)A X tr(B)tr(A)X
2
1
1
1
tr(B)tr(X)A + tr(B)tr(A)tr(X)E tr(B)(A, X)E
2
2
2
1
1
+ (tr(A)tr(X) (A, X))B tr(B)(tr(A)tr(X) (A, X))E + (A, B, X)E
2
2
= tr(A)B X + tr(B)X A + tr(X)A B
1
(tr(A)tr(B)X + tr(B)tr(X)A + tr(A)tr(X)B)
2
1
+ ((A, B)X + (B, X)A + (A, X)B) + (A, B, X)E 2B (A X)
2
= A (B X) + B (A X) + X (A B) 2B (A X)
= A (B X) B (A X) + (A B) X
e B]X
e + (A B) X.
= [A,
Lemma 3.4.2. (1) Ei Ej = 0, i 6= j.

1
(2) E1 E1 = (2E1 E2 E3 ) , more explicitly, we have
3
x2
1 x3 x2
41
x3
1
(E1 E1 ) x3 2 x1 = x3 22 2x1 .
6
x2 x1 3
x2 2x1 23
For HomC (JC ), we often denote t by :

( X, Y ) = (X, Y ),
X, Y JC .
Lemma 3.4.3. (1) For e6 C , we have

(X Y ) = X Y + X Y,
(2) For A, B JC , we have
(A B) = B A.
72
X, Y JC .
Proof. (1) it is evident from Lemma 2.3.1.

e B]
e + A B 1 (A, B)E
e B]
e A B 1 (A, B)E
(2) (A B) = [A,
= [A,
3
3

e A]
e A B 1 (A, B)E
(Lemma 3.2.3) = [B,
= B A.
3
Lemma 3.4.4. (1) For e6 C and A, B JC , we have
[, A B] = A B + A B.
X
(Ai Bi ), Ai , Bi JC .
(2) Any element e6 C is expressed by =
i
Proof. (1) [, A B]X = (A B)X (A B)X

1
1
= (B, X)A + (A, B)X 2B (A X) (A B)X (Lemma 3.4.1)
2
6
1
1
= (B, X)A + (A, B)X 2 B (A X) 2B (A X)
2
6
1
1
2B (A X) (B, X)A (A, B)X + 2B (A X) (Lemma 3.4.3)
2
6
1
1
= (B, X)A + (A, B)X 2B (A X)
2
6
1
1
+ ( B, X)A + (A, B)X 2 B (A X)
2
6
= (A B)X + (A B)X, X JC .
o
nX
(Ai Bi ) | Ai , Bi JC is an ideal of e6 C .
(2) We can see from (1) that a =
i
From the simplicity of e6 C (Theorem 3.3.1) we have a = e6 C .

Definition. We define a symmetric inner product (1 , 2 )6 in e6 C by

(1 , 2 )6 = (1 , 2 )4 + (T1 , T2 ),
where i = i + Tei , i f4 C , Ti J0 C .
Lemma 3.5.1. (1) The inner product (1 , 2 )6 of e6 C is e6 C -adjoint invariant :

([, 1 ], 2 )6 + (1 , [, 2 ])6 = 0,
, i e6 C .
(2) For e6 C , A, B JC , we have

(, A B)6 = (A, B).
Proof. (1) For = + Te, i = i + Tei , , i f4 C , T, Ti J0 C , we have
([, 1 ], 2 )6
= ([ + Te, 1 + Te1 ], 2 + Te2 )6
73
g1 g
e
= (([, 1 ] + [Te, Te1 ]) + (T
1 T ), 2 + T2 )6 (Theorem 3.2.2)
= ([, 1 ], 2 )4 + ([Te, Te1 ], 2 )4 + (T1 1 T, T2 )
= (1 , [, 2 ])4 + (2 T, T1 ) + (T1 , T2 ) (1 T, T2 ) (Lemma 2.5.2)
= (1 , [, 2 ])4 (1 T, T2 ) (T1 , T2 ) + (T1 , 2 T )

g2 g
= (1 + Te1 , [, 2 ] + [Te, Te2 ]) + (T
2 T ))6
= (1 + Te1 , [ + Te, 2 + Te2 ])6
= (1 , [, 2 ])6 .
(2) For = + Te, f4 C , T J0 C , we have

e B]
e + A B 1 (A, B)E
(, A B)6 = + Te, [A,
3
6

1
e
e
= (, [A, B])4 + T, A B (A, B)E
3
= (A, B) + (TeA, B) = (( + Te)A, B) = (A, B).
Lemma 3.5.2. In e6 C , we have
[(Ei Ei+1 ) , D] = 0, D d4 C ,
ei (a)] = 1 Fei (a),
[(Ei Ei+1 ) , A
2
1
e
[(Ei Ei+1 ) , Ai+1 (a)] = Fei+1 (a),
2
ei+2 (a)] = Fei+2 (a),
[(Ei Ei+1 ) , A
[(Ei Ei+1 ) , (Ej Ej+1 ) ] = 0,

1e
[(Ei Ei+1 ) , Fei (a)] = A
i (a),
2
1e
[(Ei Ei+1 ) , Fei+1 (a)] = A
i+1 (a),
2
ei+2 (a).
[(Ei Ei+1 ) , Fei+2 (a)] = A
Theorem 3.5.3. The Killing form B6 of the Lie algebra e6 C is given by

B6 (1 , 2 ) = 12(1 , 2 )6
= 12(1 , 2 )4 + 12(T1 , T2 )
4
= B4 (1 , 2 ) + 12(T1 , T2 )
3
= 4tr(1 2 ),
where i = i + Tei , i f4 C , Ti J0 C and B4 is the Killing form of f4 C .
Proof. Since e6 C is simple (Theorem 3.3.1), there exist k, k C such that

B6 (1 , 2 ) = k(1 , 2 )6 = k tr(1 2 ).
To determine these k, k , let = 1 = 2 = (E1 E2 ) . Then we have

(, )6 = ((E1 E2 ) , (E1 E2 ) )6 = (E1 E2 , E1 E2 ) = 2.
On the other hand, (ad)2 is calculated as follows.
i 1
h
1
e1 (ei ),
[, [, Ae1 (ei )] ] = , Fe1 (ei ) = A
2
4
74
h
i
e2 (ei )] ] = , 1 Fe2 (ei ) = 1 A
e2 (ei ),
[, [, A
2
4
e3 (ei )] ] = [, Fe3 (ei )] = A
e3 (ei ),
[, [, A
i
h
1e
1e
[, [, Fe1 (ei )] ] = , A
1 (ei ) = F1 (ei ),
2
4
h
i 1
1e
e
[, [, F2 (ei )] ] = , A2 (ei ) = Fe2 (ei ),
2
4
e3 (ei )] = Fe3 (ei ),
[, [, Fe3 (ei )] ] = [, A
the others = 0.
Hence
B6 (, ) = tr((ad)2 ) =
1

4 + 1 2 8 = 24.
4
Therefore k = 12. Next, we will calculate tr() as follows.
1
1
Fe1 (ei ) = Fe1 (ei ) = Fe1 (ei ),
2
4
1 e
1e
e
E2 = E2 = E2 , F2 (ei ) = F2 (ei ) = F2 (ei ),
2
4
E3 = 0 = 0,
Fe3 (ei ) = 0 = 0.
E1 = E1 = E1 ,
Hence
tr() = 1 2 +
1
8 2 = 6.
4
Therefore k = 4.
Lemma 3.5.4. The followings hold in e6 C .
(1) A (A A) = 0,
A JC ,
A (B C) + B (C A) + C (A B) = 0,
A, B, C JC .
(2) For A JC , A 6= 0, there exists B JC such that A B 6= 0.
Proof. (1) (, (A A) A)6 = ((A A), A) (Lemma 3.5.1.(2))
= 2( A A, A) (Lemma 3.4.3.(1)) = 2( A, A A) = 2(A, (A A))
= 2(, (A A) A)6 .
Hence (, (A A) A)6 = 0 for all e6 C , so that (A A) A = 0, that is,
A (A A) = 0 (Lemma 3.4.3.(2)). If we put A + B + C in the place of A, then
the result follows from the coefficient of .
(2) Assume that A B = 0, that is B A = 0 (Lemma 3.4.3.(2)) for all B JC .
Then for any e6 C , 0 = (, B A)6 = (B, A) (Lemma 3.5.1.(2)). Since e6 C JC =
JC (Proposition 3.3.2.(3)), we have (JC , A) = 0, so that A = 0.
3.6. Roots of e6 C
Let
Mr = {X M (3, C) | all diagonal elements of X are real},
75
and let (Mr )C be its complexification. In (Mr )C , we define a multiplication X Y

by
1
X Y = (XY + Y X ),
2
where X = t X. For = (1 , 2 , 3 ) d4 C satisfying the principle of triality (1 x)y +
x(2 y) = 3 (xy), x, y CC , we define a C-linear mapping : (Mr )C (Mr )C by
0
3 x12 2 x13
1 x12 x13
x21 2 x23 = 3 x21
0
1 x23 .
x31 x32 3
0
2 x31 1 x32
Observe that this mapping is an extension of : JC JC .
Lemma 3.6.1. For d4 C , we have
(X Y ) = X Y + X Y,
X, Y (Mr )C .
Proof. We use the following notations:

23 = 1 ,
32 = 1 ,
31 = 2 ,
13 = 2 ,
12 = 3 ,
21 = 3 .
The (i, i)-element of X Y + Y Y (note that this is contained in C) is equal to

P3
P3
R( k=1 ((ik xik )yki + y ki (ik xik )) + k=1 (xik (ki yki ) + (ki yki )xik ))
P
= 2R( k ((ik xik )yki + xik (ki yki )))
P
= 2 k ((ik xik , yki ) + (xik , ki yki ))
P
= 2 k ((xik , ik y ki ) + (xik , ik yki )) = 0.
The (i, j)-element of X Y + Y Y (i 6= j) is equal to
P3
P3
k=1 ((ik xik )ykj + y ki (jk xjk )) +
k=1 (xik (kj ykj ) + (ki yki )xjk )
P
P
= k ((ik xik )ykj + yik (kj xkj )) + k ((jk xjk )yki + xjk (ki yki ))
(If i, j, k are all distinct, from the principle of triality, we obtain
(ik xik )ykj + xik (kj ykj ) = ij (xik ykj ).
Even if i = k 6= j or i 6= k = j, considering the fact that ll = 0, we see that the
formula above is also valid, since xll C)
P
P
= k ij (xik ykj ) + k ji (xjk yki )
P
P
= k ij (xik ykj ) + k ij (xjk yki ) = (i, j)-element of (X Y ).
For T M (3, CC ), we define a C-linear mapping Te : JC JC by
1
TeX = (T X + XT ),
2
76
where T = t T .
Proposition 3.6.2. For T M (3, CC ), tr(T ) = 0, we have Te e6 C .
T + T
T T
, T2 =
. Then Te1 e6 C
2
2
(Proposition 2.3.6) e6 C . Therefore Te = Te1 + Te2
Proof. We decompose T = T1 + T2 , T1 =
(Lemma 2.4.1.(1)) and Te2 f4 C

e6 C .
Lemma 3.6.3. (1) For d4 C and R (Mr )C , tr(R) = 0, we have

f
e = R.
[, R]
(2) For H M (3, C), tr(H) = 0 and T M (3, CC ), tr(T ) = 0, we have

e Te] = 1 [H, T ] .
[H,
2
f
e
Proof. (1) (R)X
= R X = (R X) R X (Lemma 3.6.1) = (RX)
C
f
e
e
e
R(X) = [, R]X, X J . Hence, R = [, R].
e Te]X = H
e TeX TeHX
e
(2) [H,
1 e
(H(T X + XT ) Te(HX + XH ))
2
1
= (HT X + XT H + HXT + T XH T HX XH T T XH HXT )
4
(since H M (3, C), products of matrices above are associative)
1
1
= ([H, T ]X + X[H, T ]) = [H, T ] X, X JC .
4
2
=
Theorem 3.6.4. The rank of the Lie algebra e6 C is 6. The roots of e6 C relative
to some Cartan subalgebra are given by
(k l ), (k + l ), 0 k < l 3,
1
k (2 3 ),
0 k 3,
2
1
1
(0 1 + 2 3 ) (3 1 ),
2
2
1
1
( 0 + 1 + 2 3 ) (3 1 ),
2
2
1
1
(0 + 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 3 ) (1 2 ),
2
2
1
1
(0 + 1 + 2 3 ) (1 2 ),
2
2
1
1
( 0 + 1 + 2 + 3 ) (1 2 ),
2
2
1
1
(0 1 + 2 + 3 ) (1 2 ),
2
2
77
with 1 + 2 + 3 = 0.
Proof. We use the decomposition of Theorem 3.3.1.(2):
C
e6 C = f4 C e
J0 .
Let
h=
e e6 C
h = h + H

3
3

X
X
h =
k Hk = k iGk4+k , k C )

k=0
k=0

,
3

X

E
,
C,
=
0
H
=

j j
j
1
2
3

j=1
then h is an abelian subalgebra of e6 C (it will be a Cartan subalgebra of e6 C ). That

h is abelian is clear from
[h , h ] = 0,
e ] = hg
e
[h , H
H (Lemma 3.6.3.(1)) = 0 = 0,
e H
e ] = 1 [H, H ] (Lemma 3.6.3.(2)) = 0.
[H,
2
I The roots k l of d4 C ( f4 C e6 C ) are also roots of e6 C . Indeed, let

be a root of d4 C and S 4 C e6 C be an associated root vector. Then
e S] = [h , S] [S, H]
e
[h, S] = [h + H,
g (Lemma 3.6.3.(1))
= (h )S SH
= (h )S = (k l )S.
Hence k l are roots of e6 C .

II
We denote
aEkl
0
example, F23 (a) = 0
0
0
0
0
M
(3, CC ) by Fkl (a): Fkl (a) = aEkl , a CC , k 6= l. For
0
a . Then, we have
0
e Fe23 (a)]
[h, Fe23 (a)] = [h , Fe23 (a)] + [H,
1
= (h F23 (a)) + [H, F23 (a)] (Lemma 3.6.3)
2
1
= Fe23 (h a) + [H, F23 (a)]
2

1
= k + (2 3 ) Fe23 (a),
2
where a = ek + ie4+k (since h a = h (ek + ie4+k ) = k (ek + ie4+k ) = k a, and

1
[H, F23 (a)] = HF23 (a) F23 (a)H = (2 3 )F23 (a)). Hence k + (2 3 ) is a
2
78
1
root of e6 C and Fe23 (ek + ie4+k ) is its root vector. Similarly k + (2 3 ) is a
2
root of e6 C and Fe23 (ek ie4+k ) is its root vector. Again, from
[h, Fe32 (a)] =

1
k + (3 2 ) Fe32 (a),
2
a = ek + ie4+k ,
1
1
we see that k + (3 2 ) is a root of e6 C . Similarly k + (3 2 ) is a root of
2
2
e6 C and Fe32 (ek ie4+k ) is its root vector. Next, using the relation
1 e e
1
[h, Fe31 (a)] = (h F31 (a)) + [H,
F31 (a)] = Fe31 ((h )a) + (3 1 )Fe31 (a),
2
2
1 e e
1
e
e
[h, F12 (a)] = (h F12 (a)) + [H, F12 (a)] = F12 ((h )a) + (1 2 )Fe12 (a),
2
2
where
1
1
(0 1 + 2 3 )H0 + (0 + 1 + 2 3 )H1
2
2
1
1
+ (0 + 1 + 2 + 3 )H2 + (0 1 + 2 + 3 )H3 ,
2
2
1
1
h = (0 + 1 2 + 3 )H0 + (0 + 1 + 2 3 )H1
2
2
1
1
+ (0 + 1 + 2 + 3 )H2 + (0 1 + 2 + 3 )H3
2
2
h =
etc., we can obtain the remainders of roots.

1 = 0 1 , 2 = 1 2 , 3 = 2 3 ,
1
4 = 3 + (2 3 ),
2
1
1
5 = (0 1 2 + 3 ) + (3 1 ),
2
2
1
1
6 = (0 + 1 + 2 + 3 ) + (1 2 )
2
2
is a fundamental root system of the Lie algebra e6 C and
= 1 + 22 + 33 + 24 + 25 + 6
is the highest root. The Dynkin diagram and the extended Dynkin diagram of e6 C are
3
3
2 4
79
Proof. In the following, the notation n1 n2 n6 denotes the root n1 1 + n2 2 +

+ n6 6 . Now, all positive roots of e6 C are represented by
0 1
0 2
0 3
1 2
1 3
2 3
=
=
=
=
=
=
1
1
1
0
0
0
0
1
1
1
1
0
0
0
1
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 + 1
0 + 2
0 + 3
1 + 2
1 + 3
2 + 3
1
0 + (2 3 ) = 1 1 1
2
1
1 + (2 3 ) = 0 1 1
2
1
2 + (2 3 ) = 0 0 1
2
1
3 + (2 3 ) = 0 0 0
2
1
0 (2 3 ) = 1 1 1
2
1
1 (2 3 ) = 0 1 1
2
1
2 (2 3 ) = 0 0 1
2
1
3 (2 3 ) = 0 0 0
2
1
1
(0 1 + 2 3 ) + (3 1 ) =
2
2
1
1
( 0 + 1 + 2 3 ) + (3 1 ) =
2
2
1
1
(0 + 1 + 2 + 3 ) + (3 1 ) =
2
2
1
1
( 0 1 + 2 + 3 ) + (3 1 ) =
2
2
1
1
( 0 + 1 2 + 3 ) + (3 1 ) =
2
2
1
1
(0 1 2 + 3 ) + (3 1 ) =
2
2
1
1
( 0 1 2 3 ) + (3 1 ) =
2
2
1
1
(0 + 1 2 3 ) + (3 1 ) =
2
2
1
1
( 0 1 + 2 3 ) (1 2 ) =
2
2
1
1
( 0 1 2 + 3 ) (1 2 ) =
2
2
1
1
( 0 + 1 + 2 + 3 ) + (1 2 ) =
2
2
80
=
=
=
=
=
=
1
1
1
0
0
0
2
1
1
1
1
0
0 0
0 0
0 0
0 0
1 1
1 1
1 1
1 1
2
2
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0 0
1 0
1 0
1 2
3 1
2 1
0 1
2 1
2 1
1 1
2 1
2 1
1 2
2 1
2 1
0 0
0 0
1 0
1 1
1 0
1 0
0 1
1 0
1 0
1 1
2 1
1 0
1 1
1 1
1 0
0 0
0 0
0 1
1
1
1
1
1
1
1
1
( 0 + 1 2 3 ) (1 2 ) = 1 2
2
2
1
1
(0 + 1 2 + 3 ) (1 2 ) = 0 1
2
2
1
1
(0 + 1 + 2 3 ) (1 2 ) = 0 1
2
2
1
1
( 0 + 1 + 2 + 3 ) (1 2 ) = 1 2
2
2
1
1
(0 1 + 2 + 3 ) (1 2 ) = 0 0
2
2
Hence = {1 , 2 , , 6 } is a fundamental root system
of h is
hR =
3
nX
k Hk +
3
X
j=1
k=0
j Ej
1 1
1 1
1 1
2 2
1 1
0.
of e6 C . The real part hR
o
| k , j R, 1 + 2 + 3 = 0 .
The Killing form B6 of e6 C on hR is given by

3
3

X
X
j j
k k +
B6 (h, h ) = 12 2
j=1
k=0
for h =
3
X
k Hk +
3
X
j Ej
j=1
k=0
from Theorem 3.5.3, we have

B6 (h, h ) =
3
X
, h =
k Hk +
k=0
3
X
j Ej
j=1
hR . Indeed,
3
3
3
3

X

X
X
4 X
B4
k Hk ,
k Hk + 12
k Ek ,
j Ej
3
j=1
j=1
k=0
4
18
3
3
X
k=0
k k + 12
k=0
3
X
j j (Theorem 2.6.2)
j=1
3
3

X
X
j j .
= 12 2
k k +
k=0
j=1

hR ) are determined as follows.
1
1
1
(H0 H1 ), H2 =
(H1 H2 ), H3 =
(H2 H3 ),
24
24
24
1
(H3 + (E2 E3 ) ),
H4 =
24
1
((H0 H1 H2 + H3 ) + 2(E3 E1 ) ),
H5 =
48
1
H6 =
((H0 + H1 + H2 + H3 ) + 2(E1 E2 ) ).
48
Therefore, we have
H1 =
(1 , 1 ) = B6 (H1 , H1 ) = 24
81
1
1 1
2=
,
24 24
12
and the other inner products are similarly calculated. Hence, the inner product
induced by the Killing form B6 between 1 , 2 , , 6 and are given by
(i , i ) =
1
,
12
i = 1, 2, 3, 4, 5, 6,
(1 , 2 ) = (2 , 3 ) = (3 , 4 ) = (3 , 5 ) = (5 , 6 ) =
otherwise,
1
1
, (, 4 ) = , (, i ) = 0,
(, ) =
12
24
1
,
24
(i , j ) = 0,
i = 1, 2, 3, 5, 6,
e6 C .
According to Borel-Siebenthal theory, the Lie algebra e6 has three subalgebras as
(1) The first one is a subalgebra of type T D5 which is obtained as the fixed
points of an involution of e6 .
(2) The second one is a subalgebra of type C1 A5 which is obtained as the fixed
(3) The third one is a subalgebra of type A2 A2 A2 which is obtained as the
fixed points of an automorphism w of order 3 of e6 .
The Lie algebra e6 has furthermore two outer involutions , . The subalgebra
obtained as the fixed points of is type F4 and the subalgebra obtained as the fixed
points of is type C4 .
These subalgebras will be realized as subgroups of the group E6 in Theorems
3.10.7, 3.11.4, 3.13.5, 3.7.1 and 3.12.2, respectively.
3.7. Involution and subgroup F4 of E6
We shall study the following subgroup (E6 ) of E6 :
(E6 ) = { E6 | = }
= { E6 | () = } = (E6 ) .
If E6 satisfies = , then (X, Y ) = h X, Y i = h X, Y i = h X, Y i =

(X, Y ) and vice versa. Further, for E6 , the conditions (X, Y ) = (X, Y ) and
E = E are equivalent (Lemma 2.2.4). Hence (E6 ) can be also defined by
(E6 ) = { E6 | (X, Y ) = (X, Y ), X, Y JC }
= { E6 | E = E} = (E6 )E .
Theorem 3.7.1.
(E6 ) = (E6 )E
= F4 .

82
F4 . Let (E6 ) . Then, for X J we have

Proof. We shall show (E6 ) =
X = X = X, so that X J. Hence induces an R-linear transformation of
J. Therefore, the restriction |J of to J belongs to the group
F4 = { IsoR (J) | det(X) = detX, (X, Y ) = (X, Y )}
= { IsoR (J) | (X Y ) = X Y }.
Conversely, for F4 , its complexification C : JC JC , C (X1 + iX2 ) = X1 +

iX2 belongs to (E6 ) . Therefore the correspondence F4 C (E6 ) gives an
isomorphism between F4 and (E6 ) .
We denote by (E6 )0 the connected component of E6 containing the identity 1.
Lemma 3.8.1. (1) For t R, if we define a mapping 12 (t) : JC JC by
it
1 x3 x2
e 1
x3
eit/2 x2
12 (t) x3 2 x1 = x3
eit 2
eit/2 x1 ,
it/2
it/2
x1
3
x2 x1 3
e x2 e
then 12 (t) (E6 )0 . Similarly we can define 13 (t), 23 (t) (E6 )0 .
(2) For a C, if we define a mapping 1 (a) : JC JC by 1 (a)X(, x) = Y (, y),

where
1 = 1
2 + 3
(a, x1 )
2 3
+
cos |a| + i
sin |a|
2 =
2
2
|a|
3 = 2 3 + 2 + 3 cos |a| + i (a, x1 ) sin |a|,

2
2
|a|
(2 + 3 )a
|a|
2(a, x1 )a
y = x1 + i
sin2
sin |a|
1
2|a|
|a|2
2
x3 a
|a|
|a|
+i
sin
y2 = x2 cos
2
|a|
2
ax
|a|
|a|
2
y3 = x3 cos
+i
sin
2
|a|
2

sin |a|
means 1 , then 1 (a) (E6 )0 .
if a = 0, then
|a|
Proof. (1) For E1 E2 J0 , we have i(E1 E2 ) e6 (Theorem 3.2.4.(2)) and

12 (t) = exp it(E1 E2 ) . Hence 12 (t) (E6 )0 .
(2) For F1 (a) J0 , we have iFe1 (a) e6 (Theorem 3.2.4.(2)) and 1 (a) =
exp iFe1 (a). Hence 1 (a) (E6 )0 .
Proposition 3.8.2. Any element X JC can be transformed to a diagonal form

by some element (E6 )0 :
83
1
X = 0
0
0
2
0
0
0 ,
3
i C.
Moreover, we can choose (E6 )0 so that two of 1 , 2 , 3 are non-negative real

numbers.
p
Proof. (Hereafter, we use the notation || instead of ( ) for RC = C).
For a given X JC , consider a space X = {X | (E6 )0 }. Since (E6 )0 is compact
(Theorem 3.1.1), X is also compact. Let |1 |2 + |2 |2 + |3 |2 be the maximal value of
all |1 |2 + |2 |2 + |3 |2 for Y = Y (, y) X and let X0 = X(, x) be an element of
X which attains its maximal value. Then X0 is of diagonal form. Certainly, suppose
that X0 is not of diagonal form, for example, the 2 3 entry x1 of X0 is non-zero:
0 6= x1 = p + iq,
p, q C.
It is sufficient to prove in the case that 2 , 3 are real numbers (otherwise we can
apply some 12 (t1 ) and 13 (t2 ) of Lemma 3.8.1.(1)).
q
t, t > 0 and construct 1 (a(t)) (E6 )0 of
(1) Case q 6= 0. Let a(t) =
|q|
q
(a(t), x1 )
Lemma 3.8.1.(2). Since |a(t)| = t and i
= i |q|, where =
,p ,
|a(t)|
|q|
for Y ((t), y(t)) = 1 (a(t))X0 X, we have
|1 (t)|2 + |2 (t)|2 + |3 (t)|2
2

2 + 3

2
3
+
cos t |q| sin t + i sin t
= |1 |2 +
2
2
2

2 + 3

2
3
+
cos t |q| sin t + i sin t
+
2
2
+
2
2
2
3
2
3
+2
cos t |q| sin t + 2 2 sin2 t
= |1 |2 + 2
2
2
2

+ 2
2
3
2
3
2
= |1 | + 2
+2
+ |q|2 sin2 (t + t0 ) + 2 2 sin2 t
2
2
|1 |2 + |2 |2 + |3 |2 + 2|q|2 + 2 2 cos2 t0 (for some t0 R)
which is the maximal value and attains at some t > 0. This contradicts the maximum
of |1 |2 + |2 |2 + |3 |2 . Hence q = 0.
p
(2) Case p 6= 0. Let a(t) =
t, t > 0 and construct 1 (a(t)) F4 (E6 )0 of
|p|
Lemma 2.8.1. For Y ((t), y(t)) = 1 (a(t))X0 X, the maximal value of |1 (t)|2 +
|2 (t)|2 + |3 (t)|2 is |1 |2 + |2 |2 + |3 |2 + 2|p|2 (the calculation is the same as Proposition 2.8.2) which contradicts the maximum of |1 |2 + |2 |2 + |3 |2 . Hence p = 0.
Consequently, we have x1 = 0. x2 = x3 = 0 can be similarly proved constructing
2 (a), 3 (a) (E6 )0 analogous to 1 (a) of Lemma 3.8.1.(2). Hence X0 is of diagonal
form.
84
The space EIV , called the symmetric space of type EIV , is defined by
EIV = {X JC | detX = 1, hX, Xi = 3}.
Theorem 3.8.3.
In particular, E6 is connected.
E6 /F4 EIV.
Proof. For E6 and X EIV , we have X EIV . Hence E6 acts on EIV .

We shall prove that the group (E6 )0 acts transitively on EIV . To prove this, it is
sufficient to show that any element X EIV can be transformed to E EIV by
some (E6 )0 . Now, X EIV JC can be transformed to a diagonal form by
(E6 )0 :
1 0 0
X = 0 2 0 , 1 C, 2 0, 3 0
0 0 3
(Proposition 3.8.2). From the condition X EIV , we have
1 2 3 = det (X) = det X = 1, (hence i > 0, i = 1, 2, 3),

1 2 + 2 2 + 3 2 = hX, Y i = hX, Y i = 3.
This implies that 1 = 2 = 3 = 1. Certainly, from 0 (2 3 )2 = 2 2 +3 2 22 3 =
2
1 3 31 + 2
(1 1)2 (1 + 2)
3 1 2
=
=
0, we have 1 = 1. Similarly
1
1
1
2 = 3 = 1 are obtained. Hence X = E, which shows the transitivity of (E6 )0 .
Since we have EIV = (E6 )0 E, EIV is connected. Now, the group E6 acts transitively
on EIV and the isotropy subgroup of E6 at E EIV is F4 (Theorem 3.7.1). Thus
we have the homeomorphism E6 /F4 EIV . Finally, the connectedness of E6 follows
from the connectedness of EIV and F4 .
Theorem 3.9.1. The center z(E6 ) of the group E6 is isomorphic to the cyclic
group of order 3:
1
3
2
z(E6 ) = {1, 1, 1}, = +
i C.
2
2
Proof. Let z(E6 ). From the commutativity with F4 E6 , we have
E = E = E. Let denote E = Y = Y (, y) JC , then we have
Y = Y,
for all F4 .
We choose F4 such that

X = T XT 1,
85
X JC ,
1 0
where T = 0 1
0 0
have y1 = y2 = y3 = 0
0
1 0 0
0
0 , 0 1 0 and 1
1
0 0 1
0
and 1 = 2 = 3 (= ), that is,
E = Y = E,
0 1
0 0 SO(3). Then we
1 0
C,
and 3 = det(E) = detE = 1. Since it is easy to verify that 1 z(E6 ), we

have 1 z(E6 ) and 1 E = E, and so 1 z(F4 ) (Theorem 3.7.1). Since
z(F4 ) = {1} (Theorem 2.10.1), we have 1 = 1, that is, = 1. This completes
the proof.
According to the general theory of compact Lie groups, it is known that the center
of the simply connected compact simple Lie group of type E6 is the cyclic group of
order 3. Hence the group E6 has to be simply connected. Thus we have the following
theorem.
Theorem 3.9.2. E6 = { IsoC (JC ) | det (X) = det X, hX, Y i = hX, Y i} is
a simply connected compact Lie group of type E6 .
3.10. Involution and subgroup (U (1) Spin(10))/Z 4 of E6
Let the C-linear mapping : JC JC be the complexification of F4 of
Section 2.9. Then E6 and 2 = 1.
(E6 ) = { E6 | = }.
To this end, we consider the C-subspaces (JC ) and (JC ) of JC , which are the
eigenspaces of :
(JC ) = {X JC | X = X}
= {X JC | 4E1 (E1 X) = X} E1 C ,
(JC ) = {X JC | X = X}
= {X JC | E1 X = 0, hE1 , Xi = 0},
where E1 C = {E1 | C}. Then JC = (JC ) (JC ) (which is the complexification

of J = J J in Section 2.9), and (JC ) , (JC ) are invariant under the action of
(E6 ) .
Lemma 3.10.1. For (E6 ) , there exists C such that
E1 = E1 ,
( ) = 1.
Proof. By the analogous proof to that of Theorem 2.9.1, we see that

E2 , E3 J(2, CC ).
86
Indeed, we have
E2 = (F2 (1) F2 (1)) = F2 (1) F2 (1)
= (F2 (x2 ) + F3 (x3 )) (F2 (x2 ) + F3 (x3 )) (for some x2 , x3 CC )

= (x2 , x2 )E2 + (x3 , x3 )E3 F1 (x2 x3 ) J(2, CC ).
Next, we shall show

E1 6 J(2, CC ).
Suppose that E1 J(2, CC ). Then E = E1 + E2 + E3 J(2, CC ), so we can
put E = 2 E2 + 3 E3 + F1 (x1 ), 2 , 3 C, x1 CC . Hence
2 E2 + 3 E3 + F1 (x1 ) = E = (E E) = E E
= (2 E2 + 3 E3 + F1 (x1 )) (2 E2 + 3 E3 + F1 (x1 ))
= ( 2 3 ( x1 , x1 ))E1 .
This implies that 2 = 3 = x1 = 0. Hence E = 0, which is a contradiction.
Therefore E1 is of the form
E1 = E1 + 2 E2 + 3 E3 + F1 (x1 ),
6= 0.
From E1 E1 = (E1 E1 ) = 0, we have

0 = (E1 + 2 E2 + 3 E3 + F1 (x1 )) (E1 + 2 E2 + 3 E3 + F1 (x1 ))
= (2 3 (x1 , x1 ))E1 + 3 E2 + 2 E3 F1 (x1 ).
This implies that 2 = 3 = x1 = 0 and so
E1 = E1 ,
6= 0.
Finally, from
1 = hE1 , E1 i = hE1 , E1 i = hE1 , E1 i = ( )hE1 , E1 i = ( ),
we have ( ) = 1.
In order to investigate the group (E6 ) , we consider the following subgroup (E6 )E1
of E6 :
(E6 )E1 = { E6 | E1 = E1 }.
Lemma 3.10.2. (E6 )E1 is a subgroup of (E6 ) : (E6 )E1 (E6 ) .
Proof. Since (JC ) = {X JC | 4E1 (E1 X) = X} E1 C and (JC ) =
{X JC | E1 X = 0, hE1 , Xi = 0}, these spaces are seen to be invariant under the
action of (E6 )E1 . Hence for (E6 )E1 , we have = (the proof is the same as
that of Theorem 2.9.1). Thus we have (E6 ) .
87
We define a 10 dimensional R-vector space V 10
n 0
V 10 = {X JC | 2E1 X = X} = 0
0
Proposition 3.10.3.
by
0
0

o

x C, x C .

