07.07.2013 Views

L - Cimat

L - Cimat

L - Cimat

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Mercedes Siles Molina<br />

Universidad de Málaga


Part of this course follows the notes on a course on Jordan algebras<br />

delivered by Prof. Cris=na Draper Fontanals. I’m grateful to her for<br />

sharing her notes.


Bibliography<br />

N. Jacobson, Structure and representa=ons of Jordan<br />

algebras, Colloquium Publica=ons (American Mathema=cal<br />

Society) Vol. 39, 1968.<br />

N. Jacobson, Excep=onal Lie algebras, Lecture Notes in<br />

pure and applied Mathema=cs, 1971.<br />

O. Loos, Jordan pairs, Lecture Notes in Math., vol. 460,<br />

Springer-­‐Verlag, Berlin, New York, 1975.<br />

K. McCrimmon, A taste of Jordan algebras, Springer-­‐<br />

Verlag, New York, 2003.<br />

R. D. Schafer, An introduc=on to non-­‐associa=ve algebras,<br />

Academic Press, 1966.<br />

K. A. Zhevlakov, A. M. Slinko, I.P. Shestakov, A.<br />

I. Shirshov, Rings that are nearly associa=ve, Academic<br />

Press, 1982.<br />

3


F-algebra:<br />

FIRST DEFINITIONS<br />

F field<br />

An F-vector space A<br />

Equipped with a bilinear mul=plica=on A×A→A<br />

(x,y) à xy<br />

AssociaCve F-algebra:<br />

(xy)z = x(yz) for all x,y, z in A<br />

WE WILL CONSIDER NON-­‐ASSOCIATIVE ALGEBRAS<br />

4


Examples<br />

Associa=ve F-algebras:<br />

Non associa=ve F-algebras:<br />

5


Jordan algebra:<br />

An F-algebra (J, •) is a Jordan algebra if:<br />

x • y = y • x (Commuta=vity)<br />

(x 2 • y) • x = x 2 • (y • x) (Jordan iden=ty)<br />

Note: x 2 :=x • x<br />

This is the defini=on of a LINEAR Jordan algebra valid for rings of<br />

scalars containing ½. There is a more modern defini=on of quadra=c<br />

Jordan algebra over arbitrary rings of scalars.<br />

6


Examples:<br />

A associa=ve and commuta=ve F-­‐algebra<br />

A alterna=ve commuta=ve algebra<br />

A associa=ve F-­‐algebra<br />

A + = (A, •)<br />

a • b = 1/2(ab+ba)<br />

Exercise: Show that these are examples of Jordan algebras<br />

7


alterna=ve algebra<br />

8


Suppose A is an algebra<br />

The commutator is [x,y]:= xy - yx<br />

The associator is (x, y, z):= (xy)z - x(yz)<br />

With this nota=on<br />

A is commuta=ve if [x, y]:= 0 for all x, y in A<br />

A is associa=ve if (x, y, z):= 0 for all x, y, z in A<br />

A is alterna=ve if<br />

A is a Jordan algebra if<br />

• It is commuta=ve: [x, y]=0<br />

(x, x, y):= 0<br />

(x, y, y):= 0 for all x, y in A<br />

• It sa=sfies the Jordan iden=ty (x 2 , y, x) = 0<br />

9


Examples<br />

A = Mn (F), A x • y = 1/2 (xy+yx)<br />

+<br />

H(A, t ) = {x in M n (F) | x = x t }<br />

Is not a subalgebra of M n (F)<br />

Is a subalgebra of A + Is Jordan<br />

Is not associa=ve<br />

10


Examples<br />

H(M n (C ) , *) = {x in M n (C ) | x = x * }<br />

Is not a subalgebra of M n (F)<br />

Is a subalgebra of A +<br />

x • y = 1/2 (xy+yx)<br />

Is not associa=ve<br />

Is Jordan<br />

H(A, *) = {x in M n (C ) | x = x * } ≤ A +<br />

x • y = 1/2 (xy+yx)<br />

x* = x t<br />

__<br />

11


Examples<br />

H(M n (C ) , *) = {x in M n (C ) | x = x * }<br />

Is not a subalgebra of M n (F)<br />

Is a subalgebra of A +<br />

(A, *)<br />

associa=ve with<br />

an involu=on<br />

x • y = 1/2 (xy+yx)<br />

x • y = 1/2 (xy+yx)<br />

Is not associa=ve<br />

Is Jordan<br />

H(A, *) = {x in M n (C ) | x = x * } ≤ A +<br />

x* = x t<br />

__<br />

12


ObservaCons:<br />

(A, *)<br />

J ≤ A +<br />

A = H(A, *) + SK(A, *)<br />

Is Jordan Is Lie<br />

Special Jordan algebra<br />

Exceptional Jordan algebra<br />

13


WHEN and WHY did the Jordan algebras appear?<br />

It’s appearance was not to study matrices and symmetric<br />

matrices…<br />

14


1932<br />

15


Matrix Interpreta=on<br />

of Quantum Mechanics<br />

Physical observables represented<br />

Observable opera=ons<br />

λx à mul=plica=on by a real<br />

x+y à addi=on<br />

x n à powers of matrices<br />

Hermi=an matrices (x*= x)<br />

Matrix opera=ons<br />

λx à mul=plica=on by a complex<br />

x+y à addi=on<br />

xy à mul=plica=on of matrices<br />

x* à adjoint operator<br />

xy is observable only if x and y commute (are simultaneously observable)<br />

x* is “invisible”<br />

16


The Jordan program<br />

hkp://en.wikipedia.org/wiki/<br />

File:Jordan,Pascual_1963_Kopenhagen.jpg<br />

Pascual Jordan<br />

(1902-­‐1980)<br />

17


The Jordan program<br />

In 1932, Jordan proposed a program to discover a new<br />

algebraic senng for quantum mechanics<br />

Ø To study the intrinsic algebraic proper=es of hermi=an<br />

matrices, without reference to the underlying<br />

(unobservable) matrix algebra.<br />

Ø To capture the algebraic essence of the physical<br />

situa=on by formal algebraic proper=es that seemed<br />

essen=al and physically significant.<br />

Ø To consider abstract systems axioma=zated by these<br />

formal proper=es and see which other new (non-­‐matrix)<br />

systems sa=sfied the axioms.<br />

18


The Jordan opera=ons<br />

Which should be the basic observable opera=ons?<br />

How to combine hermi=an matrices to obtain<br />

another hermi=an matrix?<br />

The most natural way is linearizing (forming<br />

polynomials):<br />

product by a real scalar + addi=on + powers<br />

x2 linearization<br />

½(xy+yx) = x • y<br />

The dot product appears<br />

19


Is the dot enough?<br />

Jordan thought that all the products could be expressed using •<br />

More…<br />

x 2 = ½ x • x<br />

x n<br />

That’s why we use<br />

½ in the formula<br />

xyz + zxy (Jordan triple product, obtained linearizing x3 )<br />

x -1<br />

Not xy! (fortunately)<br />

There are opera=ons with hermi=an elements that cannot<br />

be built from the •, e.g.<br />

xyzt + tzyx<br />

Jordan didn’t realise and that was good luck!<br />

20


The Jordan axioms<br />

Which law would they obey?<br />

A property of the hermi=an matrices:<br />

power-­‐associa=ve x n • x m = x n+m<br />

The Jordan iden=ty (x 2 • y)• x = x 2 • (y• x)<br />

Another basic property of the hermi=an matrices:<br />

posi=ve-­‐definite formally real<br />

The -­‐Jordan algebra is called formally real if<br />

x 1 2 +…+ xn 2 = 0 implies x1 = … = x n = 0<br />

In fact, there are some laws sa=sfied by • for hermi=an<br />

elements that cannot be obtained from the Jordan iden=ty.<br />

21


The Jordan axioms<br />

Which law would they obey?<br />

A property of the hermi=an matrices:<br />

power-­‐associa=ve x n • x m = x n+m<br />

The Jordan iden=ty (x 2 • y)• x = x 2 • (y• x)<br />

Another basic property of the hermi=an matrices:<br />

posi=ve-­‐definite formally real<br />

The -­‐Jordan algebra is called formally real if<br />

x 1 2 +…+ xn 2 = 0 implies x1 = … = x n = 0<br />

In fact, there are some laws sa=sfied by • for hermi=an<br />

elements that cannot be obtained from the Jordan iden=ty.<br />

22


The Jordan axioms<br />

The first iden=ty<br />

Which law would<br />

that<br />

they obey?<br />

A property of the hermi=an matrices:<br />

power-­‐associa=ve xn • xm = xn+m cannot be built from the<br />

