Sei sulla pagina 1di 32

REND. SEM. MAT.

UNIVERS. POLITECN. TORINO


Vol.40°, 1 (1982)

Krzysztof Maurin

PLATO'S CAVE PARABLE AND THE DEVELOPEMENT


OF MODERN MATHEMATICS (*)

Motto: A truly meaningful philosophy of mathematics is unthinkable without relating


crucial mathematical theorems to the most profound concepts of philosophy.

Some four years ago (cf. [1]) I observed that a certain number of most
significant theorems and constructions of modern mathematics have under-
gone the following evolution and that one might even talk of a principle:
"Viewed historically, at first one knew certain" natural objects" (e.g.
spaces); then certain, "abstract objects" (or spaces) were discovered (or one
was forced to introduce them). Finally, with considerable effort and brilliance
of mind, it was proved that these objects were "simply" sub-spaces of the
well-known spaces, and that in some (favourable) cases they were indeed
isomorphic with "classical" objects. The more natural the spaces were, the
more difficult it was to prove the corresponding embedding theorems".
A little later, thanks to my contacts with some German Platonists (Georg
Picht and Carl Friedrich von Weizsacker), I realized that the evolution princi-
ple of modern mathematics that I had observed was an exact illustration of the
famous parable of the cave from the 7th Book of Plato's "Polites". The fol-

Classificazione per soggetto: 00A2 5

(*) An extended version of a lecture held on 15.6.1981 at Seminario Matematico della


Universita e del Politecnico di TORINO.
2

lowing correspondences are found:


a) Shadows on the cave's walls are classical mathematical objects: e.g. planes
in the Euclidean space, algebraic projective varieties, etc.
b) Ideas are "abstract objects", e.g. Riemann's spaces, Hodge manifolds.
c) The dramatic "descent" into the cave is the corresponding embedding
theorem.
And who is the prisoner — at first fettered, then released and dragged (by
force) into the sunlight, and finally descending again into the cave? It is mathe-
matics itself as a whole, for it is different researchers belonging to different
generations of mathematicians who have accomplished the ascent, the creation
or awakening of a great mathematical-idea, e.g. the Riemann surface, and an
embedding or uniformisation theorem which is, often several decades later,
proved by quite different mathematicians.
As I learned from Georg Picht, the descent is the most difficult and
dangerous way according to Plato (the person in question is killed in all
probability). And indeed the embedding or uniformisation theorems require
quite new means and conceptual formations: it was necessary to create
powerful and far-reaching mathematical disciplines (e.g. topology, functional
analysis, group theory...) in order to make the "descent" possible.
I have chosen four examples to illustrate this dramatic process. The
first of these consists of theorems around the idea of Riemann surface culmi-
nating in the uniformisation principle (Chapter I). Closely connected with
that idea, the theory of function fields is treated in Chapter II. Chapter III
tells about famous "imbedding theorems" of modern mathematics. The last
Chapter considers Gelfand's spectral theory and the classic spectral theorem.
The well known ton mot that "European philosophy is only a footnote
to Plato" is perhaps true, but I should venture the much truer one: modern
mathematics is only a footnote to Riemann.
I have tried to assume as little as possible in order that the present
article be accessible to physicists and philosophers as well as mathematicians.
It was a great opportunity for me to be the guest of the Istituto di Fisica
Mathematica of TORINO. May this little paper be a humble gift to all the
Italian friends who made my stay in Sabaudia (and Piemonte) an unforge-
table beautiful experience.
3

CHAPTER I

PLATO'S CAVE PARABLE AND THE IDEA


OF THE RIEMANN SURFACE

In dem Symbol des zweidimensionalen Nicht-


. -Euklidischen Kristals wird das Urbild der
Riemannschen Flache selbst..., erschaubar

H.WEYL [1913]

Plato often expressed his most profound and boldest intuitions as para-
bles, metaphers or even myths. Since the grand parable in the beginning of
the VII book of POLITEIA did not lose anything of its eternal verity, I will
quote verbatim this masterpiece of the great artist.