(E6 )E1 /Spin(9) S 9 .
In particular, the group (E6 )E1 is connected.
Proof. S 9 = {X V 10 | hX, Xi = 2} is a 9 dimensional sphere. For (E6 )E1

and X S 9 , we have X S 9 . Indeed,
2E1 X = 2E1 X = 2 (E1 X) = ( X) = (X),
hX, Xi = hX, Xi = 2.
Hence the group (E6 )E1 acts on S 9 . We shall prove that the action is transitive. To
prove this, it is sufficient to show that any element X S 9 can be transformed to
i(E2 + E3 ) S 9 by some (E6 )E1 . Now, for a given X S 9 , we can choose 23 (t0 )
of Lemma 3.8.1.(1) such that
23 (t0 )X S 8 = {X V 9 | hX, Xi = 2}
where V 9 = {X V 10 | X = X}. (Note that 23 (t0 ) (E6 )E1 because 23 (t0 )E1 =
E1 ). Since the group Spin(9) = (F4 )E1 (E6 )E1 acts transitively on S 8 (Proposition
2.7.3), there exists Spin(9) such that
23 (t0 )X = E2 E3 S 8 .
By applying 23 (/2) (E6 )E1 of Lemma 3.8.1.(1), we get
23 (/2)23 (t0 )X = i(E2 + E3 ).
This shows the transitivity. The isotropy subgroup of (E6 )E1 at i(E2 + E3 ) S 9 is
Spin(9). Indeed, if (E6 )E1 satisfies (i(E2 +E3 )) = i(E2 +E3 ), then E = E1 +
(E2 + E3 ) = E1 + (E2 + E3 ) = E, so that F4 and hence (F4 )E1 = Spin(9).
Conversely Spin(9) satisfies (i(E2 + E3 )) = i(E2 + E3 ). Thus we have the
homeomorphism (E6 )E1 /Spin(9) S 9 .
Theorem 3.10.4.
(E6 )E1
= Spin(10).

Proof. Analogously to Theorem 2.7.4, we can define a homomorphism
p : (E6 )E1 SO(10) = SO(V 10 )
by p() = |V 10 . The restriction p of p : (E6 )E1 SO(10) to (F4 )E1 coincides
with the homomorphism p : Spin(9) SO(9) of Theorem 2.7.4. In particular,
p : Spin(9) SO(9) is onto. Hence, from the following commutative diagram
88
Spin(9)
p
SO(9)
(E6 )E1
p
SO(10)
S9
=
S9
we see that p : (E6 )E1 SO(10) is onto by the five lemma. We also have Ker p =
{1, }. Indeed, Ker p leaves E2 E3 and i(E2 + E3 ) invariant, so that Ei = Ei
for i = 1, 2, 3. Hence we have Spin(8) and so Ker p . Therefore = 1 or
by Theorem 2.7.4. Hence we have the isomorphism
(E6 )E1 /{1, }
= SO(10).
Therefore the group (E6 )E1 is isomorphic to the group Spin(10) as the universal
covering group of SO(10).
For C, 6= 0, we define a
1 x3
() x3 2
x2 x1
C-linear mapping () : JC JC by
4
x2
1
x3
x2
x1 = x3 2 2 2 x1 .
3
x2 2 x1 2 3
Theorem 3.10.5. The group E6 contains a subgroup
U (1) = {12 (t)13 (t) = exp it(2E1 E2 E3 ) | t R}

= {() | C, ( ) = 1}
(where 12 (t), 13 (t) are mappings defined in Lemma 3.8.1) which is isomorphic to
the usual unitary group U (1) = { C | ( ) = 1}.
Note that U (1) is a subgroup of (E6 ) . From now on, we identify these two groups
U (1).
Lemma 3.10.6. Two subgroups U (1) and Spin(10) of (E6 ) are elementwise
commutative.
Proof. We consider the decomposition JC = E1 C J(2, C)C (JC ) , where
E1 C = {E1 | C},
J(2, C)C = {X JC | 4E1 (E1 X) = X},
(JC ) = {X JC | E1 X = 0, hE1 , Xi = 0}.
The restrictions of () U (1) to these spaces are all constant mappings:

()|E1 C = 4 1,
()|J(2, C)C = 2 1,
()|(JC ) = 1.
On the other hand, Spin(10) also induces C-linear transformations of these

spaces. From this, the commutativity of () and : () = () follows.
Theorem 3.10.7. (E6 )
= (U (1) Spin(10))/Z 4 , Z 4 = {(1, (1)), (1, (1)),
(i, (i)), (i, (i))}.
89
Proof. We define a mapping : U (1) Spin(10) (E6 ) by

(, ) = ().
Since () U (1) and Spin(10) are commutative (Lemma 3.10.6), we see that
is a homomorphism. We shall show that is onto. For (E6 ) , there exists C,
( ) = 1 satisfying
E1 = 4 E1 = ()E1
(Lemma 3.10.1). Let = ()1 , then E1 = E1 , so Spin(10) (Theorem
3.10.4). Hence, we have = () = (, ) and therefore is onto. It is easily seen
that
Ker = {(, ()1 ) | C, 4 = 1} = {(, (1 )) | = 1, i} = Z 4 .
Thus we have the isomorphism (U (1) Spin(10))/Z 4
= (E6 ) .
3.11. Involution and subgroup (Sp(1) SU (6))/Z 2 of E6
Let the C-linear mapping : JC JC be the complexification of G2 F4 .
Then E6 and 2 = 1.
(E6 ) = { E6 | = }.
As in Section 2.11, we use the decomposition
JC = J(3, H)C (H 3 )C ,
which is the complexification of the decomposition J = J(3, H) H 3 . As usual we
denote J(3, H) and {X J(3, H) | tr(X) = 0} by JH and (JH )0 , respectively.
We consider the embedding C = {x + ye1 | x, y R} C and, for an element
a = x + ye1 C, we denote by a the element x + yi C. Now, we define an R-linear
mapping k : H M (2, C) by

a
b
k(a + be2 ) =
, a, b C.
b a
The mapping k is naturally extended to R-linear mappings
k : M (3, H) M (6, C) and k : H 3 M (2, 6, C).
We adopt the following notations.
J 0 0
J = 0 J 0 M (6, C),
0 0 J
90
J=
0 1
1 0
and M = t M , M M (3, H). Then the following four properties hold.

(1)
(2)
(3)
(4)
k(M N ) = k(M )k(N ),

t (k(M )) = k(M ),
J(k(M )) = ( (k(M )))J,
k(aM ) = k(a)k(M ),
M, N M (3, H), a H 3 .
The R-linear mappings k : M (3, H) M (6, C) and k : H 3 M (2, 6, C) are extended to C-linear mappings k : M (3, H)C M (6, C) and k : (H 3 )C M (2, 6, C)
respectively by
k(M1 + iM2 ) = k(M1 ) + ik(M2 ),
k(a1 + ia2 ) = k(a1 ) + ik(a2 ),
M1 , M2 M (3, H),
a1 , a2 H 3 .
It is not difficult to see that they satisfy the four properties (1) (4) above.
We define a C-vector space S(6, C) by
S(6, C) = {S M (6, C) | t S = S}
and a C-linear mapping kJ : J(3, H)C S(6, C) by
kJ (M ) = k(M )J.
Then kJ is well-defined. Indeed, for M = M1 + iM2 J(3, H)C , we have
t
(kJ (M )) = t ((k(M ))J) = J t (k(M )) = J t (k(M1 ) + ik(M2 ))
= ( t (k(M1 )) + i t (k(M2 )))J = (k(M1 ) + ik(M2 ))J
= (k(M1 ) + ik(M2 ))J = (k(M ))J = kJ (M ).
Finally, we define Hermitian inner products hS, T i in S(6, C) and hP, Qi in M (2, 6,
C) respectively by
hS, T i = tr(( t S)T ),
hP, Qi = tr(( t P )Q).
Lemma 3.11.1. (1) k : M (3, H)C M (6, C), k : (H 3 )C M (2, 6, C) and

kJ : J(3, H)C S(6, C) are C-linear isomorphisms.
(2) hkJ (M ), kJ (N )i = 2hM, N i,
M, N J(3, H)C ,
(3) det(kJ (M )) = (detM )2 ,
M J(3, H)C .
hk(a), k(b)i = 2ha, bi,
a, b (H 3 )C .
Proof. (1) The mapping k is injective. Indeed, if

k (a + be2 ) + i(c + de2 ) = 0, a, b, c, d C,
then
a
b
b
a
+i
91
c
d
d
c
= 0.
From which, we have a = b = c = d = 0 and so a = b = c = d = 0. Now, k is a

C-linear isomorphism, because dimC (M (3, H)C ) = 36 = dimC (M (6, C)). The other
mapping can be treated analogously.
(2) For mk = ak + bk e2 and nk = ck + dk e2 , ak , bk , ck , dk C, k = 1, 2, we have,
after some calculations,
hk(m1 + in1 ), k(m2 + in2 )i = 2hm1 + in1 , m2 + in2 i.
(3) Note that det(kJ (M )) = det(k(M )) and detM C. Since we know that detS
of a skew-symmetric matrix S is expressed in terms of the square of a polynomial
with entries sij in S, we can easily see that
0
s12
s13
s14
s15 s16
0
s23
s24
s25 s26
s12
0
s34
s35 s36
s13 s23
det
0
s45 s46
s14 s24 s34
s15 s25 s35 s45

0
s56
s16 s26 s36 s46 s56 0
= (s12 s34 s56 s12 s35 s46 + s12 s36 s45 s13 s24 s56 + s13 s25 s46 s13 s26 s45
+s14 s23 s56 s14 s25 s36 + s14 s26 s35 s15 s23 s46 + s15 s24 s36 s15 s26 s35
+s16 s23 s45 s16 s24 s35 + s16 s25 s34 )2 .
Now, since kJ (M ) is skew-symmetric, using the above result, we have
0
1
n3
m3
n2
(m2 )
0
(m3 ) (n3 )
m2
(n2 )
1
(m3 )
0
2
n1
m1
n3
det(kJ (M )) = det
(n3 )
2
0
(m1 ) (n1 )
m3
n2
m2
n1
(m1 )
0
3
(m2 ) (n2 )
m1
(n1 )
3
0
= (1 2 3 1 n1 (n1 ) 1 m1 (m1 ) n3 (n3 )3 n3 m2 (n1 ) n3 (n2 ) (m1 )

m3 (m3 )3 + m3 m2 m1 m3 (n2 )n1 n2 (m3 ) (n1 ) n2 (n3 )m1 n2 (n2 )2
+ (m2 ) (m3 ) (m1 ) (m2 ) (n3 )n1 (m2 )m2 2 )2 .
On the other hand, we have
1
m3 + n 3 e 2
(m2 + n2 e2 )
det (m3 + n3 e2 )
2
m1 + n 1 e 2
m2 + n 2 e 2
(m1 + n1 e2 )
3
= 1 2 3 + (m1 + n1 e2 )(m2 + n2 e2 )(m3 + n3 e2 )
3
X
i (mi + ni e2 ) (mi + ni e2 )
+ ((m1 + n1 e2 )(m2 + n2 e2 )(m3 + n3 e2 ))
i=1
= the contents of in the parenthesis above.
We now consider a group E6,H defined by replacing C by H in the definition of

the group E6 :
92
E6,H = { IsoC ((JH )C ) | det(M ) = detM, hM, N i = hM, N i}.

Lemma 3.11.2. E6,H is a connected group of dimension 35.
Proof. As in the case E6 , the group E6,H contains a subgroup
F4,H = { E6,H | E = E},
which is also defined by
F4,H = { IsoR (JH ) | (M N ) = M N }
(see Lemma 2.2.4). Moreover, it is isomorphic to the group Sp(3)/Z 2 (Proposition
2.11.1). The Lie algebra e6,H of the group E6,H is given by
e6,H = { HomC ((JH )C ) | (M, M, M ) = 0, hM, N i + hM, N i = 0}
= { + iTe | f4,H , T (JH )0 },
(see Theorem 3.2.1), so that
e6,H = f4,H i(e

JH )0 .
Where f4,H = { e6,H | E = 0} is the Lie algebra of the group F4,H and is
isomorphic to the Lie algebra sp(3) by the mapping : sp(3) f4,H given by
(C)M = CM + M C = [C, M ],
M JH
(Proposition 2.11.1) and we see that dim e6,H = 21 + 15 = 35. As in Theorem 3.8.3,
we have a homeomorphism
E6,H /F4,H EIVH = {X (JH )C | detM = 1, hM, M i = 3}
and so we see that the group E6,H is connected.
Proposition 3.11.3.
E6,H
= SU (6)/Z 2 , Z 2 = {E, E}.
Proof. Let SU (6) = {A M (6, C) | ( t A)A = E, detA = 1} and define a

mapping : SU (6) E6,H by
(A)M = kJ 1 (A(kJ (M )) t A),
M (JH )C .
We first have to prove that (A) E6,H . Indeed, we have (det((A)M ))2 =
det(kJ ((A) M )) (Lemma 3.11.1) = det(A(kJ (M ))t A) = det(kJ (M )) = (detM )2
(Lemma 3.11.1), and so det((A)M ) = detM. On the other hand, since E6,H is
connected (Lemma 3.11.2), the sign of det((A)M ) is constant, that is, independent
of A (assuming that detM 6= 0). Therefore
det ((A)M ) = det M.
93
Next, again we have

2h(A)M, (A)N i = hkJ ((A)M ), kJ ((A)N )i (Lemma 3.11.1)
= hA(kJ (M )) t A, A(kJ (N )) t Ai = hkJ (M ), kJ (N )i (because A SU (6))
= 2hM, N i,
so (A) E6,H . Consequently the mapping : SU (6) E6,H is well-defined.
Evidently is a homomorphism. That Ker = {E, E} = Z 2 can be easily obtained.
Since the group E6,H is connected and dim SU (6) = 35 = dim E6,H (Lemma 3.11.2),
is onto. Thus we have the isomorphism SU (6)/Z 2
= E6,H .
Theorem 3.11.4.
(E6 )
= (Sp(1) SU (6))/Z 2 , Z 2 = {(1, E), (1, E)}.
Proof. We define a mapping : Sp(1) SU (6) (E6 ) by
(p, A)(M + a) = kJ 1 (A(kJ (M )) t A) + pak 1 ( t A),

M + a (JH )C (H 3 )C = JC .
We first need to prove that (p, A) (E6 ) .

Claim 1.
(p, A)1 = (p, A).
Proof. we have
2h t (p, A) (M + a), N + bi
M + a, N + b JC
= 2hM + a, (p, A)(N + b)i
= 2hM + a, kJ 1 (A(kJ (N )) tA) + pbk 1 ( tA)i
= 2hM, kJ 1 (A(kJ (N )) tA)i + 4ha, pbk 1 ( tA)i
= hkJ M, A(kJ (N )) t Ai + 2hk(pa), (kb) t Ai
= h t A(kJ (M )) A, kJ (N )i + 2hk(pa)A, kbi
= 2hkJ 1 ( t A(kJ (M )) A), N i + 4hpak 1 (A), bi
= 2hkJ 1 ( t A(kJ (M )) A + pak 1 (A)), N + bi

= 2h(p, t A)(M + a), N + bi,
and so we have t (p, A) = (p, t A), which implies that t (p, A)1 = (p, A).
Claim 2.
(p, A) (E6 ) .
Proof. We will first show that = (p, A) satisfies

X Y = t 1 (X Y ),
X, Y JC .
Recall the equality

1

1
(M + a) (N + b) = M N (a b + b a) (aN + bM ).
2
2
The equality M N = t 1 (M N ) is evident from det(M ) = detM (Proposition

3.11.3). Now, using the equalities
kJ 1 (M ) = k 1 (M J) = k 1 (J( M )),
94
k 1 (M ) = k 1 (JM J),
we have
(a) (b) = (pak 1 ( t A)) (pbk 1 ( t A))
= k 1 (A)a bk 1 ( t A),
(a b) = (kJ 1 (A(kJ (a b)) t A))
= k 1 (J (A(k( (a b)))J t A))
= k 1 (J (AJ k(a b) t A))
= k 1 (J AJk(a b) t A)
= k 1 (A k(a b)J t AJ) = k 1 (A)a bk 1 ( tA),
(a)(N ) = (pak 1 ( t A))kJ 1 (A(kJ (N )) tA)
= pak 1 ( t A)k 1 (A(k(N ))J t AJ)
= paN k 1 (J t AJ) = paN k 1 ( t A),
(aN ) = (p (aN )k 1 ( t A)) = paN k 1 ( t A).

We have therefore shown that (p, A) E6 , Clearly, (p, A) = (p, A), so that
(p, A) (E6 ) .
We will return to the proof of Theorem 3.11.4. Evidently is a homomorphism.
We shall now show that is onto. Let (E6 ) . Since the restriction =
|(JH )C of to (JH )C belongs to E6,H , there exists A SU (6) such that = (A)
(Proposition 3.11.3). If we put = (1, A)1 , then |(JH )C = 1. Hence, by the
same argument as in Theorem 2.11.2, there exists p Sp(1) such that = (p, E),
and we obtain
= (1, A) = (1, A)(p, E) = (p, A).
Therefore is onto. Ker = {(1, E), (1, E)} = Z 2 can be easily obtained. Thus
we have the isomorphism (Sp(1) SU (6))/Z 2
= (E6 ) .
3.12. Involution and subgroup Sp(4)/Z 2 of E6
We consider an involutive complex conjugate linear transformation of JC , and
we shall study the following subgroup (E6 ) of E6 :
(E6 ) = { E6 | = }
= { E6 | () = } = (E6 ) .
For this end, we consider R-vector subspaces (JC ) and (JC ) of JC , which are
eigenspaces of , respectively by
(JC ) = {X JC | X = X}
(
1 m3 m2
0
= m3 2 m1 + i a3 e4
a2 e 4
m2 m1 3
95
a3 e 4
0
a1 e4
a2 e4
a1 e 4
0

)

i R

m i , ai H
= {M + iF (ae4 ) | M J(3, H), a H 3 }

= JH iH 3 ,
(JC ) = {X JC | X
(
1 m3
= i m3 2
m2 m1
= X}

m2
0
m1 + a3 e4
3
a2 e 4
a3 e 4
0
a1 e4
= {iM + F (ae4 ) | M J(3, H), a H 3 }
a2 e4
a1 e 4
0
= iJH H 3 ,

)

i R

m i , ai H
The spaces (JC ) and (JC ) are invariant under the action of (E6 ) and we have
the decomposition of JC :
JC = (JC ) (JC ) = (JC ) i(JC ) .
In particular, JC is the complexification of (JC ) : JC = ((JC ) )C .
In the R-vector space
J(4, H) = {P M (4, H) | P = P },
we define the Jordan multiplication P Q and an inner product (P, Q) respectively
by
1
P Q = (P Q + QP ), (P, Q) = tr(P Q).
2
The group Sp(4) acts on J(4, H) by the mapping : Sp(4) J(4, H) J(4, H),
(A, P ) = AP A . Then we have
A(P Q)A = AP A AQA ,
(AP A , AQA ) = (P, Q),
A Sp(4), P, Q J(4, H).
The quaternion projective space HP3 is defined by

HP3 = {P J(4, H) | P 2 = P, tr(P ) = 1}
n
o
= AE1 A | A Sp(4), E1 = diag(1, 0, 0, 0) M (4, H) .
Finally, in the complexification J(4, H)C of J(4, H), we extend naturally the Jordan
multiplication P Q and the inner product (P, Q) and further define a Hermitian
inner product hP, Qi by
hP, Qi = ( P, Q).
The action of Sp(4) on J(4, H) is also naturally extended to J(4, H)C :
A(X1 + iX2 )A = AX1 A + iAX2 A ,
A Sp(4), X1 , X2 J(4, H).
Then we have
hAP A , AQA i = hP, Qi,
96
P, Q J(4, H)C .
We denote by J(4, H)0 the space {P J(4, H) | tr(P ) = 0} and by J(4, H)0
complexification.
its
Definition. We define a C-linear mapping g : JC J(4, H)0 C by

1
tr(M )
ia
, M + a (JH )C (H 3 )C = JC .
g(M + a) = 2
1
ia
M tr(M )E
2
The restriction of g to (JC ) is given by

1
tr(M ) a
g(M + ia) =
tr(M )E,
a
M
2
M + ia JH iH 3 = (JC ) .
Note that the mapping g in the above definition is the complexification of this
restriction.
Lemma 3.12.1. The mapping g : JC J(4, H)0 C is a C-linear isomorphism
and satisfies
1
gX gY = g((X Y )) + (X, Y )E,
4
(gX, gY ) = (X, Y ),
X, Y JC .
Moreover, g is an isometry with respect to the inner product hX, Y i:

hgX, gY i = hX, Y i,
X, Y JC .
The restriction of g to (JC ) induces an R-linear isomorphism g : (JC ) J(4, H)0 .

Proof. It is not difficult to see that g is well-defined and that it is injective. Since
C
dimC JC = 27 = dimC J(4, H)0 , g is a C-linear isomorphism. Now, for X = M + a,
3
Y = N + b (JH )C (H )C = JC , we have
g((X Y )) = g(((M + a) (N + b)))
= g((M a) (N b))
1
1
= g((M N (a b + b a)) + (aN + bM ))
2
2
1
1
i
tr(M N ) (a, b)
(aN + bM )
2
2
=2 i
1
1
(aN + bM )
M N (a b + b a) (tr(M N ) (a, b))E
2
2
2
=

1
1
(M, N ) (a, b) E
= g(M + a) g(N + b)
4
2
1
= g(M + a) g(N + b) ((M + a), (N + b))E
4
1
= gX gY (X, Y )E.
4
97
1
Thus the first equality gX gY = g((X Y )) + (X, Y )E is proved. Taking the
4
traces of the both sides, we have
(gX, gY ) = (X, Y ),
X, Y JC .
(i)
It is easily seen that the restriction g : (JC ) J(4, H)0 of g is an R-isomorphism.

Finally, if we note that (X, Y ) = hX, Y i for X, Y (JC ) , then we can easily show
that
hg(X1 + iX2 ), g(Y1 + iY2 )i = hX1 + iX2 , Y1 + iY2 i,
Xi , Yi (JC )
from (i). Thus Lemma 3.12.1 is proved.

Theorem 3.12.2.
(E6 )
= Sp(4)/Z 2 , Z 2 = {E, E}.
Proof. We define a mapping : Sp(4) (E6 ) by

(A)X = g 1 (A(gX)A ),
X JC .
We first have to prove that (A) (E6 ) . Let Z = (A)X and use Lemma 3.12.1,
then we have
and
3det((A)X) = 3detZ = (Z Z, Z) = (g((Z Z)), gZ)

1
= gZ gZ (Z, Z)E, gZ
4

1
= gZ gZ (gZ, gZ)E, gZ
4

1
= A(gX)A A(gX)A (A(gX)A , A(gX)A )E, A(gX)A
4
1
= (gX gX (gX, gX)E, gX)
4
1
= (gX gX (X, X)E, gX)
4
= (g((X X)), gX) = (X X, X) = 3detX,
h(A)X, (A)Y i = hg(A)X, g(A)Y i
= hA(gX)A , A(gY )A i = hgX, gY i = hX, Y i.
Hence (A) E6 . To prove (A) (E6 ) , namely, (A) = (A) , it is sufficient

to show that
(A) X = (A)X, X (JC ) ,
since JC = ((JC ) )C . However this is evident. Indeed, if X (JC ) , then gX
J(4, H)0 , so that (A)X (JC ) . Hence (A) X = (A)X = (A)X.
Evidently is a homomorphism. We shall show that is onto. For (E6 ) , we
have
3
(g(E))2 = g(E) + E.
4
98
Certainly,
1
(g(E))2 = g(E) g(E) = g((E E)) + (E, E)E (Lemma 3.12.1)
4
1
1
= g( E) + h E, EiE = g( E) + h E, EiE
4
4
1
3
1
= g(E) + hE, EiE = g(E) + hE, EiE = g(E) + E.
4
4
4
Now, let
1
(2g(E) + E),
4
then we have P J(4, H) and P 2 = P , tr(P ) = 1, that is, P HP 3 . Indeed,
P =
1
1
(4(g(E))2 + 4g(E) + E) = (2g(E) + E) = P,
16
4
1
1
tr(P ) = tr(2g(E)) + tr(E) = 0 + 1 = 1.
4
4
P2 =
Hence there exists A Sp(4) such that

P = AE1 A .
1
Since gE = 2E1 E, we have
2

1
1
(A)E = g 1 (A(gE)A ) = g 1 A 2E1 E A = g 1 2AE1 A E
2
2
= g 1 (g(E)) = E.
Putting = (A)1 , we have E = E, and so F4 (Theorem 3.7.1). Also
satisfies = and = (Theorem 3.7.1), hence = , and therefore
(F4 ) . From Theorem 2.11.2, there exist p Sp(1) and D Sp(3) such that
(M + a) = DM D + paD , M + a JH H 3 = J.

p 0
Let B =
, then B Sp(4) and we have
0 D
= (B).
Certainly, for M + a (JH )C (H 3 )C = JC ,
(B)(M + a) = g 1 (Bg(M + a)B )
1

tr(M )
ia
p
p
0
1
2
(
= g
1
0
0 D
ia
M tr(M )E
2
1

tr(M )
ipaD
= g 1 2
1
iDa p DM D tr(M )E
2
= DM D + paD = (M + a).
99
0
D

Hence we have
= (A) = (A)(B) = (AB),
AB Sp(4),
so that is onto. Ker = {E, E} = Z 2 can be easily obtained. Thus we have the
isomorphism Sp(4)/Z 2
= (E6 ) .
3.13. Automorphism w of order 3 and subgroup (SU (3)SU (3)SU (3))/Z 3
of E6
Let the C-linear mapping w : JC JC be the complexification of w G2 F4
of Section 2.12. Then w E6 and w3 = 1.
We shall study the following subgroup (E6 )w of E6 :
(E6 )w = { E6 | w = w}.
As in Section 2.12, we identify
J(3, C)C M (3, C)C = JC .
For convenience, we denote J(3, C) and {X J(3, C) | tr(X) = 0} by JC and (JC )0 ,
respectively.
The group E6,C is defined to be obtained by replacing JC with (JC )C in the
definition of the group E6 :
E6,C = { IsoC ((JC )C ) | det (X) = det X, hX, Y i = hX, Y i}.
As in the case E6 , the group E6,C contains a subgroup
F4,C = { E6,C | E = E},
which is also defined by the group F4,C = { IsoR (JC ) | (X Y ) = X Y },
moreover, it is isomorphic to the group (SU (3)/Z 3 ) Z 2 (Proposition 2.12.1). The
Lie algebra e6,C of the group E6,C is
e6,C = { HomC ((JC )C ) | (X, X, X) = 0, hX, Y i + hX, Y i = 0}
= { + iTe | f4,C , T (JC )0 }
(Theorem 3.2.4), where f4,C = { e6,C | E = 0} is the Lie algebra of the group
F4,C . In particular, the dimension of e6,C is
dim(e6,C ) = 8 + 8 = 16.
As in Theorem 3.8.3, we see that the space
EIVC = {X (JC )C | detX = 1, hX, Xi = 3}
100
is connected and we have the homeomorphism

E6,C /F4,C EIVC .
Lemma 3.13.1. E6,C has at most two connected components (in reality has two
connected components (Proposition 3.13.4)).
Proof. From the exact sequence 0 (F4,C ) 0 (E6,C ) 0 (EIVC ), that is,
Z 2 0 (E6,C ) 0 (Proposition 2.12.1), we see that 0 (E6,C ) is 0 or Z 2 .
We define mappings h : C C C C and h : M (3, C) M (3, C) M (3, C)C
respectively by
h(a, b) =
h(A, B) =
a+b
ab
+i
e1 = a + b,
2
2
AB
A+B
+i
e1 = A + B,
2
2
1 + ie1
.
2
Lemma 3.13.2. The mappings h : C C C C and h : M (3, C) M (3, C)

M (3, C)C satisfy the following four conditions.
(1) Both are C-linear isomorphisms, that is, they are injective and satisfy
h(a, b) + h(a , b ) = h(a + a , b + b ), h(ca, cb) = ch(a, b), c C,
h(A, B) + h(A , B ) = h(A + A , B + B ), h(cA, cB) = h(A, B), c C.
(2) h(a, b)h(a , b ) = h(aa , bb ), h(A, B)h(A , B ) = h(AA , BB ).

(3) h(a, b) = h(b, a), h(a, b) = h(b, a).
h(A, B) = h(B, A), h(A, B) = h(B, A), h(A, B) = h(B , A ).

(4) det (h(A, B)) = h(det A, det B).
Proof. It is easy to prove, noting that 2 = , 2 = , + = 1.
Lemma 3.13.3.
e6,C
= su(3) su(3).
Proof. The mapping C : su(3) su(3) e6,C ,

C (C, D)X = h(C, D)X + Xh(C, D) ,
X (JC )C
gives an isomorphism as Lie algebras. This is the direct consequence of the following
Proposition 3.13.4, so we will omit its proof here.
We define an action of the group Z 2 = {1, } on SU (3) SU (3) by
(A, B) = (B, A),
and let (SU (3) SU (3)) Z 2 be the semi-direct product of the groups SU (3) SU (3)
and Z 2 under this action.
101
((SU (3) SU (3))/Z 3 ) Z 2 , Z 3 = {(E, E), (1 E,

Proposition 3.13.4. E6,C =
1
3
2
2
1 E), (1 E, 1 E)}, 1 = +
e1 .
2
2
Proof. We define a mapping : (SU (3) SU (3)) Z 2 E6,C by
((A, B), 1)X = h(A, B)Xh(A, B) ,
((A, B), )X = h(A, B)Xh(A, B) ,
X (JC )C .
First we have to show that = ((A, B), 1) E6,C . Using det(h(A, B)) = h(detA,
detB) (Lemma 3.13.2.(4)) = h(1, 1) = 1 and h(A, B) h(A, B) = h(A , B ) h(A, B)
(Lemma 3.13.2.(3)) = h(A A, B B) = h(E, E) = E, we have
det (X) = (det (h(A, B)))(det X)(det (h(A, B) )) = det X,
hX, Y i = hh(A, B)Xh(A, B) , h(A, B)Y h(A, B) i
= ( h(A, B) X h(A, B) , h(A, B)Y h(A, B) )
= ( X, Y ) = hX, Y i.
Hence E6,C . Since ((E, E), ) = G2,C (= Aut(C)) F4,C E6,C , we also
have ((A, B), ) = ((A, B), 1)((E, E), ) E6,C . Next, we shall show that is a
homomorphism. Indeed, for instance,
((A, B), )((C, D), 1)X
= ((A, B), )(h(C, D)Xh(C, D) )
= h(A, B)h(C, D)Xh(C, D) h(A, B)
= h(A, B)h(D, C)Xh(D, C) h(A, B) (Lemma 3.13.3.(3))
= h(AD, BC)Xh(AD, BC) = ((AD, BC), )X
= ((A, B)(C, D), )X,
X (JC )C .
That Ker = {(E, E), (1 E, 1 E), (1 2 E, 1 2 E)} 1 = Z 3 1 can be easily obtained. In particular, Ker is discrete. Hence induces an injective homomorphism
: su(3) su(3) e6,C .
In particular, dim(su(3) su(3)) = dim(e6,C ), so is an isomorphism ( coincides
with C of Lemma 3.13.3). Hence, induces the surjection : SU (3) SU (3)
(E6,C )0 (which denotes the connected component of E6,C containing the identity 1).
However = ((E, E), ) 6 (E6,C )0 . Certainly, for any A, B SU (3),
h(A, B)Xh(A, B) = X,
X (JC )C
does not hold. Therefore E6,C has just two connected components (Lemma 3.13.1).
Consequently, : (SU (3) SU (3)) Z 2 E6,C is onto and we have the isomorphism
((SU (3) SU (3))/Z 3 ) Z 2
= E6,C .
102
Theorem 3.13.5. (E6 )w

= (SU (3) SU (3)
SU (3))/Z 3 , Z 3 = {(E, E, E),
1
3
(1 E, 1 E, 1 E), (1 2 E, 1 2 E, 1 2 E)}, 1 = +
e1 .
2
2
Proof. We define a mapping : SU (3) SU (3) SU (3) (E6 )w by
(P, A, B)(X + M ) = h(A, B)Xh(A, B) + P M h(A, B) ,
X + M (JC )C M (3, C)C = JC .
We first have to show that = (P, A, B) (E6 )w , that is, leaves
1
1
(X + M ) (Y + N ) = (X Y (M N + N M )) (M Y + N X + M N ),
2
2
hX + M, Y + N i = hX, Y i + 2hM, N i
invariant, namely, satisfies

((X + M ) (Y + N )) = (X + M ) (Y + N ),
h(X + M ), (Y + N )i = hX + M, Y + M i.
Now, for X + M, Y + N (JC )C M (3, C)C = JC , the relations

(X Y ) = X Y,
hX, Y i = hX, Y i
are already shown in Proposition 3.13.4. Next,

(P M h(A, B) ) (P N h(A, B) ) = h(A, B)M P P N h(A, B)
= (h(A, B) (M N )h(A, B) ),
(P M h(A, B) )(h(A, B)Y h(A, B) ) = (P (M Y ) h(A, B) ),
(P M h(A, B) ) (P N h(A, B) ) = t Pe(M N ) t (h(A, B) )
= P (M N )h(A, B) = (P (M N ) h(A, B) ).
Furthermore, we have
hX, M i = 0 = hX, M i,
hM, N i = ( M, N ) = (P M h(A, B) , P N h(A, B) )
= ( M, N ) = hM, N i,
and so E6 . Clearly, w = w, so that (E6 )w . Evidently is a homomorphism. We shall now show that is onto. Let (E6 )w . The restriction of
to (JC )w = {X JC | wX = X} = (JC )C belongs to E6,C : E6,C . Hence, there
exist A, B SU (3) such that
X = h(A, B)Xh(A, B)
or X = h(A, B)Xh(A, B) ,
X (JC )C
(Proposition 3.13.4). In the former case, let = (E, A, B)1 , then |(JC )C = 1.
Hence G2 , moreover (G2 )e1 = (G2 )w = SU (3) (Theorem 1.9.4). Thus there
exists P SU (3) such that
(X + M ) = X + P M = (P, E, E)(X + M ),
103
X + M (JC )C M (3, C)C = JC .
Therefore we have
= (E, A, B) = (E, A, B)(P, E, E) = (P, A, B).
In the latter case, consider the mapping 1 : JC JC , 1 (X + M ) = X + M , X +
M (JC )C M (3, C)C = JC and remember that 1 G2 F4 E6 . Let
= 1 (E, A, B)1 , then E6 and |(JC )C = 1. Hence (G2 )e1 = (G2 )w
(Theorem 1.9.4) (E6 )w . Since and (E, A, B) (E6 )w , we have 1 (E6 )w , so
that 1 (G2 )w which is a contradiction (Theorem 1.9.4). Therefor we see that
is onto. That Ker = {(E, E, E), (1 E, 1 E, 1 E), (1 2 E, 1 2 E, 1 2 E)} = Z 3 can
be easily obtained. Thus we have the isomorphism (SU (3) SU (3) SU (3))/Z 3
=
w
(E6 ) .
Remark 1. Since (E6 )w is connected as the fixed points subgroup of E6 by
an automorphism w of order 3 of the simply connected group E6 , the fact that :
SU (3) SU (3) SU (3) (E6 )w is onto can be proved as follows. The elements
G01 , G23 ,
G45 , G67 , G26 + G37 , G27 + G36 ,
G24 + G35 , G25 + G34 , G46 + G57 , G47 + G56 ,
e1 (1), A
e2 (1), A
e3 (1), A
e1 (e1 ), A
e2 (e1 ), A
e3 (e1 ),
A
(E1 E2 ) ,
(E2 E3 ) ,
Fe1 (1),
Fe1 (e1 ),
Fe2 (1), Fe3 (1),

Fe2 (e1 ), Fe3 (e1 )
forms an R-basis of (e6 )w . So dim((e6 )w ) = 16 + 8 = 24 = 8 + 8 + 8 = dim(su(3)

su(3) su(3)), Hence is onto.
Remark 2. The group E6 has a subgroup which is isomorphic to the group
((SU (3) SU (3) SU (3))/Z 3 ) Z 2 which is the semi-direct product of the groups
(SU (3) SU (3) SU (3))/Z 3 and Z 2 (the action of Z 2 = {1, 1 } on the group
SU (3) SU (3) SU (3) is 1 (P, A, B) = (P , B, A)).
3.14. Complex exceptional Lie group E6 C
Theorem 3.14.1. The polar decomposition of the Lie group E6 C is given by
E6 C E6 R78 .
In particular, E6 C is a simply connected complex Lie group of type E6 .
Proof. Evidently E6 C is an algebraic subgroup of IsoC (JC ) = GL(27, C). The
conjugate transposed mapping of E6 C with respect to the inner product
hX, Y i is = t E6 C . Hence, from Chevalleys lemma, we have
E6 C (E6 C U (JC )) Rd = E6 Rd ,
104
d = 78.
Since E6 is simply connected (Theorem 3.9.2), E6 C is also simply connected. The Lie
algebra of the group E6 C is e6 C , so E6 C is a complex simple Lie group of type E6 .
3.15. Non-compact exceptional Lie groups E6(6) , E6(2) , E6(14) and E6(26)
of type E6
by
We define Hermitian inner products hX, Y i and hX, Y i in J(3, CC ) respectively

hX, Y i = hX, Y i,
hX, Y i = hX, Y i.
Now, we define groups E6(6) , E6(2) , E6(14) and E6(26) respectively by

E6(6) = { IsoR (J(3, C )) | det (X) = det X},
E6(2) = { IsoC (J(3, CC )) | det (X) = det X, hX, Y i = hX, Y i },
E6(14) = { IsoC (J(3, CC )) | det (X) = det X, hX, Y i = hX, Y i },
E6(26) = { IsoR (J(3, C)) | det (X) = det X}.