Jordan product was<br />

The Jordan iden=ty (x 2 • y)• x = x 2 • (y• x)<br />

founded in 1963.<br />

Another basic property of the hermi=an matrices:<br />

posi=ve-­‐definite formally real<br />

It was needed to wait for<br />

The -­‐Jordan algebra is called formally real if<br />

x 1 2 +…+ xn 2 = 0 implies x1 = … = x n = 0<br />

30 years<br />

In fact, there are some laws sa=sfied by • for hermi=an<br />

elements that cannot be obtained from the Jordan iden=ty.<br />

23


There were three basic examples of linear Jordan algebras<br />

known by Jordan<br />

The FIRST example:<br />

Full algebras<br />

A à -associa=ve algebra, A + = (A, •) is a Jordan<br />

algebra<br />

x • y = 1/2 (xy+yx)<br />

It is not necessarily formally real, for instance, for<br />

A = M n (D), n > 1<br />

24


There were three basic examples of linear Jordan algebras<br />

known by Jordan<br />

The SECOND example:<br />

HERMITIAN algebras<br />

(A, *) à - associa=ve algebra with involu=on,<br />

H(A, *) := {x in A | x* = x}<br />

is a Jordan algebra x • y = 1/2 (xy+yx)<br />

*: A→A, linear map<br />

(a*)* = a<br />

(ab)*=b*a*<br />

H(A, *) ≤ A+<br />

25


There were three basic examples of linear Jordan algebras<br />

known by Jordan<br />

The SECOND example:<br />

HERMITIAN algebras<br />

(A, *) à - associa=ve algebra with involu=on,<br />

H(A, *) := {x in A | x* = x}<br />

is a Jordan algebra<br />

A parCcular case:<br />

à (D,-) unital associa=ve algebra with involu=on a ā<br />

à A= M n (D) with * the conjugate traspose, that is, if x = (a ij ),<br />

then x*=(ā ji ). H(A,*) = H n (D, -)<br />

If D = R, C or H with their usual involu=ons then Hn (D, -­‐) is<br />

formally real.<br />

Exercise<br />

26


There were three basic examples of linear Jordan algebras<br />

known by Jordan<br />

The THIRD example:<br />

SPIN factor<br />

< , > euclidean inner product on n<br />

JSpinn := 1 n<br />

is a formally real Jordan algebra: Jordan spin factor<br />

1 is the unit element<br />

v • w = 1 for all v, w in<br />

n<br />

27


P. Jordan, J. v. Neumann, E. Wigner, On an algebraic<br />

generaliza=on of the quantum mechanical formalism, Annals<br />

of Math. Vol. 35 (1), 29-­‐64 (1934).<br />

Theorem<br />

CLASSIFICATION (I)<br />

hkp://www.jstor.org/stable/1968117<br />

Every finite-­‐dimensional formally real Jordan algebra is a direct sum<br />

of a finite number of simple ideals.<br />

Every finite-­‐dimensional simple formally real Jordan algebra is isomorphic to<br />

one of:<br />

H n (R, -), H n (C, -), H n (H, -), H 3 (O, -) and JSpin n<br />

28


Special and excep=onal Jordan algebras<br />

J à Jordan algebra<br />

J is special if it is isomorphic to a subalgebra of A + ,<br />

for some associa=ve algebra A.<br />

Examples<br />

-<br />

-<br />

-<br />

J is excepConal if it is not special.<br />

H n (R, -­‐), H n (C, -­‐), H n (H, -­‐) are special Jordan algebras.<br />

Jspin n is a special Jordan algebra.<br />

H 3 (O,-­‐) seems to be exceptional, is it?<br />

Exercise:<br />

Show that Jspinn is a subalgebra of A + , for A the<br />

associative) Clifford algebra generated by elements<br />

v1 ,…,vn such that v 2<br />

i = 1, vivj + vjvi = 0).<br />

29


Albert’s Theorem<br />

The surprise<br />

The algebra H 3 (O, -­‐) is an excep=onal Jordan algebra<br />

of dimension 27.<br />

Now it is called the Albert algebra (as well as some<br />

extensions).<br />

Exercise:<br />

Show it.<br />

30


Jordan’s idea<br />

The end<br />

(Physical end)<br />

Jordan wanted to find a formally real Jordan algebra of infinite<br />

dimension, not special (since for him the (associa=ve) product xy was<br />

not relevant).<br />

His idea when classifying the simple finite-­‐dimensional ones was to find<br />

a family of excep=onal formally real Jordan algebras depending on a<br />

parameter, so that they could provide a clue to find one infinite-­‐<br />

dimensional formally real Jordan algebras (perhaps by taking some kind<br />

of limit with the parameter).<br />

But with only one example in the list, this list was not of help to find<br />

the excep=onal searched system.<br />

The reality<br />

(1979) Zelmanov proved that such an algebra does not exist. In fact, he<br />

proved that the examples of Jordan algebras in arbitrary dimension are<br />

essen=ally generaliza=ons of this one finite dimensional).<br />

31


Jordan’s idea<br />

The end<br />

(Physical end)<br />

With the only exception of the<br />

Jordan wanted to find a formally real Jordan algebra of infinite<br />

dimension, not special (since for him the (associa=ve) product xy was<br />

not relevant).<br />

Albert algebra, Jordan algebras<br />

His idea when classifying the simplefinite-­‐dimensional ones was to find<br />

a family of excep=onal formally real Jordan algebras depending on a<br />

parameter, so that they could provide a clue to find one infinite-­‐<br />

dimensional formally real Jordan algebras (perhaps by taking some kind<br />

of limit with the parameter).<br />

But with only one example in the list, this list was not of help to find<br />

the excep=onal searched system.<br />

hide an associative structure:<br />

Its Jordan product comes from<br />

The reality<br />

an associative one!!<br />

(1979) Zelmanov proved that such an algebra does not exist. In fact, he<br />

proved that the examples of Jordan algebras in arbitrary dimension are<br />

essen=ally generaliza=ons of this one finite dimensional).<br />

32


Only in the sense that he has not had capture all the proper=es of<br />

the hermi=an matrices.


IDENTITIES<br />

An idenCty in an algebra A is a polynomial f(x 1 , x 2 ,…, x n ) € F[x 1 , x 2 ,…, x n ]<br />

such that f(a 1 , a 2 ,…, a n ) = 0 for all a 1 , a 2 ,…, a n € A.<br />

Example: xy-yx and (x 2 y)x-x 2 (yx) are iden==es in any Jordan algebra (char<br />

F≠ 2). (Moreover, the variety of Jordan algebras is defined by such iden==es.)<br />

Luiz A. Peresi will speak about<br />

iden==es in free algebras.


IDENTITIES<br />

An idenCty in an algebra A is a polynomial f(x 1 , x 2 ,…, x n ) € F[x 1 , x 2 ,…, x n ]<br />

such that f(a 1 , a 2 ,…, a n ) = 0 for all a 1 , a 2 ,…, a n € A.<br />

Example: xy-yx and (x 2 y)x-x 2 (yx) are iden==es in any Jordan algebra (char<br />

F≠ 2). (Moreover, the variety of Jordan algebras is defined by such iden==es.)<br />

Shirshov and Macdonald’s Theorem<br />

If f(x 1 , x 2 , x 3 ) € F[x 1 , x 2 , x 3 ] is of degree 0 or 1 in x 3 , and is<br />

an iden=ty in every special Jordan algebra with 1, then f<br />

is an iden=ty in every Jordan algebra.<br />

Corollary<br />

Every Jordan algebra generated by two elements is special.<br />

35


IDENTITIES<br />

An idenCty in an algebra A is a polynomial f(x 1 , x 2 ,…, x n ) € F[x 1 , x 2 ,…, x n ]<br />

such that f(a 1 , a 2 ,…, a n ) = 0 for all a 1 , a 2 ,…, a n € A.<br />

Example: xy-yx and (x 2 y)x-x 2 (yx) are iden==es in any Jordan algebra (char<br />

F≠ 2). (Moreover, the variety of Jordan algebras is defined by such iden==es.)<br />