"Next, then," I said, "take the following parable of education and ignorance as a
picture of the condition of our nature. Imagine mankind as dwelling in an underground
cave with a long entrance open to the light across the whole width of the cave; in this they
have been from childhood, with necks and legs fettered, so they have to stay where they
are. They cannot move their head around because of the fetters, and they can only look
forward, but light comes to them from a fire burning behind them higher up at a distance.
Between the fire and the prisoners is a road above their level, and along it imagine a low
wall has been built, as puppet showmen have screens in front of their people over which
they work their puppets."
"I see," he said.
"See, then, bearers carrying along this wall all sorts of articles which they hold
projecting above the wall, statues of men and other living things, made of stone or wood
and all kinds of stuff, some of the bearers speaking and some silent, as you might expext."
"What a remarkable image," he said, "and what remarkable prisoners!"
"Just like ourselves," I said. "For, first of all, tell me this: What do you think such
people would have seen of themselves and each other except their shadows, which the
fire cast on the opposite wall of the cave?"
"I don't see how they could see anything else," said he, "if they were compelled to
keep their heads unmoving all their lives!"
"Very well, what of the things being carried along? Would not this be the same?"
"Of course it would."
4

"Suppose the prisoners were able to talk together, don't you think that when they
named the shadows which they saw passing they would believe they were naming things?"
"Necessarily."
"Then if their prison had an echo from the opposite wall, whenever one of the
passing bearers uttered a sound, would they not suppose that the passing shadow must be
making the sound? Don't you think so?"
"Indeed I d o , " he said.
"If so," said I, "such persons would certainly believe that there were no realities
except those shadows of handmade things."
"So it must be," said he.
"Now consider," said I, "what their release would be like, and their cure from these
fetters and their folly; let us imagine it might naturally be something like this. One
might be released, and compelled suddenly to stand up and turn his neck around, and
to walk and look towards the firelight; all this would hurt him, and he would be too much
dazzled to see distinctly those things whose shadows he had seen before. What do you
think he would say, if someone told him that what he saw before was foolery, but now he
saw more rightly, being a bit nearer reality and turned towards what was a little more real?
What if he were shown each of the passing things, and compelled by questions to answer
what each one was? Don't you think he would be puzzled, and believe what he saw before
was more true than what was shown to him now?"
"Far more," he said.
"Then suppose he were compelled to look towards the real light, it would hurt his
eyes, and he would escape by turning them away to the things which he was able to look
at, and these he would believe to be clearer than what was being shown to him."
"Just so," said he.
"Suppose, now," said I, "that someone should drag him thence by force, up the
rough ascent, the steep way up, and never stop until he could drag him out into the light of
the sun, would he not be distressed and furious at being dragged; and when he came into
the light, the brilliance would fill his eyes and he would not be able to see even one of the
things now called real?"
"That he would not," said he, "all of a sudden."
"He would have to get used to it, surely, I think, if he is to see the things above. First
he would most easily look at shadows, after that images of mankind and the rest in water,
lastly the things themselves. After this he would find it easier to survey by night the
the heavens themselves and all that is in them, gazing at the light of the stars and moon,
rather than by day the sun and the sun's light."
"Of course."
"Last of all, I suppose, the sun; he could look on the sun itself by itself in its own
place, and see what it is like, not reflections of it in water or as it appears in some alien
setting."
"Necessarily." said he.
"And only after all this he might reason about it, how this is he who provides
seasons and years, and is set over all there is in the visible region, and he is in a manner
the cause of all things which they saw."
"Yes, it is clear," said he, "that after all that, he would come to this last."
5

"Very good. Let him be reminded of his first habitation, and what was wisdom in
that place, and of his fellow-prisoners there; don't you think he would bless himself for
the change, and pity them?"
"Yes, indeed."
"And if there were honours and praises among them and prizes for the one who saw
the passing things most sharply and remembered best which of them used to come before
and which after and which together, and from these was best able to prophesy accordingly
what was going to come do you believe he would set his desire on that, and envy those who
were honoured men or potentates among them? Would he not feel as Homer says, and
heartily desire rather to be serf of some landless man- on earth and to endure anything in
the world, rather than to opine as they did and to live in that way?"
"Yes indeed," said he, "he would rather accept anything than live like that."
"Then again," I said, "just consider; if such a one should go down again and sit on
his old seat, would he not get his eyes full of darkness coming in suddenly out of the sun?"
"Very much so," said he.
"And if he should have to compete with those who had been always prisoners, by
laying down the law about those shadows while he was blinking before his eyes were
settled down — and it would take a good long time to get used to things — wouldn't they
all laugh at him and say he had spoiled his eyesight by going up there, and it was not
worth-while so much as to try to go up? And would they not kill anyone who tried to
release them and take them up, if they could somehow lay hands on him and kill him?"
"That they would!" said he.