These groups can also be defined by
E6(6)
= (E6 C ) .
= (E6 C ) , E6(26)
= (E6 C ) , E6(14)
= (E6 C ) , E6(2)
Theorem 3.15.1. The polar decompositions of the Lie groups E6(6) , E6(2) , E6(14)
and E6(26) are respectively given by
E6(6) Sp(4)/Z 2 R42 ,
E6(2) (Sp(1) SU (6))/Z 2 R40 ,
E6(14) (U (1) Spin(10))/Z 4 R32 ,
E6(26) F4 R26 .
Proof. These are the facts corresponding to Theorems 3.12.2, 3.11.4, 3.10.7 and
3.7.1.
Theorem 3.15.2. The centers of the groups E6(6) , E6(2) , E6(14) and E6(26) are
given by
z(E6(6) ) = {1}, z(E6(2) ) = Z 3 , z(E6(14) ) = Z 3 , z(E6(26) ) = {1}.
105
4. Exceptional Lie group E7

4.1. Freudenthal vector space PC
Definition. We define a C-vector space PC , called the Freudenthal C-vector
space, by
PC = JC JC C C.

X
Y
PC is a 56 dimensional C-vector space. An element of PC is often denoted by
(X, Y, , ) or sometimes by X +Y. + +. . In PC , we define an inner product (P, Q), a

positive definite Hermitian inner product hP, Qi and a skew-symmetric inner product
{P, Q} respectively by
(P, Q) = (X, Z) + (Y, W ) + + ,
hP, Qi = hX, Zi + hY, W i + ( ) + ( ),
{P, Q} = (X, W ) (Z, Y ) + ,

where P = (X, Y, , ), Q = (Z, W, , ) PC .
For e6 C , A, B JC , C, we define a C-linear mapping (, A, B, ) :

PC PC by
1

2B
0 A
X
X
3
Y
Y
1

t + B 0
(, A, B, ) = 2A

3

0
A
B
0
0
1
X X + 2B Y + A
3
2A X t Y + 1 Y + B
=
.
3
(A, Y ) +
(B, X)
Definition. For P = (X, Y, , ), Q = (Z, W, , ) PC , we define a C-linear

mapping P Q : PC PC by
= (X W + Z Y )
A = (2Y W Z X)
4
P Q = (, A, B, ),
1
B =
(2X Z W Y )
= 1 ((X, W ) + (Z, Y ) 3( + )).

8
106
Lemma 4.1.1. For P, Q, R PC , we have
(1) P Q = Q P .
3
(2) (P Q)P (P P )Q + {P, Q}P = 0.
8
1
1
1
(3) (P R)Q (Q R)P + {Q, R}P {P, R}Q {P, Q}R = 0.
8
8
4
Proof. (1) is evident.
(2) For P = (X, Y, , ), Q = (Z, W, , ) PC , we have
(P Q)P
1
1
1
= (X W + Z Y ), (2Y W Z X), (2X Z W Y ),
2
4
4

1
((X, W ) + (Z, Y ) 3( + )) (X, Y, , )
8
= (using the formula X Y of Lemma 3.4.1), etc.)

1
1
1
= X Y, (Y Y X), (X X Y ), ((X, Y ) 3) (Z, W, , ),
2
2
4
3
((X, W ) (Z, Y ) + )(X, Y, , )
8
3
= (P P )Q {P, Q}P.
8
(3) In the equality (2), put P + R in the place of P , then we have
3
3
2(P R)Q (P Q)R (R Q)P + {Q, R}P {P, Q}R = 0.
8
8
Exchanging P with Q, we see that
3
3
2(Q R)P (Q P )R (R P )Q + {P, R}Q {Q, P }R = 0.
8
8
Taking ((i)(ii))3, we have
1
1
1
(P R)Q (Q R)P + {Q, R}P {P, R}Q {P, Q}R = 0.
8
8
4
We define a space MC , called the complex Freudenthal manifold, by
MC = {P PC | P P = 0, P 6= 0}

(
)
P = (X, Y, , ) PC X Y = 0, (X, Y ) = 3,
=
.

X X = Y, Y Y = X
P 6= 0
Lemma 4.1.2. The following elements (assuming 6= 0, =

6 0) of PC
1

X
(Y Y )
0
0
1 (X X)

, 1 =
,
, 1. =
1
2 det X
1
0
1
det Y
2
107
(i)
(ii)
belong to MC .
Proof. This is clear from X (X X) = 0 (Lemma 3.5.4.(1)) and (X X)
(X X) = (det X)X (Lemma 2.1.1.(3)).
Definition. We define the groups E7 C and E7 respectively by
E7 C = { IsoC (PC ) | (P Q)1 = P Q},
E7 = { IsoC (PC ) | (P Q)1 = P Q, hP, Qi = hP, Qi}.
E7 is a subgroup of E7 C .
Theorem 4.2.1. E7 is a compact Lie group.
U (56) = U (PC ) = { IsoC (PC ) | hP, Qi = hP, Qi}.
Proposition 4.2.2. For E7 C (and so, for E7 ), we have
(1) MC = MC .
(2) {P, Q} = {P, Q},
P, Q PC .
Proof. (1) It is sufficient to prove that MC MC . Now, for P MC , we have

P P = (P P )1 = 01 = 0. Hence P MC .
8
(2) {P, Q}P = ((P P )Q (P Q)P ) (Lemma 4.1.1.(2))
3
8
= ((P P )Q (P Q)P ) = {P, Q}P.
3
It follows {P, Q} = {P, Q}.
Before we investigate the Lie algebra e7 of the group E7 , we shall first study the
Lie algebra e7 C of the group E7 C .
Theorem 4.3.1. The Lie algebra e7 C of the group E7 C is given by
e7 C = {(, A, B, ) HomC (PC ) | e6 C , A, B JC , C}.
The Lie bracket [ , ] in e7 C is given by
[(1 , A1 , B1 , 1 ), (2 , A2 , B2 , 2 )] = (, A, B, ),
where
108

= [1 , 2 ] + 2A1 B2 2A2 B1

2
2
A = 1 + 1 A2 2 + 2 A1
3
3 2

2
B = 1 + 1 B2 + t 2 + 2 B1
3
3
= (A1 , B2 ) (B1 , A2 ).
In particular, the dimension of e7 C is
dimC (e7 C ) = 78 + 27 2 + 1 = 133.

Proof. Before we show that e7 C is the Lie algebra of the group E7 C , we first
check the form of the Lie bracket in e7 C . For i e7 C and P PC , we have
[1 , 2 ]P = 1 2 P 2 1 P
= (using the formula of A B (Lemma 3.4.1) etc.)
= P.
This is that of the theorem.
We now determine the Lie algebra e7 C of the group E7 C . Since E7 C satisfies
(
P P = 0,
P MC
{P, Q} = {P, Q}, P, Q PC (Proposition 4.2.2.(2)),
if HomC (PC ) belongs to e7 C , then satisfies

(
P P = 0,
P MC
{P, Q} + {P, Q} = 0, P, Q P .
(i)
(ii)
Since e7 C is a C-linear mapping of PC = JC JC C C, is of the form
g
k
=
c
b
l
h
a
d
C
B
A
D
,
g, h, k, l HomC (JC ),
a, b, c, d HomC (JC , C),
A, B, C, D JC ,
, , , C.
For 0 6= r C, we define a C-linear isomorphism fr : PC PC by

fr (X, Y, , ) = (X, rY, r2 , r1 ),
then it is easy to see that fr satisfies
rfr (P Q)fr 1 = fr P fr Q,
P, Q PC .
Hence we see that for E7 C we have fr fr 1 E7 C . Therefore, for e7 C , we

have
109
e7 C fr fr 1
g
rk
= 2
r c
r1 b
r1 l r2 C
h
r1 B
ra
r2 d r3 k
for any 0 6= r C. Hence is decomposed as
rA
r2 D
3
r
i e7 C ,
= 3 + 2 + 1 + 0 + 1 + 2 + 3 ,
where
3
0
0
=
0
0
0
0
=
0
0
0
0
=
0
b
0
0
0
0
, 3 =
0
0
0
0
0
0
0
0
, 2 =
0
c
0
0
0
0
0
k
, 1 =
0
0
0
0
g 0 0 0
0 h 0 0
0 =
.
0 0 0
0 0 0
0 0
0 0
0 0
0
0 C
0 0
0 0
d 0
l 0
0 B
0 0
0 0
0
0
0
0
0
0
0
0
0
0
a
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
,
0
D
,
0
0
A
0
,
0
0
The relation 3 (0, 0, 1, 0) (0, 0, 1, 0) = 0 implies = 0, hence 3 = 0. Similarly

3 = 0. The relation 2 (0, 0, 1, 0) (0, 0, 1, 0) = 0 implies C = 0. Moreover, the
relation 2 P P = 0 for P = (Y Y, Y, 1, det Y ) MC , that is, (0, 0, 0, d(Y ))
(Y Y, Y, 1, det Y ) = 0 implies d = 0. Hence 2 = 0. Similarly 2 = 0. The relation
1 P P = 0 for P = (Y Y, Y, 1, det Y ) MC , that is, (l(Y ), B, 0, b(Y Y ))
(Y Y, Y, 1, det Y ) = 0 implies
l(Y ) = 2B Y,
Y JC .
Next, the relation 1 P P = 0 for P = (X, X X, det X, 1) MC , that is,

(l(X X), (det X)B, 0, b(X)) (X, X X, det X, 1) = 0, (from the 4-th condition)
we have
2(B (X X), X X) + (det X)(X, B) = 3(det X)b(X),
hence, 3(det X)(B, X) = 3(det X)b(X), and so we have
b(X) = (B, X),
X JC .
(Since b is continuous, the above is also valid for X JC such that det X = 0).
Similarly, using 1 , we have
110
k(X) = 2A X,
a(Y ) = (A, Y ),
X, Y JC .
The relation 0 P P = 0 for P = (X, X X, det X, 1) MC , that is, (g(X), h(X

X), (det X), ) (X, X X, det X, 1) = 0 implies
2g(X) X = X X + h(X X),
2h(X X) (X X) = (det X)(X + g(X)),
(g(X), X X) + (h(X X), X) = 3( + )det X.
(i)
(ii)
(iii)
1
Putting = g ( + 2)1, then, using (i) and (iii), we have
3
3(X, X, X) = 3(g(X), X, X) ( + 2)(X, X, X)
= (X X + h(X X), X) + (g(X), X X) 3( + 2)det X
= 3det X + 3( + )det X 3( + 2)det X = 0.
Therefore
e6 C .
1
Similarly = h (2 + )1 e6 C . Furthermore, from (ii), we have
3

1
2 (X X) + (2 + )(X X) (X X)
3

1
= (detX) X + X + ( + 2)X ,
3
so 2(X X) (X X) = (det X)X, and so (for a while, instead of t , we use

again) ((X X) (X X)) = (det X)X (Lemma 3.4.3.(1)), hence (det X) X =
(det X)X. Therefore, we have X = X, X JC , (even if for X JC such that
det X = 0), that is,
= .
Finally, the relation {(0, 0, 1, 0), (0, 0, 0, 1)} + {(0, 0, 1, 0), (0, 0, 0, 1)} = 0 implies
+ = 0. Thus we see that e7 C is of the form
1

2B
0 A
2A
B
0
=
0
A
0
B
0
0
C
= (, A, B, ), e6 , A, B JC , C.
Conversely, we shall show that = (, A, B, ), e6 C , A, B JC , C belongs
to e7 C , that is, exp t E C for all t C. For this purpose, we prove the following
proposition.
111
Proposition 4.3.2. For = (, A, B, ), e6 C , A, B JC , C and

P, Q PC , we have
[, P Q] = P Q + P Q.
Proof. It is sufficient to show that [, P P ] = 2P P. For P = (X, Y, , )
P , we have
C
[, P P ]
h

i
1
1
1
= (, A, B, ), X Y, (Y Y X), (X X Y ), ((X, Y ) 3)
2
2
4
= (using (X Y ) = X Y + X Y (Lemma 3.4.3.(1)), [, A B]
= A B + A B (Lemma 3.4.4.(1)), the formula about A B (Lemma
3.4.1) etc.)
= 2P P.
We will now return to the proof of Theorem 4.3.1. For = (, A, B, ), e6 C ,
A, B JC , C and t C, we have
(exp t)(P Q)(exp t)1
= (exp t(ad))(P Q) ((ad)1 = [, 1 ], 1 e7 C )
X
1
(t(ad))n (P Q)
=
n!
n=0
n X

X
t
n! k
=
P l Q (Proposition 4.3.2)
n!
k! l!
n=0
k+l=n
X

X
tk tl k
=
P l Q
k! l!
n=0
k+l=n
X
X

1
1
=
(t)k P
(t)l Q
k!
l!
k=0
l=0
= (exp t)P (exp t)Q.
Hence exp t E7 C , so that e7 C . Thus the proof of Theorem 4.3.1 is completed.

Definition. We define a C-linear transformation of PC by
(X, Y, , ) = (Y, X, , ).
For HomC (PC ), we also denote by t the transpose of with respect to the
inner product (P, Q): (t P, Q) = (P, Q).
Lemma 4.3.3. (1) E7 and satisfies 2 = 1.
(2) For P, Q PC , we have
(P, Q) = {P, Q} = {P, Q},

112
hP, Qi = { P, Q}.
(3) For E7 C , we have
t 1
= 1 .
(4) For E7 C , we have

E7
if and only if
= .
(5) For (, A, B, ) e7 C , we have

(, A, B, )1 = (t , B, A, ).
Proof. (1) and (2) are evident.
(3) (P, Q) = {P, Q} = {P, Q} (Proposition 4.2.2.(2)) = (P, Q) (Lemma
4.3.3) = (P, t Q) implies = t , hence t 1 = 1 .
(4) If E7 C satisfies = , then hP, Qi = { P, Q} = { P, Q}
= { P, Q} (Proposition 4.2.2.(2)) = hP, Qi (Lemma 4.3.3), P, Q PC . Hence,
E7 . The converse implication can also be easily proved.
(5) is easily checked by direct calculations.
Theorem 4.3.4. The Lie algebra e7 of the group E7 is given by

e7 = {(, A, A, ) | e6 , A JC , iR},
where
2A
(, A, A, ) =
0
A
2 A
1
+
3
A
0
0
A
The Lie bracket [1 , 2 ] in e7 is given by
0
.
[(1 , A1 , A1 , 1 ), (2 , A2 , A2 , 2 )] = (, A, A, ),
where
= [1 , 2 ] 2A1 A2 + 2A2 A1

2
2
A = 1 + 1 A2 2 + 2 A1
3
3
= hA1 , A2 i hA2 , A1 i.
In particular, the dimension of e7 is
dim(e7 ) = 78 + 54 + 1 = 133.
Proof. For e7 C ,
e7
if and only if 1 = .
113
Therefore the theorem follows from Theorem 4.3.1, Lemma 4.3.3.(5) and (, A,
B, ) = ( , A, B, ).
Proposition 4.3.5. The complexification of the Lie algebra e7 is e7 C .
Proof. For e7 C , the conjugate transposed mapping of with respect to
the inner product hP, Qi of PC is = e7 C , and for e7 C , belongs to
e7 if and only if = . Now, any element e7 C is represented as
=

+
+i
,
2
2i
+
,
e7 .
2
2i
Hence e7 C is the complexification of e7 .

Theorem 4.4.1. The Lie algebra e7 C is simple and so e7 is also simple.
Proof. We use the decomposition of e7 C of Theorem 4.3.1
e7 C = e6 C NC ,
where e6 C = {(, 0, 0, 0) e7 C | e6 C } and NC = {(0, A, B, ) e7 C | A, B
JC , C}. Let p : e7 C e6 C and q : e7 C NC be projections of e7 C = e6 C NC .
Now, let a be a non-zero ideal of e7 C . Then p(a) is an ideal of e6 C . Indeed, if p(a),
then there exists (0, A, B, ) NC such that (, A, B, ) a. For any 1 e6 C ,
we have
a [(1 , 0, 0, 0), (, A, B, )] = ([1 , ], 1 A, 1 B, 0)
(Theorem 4.3.1), hence [1 , ] p(a).
We shall show that either e6 C a 6= {0} or NC a 6= {0}. Assume that e6 C a =

{0} and NC a = {0}. Then the mapping p|a : a e6 C is injective because NC a =
{0}. Since p(a) is a non-zero ideal of e6 C and e6 C is simple, we have p(a) = e6 C . Hence
dimC (a) = dimC (p(a)) = dimC (e6 C ) = 78. On the other hand, since e6 C a = {0},
q|a : a NC is also injective. Hence we have dimC (a) dimC (NC ) = 27+27+1 = 55.
This leads to a contradiction.
(1) Case e6 C a 6= {0}. From the simplicity of e6 C , we have e6 C a = e6 C , hence
a e6 C . On the other hand, we have
a [a, e7 C ] [(e6 C , 0, 0, 0), (0, JC , 0, 0)]
= (0, e6 C JC , 0, 0) = (0, JC , 0, 0) (Proposition 3.3.2.(3)).
Similarly we have (0, 0, JC , 0) a. Moreover, from

a [(0, E1 , 0, 0), (0, 0, E1 , 0)] = (2E1 E1 , 0, 0, 1),
114
we have (0, 0, 0, 1) a, and so a NC . Hence a e6 C NC = e7 C .
(2) Case NC a 6= {0}. Let (0, A, B, ) be a non-zero element of NC a.
(i) Case (0, A, B, ), A 6= 0. Choose B1 JC such that A B1 6= 0 (Lemma

3.5.4.(2)), and choose e6 C such that [A B1 , ] 6= 0 (since e6 C is simple, such a
exists because the center of e6 C is zero). Now, we have
h

3 i
a (0, A, B, ), 0, 0, 0,
= (0, A, B, 0),
2
a [(0, A, B, 0), (0, 0, B1 , 0)] = (2A B1 , 0, 0, (A, B1 )),
a [(2A B1 , 0, 0, (A, B1 )), (, 0, 0, 0)] = (2[A B1 .], 0, 0, 0),
(ii) Case (0, A, B, ), B 6= 0. This case is also reduced to the case (1) in a
similar way to (i).
(iii) Case (0, 0, 0, ), 6= 0. If we choose 0 6= A JC , then we have
2

a [(0, 0, 0, ), (0, A, 0, 0)] = 0, A, 0, 0 .
3
Hence this case is also reduced to the case (1).

Thus we have a = e7 C .
Proposition 4.4.2. (1) PC is an e7 C -irreducible C-module.

nX
o
(2) e7 C PC =
k Pk | k e7 C , Pk PC = PC .
k
Proof. (1) Let W be a non-zero e7 C -invariant C-submodule of PC . We first prove

that if (0, 0, 0, 1) W , then we have W = PC . Indeed,
W
W
W
W
(0, X, 0, 0)(0, 0, 0, 1) = (X, 0, 0, 0),

(0, E2 , 0, 0)(E3 , 0, 0, 0) = (0, E1 , 0, 0),
(0, E1 , 0, 0)(0, E1 , 0, 0) = (0, 0, 1, 0),
(0, 0, Y, 0)(0, 0, 1, 0) = (0, Y, 0, 0).
Hence we have W = PC . Now, let P = (X, Y, , ) be a non-zero element of W .

(i) Case W P = (X, Y, , ), X 6= 0. Then we have
(a)
W (0, 0, 0, 3)(X, Y, , ) = (X, Y, 3, 3),
(b)
W (0, 0, 0, 3)(X, Y, 3, 3) = (X, Y, 9, 9).
(c)
Taking ((a)(b))2, ((a)(c))8, we have (X, 0, , 2) W , (0, 0, , ) W , respectively. Consequently (X, 0, 0, 3) W . Next, if we choose X1 JC such that
(X1 , X) 6= 0, from
W (0, 0, X1 , 0)(X, 0, 0, 3) = (0, 0, 0, (X1 , X)),
115
we have (0, 0, 0, 1) W . Hence this case is reduced to the first case.
(ii) Case P = (0, Y, , ), Y 6= 0. If we choose B JC such that B Y 6= 0, from

W (0, 0, B, 0)(0, Y, , ) = (2B Y, B, 0, 0).
Hence this case is reduced to the case (i).

(iii) Case P = (0, 0, , ), 6= 0. For 0 6= B JC , we have
W (0, 0, B, 0)(0, 0, , ) = (0, B, 0, 0).
Hence this case is also reduced to the case (ii).
Thus we have W = PC .
(2) Since e7 C PC is an e7 C -invariant C-submodule of PC , e7 C PC = PC follows
from the irreducibility of PC (above (1)).
X
(Pi Qi ), Pi , Qi
Lemma 4.4.3. Any element e7 C is expressed by =
i
PC .
nX
(Pi
Proof. Since [, P Q] = P Q + P Q (Proposition 4.3.2), a =
i
o
Qi ) | Pi , Qi PC is an ideal of e7 C . From the simplicity of e7 C (Theorem 4.4.1), we
have a = e7 C .

Definition. We define a symmetric inner product (1 , 2 )7 in e7 C by
8
(1 , 2 )7 = 2(1 , 2 )6 4(A1 , B2 ) 4(A2 , B1 ) 1 2 ,
3
where i = (i , Ai , Bi , i ) e7 C .
Lemma 4.5.1. (1) The inner product (1 , 2 )7 of e7 C is e7 C -adjoint invariant :
([, 1 ], 2 )7 + (1 , [, 2 ])7 = 0,
, i e7 C .
(2) For e7 C , P, Q PC , we have

(, P Q)7 = {P, Q}.
Proof . (1) ([, 1 ], 2 )7
[, 1 ] + 2A B1 2A1 B
2

2
2
!
+ A1 1 + 1 A
A
2
3
3
=

,
2
2
t + B1 + t 1 + 1 B B2 7
3
3
2
(A, B1 ) (B, A1 )
116
= (using ([, A B] = A B + A B (Lemma 3.4.4) etc.)

= (1 , [, 2 ])7 .
(2) For P = (X, Y, , ), Q = (Z, W, , ) PC , we have
1
(X W + Z Y )

2
(2Y
X)

4
(, P Q)7 = ,
B
(2X Z W Y )
7
1
((X, W ) + (Z, Y ) 3( + ))
8
= (, X W + Z Y )6 (A, 2X Z W Y ) + (2Y W Z X, B)
1
((X, W ) + (Z, Y ) 3( + ))
3
= (X, W ) + (Z, Y ) 2(A, X, Z) + (A, W ) + (A, Y ) + 2(Y, W, B)
1
1
(Z, B) (X, B) (X, W ) (Z, Y ) + ( + )
3
3
1

X X + 2B Y + A
Z
3
(
)
1
W
2A X t Y + Y + B ,
=
= {P, Q}.
(A, Y ) +
(B, X)

B7 (1 , 2 ) = 9(1 , 2 )7
= 18(1 , 2 )6 + 36(A1 , B2 ) + 36(A2 , B1 ) + 241 2

3
= B6 (1 , 2 ) + 36(A1 , B2 ) + 36(A2 , B1 ) + 241 2
2
= 3tr(1 2 ),
where i = (i , Ai , Bi , i ) e7 C and B6 is the Killing form of e6 C .

Proof. Since e7 C is simple (Theorem 4.4.1), there exist k, k C such that
B7 (1 , 2 ) = k(1 , 2 )7 = k tr(1 2 ).
To determine these k, k , let 0 = 1 = 2 = (0, 0, 0, 1). Then we have
8
(0 , 0 )7 = .
3
Hence
h
2
i
4 4

2
[0 , [0 , (, A, B, )] ] = 0 , 0, A, B, 0 = 0, A, B, 0 .
3
3
9 9
B7 (0 , 0 ) = tr((ad0 )2 ) =
117
4
27 2 = 24.
9
Therefore k = 9. Next, from
we have
X Y
X Y

0 0 (X, Y, , ) = 0 , , , =
, , , ,
3 3
9 9
tr(02 ) =
1
27 2 + 1 + 1 = 8.
9
Therefore k = 3.
Lemma 4.5.3. For P PC , P 6= 0, there exists Q PC such that P Q 6= 0.
Proof. Asumme that P Q = 0 for all Q PC . Then for any e7 C ,
0 = (, P Q)7 = (, Q P )7 = {Q, P } (Lemma 4.5.1.(2)). Since e7 C PC = PC
(Proposition 4.4.2.(2)), we have {PC , P } = 0, so that P = 0.
4.6. Roots of e7 C
Theorem 4.6.1 The rank of the Lie algebra e7 C is 7. The roots of e7 C relative to
some Cartan subalgebra h are given by
(k l ), (k + l ), 0 k < l 3,
1
k (2 3 ),
0 k 3,
2
1
1
(0 1 + 2 3 ) (3 1 ),
2
2
1
1
( 0 + 1 + 2 3 ) (3 1 ),
2
2
1
1
(0 + 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 3 ) (1 2 ),
2
2
1
1
(0 + 1 + 2 3 ) (1 2 ),
2
2
1
1
( 0 + 1 + 2 + 3 ) (1 2 ),
2
2
1
1
(0 1 + 2 + 3 ) (1 2 ),
2
2

2
j + , 0 j 3,
1 3 2
k
1 , 0 k 3,
2
3
1
1
2
(0 1 + 2 3 )
2 ,
2
3
21
1
2
( 0 + 1 + 2 3 )
2 ,
2
2
3
118
1
1
2
(0 + 1 + 2 + 3 )
2 ,
2
3
21
1
2
( 0 1 + 2 + 3 )
2 ,
2
2
3
1
2
1
3 ,
( 0 1 + 2 3 )
2
3
12
1
2
(0 + 1 + 2 3 )
3 ,
2
2
3

1
1
2
( 0 + 1 + 2 + 3 )
3 ,
2
2
3
1
1
2
(0 1 + 2 + 3 )
3
2
2
3
with 1 + 2 + 3 = 0.
Proof. We use the decomposition of of e7 C in Theorem 4.3.1
e7 C = e6 C JC JC C.
Let
3
3

X
n X
j Ej , 0, 0, e7 C
k Hk +
h=
k=0
j=1

o
k , C

j C, 1 + 2 + 3 = 0
(where Hk = iGk4+k ), then h is an abelian subalgebra of e7 C (it will be a Cartan sub3

3
X
X
j Ej .
algebra of e7 C ). In the following calculations, we put h =
k Hk , H =
k=0
j=1
I The roots of e6 C are also roots of e7 C . Indeed, let be a root of e6 C and

S e6 C e7 C be a root vector belonging to . Then
e 0, 0, ), (S, 0, 0, 0)]
[h, S] = [(h + H,
e S], 0, 0, 0) = ((h + H)S,
e
= ([h + H,
0, 0, 0) = (h)S.
Hence is a root of e7 C .
e 0, 0, ), (0, Ej , 0, 0)]
II [(h + H,

e + 2 Ej , 0, 0 = j + 2 (0, Ej , 0, 0),
= 0, h + H
3
3

e 0, 0, ), (0, 0, Ej , 0)] = 0, 0, (h + H)
e 2 Ej , 0
[((h + H,
3

2
2
e
= 0, 0, h H Ej , 0 = j (0, 0, Ej , 0).
3
3

2
C
Hence j + , 0 j 3, are roots of e7 .
3
e 0, 0, ), (0, F1 (a), 0, 0)]
III [(h + H,
a = ek ie4+k

2
e + F1 (a), 0, 0
= 0, h + H
3

2
1
= 0, F1 (h a) + (2 + 3 )F1 (a) + F1 (a), 0, 0
2
3
119
1
2
k 1 + (0, F1 (a), 0, 0).
2
3
2
1
Hence k 1 + , 0 k 3, are roots of e7 C . The remainders of roots can be
2
3
similarly found (h , h in Theorem 3.6.4 will be used).
=

1 = 0 1 , 2 = 1 2 , 3 = 2 3 ,
1
1
4 = (0 1 2 + 3 ) + (3 1 ),
2
2
1
1
5 = (0 + 1 + 2 + 3 ) + (1 2 ),
2
2
3
3
6 = 2 + , 7 = 3
2
2
= 1 + 22 + 33 + 44 + 35 + 26 + 27
1
7
1
4
4
2

+ n7 7 . Now, all positive roots of e7 C are represented by
0 1
0 2
0 3
1 2
1 3
2 3
=1
=1
=1
=0
=0
=0
0
1
1
1
1
0
0
0
1
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 + 1
0 + 2
0 + 3
1 + 2
1 + 3
2 + 3
1
0 + (2 3 ) = 1
2
1
1 + (2 3 ) = 0
2
1
2 + (2 3 ) = 0
2
1
3 + (2 3 ) = 0
2
120
=1
=1
=1
=0
=0
=0
2
1
1
1
1
0
2
2
1
2
1
1
1 1
1 1
1 1
1 1
1 1
1 1
0 1
1 1
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
0 (2 3 ) = 1 1 1 1
2
1
1 (2 3 ) = 0 1 1 1
2
1
2 (2 3 ) = 0 0 1 1
2
1
3 (2 3 ) = 0 0 0 1
2
1
1
(0 1 + 2 3 ) + (3 1 ) = 0
2
2
1
1
( 0 + 1 + 2 3 ) + (3 1 ) = 1
2
2
1
1
(0 + 1 + 2 + 3 ) + (3 1 ) = 0
2
2
1
1
( 0 1 + 2 + 3 ) + (3 1 ) = 1
2
2
1
1
( 0 + 1 2 + 3 ) + (3 1 ) = 1
2
2
1
1
(0 1 2 + 3 ) + (3 1 ) = 0
2
2
1
1
( 0 1 2 3 ) + (3 1 ) = 1
2
2
1
1
(0 + 1 2 3 ) + (3 1 ) = 0
2
2
1
1
( 0 1 + 2 3 ) (1 2 ) = 1
2
2
1
1
( 0 1 2 + 3 ) (1 2 ) = 1
2
2
1
1
( 0 + 1 + 2 + 3 ) + (1 2 ) = 0
2
2
1
1
( 0 + 1 2 3 ) (1 2 ) = 1
2
2
1
1
(0 + 1 2 + 3 ) (1 2 ) = 0
2
2
1
1
(0 + 1 + 2 3 ) (1 2 ) = 0
2
2
1
1
( 0 + 1 + 2 + 3 ) (1 2 ) = 1
2
2
1
1
(0 1 + 2 + 3 ) (1 2 ) = 0
2
2
2
1 = 1 2 3 4 2
3
2
2 + = 0 0 0 0 0
3
2
3 = 0 0 0 0 0
3
121
1 0
1 0
1 0
1 0
0 1
1 0
0 0
2 3
3 2
1 1
1 2
3 2
1 1
1 2
3 2
1 1
2 2
3 2
1 1
0 0
1 0
0 0
1 1
1 0
0 0
1 1
1 0
0 0
1 2
2 1
1 1
1 1
2 1
1 1
0 0
0 1
0 0
2 2
2 1
1 1
1 1
2 1
1 1
1 2
2 1
1 1
2 3
4 3
2 2
0 1
2 1
1 1
1 2
1 0
0 1
1
0 1 +
2
1
1 1 +
2
1
2 1 +
2
1
3 1 +
2
1
0 + 1
2
1
1 + 1
2
1
2 + 1
2
1
3 + 1
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
=1 1
1 1
1 1
=0 1
1 1
1 1
=0 0
1 1
1 1
=0 0
0 1
1 1
=1 1
1 1
1 0
=0 1
1 1
1 0
=0 0
1 1
1 0
=0 0
0 1
1 0
0 1
1 0
0 1
2 3
3 2
1 2
1 2
3 2
1 2
1 2
3 2
1 2
2 2
3 2
1 2
0 0
1 0
0 1
1 1
1 0
0 1
1 1
1 0
0 1
1 2
2 1
0 1
1 1
2 1
0 1
0 0
0 1
1 0
2 2
2 1
0 1
1
2
1
(0 1 + 2 3 ) + 2 = 0
2
2
3
1
2
1
( 0 + 1 + 2 3 ) + 2 = 1
2
2
3
1
2
1
(0 + 1 + 2 + 3 ) + 2 = 0
2
2
3
1
2
1
( 0 1 + 2 + 3 ) + 2 = 1
2
2
3
1
1
2
( 0 + 1 2 + 3 ) + 2 = 1
2
2
3
1
2
1
(0 1 2 + 3 ) + 2 = 0
2
2
3
1
2
1
( 0 1 2 3 ) + 2 = 1
2
2
3
1
1
2
(0 + 1 2 3 ) + 2 = 0
2
2
3
1
2
1
( 0 1 + 2 3 ) + 3 = 1
2
2
3
1
2
1
( 0 1 2 + 3 ) + 3 + = 1
2
2
3
1
2
1
( 0 + 1 + 2 + 3 ) 3 = 0
2
2
3
1
2
1
( 0 + 1 2 3 ) + 3 = 1
2
2
3
1
1
2
(0 + 1 2 + 3 ) + 3 = 0
2
2
3
1
2
1
(0 + 1 + 2 3 ) + 3 = 0
2
2
3
1
2
1
( 0 + 1 + 2 + 3 ) + 3 = 1
2
2
3
1
2
1
(0 1 + 2 + 3 ) + 3 = 0
2
2
3
122
1 2
1 0
2 2
1 0
3 4
3 1
1 2
1 0
1.
Hence = {1 , 2 , , 7 } is a fundamental root system of e7 C . The real part hR

of h is
hR
3
3

n X
X
=
j Ej , 0, 0, e7 C
k Hk +
j=1
k=0
and the Killing form B7 of e7 C on hR is given by

o
k , R

j R, 1 + 2 + 3 = 0
3
3

X
X
j j + 4
B7 (h, h ) = 6 6
k k + 3
j=1
k=0
3
3
3
3
X

X
X

X
for h =
j Ej , 0, 0, , h =
k Hk +
j Ej , 0,
k Hk +
0,
j=1
k=0
j=1
k=0
hR . Indeed, from Theorem 4.5.2, we have