Shirshov and Macdonald’s Theorem<br />

If f(x1 , x2 , x3 ) € F[x1 , x2 , x3 ] is of degree 0 or 1 in x3 , and is<br />

anExercise: iden=ty in every special Jordan algebra with 1, then f<br />

2<br />

is an iden=ty Show in every it. (UJordan x =2Lx -­‐ L algebra. x ) 2<br />

Immediate applicaCon: we may reduce to special Jordan algebras to prove<br />

iden==es such as Macdonald’s iden=ty:<br />

36


An s-­‐iden=ty<br />

2{xzx}•{y{zy 2 z}x}-2{yzy}•{x{zx 2 z}y} =<br />

={x{z{x{yzy}y}z}x}-{y{z{y{xzx}x}z}y}<br />

This iden=ty is valid in any special Jordan algebra.<br />

4{xzx} • {y{zy 2 z}x} -­‐ 4{yzy} • {x{zx 2 z}y}<br />

= xzx(yzy 2 zx + xzy 2 zy) + (yzy 2 zx + xzy 2 zy)xzx -­‐ yzy(yzx 2 zy + yzx 2 zx) -­‐ (xzx 2 zy + yzx 2 zx)yzy<br />

= xzxyzy 2 zx + xzy 2 zyxzx -­‐ yzyxzx 2 zy -­‐ yzx 2 zxyzy<br />

= xz(xyzy 2 + y 2 zyx)zx -­‐ yz(yxzx 2 + x 2 zxy)zy<br />

= 2{x{z{x{yzy}y}z}x} -­‐ 2{y{z{y{xzx}x}z}y}<br />

But not in any Jordan algebra!<br />

That kind of iden==es are called s-­‐idenCCes.<br />

Concretely, that one is Glennie’s idenCty G 9 , discovered in 1963.<br />

37


Excep=onality of Albert’s Algebra<br />

It is enough to check that it does not sa=sfy:<br />

2{xzx}•{y{zy 2 z}x}-2{yzy}•{x{zx 2 z}y} = {x{z{x{yzy}y}z}x}-{y{z{y{xzx}x}z}y}<br />

Exercise:<br />

Take u,v,w € O such that (u,v,w) ≠ 0 and check that the<br />

following choice of elements in Albert’s algebra does not<br />

sa=sfy the iden=ty<br />

38


(From McCrimmon’s book)<br />

Jordan’s program<br />

Jordan thought that there were no s-­‐iden==es<br />

Jordan’s goal was to capture the algebraic behavior of hermi=an operators in<br />

the Jordan axioms. But Albert’s algebra does not sa=sfies all the algebraic<br />

proper=es of the Copenhagen model. Jordan was wrong in thinking that his<br />

axioms had captured the hermi=an essence – he overlooked some algebraic<br />

proper=es of hermi=an matrices.<br />

First, he missed some algebraic opera=ons which could not be built from the<br />

bullet: the symmetric n-­‐tad products {x 1 ,…,x n } := x 1 … x n + x n … x 1 cannot be<br />

expressed in terms of the bullet for n ≥ 4. (Their inclusion in the axioms would<br />

have excluded both Albert’s algebras and spin factors, landing back in<br />

Copenhagen with nothing but hermi=an algebras.)<br />

Secondly, Jordan missed some laws for the bullet which cannot be derived<br />

from the Jordan iden=ty, the s-­‐iden==es, which are in fact just the algebraic<br />

iden==es sa=sfied by all hermi=an matrices which are not consequences of<br />

the Jordan axioms.<br />

39


Algebraists become interested...<br />

A big amount of unexpected connec=ons to other branches of<br />

mathema=cs appeared!<br />

Jordan -­‐ von Neumann – Wigner’s Theorem was<br />

extended to finite-­‐dimensional Jordan algebras over an<br />

arbitrary algebraically closed field of characterisCc ≠ 2:<br />

none new algebra really appeared.


Let A be an F-­‐algebra<br />

Some defini=ons<br />

A is unital if there is a unit element 1€A such that 1x = x1 = x for all x.<br />

Any algebra A can be embedded in a unital algebra F1 A=:Â<br />

A subalgebra B of A is an F-vector subspace such that BB B.<br />

An ideal I A is an F-vector subspace such that IA+AI I.<br />

If I A, we may consider the quotient algebra =A/I: all cosets [x]=x+I<br />

with operations α[x]:=[αx], [x]+[y]:=[x+y] and [x][y]:=[xy].<br />

A is simple if it has not proper ideals and AA ≠ 0.<br />

A is semisimple if it is a finite direct sum of simple ideals.<br />

The center Cent(A) = {c € A| c x = x c, c(xy)=(cx)y=x(cy), x, y € A}<br />

It is a ring of scalars, that is, associative and commutative.<br />

If A is unital, A can be considered as an algebra over its center<br />

A is simple-central if it is simple, unital and Cent(A)=F1.<br />

IMPORTANT BUILDING BLOCKS<br />

FOR THE STRUCTURE THEORY<br />

41


Let F be an algebraically closed field, char F ≠ 2.<br />

Let J be a finite-­‐dimensional Jordan algebra.<br />

Then:<br />

• There is an unique maximal nilpotent ideal, Rad J. The algebra J is<br />

semisimple iff Rad J=0. The quo=ent J/Rad(J) is semisimple.<br />

• If J is semisimple, it has 1, and its decomposi=on as a direct sum of<br />

simple ideals is unique: the simple summands are precisely the minimal<br />

ideals.<br />

• If J is simple, it is simple-­‐central over F.<br />

• J is simple if and only if it is one of this list:<br />

-<br />

-<br />

-<br />

CLASSIFICATION (II)<br />

F, the ground field<br />

JSpin n (F), the spin factor<br />

H n (C), hermitian matrices (=H(C,*), for C a (split)<br />

composition algebra (there are exactly 4) (n=3 if dim C=8))<br />

42


Some details about the list<br />

The spin factor (direct generalization): if is a nondegenerate<br />

symmetric bilinear form in F n , then we define JSpin n (F)=F1 F n as the<br />

Jordan algebra given by:<br />

◊ 1 is the unit element<br />

◊◊ If v,w € F n , v • w:= 1<br />

Exercise: Check that the spin factor is special.<br />

Show that it is simple if n≠ 1.<br />

Concerning the hermitian matrices, which is the coordinate<br />

algebra, the composition algebra? Which are they? (The 4<br />

ones jointly their involutions…)<br />

More about composi=on algebras<br />

43


The n-­‐Squares Problem<br />

Everything started by a number Theory problem…<br />

It was known from centuries that the sums of two squares can be composed:<br />

For us, it is enough to think that the module of the product of two complex<br />

numbers is the product of the modules, and apply to z 1 =x 0 +ix 1 and z 2 =y 0 +iy 1 ).<br />

Euler found (in 1748) a formula for the sum of four squares.<br />

And Cayley (in 1845) found another one for sums of eight squares, but there is no<br />

enough place here! (neither for copying nor for remembering, except if we look<br />

for some more conceptual approach to understand these formulae!)<br />

44


Forms adminng composi=on<br />

A composition algebra is a unital F-algebra A (for F an arbitrary field) with a<br />

nondegenerate quadratic form n: A→F admitting composition, that is,<br />

n(xy) = n(x)n(y)<br />

Nondegenerate means that n(x,A)=0 iff x=0, where n is here the polar form.<br />

The existence of an<br />

m-­‐squares formula<br />

m=2<br />

m=4<br />

m=8<br />

The existence of an m-­‐dimensional<br />

composi=on R-­‐algebra<br />

C (complex numbers)<br />

H (quaternions)<br />

O (octonions)<br />

These three algebras are DIVISION algebras, which for<br />

composi=on algebras is equivalent to:<br />

** There are no (≠ 0) isotropic elements (n(x)=0)<br />

and to<br />

** There are no zero-­‐divisors (x, y≠ 0 such that xy=0)<br />

45


Recalling H and O<br />

46


Recalling H and O<br />

47


C, H and O are real composi=on algebras, but not all of them.<br />

Anyway, we need composi=on algebras over algebraically closed fields.<br />

Thus we try to build composiCon algebras.<br />

The Cayley-­‐Dickson process<br />

Assume that A is an associa=ve composi=on algebra: it is unital and n: A→F is<br />