We have to assume some notions, but in order "to fix the notations" let
us recall, that by C is denoted the (Gauss) plane of complex numbers;
C : = C U | o o j is the Riemann sphere: a one point compactification of C.
The sphere looks locally like the plane: it is a 2-dimensional manifold.
Holomorphicity is a local property: A continuous function f.X—>C
is holomorphic at z G X if it is complex different lable at z 0 (there exists
df/dz(z0) = lim ) A(X) - denotes the set of holomorphic
h->o h
functions on X.

1.1. Examples. 1. P(z) : = a0z + ... -V a^zk — a polynomial of degree k


(ak =£ 0) is holomorphic on the whole C.
2. The exponential function z—>ez—:exp(z) is holomorphic on C,
thus expEA(C).
3. Rational functions are quotients of polynomials
P(z) a0+ ... + cikzk
aj, brGC
Q(z) bx + ... + bxzl
6

They are defined on X = C - {set of zeros of Q{, zeros of Q are poles of


P/Q.
4. Meromorphic functions are holomorphic on X with the exception
of a discreet set of poles. Hence a meromorphic function / can be defined as
holomorphic map f:X—+C.
5. Algebraic function is defined locally as a solution of the algebraic
equation
W(/,z) =fn + rrf"-1 + ... + rn = 0 , where rfc
are stational functions of z.

Example / 2 - z = 0, " / = ± > / z " / 2 - ( z -tf)(z-£).( 2 -<7) = 0, where


#, &, c are distinct complex numbers. It was Riemann who considered
algebraic functions globally. Trying to understand these functions he made
the revolutionary discovery — or construction — of a many sheeted surface
extended of over his sphere C U jooj# \y e will now try to ascend to the IDEA
OF RIEMANN SURFACE on the lines of the monumental work of the young
Hermann WEYL "Idee der Riemannschen Flache" (1913-10553). In order to
facilitate formulations let us recall that a differentiable surface Y is a space
which locally looks like the plane R2 (= C): mappings YDUBy—•
—>{f(y) EOC C are called maps or coordinate systems. With the help of local
maps we can introduce (local) notions on the surface Y. For instance a
function / defined on a neighourhood of y G o is holomorphic if the
function foy~l:o—>C is holomorphic. A complex surface is a 2-dimen-
sional manifold on which grow holomorphic functions. In similar way we can
define the holomorphic mapping F: X —> Y of the complex manifold X
into the complex manifold Y. It is very important to underline that manifolds
are abstract notions and should not be imagined as placed or imbedded in
the 3-dimensional Euclidean space R3. Nothing prevents one from consider-
ing w-dimensional real (complex) manifolds: topological spaces which
locally are like Rn (Cn). We recall that a topological space (X, B) is such a
point set, X, that each of its elements (points) x has a family B(x) of
neighbourhoods, x EX. These families B(x), x GX — so called basis of the
topology of X — have natural properties ("HAUSDORFF axiomes"):
i) If x G UEiB(y), then there exists such V G B(x), that V C U
ii) If UltU2€B(x) then there exists UEB(x) such that U^UlC\U2.
The open subset o is a sum of elements of the Basis B. The family of all
open subsets is the topology of (X,B).
7

It is nonsense to speak about continuity of a map f-.X —>Y of sets


without topology. But if (X,BX) and (Y,B2) are topological spaces, then
f:X—*Y is continuous (by definition) if f'1(02) is open in X for each
open 02 set in Y-, f is homeomorphic if it is one to one, continuous and
the inverse map f'1: Y-^X is continuous to. A topological space X is
connected if it consists of precisely one piece: if any two points xlfx2€X
can be connected by a curve c [o, 1] —>X, c(o) = xlf c(l) = x2, where c
is a continuous map of the interval 0 < £ < 1 , into X. X is compact if
from any covering by open set UO,- = X , one can chose a finite covering:
oilu...uoik=.x.
1.2. Examples. 1. Any bounded closed subset of C.
2. S2 = C. ^
3. A torus T in C are compact space.
All spaces 1-3 are connected.
1.3. Definition. A Riemann surface is a complex connected surface (2-di-
mensional complex manifold).
Compact spaces have many nice properties:
1. Each continuous function f.X—>C(R) on a compact space is
bounded.
2. Let X be a compact Riemann surface and let Y be any Riemann
surface. Then every non constant holomorphic mapping f:X—>Y is a
surjection: f(X) = Y. All holomorphic functions on X are constant:
A(X) = C.
3. Each meromorphic function h : C —> C is rational (since it has only
a finite number of poles).

Function germs. Sheaves. Coverings. Analytic continuation.