X
X
X
3 X
+ 24
j Ej
k Hk +
j Ej ,
B6
k Hk +
2
j=1
j=1
3
B7 (h, h ) =
k=0
k=0
3
3

X
3 X
= 12 2
j j + 24 (Theorem 3.6.5)
k k +
2
j=1
k=0
3
3

X
X
j j + 4 .
= 6 6
k k + 3
k=0
j=1

h) are determined as follows.
H1 =
H2 =
H3 =
H4 =
H5 =
H6 =
H7 =
Thus we have
1
(H0 H1 , 0, 0, 0),
36
1
(H1 H2 , 0, 0, 0),
36
1
(H2 H3 , 0, 0, 0),
36
1
((H0 H1 H2 + H3 ) + 2(E3 E1 ) , 0, 0, 0),
72
1
((H0 + H1 + H2 + H3 ) + 2(E1 E2 ) , 0, 0, 0),
72
1
3
(E1 + 2E2 E3 ) , 0, 0, ,
54
2
1
3
(E1 + E2 2E3 ) , 0, 0, .
54
2
1 1
1
2=
36 36
18
and the other inner products are similarly calculated. Consequently, the inner product
(1 , 1 ) = B7 (H1 , H1 ) = 36
123
(i , i ) =
1
,
18
i = 1, 2, 3, 4, 5, 6, 7,
(1 , 2 ) = (2 , 3 ) = (3 , 4 ) = (4 , 5 ) = (4 , 7 ) = (5 , 6 ) =
(i , j ) = 0,
(, ) =
1
,
36
otherwise,
1
,
18
(, 6 ) =
1
,
36
(, i ) = 0, i = 1, 2, 3, 5, 7,
e7 C .
According to Borel-Siebenthal theory, the Lie algebra e7 has four subalgebras as
(1) The first one is a subalgebra of type T E6 which is obtained as the fixed
(2) The second one is a subalgebra of type A1 D6 which is obtained as the fixed
(3) The third one is a subalgebra of type A7 which is obtained as the fixed points
of an involution of e7 .
(4) The fourth one is a subalgebra of type A2 A5 which is obtained as the fixed
points of an automorphism w of order 3 of e7 .
4.10.2, 4.11.15, 4.12.5 and 4.13.5, respectively.
4.7. Subgroups E6 and U (1) of E7
We shall study the following subgroup (E7 )(0,0,1,0) of E7 :
(E7 )(0,0,1,0) = { E7 | (0, 0, 1, 0) = (0, 0, 1, 0)}.
Lemma 4.7.1. If E7 satisfies (0, 0, 1, 0) = (0, 0, 1, 0), then also satisfies
(0, 0, 0, 1) = (0, 0, 0, 1), and conversely.
Proof. If E7 satisfies (0, 0, 1, 0) = (0, 0, 1, 0), then we have
(0, 0, 0, 1) = ( (0, 0, 1, 0)) = (0, 0, 1, 0) = (0, 0, 1, 0) = (0, 0, 0, 1).
The converse can be similarly proved.
Theorem 4.7.2.
(E7 )(0,0,1,0)
= E6 .
Proof. We associate an element E6 = { IsoC (JC ) | det (X) = det X, hX,

Y i = hX, Y i} with the element
124
0
0
e=
0
0
0
0
0
0
1
0
0
0
(E7 )(0,0,1,0) E7 .
0
1
We first have to prove that

e E7 . For P = (X, Y, , ), Q = (Z, W, , ) PC , we
have
eP
eQ = (X, Y, , ) (Z, W, , )
= (using (X Y )1 = X Y, (1 ) = ( ) ( 1 ) etc.)
=
e(P Q)e
1
and he
P,
eQi = hP, Qi is evident. Hence
e E7 , moreover
e (E7 )(0,0,1,0) .
Conversely, suppose that E7 satisfies (0, 0, 1, 0) = (0, 0, 1, 0) and (0, 0, 0, 1) =

(0, 0, 0, 1) (Lemma 4.7.1). Then is of the form
=
0
0
1
0
0
0
0
1
0
0
0
,
0
1
, 1 , , HomC (JC ).
Indeed, the fact that the left bottom parts are 0 follows from
1i
1i
1i
= hX,
= hX,
= 0,
hX,
1.i = hX,
1.i = hX,
1.i = 0.
hX,
Now, since
1
X + (X X)
X X
1
X
+
(X
X)
1
C
,
=
M
1 detX
1

2
det
X
we can see that

1
1
1
X + (X X) X + (X X) = X + 1 (X X)
holds for all 0 6= C. Comparing the coefficients of of both sides, we have = 0.

1
1
Similarly, from (Y Y ), Y, , 2 det Y MC , we have = 0. Furthermore, since
MC (X, X X, det X, 1) = (X, 1 (X X), det X, 1),

we have
X X = 1 (X X),
(X, 1 (X X)) = 3det X.

125
and so
1
1
(X, X X) = (X, 1 (X X)) = det X,
3
3
C
Y i = hX,
Y i implies hX, Y i =
which implies that E6 . The equality hX,
hX, Y i and therefore E6 . Moreover from the relation
det (X) =
1 (X X) = X X = (X X),
we obtain 1 = . Indeed, putting X X instead of X, we have
(det X)1 X = (det X) X.
If det X 6= 0, then we have 1 X = X, Since 1 and are linear mappings (of
course are continuous mappings), we have 1 X = X even if det X = 0. Thus, the
proof of Theorem 4.7.2 is completed.
For C, 6= 0, we define a C-linear transformation 1 () : PC PC by
1 ()(X, Y, , ) = (1 X, Y, 3 , 3 ).
Then 1 () E7 C .
Theorem 4.7.3. The group E7 contains
U (1) = {1 () | C, ( ) = 1}
as a subargroup. This subgroup is isomorphic to the usual unitary group U (1) = {
C | ( ) = 1}.
Proof. It is easy to check that 1 () E7 .
We denote by (E7 )0 the connected component of E7 containing the identity 1.
by
Lemma 4.8.1. For a C, if we define a mapping i (a) : PC PC , i = 1, 2, 3
1 + (cos |a| 1)pi
2 a sin |a| Ei
|a|
i (a) =
sin |a|
Ei
a
|a|
2a
sin |a|
Ei
|a|
1 + (cos |a| 1)pi

a
sin |a|
Ei
|a|
0
0
a
sin |a|
Ei
|a|
cos |a|
0
sin |a|
Ei
|a|
cos |a|

sin |a|
if a = 0, then
means 1 , then i (a) (E7 )0 , where pi : JC JC is the
|a|
C-linear mapping defined by
126
1
pi x3
x2
x3
2
x1

1
x2
x1 = i3 x3
3
i2 x2
i3 x3
2
i1 x1
i2 x2
i1 x1 ,
3
where ij is the Kronecker delta symble. The mappings 1 (a1 ), 2 (a2 ), (a3 ), (ai C)
are commutative for each other.
Proof. For
0
2 aEi
i (a) = (0, aEi , aEi , 0) =
0
aEi
2aEi
0
aEi
0
0
aEi
0
0
aEi
0
e7
0
0
(Theorem 4.3.4), we have i (a) = exp i (a). Hence i (a) (E7 )0 . The relation
[i (ai ), j (aj )] = 0 shows that i (ai ) and j (aj ) are commutative.
Proposition 4.8.2. Any element P MC , P 6= 0 can be transformed to a
diagonal form by some element (E7 )0 :
P = (X, Y, , ),
X, Y are diagonal, > 0.
Proof. Let P = (X, Y, , ) MC . We shall first show that P can be transformed

to a diagonal form with 6= 0.
1
(1) Case P = (X, Y, , ), 6= 0. In this case, X = (Y Y ). Choose E6
such that Y is diagonal (Proposition 3.8.2), then

X =
1
1
(Y Y ) = Y Y
is also diagonal.
(2) Case P = (X, Y, 0, ), Y 6= 0. Choose
1 0
Y = 0 2
0 0
E6 so that
0
0 , i C
3
(Proposition 3.8.2). Since Y 6= 0, some i is non-zero : i 6= 0. Applying

i (/2) (E7 )0 of Lemma 4.8.1 on P , we get

X
1 pi 2Ei
0
Ei
1 pi Ei 0 Y
2Ei
i (/2)P =
= , i 6= 0,
0
Ei
0
0
0
i
Ei
0
0
0
so that this case is reduced to the case (1).
(3) Case P = (X, 0, 0, ), X 6= 0. Choose E6 so that X = 1 E1 +2 E2 +3 E3 ,

i C (Proposition 3.8.2). Since X 6= 0, some i is non-zero : i 6= 0. Then
i+1 (/2)P = (, i Ei+2 + i+2 Ei , 0, ),
127
i 6= 0,
so that this case is reduced to the case (2).

(4) Case P = (0, 0, 0, ), 6= 0. Then
1 (/2)P = (E1 , 0, 0, 0),
6= 0,
so that this case is also reduced to the case (3).

Consequently, any element P can be transformed to a diagonal form with 6= 0.
Furthermore, by applying some 1 () U (1) (E7 )0 of Theorem 4.7.3 on it, then
becomes > 0. Thus the proof of Proposition 4.8.2 is completed.
Remark. In Proposition 4.8.2, the condition P MC does not need. That is,
any element P PC can be transformed to a diagonal form by some E7 . (See
Miyasaka, Yasukura and Yokota [23]).
We define a space M1 , called the compact Freudenthal manifold, by
M1 = {P PC | P P = 0, hP, P i = 1}.
E7 /E6 M1 .
Theorem 4.8.3.
In particular, the group E7 is connected.
Proof. For E7 and P M1 , we have P M1 . Hence E7 acts on M1 . We

shall prove that the group (E7 )0 acts transitively on M1 . To prove this, it is sufficient
to show that any element P M1 can be transformed to (0, 0, 1, 0) M1 by some
(E7 )0 . Now, P M1 can be transformed to a diagonal form by (E7 )0 :

0
0
1 0 0

1 2 3
1
0
P =
3 1
0 , 0 2 0 , , 2 1 2 3 , > 0
0
0
1 2
0 0 3
(Proposition 4.8.2). From the condition hP, P i = hP, P i = 1, we have
1
1
(|2 3 |2 + |3 1 |2 + |1 2 |2 ) + (|1 |2 + |2 |2 + |3 |2 ) + 2 + 4 |1 2 3 |2 = 1,
2
that is,

1+
|1 |2
|2 |2
|3 |2
1
1
+
1
+
= 2.
2
2
2
Choose r1 , r2 , r3 R, 0 ri <
, such that
2
tan ri =
|i |
,
i = 1, 2, 3,
then (i) becomes

= cos r1 cos r2 cos r3 .
128
(i)
By letting
ai =
i
ri ,
|i |
i = 1, 2, 3
(if i = 0, then ai means 0), we have

ri = |ai |,
i =
ai
1 i |i |
ri
=
tan ri cos r1 cos r2 cos r3 .
|ai | |i |
|ai |
Therefore, we see that P is of the form
sin |a2 | sin |a3 |

a3
0
0
cos |a1 |a2
|a2 |
|a3 |
sin
|a
|
sin
|a
|
1
3
0
a1
cos |a2 |a3
0
|a
|
|a
|
1
3
sin
|a
|
sin
|a
|
1
2
0
0
a1
a2
cos |a3 |
|a1 |
|a2 |
sin |a1 |
0
0
a1 |a | cos |a2 | cos |a3 |

1

sin |a2 |

0
cos |a1 |a2
cos |a3 |
0

|a2 |

sin
|a
|
3
0
0
cos |a1 | cos |a2 |a3
|a3 |
cos |a1 | cos |a2 | cos |a3 |
sin |a1 | sin |a2 | sin |a3 |

a2
a3
a1
|a1 |
|a2 |
|a3 |
= 3 (a3 )2 (a2 )1 (a1 )(0, 0, 1, 0),
hence we have
1 (a1 )1 2 (a2 )1 3 (a3 )1 P = (0, 0, 1, 0).
This shows the transitivity of (E7 )0 . Since we have M1 = (E7 )0 (0, 0, 1, 0), M1 is
connected. Now, the group E7 acts transitively on M1 and the isotropy subgroup of
E7 at (0, 0, 1, 0) M1 is E6 (Theorem 4.7.2). Therefore we have the homeomorphism
E7 /E6 M1 . Finally, the connectedness of E7 follows from the connectedness of M1
and E6 .
Theorem 4.9.1. The center z(E7 ) of the group E7 is the cyclic group of order 2:
z(E7 ) = {1, 1}.
Proof. Let z(E7 ). From the commutativity with E6 E7 , we have
(0, 0, 1, 0) = (0, 0, 1, 0) = (0, 0, 1, 0). If we denote (0, 0, 1, 0) = (X, Y, , )
PC , then from (X, Y, , ) = (X, Y, , ), we have
X = X, Y = Y
129
for all E6 .
Hence X = Y = 0. Therefore, (0, 0, 1, 0) is of the form

(0, 0, 1, 0) = (0, 0, , ).
From the condition (0, 0, 1, 0) MC , we have = 0. Suppose = 0, then
(0, 0, 1, 0) = (0, 0, 0, ), 6= 0. Also from the commutativity with 1 () U (1) E7
(Theorem 4.7.3), we have
(0, 0, 0, 3 ) = 1 ()(0, 0, 0, ) = 1 ()(0, 0, 1, 0)
= 1 ()(0, 0, 1, 0) = (0, 0, 3 , 0) = (0, 0, 0, 3 ),
and so 3 = 3 for all . But this is a contradiction. Hence 6= 0, = 0,
that is, (0, 0, 1, 0) = (0, 0, , 0). Similarly we have (0, 0, 0, 1) = (0, 0, 0, ). Since
{(0, 0, 1, 0), (0, 0, 1, 0)} = 1, we have = 1, and therefore
(0, 0, 1, 0) = (0, 0, , 0),
(0, 0, 0, 1) = (0, 0, 0, 1 ).
Moreover, from the commutativity with E7 ,

(0, 0, 0, ) = (0, 0, , 0) = (0, 0, 1, 0)
= (0, 0, 1, 0) = (0, 0, 0, 1) = (0, 0, 0, 1).
Hence = 1 , that is, = 1. In the case = 1, we have E6 (Theorem 4.7.2),
so that z(E6 ) = {1, 1, 21} (Theorem 3.9.1), that is,

1
0
0 0
1
0 1 0 0
=
, = 1, or 2 .
0
0
1 0
0
0
0 1
Again from the commutativity with ,
(0, X, 0, 0) = ( X, 0, 0, 0) = (X, 0, 0, 0)
= (X, 0, 0, 0) = (0, X, 0, 0) = (0,
X, 0, 0),
for all X JC , which shows that = , hence = 1. Therefore = 1. In the

case = 1, we have z(E6 ), so that by the similar argument as above we have
= 1. Thus we have z(E7 ) = {1, 1}.
According to a general theory of compact Lie groups, it is known that the center
of the simply connected compact simple Lie group of type E7 is the cyclic group of
order 2. Hence the group E7 has to be simply connected. Thus we have the following
theorem.
Theorem 4.9.2. E7 = { IsoC (PC ) | (P Q)1 = P Q, hP, Qi =
hP, Qi} is a simply connected compact Lie group of type E7 .
4.10. Involution and subgroup (U (1) E6 )/Z 3 of E7
130
Definition. We define a C-linear transformation of PC by

(X, Y, , ) = (iX, iY, i, i).
Then, = 1 (i) U (1) E7 , 2 = 1 z(E7 ) (Theorem 4.9.1) and so 4 = 1.
Lemma 4.10.1. is conjugate to in E7 .
i i i
Proof. For = 1
2
3
(Lemma 4.8.1), we have = 1 .
4
4
4
induces an involutive automorphism e

: E7 E7 by
e
() = 1 ,
E7 .
We shall now study the following subgroup (E7 ) of E7 :

(E7 ) = { E7 | = }
= { E7 | = } = (E7 ) .
2
2

Theorem
4.10.2 (E7 ) = (U (1) E6 )/Z 3 , Z 3 = {(1, 1), (, 1), ( , 1)},
1
3
= +
i C.
2
2
Proof. We define a mapping : U (1) E6 (E7 ) by
(, ) = 1 (),
1 1 0
1
0
1 () =
0
0
0
0
0
0
3
0
0
0
, =
0
0
3
0
0

0
0
0
0
1
0
0
0
.
0
1
Evidently (, ) (E7 ) . Since 1 () and are commutaive, is a homomorphism.

We shall prove that is onto. Let (E7 ) . From = , is seen to be of the
form
, HomC (JC ),
0 M 0
0 0 N
a, b HomC (JC , C),
=
a 0 0 ,
M, N JC ,
0 b 0
, C.
The condition (0, 0, 1, 0), (0, 0, 0, 1) MC implies that
M = 0,
N = 0.
We shall first show that M = N = 0. Suppose that M 6= 0, = 0. Then, the

condition {(0, 0, 1, 0), (0, 0, 0, 1)} = {(0, 0, 1, 0), (0, 0, 0, 1)} = 1 implies that
(M, N ) = 1.
Hence we have N 6= 0, = 0. From
131
(i)
we have
1
1X X

C
M
=

1
2 detX
1
(detX)M
2
1
(X X) + N
,
a(X)
1
b(X X)
X +

1

1
(X X) + N
(X X) + N = a(X) X + 2 (detX)M ,

1
1
1
X + 2 (detX)M, (X X) + N = 3a(X) b(X X)
hold for all 0 6= C. Comparing the coefficients of , we have
2(X X) N = a(X)X
(X X) (X X) = a(X)(det X)M
(X, (X X)) + det X = 3a(X)b(X X) (use (i)).
(ii)
(iii)
(iv)
Therfore, using (i) (iv), we have
a(X)detX = a(X)(det X)(M, N ) = ((X X) (X X), N )

1
= ((X X), (X X) N ) = a(X)((X X), X)
2
1
= a(X)(3a(X)b(X X) det X).
2
Hence we have a(X)det X = a(X)2 b(X X). Furthermore we have
det X = a(X)b(X X).
(v)
Indeed, from = 0, we deduce that a 6= 0. Since a : JC C is a linear form, the

set {X JC | a(X) 6= 0} is dense in JC and the correspondence det X and b(X X)
is continuous with respect to X, (v) is also valid for X JC such that a(X) = 0.
Now, since a 6= 0 and b 6= 0, (v) contradicts the irreducibility of the determinant
det X with respect to the variables of its components. Consequently we have shown
that M = 0. Similarly we can prove that N = 0. Therefore (0, 0, 1, 0) = (0, 0, , 0),
1.} = 1, h1,
1i
= 1, we deduce
(0, 0, 0, 1) = (0, 0, 0, ). From the condition {1,
that
(0, 0, 1, 0) = (0, 0, , 0), (0, 0, 0, 1) = (0, 0, 0, 1 ),
C, ( ) = 1.
If we choose C such that 3 = and let = 1 ()1 , then (0, 0, 1, 0) =

(0, 0, 1, 0), (0, 0, 0, 1) = (0, 0, 0, 1). Hence, E6 (Theorem 4.7.2) and we have
= 1 (),
U (1), E6 .
132
This shows is onto. That Ker = {(1, 1), (, 1), ( 2 , 2 1)} = Z 3 is easily obtained. Thus we have the isomorphism (U (1) E6 )/Z 3
= (E7 ) .
Remark. (E7 ) is connected as a fixed points subgroup under the involution of
the simply connected Lie group E7 . Hence, to show that : U (1) E6 (E6 ) is
onto, it is sufficient to show that : u(1) e6 (e7 ) is onto, which is easily shown.
4.11. Involution and subgroup (SU (2) Spin(12))/Z 2 of E7
We define an involutive C-linear transformation of PC by
(X, Y, , ) = (X, Y, , ).
which is the extension of the C-linear transformation of JC . This is the same as
F4 regarding as F4 E6 E7 .
We shall now study the following subgroup (E7 ) of E7 :
(E7 ) = { E7 | = }.
To this end, we define two C-linear mappings , : PC PC by
= (2E1 E1 , 0, 0, 1),
= (0, E1 , E1 , 0).
The explicit forms of and are given by

X
1 X
Y Y
= 1 , 1 X = (E1 , X) 4E1 (E1 X),
X
2E1 Y + E1
Y 2E X + E1
= 1
.
(E1 , Y )
(E1 , X)
More precisely, and are of the form
1 y3 y 2
1 x3 x2

(X, Y, , ) = x3 2 x1 , y 3 2 y1 , ,
x2 x1 3
y2 y 1 3
0
1
0
0
1 0

= 0
2 x1 , 0 2 y1 , , ,
0
x1 3
0 y 1 3

0
0
0
0

x1 , 1 , 1 .
(X, Y, , ) = 0 3 y1 , 0 3
2
0 x1
0 y1 2

=
=
Lemma 4.11.1. (1)
= ,
= ,
= .
133
(2) If E7 satisfies = , then also satisfies = .

Proof. (1) These are checked by direct calculations.
(2) Since = exp i, we have = (exp i) = (exp i) = .
We shall first study the following subgroups (E7 ), and ((E7 ), )(0,E1 ,0,1) of E7 :
(E7 ), = { E7 | = , = },
((E7 ), )(0,E1 ,0,1) = { (E7 ), | (0, E1 , 0, 1) = (0, E1 , 0, 1)}.

Proposition 4.11.2. The Lie algebras (e7 ) , (e7 ), , ((e7 ), )(0,E1 ,0,1) of the
groups (E7 ) , (E7 ), , ((E7 ), )(0,E1 ,0,1) are respectively given by
(1) (e7 ) = { e7 | = }
= {(, A, A, ) e7 | (e6 ) , A (JC ) }.
(2) (e7 ), = { e7 | = , = }

(
(e6 ) , A (JC ) , (E1 , A) = 0, )

= (.A, A, ) e7
.
3
= (E1 , E1 )
2
(3) ((e7 ), )(0,E1 ,0,1) = { (e7 ), | ((0, E1 , 0, 1)) = 0}

e , E = 0,
6
1

= (, A, A, 0) e7
.
A JC , 2E1 A = A
Proof. (1) It is not difficult to prove and so is omitted here.
(2) Suppose that = (, A, A, ) e7 satisfies = and = .

Comparing the -term of P = P, P = (X, Y, , ) PC , we have
(A, Y ) = (A, 1 Y ),
Y JC .
Let Y = E1 , then we have (A, E1 ) = 0. Next, comparing the -term of = , we

have
1
(i)
(E1 , X) (E1 , X) = (E1 , X).
3
Since (e6 ) , we can set E1 = kE1 , k iR (Lemma 3.10.1). Hence let X = E1 in
2
(i), then we have k = . Conversely, if = (, A, A, ) e7 has the condition
3
above, then from the following Lemma 4.11.3, we can see that satisfies =
and = .
Lemma 4.11.3. In JC , the following hold.
(1) For A (JC ) , we have 1 (A X) = 1 A 1 X, X JC .
(2) For (e6 ) , we have 1 = 1 .
(3) For A (JC ) , (E1 , A) = 0, we have 1 A = A and

4 A (E1 X) + (E1 , X)A = 4E1 (A X) hA, XiE1 ,
134
X JC .
We shall now return to the proof of (3) of Proposition 4.11.2.

If = (, A, A, ) (e7 ), satisfies ((0, E1 , 0, 1)) = 0, then
= (A, E1 ) = (A, E1 ),
so that = . Together with = , we have = 0 and E1 = 0, furthermore,
we have 2A E1 = A (in this case, A (JC ) and (E1 , A) = 0 automatically hold).
For iR, we define a C-linear mapping () : JC JC by
() = 2E1 E1 ,
that is,
1
() x3
x2
x3
2
x1
x2
41
x1 = x3
3
3
x2
(Lemma 3.4.2.(2)). Then () (e6 ) .
x3
22
2x1
x2
2x1
23
Proposition 4.11.4. (1) a1 = {((), aE1 , aE1 , ) | a C, iR}

is a Lie subalgebra of (e7 ) and isomorphic to the Lie algebra su(2).
(2) The Lie algebra (e7 ) is isomorphic to the direct sum of Lie algebras a1 and
(e7 ), :
(e7 )
= a1 (e7 ), .
by
Proof. (1) The mapping : a1 su(2) = {D M (2, C) | (tD) = D} defined

(((), aE1 , aE1 , )) =
gives an isomorphism as Lie algebras. Indeed, this is clear from

i

h
b
b( a) a( b)
2(b a)
,
=
,
a
b
2 (b a) a( b) b( a)
and
h
i
((), aE1 , aE1 , ), ((), bE1 , bE1 , )
= ((b a) a( b)), 2(b a)E1 , 2 (b a)E1 , ( a)b a( b)).

(2) Using (1) above and Proposition 4.11.2.(2), the following decomposition of
(e7 ) ,
( )
( )
A
aE1
A aE1
,
(e7 )
=
+
a1 (e7 ) ,
A
aE1
A + aE1

where =
1
1
+ (E1 , E1 ), a = (E1 , A), gives an isomorphism as Lie algebras.
3
2
135
Lemma 4.11.5. For a C, we have

23 (a) = 2 (a)3 ( a) ((E7 ), )(0,E1 ,0,1) ,
where i (a) E7 , i = 2, 3 are defined in Lemma 4.8.1.
Proof. Since (0, aEi , aEi , 0) (e7 ), , we have i (a) = exp (0, aEi ,
aEi , 0) (E7 ), , i = 2, 3. Since 2 (a) and 3 ( a) are commutative, we have
23 (a) = 2 (a)3 ( a)
= exp (0, aE2 aE3 , aE2 + aE3 , 0).
Since (0, aE2 aE3 , aE2 + aE3 , 0) ((e7 ), )(0,E1 ,0,1) , we have 23 (a)
((E7 ), )(0,E1 ,0,1) .
We recall the group
Spin(10) = { E6 | E1 = E1 }
= { E6 | = , E1 = E1 } E7
(Lemma 3.10.4). The group Spin(10) acts transitively on the 9 dimensional sphere
0
n 0 0

o

0
S9 =
x , 0, 0, 0 C, x C, xx + ( ) = 1 .
0 x
Lemma 4.11.6. For ((E7 ), )(0,E1 ,0,1) , we have
(0, E1 , 0, 1) = (0, E1 , 0, 1)
if and only if
(0, 0, 1, 0) = (0, 0, 1, 0).
In particular, we have
{ ((E7 ), )(0,E1 ,0,1) | (0, E1 , 0, 1) = (0, E1 , 0, 1)}
= Spin(10).
Proof. If (E7 ), satisfies (0, E1 , 0, 1) = (0, E1 , 0, 1) and (0, E1 , 0, 1) =
(0, E1 , 0, 1), then we have (0, 0, 0, 1) = (0, 0, 0, 1) and (0, E1 , 0, 0) = (0, E1 , 0, 0),
which imply that (0, 0, 1, 0) = (0, E1 , 0, 0) = (0, E1 , 0, 0) = (0, E1 , 0, 0) =
(0, 0, 1, 0). The converse can be similarly proved. If E7 satisfies (0, 0, 1, 0) =
(0, 0, 1, 0), then E6 (Theorem 4.7.2), and from the condition E1 = E1 , we obtain
Spin(10), (Theorem 3.10.4). The converse also holds.
We define an 11 dimensional R-vector space V 11 by
V 11 = {P PC | P = P, P
n 0 0
0
=
x ,0
0 x
0
= P, P (0, E1 , 0, 1, 0) = 0}
0 0

o

0 0 , 0, x C, C, iR
0 0
136
with the norm (P, P ) given by

(P, P ) =
Proposition 4.11.7.
1
{P, P } = xx + ( ) + ( ).
2
((E7 ), )(0,E1 ,0,1) /Spin(10) S 10 .
In particular, the group ((E7 ), )(0,E1 ,0,1) is connected.
Proof. S 10 = {P V 11 | (P, P ) = 1} is a 10 dimensional sphere. For

((E7 ), )(0,E1 ,0,1) and P S 10 , we have P S 10 (Proposition 4.2.2, Lemma 4.3.3).
Hence the group ((E7 ), )(0,E1 ,0,1) acts on S 10 . We shall prove that this action is
transitive. To prove this, it is sufficient to show that any element P S 10 can be
transformed to (0, iE1 , 0, i) S 10 by some ((E7 ), )(0,E1 ,0,1) . Now, for a given
0
0 0
0 0

P = 0
x , 0 0 0 , 0, S 10 ,
0 x
0 0 0
choose a R, 0 a <
, such that
4
tan 2a =
2
.

If = 0, then we choose a =
. Applying 23 (a) of Lemma 4.11.5 on P , then
4
the -part of 23 (a)P becomes
2 sin2 a(E2 , E3 X) + sin a cos a (E3 , Y ) sin a cos a cos2 a
= sin2 a + ( ) sin a cos a cos2 a
1
= ( ) sin 2a cos 2a = 0.
2
Hence we have
23 (a)P S 9 .
Since the group Spin(10) acts transitively on S 9 (Proposition 3.10.3), there exists
Spin(10) = (E6 )E1 ((E7 ), )(0,E1 ,0,1) such that
23 (a)P = (i(E2 + E3 ), 0, 0, 0).
Again, applying 23 (/4) ((E7 ), )(0,E1 ,0,1) of Lemma 4.11.5 on the above, we
have
23 (/4)23 (a)P = (0, iE1 , 0, i).
This shows the transitivity of ((E7 ), )(0,E1 ,0,1) . The isotropy subgroup of
((E7 ), )(0,E1 ,0,1) at (0 iE1 , 0, i) is Spin(10) (Lemma 4.11.6). Thus we have the
homeomorphism ((E7 ), )(0,E1 ,0,1) /Spin(10) S 10 .
Theorem 4.11.8.
((E7 ), )(0,E1 ,0,1)

= Spin(11).
137

Proof. Analogous to Theorem 3.10.4, we can define a homomorphism p :
((E7 ), )(0,E1 ,0,1) SO(11) = SO(V 11 ) by p() = |V 11 . The restriction p of p
to (E6 )E1 coincides with the homomorphism p : Spin(10) SO(10) = SO(V 10 )
(where V 10 = {P V 11 | P = (X, 0, 0, 0)}). In particular, p : Spin(10) SO(10) is
onto. Hence, from the following commutative diagram
1
Spin(10)
p
SO(10)
((E7 ), )(0,E1 ,0,1)

p
SO(11)
S 10
=
S 10
we see that p : ((E7 ), )(0,E1 ,0,1) SO(11) is onto by the five lemma. Using the five
lemme again, we see that Ker p coincides with Ker p . Hence Ker p = {1, } (Theorem
3.10.4). Thus we have the isomorphism
((E7 ), )(0,E1 ,0,1) /{1, }
= SO(11).
Therefore the group ((E7 ), )(0,E1 ,0,1) is isomorphic to the group Spin(11) as the
universal covering group of SO(11).
Lemma 4.11.9. For t R, we define a mapping (t) : PC PC by
1 y3 y 2
1 x3 x2

(t) x3 2 x1 , y 3 2 y1 , ,
x2 x1 3
y2 y 1 3
2it
2it
it
it
e
1 eit y3 eit y 2

e 1 e x3 e x2
2
x1 , eit y 3
2
y1 , e2it , e2it ,
= eit x3
x1
3
y1
3
eit x2
eit y2
then (t) (E7 ), .
Proof. For = it iR, let () = 2E1 E1 (e6 ) . Then, ((), 0, 0, 2)

(e7 ), (Proposition 4.11.2) and (t) = exp ((), 0, 0, 2). Hence we have (t)
(E7 ), .
We define a 12 dimensional R-vector space V 12 by
V 12 = {P PC | P = P, P

n 0 0
0
x ,0
=
0 x
0
= P}
0
0
0
with the norm (P, P ) given by

(P, P ) =
Proposition 4.11.10.
0
o

0 , 0, x C, , C
0
1
{P, P } = xx + ( ) + ( ).
2
(E7 ), /Spin(11) S 11 .
138
In particular, the group (E7 ), is connected.

Proof. S 11 = {P V 12 | (P, P ) = 1} is an 11 dimensional sphere. For
(E7 ), and P S 11 , we have P S 11 (Proposition 4.2.2, Lemma 4.3.3). Hence
the group (E7 ), acts on S 11 . We shall prove that this action is transitive. To
prove this, it is sufficient to show that any element P S 11 can be transformed to
(0, E1 , 0, 1) S 11 by some (E7 ), . Now, for a given
0
0 0
0 0

x , 0 0 0 , 0, S 11 ,
P = 0
0 x
0 0 0
we choose t R such that e2it iR. Applying (t) of Lemma 4.11.9 on P , we get
(t)P S 10 .
Since the group Spin(11) acts transitively on S 10 (Proposition 4.11.7), there exists
Spin(11) = ((E7 ), )(0,E1 ,0,1) such that
(t)P = (0, iE1 , 0, i).
If we further apply (/4) (E7 ), of Lemma 4.11.9 on the above, then we have
(/4)(t)P = (0, E1 , 0, 1).
This shows the transitivity of (E7 ), . The isotropy subgroup of (E7 ), at (0, E1 , 0, 1)
is Spin(11) (Theorem 4.11.8). Thus we have the homeomorphism (E7 ), / Spin(11)
S 11 .
(E7 ),
= Spin(12).
Theorem 4.11.11.