a nondegenerate norm admitting composition. The standard conjugation given<br />

by ā = n(a,1)1- a verifies<br />

Take 0≠λ, and define in CD(A,λ) = A Au (two formal copies of A) with<br />

the following product and norm:<br />

The obtained algebra is AGAIN a composi=on algebra.<br />

The process is called the Cayley-­‐Dickson process.<br />

48


The composi=on algebras<br />

Generalized Hurwitz Theorem<br />

Every composi=on algebra over a field F is isomorphic to one of the<br />

following types:<br />

q F (if char F≠2). It is a field.<br />

q K(µ) := F1+Fv with v 2 =v+µ and 4µ+1≠0. It is commuta=ve and<br />

associa=ve.<br />

q Q(µ, β):= CD(K(µ), β). These are called quaternion algebras. They<br />

are associa=ve but not commuta=ve.<br />

q C(µ, β, γ) := CD(Q(µ , β), γ). These are called Cayley algebras or<br />

octonion algebras. They are not associa=ve (but alterna=ve).<br />

49


The split composi=on algebras<br />

A composi=on algebra is split if it has zero divisors, or equivalently, isotropic<br />

elements. Over algebraically closed fields, every composiCon algebra is split. And<br />

it is not difficult to check that (over any F) there is only one split composi=on<br />

algebra of each dimension 1,2,4,8:<br />

♣ F<br />

♣ K(0). Isomorphic to F⊕F with the exchange involu=on and n(a):=a 2<br />

♣ Q(0,1). Isomorphic to M 2 (F) with involu=on the conjugate transpose<br />

and n(a)=det(a).<br />

♣ C(0,1,1). It is isomorphic to the Zorn matrices<br />

50


CLASSIFICATION (II REVISTED)<br />

The simple finite-­‐dimensional Jordan algebras over an algebraically<br />

closed field F (char F ≠ 2) are:<br />

F<br />

JSpin n (F), n ≥2<br />

H n (F), n ≥3<br />

H n (F⊕F) n ≥3<br />

H n (M 2 (F)), n ≥3<br />

H 3 (C(0,1,1))<br />

51


What happens if the characteris=c of the field is 2?<br />

Or simply, if we consider a ring of scalars instead of a field, like Z?


Some important Operators<br />

If x€J, the left and right mutiplication operators:<br />

L x : J → J , L x (y) = xy<br />

R x : J → J , R x (y) = yx<br />

In these terms, J is a Jordan algebra if and only if<br />

L x = R x<br />

[R x2 , R x ]=0<br />

Now take the quadratic operator<br />

U x : J → J , U x (y) = 2x • (x • y) - x 2 • y<br />

That is, U x = 2 L x 2 – Lx2<br />

Note: if J is special (there is a subjacent product xy), then U x (y) = xyx<br />

53


About these and other “natural” operators<br />

The U-operator satisfies the so called Fundamental Formula:<br />

We define the Jordan triple product:<br />

{x, y, z} = 2 (x • (y • z) + (x • y) • z - (x • z) • y)<br />

Note: if J is special (there is a subjacent product xy), then {x, y, z} = xyz + zyx<br />

We have a related operator:<br />

V x,y : J → J , V x,y (z) = {x,y,z}<br />

Which, at the same =me, can be wriken in terms of the U-­‐operator:<br />

If U x,y := U x+y – U x – U y , then {x,y,z} = U x,z (y)<br />

54


The quadra=c program<br />

ObjecCve: give a defini=on of Jordan algebra valid also for fields of<br />

charateris=c 2, and for arbitrary rings of scalars, as the integers.<br />

Idea: switch from the Jordan product x• y = ½ (xy+yx)<br />

to the product U x (y) = xyx<br />

Search<br />

A quadratic axiomatization such that:<br />

• Agrees with the usual one over scalars containing ½<br />

• Admittes all basic examples in characteristic 2<br />

• Admittes essentially nothing new in simple algebras<br />

55


(Unital) quadra=c Jordan algebras<br />

McCrimmon, 1967<br />

A unital quadratic Jordan algebra J (over a ring of scalars Φ) is a<br />

Φ-module with<br />

A quadratic map U: J → End Φ (J)<br />

x U x<br />

and a unit element 1 such that the following operator identities hold in all<br />

scalar extensions<br />

This is equivalent to the fact that the linearizations are valid in J.<br />

56


About this connec=on<br />

Kantor : “There are no Jordan algebras, there are only Lie algebras.”<br />

58


About this connec=on<br />

Kantor : “There are no Jordan algebras, there are only Lie algebras.”<br />

K. McCrimmon: “Of course, this can be turned around: nine Cmes<br />

out of ten, when you open up a Lie algebra you find a Jordan<br />

algebra inside which makes it thick.”<br />

59


The deriva=on algebra<br />

A derivaCon of a Jordan algebra J is an endomorphism D of J such that<br />

D(xy) = D(x)y+xD(y). The deriva=ons form a Lie algebra, der(J).<br />

Example: If J=H(A,*) for A associa=ve with involu=on, each<br />

x € A such that x*=-­‐x provides the deriva=on ad x= L x -­‐ R x €<br />

der(J).<br />

60


The deriva=on algebra<br />

A derivaCon of a Jordan algebra J is an endomorphism D of J such that<br />

D(xy) = D(x)y+xD(y). The deriva=ons form a Lie algebra, der(J).<br />

Example: If J=H(A,*) for A associa=ve with involu=on, each<br />

x € A such that x*=-­‐x provides the deriva=on ad x= L x -­‐ R x €<br />

der(J).<br />

In fact, Skew(A,*) ≈ ad{x € A| x*=-­‐x} = der (J)<br />

Example: If J=Albert algebra, der(J)=F 4 ! now we<br />

check it<br />

61


The deriva=on algebra<br />

A derivaCon of a Jordan algebra J is an endomorphism D of J such that<br />

D(xy) = D(x)y+xD(y). The deriva=ons form a Lie algebra, der(J).<br />

Example: If J=H(A,*) for A associa=ve with involu=on, each<br />

x € A such that x*=-­‐x provides the deriva=on ad x= L x -­‐ R x €<br />

der(J).<br />

In fact, Skew(A,*) ≈ ad{x € A| x*=-­‐x} = der (J)<br />

Example: If J=Albert algebra, der(J)=F 4 !<br />

The exceptional Lie algebras are related to<br />

the exceptional Jordan algebras<br />

now we<br />

check it<br />

62


Some deriva=ons<br />

In (x,y,x 2 )=0, replace x by x+λz:<br />

Now replace x by x+λw:<br />

Rewrite:<br />

Interchange x and y and subtract:<br />

Note that D € End(J) is a deriva=on iff [R x ,D]= R D(x)<br />

63


Some deriva=ons<br />

In (x,y,x 2 )=0, replace x by x+λz:<br />

Now replace x by x+λw:<br />

Rewrite:<br />

Interchange x and y and subtract:<br />

Note that D € End(J) is a deriva=on iff [R x ,D]= R D(x) , hence<br />

[R x ,R y ] € Der(J)<br />

64


Some deriva=ons<br />

In (x,y,x 2 )=0, replace x by x+λz:<br />

Now replace x by x+λw:<br />

Rewrite:<br />

Interchange x and y and subtract:<br />

Note that D € End(J) is a deriva=on iff [D,R x ]= R D(x) , hence<br />

[R x ,R y ] € Der(J), moreover,<br />

{Σ[R xi ,R yi ] |x i ,y i € J}=[R J ,R J ] is a Lie subalgebra of Der (J).<br />

65


Albert algebra<br />

An arbitrary element in the Albert algebra, say J, is, for α 1 , α 2 , α 3 € F<br />

and a,b,c € C:<br />

With the mul=plica=on table summarized as follows:<br />

66


for<br />

More proper=es of the Albert algebra<br />

Every element x € J as above verifies a cubic equa=on<br />

x 3 -­‐ T(x) x 2 + Q(x) x-­‐ N(x)1 = 0<br />

T(x) = α 1 + α 2 + α 3<br />

Q(x) = α 1 α 2 + α 2 α 3 + α 3 α 1 -­‐ n(a) -­‐ n(b) -­‐ n(c)<br />

N(x) = α 1 α 2 α 3 -­‐ α 1 n(α) -­‐ α 2 n(b) -­‐ α 3 n(c) + t(abc)<br />

As 1=e 1 +e 2 +e 3 is a decomposi=on into orthogonal idempotents, we have the<br />