All these general (and "abstract") notions were implicitely contained in


the fundamental Riemann doctoral thesis but explicitely they were formulated
70 years later (by Weyl, Hausdorff, Frechet...): the time was not ripe then!
We will now outline the construction of the concrete Riemann surface.
Let X be a connected open subset of C (or any Riemann surface). Let
us consider all local holomorphic functions defined on open subsets of X.
Two holomorphic (or meromorphic) functions fk-Uk—• C, & = 1,2,...
8

define the same germ at x EUiH U2 if there exists a neighbourhood


0 C C/j fl U2 of a; such that f\(z) = /2(z) for all yGO. This is plainly an
equivalence relation, which we denote by #. A germ of / at #-denoted by
fx is the class of all meromorphic functions g, g~f. We write px(f) • = / .
Let us denote by Ax (Mx) the space a of all holomorphic (meromor-
phic) germs at all points x E X The triple (Ax,p, X), where the projection
p :AX—+X is defined by p(fx): = x is called the sheaf of holomorphic
germs over X.
1.4. Proposition (Weyl). Ax (or Mx) is a topological manifold and
p : Ax —• X is a local homeomorphism. Whence A x has the structure of a
complex manifold of two (real) dimensions.
1.5. Corollary. Each connected component of Ax (or Mx) is a Riemann
surface.
Proof (construction of the basis of a topology) is quite natural: Let
0 G B(x) be a neighbourhood of x let [0,/] = {px(f) C y4x : # G 0}. One
must check that the family {[0,/] :0} open sets in X and / A(0) is a
basis of the topology of Ax. Plainly p: [0,/] —> 0 is a one to one map.
Hence p: Ax —• x is a local homeomorphism.
A holomorphic (or meromorphic) function is an organism: a perturba-
tion on any neighbourhood disturbs the whole function.
1.6. Identity theorem (Weierstrass). Let X be a connected open subset
of X (or any Riemann surface). If two meromorphic functions f~g> for
any x then f = g.
Remark. It is interesting that continuous (and even differentiable) func-
tions do not satisfy the Identity-theorem.
Example / ^N?
__

This important property of holomorphic functions led Weierstrass and


Riemann to the notion of
1.7. Analytic continuations.
Let X be a domain in C, and let /: [o, 1] —> X be a path connecting a
and b :/(o) =a, 1(1) = b. Let $£Aa and \jj G ^ be holomorphic germs
at a and b. Let a = t0<tt< ...<tn-x<tn = b and /([**-i,**]) C Ukt
9

* = 1,•...,», fk€A(Uk) and fk^fk+u where 2;- = /(*/)€£/; O [/;+1, for


/ = 0,1,...,» —1. When this situation occurs one says that the germ \p is
obtained from the germ ^ by an analytic continuation along the path /. We
can reformulate this notion in the following way: \/J is obtained of by an
analytic continuation of the germ «p if both germs <p, \jj are in the same
component of Ax\ there exists a path / joining «/? with \jj such that
p(l) = l
The preceding notion of analytic continuation leads to the notion of the
maximal continuation of a meromorphic germ <p: we take Y — the connected
component ^Ax containing <p and we obtain the triple (Y,p,F), where Y
is a Riemann surface, p :Y —>X, F: Y —> C is a holomorphic function such
that if y — fx is a germ represented by / , where x = p(y), then F(y) : = /(#).
We have thus the diagram
F: Y • C