Proof. Analogous to Theorem 4.11.8, we can define a homomorphism
p : (E7 ), SO(12) = SO(V 12 )
by p() = |V 12 . The restriction p of p to (E6 ), coincides with the homomorphism
p : Spin(11) SO(11) of Theorem 4.11.8. In particular, p : Spin(11) SO(11) is
onto. Hence from the following commutative diagram
1
Spin(11)
p
SO(11)
(E7 ),
p
SO(12)
S 11
=
S 11
we see that p : (E7 ), SO(12) is onto by the five lemma. Using the five lemme
again we see that Ker p coincides with Ker p . Hence Ker p = {1, } (Theorem 4.11.8).
Thus we have the isomorphism
(E7 ), /{1, }
= SO(12).
139
Therefore the group (E7 ), is isomorphic to the group Spin(12) as the universal
covering group of SO(12).
Theorem 4.11.12. The center z(Spin(12)) of Spin(12) is
z(Spin(12)) = {1, 1, }
= {1, 1} {1, }
= Z2 Z2.
And we have
Spin(12)/{1, }
= SO(12),
Spin(12)/{1, 1}
= Spin(12)/{1, }
= Ss(12).
Theorem 4.11.13. The group (E7 ) contains a subgroup
2 (SU (2)) = {2 (A) E7 | A SU (2)}
which is isomorphic to the special unitary group SU (2) = {A M (2, C) | t ( A)A =
E, det A = 1}. Here, for A SU (2), a mapping 2 (A) : PC PC is defined by
1 y3 y 2
1 x3 x2

2 (A) x3 2 x1 , y 3 2 y1 , ,
3
x2 x1 3
y2 y

1
1 y3 y 2
1 x3 x2

= x3 2 x1 , y3 2 y1 , , ,
x2 x1 3
y2 y 1 3
where

1
3
3
1
2
2
,
,
=A
=A
,
,
=A
=A
2
2
1
3
3
1

x3
x2
x1
x3
x2
x1
.
,
,
=
=
= A
y3
y3
y2
y2
y1
y1
Proof. The action of ((), aE1 , aE1 , ) a1 (() = 2E1 E1 , iR, a

C) on PC is
((), aE1 , aE1 , )(X, Y, , ) = (X , Y , , )
where

a
1
1
=
,
=
,
1
a
1

a
3
3
a
2
2
,
=
,
=
2
a
2
3
a
3

x1
a
x1
x2
x3
0
=
,
=
=
.
y1
a
y2
y1
y3
0

a
Therefore, for A = exp
SU (2), we have
a

exp(((), aE1 , aE1 , )) = 2 (A) 2 (SU (2)) (E7 ) .

140
Lemma 4.11.14. The group (E7 ) is connected.

Proof. The group (E7 ) is the fixed points set obtained by the involutive automorphism of the simply connected Lie group E7 , hence (E7 ) is connected.
Theorem 4.11.15. (E7 )
= (SU (2) Spin(12))/Z 2 , Z 2 = {(E, 1), (E, )}.
Proof. We define a mapping : SU (2) Spin(12) (E7 ) by
(A, ) = 2 (A).
Since the Lie algebras a1 and (e7 ), of SU (2) and Spin(12) are elementwise commutative (Proposition 4.11.4.(2)), 2 (A) SU (2) and Spin(12) are commutative
: 2 (A) = 2 (A). Hence is a homomorphism. We shall show that is onto.
Since the group (E7 ) is connected (Lemma 4.11.14), to prove this, it is sufficient to
show that its differential mapping : a1 (e7 ), (e7 ) is onto. However this
has been already shown in Proposition 4.11.4.(2). Ker = {(E, 1), (E, 2 (E))} =
{(E, 1), (E, )} = Z 2 (Theorem 4.11.13) is easily obtained. Thus we have the
isomorphism (SU (2) Spin(12))/Z 2
= (E7 ) .
Remark. We can give an elementary proof of Lemma 4.11.14 once we have proved
the following three claims.
Claim 1. Any element X (JC )
(E6 ) :
1
X = 0
0
can be transformed to a diagonal form by some

0
2
0
0
0 ,
3
i C.
Moreover we can choose (E6 ) so that 2 0, 3 0.

e1 (a)(a C) (e6 ) , then
Proof. Recall that i(E1 E2 ) , i(E1 E3 ) , iFe1 (a), A
we can prove analogously as in Proposition 3.8.2.
Claim 2. Any element P (MC ) = {P MC | P = P } can be transformed to

a diagonal form by some ((E7 ) )0 (the connected component of (E7 ) containing
the identity 1):
P = (X, Y, , ),
X, Y are diagonal, > 0.
Proof. Recall (0, aEi , aEi , 0) (e7 ) , i = 1, 2, 3, then we can prove analogously as in Proposition 4.8.2.
Claim 3.
(E7 ) /(E6 ) (M1 ) = {P M1 | P = P }.
In particular, the group (E7 ) is connected.
Proof. Remark that i (a) of Lemma 4.8.1 belongs to ((E7 ) )0 , then this claim is
proved analogously as Theorem 4.8.3. The connectedness of (E7 ) follows from the
connectedness of (E6 )
= (U (1) Spin(10))/Z 4 (Theorem 3.10.7) and (M1 ) .
141
4.12. Involution and subgroup SU (8)/Z 2 of E7

We consider the involutive complex conjugate transformation of PC ,
(X, Y, , ) = ( X, Y, , ),
where is the same as G2 F4 E6 E7 .
(E7 ) = { E7 | = }
= { E7 | = } = (E7 ) .
To this end, we consider R-vector subspaces (PC ) , (PC ) of PC , which are

eigenspaces of , respectively by
(PC ) = {P PC | P = P }
= {(X, Y, , ) PC | X, Y (JC ) , , R},
(PC ) = {P PC | P = P }
= {(X, Y, , ) PC | X, Y (JC ) , , iR},
= i(PC ) .
These spaces (PC ) , (PC ) are invariant under the action of (E7 ) and we have
the decomposition
PC = (PC ) (PC ) = (PC ) i(PC ) .
In paticular, PC is the complexification of (PC ) : PC = ((PC ) )C .
Analogious to Section 3.11, we can define the R-linear mapping k : M (4, H)
M (8, C),

a b
k a + be2 =
, a, b C.
b a
Lemma 4.12.1. Any element B su(8) is uniquely expressed by
B = k(D) + e1 k(T ),
Proof. For B su(8), let D1 =
have
B = D1 + e1 T 1 ,
D sp(4), T J(4, H)0 .
B JBJ
B + JBJ
M (8, C), then we
, T1 =
2
2e1
D1 = D1 , JD1 = D1 J,
T1 = T1 , JT1 = T 1 J, tr(T1 ) = 0.
Then, D = k 1 (D1 ), T = k 1 (T1 ) M (4, H) are the required elements. To prove

the uniqueness of the expression, it is sufficient to show that
D1 + e1 T1 = 0, D1 k(sp(4)), T1 k(J(4, H)0 )
142
implies
C1 = T1 = 0.
Certainly, from the condition, we have JD1 + e1 JT1 = 0, so D1 J + e1 T 1 J = 0, and

so that D1 J e1 T1 J = 0, that is, D1 e1 T1 = 0. Together with the first equation,
we have D1 = T1 = 0.
After this, we will use the C-linear mapping g : JC J(4, H)C , g(M + a) =
1
tr(M )
ia
2
, the homomorphism : Sp(4) (E6 ) , (A)X =
1
ia
M tr(M )E
2
1
g (A(gX)A ), X JC and its differential mapping : sp(4) (e6 ) , (D)X =

g 1 (D(gX) + (gX)D ), X JC which are defined in Section 3.12.
Proposition 4.12.2. The Lie algebra (e7 ) of the group (E7 ) is
(e7 ) = { e7 | = }
= {(, A, A, 0) e7 | (e6 ) , A (JC ) }
= {( (D), g 1 (T ), g 1(T ), 0) e7 | D sp(4), T J(4, H)0 }.

The Lie bracket [1 , 2 ] in (e7 ) is given by
[(1 , A1 , A1 , 0), (2 , A2 , A2 , 0)] = (, A, A, 0),
where
= [1 , 2 ] 2A1 A2 + 2A2 A1 ,
A = 1 A2 2 A1 .
Proof. It is not difficult to verity them..

Analogous to Section 3.11, we define a C-vector space S(8, C) by
S(8, C) = {S M (8, C) | t S = S},
and a C-linear mapping kJ : J(4, H)C S(8, C)C by
kJ (M1 + iM2 ) = k(M1 )J + ik(M2 )J, M1 , M2 J(4, H),

0 1
where J = diag(J, J, J, J) M (8, C), J =
.
1 0
Definition. We define a C-linear isomorphism : PC S(8, C)C by

(X, Y, , ) = kJ gX E + e1 kJ g(Y ) E .
2
2
Proposition 4.12.3.
(e7 )
= su(8).
This isomorphism is given by the mapping : su(8) (e7 ) ,

(B)P = 1 (B(P ) + (P )tB),
143
P PC .
Proof. We first prove that for B su(8) we have (B) (e7 ) .

(i) For B = k(D), D sp(4), we have
P = (X, Y, , )

k gX E J + e1 k g(Y ) E J
2
2

k(D)k gX E J + e1 k(D)k g(Y ) E J
2
2

t
t
+k gX E J k(D) + e1 k g(Y ) E J k(D)
2
2

= k D gX E J + e1 k D g(Y ) E J
2
2

+k gX E D J + e1 k g(Y ) E D J
2
2
= k(D(gX) + (gX)D )J + e1 k(D(g(Y ) + (g(Y ))D )J
= k(g( (D)X))J + e1 k(g( (D)(Y )))J
(recall (C)X = g 1 (C(gX) + (gX)C ))
(D)X
(D)X
0

0
(D)
(D)Y
=
=
0
0
0
0
0
0
= (( (D), 0, 0, 0)P ).
0
0
0
0

0
X

0 Y

0
Hence, we have (k(D)) = ( (D), 0, 0, 0) (e7 ) .
(ii) For B = e1 k(T ), T J(4, H)0 (denote T = gA, A (JC ) )
P = (X, Y, , )

k gX E J + e1 k g(Y ) E J
2
2

e1 k(T )k gX E J k(T )k g(Y ) E J
2
2

t
t
+e1 k gX E J k(T ) k g(Y ) E J k(T )
2
2
= k(T g(Y ) g(Y )T + T )J + e1 k(T (gX) + (gX)T T )J
= k(2gA g(Y ) + gA)J + e1 k(2gA gX gA)J

1
= k 2g(A Y ) (A, Y )E + gA J
2

1
+e1 k 2g(A X) + (A, X)E gA J (Lemma
2
2A Y + A
0
2A
0

0
A
2A X A
2A
=
=
(A, Y )
0
A
0
(A, X)
A
0
0
= ((0, A, A, 0)P ).
Hence we have (e1 k(T )) = (0, A, A, 0) (e7 ) .

144
3.12.1)

A
X

0 Y

0
Consequently we see that the mapping : su(8) (e7 ) is well-defined. We can

easily check that is onto. Finally, we have to prove that is a homomorphism
as Lie algebras. However this follows from the following Theorem 4.12.5, so that we
shall omit the proof.
Proof. The group (E7 ) is the fixed points set by the involution of the simply
connected Lie group E7 , hence (E7 ) is connected.
Theorem 4.12.5.
(E7 )
= SU (8)/Z 2 , Z 2 = {E, E}.
Proof. We define a mapping : SU (8) (E7 ) by

(A)P = 1 (A(P ) tA),
P PC .
We first prove (A) (E7 ) . To prove this, for the differential mapping : su(8)
(e7 ) of ,
(D)P = 1 (D(P ) + (P )t D), P PC ,
it is sufficient to show that is well-defined, that is, (D) (e7 ) . However
this fact is already shown in Proposition 4.12.3. Evidently : SU (8) (E7 ) is a
homomorphism. Since : su(8) (e7 ) is onto and (E7 ) is connected (Lemma
4.12.4), : SU (8) (E7 ) is also onto. Ker = {E, E} = Z 2 is easily obtained.
Thus we have the isomorphism SU (8)/Z 2
= (E7 ) .
Remark. Without using Lemma 4.12.4, the fact that the mapping : SU (8)
(E7 ) is onto will be followed from two claims.
Claim 1. For a R, i (a) of Lemma 4.8.1 belongs to (SU (8)).
Proof. i (a) = exp((0, aEi , aEi , 0)) exp (su(8)) (Proposition 4.12.3) =
(exp(su(8))) (SU (8)).
Claim 2. Any element P (MC ) = {P MC | P = P } can be transformed

to a diagonal form by some (SU (8)):
P = (X, Y, , ),
X, Y are real diagonal, > 0.
Proof. Let P = (X, Y, , ) (MC ) . If 6= 0. Then

Y = Y,
X=
1
(Y Y ),
= ,
= .
Since Y (JC ) , we have Y (JC ) , so that g(Y ) J(4, H)0 . Hence, there
exists D Sp(4) such that
D(g(Y ))D
is real diagonal.
145
Then
(D)Y = g 1 (D(g(Y ))D ) is real diagonal,
1
1
(D)X = (D) Y Y = ((D)Y (D)Y ) is real diagonal.
In the case = 0, by the same proof of Proposition 4.8.2, we can choose (SU (8))
so that
P = (X, Y, , ), X, Y are real diagonal, 0 6= R.
If < 0, apply 1 () of Claim 1 on it, then becomes > 0.
Now, we will return to the proof of the surjection of : SU (8) (E7 ) using
Claims 1, 2. For a given (E7 ) , consider the element P = 1 (MC ) . We
first transform P to a diagonal form (Claim 2) by some (SU (8)), and we have,
in a similar way to Theorem 4.8.3,
1 (a1 )1 2 (a2 )1 3 (a3 )1 1 = 1,

i
ri (i is a diagonal element of Y ). Since i R, we have i (ai )
|i |
(SU (8)) (Claim 1). If we put
e = 1 (a1 )1 2 (a2 )1 3 (a3 )1 , then,
e E6
(Theorem 4.7.2) and
e satisfies
e=
e . Hence
e (E6 ) = (Sp(4)) (Theorem
3.12.2) (SU (8)), Therefore = 1 3 (a3 )2 (a2 )1 (a1 )e
(SU (8)). This shows
that is onto.
where ai =
E7
4.13. Automorphism w of order 3 and subgroup (SU (3) SU (6))/Z 3 of

We define a C-linear transformation w of order 3 of PC by
w(X, Y, , ) = (wX, wY, , ).
This w is the same as w G2 F4 E6 E7 .

We shall study the following subgroup (E7 )w of E7 :
(E7 )w = { E7 | w = w}.
We consider the group E7,C replaced with C in the place C in the definition of
the group E7 :
E7,C = { IsoC ((PC )C ) | (P Q)1 = P Q, hP, Qi = hP, Qi}.
As in Section 4.7, the group E7,C contains a subgroup
E6,C = { E7,C | (0, 0, 1, 0) = (0, 0, 1, 0)},
146
which is isomorphic to the group ((SU (3) SU (3))/Z 3 ) Z 2 (Proposition 3.13.4).

The Lie algebra e7,C of the group E7,C is given by
e7,C = {(, A, A, ) | e6,C , A (JC )C , iR}
(Theorem 4.3.4). In particular, the dimension of e7,C is
dim e7,C = 16 + 18 + 1 = 35.
As in Theorem 4.8.3, we see that the space
(MC )1 = {P (MC )C | P P = 0, hP, P i = 1}
is connected and we have the homeomorphism
E7,C /E6,C (MC )1 .
Lemma 4.13.1. E7,C has at most two connected components (in reality has two
connected components).
Proof. From the exact sequence 0 (E6,C ) 0 (E7,C ) 0 ((MC )1 ), that is,
Z 2 0 (E7.C ) 0 (Proposition 3.13.4), we see that 0 (E7,C ) is 0 or Z 2 .
Let h : C C be the R-linear isomorphism defined by
h (a + bi) = a + be1 ,
a, b R.
Now, let V, W be C- and C-vector spaces, respectively. A linear mapping f : V

W is called a C-C-linear mapping if
f (av) = h (a)f (v),
a C, v V.
Similarly, a C-C-linear mapping g : W V is defined.

Definition. Let h : C C C be a C-C-linear mapping defined by
h (a + bi) = a + be1 ,
a, b C.
Now, let 3 (C 6 ) be the third exterior product of C-vector space C 6 and we define
a C-C-linear isomorphism f : (PC )C 3 (C 6 ) by
1 y3 y 2
1 x3 x2

X
f x3 2 x1 , y 3 2 y1 , , =
xijk ei ej ek
x2 x1 3
y2 y 1 3
i<j<k

{e1 , e2 , , e6 } is the canonical basis of C 6 and xijk C are skew-symmetric tensor:

i j k
xi j k = sgn
xijk , where
i j k
147
x156 = h (1 ),
x164 = h (x3 ),
x145 = h (x2 ),
x256 = h (x3 ),
x264 = h (2 ),
x245 = h (x1 ),
x356 = h (x2 ),
x364 = h (x1 ),
x345 = h (3 ),
x423 = h (1 ),
x431 = h (y3 ),
x412 = h (y 2 ),
x523 = h (y 3 ),
x531 = h (2 ),
x512 = h (y1 ),
x623 = h (y2 ),
x631 = h (y 1 ),
x612 = h (3 ),
x123 = h (),
x456 = h ().
The inverse mapping f 1 : 3 (C 6 ) (PC )C of f

h(x156 )
h(x256 , x164 )
h(x356 , x145 )
X

h(x423 )
f 1
xijk ei ej ek =
h(x523 , x431 )
i<j<k
h(x623 , x412 )
is given by
h(x164 , x256 )
h(x264 )
h(x364 , x245 )
h(x431 , x523 )
h(x531 )
h(x631 , x512 )
h(x123 )
h(x456 )

h(x145 , x356 )
h(x245 , x364 )
h(x345 )

h(x412 , x623 )
,
h(x512 , x631 )
h(x612 )
where h : C C C C , h : C C are C-C-linear mappings defined respectively by

a+b
(b a)e1
+i
, a, b C,
2
2
h(a + be1 ) = a + bi, a, b R.
h(a, b) =
It is easy to see that

f (h(a)P ) = a(f P ),
a C, P (PC )C .
The group SU (6) acts naturally on 3 (C 6 ), that is, the action of A SU (6) on
a b c 3 (C 6 ) is defined by
A(a b c) = Aa Ab Ac.
Hence, the action of D su(6) on 3 (C 6 ) is given by
D(a b c) = Da b c + a Db c + a b Dc.
Lemma 4.13.2. (1) Any element D su(6) is uniquely expressed by

E
B
L
0
D=
+
, B, C su(3), L M (3, C), e1 R.
L C
3 0 E
148
(2) The Lie algebra e7,C is isomorphic to the Lie algebra su(6) as Lie algebras:
e7,C
= su(6).
This isomorphism is given by the mapping C : su(6) e7,C ,

E
B
L
0
+
C
= (C (B, C), h(L), h(L), ie1 )
L C
3 0 E
where C : su(3) su(3) e6,C is defined by C (B, C)X = h(B, C)X + Xh(B, C) ,
X (JC )C (Lemma 3.13.3).

B
L
Proof. (1) For D =
su(6), B , C u(3), L M (3, C), we let
L C
= tr(B ) = tr(C ),
B = B
E,
3
C = C +
E,
3
then we have the result.

(2) This is the direct consequence of the following Proposition 4.13.3, so we will
omit its proof.
We define the action of the group Z 2 = {1, } on the group SU (6) by

0
E
A = (AdJ3 )A, J3 =
,
E 0
that is,
A =
A11
A21
A12
A22
0
E
E
0

A11
A21
A12
A22

0
E
E
0
1
A22 A21
A12 A11
where E, Aij M (3, C), and let SU (6) Z 2 be the semi-direct product of the groups
SU (6) and Z 2 under this action.
Proposition
4.13.3. E7,C
= (SU (6)/Z 3 ) Z 2 , Z 3 = {E, 1 E, 1 2 E}, 1 =
1
3
+
e1 .
2
2
Proof. We define a mapping : SU (6) Z 2 E7,C by
(A, 1)P = f 1 (A(f P )),
(A, )P = f 1 (A(f P )),
P (PC )C .
We first have to show that (A, 1) E7,C . To prove this, it is sufficient to show that
the differential mapping : su(6) e7,C of :
(D)P = f 1 (D(f P )),
P (PC )C
coincides with the mapping C : su(6) e7,C of Lemma 4.13.2. We put

E
0
B
L
D=
+
L C
3 0 E
149
b11
b12
b13
l11
l12
l13
b12
b22
b23
l21
l22
l23
b13
b23
b33
l31
l32
l33
l11
l21
l31
c11
c12
c13
l12
l22
l32
c12
c22
c23
l13
l23
l33
c13
c23
c33

E
+
3 0
0
E
su(6),
where bii = bii , cii = cii , b11 + b22 + b33 = c11 + c22 + c33 = 0, = .
(1) For P = (0, 0, 1, 0), (D)P is calculated as follows.
P = (0, 0, 1, 0)
f
e1 e2 e3
D
De1 e2 e3 + e1 De2 e3 + e1 e2 De3

e1 e2 e3 l11 e4 e2 e3 l12 e5 e2 e3 l13 e6 e2 e3
= b11 +
3


+ b22 +
3

+ b33 +
3
0 0 0
0 0 0
0
0
0

h(L)
h(l11 )
h(l21 , l12 ) h(l31 , l13 )

f 1
,
l
)
h(l
)
h(l
,
l
)
h(l
12
21
22
32
23
ie1
h(l33 )
h(l13 , l31 ) h(l23 , l32 )
h()
ie1
(B, C) +
C
3
2h(L)
=
0
h(L)
2 h(L)
C (B, C)
h(L)
0
= (C (B, C), h(L), h(L), ie1 )P .
ie1
3
h(L)
ie1
0

0

0

0

1
0
0
ie1
h(L)
(2) For P = (E1 , 0, 0, 0), (D)P is calculated as follows.

P = (E1 , 0, 0, 0)
f
e1 e5 e6
D
De1 e5 e6 + e1 De5 e6 + e1 e5 De6

e1 e5 e6 b12 e2 e5 e6 b13 e3 e5 e6 l11 e4 e5 e6
= b11 +
3

e1 e5 e6
+l22 e1 e2 e6 + l32 e1 e3 e6 + c12 e1 e4 e6 + c22
3

+l23 e1 e5 e2 + l33 e1 e5 e3 + c13 e1 e5 e4 + c33
e1 e5 e6
3
150

h(c12 , b12 ) h(c13 , b13 )
h b11 c11
3

h(b12 , c12 )
0
0
0
0
h(b13 , c13 )
f 1
0
0
0

0
h(l
)
h(l
33
23 , l32 )
0 h(l32 , l23 )
h(l22 )
h(l11 )
ie1
E1
E1
3
2h(L) E1
= (C (B, C), h(L), h(L), ie1 )
0
0
( h(L), E )
C (B, C)E1 +
(3) For P = (F1 (1), 0, 0, 0), (D)P is calculated as follows.

P = (F1 (1), 0, 0, 0)
f
e2 e4 e5 + e3 e6 e4
D
(De2 e4 e5 + e2 De4 e5 + e2 e4 De5 )
+(De3 e6 e4 + e3 De6 e4 + e3 e6 De4 )

e2 e4 e5 b23 e3 e4 e5 l23 e6 e4 e5
= b12 e1 e4 e5 + b22 +
3

+l11 e2 e1 e5 + l31 e2 e3 e5 + c11
e2 e4 e5 c13 e2 e6 e5
3

+l12 e2 e4 e1 + l32 e2 e4 e3 + c22

e3 e6 e4 c23 e2 e4 e6
3

+ b13 e1 e6 e4 + b23 e2 e6 e4 + b33 +

e3 e6 e4 l32 e5 e6 e4
3

+l13 e3 e1 e4 + l23 e3 e2 e4 + c23 e3 e5 e4 + c33
e3 e6 e4
3

e3 e6 e4 c12 e3 e6 e5
+l11 e3 e6 e1 + l21 e3 e6 e2 + c11
3

0
h(b13 , c13 )
0

h(b23 + c23 ) h(b22 c33 , b33 + c33 +

3
3
h(b23 + c23 )
h(c12 , b12 )
h(l23 + l32 ) h(l13 , l31 )
0
h(l11 )
h(l21 , l12 )
0
h(l23 + l32 )
151
ie1
F1 (1)
F1 (1)
3
2h(L) F1 (1)
= (C (B, C), h(L), h(L), ie1 )
0
0
(h(L), F (1))
C (B, C)F1 (1) +
(4) For the other generator of PC C , that is, P = (X, 0, 0, 0), (0, X, 0, 0) where
X = Ei , Fi (1), Fi (e1 ), i = 1, 2, 3 and P = (0, 0, 0, 1), we have also
f 1 (D(f P )) = (C (B, C), h(L), h(L), ie1 )P.
Thus we see that (A, 1) E7,C for A SU (6). Since (E, )P = P , we see
(E, ) = G2,C (= Aut(C)) F4,C E6,C E7,C . We shall show that
: SU (6) Z 2 E7,C is a homomorphism. For this purpose, we first show
f 1 (A(f P )) = f 1 (((AdJ3 )Af P )),
A SU (6), P (PC )C .
Furthermore, to show this, it is sufficient to show that for D su(6) instead of

A SU (6). Now,
f 1 (D(f P )) = (C (B, C), h(L), h(L), ie1 )P
= (C (B, C), h(L), h(L), ie1 )P
= (C (C, B), h(L ), h(L ), ie1 )P

= f 1 (((AdJ3 )D)(f P )).
is a homomorphism. Indeed, for example,

(A, )(B, 1)P = (A, )(f 1 (B(f P )))
= f 1 (Af (f 1 (B(f P ))) = f 1 (Af (f 1 ((AdJ3 )B)(f P )))
= f 1 ((A(B))(f P )) = (A(B), )P.
Thus, for A SU (6), we have (A, ) = (A, 1)(E, ) E7,C . Since induces a
surjection : su(6) e7,C , : SU (6) (E7,C )0 (which is the connected component
of E7,C containing the identity 1) is onto. However = (E, ) 6 (E7,C )0 . Indeed,
for any A SU (6),
X
(Aa Ab Ac) = a b c, a, b, c C 6
does not hold. Therefore E7,C has just two connected components (see Lemma 4.13.1).
Hence : SU (6) Z 2 E7,C is onto. Ker = {E, 1 E, 1 2 E} 1 = Z 3 1 easily
obtained. Thus we have the isomorphism (SU (6)/Z 3 ) Z 2
= E7,C .
We identify (PC )C (M (3, C)C M (3, C)C ) with PC (using the identification
JC C M (3, C)C with JC in Section 3.13) by
((X, Y, , ), (M, N )) = (X + M, Y + N, , ).
152
Further, we define a C-C-linear mapping : M (6, C) M (3, C)C M (3, C)C by

M11 M12
M21 M22
(M M )e
M21 + M12 (M22 + M11 )e1
M22 M11
21
12 1
,
=
+i
,
+i
2
2
2
2
where Mij M (3, C). The inverse mapping 1 : M (3, C)C M (3, C)C M (6, C)
of is given by

N2 N1 e1 M2 + M1 e1
1
, Mi , Ni M (3, C).
(M1 + iM2 , N1 + iN2 ) =
M 2 M 1 e1 N2 N1 e1
f M (6, C), we have
Lemma 4.13.4. For D su(6) and M
fD ) = (D)(M
f).
(M
Proof. Let

E
B
L
0
+
su(6),
L C
3 0 E

f = N2 N1 e1 M2 + M1 e1 , Mi , Ni M (3, C),
M
M 2 M 1 e1 N2 N1 e1
M = M1 + iM2 , N = N1 + iN2 .
D=
Then we have
f)
(D)(M
= (C (B, C), h(L), h(L), ie1 )(M, N )
1
2 h(L)
C (B, C) + ie1
3
1
2h(L)
C (B, C) ie1
=
0
h(L)
h(L)
0
1
C (B, C)M + ie1 M 2 h(L) N
3
= 2h(l) M + C (B, C) N ie1 N
(h(L), N )
( h(L), M )
1
M h(B, C) + N h(L) + ie1 M
3
=
ie
N
M
h(L)
h(B,
C)
153
0
h(L)
ie1
0
M

N
0
0
0
0
ie1
h(L)
(using C (B, C)M = M h(B, C) = M h(B, C) and 2 h(L) N = N h(L) etc.)

1
by simple calculations

N2 N1 e1 M2 + M1 e1
B
=
L
M 2 + M 1 e1 N2 + N1 e1
by
L
C
E
0
0
E

fD .
=M
Definition. We define a C-C-linear isomorphism f : PC 3 (C 6 ) M (6, C)

f (PC + (M + N )) = f (PC ) + 1 (M + N ),
PC + (M + N ) (PC )C (M (3, C)C M (3, C)C ) = PC .
The group SU (3) SU (6) acts on 3 (C 6 ) M (6, C) by
X
X
f) =
fA ,
(Q, A)( (a b c) + M
(Aa Ab Ac) + QM
f means
where QM
Q 0
0 Q

M11
M21
M12
M22
QM11
QM21
QM12
QM22
, Mij M (3, C).
Theorem 4.13.5. (E7

)w
= (SU (3) SU (6))/Z 3 , Z 3 = {(E, E), (1 E, 1 E),
1
3
(1 2 E, 1 2 E)}, 1 = +
e1 .
2
2
Proof. We defined a mapping : SU (3) SU (6) (E7 )w by
(Q, A)P = f 1 ((Q, A)(f P )),
P PC .
We first have to prove that (Q, A) E7 . To prove this, since (Q, E) (E6 )w
(E7 )w , it suffices to show that (E, A) E7 . Moreover, it is sufficient to show that,
for the differential mapping : su(3) su(6) e7 of , (0, D) coincides with
(C (B, C), h(L), h(L), ie1 ) e7 .
However, this is already shown in Proposition 4.13.3 and Lemma 4.13.4. Since
(PC )w = {P PC | wP = P } = (PC )C , obviously we have w(Q, A) = (Q, A)w,
hence (Q, A) (E7 )w . Evidently is a homomorphism. We shall show that is
onto. Let (E7 )w . Since the restriction to (PC )w = (PC )C of belongs to
E7,C , there exists A SU (6) such that
P = f 1 (A(f P ))
or P = f 1 (A(f P )),
P (PC )C
(Proposition 4.13.3). In the former case, let = (E, A)1 , then |(PC )C = 1,
hence G2 . Furthermore, (G2 )w = SU (3) (Theorem 1.9.4), so there exists
Q SU (3) such that
(PC + (M + N )) = PC + Q(M + N ) = PC + (QM + QN )
= (Q, E)(PC + (M + N )),
154
PC + (M + N ) PC
Hence we have
= (E, A) = (E, A)(Q, E) = (Q, A).
In this case, this shows that is onto. In the latter case, consider the mapping
1 : PC PC , 1 (PC + (M + N )) = PC + (M + N ),
PC + (M + N ) PC .
Then, 1 G2 F4 E6 E7 . From the same argument as Section 1, we have 1

(E7 )w , hence 1 (G2 )w = SU (3). However this is a contradiction (Theorem 1.9.4).
Therefore that is onto is shown. Ker = {(E, E), (1 E, 1 E), (1 2 E, 1 2 E)} = Z 3
is easily obtained. Thus we have the isomorphism (SU (3) SU (6))/Z 3
= (E7 )w .
Remark 1. The group E7 has a subgroup which is isomorphic to the semi-direct
product ((SU (3) SU (6))/Z 3 ) Z 2 (the action of the group Z 2 = {1, } to the group
SU (3) SU (6) is (Q, A) = (Q, Ad(J3 )A)).
Remark 2. Since (E7 )w is connected, the fact that : SU (3) SU (6) (E7 )w
is onto can be proved as follows. The elements
G01 , G23 , G45 , G67 , G46 + G47 , G47 G56 ,
G24 + G35 , G25 G34 , G26 + G37 , G27 G36 ,
el (1), A
el (e1 ), Fel (1), Fel (e1 ), (E1 E2 ) , (E2 E3 )
A
l , E
l , 1, l = 1, 2, 3
Fl (1), Fl (e1 ), Fl (1), Fl (e1 ), E
forms an R-basis of (e7 )w . So, dim(e7 )w = 10 + 14 + 6 3 + 1 = 43 = 8 + 35 =
dim(su(3) su(6)). Hence is onto.
E7 C E7 R133 .
In particular, E7 C is a simply connected complex Lie group of type E7 .
Proof. Evidently E7 C is an algebraic subgroup of IsoC (PC ) = GL(78, C). If
E7 C , then, the complex conjugate transpose with repect to the inner product
hX, Y i: hX, Y i = hX, Y i is = 1 1 E7 C . Hence, from Chevalleys
lemma, we have
E7 C (E7 C U (PC )) Rd = E7 Rd ,
d = 133.
Since E7 is simply connected (Theorem 4.9.2), E7 C is also simply connected. The Lie
algebra of the group E7 C is e7 C , so E7 C is a complex simple Lie group of type E7 .
4.15. Non-compact exceptional Lie groups E7(7) , E7(5) and E7(25) of
type E7
155
Let
P = J(3, C) J(3, C) R R,
P = J(3, C ) J(3, C ) R R.
For P, Q P or P , we define an R-linear mapping P Q : P P or P P as
similar to Section 4.1. And we define a Hermitian inner product hP, Qi in PC by
hP, Qi = hP, Qi.
Now, we define groups E7(7) , E7(5) and E7(25) by
E7(7) = { IsoR (P ) | (P Q)1 = P Q},
E7(5) = { IsoC (PC ) | (P Q)1 = P Q, hP, Qi = hP, Qi },
E7(25) = { IsoR (P) | (P Q)1 = P Q}.

E7(7)
= (E7 C ) ,
E7(5)
= (E7 C ) ,
E7(25)
= (E7 C ) .
Theorem 4.15.1. The polar decompositions of the groups E7(7) , E7(5) and
E7(25) are respectively given by
E7(7) SU (8)/Z 2 R70 ,
E7(5) (SU (2) Spin(12))/Z 2 R64 ,
E7(25) (U (1) E6 )/Z 3 R54 .
Proof. These are the facts corresponding to Theorems 4.12.5, 4.11.15 and 4.10.2.
Theorem 4.15.2. The centers of the groups E7(7) , E7(5) and E7(25) are the
group of order 2:
z(E7(7) ) = Z 2 ,
z(E7(5) ) = Z 2 ,
156
z(E7(25) ) = Z 2 .
5. Exceptional Lie group E8

5.1. Lie algebra e8 C
Theorem 5.1.1. In a 133 + 56 2 + 3 = 248 dimensional C-vector space
e8 C = e7 C PC PC C C C,
if we define a Lie bracket [R1 , R2 ] by
[(1 , P1 , Q1 , r1 , s1 , t1 ), (2 , P2 , Q2 , r2 , s2 , t2 )] = (, P, Q, r, s, t),
where
then e8 C
= [1 , 2 ] + P1 Q2 P2 Q1
P = 1 P2 2 P1 + r1 P2 r2 P1 + s1 Q2 s2 Q1
Q = 1 Q2 2 Q1 r1 Q2 + r2 Q1 + t1 P2 t2 P1
1
1
r = {P1 , Q2 } + {P2 , Q1 } + s1 t2 s2 t1
8
8
{P1 , P2 } + 2r1 s2 2r2 s1

s=
t = 1 {Q1 , Q2 } 2r1 t2 + 2r2 t1 ,

4
is a C-Lie algebra.
Proof. Among the definition of the Lie algebra, the relations

[R1 , R2 + R3 ] = [R1 , R2 ] + [R1 , R3 ],
[kR1 , R2 ] = k[R1 , R2 ], k C,
[R1 , R2 ] = [R2 , R1 ]
are evident, and we are left to show the Jacobi identity, which can be proved by direct
calculations as follows.
[R1 , [R2 , R3 ] ] + [R2 , [R3 , R1 ] ] + [R3 , [R1 , R2 ] ]
= (using [, P Q] = P Q + P Q (Proposition 4.3.2),
1
1
1
(P R)Q (Q R)P + {Q, R}P {P, R} {P, Q}R = 0
8
8
4
(Lemma 4.1.1.(3), {P, Q} + {P, Q} = 0 (Proposition 4.2.2.(2)) etc.)
= 0.
We use the following notation in e8 C :
= (, 0, 0, 0, 0, 0),
Q = (0, 0, Q, 0, 0, 0),
s = (0, 0, 0, 0, s, 0),
P = (0, P, 0, 0, 0, 0),
r = (0, 0, 0, r, 0, 0),
t = (0, 0, 0, 0, 0, t).
157
Theorem 5.2.1. The C-Lie algebra e8 C is simple.