generalized Peirce decomposi=on:<br />

J = J 11 + J 22 + J 33 + J 23 + J 13 + J 12<br />

< , >: J × J → F, = tr R xy is a trace form,<br />

< , , >: J × J × J →J, = is trilinear.<br />

J ii = Fe i<br />

J ij = {a (k) |a € C} = C (k)<br />

Exercise: Let d €End J.<br />

Show that d is a derivaCon iff d leaves both < , > and<br />

< , , > invariant<br />

67


Der(Albert algebra)= f 4<br />

Jordan algebras<br />

Lie algebras<br />

DerivaCon algebras of Jordan algebras<br />

provides models for Lie algebras<br />

DerivaCon algebras of excepConal Jordan algebras<br />

provides models for excepConal Lie algebras<br />

68


J. E. Humphreys, Introduc=on to Lie algebras and<br />

representa=on theory, Springer-­‐Verlag, New York,Heidelberg,<br />

Berlin 1972.<br />

ClassificaCon of irreducible root systems by using Dynkin diagrams<br />

69


PROOF:<br />

First step: dim Der(J)=52<br />

Der(Albert algebra)= f 4<br />

D 0 := {d €Der(J) | d(e i )=0 for all i} ≅ so(C,n) (≅d 4 , simple dim 28)<br />

If U€ so(C,n) , D U :J → J given by D U (e i ):=0, D U (a (i) ) := U i (a) (i) is a<br />

deriva=on, for U 1 = SUS -­‐1 , U 2 =U’ and U 3 =U’’. These endomorphisms<br />

are given by the triality: U(xy)=U’(x)y+xU’’(y) and S(a)=ā denotes the<br />

conjuga=on in C.<br />

Di := {[Ra(i) , Rej-­‐ek ]| a€C} ( Der(C,n)); (i,j,k cyclic permuta=on of<br />

1,2,3) (dim 8, since ≅ C as vector space, as {[R a(i) , R ej-­‐ek ](e j )= -­‐1/2a(i) )<br />

Now check that Der (J)= D 0 + D 1 + D 2 + D 3 (for use that de i € C (j) +C (k)<br />

using the generalized Peirce decomposi=on) and that the sum is direct.<br />

70


f 4 is the algebra of deriva=ons of the<br />

Second step: Der(J) is simple<br />

Albert algebra (bis)<br />

Computa=ons show that [D i , D i ] D 0 and [D i , D j ] D k (i,j,k dis=nct)<br />

Lemma: A fin-­‐dim arbitrary algebra A = S + I, dim S > dim I, S simple, SI=I IS.<br />

Then every proper ideal of A is contained in I.<br />

Finally if B J, B D 1 + D 2 + D 3 . Take d=d 1 +d 2 +d 3 € B. Fix i, there exists<br />

b € C with d i =[R b(i),R ej-­‐ek ]. Thus [d i , [R a(i),R ej-­‐ek ]](c(i) )=(-­‐n(b,c)a+n(a,c)b) (i)<br />

=0, since [d i , [R a(i),R ej-­‐ek ]] €D 0 (D 1 + D 2 + D 3) =0, so n(a,c)b=n(b,c)a for<br />

all a,c so that b=0.<br />

Corollary: Der J=[R J , R J ]=Inder J.<br />

Proof: [R J , R J ] is an ideal of Der J, which is simple.<br />

71


The structure algebras<br />

The structure algebra is the Lie algebra str(J):= der(J) + R J<br />

The inner structure algebra is the Lie algebra Instr(J):= Inder(J) + R J<br />

Example: If J=H(A,*) for A associa=ve with involu=on, then str(J)=A -­‐<br />

Example: If J=Albert algebra, str(J)=F+E6 !<br />

now we<br />

check it<br />

The exceptional Lie algebras are related to<br />

the exceptional Jordan algebras<br />

72


PROOF:<br />

Str(Albert alg)= e 6 (+center)<br />

First step: dim R J0 +[R J , R J ]=78<br />

Direct sum as 1 € J<br />

Second step: it is simple<br />

If B is an ideal of RJ0 +[RJ , RJ ], by lemma, B RJ0 . Take Rc € B,<br />

then [Rc ,Rw ] € RJ0 [RJ , RJ ] =0, so c is in the center=F and in F J0 =0<br />

E6 B6 C6 73


PROOF:<br />

Str(Albert alg)= e 6 (+center)<br />

First step: dim R J0 +[R J , R J ]=78<br />

Direct sum as 1 € J<br />

Second step: it is simple<br />

If B is an ideal of R J0 +[R J , R J ], by lemma, B R J0 . Take R c € B,<br />

then [R c ,R w ] € R J0 [R J , R J ] =0, so c is in the center=F and in F J 0 =0<br />

One more thing: J is an irreducible module of dimension 27<br />

As J simple, J is RJ -­‐irreducible and RJ0 -­‐irreducible, so that alg< RJ0 > -­‐irreducible<br />

E6 B6 C6 74


Tits-­‐Kantor-­‐Koecher construc=on<br />

If J is a Jordan algebra, we can build the Lie algebra (obviously Z-­‐graded)<br />

L:= TKK (J) := L -­‐1 ⊕ L 0 ⊕ L 1<br />

Where L -­‐1 = J, L 1 = J, L 0 = Instr(J):= V J, J (inner structure algebra, which<br />

is a Lie algebra since [V x,y , V u,v ] = V {x,y,u},v – V u,{y,x,v} )<br />

with an=commuta=ve product given by<br />

if T € Instr(J), x,y € J and the involu=on is V x,y *= V y,x .<br />

Thus TKK (J) is a Lie algebra with involu=on x 1 + V x,y + y -­‐1<br />

y 1 + V y,x + x -­‐1<br />

75


e 7<br />

TKK (Albert algebra) =The excep=onal Lie algebra e 7<br />

Proof: Simple, dim 133.<br />

But TKK-­‐construc=on was not only important<br />

because of its connec=on to Lie algebras.<br />

76


A consequence of this Tits-­‐Kantor-­‐Koecher construc=on was the<br />

enlargement of the family of Jordan objects.<br />

77


Jordan triples<br />

Meyberg (in 1969) thought what was necessary to add to a Jordan algebra J in<br />

order to get that TKK(J) is a Lie algebra.<br />

DefiniCon:<br />

T is a (linear) Jordan triple system if there is a trilinear product<br />

{.,.,.}:T×T× T → T such that:<br />

{x,y,z} = {z,y,x}<br />

{x,y,{u,v,w}} = {{x,y,u},v,w}-{u,{y,x,v},w}+{u,v,{x,y,w}}<br />

(Remark: there is also another axioma=za=on in the pressence of ½ and 1/3)<br />

Example: If J is a Jordan algebra, also J is a Jordan triple system with<br />

{x, y, z} = 2 (x • (y • z) + (x • y) • z - (x • z) • y)<br />

(respectively {x, y, z} = U x,z (y)), as usual. That is, simply forget the bilinear<br />

product (respec=vely the U-­‐quadra=c product).<br />

Example: M p×q (R) for p≠q with the product {x,y,z} = x y t z + z y t x is a Jordan<br />

triple system which does not come from a bilinear product.<br />

78


Jordan pairs<br />

Meyberg himself suggested that the en=re TKK construc=on would work for<br />

“pairs”, that is, for two independent spaces J + , J -­‐ ac=ng on each other:<br />

DefiniCon:<br />

(V + ,V -­‐ ) is a (linear) Jordan pair if there are two trilinear maps {.,.,.}:V + ×V -­‐ × V + → V +<br />

such that for each s €{+,-­‐}, x s ,u s ,w s €V s , y -s ,v -s €V -s<br />

Example: V + = M p×q (R) , V -­‐ = M q×p (R)<br />

{.,.,.}:V -­‐ ×V + × V -­‐ → V -­‐<br />

{x s ,y -s ,z s } = {z s ,y -s ,x s }<br />

{x s ,y -s ,{u s ,v -s , w s }} = {{x s ,y -s ,u s },v -s , w s }-{u s ,{y -s ,x s ,v -s },w s }+{u s ,v -s ,{x s ,y -s ,w s }}<br />