X
One sees easily, that the maximal analytic continuation of a germ <p of
a local function in X is uniquely defined. Riemann recognised that the sur-
face Y >X is the natural "Lebensraum" (living space) of holomorphic
and meromorphic functions: subdomains of C as domains of definition of a
function are accidental. The true domains of complex analysis are surfaces Y
which cover (a part of) C.
1.8. Examples of Riemann surfaces are all open subsets of the sphere C =
= CU{ooj (C = R2):
1. The complex plane C (with the trivial atlas consisting of the single
map: the identity id c ).
2. The Riemann sphere C, which is isomorphic to P 1 (the 1-dimen-
sional complex projective space P 1 = P(C2) = } [zt, z2] • \zx, z2) € C2 -
-(0,0)) — the set of equivalence classes [2i,2 2 ] of pairs (2j,2 2 ) with
respect to the relation: ( 2 j , 2 2 ) ~ (2i,2 2 ) if 2 1 =tf2 , 1 , z2=az2, with
complex a). The atlas consists of (is generated by) the two maps:
b1--P1-\ll>0])-+C , h1([z1,z2]) = z1/z2
* V ^ - i [ 0 , l ] } - + C , b2([zl,z2]) = z2/zl
P 1 is biholomorphic to C since [2^22]—•2 1 /2 2 is biholomorphic.
10
/7<2 -)- h
Automorphisms (i.e. biholomorphic maps) of C are z —> XT'
where a,b,c,d€:R and ad — cb^0.
3. D = {zeC:\z\<l] - the unit disc in C.
4. H = \z 6 C : Imz > 0) — the upper half plane. Automorphisms of H
are the maps z —> — , where a,|3,7,6 E C and det ( ? J : = = a 5 -
— 7 0 = 1. The "surfaces" D and H are biholomorphic H3z —» £D
is a biholomorphic map.
Definition. A concrete Riemann surface is a Riemann surface provided
with a meromorphic function p :X —> C: i.e. cover of a sphere C = C U j°°].
It is easy to prove, that all the sets p~1(z1), zEC, have the same car-
dinality, which is called the number of sheets of the surface Y. If Y is
compact then this number is finite.
One can prove that the surface of any algebraic function / defined by
an irreducible polynomial W(f,z) of order n in / is precisely ra-sheeted:
this is the resolution of the puzzle of "many valued functions". But why is
the covering p : Y—+X many sheeted?
Answer: there is no reason why analytic continuations of a germ #
along different paths lt, l2 with the same end points a, b should produce the
same germ. Indeed we have the famous
1.9. Monodromy theorem (Weierstrass) gives a sufficient condition. If there
is a continuous deformation of the path lx into l2, then both analytic con-
tinuations produce the same germ \jj at b.
This theorem led Henri Poincare, the greatest French mathematician and
physicist, to the notion of
Homotopy. Two curves (pr maps) lx and l2 are homotopic if there
exists a 1-parameter continuous family Ls, s G [1,2] such that Lx=lx and
L2 = i>2'
Poincare proved that closed paths (loops) with the same end point x in
a connected space X form a group 7r1(X)= 71'^X,^) — the fundamental
group of X.
1.10. Definition. The connected manifold X is simply connected if
IIJCX) consists of only the neutral element 1, i.e. if each loop can be con-
tracted to a point.

1.11. Examples. 1. The spaces C,D,H,C are simple connected.


11

2. The torus T is not simply connected: the group .TT 1 (T) has two
generators, namely n1 (T) = Z@ Z.
3. C * : = C - ( o } , D * : = { z G C : | z | < l } and anulus {zECirK
< | z | < 1} have the fundamental group n ^ X ) = Z (Z — additive group of
integers).
Remark. One can prove that these are the only Riemann surfaces (up
to biholomorphic equivalence) with Abelian fundamental groups.
One of the most important discoveries of Riemann was the

1.12. Theorem (Riemann). Let

(•) P ( / , z ) = / » + rj"-1 4- ... + rn <ZM(C) [f]

be an irreducible of polynomial with rational coefficients rlf...,rn on C,


then the Riemann surface Y of the algebraic function / defined by (*) is
compact and the covering p-.Y—>C has precisely n sheets. Conversely:
each compact w-sheeted covering p-.Y —>C of the sphere C is the Riemann
surface of an algebraic function / satisfying an irreducible equation (*•) of
order n. P is defined by / uniquely (modulo a constant factor!).
Remark. The surface Y is obtained as the component of an algebraic
germ of a solution of (*) completed by n algebraic branching points. We
shall return to algebraic functions in the next Chapter.

Universal covering surface. Transformation groups.

For each meromorphic function on a Riemann sphere X we have con-


structed a holomorphic covering p : Y —> X, where Y is another Riemann
surface. Natural questions arise:
a) What are all the coverings of a Riemann surface X ?
b) Is there the strongest covering X—>X of X?
c) What are the automorphisms of a covering: Y —>X}
Question b) was already answered by Herman Amandus Schwarz in 1870,
whose memoirs had a catalytic effect on the young Henri Poincare, leading
him to his magnificent
Theory of covering surfaces. If p : Y —>X is a holomorphic covering
then biholomorphisms y-Y—>Y such that (*) poj = y i.e. which map
1
every fiber p~ (x) into itself are of great interest: they are called automor-
12