Proof. We use the decomposition of e8 C :
e8 C = e7 C KC ,
where KC = PC PC C C C. Let p : e8 C e7 C and q : e8 C KC be
projections of e8 C = e7 C KC . Now, let a be a non-zero ideal of e8 C . Then p(a) is
an ideal of e7 C . Indeed, if p(a), then there exists (0, P, Q, r, s, t) KC such that
(, P, Q, r, s, t) a. For any 1 e7 C , we have
a [1 , (, P, Q, r, s, t)] = ([1 , ], 1 P, 1 Q, 0, 0, 0),
hence [1 , ] p(a).
We shall show that either e7 C a 6= {0} or KC a 6= {0}. Assume that e7 C a = {0}

and KC a = {0}. Then the mapping p|a : a e7 C is injective because KC a = {0}.
Since p(a) is a non-zero ideal of e7 C and e7 C is simple, we have p(a) = e7 C . Hence
dimC (a) = dimC (p(a)) = dimC (e7 C ) = 133. On the other hand, since e7 C a = {0},
q|a : a KC is also injective. Hence we have dimC (a) dimC (KC ) = 562+3 = 115.
This leads to a contradiction.
(1) Case e7 C a 6= {0}. From the simplicity of e7 C , we have e7 C a = e7 C , hence
a e7 C . On the other hand, we have
a [(0, 0, 0, 1), (0, 0, 1, 0)] = (0, 0, 1, 0),
a [(0, 0, 0, 1), (0, 0, 0, 1)] = (0, 0, 0, 1),
a [(0, 0, 1, 0) , (0, 0, 0, 4)] = 1 ,
a [(0, 0, 0, 1) , (0, 0, 4, 0)] = 1 ,
a [1 , 1 ] = 1,
a [1 + 1 , Q + P ] = P + Q ,
Therefore, a e7 C KC = e8 C which implies a = e8 C .
(2) Case KC a 6= {0}. Let R = (0, P, Q, r, s, t) be a non-zero element of KC a a.

(i) Case R = (0, P, Q, r, s, t), P 6= 0. We have
a [1, [1 , [1, R] ] ] = [1, [1 , (0, P, Q, 0, 2s, 2t)] ]

= [1, (0, 0, P, 2s, 0, 0)] = (0, 0, P, 0, 0, 0) = P .
We choose P1 PC so that P P1 6= 0 (Lemma 4.5.3) and choose e7 C so that
[, P P1 ] 6= 0. (Since e7 C simple, the center of e7 C consists only of 0, so such
exists). Then we have
a [, [P1 , P ] ] = [, P P1 ].
158

(ii) Case R = (0, P, Q, r, s, t), Q 6= 0. The argument is similar to (i).
(iii) Case R = (0, 0, 0, r, s, t), r 6= 0. For 0 6= P PC , we have
a [P , [1 , [1 , R] ] ] = [P , [1 , (0, 0, 0, s, 0, 2r)] ]
= [P , (0, 0, 0, 2r, 2s, 0)] = (0, 2rP, 0, 0, 0, 0).
Hence this case is reduced to the case (ii) above.
(iv) Case R = (0, 0, 0, 0, s, t), s 6= 0. We have
a [1 , R] = (0, 0, 0, s, 0, 0).
Hence this case is reduced to the case (iii) above.
(v) Case R = (0, 0, 0, 0, 0, t), t 6= 0. The argument is similar to (iv).
Consequently, we have a = e8 C , which proves the simplicity of e8 C .
For R e8 C we denote adR : e8 C e8 C , that is, adR(R1 ) = [R, R1 ], by (R) =

adR. Since e8 C is simple (Theorem 5.2.1), we obtain an isomorphism of Lie algebras
e8 C
= (e8 C ) = {(R) | R e8 C }
by assigning (R) to R. Moreover, e8 C is isomorphic to the algebra
Der(e8 C ) = { HomC (e8 C ) | [R1 , R2 ] = [R1 , R2 ] + [R2 , R2 ]}.
Hereafter we often denote (R) by R identifying e8 C
= Der(e8 C ).
Definition. We define a symmetric inner product (R1 , R2 )8 in e8 C by
(R1 , R2 )8 = (1 , 2 )7 {Q1 , P2 } + {P1 , Q2 } 8r1 r2 4t1 s2 s1 t2 ,
where Ri = (i , Pi , Qi , ri , si , ti ) e8 C .
Lemma 5.3.1. The inner product (R1 , R2 )8 of e8 C is e8 C -adjoint invariant :
([R, R1 ], R2 )8 + (R1 , [R, R2 ])8 = 0,
Proof. ([R, R1 ], R2 )8
159
R, Ri e8 C .
[, 1 ] + P Q1 P1 Q
2
P1 1 P + rP1 r1 P + sQ1 s1 Q
P
Q1 1 Q rQ1 + r1 Q + tP1 t1 P 2 !
Q2
1
1
{P,
Q
}
+
{P
,
Q}
+
st
s
t
=
1
1
1
1

8
8
r2 8

{P, P1 } + 2rs1 2r1 s
s2
4
1
t2
{Q, Q1 } 2rt1 + 2r1 t
4
= (using ([, 1 ], 2 )7 + (1 , [, 2 ])7 = 0 (Lemma 4.5.1.(1)),
(, P Q)7 = {P, Q} (Lemma 4.5.1.(2)) etc.)
= (R1 , [R, R2 ])8 .
B8 (R1 , R2 )
= 15(R1 , R2 )8
= 15(1 , 2 )7 + 15{Q1, P2 } 15{P1 , Q2 } + 120r1 r2 + 60t1 s2 + 60s1 t2
5
= B7 (1 , 2 ) + 15{Q1, P2 } 15{P1 , Q2 } + 120r1 r2 + 60t1 s2 + 60s1 t2 ,
3
where Ri = (i , Pi , Qi , ri , si , ti ) e8 C and B7 is the Killing form of e7 C .
Proof. Since e8 C is simple (Theorem 5.2.1), there exist k C such that
B8 (R1 , R2 ) = k(R1 , R2 )8 ,
Ri e8 C .
To determine k, let R1 = R2 = (0, 0, 0, 1, 0, 0) = 1. Then, we have

(1, 1)8 = 8.
On the other hand, since
[1, [1, (, P, Q, r, s, t)] ] = [1, (0, P, Q, 0, 2s, 2t)] = (0, P, Q, 0, 4s, 4t),
we have
B8 (1, 1) = 56 2 + 4 2 = 120.
Therefore k = 15. Thus we have B8 (R1 , R2 ) = 15(R1 , R2 )8 .
Definition. The group E8 C is defined to be the automorphism group of the Lie
algebra e8 C :
E8 C = { IsoC (e8 C ) | [R1 , R2 ] = [R1 , R2 ]}.
160
Theorem 5.4.1. The group E8 C is connected.

Proof. Denote by Inn(e8 C ) the subgroup generated by the inner automorphisms
exp((R)), R e8 C in the automorphism group Aut(e8 C ) = E8 C of the Lie algebra
e8 C . It is known that Aut(e8 C )/Inn(e8 C ) = {1} holds for the C-algebra of E8 type
(see for example, Matsushima [19]), that is,
Aut(e8 C ) = Inn(e8 C ).
Since Inn(e8 C ) is connected, E8 C which is equal to Inn(e8 C ) is also connected.
Remark. We can also prove the connectedness of the group E8 C from the following fact. For R e8 C , we define a C-linear mapping R R : e8 C e8 C by
(R R)R1 = (R)2 R1 +
1
B8 (R, R1 )R,
30
R1 e8 C ,
and we define a space WC by

WC = {R e8 C | R R = 0, R 6= 0}.
Then we have
E8 C /(E8 C )1 WC ,
where (E8 C )1 = { E8 C | 1 = 1 } = exp((0, 0, PC , 0, 0, C))E7 . The connectedness of E8 C follows from the connectedness of (E8 C )1 and WC . (See Imai and
Yokota [13]).
The Lie algebra of the group E8 C is Der(e8 C )
= e8 C , and therefore we have shown
C
that E8 is a complex Lie group of type E8 , since we will show in Theorem 5.6.2
that e8 C is a Lie algebra of type E8 . It is known by the general theory of Lie groups,
that if a complex Lie group of type E8 is connected, then it is simply connected, and
hence we have obtained the following result.
Theorem 5.4.2. E8 C is a simply connected complex Lie group of type E8 .
We define C-linear transformations , of e8 C respectively by
(, P, Q, r, s, t) = (1 , P, Q, r, s, t),
(, P, Q, r, s, t) = (, Q, P, r, t, s),
where in the right hand side is the same as E7 defined in Section 4.3. The
mappings and preserve the Lie bracket in e8 C , that is, , Aut(e8 C ) = E8 C .
We set
e = = .
161
Finally, we denote by the complex cojugation in e8 C , that is,

(, P, Q, r, s, t) = ( , P, Q, r, s, t),
where in the right hand side is the usual complex cojugation in the complexification.
Definition. We define a Hermitian inner product hR1 , R2 i in e8 C by
hR1 , R2 i =
1
e 1 , R2 ).
B8 ( R
15
Proposition 5.5.1. The Hermitian inner product hR1 , R2 i in e8 C is positive

definite.
e 1 = ( 1 1 ,
Proof. Let Ri = (i , Pi , Qi , ri , si , ti ) e8 C , i = 1, 2. Since R
Q1 , P1 , r1 , t1 , s1 ), we have, by Theorem 5.3.2, that
hR1 , R2 i
= ( 1 1 , 2 )7 + hP1 , P2 i + hQ1 , Q2 i + 8( r1 )r2 + 4( s1 )s2 + 4( t1 )t2 .
Hence, it is sufficient to show that ( 1 1 , 2 )7 is positive definite. Let i =
(i , Ai , Bi , i ), i = 1, 2. Since 1 1 = ( t 1 , B1 , A1 , 1 ), we have
8
( 1 1 , 2 )7 = 2( t 1 , 2 )6 + 4hA1 , A2 i + 4hB1 , B2 i + ( 1 )2 .
3
Therefore, it is enough to show that ( t 1 , 2 )6 is positive definite. Let i = i + Tei
e6 C , i f4 C , Tei J0 C , i = 1, 2. Since t 1 = 1 + Te1 , we have
( t 1 , 2 )6 = ( 1 , 2 )4 + hT1 , T2 i.
Consequently, it is sufficient to show that ( 1 , 2 )4 is positive definite, which can

be seen, however, from the fact that the following set
e1 , Fe2 (ei )],

2[E
e1 , Fe3 (ei )],

2[E
1
[Fe1 (ei ), Fe1 (ej )],
2
e3 , Fe1 (ei )],

2[E
0 i 7,
0i<j7
forms an orthonormal C-basis of f4 C with respect to the inner product ( 1 , 2 )4 .

Thus the proposition is proved.
Definition. We define a group E8 by
E8 = { E8 C | hR1 , R2 i = hR1 , R2 i}
e = }.
e
= { E8 C |
Theorem 5.5.2. The group E8 is a compact Lie group.
162

U (248) = U (e8 C ) = { IsoC (e8 C ) | hR1 , R2 i = hR1 , R2 i}.
Theorem 5.5.3. The Lie algebra e8 of the group E8 is
e = R}
e8 = {R e8 C | R
= {(, P, P, r, s, s) e8 C | e7 , P PC , r iR, s C}.
Proof. For R = (, P, Q, r, s, t) e8 C , since

e = ( 1 , Q, P, r, t, s).
R
e = R is equivalent to = , Q = P, r = r, t = s,
the condition R
hence we have the theorem.
Proposition 5.5.4. The complexification of the Lie algebra e8 is e8 C . Hence e8
is simple.
Proof. For R e8 C , the conjugate transposed mapping R of R with respect to
e
e e8 C , and for R e8 C , R belongs
the inner product hR1 , R2 i of e8 C is R = R
to e8 if and only if R = R. Now, any element R e8 C is represented by
R=
R R
R + R
+i
,
2
2i
R R R + R
,
e8 .
2
2i
Hence e8 C is the complexification of e8 . Since e8 C is simple (Theorem 5.2.1), e8 is

also simple.
Analogously as in e8 C , for R e8 , we identify R with (R) and regard e8
= (e8 ).
E8 C E8 R248 .
In particular, the group E8 is simply connected.
Proof. Evidently E8 C is an algebraic subgroup of IsoC (e8 C ) = GL(248, C). For
E8 C , the conjugate transposed mapping of with respect to the inner product
e 1
e E8 C . Therefore, by Chevalleys lemma, we have
hR1 , R2 i is =
E8 C (E8 C U (e8 C )) Rd = E8 Rd ,
d = 248.
Since E8 C is simply connected (Theorem 5.4.2), E8 is also simply connected.

5.6 Roots of e8 C
Theorem 5.6.1. The rank of the Lie algebra e8 C is 8. The roots of e8 C relative
to some Cartan subalgebra are given by
163
(k l ), (k + l ), 0 k < l 3,
1
0 k 3,
k (2 3 ),
2
1
1
(0 1 + 2 3 ) (3 1 ),
2
2
1
1
(0 + 1 + 2 3 ) (3 1 ),
2
2
1
1
(0 + 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 + 3 ) (3 1 ),
2
2
1
1
( 0 1 + 2 3 ) (1 2 ),
2
2
1
1
(0 + 1 + 2 3 ) (1 2 ),
2
2
1
1
( 0 + 1 + 2 + 3 ) (1 2 ),
2
2
1
1
(0 1 + 2 + 3 ) (1 2 ),
2
2

2
0 k < l 3,
k + ,
3
1

2
k
1 , 0 k 3,
2
3
1
2
1
2 ,
(0 1 + 2 3 )
2
3
21
1
2
( 0 + 1 + 2 3 )
2 ,
2
2
3
1
1
2
(0 + 1 + 2 + 3 )
2 ,
2
3
21
1
2
( 0 1 + 2 + 3 )
2 ,
2
2
3
1
1
2
( 0 1 + 2 3 )
3 ,
2
3
12
1
2
(0 + 1 + 2 3 )
3 ,
2
2
3

2
1
1
3 ,
( 0 + 1 + 2 + 3 )
2
3
12
1
2
(0 1 + 2 + 3 )
3 ,
2
2
3

1
j + r ,
1 j 3,
1 3 1
k
1 + r, 0 k 3,
2
3
1
1
1
( 0 + 1 2 3 )
2 + r,
2
3
12
1
1
2 + r,
( 0 + 1 2 3 )
2
2
3
164
1
(
2
1
(
2
1
(
2
1
(
2
1
(
2
1
(
2
with 1 + 2 + 3 = 0.
1
2 + r,
3
12
1
0 + 1 2 + 3 )
2 + r,
2
3
1
1
0 + 1 + 2 + 3 )
3 + r
3
12
1
0 + 1 2 3 )
3 + r,
2
3
1
1
0 1 + 2 + 3 )
3 + r,
3
12
2
0 1 2 + 3 )
3 + r,
2
3
2r,
r
0 + 1 + 2 3 )
1
Proof. Let h7 be the Cartan subalgebra of e7 C given in Theorem 4.6.1, then

h = {((h), 0, 0, r, 0, 0) | (h) h7 , r C}
is an abelian subalgebra of e8 C (it will be a Cartan subalgebra of e8 C ).
I The roots of e7 C are also roots of e8 C . Indeed, we have
[((h), 0, 0, r, 0, 0), (, 0, 0, 0, 0, 0)] = ([(h), ], 0, 0, 0, 0, 0).
II We have
[((h), 0, 0, r, 0, 0), (0, P, 0, 0, 0, 0)] = (0, ((h) + r)P, 0, 0, 0, 0),
and using the same notation as in Theorem 4.6.1, we also have
e 0, 0, ) + r1)(X, Y, , )
((h + H,

e + 1 + r Y, ( + r), ( + r) .
e 1 + r X, h H
= h + H
3
3
By putting Y = 0, = = 0, we obtain
1
the root k + r
3
1
1
the root k 1 + r
2
3
by letting X = Ek ,
by letting X = F1 (a), a = ek ie4+k .
We can also obtain roots by letting X = F2 (a), F3 (a).

By putting X = 0, = = 0 and further Y = Ek , Y = Fi (a), we can again roots.
By putting X = Y = 0, = 1, = 0, we can again the root + r. Similarly, we can
obtain the roort + r.
By using
[((h), 0, 0, r, 0, 0), (0, 0, Q, 0, 0, 0)] = (0, 0, ((h) r)Q, 0, 0, 0),
165
we can also obtain roots.

III From the relations
[((h), 0, 0, r, 0), (0, 0, 0, 0, 1, 0)] = (0, 0, 0, 0, 2r, 0),
[((h), 0, 0, r, 0), (0, 0, 0, 0, 0, 1)] = (0, 0, 0, 0, 0, 2r),
we obtain the roots 2r and 2r.
1
1
(0 1 2 3 ) + (3 1 ),
2
2
1
2 = 1 r, 3 = 2r,
3
1
1
4 = 2 r, 5 = 3 (2 3 ),
3
2
6 = 2 3 , 7 = 1 2 , 8 = r
1 =

= 21 + 42 + 63 + 54 + 45 + 36 + 27 + 38
8
2

+ n8 8 . Now, all positive roots of e8 C are expressed by
0 1
0 2
0 3
1 2
1 3
2 3
=2
=2
=2
=0
=0
=0
3
3
3
0
0
0
4
4
4
0
0
0
3
3
3
0
0
0
2
2
2
0
0
0
1
1
2
0
1
1
0
1
1
1
1
0
2
2
2
0
0
0
1
0 + (2 3 ) = 2
2
1
1 + (2 3 ) = 0
2
0 + 1
0 + 2
0 + 3
1 + 2
1 + 3
2 + 3
=2
=2
=2
=0
=0
=0
4
4
4
1
1
1
6
6
6
2
2
2
6 5
3 2
1 3
2 2
1 1
1 1
166
5
5
5
2
2
2
4
4
4
2
2
2
3
3
2
2
1
1
2
1
1
1
1
0
3
3
3
1
1
1
1
2 + (2 3 ) = 0 1 2 2
2
1
3 + (2 3 ) = 0 1 2 2
2
1
0 (2 3 ) = 2 3 4 3
2
1
1 (2 3 ) = 0 0 0 0
2
1
2 (2 3 ) = 0 0 0 0
2
1
3 (2 3 ) = 0 0 0 0
2
1
1
(0 + 1 + 2 3 ) + (3 1 ) = 1
2
2
1
1
(0 + 1 2 + 3 ) + (3 1 ) = 1
2
2
1
1
(0 1 + 2 + 3 ) + (3 1 ) = 1
2
2
1
1
(0 1 2 3 ) + (3 1 ) = 1
2
2
1
1
(0 + 1 + 2 3 ) (3 1 ) = 1
2
2
1
1
(0 + 1 2 + 3 ) (3 1 ) = 1
2
2
1
1
(0 1 + 2 + 3 ) (3 1 ) = 1
2
2
1
1
(0 1 2 3 ) (3 1 ) = 1
2
2
1
1
(0 1 + 2 3 ) + (1 2 ) = 1
2
2
1
1
(0 1 2 + 3 ) + (1 2 ) = 1
2
2
1
1
(0 + 1 2 3 ) + (1 2 ) = 1
2
2
1
1
(0 + 1 + 2 + 3 ) + (1 2 ) = 1
2
2
1
1
(0 1 + 2 3 ) (1 2 ) = 1
2
2
1
1
(0 1 2 + 3 ) (1 2 ) = 1
2
2
1
1
(0 + 1 2 3 ) (1 2 ) = 1
2
2
1
1
(0 + 1 + 2 + 3 ) (1 2 ) = 1
2
2
2
1 + = 0 1 1 0 0
3
2
2 + = 0 0 1 1 0
3
2
3 = 0 1 1 1 0
3
167
1 1
0 1
1 0
0 1
3 2
1 2
1 1
1 0
1 1
0 0
1 0
0 0
1 2
2 2
2 1
1 2
2 2
1 1
1 2
2 2
1 0
0 0
0 0
0 0
3 4
3 2
2 1
3 4
3 2
1 1
3 4
3 2
1 0
2 2
1 0
0 0
2 2
1 1
1 0
2 2
1 1
0 0
2 2
1 1
1 1
3 4
3 3
2 1
1 2
2 1
1 0
1 2
2 1
0 0
1 2
2 1
1 1
2 4
4 3
2 1
0 1
0 1
0 0
1
0 + 1
2
1
0 1 +
2
1
1 + 1
2
1
1 1 +
2
3
2
3
2
3
2
1
2 + 1
2
1
2 1 +
2
1
3 + 1
2
1
3 1 +
2
3
2
3
2
3
2
1
(0 + 1 + 2 3 ) +
2
1
(0 + 1 2 + 3 ) +
2
1
(0 1 + 2 + 3 ) +
2
1
(0 1 2 3 ) +
2
1
(0 + 1 + 2 3 )
2
1
(0 + 1 2 + 3 )
2
1
(0 1 + 2 + 3 )
2
1
(0 1 2 3 )
2
1
(0 + 1 + 2 + 3 ) +
2
1
(0 + 1 2 3 ) +
2
1
(0 1 + 2 3 ) +
2
1
(0 1 2 + 3 ) +
2
1
(0 + 1 + 2 + 3 )
2
1
(0 + 1 2 3 )
2
1
(0 1 + 2 3 )
2
1
(0 1 2 + 3 )
2
=2
4 5
3 2
1 2
=2
3 5
3 2
1 3
=0
1 1
1 1
1 0
=0
0 1
1 1
1 1
=0
1 1
1 1
0 0
=0
0 1
1 1
0 1
=0
1 1
1 0
0 0
=0
0 1
1 0
0 1
1
2
2
1
2
2
1
2
2
1
2
2
1
2 +
2
1
2 +
2
1
2
2
1
2 +
2
1
3
2
1
3
2
1
3
2
1
3
2
1
3 +
2
1
3 +
2
1
3 +
2
1
3 +
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
3
2
168
=1
2 3
3 2
2 1
=1
2 3
3 2
1 1
=1
2 3
3 2
1 0
=1
1 1
1 0
0 0
=1
2 3
2 2
2 1
=1
2 3
2 2
1 1
=1
2 3
2 2
1 0
=1
1 1
0 0
0 0
=1
2 3
3 3
2 1
=1
1 1
1 1
1 1
=1
1 1
1 1
1 0
=1
1 1
1 1
0 0
=1
3 5
4 3
2 1
=1
2 3
2 1
1 1
=1
2 3
2 1
1 0
=1
2 3
2 1
0 0
1
1 + r = 0
3
1
2 + r = 0
3
1
3 + + r = 0
3
1
1 r = 0
3
1
2 + r = 0
3
1
3 r = 0
3
1 1
0 0
0 0
0 1
1 0
0 0
1 2
1 0
0 1
1 0
0 0
0 0
0 0
1 0
0 0
1 1
1 0
0 1
1
+r =2 4 6 4
3
1
r =2 4 5 4
3
1
+r =2 3 5 4
3
1
r =2 3 4 4
3
1
+r =0 1 2 1
3
1
r =0 1 1 1
3
1
+r =0 0 1 1
3
1
r =0 0 0 1
3
1
+r =0 1 2 1
3
1
r =0 1 1 1
3
1
+r =0 0 1 1
3
1
r =0 0 0 1
3
1
+r =0 1 2 1
3
1
r =0 1 1 1
3
1
+r =0 0 1 1
3
1
r =0 0 0 1
3
1
1
1
(0 + 1 + 2 3 ) + 2 + + r = 1
2
2
3
1
1
1
(0 + 1 + 2 3 ) + 2 + r = 1
2
2
3
1
0 + 1 +
2
1
0 + 1 +
2
1
0 1
2
1
0 1
2
1
1 + 1 +
2
1
1 + 1 +
2
1
1 1
2
1
1 1
2
1
2 + 1 +
2
1
2 + 1 +
2
1
2 1
2
1
2 1
2
1
3 + 1 +
2
1
3 + 1 +
2
1
3 1
2
1
3 1
2
169
2 1
2 1
2 1
2 1
1 1
1 1
1 1
1 1
1 0
1 0
1 0
1 0
0 0
0 0
0 0
0 0
4 3
2 2
1 2
3 3
2 2
1 2
1
(0 + 1 + 2 3 )
2
1
(0 + 1 + 2 3 )
2
1
(0 + 1 2 + 3 ) +
2
1
(0 + 1 2 + 3 ) +
2
1
(0 + 1 2 + 3 )
2
1
(0 + 1 2 + 3 )
2
1
(0 1 + 2 + 3 ) +
2
1
(0 1 + 2 + 3 ) +
2
1
(0 1 + 2 + 3 )
2
1
(0 1 + 2 + 3 )
2
1
(0 1 2 3 ) +
2
1
(0 1 2 3 ) +
2
1
(0 1 2 3 )
2
1
(0 1 2 3 )
2
1
(0 + 1 2 3 ) +
2
1
(0 + 1 2 3 ) +
2
1
(0 + 1 2 3 )
2
1
(0 + 1 2 3 )
2
1
(0 1 + 2 3 ) +
2
1
(0 1 + 2 3 ) +
2
1
(0 1 + 2 3 )
2
1
(0 1 + 2 3 )
2
1
(0 + 1 + 2 + 3 ) +
2
1
(0 + 1 + 2 + 3 ) +
2
1
2
2
1
2
2
1
2 +
2
1
2 +
2
1
2
2
1
2
2
1
2 +
2
1
2 +
2
1
2
2
1
2
2
1
2 +
2
1
2 +
2
1
2
2
1
2
2
1
3 +
2
1
3 +
2
1
3
2
1
3
2
1
3 +
2
1
3 +
2
1
3
2
1
3
2
1
3 +
2
1
3 +
2
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
170
=1 2
3 2
2 2
1 1
=1 2
2 2
2 2
1 1
=1 2
4 3
2 1
1 2
=1 2
3 3
2 1
1 2
=1 2
3 2
2 1
1 1
=1 2
2 2
2 1
1 1
=1 2
4 3
2 1
0 2
=1 2
3 3
2 1
0 2
=1 2
3 2
2 1
0 1
=1 2
2 2
2 1
0 1
=1 1
2 1
0 0
0 1
=1 1
1 1
0 0
0 1
=1 1
1 0
0 0
0 0
=1 1
0 0
0 0
0 0
=1 1
2 1
1 1
1 1
=1 1
1 1
1 1
1 1
=1 2
3 2
1 1
1 1
=1 2
2 2
1 1
1 1
=1 1
2 1
1 1
0 1
=1 1
1 1
1 1
0 1
=1 2
3 2
1 1
0 1
=1 2
2 2
1 1
0 1
=1 2
4 3
3 2
1 2
=1 2
3 3
3 2
1 2
1
1
(0 + 1 + 2 + 3 ) 3
2
2
1
1
(0 + 1 + 2 + 3 ) 3
2
2
1
1
(0 1 2 + 3 ) + 3 +
2
2
1
1
(0 1 2 + 3 ) + 3 +
2
2
1
1
(0 1 2 + 3 ) 3
2
2
1
1
(0 1 2 + 3 ) 3
2
2
r =0 0
2r
=0 0
+r =0 0
1
+r
3
1
r
3
1
+r
3
1
r
3
1
+r
3
1
r
3
0 0
1 0
1 0
=1 3
5 4
3 2
1 2
=1 3
4 4
3 2
1 2
=1 1
2 1
1 0
0 1
=1 1
1 1
1 0
0 1
=1 2
3 2
1 0
0 1
=1 2
2 2
1 0
0 1
0
0
0
0 0
0 0
0 0
1
0
1.
Hence = {1 , 2 , , 8 } is a fundamental root system of e8 C . The real part hR

of h is
hR = {((h), 0, 0, r, 0) | (h) (h7 )R , r R},
3
3

X
X

j Ej , 0, 0, (h7 )R (Theorem 4.6.2) . The
k Hk +
where (h) = {
j=1
k=0
Killing form B8 of e8 C on hR is given by

B8 (e
h, e
h ) = 60
3
X
k k + 30
3
X
j j + 40 + 120rr ,
j=1
k=0
3
X
e
e
for h = ((h), 0, 0, r, 0), h = ((h ), 0, 0, r , 0) h, where (h) =
k Hk +
3
X
j=1
j Ej
k=0
3
3

X
X
, 0, 0, , (h ) = (
j Ej , 0, 0, (h7 )R , Indeed,
k Hk +
j=1
k=0
5
B8 (e
h, e
h ) = B7 ((h), (h )) + 120rr (Theorem 5.3.2)
3
3
3

X
5 X
j j + 4 + 120rr (Theorem 4.6.2)
k k + 3
= 6 6
3
j=1
k=0
= 60
3
X
k=0
k k + 30
3
X
j j + 40 + 120rr .
j=1
Now, the canonical elements Hi hR associated with i (B8 (H , H) = (H), H

hR ) are determined as follows.
1

H1 =
(H0 H1 H2 H3 ) + 2(E3 E1 ) , 0, 0, 0 , 0, 0, 0, 0, 0 ,
120

1
1
1
, 0, 0,
(2E1 E2 E3 ) , 0, 0,
, 0, 0 ,
H2 =
90
120
120
171

1
H3 = 0, 0, 0, , 0, 0 ,
60

1
1
1
, 0, 0,
(E1 + 2E2 E3 ) , 0, 0,
, 0, 0 ,
H4 =
90
120
120

1
(H3 (E2 E3 ) ), 0, 0, 0 , 0, 0, 0, 0, 0 ,
H5 =
60
1

H6 =
(H2 H3 ), 0, 0, 0 , 0, 0, 0, 0, 0 ,
60

1
(H1 H2 ), 0, 0, 0 , 0, 0, 0, 0, 0 ,
H7 =
60

1
1
, 0, 0,
, 0, 0 .
H8 = 0, 0, 0,
40
120
Thus we have
(1 , 1 ) = B7 (H1 , H1 ) = 60
1 1
1
1 1
4 + 30
8=
,
120 120
120 120
30
and the other inner products are similarly calculated. Consequently, the inner product
(i , i ) =
1
,
30
i = 1, 2, 3, 4, 5, 6, 7, 8,
1
1
, i = 1, 2, , 6, (3 , 8 ) = ,
60
60
(i , j ) = 0, otherwise,
1
1
(, ) =
, (, 7 ) = , (, i ) = 0, i = 1, 2, 3, 4, 5, 6, 8,
30
60
(i , i+1 ) =
e8 C .
According to Borel-Siebenthal theory, the Lie algebra e8 has five subalgebras as
(1) The first one is a subalgebra of type A1 E7 which is obtained as the fixed
(2) The second one is a subalgebra of type D8 which is obtained as the fixed points
e of e8 .
of an involution
(3) The third one is a subalgebra of type A2 E6 which is obtained as the fixed
points of an automorphism w of order 3 of e8 .
(4) The fourth one is a subalgebra of type A8 which is obtained as the fixed points
of an automorphism w3 of order 3 of e8 .
(5) The fifth one is a subalgebra of type A4 A4 which is obtained as the fixed
points of an automorphism z5 of order 5 of e8 .