Example: If T is a Jordan triple system, it can be doubled into a pair,<br />

V + = T , V -­‐ = T, {x s ,y -s ,z s } = {x,y,z}<br />

Example: V + =Hom F (W,U), V -­‐ = Hom F (U,W). Hence not any pair comes from a<br />

triple system (these two spaces have not neccesarily the same dimension).<br />

79


Jordan pairs (quadra=c version)<br />

DefiniCon:<br />

A pair V=(V + ,V -­‐ ) of F-­‐vector-­‐spaces is a (quadra=c) Jordan pair if there are two<br />

quadra=c maps Q + : V + → Hom F (V -­‐ ,V + ) and Q -­‐ : V -­‐ → Hom F (V + ,V -­‐ ) which sa=sfy the<br />

following iden==es in all scalar extensions:<br />

(JP1) {x,y,Q s (x)z}= Q s (x){yxz}<br />

(JP2) {Q s (x)y,y,z}= {x,Q -­‐s (y)x,z}<br />

(JP3) Q s (Q s (x)y)= Q s (x)Q -­‐s (y) Q s (x)<br />

80


(Loos’ book)<br />

Mo=va=on for Jordan pairs<br />

DefiniCon:<br />

An element € J is inver=ble if it has a Jordan inverse, that is, y such that<br />

x • y = 1, x 2 • y = x.<br />

This happens iff U x is an inver=ble operator.<br />

DefiniCon:<br />

If u € J is inver=ble, we may consider the u-­‐isotope J (u) :<br />

the same vector space J but with<br />

x • u y = x • (u • y) + (x • u) • y – u • (x • y)<br />

The descrip=on is simple using the quadra=c product: U x (u) =Ux U u<br />

DefiniCon:<br />

Two Jordan algebras J and J’ are isotopic if there is an automorphism between J<br />

and J’ (u) (called isotopism)<br />

Several important theorems in the theory<br />

of Jordan algebras hold only up to isotopy<br />

81


Mo=va=on for Jordan pairs<br />

The autotopism group of J, usually called the structure group and<br />

denoted by Str(J), plays a more important role than the<br />

automorphism group.<br />

There ought to be some algebraic object associated to J which<br />

somehow incorporates all homotopes of J and whose<br />

automorphism group is just the structure group of J.<br />

This object is the Jordan pair (J,J) associated with J .<br />

Automorphism group<br />

Structure group<br />

Lie algebra<br />

Deriva=on algebra<br />

Structure algebra<br />

82


Take T= J with the<br />

product {x,y,z}<br />

Rela=ons among these objects<br />

Jordan Triple System T<br />

T=V + ⊕V -­‐<br />

V=(T,T)<br />

Jordan Algebra J Jordan Pair V =(V + ,V -­‐ )<br />

And proper=es go up and down and conversely very smoothly<br />

83


Not only<br />

Jordan algebras Lie algebras<br />

but<br />

Jordan excep=onal algebras Lie excep=onal algebras


Freudenthal-­‐Tits magic square<br />

It combines alterna=ve and Jordan algebras in a very exci=ng<br />

characteriza=on of the excep=onal simple Lie algebras (Tits).<br />

char F≠ 2,3<br />

U composiCon algebra<br />

J central simple Jordan algebra of degree 3<br />

L (U, J) = Der (U) ⊕ U 0<br />

J 0 ⊕ Der (J)<br />

• [Der(U), Der(J)] =0<br />

• Der (U) and Der (J) are Lie subalgebras<br />

• [a x, d] = d(a) x<br />

• [a x, D] = a D(x)<br />

• [a x, b y] = 1/12 D a,b + (a * b) (x * y) -­‐ (a,b) [R x ,R y ]<br />

If a,b €U 0 , x,y€ J 0 , d € Der(U), D € Der (J)<br />

Where (a,b) is the bilinear form in U, a * b=ab-­‐(a,b)1U is the projec=on of the<br />

product inU0 , =t(xy)=3/(dim J) trace Rxy , x * y=xy-­‐1/31J is the<br />

projec=on of the product in J 0 , D a,b =[L a ,L b ]+[L a ,R b ]+[R a ,R b ] €Der (U)<br />

85


F algebraically closed, char F=0<br />

Magic Square (II)<br />

Recall that there are 4 composi=on (split) algebras: F, K=K(0) ,Q=Q(0,1), C =C(0,1,1)<br />

U<br />

F<br />

K<br />

Q<br />

C<br />

J F<br />

H3 (F) H3 (K) H3 (Q) H3 (C)<br />

86


F algebraically closed, char F=0<br />

Magic Square (II)<br />

Recall that there are 4 composi=on (split) algebras: F, K=K(0) ,Q=Q(0,1), C =C(0,1,1)<br />

U<br />

F<br />

K<br />

Q<br />

C<br />

J F H3 (F) H3 (K) H3 (Q) H3 (C)<br />

0<br />

0<br />

sl(2)<br />

sl(2) sl(3) sp(6)<br />

sl(3) sl(6)<br />

sp(6)<br />

sl(3)⊕sl(3)<br />

sl(6)<br />

o(12)<br />

g2 f e<br />

4<br />

6 e7 e8 f 4<br />

e 6<br />

e 7<br />

87


F algebraically closed, char F=0<br />

Magic Square (II)<br />

Recall that there are 4 composi=on (split) algebras: F, K=K(0) ,Q=Q(0,1), C =C(0,1,1)<br />

U<br />

F<br />

K<br />

Q<br />

C<br />

J F H3 (F) H3 (K) H3 (Q) H3 (C)<br />

0<br />

0<br />

sl(2)<br />

sl(2) sl(3) sp(6)<br />

sl(3) sl(6)<br />

sp(6)<br />

sl(3)⊕sl(3)<br />

sl(6)<br />

o(12)<br />

g2 f e<br />

4<br />

6 e7 e8 f 4<br />

e 6<br />

e 7<br />

88


(char F =0 for avoiding separability ques=ons)<br />

89


Aim:<br />

Any Jordan algebra is power-­‐associa=ve<br />

Define powers: x 1 := x, x i+1 := x i • x<br />

(there is no ambiguity, x i (•)=x i if J is special)<br />

x i • x j = x i+j for all x€ J<br />

Hint: To prove this, first show that R xi belongs to the (associa=ve and<br />

commuta=ve) algebra spanned by R x and R x2, by using an induc=ve argument<br />

since<br />

Now apply R xi R xj = R xjR xi to x, and finish by induc=on on i.<br />

90


Idempotents and Peirce decomposi=on<br />

DefiniCon: e € J is said to be an idempotent if e 2 =e.<br />

Denote: J i ={x € J | xe=ix }<br />

J = J 0 ⊕ J 1/2 ⊕ J 1 is the Peirce decomposiCon associated to e.<br />

Proof: Check that 2R e 3 -­‐3Re 2 +Re =0 to conclude that R e is diagonalizable<br />

with eigenvalues € {0, 1/2, 1}.<br />

ObservaCon: trace R e ≠0 = dim J 1 +1/2 dim J 1/2<br />

Behaviour<br />

Remark: There is a generalized Peirce decomposi=on when 1=e 1 +…+e t<br />

is a decomposi=on into pairwise orthogonal idempotents.<br />

91


The nilpotent Radical<br />

DefiniCons:<br />

A power-­‐associa=ve algebra A is nil if every element is nilpotent (for each x€A,<br />

there is n€ N such that x n =0). A is nilpotent if there is n€ N such that any<br />

product x 1 …x n of n elements in A is 0, no maker how we associate these<br />

elements. A is solvable if A (n) =0 for some n€ N. (where A (1) =A, A (r+1) =(A (r) ) 2 ).<br />

nilpotent solvable nil<br />

If J is finite-­‐dimensional, nil implies nilpotent:<br />

z Solvable implies J* nilpotent (equivalent to J nilpotent by an easy<br />

nonassocia=ve result)<br />

It can be proved by induc=on on dim J (clue: J=Fw+C J 2 with C solvable)<br />

z Nil implies solvable by induc=on on dim J. Take 0≠B≤ J maximal soluble<br />

and apply a technical lemma: if xB B,x 2 B B then x n B B, in par=cular it<br />

happens if x€u 2 B with uB B.<br />

Hence we can define Rad J= the maximal nilpotent ideal of J<br />

92


A counterexample<br />

Exercise: find a solvable not nilpotent Jordan algebra<br />

(obviously infinite-­‐dimensional)<br />

Hint: (Shestakov’s example) (char F≠2)<br />

Take V a vector space with a countable basis.<br />

Consider the exterior algebra Λ(V), and let Λ 0 (V) be the subalgebra generated<br />

by V.<br />

Then J= Λ 0 (V) ⊕V with product (u+x)(v+y)=uy+vx € Λ 0 (V) is the algebra we are<br />

looking for.<br />

93


Semisimple =sum of simple ideals<br />

DefiniCon: J is semisimple if Rad J=0. (Always J/Rad J is semisimple)<br />

DefiniCon: A trace-­‐form ( , ):A×Aà F is a symmetric bilinear form such that<br />

(xy,z)=(x,yz).<br />

ObservaCon: If B A then B⊥ ={y∈A| (B,y)=0} A.<br />

Example: (x,y)= trace R xy is a trace form in a Jordan algebra J<br />

Theorem: If J is finite-­‐dimensional (char F=0 for the proof), Rad J = J ⊥<br />