5.7.6, 5.8.7, 5.10.2, 5.11.7 and 5.12.5, respectively. As for Theorems 5.10.2, 5.11.7 and
5.12.5, we refer to Gomyo [9].
172
5.7. Involution and subgroup (SU (2) E7 )/Z 2 of E8

We shall first study the following subgroup (E8 C )1,1 ,1 of E8 C :
(E8 C )1,1 ,1 = { E8 C | 1 = 1, 1 = 1 , 1 = 1 }.
(E8 C )1,1 ,1
= E7 C .
Proposition 5.7.1.
Proof. For E7 C , we define a C-linear mapping e : e8 C e8 C by
Ad
0
0
e =
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
,
0
0
1
where (Ad) = 1 , e7 C . It is easy to check that e (E8 C )1,1 ,1 . Conversely, if E8 C satisfies 1 = 1, 1 = 1 and 1 = 1 , then is of the
form
1 HomC (e7 C ),
1 12 13 0 0 0
21 2 23 0 0 0
2 , 3 , 23 , 32 HomC (PC ),
21 , 31 HomC (e7 C , PC ),
31 32 3 0 0 0
=
,
a1
b1
c1 1 0 0
12 , 13 HomC (PC , e7 C ),
a2
b2
c2 0 1 0
ai HomC (e7 C , C),
a3
b3
c3 0 0 1
bi , ci HomC (PC , C).
From the relation [, 1] = [, 1] = [, 1] = 0, that is,
0 = [(1 , 21 , 31 , a1 , a2 , a3 ), (0, 0, 0, 1, 0, 0)]
= (0, 21 , 31 , 0, 2a2 , 2a3 ),
we obtain 21 = 31 = 0 and a2 = a3 = 0. Furthermore, from [, 1 ] = [, 1 ] =
[, 1 ] = 0, that is,
0 = [(1 , 0, 0, a1 , 0, 0), (0, 0, 0, 0, 1, 0)] = (0, 0, 0, 0, 2a1, 0),
we obtain a1 = 0. Using the fact that [P , 1] = [P , 1] = [P , 1] = P ,
that is,
(12 P, 2 P, 32 P, b1 P, b2 P, b3 P )
= [(12 P, 2 P, 32 P, b1 P, b2 P, b3 P ), (0, 0, 0, 1, 0, 0)]
= (0, 2 P, 32 P, 0, 2b2P, 2b3 P ),
we obtain 12 = 32 = 0 and b1 = b2 = b3 = 0. Similarly, from [Q , 1] =
[Q , 1] = [Q , 1] = Q , we obtain 13 = 23 = 0 and c1 = c2 = c3 = 0.
173
Thus we have seen that is of the form
1
0
=
0
0
0
2
0
0
0
0
3
0
0
0
.
0
E

1
By applying on [(0, P, 0, 0, 0, 0), (0, 0, Q, 0, 0, 0)] = P Q, 0, 0, {P, Q}, 0, 0 , we
8
obtain
1 (P Q) = 2 P 3 Q,
{2 P, 3 Q} = {P, Q},
(i)

1
since [(0, 2 P, 0, 0, 0, 0), (0, 0, 3Q, 0, 0, 0)] = 1 (P Q), 0, 0, {P, Q}, 0, 0 . Again,
8

1
by applying on [(0, P, 0, 0, 0, 0), (0, Q, 0, 0, 0, 0)] = 0, 0, 0, {P, Q}, 0, 0 , we obtain
4
{2 P, 2 Q} = {P, Q}.
(ii)
Further, by applying on [(, 0, 0, 0, 0, 0), (0, P, 0, 0, 0, 0)] = (0, P, 0, 0, 0, 0), we obtain

(1 )(2 P ) = 2 (P ).
(iii)
From (i), (ii), we have

{2 P, 3 Q} = {2 P, 2 Q},
P, Q PC ,
hence 2 = 3 , which we denote by . If we put 1 P in (iii) in the place of P , we

obtain
1 = 1 .
Therefore, from (i), we have
(P Q) 1 = P Q,
which implies that E7 C . Thus the proof of Theorem 5.4.1 is completed.
Definition. We define a C-linear mapping : e8 C e8 C by
(, P, Q, r, s, t) = (. P, Q, r, s, t).
Then E8 and 2 = 1. Note that is the central element 1 of E7 regarding as
an element of E8 (see Theorem 5.7.3).
(E8 ) = { E8 | = }.
174
Lemma 5.7.2. If E8 satisfies 1 = 1 , then it also satisfies 1 = 1 and

1 = 1 .
Proof. Let 1 = (, P, Q, r, s, t). From the relation [1, 1 ] = [1, 1 ] =

[1, 1 ] = 21 = 21. we have
21 = [(, P, Q, r, s, t), 1 ] = (0, 0, P, s, 0, 2r),
from which we obtain P = 0, s = 0, r = 1. Further, h1, 1i = h1, 1i = 8, that is,
h, i + hQ, Qi + 8 + 4( t)t = 8, which implies that = 0, Q = 0, t = 0. Therefore
1 = 1. By applying on [1 , 1 ] = 1, we obtain 1 = 1 by a similar method to
the above.
To find the structure of the group (E8 ) , we shall first study the following subgroup
(E8 )1 of E8 :
(E8 )1 = { E8 | 1 = 1 }.
Theorem 5.7.3.
(E8 )1
= E7 .
Proof. (E8 )1 = (E8 )1,1 ,1 (Lemma 5.7.2)

e = }
e
= { (E8 C )1,1 ,1 |
e = }
e (Proposition 5.7.1)
= { E7 C |
= { E7 C | = } (by the correspondence to Proposition 5.7.1)

= E7 (Lemma 4.3.3.(4)).
Remark. We define a space W1 by

W1 = {R e8 C | R R = 0, hR, Ri = 4}
(see Remark of Theorem 5.4.1), then, we obtain a homeomorphism
E8 /E7 W1 .
(See Yokota, Imai and Yasukura [53]).
Theorem 5.7.4. The group (E8 ) contains a subgroup
3 (SU (2)) = {3 (A) E8 | A SU (2)}
t
which is isomorphic
to the

group SU (2) = {A M (2, C) | ( A)A = E, detA = 1},
a b
where, for A =
SU (2), 3 (A) : e8 C e8 C is defined by
b a
1 0
0
0
0
0
0 a1 b1
0
0
0
0 b1 a1
0
0
0
.
3 (A) =
0
( a)a ( b)b ( a)b a( b)
0 0
0 0
0
2a( b)
a2
( b)2
0 0
0
2( a)b
b2
( a)2
175

a b
i
= exp
b a
exp((0, 0, 0, i, , )) (E8 ) .
Proof. For A =
SU (2), we have (A) =

Proof. (E8 ) is connected as a fixed points subgroup of the involutive automorphism of the simply connected Lie group E8 .
Theorem 5.7.6.
(E8 )
= (SU (2) E7 )/Z 2 , Z 2 = {(E, 1), (E, 1)}.
Proof. We define a mapping : SU (2) E7 (E8 ) by

(A, ) = 3 (A).
Since 3 (A) 3 (SU (A)) and E7 commute, is a homomorphism. Since (E8 )
is connected (Lemma 5.7.5), to prove that is onto, it is sufficient to show that the
differential mapping : su(8) e7 (e8 ) of is onto. But which is not difficult to
see. Indeed, we have
(e8 ) = {(R) (e8 ) | (R) = (R)}
= {R e8 | R = R}
= {(, 0, 0, r, s, s) | e7 , r iR, s C}.
Ker = {(E, 1), (E, 1)} = Z 2 is easily obtained. Thus we have the isomorphism
(SU (2) E7 )/Z 2
= (E8 ) .
Remark. We can prove directly that is onto without using the connectedness
of (E8 ) (Lemma 5.7.5), (see Imai and Yokota [13]).
e and subgroup Ss(16) of E8
5.8. Involution
e : e8 C e8 C by
We define a C-linear mapping
e
(,
P, Q, , s, t) = (1 , Q, P, r, t, s).
e E8 and ()
e 2 = 1.
Then
We shall study the following subgroup (E8 )e

of E8 :
e
e
= { E8 |
= }
(E8 )e
= { E8 | = } = (E8 ) .
We define an R-linear mapping l : M (8, C) M (16, R) by

xkl ykl
, xkl , ykl R.
l (xkl + iykl ) =
ykl xkl
176
Further, we define I, J M (16, R) by

1 0
I = diag(I, , I), I =
,
0 1
J = diag(J, , J), J =
0 1
1 0
then IJ = JI, and for X, Y M (8, C), we have

(1) l(XY ) = l(X)l(Y ),
(2) Il(X) = l( X)I, Jl(X) = l(X)J,
(3) t l(X) = l( tX).
Lemma 5.8.1. (1) l(u(8)) = {B so(16) | JB = BJ},
l(S(8, C))I = {B so(16) | JB = BJ}.
(2) Any element B so(16) is uniquely expressed by
B = l(D ) + l(S)I,
D u(8), S S(8, C)
= l(D) + l(S)I + l(icE), D su(8), S S(8, C), c R.
Proof. (1) If D u(8), then we have Jl(D) = l(D)J and t l(D) = l( tD) = l(D).
Conversely, suppose that B so(16) satisfies JB = BJ. Let B = l(D), D M (8, C).
Then, the relation
l(D) = B = t B = t l(D) = l( t D)
implies t D = D, that, is, D u(8). Next, for S S(8, C), we have
Jl(S)I = l(S)JI = l(S)IJ,
t
(l(S)I) = t I t (l(S)) = Il( t S) = Il( S) = l(S)I.
Conversely, suppose that B so(16) satisfies JB = BJ. Since the element BI
satisfies JBI = BIJ, we let BI = l(S), S M (8, C). Then the relation
l(S)I = B = t B = t I t (l(S)) = Il( t S) = l(t S)I
implies S = t S, that is, S S(8, C).
(2) Let B =
B + JBJ
B JBJ
+
and use (1) above.
2
2
The following Lemmas 5.8.2 and 5.8.3 are properties of the mappings : (PC )
S(8, C) and : su(8) (e6 ) of Section 4.12, and will be used in the proof of
Theorem 5.8.4.
Lemma 5.8.2 The Lie isomorphism : sp(4) (e6 ) defined by ( D)X =
g 1 (D(gX) + (gX)D ), X JC of Theorem 3.1.2 satisfies
([gX1 , gX2 ]) = 2(X1 X2 X2 X1 ),
177
X1 , X2 JC .
Proof. Note that [gX1 , gX2 ] sp(4). Now, since

g(2(X1 X2 )X)
X JC
1
= g((X2 , X)X1 + (X1 , X2 )X 4X2 (X1 X)) (Lemma 3.4.1)
3
1
= (X2 , X)gX1 + (X1 , X2 )gX 4gX2 (gX1 gX) + (X1 , X)gX2
3
+(X2 , X1 , X)E (Lemma 3.12.1),
we have
g(2(X1 X2 X2 X1 )X)
= 4gX2 (gX1 gX) + 4gX1 (gX2 gX)
= gX2 gX1 gX gX2 gXgX1 gX1 gXgX2 gXgX1 gX2

+gX1 gX2 gX + gX1 gXgX2 + gX2 gXgX1 + gXgX2gX1
= [gX1 , gX2 ]gX gX[gX1, gX2 ] = g(( [gX1 , gX2 ])X).

Consequently, since g is injective, we have the lemma.
Lemma 5.8.3. For S, S1 , S2 S(8, C), we have
(1)
(2)
(3)
1 (S) = 1 (iS).
tr(S1 S2 S2 S1 ) = 4i{1S1 , 1 S2 }.

1
(S1 S2 S2 S1 ) tr(S1 S2 S2 S1 )E
8
= 4(1 S1 1 S2 1 S2 1 S1 ).
Proof. (1) Let 1 S = P = (X, Y, , ). Then

1 S = (X, Y, , ) = (Y, X, , )

E
J
= k g(Y ) E + ik g (X)
2
2

= i k gX E + ik g(Y ) E J
2
2
= i(X, Y, , ) = iP = iS.
(2),(3) Let 1 Si = Pi = (Xi , Ti , i , i ), i = 1, 2. Noting
1
S1 S2 S2 S1 (S1 S2 S2 S1 ) su(8),
8
we have
S1 S2 = P1 P2 = (X1 , Y1 , 1 , 1 ) (X2 , Y2 , 2 , 2 )

1
2
1
= k gX1 E + ik g(Y1 ) E J k gX2 E
2
2
2

2
+ik g(Y2 ) E J
2
178

1
1
2
= k gX1 E + ik g(Y1 ) E
k gX2 E
2
2
2

2 2
ik g(Y2 ) E J
2

2
1
2
1
= k gX1 E gX2 E + g(Y1 ) E g(Y2 ) E
2
2
2
2

1
2
1
2
ik g(Y1 ) E gX2 E gX1 E g(Y2 ) E .
2
2
2
2
Therefore
S1 S2 S2 S1
= k(gX1 gX2 gX2 gX1 + g(Y1 )g(Y2 ) g(Y2 )g(Y1 ))
ik(g(Y1 )gX2 g(Y2 )gX1 gX1 g(Y2 ) + gX2 g(Y1 ))

i
1 gX2 + 2 gX1 + 1 g(Y2 ) 2 g(Y1 )) + (1 2 2 1 )E
2
= k([gX1 , gX2 ] [g(Y1 ), g(Y2 )])
(denote
+ik(g(2X1 Y2 2X2 Y1 + 1 X2 2 X1 1 Y2 + 2 Y1 ))
i
+ ((X1 , Y2 ) (X2 , Y1 ) + 1 2 2 1 ))E (Lemma 3.12.1)
2
D = [gX1 , gX2 ] [g(Y1 ), g(Y2 )] sp(4),
A = 2X1 Y2 2X2 Y1 + 1 X2 2 X1 1 X2 + 2 Y1 JC )
i
= kD + ik(gA) + {P1 , P2 }E.
2
Taking the trace of both sides, we obtain
tr(S1 S2 S2 S1 ) = 4i{P1 , P2 } = 4i{1 S1 , 1 S2 }.
and the expression above equal to
1
S1 S2 S2 S1 tr(S1 S2 S2 S1 ) = kD + ik(gA),
8
Therefore,

X2
Y1
X1 Y2
P1 P2 =
2
1
2
1
1
(Y1 Y2 X2 X1 )
(2X1 Y2 + 1 X2 + 2 Y1 )
4
=
.
1
(2Y1 X2 + 1 Y2 + 2 X1 )
4
1
((Y1 , Y2 ) (X2 , X1 ) 3(1 2 2 1 ))

8
179
P1 P2 P2 P1
1
(X1 X2 + X2 X1 Y1 Y2 + Y2 Y1 )
(2X1 Y2 2X2 Y1 + 1 X2 2 X1 + 2 Y1 1 Y2 )
= 4
(2X2 Y1 2X1 Y2 + 2 X1 1 X2 + 1 Y2 2 Y1 )
4
0
1

1
1
= D, A, A, 0 (Lemma 5.8.2)
4
4
4
1
= (kD + ik(gA))
4

1
1
= S1 S2 S2 S1 tr(S1 S2 S2 S1 )E .
4
8
is given by
of the Lie group (E )e
Theorem 5.8.4. The Lie algebra (e8 )e
8
e
e
(e8 )e
= { (e8 C ) |
= }
= {(, P, P, 0, s, s) | (e7 ) , P (PC ) , s R}
is isomorphic to the Lie algebra so(16) by the mapping : so(16) (e8 )e

and (e8 )e
defined by
(l(D) + l(S)I + l(icE)) = ( D, 21 S, 21 S, 0, 2c, 2c),

where D su(8), S S(8, C), c R, and : su(8) (e7 ) , : (PC )tau
S(8, C) are mappings defined in Section 4.12.
Proof. It is not difficult to see that the first half of the theorem and that is
onto. We will prove that preserve the Lie bracket.
(1) [l(D1 ), l(D2 )] = l[D1 , D2 ]
= ( [D1 , D2 ], 0, 0, 0, 0, 0) = ([ D1 , D2 ], 0, 0, 0, 0, 0)
= [( D1 , 0, 0, 0, 0, 0), ( D2 , 0, 0, 0, 0, 0)]
= [l(D1 ), l(D2 )].
(2) [l(D), l(S)I)] = (l(DS S D)I) = (l(DS + S t D)I)
= (0, 21 (DS + S t D), 21 (DS + S t D), 0, 0, 0).
On the other hand,
[l(D), (l(S)I)]
= [( D, 0, 0, 0, 0, 0), (0, 21S, 21 S, 0, 0, 0)]
= (0, 2( D)1 S, 2( D)1 S, 0, 0, 0).
Since ( D) = ( D) and ( D)1 S = 1 (DS + S t D), they are equal.
180
(3) [l(D), l(icE)] = [D, icE] = 0 = 0

= [( D, 0, 0, 0, 0, 0), (0, 0, 0, 0, 2c, 2c) = [l(D), l(icE)].
(4)
[l(S1 )I, l(S2 )I] = l(S1 S2 S2 S1 )

1
= l S1 S2 S2 S1 tr(S1 S2 S2 S1 )E
8

1
+l tr(S1 S2 S2 S1 )E
8

1
= S1 S2 S2 S1 tr(S1 S2 S2 S1 )E , 0, 0, 0,
8

i
i
tr(S1 S2 S2 S1 ), tr(S1 S2 S2 S1 )
4
4
= (4(1 S1 1 S2 1 S2 1 S1 ), 0, 0, 0,
{1 S1 , 1 S2 } {1 S1 , 1 S2 }) (Lemma 5.8.3).
On the other hand,
[l(S1 )I, (l(S2 )I]

= [(0, 21 S1 , 21 S1 , 0, 0, 0), (0, 21S2 , 21 S2 , 0, 0, 0)]

= 21 S1 21 S2 21 S2 21 S1 , 0, 0,
1
({21 S1 , 21 S2 } + {21 S2 , 21 S1 }),
8

1
1
{21 S1 , 21 S2 }, {21 S1 , 21 S2 } ,
4
4
which equals to the above.
(5) [(icE), l(S)I] = (2l(icS)I)

= (0, 41 (icS), 41 (icS), 0, 0, 0)
= (0, 41 (cS), 41 (cS), 0, 0, 0) (Lemma 5.8.3)
= [(0, 0, 0, 0, 2c, 2c), (0, 21S, 21 S, 0, 0, 0)]
= [l(icE), (l(S)I)].
Finally,
(6) [l(ic1 E), l(ic2 E)] = l[ic1 E, ic2 E] = 0 = 0
= (0, 0, 0, 0, 2c1, 2c1 ), (0, 0, 0, 0, 2c2, 2c2 ))]
= [l(ic1 E), l(ic2 E)].
Thus we have proved Theorem 5.8.4.
is connected as a fixed point subgroup under the involution

The group (E8 )e
e
of the simply connected Lie group E8 . Therefore, by Theorem 5.8.4, (E8 )e
is
isomorphic to one of the following groups
Spin(16),
SO(16),
Ss(16),
SO(16)/Z 2 .
Ss(16), below we will give an outline of the proof.

Precisely we have (E8 )e
=
181
We will use the Lie algebra

e8(8) = (e8 C ) = {R e8 C | R = R}
e
(see Section 5.13). Now, consider the eigenspace decomposition of e8 C by :
e8 C = (e8 C )e
(e8 C )e
,
e
e
e
(e8 C )e
= { Der(e8 C ) |
= }
= R} = (e8 C )e
,
= {R e8 C | R
e
e
e
= R},
(e8 C )e
= { Der(e8 C ) |
= }
= {R e8 C | R
Since we have
,
((e8 C )e
) = ((e8 C ) )e
= (e8(8) )e
= (e8(8) )e
e
((e8 C )e
) = ((e8 C ) )e
= (e8(8) )e
,
e
we obtain the following decomposition of e8(8) :
e8(8) = (e8(8) )e
(e8(8) )e
.

Since (e8(8) )e
= Ss(16) (Theorem 5.8.4), this is the Cartan decompposition of e8(8) .
e
to
] (e8(8) )e
, we obtain a representation of (e8(8) )e
Since [(e8(8) ) , (e8(8) )e
(e8(8) )e
:
(R)R1 = [R, R1 ], R (e8(8) )e

, R1 (e8(8) )e
.
which is irreducible. (See, for example, (8.5.1) of Goto and Grosshans [11]). Furthermore, the complex representation C of to (((e8(8) )e
)C = (e8 C )e
is also
irreducible, since (e8(8) )e

is
simple
(see
(8.8.3)
of
the
same
book).
The
following
lemma follows from above mentioned results.
2:
Lemma 5.8.5. The representation of the group (E8 )e

to (e8(8) )e
is irreducible.
Proposition 5.8.6. The center z((E8 )e

) of the group (E8 )e
is a group of order
) = {1, e
}.
z((E8 )e
e
Proof. Evidently, {1, }
z((E8 )e
). Conversely, let z((E8 )e
). Since
e
C
the representation of (E8 ) to (e8 )e
irreducible
(Lemma
5.8.5),
we
see,
by using
is constant.
Schurs lemma in the theory of groups, that the action of on (e8 C )e
Therefore, there exists an element k C such that

R = kR,
R (e8 C )e
.
Since the Killing form B8 (R, R ) is invariant under : B8 (R, R ) = B8 (R, R ), we

have
k 2 B8 (R, R ) = B8 (R, R ) = B8 (R, R ), R, R (e8 C )e
,
182

which implies that k 2 = 1. By Theorem 5.8.4, we have (e8 C )e
= so(16, C) which is
C e
C e
simple, and hence we see that (e8 ) is generated by (e8 ) :
nX
o
[Rk , Rl ] | Rk , Rl (e8 C )e
=
(e8 C )e
.
k,l
Consequently, satisfies k 2 1 = 1 on (e8 C )e

, that is, the identity mapping. When
e Thus we have proved the

k = 1, we have = 1 , when k = 1, we have = .
theorem.
It follows from Proposition 5.8.6, that (E8 )e

is isomorphic to one of the following
SO(16),
Ss(16).
There are only two, up to equivalence, complex irreducible representations of the

Lie algebra so(16) of dimension 128. In fact, one can obtain the following table, by
calculating the dimension of dominant roots by virtue of Weyls dimension formula
(see (7.5.9) of Goto and Grosshans [11]):
1
16
21
135
2 22 3
4
5
6
7 8
120 5304 560 1820 4368 8008 128 128
hence the dominant root of dimension 128 is either 7 or 8 . (Here, 1 , 2 , , 8

are fundamental weights). On the other hand, Spin(16) has two complex irreducible
representation 16 + and 16 , called spinor representations. Furthermore, both of
has
16 + and 16 are not representation of SO(16). Now, by Lemma 5.8.5, (E8 )e
C e
a complex irreducible representation (e8 )

of dimension 128, which implies that
must be Ss(16). Thus we have proved the following

is not SO(16). So (E8 )e
(E8 )e
theorem.
Theorem 5.8.7.

(E8 )e
= Ss(16).
Remark. We define an involution C-linear transformation of e8 by

(, P, Q, r, s, t) = (, P, Q, r, s, t).
This is the same as F4 E6 E7 E8 . We define a subgroup (E8 ) by
(E8 ) = { E8 | = }.
Then we have
(E8 )
= Ss(16).
Indeed, we can prove that the Lie algebra (e8 ) of the group (E8 ) is isomorphic to
the Lie algebra so(16) = {X M (16, R) | t X = X}. Hnece as the same the case the
group (E8 )e
, the group (E8 ) have to the semi-spinor group Ss(16). However the
e
proof of (E8 )
= (E8 ) is not concretely. M. Gomyo [10] find the group
= Ss(16)
183
Ss(16) explicitly in the group E8 (although the definition of the group E8 is different
from E8 in Section 5.5).
Theorem 5.9.1. The center z(E8 ) of the group E8 is trivial :
z(E8 ) = {1}.
Proof. Let z(E8 ). The relation = implies that (SU (2) E7 )

=
(SU (2) E7 )/Z 2 (Theorem 5.7.6), and so z((SU (2) E7 )). Therefore by
Theorem 4.9.1 we have
= (E, 1) = 1 or = (E, 1) = .
However
/ z(E8 ) (Theorem 5.7.6). Hence = 1.
5.10. Automorphism w of order 3 and subgroup (SU (3) E6 )/Z 3 of E8
In this section (also in the following Sections 5.11 and 5.12), we use the same
e w, even if these are different from those used in the
notation as e8 C , hR1 , R2 i, ,
proceeding sections.
We consider a 27 3 = 78 dimensional C-vector space
X1
o
n

(JC )3 = X = X2 Xi JC .
X3
In (JC )3 , we define an inner product (X, Y ), a Hermitian inner product hX, Y i, a

cross product X Y , an element X Y of sl(3, C) and an element X Y of e6 C
respectively by
(X, Y ) = (X1 , Y1 ) + (X2 , Y2 ) + (X3 , Y3 ) C,
hX, Y i = hX1 , Y1 i + hX2 , Y2 i + hX3 , Y3 i C,
X2 Y3 Y2 X3
X Y = X3 Y1 Y3 X1 (JC )3 ,
X1 Y2 Y1 X2
(X1 , Y1 ) (X1 , Y2 ) (X1 , Y3 )

1
X Y = (X2 , Y1 ) (X2 , Y2 ) (X2 , Y3 ) (X, Y )E sl(3, C),
3
(X3 , Y1 ) (X3 , Y2 ) (X3 , Y3 )
X Y = X1 Y1 + X2 Y2 + X3 Y3 e6 C ,

X1
Y1

where X = X2 , Y = Y2 (JC )3 . Further, for HomC (JC ), D = dij
X3
Y3
X1
M (3, C) and X = X2 (JC )3 , elements X, DX (JC )3 are naturally defined
X3
184
by
X1
X = X2 ,
X3
d11 X1 + d12 X2 + d13 X3

DX = d21 X1 + d22 X2 + d23 X3 .
d31 X1 + d32 X2 + d33 X3
Theorem 5.10.1. In an 8 + 78 + 27 3 + 27 3 = 248 dimensional C-vector

space
e8C = sl(3, C) e6 C (JC )3 (JC )3 ,
we define a Lie bracket [R1 , R2 ] by
[(D1 , 1 , X 1 , Y 1 ), (D2 , 2 , X 2 , Y 2 )] = (D, , X, Y ),
where
1
1
D = [D1 , D2 ] + X 1 Y 2 X 2 Y 1
4
4
1
1
= [1 , 2 ] + X 1 Y 2 X 2 Y 1
2
2
X = 1 X 2 2 X 1 + D1 X 2 D2 X 1 Y 1 Y 2
Y = t 1 Y 2 + t 2 Y 1 t D1 Y 2 + t D2 Y 1 + X 1 X 2 ,
then e8C becomes a C-Lie algebra of type E8 .
C
Proof. Let ee8 = e7 C PC PC C C C be the C-Lie algebra constructed
C
in Theorem 5.1.1. We define a mapping f : ee8 e8 C by
f ((, A, B, ), (X, Y, , ), (Z, W, , ), r, s, t))
1
1
2
2
2
2A
2B

1
1
= r
, , Z , Y ,
t
2
3
1
X
W
1
s
+r
2
3
then we can prove that f is an isomorphism as Lie algebras by straightforward calC
culations. Thus we have the isomorphism ee8
= e8 C .
C
Using the Killing form of ee8 which is obtained in Theorem 5.3.2, we see that the
Killing form B8 of e8 C is given by
5
B8 (R1 , R2 ) = 60tr(D1 D2 ) + B6 (1 , 2 ) + 15(X 1 , Y 2 ) + 15(X 2 , Y 1 )
2
(Ri = (Di , i , X i , Y i ) e8 C ), where B6 is the Killing form of e6 C . We define a

e of e8 C by
complex conjugate transformation
e
(D,
, X, Y ) = ( tD, t , Y , X).
And we define a Hermitian inner product hR1 , R2 i in e8 C by

e 2 ).
hR1 , R2 i = B8 (R1 , R
185
Then we have
5
hR1 , R2 i = 60tr(D1 ( t D2 )) + B6 (1 , t2 ) + 15hX 1 , X 2 i + 15hY 1 , Y 2 i.
2
Now, as in Theorem 5.5.3, we see that
E8 = { Aut(e8 C ) | hR1 , R2 i = hR1 , R2 i}
is a simply connected compact Lie group of type E8 .
We define a C-linear transformation w of e8 C by
w(D, , X, Y ) = (D, , X, 2 Y ),
3
1
i C. Then w E8 and w3 = 1.
where = +
2
2
Now, we shall study the following subgroup (E8 )w of E8 :
(E8 )w = { E8 | w = w}.
Theorem 5.10.2. (E8 )w
= (SU (3) E6 )/Z 3 , Z 3 = {(E, 1), (E, 2 1), ( 2 E,
1)}.
Proof. We first define a mapping 1 : SU (3) (E8 )w by
1 (A)(D, , X , Y ) = (ADA1 , , AX, tA1 Y ).
We have to prove that 1 (A) (E8 )w . Indeed, since the action of D1 = (D1 , 0, 0, 0)
su(3) sl(3, C) e8 C is given by
(adD1 )(D, , X, Y ) = ((adD1 )D, 0, D1 X, tD1 Y ),
for A = exp D1 , we have 1 (A) = exp(ad(D1 )) Aut(e8 C ). And from
tr(AD1 tA( t (AD2 tA))) = tr(AD1 ( tD2 )A1 ) = tr(D1 ( t D2 )),
hAX, AY i = hX, Y i,
we see that 1 (A) E8 . Evidently, w1 (A) = 1 (A)w, hence, 1 (A) (E8 )w . Next,
we define a mapping 2 : E6 (E8 )w by
2 ()(D, , X , Y ) = (D, 1 , X, t 1 Y ).
We have to prove that 2 () (E8 )w . Indeed, since the action of an element =
(0, , 0, 0) e6 e6 C e8 C is given by
(ad )(D, , X, Y ) = (0, (ad ), X, t Y ),
for = exp , we have 2 () = exp(ad( )) Aut(e8 C ). And from
186
B6 (1 1 , t (2 1 ) ) = B6 (1 1 , t 2 1 ) = B6 (1 , t 2 ),
hX, t 1 Y i = hX, Y i,
we see that 2 () E8 . Evidently, w2 () = 2 ()w, hence, 2 () (E8 )w . Now,
we define a mapping : SU (3) E6 (E8 )w by
(A, ) = 1 (A)2 ().
Since 1 (A) and 2 () commute, is a homomorphism. It is not difficult to show
that Ker = {(E, 1), (E, 2 1), ( 2 E, 1)} = Z 3 . Finally, since (E8 )w is connected
as the fixed points subgroup by automorphims w of the simply connected group E8
and
e = R}
(e8 )w = {R e8 C | wR = R, R
= {(D, , 0, 0) e8 C | D su(3), e6 }
= su(3) e6 ,
is onto. Thus we have the isomorphism (SU (3) E6 )/Z 3

= (E8 )w .
5.11. Automorphism w3 of order 3 and subgroup SU (9)/Z 3 of E8

In order to construct another C-Lie algebra of type E8 , we investigate the properties of the exterior C-vector space k (C n ). Let e1 , , en be the canonical C-basis
of n-dimensional C-vector space C n and (x, y) the inner product in C n satisfying
(ei , ej ) = ij . In k (C n ), we define an inner product by

(x1 xk , y 1 y k ) = det (xi , y j ) , k 1,
(a, b) = ab,
a, b 0 (C n ) = C.
Then, ei1 eik , i1 < < ik forms an orthonormal C-basis of k (C n ). For

u k (C n ), we define an element u nk (C n ) by
(u, v) = (u v, e1 en ),
v nk (C n ).
Then, induces a C-linear isomorphism

: k (C n ) nk (C n )
and satisfies the following identity:
2 u = (1)k(nk) u,
u k (C n ).
The group SL(n, C) naturally acts on k (C n ) as

A(x1 xk ) = Ax1 Axk ,
187
A1 = 1.
Hence the Lie algebra sl(n, C) acts on k (C n ) as

D(x1 xk ) =
k
X
j=1
x1 Dxj xk ,
D1 = 0.
Lemma 5.11.1. For A SL(n, C), D sl(n, C) and u, v k (C n ), we have
(1) (Au, tA1 v) = (u, v),

(2) (Au) = tA1 (u),
(Du, v) + (u, t Dv) = 0.
(Du) = tD1 (u).
For u, v k (C n ) (1 k n), we define a C-linear mapping u v of C n by

(u v)x = (v (u x)) + (1)nk
nk
(u, v)x,
n
x Cn.
Since tr(u v) = 0, u v can be regarded as element of sl(n, C) with respect to the

canonical basis of C n .
Lemma 5.11.2. For A SL(n, C), D sl(n, C) and u, v k (C n ), we have
(1) A(u v)A1 = Au t A1 v, [D, u v] = Du v + u (tDv).
(2) t (u v) = v u,
(u v) = (u) (v).
(3) tr(D(u v)) = (1)nk (Du, v).
Now, we construct another C-Lie algebra e8 C of type E8 .

Theorem 5.11.3. In an 80 + 84 + 84 = 248 dimensional C-vector space
e8 C = sl(9, C) 3 (C 9 ) 3 (C 9 ),
we define a Lie bracket [R1 , R2 ] by
[(D1 , u1 , v 1 ), (D2 , u2 , v 2 )] = (D, u, v),
where
D = [D1 , D2 ] + u1 v 2 u2 v 1
u = D1 u2 D2 u1 + (v1 v 2 )
v = t D1 v 2 + t D2 v 1 (u1 u2 ),
then e8 C becomes a C-Lie algebra of type E8 .
Proof. In order to prove the Jacobi identity, we need the following Lemma.
Lemma 5.11.4. For u, v, w 3 (C 9 ), we have
(1) u (v w) + v (w u) + w (u v) = 0,
(2) (u w)v (v w)u + ((u v) w) = 0.
188
Proof. Let u = u1 u2 u3 , v = u4 u5 u6 and w = u7 u8 u9 . For

x, y C 9 , we have
2
((u v)x, y) = ((v (u x)), y) + (u, v)(x, y)
3
2
= (x u, y v) + (u, v)(x, y)
3
= (x u2 u3 , v)(u1 , y) (x u1 u3 , v)(u2 , y)
1
+(x u1 u2 , v)(u3 , y) (u, v)(x, y).
3
Hence
(u v)x = (x u2 u3 , v)u1 + (u1 x u3 , v)u2
1
+(u1 u2 x, v)u3 (u, v)x.
3
(i)
Using this identity,

(u (v w) + v (w u) + w (u v))x
9
X
(u1 uj1 x uj+1 u9 , e1 e9 )uj
=
j=1
Denote x =
P9
(u1 u9 , e1 e9 )x = (ii).
i=1
xi ei , uj =
P9
k=1

ujk ek and U = ujk M (9, C). Hence we have
(u1 uj1 x uj+1 u9 , e1 e9 ) =

(u1 u9 , e1 e9 ) = detU,
9
X
k=1
u
ejk xk ,
where u
ejk is the cofactor of ujk of the matrix U . Therefore
X
X
(ii) =
xk u
ejk uj (detU )x =
xk u
ejk uji ei (detU )x
j,k
X
j,k
i,j,k
xk (detU )ki ei (detU )x = 0.
Thus (1) is proved. Next, let u = u1 u2 u3 and v = v 1 v 2 v 3 . Using (i), for

any a 3 (C 9 ), we have
((u w)v (v w)u, a)
= (((u w)v 1 ) v 2 v 3 , a) (((u w)v 2 ) v 1 v 3 , a)
+(((u w)v 3 ) v 1 v 2 , a) (((v w)u1 ) u2 u3 , a)
+(((v w)u2 ) u1 u3 , a) (((v w)u3 ) u1 u2 , a)

3 X
3
X
(ui ui+1 v j , w)(ui+2 v i+1 v j+2 , a)
= (u, w)(v, a) +
i=1 j=1
189
+(v, w)(u, a)
3 X
3
X
i=1 j=1
(ui v j v j+1 , w)(ui+1 ui+2 v j+2 , a)
= (u v, w a) = (((u v) w), a).
Thus (2) is proved.
From Lemmas 5.11.1, 5.11.2 and 5.11.4, we can prove that e8 C is a C-Lie algebra.
Therefore Theorem 5.11.3 is proved.
We will show that the Lie algebra e8 C given in Theorem 5.11.3 is also the Lie
algebra of type E8 . Since we can not give explicit isomorphism between this e8 C and
e8 C of Theorem 5.1.1, we shall show that this e8 C is simple.
Theorem 5.11.5. e8 C = sl(9, C) 3 (C 9 ) 3 (C 9 ) is a simple C-Lie algebra of
type E8 .
Proof. We use the decomposition
e8 C = sl(9, C) q,
q = 3 (C 9 ) 3 (C 9 ).
For a subset I = {i, j, k} (i < j < k) of {1, 2, , 9}, we put

eI = ei ej ek 3 (C 9 ).
Now, let a be a non-zero ideal of g = e8 C .
(1) Case sl(9, C) a = {0} and q a = {0}. Let p : g sl(9, C) be the projection.
If p(a) = 0, then a is contained in q, which contradicts q a = {0}. Hence, p(a) is a
P8
non-zero ideal of sl(9, C), so we have p(a) = sl(9, C). For an element D = i=1 Hi

P
P
sl(9, C), Hi = Eii E99 , there exists an element (u, v) =
J vJ eJ q
I u I eI ,
such that (D, u, v) a. Since [(D, 0, 0), (X, u, v)] = (0, Du, t Dv) q a = {0}, we
have
X
X
X
0 = Du =
uI DeI = 3
u I eI 6
u I eI ,
I
0 = Dv = 3
I69
vJ eJ + 6
J69
I9
vJ eJ ,
J9
i.e., uI = 0 and vJ = 0. Then, 0 6= (D, u, v) = (D, 0, 0) sl(9, C) a = {0}. This is

a contradiction.
(2) Case sl(9, C) a 6= {0}. Since sl(9, C) a is a non-zero ideal of sl(9, C), we
have sl(9, C) a. For any ei ej ek 3 (C 9 ), put
D=
1
(Eii + Ejj + Ekk ) Ell ,
3
l 6= i, j, k.
Since (D, 0, 0) sl(9, C) a, we see that

(0, ei ej ek , 0) = [(D, 0, 0), (0, ei ej ek , 0)] a,
(0, 0, ei ej ek ) = [(D, 0, 0), (0, 0, ei ej ek )] a.
190
It follows that q a. Hence we have a = g.
(3) Case q a 6= {0}. Let R = (0, u, v) be a non-zero element of q a. In the

P
case u 6= 0, we put u = I uI eI . Without loss of generality, we may assume that
u{123} = 1. Putting Sij = (Eii Ejj , 0, 0) g and T = (0, 0, e1 e2 e4 ) g, we
have
0 6= ad(T )ad(S37 )ad(S27 )ad(S17 )ad(S36 )ad(S25 )ad(S14 )R
= (E34 , 0, 0) sl(9, C) a.
Then we can reduce this case to the case (2). In case v 6= 0, we ca similarly reduce
to the case (2).
Thus the simplicity of g has been proved. Since the dimension of g is 248, we see that
g is a Lie algebra of type E8 .
Proposition 5.11.6. The Killing form B8 of the Lie algebra e8 C = sl(9, C)
3 (C 9 ) 3 (C 9 ) is given by
B8 ((D1 , u1 , v 1 ), (D2 , u2 , v 2 )) = 60(tr(D1 D2 ) + (u1 , v 2 ) + (u2 , v 1 )).
Proof. We consider a symmetric bilinear form B of e8 C :
B((D1 , u1 , v 1 ), (D2 , u2 , v 2 )) = tr(D1 D2 ) + (u1 , v2 ) + (u2 , v 1 ).
Using Lemmas 5.11.2, 5.11.4, we see that B is e8 C -adjoint invariant. Since e8 C is
simple, there exists k C such that B8 (R1 , R2 ) = kB(R1 , R2 ) for all Ri e8 C . To
determined k, let R = R1 = R2 = (E11 E22 , 0, 0) e8 C . Then we have
B8 (R, R) = 120,
B(R, R) = 2.
Therefore k = 60.
e of e8 C by
We define a complex-conjugate linear transformation
e
(D,
u, v) = ( t D, v, u).
and we define a Hermitian inner product hR1 , R2 i in e8 C by

e 2 ).
hR1 , R2 i = B8 (R1 , R
Then we have
hR1 , R2 i = 60(tr(D1 t D2 ) + (u1 , u2 ) + (v 2 , v 1 )).

As in Theorem 5.5.3,
E8 = { Aut(e8 C ) | hR1 , R2 i = hR1 , R2 i}
is a simly connected compact simple Lie group of type E8 .
191
We define a C-linear transformation w3 of e8 C by

w3 (D, u, v) = (D, u, 2 v),
1
3
i C. Then, w3 E8 and w3 3 = 1.
where = +
2
2
Now, we study the following subgroup (E8 )w3 of E8 :
(E8 )w3 = { E8 | w3 = w3 }.
(E8 )w3
= SU (9)/Z 3 , Z 3 = {E, E, 2 E}.
Theorem 5.11.7.
Proof. We define a mapping : SU (9) (E8 )w3 by

(A)(D, u, v) = (ADA1 , Au, t A1 v).
is well-defined : (A) (E8 )w3 . Indeed, for A = exp X, X su(9), we have
exp(ad(X, 0, 0))(D, u, v) = (exp(ad(X)D, (exp X)u, (exp(t X))v)
= (Ad(exp X)D, (exp X)u, t (exp X)1 v)
= (exp X)(D, u, v).
Hence (A) E8 . Clearly w3 (A) = (A)w3 . Therefore (A) (E8 )w3 . Obviously
is a homomorphism. We shall show that is onto. Since the Lie algebra (e8 )w3 of
the group (E8 )w3 is
e = R, w3 R = R} = {(D, 0, 0) e8 C | D su(9)}
(e8 )w3 = {R e8 C | R
= su(9),
the differential of is onto. Since (E8 )w3 is connected, is also onto. It is not
difficult to see that ker = {E, E, 2 E} = Z 3 . Thus we have the isomorphism
SU (9)/Z 3
= (E8 )w3 .
E8
5.12. Automorphism z5 of order 5 and subgroup (SU (5) SU (5))/Z 5 of

We shall construct one more C-Lie algebra of type E8 .
Theorem 5.12.1. In a 48 + 50 4 = 248 dimensional C-vector space
e8 C = g0 g1 g2 g2 g1 ,
(suffices are considered mod 5) where

g0 = sl(5, C) sl(5, C),
g1 = C (C 5 ) = g1 , g2 = 2 (C 5 ) C 5 = g2 ,
5
we define a Lie bracket [R1 , R2 ] as follows.

192
[g0 , g0 ] g0
[g0 , g1 ] g1
[g0 , g2 ] g2
[g0 , g2 ] g2
[g0 , g1 ] g1
[g1 , g1 ] g0
[g2 , g2 ] g0
[g1 , g1 ] g2
[g1 , g1 ] g2
[g2 , g2 ] g1
[g2 , g2 ] g1
[g1 , g2 ] g2
[g1 , g2 ] g2
[g2 , g1 ] g1
[g2 , g1 ] g1
[(C1 , D1 ), (C2 , D2 )] = ([C1 , C2 ], [D1 , D2 ]),

[(C, D), x a] = (Cx) a + x (Da),
[(C, D), b y] = (Cb) y + b ( t Da),
[(C, D), c z] = (t Cc) z + c (Dz),
[(C, D), w d] = (t Cw) z + c ( t Dz),
[x a, w d] = ((a, d)x w, (x, w)a d),
[b y, c z] = ((y, z)b c, (b, c)z y),
[x1 a1 , x2 a2 ] = (x1 x2 ) (a1 a2 ),
[b1 y 1 , b2 y 2 ] = (b1 b2 ) (y 1 y 2 ),
[x a, b y] = (b x) (a y),
[b y, w d] = (b w) (d y).
Then e8 C becomes a C-Lie algebra.

Proof. In order to prove the Jacobi identity, we need the following Lemma.
Lemma 5.12.2. For x, y, z 1 (C 5 ) = C 5 and a, b, c 2 (C 5 ), we have
(1) a (b c) + b (c a) + c (a b) = 0,
(2) (a (b x)) + (b (a x)) + x (a b) = 0,
(3) ((x y) z) = (x, z)y (y, z)x,
(4) x (a y) + (y +(a x)) (x, y)a = 0,
(5) (a (b x)) (b (a x)) (a, b)x = 0,

(6) a (b x) + b (a x) x (a b) = 0,
(7) (a x) y (a y) x + a (x y) = 0,
1
(8) (a b)c = ((a c) b) (a, b)c (b, c)a,
5
3
(9) (x y)a = (y (x a)) + (x, y)a.
5
Proof. (1) Let a = a1 a2 , b = a3 a4 , c = a5 a6 and ai =
((a (b c)), x) = (a, (b c) x)
= (a1 , (b c))(a2 , x) (a2 , (b c))(a1 , x)
P5
j=1
aij ej . Since
= (a1 b c, e1 e5 )(a2 , x) (a2 b c, e1 e5 )(a1 , x),

we have
(a (b c) + b (c a) + c (a b))
193
5 X
6
X
j=1 i=1
(1)i (a1 ai1 ai+1 a6 , e1 e5 )aij ej
5
X
j=1
a1j
a2j
det
a6j
a11
a21
a61
a15
a25
ej = 0.
a65
(2) Using (1), we have

((a (b x)) + (b (a x)) + x (a b), c)
= (b x, c a) + (a x, b c) + (c x, a b)
= (x, b (c a) + a (b c) + c (a b)) = 0.
(3) For any v 1 (C 5 ) = C 5 , we have
(((x y) z), v) = (x y, z v) = (x, z)(y, v) (y, z)(x, v).
This shows (3).
(4),(5) Let a = a1 a2 . Since
(x a, y b) = (x, y)(a, b) (a1 , y)(x a2 , b) + (a2 , y)(x a1 , b),
(b x, a y) = (a, y (b x)) = (a1 , y)(x a2 , b) (a2 , y)(x a1 , b),
we have
(x a, y b) + (b x, a y) = (x, y)(a, b).
Using this identity, we have
(x (a y) + (y (a x)) (x, y)a, b)
= (b x, a y) + (x a, y b) (x, y)(a, b) = 0,
((a (b x)) (b (a x)) (a, b)x, y)
= (x b, y a) + (a x, b y) (x, y)(a, b) = 0.
(6) Since
1
4
((x y)z, v) = (x z, y v) + (x, y)(z, v) = (y, z)(x, v) (x, y)(z, v),
5
5
we have
1
(x y)z = (y, z)x (x, y)z.
5
For v, w 1 (C 5 ) = C 5 , we have
((a (b x))v, w)
3
= (v a, w (b x)) ((a b), x)(v, w)
5
194
(i)
= (v, w)(a, (b x)) (a1 , w)(w a2 , (b x))

3
+(a2 , w)(w a1 , (b x)) ((a b), x)(v, w)
5
2
= ((a b), x)(v, w) (a1 , w)(b x w a2 , e1 e5 )
5
+(a2 , w)(b x w a1 , e1 e5 ),
((b (a x))v, w)
3
= (x a, w (b v)) ((a b), x)(v, w)
5
= (v, w)((a b), v) + (a1 , w)(b x w a2 , e1 e5 )
3
(a2 , w)(b x w a1 , e1 e5 ) ((a b), x)(v, w).
5
Using (i), we have
1
(a (b x))v + (b (a x))v = (x, w) (a b) ((a b), x)v
5
= (x (a b))v.
(7) We have
3
((a (x y))v, w) = (x y w, v a) (a, x y)(v, w)
5
2
= (x, v)(y w, a) (y, v)(x w, a) + (a, x y)(v, w)
5
2
= (x, v)((a y), w) (y, v)((a x), w) + (a, x y)(v, w).
5
On the other hand, using (i), we have
1
((a x) y)v = (y, v) (a x) ((a x), v)y,
5
1
= (y, v) (a x) (a, x v)y,
5
1
((a x) y)v = (x, v) (a y) (a, x v)y.
5
Hence (7) is proved.
(8) Let a = a1 a2 and c = c1 c2 . Since
3
((a b)v, w) = (a v, b w) (a, b)(v, w)
5
2
= (a1 , w)(x a2 , b) + (a2 , w)(x a1 , b) + (a, b)(v, w),
5
we have
(a b)c = (a1 c1 , b)a1 c2 + (a1 c2 , b)a2 c1
4
+(a2 c1 , b)a1 c2 (a2 c2 , b)a1 c1 + (a, b)c.
5
195
On the other hand, for d 2 (C 5 ), we have

(((a c) b), d) = (a c, b d)
= (a, b)(c, d) (a1 c1 , b)(a2 c2 , d) + (a1 c2 , b)(a2 c1 , d)
+(a2 c1 , b)(a1 c2 , d) (a2 c2 , b)(a1 c1 , d) + (c, b)(a, d).
(9) Using (i), we have
2
(x y)a = (y, a1 )x a2 (y, a2 )x a1 (x, y)a.
5
( (y (x a)), b) = (x a, y b)
= (y, a1 )(x a2 , b) (y, a2 )(x a1 , b) (x, y)(a, b).
From Lemmas 5.11.1, 5.11.2 and 5.12.2, we can prove that e8 C be comes a C-Lie
algebra. Furthermore we have the following theorem.
Theorem 5.12.3. The Lie algebra e8 C = g0 g1 g2 g2 g1 is a C-simple
Lie algebra of type E8 .
Proof. We shall show that e8 C is simple. For this end, we use the decomposition
g = e8 C = g01 g12 q, where
g01 = {(C, 0) g0 | C sl(5, C)}
= sl(5, C),
g = {(0, D) g | D sl(5, C)}
= sl(5, C),
02
q = g1 g2 g2 g1 .
Now, let a be a non-zero ideal of g. There are three cases to be considered.

(1) Case g01 a = {0}, g02 a = {0} and q a = {0}. Let pi : g g0i
(i = 1, 2) denote the projection. If p1 (a) = {0} and p2 (a) = {0}, then a is contained
in q, which contradicts q a = {0}. Hence, without loss of generality, we may
P
assume that p1 (a) = g01 , because g01 is a simple Lie algebra. For C = 4i=1 Hi
sl(3, C) where Hi = Eii E55 , there exists (D, g1 , g2 , g2 , g1 ) g01 q such that
(C, D, g1 , g2 , g2 , g1 ) a. Since
[(C, 0), (C, D, g1 , g2 , g2 , g1 )]
= (0, 0, [C, g1 ], [C, g2 ], [C, g2 ], [C, g1 ]) q a = {0},
we have [C, gi ] = 0 (i = 1, 2, 2, 1). Since any eigenvalue of adX is not 0, we have
gi = 0. Then we have (C, D) g0 a. Since
[(C, D), (E45 , 0)] = (5E45 , 0) g01 a,
196
we have g01 a 6= {0}. This is a contradiction.
(2) Case g01 a 6= {0} or g02 a =

6 {0} . We may assume that g01 a 6= {0}.
Since g01 is simple, we have g01 a. Since [g01 , gi ] = gi (i = 1, 2, 2, 1), we have
q a. Since
a [g1 , g1 ] [e1 (e1 e2 ), e1 (e1 e3 )] = (0, E23 ),
we have g02 a 6= {0}. It follows that g02 a. Hence we have a = g.
(3) Case q a 6= {0}. Let R = (g1 , g2 , g2 , g1 ) (gi gi ) be a non-zero element

P
of q a. In the case g1 6= 0, we put g1 = i,j<k gijk ei (ej ek ). Without loss of
generality, we may assume that g112 6= 0. Putting Sijkl = (Eii Ejj , Ekk Ell ) g0
and T = e2 e1 e2 g1 , we have
ad(T )ad(S1523 )ad(S1415 )ad(S1314 )ad(S1213 )R = (E12 , 0) g01 a.
Then we can reduce this case to the case (2). In the case gi 6= 0 (i = 2, 2, 1), we
can similarly reduce to the case (2).
Thus the simplicity of g has been proved. Since the dimension of g is 248, we see
that g is a C-Lie algebra of type of E8 .
Proposition 5.12.4. The Killing form B8 of the Lie algebra e8 C = sl(5, C)
sl(5, C) g1 g2 g2 g1 is given by
B8 (R1 , R2 ) = 60(tr(C1 C2 ) + tr(D1 D2 ) (x1 , w2 )(a1 , d2 ) (x2 , w1 )(a2 , d1 )
(y 1 , z 2 )(b1 , c2 ) (y 2 , z 1 )(b2 , c1 )),
where Ri = (Ci , Di , xi ai , bi y i , ci z i , w i di ) e8 C .
Proof. We consider a symmetric bilinear form B of e8 C :
B(R1 , R2 ) = tr(C1 C2 ) + tr(D1 D2 ) (x1 , w2 )(a1 , d2 ) (x2 , w 1 )(a2 , d1 )
(y 1 , z 2 )(b1 , c2 ) (y 2 , z 1 )(b2 , c1 ).
Using Lemmas 5.12.1, 5.12.2, we see that B is e8 C -adjoint invariant. Since e8 C is
simple, there exists k C such that B8 (R1 , R2 ) = kB(R1 , R2 ) for all Ri e8 C . To
determined k, let R = R1 = R2 = (E11 E22 , 0, 0, 0, 0, 0) e8 C . Then we have
B8 (R, R) = 120,
B(R, R) = 2.
Therefore k = 60.
e of e8 C by
We define a complex-conjugate linear transformation
e
(C,
D, x a, b y, c z, w d)
= ( t C, t D, w d, c z, b y, x a).
197
Further we define a positive dfinite Hermitian inner product hR1 , R2 i in e8 C by

e 2 ).
hR1 , R2 i = B8 (R1 , R
Then, we have
hR1 , R2 i = 60(tr(C1 t C2 ) + tr(D1 t D2 ) + (x1 , x2 )(a1 , a2 ) + (y 1 , y 2 )(b1 , b2 )

+(z 1 , z 2 )(c1 , c2 ) + (w 1 , w 2 )(d1 , d2 )).
As in Theorem 5.5.3,
E8 = { Aut(e8 C ) | hR1 , R2 i = hR1 , R2 i}
is a simply connected compact simple Lie group of type E8 .
Let = exp(2i/5) C and we define a C-linear transformation z5 of e8 C by
z5 (C, D, g1 , g2 , g2 , g1 ) = (C, D, (g1 ), 2 (g2 ), 3 (g2 ), 4 (g1 )).
Then z5 E8 and z5 5 = 1.
Now, we study the following subgroup (E8 )z5 of E8 :
(E8 )z5 = { E8 | z5 = z5 }.
(SU (5) SU (5))/Z 5 , Z 5 = {(E, E), (E, 2 E),
Theorem 5.12.5. (E8 )z5 =
( 2 E, 4 E), ( 3 E, E), ( 4 E, 3 E)}, = exp(2i/5).
Proof.@We define mappings 1 , 2 : SU (5) E8 respectively by
1 (A)(C, D, x a, b y, c z, w d)
= (ACA1 , D, (Ax) a, (Ab) y, (t A1 c) z, (t A1 w) d).
2 (B)(C, D, x a, b y, c z, w d)
= (C, BDB 1 , x (Ba), b (t B 1 y), c (Bz), w (t B 1 d)).
1 and 2 are well-defined: 1 (A), 2 (B) E8 . Indeed, for Z su(5), we have
(Z, 0) g0 and
exp(ad(Z, 0))(C, D, x a, b y, c z, w d)
= (exp(ad(Z))C, D, ((exp Z)x) a,
((exp Z)b) y, ((exp(t Z))c z, ((exp(t Z))w) d)
= (Ad(exp Z)C, D, ((exp Z)x) a,
((exp Z)b) y, (t (exp Z)1 c) z, (t (exp Z)1 w) d)
= 1 (exp Z)(C, D, x a, b y, c z, w d).
Hence 1 (A) Aut(e8 C ) = E8 C . Using Lemma 5.11.1, we have
h(A)R1 , 1 (A)R2 i = hR1 , R2 i.
198
Therefore 1 (A) E8 . Similarly 2 (B) E8 .
Now, we define a mapping : SU (5) SU (5) E8 by

(A, B) = 1 (A)2 (B).
Since 1 (A) and 2 (B) commute, is a homomorphism. We shall show that is

onto. Since (e8 )z5 = su(5) su(5), the differential is onto. It is not difficult to see
that
Ker = {(E, E), (E, 2 E), ( 2 E, 4 E), ( 3 E, E), ( 4 E, 3 E)} = Z 5 .
Further, since (E8 )z5 is connected, is onto. Thus we have the isomorphism (SU (5)
SU (5))/Z 5
= (E8 )z5 .
5.13. Non-compact exceptional Lie groups E8(8) and E8(24) of type E8
Let
e8(8) = e7(7) P P R R R,
e8(24) = e7(25) P P R R R,
(where e7(7) = (e7 C ) ),

(where e7(25) = (e7 C ) ).
For R1 , R2 e8(8) or e8(24) , we define a Lie bracket [R1 , R2 ] as similar to e8 C of

Section 5.1. Now, we define groups E8(8) and E8(24) by
E8(8) = { IsoR (e8(8) ) | [R1 , R2 ] = [R1 , R2 ]},
E8(24) = { IsoR (e8(24) ) | [R1 , R2 ] = [R1 , R2 ]}.

E8(8)
= (E8 C ) ,
E8(24)
= (E8 C ) .
Theorem 5.13.1. The polar decompositions of the Lie groups E8(8) and E8(24)
are respectively given by
E8(8) Ss(16) R128 ,
E8(24) (SU (2) E7 )/Z 2 R112 .

Proof. These are the facts corresponding to Theorems 5.8.7 and 5.7.6.
Theorem 5.13.2. The centers of the groups E8(8) and E8(24) are trivial:
z(E8(8) ) = {1},
z(E8(24) ) = {1}.
199
References
[1] K. Abe and I. Yokota, Realization of spaces E6 /(U (1)Spin(10)), E7 /(U (1)E6 ),
E8 /(U (1)E7 ) and their volumes, Tokyo J. Math. 20(1997), 73-86.
[2] K. Abe and I. Yokota, Volumes of compact symmetric spaces, Tokyo J. Math.
20(1997), 87-105.
[3] M.F.Atiyah, R.Bott and A.Shapiro,
1(1964).
Clifford modules, Topology 3, Suppl.
[4] A. Borel and J. de Siebenthal, Les sous-groupes fermes de rang maximum des
groupes de Lie clos, Comment. Math. 23(1949-50), 200-221.
[5] C. Chevalley, Theory of Lie groups I, Princeton Univ. Press, 1946.
[6] H. Freudenthal, Oktaven, Ausnahmegruppen und Oktavengeometry, Math. Inst.
Rijkuniv. te Utrecht, 1951.
[7] H. Freudenthal, Beziehung der E7 und E8 zur Oktavenevene I, II, X, Nedel,
Akad. Weten. Proc.Ser. A.57(1954), 218-230, 363-365, 66(1963), 457-471.
[8] F. Gantmacher. On the classifications of real simple Lie groups, Rec. Math. J.
5(1939).
[9] S. Gomyo, Realization of maximal subgroups of rank 8 of the simply connected
compact simple Lie group of type E8 , Tsukuba J. Math. 21(1997), 595-616.
[10] S. Gomyo, Realizations of subgroups of type D8 of connected exceptional simple
Lie groups of type E8 , Tsukuba J. Math. 23(1999), 585-615.
[11] M. Goto and F. D. Grosshans, Semi-simpleLie algebras, Dekker, 1978.
[12] T. Imai, On bounded symmetric domains of exceptional type, J. Fac. Sci. Shinshu
Univ. 16(1981), 65-96.
[13] T. Imai and I. Yokota, Non-compacct simple Lie group E8(24) of type E8 , J.
Fac. Sci. Shinshu Univ. 15(1980), 53-76.
[14] T. Imai and I. Yokota, Another definitions of exceptional simple Lie groups of
type E7(25) and E7(133) , J. Fac. Sci. Shinshu Univ. 15(1980), 47-52.
[15] T. Imai and I. Yokota, Simply connected compact simple Lie group E7(133)
of type E7 , J. Math. Kyoto Univ. 21(1981), 383-395.
[16] T. Imai and I. Yokota, Simply connected compact simple Lie group E8(248) of
type E8 , J. Math. Kyoto Univ. 21(1981), 741-762.
200
[17] S. Kaneyuki, On the subalgebra g0 and gev of semisimple Lie algebra, J. Math.
Soc.Japan, 45(1993), 1-19.
[18] Y. Matsushima, Some remarks on exceptional Lie group F4 , Nagoya Math. J.
4(1954), 83-88.
[19] Y. Matsushima, Theory of Lie algebras (in Japaneae), Kyoritsu-shuppan, Tokyo,
1957.
[20] K. Mitsuishi and I. Yokota, Volume formula of compact simple Lie groups and
values of , J. Fac. Sci. Shinshu Uni. 29(1994), 69-70.
[21] T. Miyasaka, Adjoint orbit types of compact exceptional Lie group G2 in their
Lie algebra, Math. J. Okayama Univ. 43(2001), 17-180.
[22] T. Miyasaka, O. Schukuzawa and I. Yokota, Spinor-generators of compact exceptional Lie groups F4 , E6 and E7 , Tsukuba J. Math. 22(1998), 705-721.
[23] T. Miyasaka, O. Yasukura and I. Yokota, Diagonalization of an element P of
PC by the compact Lie group E7 , Tsukuba J. Math. 22(1998), 687-703.
[24] T. Miyasaka and I. Yokota, Orbit types of the compact Lie group E7 in the
Freudenthal vector space PC , Tsukuba J. Math. 23(1999), 229-234.
[25] T. Miyashita, Decomposition of spinor groups by the involution in exceptional
Lie groups, Math. J. Okayama Univ, 44(2001), 151-158.
[26] T. Miyashita, Fixed points subgroups by two involutive automorphisms ,
of compact exceptional Lie groups F4 , E6 and E7 , Tsukuba J. Math. 27(2003),
199-215.
[27] T. Miyashita and I. Yokota, 2-graded decompositions of exceptional Lie algebra g
and group realizations of gev , g0 , Part III, G = E8 , Japanese J. Math. 26-1(2000),
31-50.
[28] T. Miyashita and I. Yokota, Fixed points subgroups G, by two involutive

automorphisms , of compact exceptional Lie groups F4 , E6 and E7 , Tsukuba
J. Math. 24(2001), 135-149.
[29] T. Miyashita and I. Yokota, An explicit isomorphism between (e8 ), and so(8)
so(8) as Lie algebras, Math. J. Toyama Univ. 24(2001), 151-158.
[30] T. Miyashita and I. Yokota, 3-graded decompositions of exceptional Lie algebra
g and group realizations of gev , g0 and ged , Part II, G = E7 , Case 1, J. Kyoto
Univ.46-2(2006), 383-413, Case 2, 3, 46-4(2006), 805-832, Case 5, 47-1(2007),
121-128.
201
[31] T. Miyashita and I. Yokota, 3-graded decompositions of exceptional Lie algebra

g and group realizations of gev , g0 and ged , Part III, G = E8 , J. Kyoto Univ. to
appear.
[32] P. K. Rasevskii, A theorem on the connectedness of a subgroup of a simply
connected Lie group commuting with any its automorphis A theorem on the
connectms, Trans. Moscow Math. Soc. 30(1974), 3-22.
[33] J. Sekiguchi, Fundamental group of semisimple symmetric spaces, Advanced
Studies to pure Math. 14(1988), 519-529.
[34] O. Shukuzawa, Embeddings of unit spheres into exceptional groups of type G2
and homeomorphisms (G2 )C /SL(3, C) (S C )6 , G2(2) /SL(3, R) S3,4 , Math.
J. Toyama Univ. 24(2001), 159-171.
[35] O. Shukuzawa, Diagonalization of elements of Freudenthal R-vector space and
split Freudenthal R-space, Tsukuba J. Math. 27(2003), 113-128.
[36] O. Shukuzawa, Orbit space of split Freudenthal vector space, Yokohama Math.
J. 50(2003), 1-10.
[37] O. Shukuzawa and I. Yokota, Non-compact simple Lie group E6(6) of type E6 ,
J. Fac. Sci. Shinshu Univ. 14(1979), 1-13.
[38] O. Shukuzawa and I. Yokota, Non-compact simple Lie group E6(14) and E6(2)
of type E6 , J. Fac. Sci. Shinshu Univ. 14(1979), 15-28.
[39] I. Yokota, Embedding of projective spaces into elliptic projective Lie groups,
Proc. Japan Acad. 35(1959), 281-283.
[40] I. Yokota, A note on Spin(9), J. Fac. Sci. Shinshu Univ. 3(1968), 61-70.
[41] I. Yokota, Exceptional Lie group F4 and its representation rings, J. Fac. Sci.
Shinshu Univ. 3(1968), 35-60.
[42] I. Yokota, On a non-compact simple Lie group F4,1 of type F4 , J. Fac. Sci.
Shinshu Univ. 10(1975), 71-80.
[43] I. Yokota, Non-compat simple Lie group G2 of type G2 , J. Fac. Sci. Shinshu
Univ. 12(1977), 45-52.
[44] I. Yokota, Non-compact simple Lie group F4,2 of type F4 , J. Fac. Sci. Shinshu
Univ. 12(1977), 53-64.
[45] I. Yokota, Subgroups of type A2 -series in the exceptional Lie group G2 , J. Fac.
Sci. Shinshu Univ. 14(1979), 87-94.
202
[46] I. Yokota, Simply connected compact simple Lie group E6(78) of type E6 and
its involutive automorphisms, J. Math. Kyoto Univ. 20(1980), 447-473.
[47] I. Yokota, Subgroup (SU (8))/Z 2 of compact simple Lie group E7 and noncompact simple Lie group E7(7) of type E7 , Math. J. Okayama Univ. 24(1982),
53-71.
[48] I. Yokota, Realization of automorphisms of order 3 and G of compact exceptional Lie groups G, I, G = G2 , F4 , E6 , J. Fac. Sci. Shinshu Univ. 20(1985),
131-144.
[49] I. Yokota, Realization of involutive automorphisms of exceptional Lie groups
G, Part I, G = G2 , F4 and E6 , Part II, G = E7 , Part III, G = E8 , Tsukuba J.
Math. 14(1990), 185-223, 14(1990), 379-404, 15(1991), 301-314.
[50] I. Yokota, Orbit types of the compact Lie group E6 in the complex exceptional
Jordan algebra JC , Proc. Symp. on non-associative algebras, Japan, 1991, 353359.
[51] I. Yokota, 2-graded decompositions of exceptional Lie algebra g and group realizations of gev , g0 , Part I, G = G2 , F4 , E6 , Part II, G = E7 , Japan. J. Math.
24(1998), 258-296, 25(1999), 155-179.
[52] I. Yokota, 3-graded decompositions of exceptional Lie algebra g and group realizations of gev , g0 and ged , Part I, G = G2 , F4 , E6 , J. Math. Kyoto Univ.
41-3(2001), 449-494.
[53] I. Yokota, T, Imai and O. Yasukura, On the homogeneous space E8 /E7 , J. Math.,
Kyoto Univ. 23(1983), 467-473.
[54] I. Yokota, T. Ishihara and O. Yasukura, Subgroup ((SU (3) SU (6))/Z2 ).Z2
of the simply connected compact simple Lie group E7 , J. Math. Kyoto Univ.
23(1983), 715-737.
[55] I. Yokota and O. Shukuzawa, Automorphisms of order 4 of the simply connected
compact Lie group E6 , Tsukuba J. Math. 15(1991), 451-463.
[56] I. Yokota and O. Yasukura, Subgroup (SU (2) Spin(12))/Z 2 of compact simple
Lie group E7 and non-compact simple group E7, of type E7(5) , Hiroshima Univ.
J. 12(1982), 59-76.
[57] I. Yokota and O. Yasukura, Non-compact simple Lie group E8(8) , Tsukuba J.
Math. 10(1986), 331-349.
[58] I. Yokota, Groups and Topology (in Japanese), Shokabo, Yokyo, 1971.
203
[59] I. Yokota, Classical linear simple Lie groups (in Japaneae), Gendai-sugakushya,
Kyoto, 1990.
[60] I. Yokota, Exceptional linear simple Lie groups (in Japaneae), Gendai-sugakusya,
Kyoto, 1992.
has some errors
and are written by the same notaion
204

Exceptional Lie Groups Explained

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Exceptional Lie Groups Explained

Uploaded by

Copyright:

Available Formats

Exceptional Lie groups

arXiv:0902.0431v1 [math.DG] 3 Feb 2009

Exceptional Lie group G2

1.1. Cayley algebra C

Complex exceptional Lie group G2 C

Exceptional Lie group F4

2.1. Exceptional Jordan algebra J

2.13. Complex exceptional Lie group F4

Exceptional Lie group E6

3.1. Compact exceptional Lie group E6

3.14. Complex exceptional Lie group E6

Exceptional Lie group E7

4.1. Freudenthal vector space PC

4.14. Complex exceptional Lie group E7

Exceptional Lie group E8

5.1. Lie algebra e8 C

SO(n) = {A O(n) | detA = 1} (spcecial orthogonal group),

U (n) = {A M (n, C) | A A = E} or {A M (n, C) | (t A)A = E} (unitary

Exceptional Lie group G2

and defined similarly on the other lines. For example, e1 e6 = e7 , e4 e7 = e3 etc. We

(xy, xy) = (x, x)(y, y), |xy| = |x||y|.

8 b(ax) + a(bx) = 2(a, b)x = (xa)b + (xb)a.

For example, we have

(ax)y + x(ya) = a(xy) + (xy)a,

To prove this, it is sufficient to show

From (1)(1) = (1 1) = 1 and 1 6= 0, we have 1 = 1, while from (ei )(ei ) =

Theorem 1.2.2. G2 is a compact Lie group.

Proof. G2 is a compact Lie group as a closed subgroup of the orthogonal group

Then Gij D4 and {Gij | 0 i < j 7} forms an R-basis of D4 . Furthermore we

Lemma 1.3.1. For i, j = 0, 1, , 7, i 6= j, we have Fij D4 , and when i < j,

G01 + G23 + G45 + G67

= G02 G13 G46 + G57

2F04 = G04 G15 + G26 G37

2F37 = G04 G15 + G26 + G37 ,

2F06 = G06 G17 G24 + G35

2F17 = G06 + G17 G24 + G35

2F24 = G06 G17 + G24 + G35

2F35 = G06 + G17 + G24 + G35 ,

2F25 = G07 G16 + G25 G34

2F34 = G07 G16 G25 + G34 .

G03 + G12 + G47 + G56

In particular, {Fij | 0 i < j 7} forms an R-basis of D4 .

Definition. We define R-linear mappings , , : D4 D4 respectively by

Lemma 1.3.2. The mappings , , are automorphisms of the Lie algebra D4 :

In the following calculations, let i, j, k, l be all distinct and i, j, k, l 6= 0. For x C,

[Fi0 , Fkl ]x = (Fi0 Fkl Fkl Fi0 )x =

Hereafter, we denote by C0 the subset {a C | a = a} of C.

(2) (La )x = La x = ax = xa = xa = Ra x, x C. Hence La = Ra . The

(3) It is sufficient to show that these relations hold for a = ei , i = 1, , 7. We

Lei = (2Fi0 ) = 2F0i = 2G0i (Lemma 1.3.1) = 2Gi0 = Tei ,

Tei = (2Gi0 ) = 2Fi0 = Lei ,

Proof. Let D be the Lie subalgebra of D4 generated by {La | a C0 }. Since

Proof. Since {La | a C0 } generates D4 (Lemma 1.3.4), it is sufficient to check

gives an isomorphism as groups.

Also such D2 , D3 are uniquely determined for D1 and we have

Proof. If a C0 , then La , Ra , Ta D4 (Lemma 1.3.3.(1)) and the equality

Similarly, for b C0 , we have

Applying Ta on (ii) and using (i), we have

Putting x = 1, we have D2 y = D3 y, so that D2 = D3 (= D). Therefore

We therefore see that p R, so that p = 0 since p C0 . Hence Dx = xp = 0,

(D2 x)y + x(D3 y) = (D1 )(xy),

Proof. For D2 D4 , there exist D3 , D1 D4 such that