Proof: for ⊆, see that J ⊥ is nilideal, checking that R b is nilpotent when b∈J is<br />

nilpotent. (It also uses that no nil implies with idempotents).<br />

Corollary: If J is fin-­‐dim semisimple (char F=0), then it is sum of simple<br />

ideals.<br />

Proof: use Diedonné’s Theorem, since Rad J=0=J ⊥ implies nondegeneracy of<br />

the trace form.<br />

94


Semisimple implies unital<br />

DefiniCon: e∈ J an idempotent is called principal if J 0 is nil (it does not have<br />

idempotents).<br />

Remark: Any finite-­‐dimensional Jordan algebra not nil contains a principal<br />

idempotent.<br />

Proof: Take e idempotent. If it is not principal, take u idempotent in J 0 (e) and<br />

check that e’=e+u is an idempotent with J 1 (e) ⊆J 1 (e’) (but ≠).<br />

Theorem: J semisimple (char F=0) ⇒ 1∈J<br />

Proof: Take e a principal idempotent and see J 1 (e)=J. For if x∈J 0 ⊕J 1/2 then<br />

(x,y)=tr R xy =(computa=ons with the trace)=0 for all y so that x=0 (by<br />

nondegeneracy).<br />

95


Degree<br />

DefiniCon: an idempotent e ∈ J is primiCve if e cannot be wriken as the<br />

sum of two orthogonal idempotents (iff e is the unique idempotent in J 1 ).<br />

DefiniCon: e is absolutely primiCve if it is primi=ve in J K =J⊗ F K for any scalar<br />

extension K|F.<br />

DefiniCon: a fininite-­‐dimensional central simple Jordan algebra J is reduced<br />

if 1=e 1 +…+e t for pairwise orthogonal absolutely primi=ve idempotents e i ∈ J.<br />

In this case J ii =Fe i and the number t is unique and is called the degree of J<br />

(or the capacity of J).<br />

96


CLASSIFICATION (III)<br />

Simple-­‐central finite dimensional reduced Jordan<br />

algebras<br />

Consider an arbitrary field F with char F≠2, t=degree<br />

∃ A central simple ass. alg. s.t J= A +<br />

∃ A simple ass. alg. with<br />

involu=on * s.t J= H(A, *)<br />

(≤ A + )<br />

∃ ( , ): V × Và F nondegenerate bilinear form (dim V >1) s. t. J= F1+V<br />

x•y=(x,y)1<br />

J= H(M 3 (C), *); (a ij )*=(ā ji )<br />

A I<br />

Second kind A II<br />

First kind<br />

E t=3<br />

J K ≅ Mt (K) +<br />

K=alg.closure of F<br />

JK =H(Mt (K), * ); x*=xt JK =H(M2t (K), * ); x*=c -­‐1xtc c skew<br />

A<br />

97<br />

B<br />

C<br />

D t=2


RECENT RESEARCH<br />

DERIVATIONS.<br />

ALGEBRAS OF QUOTIENTS OF LIE ALGEBRAS<br />

98


THE ASSOCIATIVE CASE<br />

Y. Utumi, On quo=ents rings, Osaka J. Math. 8 (1956), 1-­‐18.<br />

Equivalently:


Examples:<br />

1.-­‐<br />

2.-­‐<br />

3.-­‐<br />

4.-­‐<br />

THE ASSOCIATIVE CASE


THE NON-ASSOCIATIVE CASE


C. MARTÍNEZ, The ring of fractions of a Jordan algebra, J. Algebra<br />

237 (2001), 798-­‐‑812.<br />

Associative seKing<br />

algebra of quotients<br />

UTUMI’s ring of quotients


MSM, Algebras of quotients of Lie algebras,<br />

J. Pure Appl. Algebra 188 (2004), 175-­‐‑188.<br />

L ⊆ Q Lie algebras. Q is an algebra of quotients of L<br />

if the following equivalent conditions are satisfied:<br />

∀ p, q ∈ Q, p ≠ 0, there exists x ∈ L such that [x, p] ≠ 0<br />

and [x, q], [x, ad y 1 , … ad y n q], ∈ L, with y 1 , …, y n ∈ L.<br />

∀ 0 ≠ q ∈ Q there exists an ideal I of L with zero annihilator<br />

such that 0 ≠ [I, q] ⊆ L.<br />

Qm (L) : = lim Der (I, L)<br />

I ∈ Ie (L)


Nonassociative quotients<br />

E. GARCÍA, M. GÓMEZ LOZANO, Centers, centroid, extended centroid<br />

and quotiens of Jordan systems, Comm. Algebra 34 (2006), 4311-­‐‑4326.<br />

J. A. ANQUELA, T. CORTÉS, E. GARCÍA, M. GÓMEZ LOZANO, Jordan<br />

centers and Martindale-­‐‑like covers, J. Algebra 305 (2006), 615-­‐‑628.<br />

F. PERERA, MSM, Associative and Lie algebras of quotients,<br />

Pub. Mat. 52 (2008), 129-­‐‑149.<br />

F. MONTANER, Algebras of quotients of Jordan algebras<br />

J. Algebra. 52 (2008), 129-­‐‑149.<br />

M. CABRERA, J. SÁNCHEZ ORTEGA, Lie quotients for skew Lie<br />

algebras, Algebra colloq. 16 (2009), 267-­‐‑274.<br />

M. BRESAR, F. PERERA, J. SÁNCHEZ ORTEGA, MSM, Computing the<br />

maximal algebra of quotients of a Lie algebra, Forum Math. 21(2009), 601-­‐‑620.


L ⊆ Q Lie algebras. Q is an algebra of quotients of L<br />

if the following equivalent conditions are satisfied:<br />

∀ p, q ∈ Q, p ≠ 0, there exists x ∈ L such that [x, p] ≠ 0<br />

and [x, q], [x, ad y 1 , … ad y n q], ∈ L, with y 1 , …, y n ∈ L.<br />

∀ 0 ≠ q ∈ Q there exists an ideal I of L with zero annihilator<br />

such that 0 ≠ [I, q] ⊆ L.<br />

Qm (L) : = lim Der (I, L)<br />

I ∈ Ie (L)


L = Lσ ,<br />

σ ∈ G<br />

⊕ Q = Qσ σ ∈ G<br />

⊕<br />

gr-­‐‑ gr-­‐‑<br />

L ⊆ Q Lie algebras. Q is an algebra of quotients of L<br />

if the following equivalent conditions are satisfied:<br />

h(Q)<br />

h(L)<br />

∀ p, q ∈ Q, p ≠ 0, there exists x ∈ L such that [x, p] ≠ 0<br />

and [x, q], h(Q)<br />

[x, ad y1 , … ad yn q], ∈ L, with y1 , …, yn ∈ L.<br />

gr-­‐‑<br />

∀ 0 ≠ q ∈ Q there exists an ideal I of L with zero annihilator<br />

such that 0 ≠ [I, q] ⊆ L.<br />

Q m (L) : = lim<br />

gr-­‐‑<br />

J. SÁNCHEZ ORTEGA, MSM, Algebras of<br />

quotients of graded Lie algebras, J. Algebra<br />

323 (2010), 2002-­‐‑2015.<br />

I ∈ Ie (L)<br />

gr-­‐‑<br />

Der (I, L)<br />

gr-­‐‑


Abstract properties<br />

F. PERERA, MSM, Associative and Lie algebras of quotients,<br />

Pub. Mat. 52 (2008), 129-­‐‑149.<br />

M. CABRERA, J. SÁNCHEZ ORTEGA, Lie quotients for skew Lie<br />

algebras, Algebra colloq. 16 (2009), 267-­‐‑274.<br />

Computation of maximal algebras of quotients<br />

M. BRESAR, F. PERERA, J. SÁNCHEZ ORTEGA, MSM, Computing the<br />

maximal algebra of quotients of a Lie algebra Forum Math. 21 (2009), 601-­‐‑620.<br />

Is Q m (L) = Q m (I) for every essential ideal I of L?<br />

Computation of Q m (A -­‐‑ /Z A )<br />

Is Q m (L) = Q m (Q m (L))?


NOT in general<br />

Computation of maximal algebras of quotients<br />

M. BRESAR, F. PERERA, J. SÁNCHEZ ORTEGA, MSM, Computing the<br />

maximal algebra of quotients of a Lie algebra Forum Math. 21 (2009), 601-­‐‑620.<br />

Is Q m (L) = Q m (I) for every essential ideal I of L?<br />

Computation of Q m (A -­‐‑ /Z A )<br />

Is Q m (L) = Q m (Q m (L))?<br />

YES when L is ssp


Der m (A) : = lim<br />

ϕ : Der m (A) Q m (A -­‐‑ /Z A )<br />

δ I<br />

I ∈ I e (A)<br />

⟝<br />

Der (I, A)<br />

Computation of Q m (A -­‐‑ /Z A )<br />

δ I<br />

Homomorphism of Lie algebras<br />

Lie ideals, Lie derivations of A<br />

Ker (ϕ) = {δ I∣δ(I) ⊆ Z(A)}<br />

-­‐‑ /ZA Associative ideals, associative derivationsof A


H. BIERWIRTH, MSM, Lie ideals of graded associative algebras,<br />

Israel J. Math 191, 111-­‐‑136 (2012).<br />

Theorem<br />

A gr-­‐‑prime Lie algebra<br />

ϕ : Der gr-­‐‑m (A) Q gr-­‐‑m (A -­‐‑ /Z A )<br />

δ I<br />

⟝<br />

δ I<br />

Isomorphism of graded Lie algebras if A does not satisfy<br />

[[x 2 , y], [x, y]] = 0 for x, y ∈ h(A).


F. PERERA, MSM, Strongly nondegenerate Lie algebras,<br />

Proceedings of the Amer Math. Soc 136, 4115-­‐‑4124 (2008).<br />

Theorem<br />

Application<br />

A gr-­‐‑semiprime Lie algebra 2 and 3 -­‐‑ torsion free<br />

Der gr (A) / I is a gr-­‐‑strongly nondegenerate Lie algebra<br />

I = {δ ∈ Dergr (A) ∣δ(A) ⊆ Z(A)} = AnnDer (A) (Inn(A))<br />

gr


Theorem<br />

Application<br />

A gr-­‐‑semiprime Lie algebra 2 and 3 -­‐‑ torsion free<br />

Der gr (A) / I is a gr-­‐‑strongly nondegenerate Lie algebra<br />

I = {δ ∈ Der gr (A) ∣δ(A) ⊆ Z(A)} = Ann Der (A) (Inn(A))<br />

gr<br />

Corollary<br />

A gr-­‐‑prime Lie algebra , non commutative,<br />

2 and 3 -­‐‑ torsion free<br />

Der gr (A) is a gr-­‐‑strongly nondegenerate Lie algebra


Theorem<br />

A gr-­‐‑semiprime Lie algebra 2 and 3 -­‐‑ torsion free<br />

Der gr (A) / I is a gr-­‐‑strongly nondegenerate Lie algebra<br />

I = {δ ∈ Dergr (A) ∣δ(I) Z(A)} = AnnDergr(A) (Inn(A))<br />

L Lie algebra<br />

for x homogeneous<br />

Corollary<br />

Application<br />

L is gr-­‐‑strongly nondegenerate if [x, [x, L]] = 0<br />

implies x = 0.<br />

A gr-­‐‑prime Lie algebra , non commutative,<br />

2 and 3-­‐‑ torsion free<br />

Der gr (A) is a gr-­‐‑strongly nondegenerate Lie algebra


Theorem<br />

A gr-­‐‑semiprime Lie algebra 2 and 3 -­‐‑ torsion free<br />

Der gr (A) / I is a gr-­‐‑strongly nondegenerate Lie algebra<br />

I = {δ ∈ Dergr (A) ∣δ(I) Z(A)} = AnnDergr(A) (Inn(A))<br />

A graded semiprime associative algebra<br />

Corollary A -­‐‑ /ZA is graded strongly nondegenerate<br />

Applications<br />

A gr-­‐‑prime Lie algebra , non commutative,<br />

2 and 3-­‐‑ torsion free<br />

Der gr (A) is a gr-­‐‑strongly nondegenerate Lie algebra


L graded semiprime Lie algebra<br />

Q gr-­‐‑m (L)<br />

≈ Q m (L)<br />

?


L graded semiprime Lie algebra ?<br />

is a monomorphism<br />

ϕ : Q gr-­‐‑m(L) Q m(L)<br />

δ I<br />

⟝<br />

δ I<br />

L = A -­‐‑ /Z A


A graded semiprime associative algebra<br />

ϕ : Q gr-­‐‑m (A -­‐‑ /Z A ) Q m (A -­‐‑ /Z A )<br />

δ I<br />

⟝<br />

Is it an isomorphism?<br />

YES when…<br />

δ I<br />

A prime<br />

Grading given by an ordered group<br />

Grading finite<br />

is a monomorphism<br />

J. SÁNCHEZ ORTEGA, MSM, Finite gradings of Lie algebras,<br />

J. Algebra 372, 161-­‐‑171 (2012).


Farnsteiner (1988)<br />

Proposition<br />

⊕ (nnassociative)<br />

R = R σ<br />

σ ∈ G<br />

finite grading<br />

Corollary<br />

σ ∈ G<br />

⊕<br />

R = R σ<br />

finite grading<br />

Der (I, R) = Der gr (I, R)<br />

I graded ideal<br />

Der (R) = Der gr (R)


Theorem<br />

L = Lσ σ ∈ G<br />

⊕<br />

G abelian ordered group<br />

is a strongly nondegenerate lie algebra<br />

Finite grading (τ the maximum)<br />

L τ ∩ I ≠ 0 for every nonzero ideal I of L<br />

Q m (L) = Q gr-­‐‑m (L) and l (Supp(L))= l (Supp(Q m (L)))


Theorem<br />

A = Aσ σ ∈ G<br />

⊕<br />

G abelian ordered group<br />

Finite grading<br />

is an associative prime algebra<br />

Q m (A -­‐‑ /Z A ) = Q gr-­‐‑m (A -­‐‑/Z A ) = Der m(A)<br />

and l (Supp(L)) = l (Supp(Q m (L)))


USE OF ALGEBRAS OF<br />

QUOTIENTS<br />

121


H. BIERWIRTH, MSM, Structure of strongly nondegenerate<br />

Prime Lie algebras (preprint).<br />

Theorem<br />

122


JORDAN AND LIE QUOTIENTS<br />

123


JORDAN AND LIE QUOTIENTS<br />

Let V= (V + , V -­‐ ) be a Jordan pair<br />

L:= TKK (V) := L -­‐1 ⊕ L 0 ⊕ L 1<br />

where L -­‐1 = V-­‐, L 1 = V+, L 0 = Instr(V):= (inner structure<br />

algebra, which is a Lie algebra) is the TKK Lie algebra of V.<br />

Defini=on:<br />

V ≤ W. W is said to be a Jordan pair of quoCents of V if for every<br />

nonzero q € W σ there exists an ideal I of V such that<br />

being nonzero some of the two sets.<br />

124


Theorem<br />

The following are equivalent condi=ons:<br />

(i) W is a Jordan pair of quo=ents of V.<br />

(ii) TKK(W) is a Lie algebra of quo=ents of TKK(V).<br />

Theorem<br />

(i) Let V be a strongly-­‐nondegenerate Jordan pair. Then:<br />

(ii) If L= L -­‐1 ⊕ L 0 ⊕L 1 is a strongly nondegenerate Lie algebra such that<br />

Q m (L) is Jordan 3-­‐graded then:<br />

where V= (L -­‐1 , L 1 ) is the Jordan pair associated to L<br />

125


Analogue result for Jordan triple systems<br />

Theorem<br />

(i) Let V be a strongly-­‐nondegenerate Jordan pair. Then:<br />

(ii) If L= L -­‐1 ⊕ L 0 ⊕L 1 is a strongly nondegenerate Lie algebra such that<br />

Q m (L) is Jordan 3-­‐graded then:<br />

where V= (L -­‐1 , L 1 ) is the Jordan pair associated to L<br />

126


For Jordan algebras<br />

Theorem<br />

Let J be a strongly nondegenerate Jordan algebra over a ring with 1/6.<br />

Then:<br />

Where J T is the JTS associated to J and<br />

V(J T )= (J T , J T ) is the Jordan pair associated to J T .<br />

127


Kantor : “There are no Jordan algebras, there are only Lie algebras.”<br />

K. McCrimmon: “Of course, this can be turned around: nine Cmes<br />

out of ten, when you open up a Lie algebra you find a Jordan<br />

algebra inside which makes it thick.”<br />

There is a lot of work to do to proof that both worlds are the same.


GRACIAS

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!