phisms of the covering p : Y—+X. Since the inverse 7" 1 and superposition
7i°72 °f automorphisms obviously satisfy (*). The set of all such auto-
morphisms form a group denoted by Aut(Y/X) or Deck(K >X), since
automorphisms (*) are called "decktransformations" (from the German
Decktransformation). The most interesting covering is p:X—>X, where
X is simply connected (^(X) = 1), this covering X iscalled the universal
covering of X. It can be shown that the universal covering is the strongest
/x/
Pi P
one: it covers every covering Y-^X of X:X >Y >X. Clearly
Aut (X/X) D Aut (Y/X). Poincare proved the following fundamental
1.13. Theorem (Poincare). i) If X—>X is the universal covering of X
then the group Aut (X/X) is isomorphic to the fundamental group irx (X)
of X. ii) To each covering Y —> X there exists a subgroup r = TY of
Aut (X/X) which is isomorphic to Aut(F/X). iii) The universal covering
surface X has the following universal property: every other covering Y of
X is obtained as the orbit space of some (discreet) subgroup T C Aut (X/X) :
: Y = X/T. The last point needs an explanation:
Transformation group and its orbits. If E is a set and G C Aut (E) is
a subgroup of its automorphisms. Then for each e £ E the G - orbit of the
point e is the set Ge =[ge :gGG]. Transformation groups are the oldest
groups in mathematics: they were used in aniquity in the theory of Platonic
solids, in ornaments... But the notion of the group appeared in mathematics
very late: in the Nineteenth century in the work of the brilliant Evariste
Galois (cf. the next Chapter) Felix Klein (1849-1925) and his great Norvegian
friend SophusLie (1842-1899) began the systematic study of transformation
groups. The young Klein, in his famous "Erlanger Program" (1872), put
transformation groups as a corner stone of the geometry.
If G is a transformation group of a space E then the set of the G-
-orbits is denoted by E/G and called the quotient (space) of E by G. Let
us return to the Poincare theory. We have a
1.14. Corollary. Every Riemann surface X is biholomorphically isomor-
phic to the quotient X/TTI (X) of the universal covering surface X of X by
its fundamental group:
X s X/TTJ (X) .

In this way the knowledge of all simply connected Riemann surfaces is


of paramount importance. This program seems hopeless: the world of simply
connected Riemann surfaces is infinite. Therefore the solution of this funda-
13

mental problem is striking: there are only three simply connected Riemann
surfaces (already mentioned in section 1.8). This answer was intuitively
"known" to Riemann but the solution was given 70 years later in the famous
Uniformisation theory of Riemann surfaces developped independently by
Koebe and Poincare in 1907. The only point is now the
Uniformisation of simply connected surfaces. The fundamental result
(called wrongly Riemann Mapping Theorem) is the following.
1.15. Theorem (Koebe-Poincare). Let Y be a simply connected Riemann
surface. Then Y is isomorphic to one and only one of the following three
(caninic domains of the Riemann Sphere):
a) The Riemann Sphere C (''elliptic" case).
b) C ("parabolic" case).
c) H — the upper half plane (= D) — ("hyperbolic" case).
Therefore all Riemann surfaces can be separated into three classes: hyperbolic,
parabolic and elliptic according to their universal covering: H,C or C (= P 1 ).
By far the most interesting compact surfaces are hyperbolic Riemann
surfaces, since non hyperbolic surfaces are only C and complex tori T.
Remark. The important case of subdomains of C is much easier and
was solved by Koebe and Osgood. It is relatively simple. But the proof of
general uniformisation theorem is very deep and difficult, and was solved with
methods suggested by Riemann, namely potential theory. Perhaps it is quite
natural: holomorphic functions: solutions of the Laplace equation
d2u du2
Au = — 2- + — 2- = 0
bx by
since the real and imaginary part of every holomorphic functions are potential
functions. Here we see the deep influence of physics upon mathematics!
Both of these preceding theorems of Poincare force us to investigate.

Automorphisms of the upper half plane H.

The automorphism group Aut (H) of the upper half plane was known
already to Mobius: it has a quite simple form given by the
1.16. Theorem. The group Aut(//) is isomorphic to the group SL(2yR)
of 2 X 2 real matrices I , I with determinant 1: ad — cb = 1.
14

Remark. To be precise, we should take the group PSL(2 ,R) : =


= SL{2,R)/[±l\ where -1 = ("Q ^) .
But the group G = SL(2,R) acts on H transitively but not freely: it
lhas fixed points.
The following simple fact is elucidating.
1.17. Theorem. H is the left quotient of SL(2,R) by the (compact)
subgroup K having fixed the point i = y/~J. K is isomorphic to the (ortho-
gonal) group 0(2) of rotations of the plane C
H = K\G .
Here we should take the occasion to say something about the geometry of H.
Non-euclidean geometry. As was discovered by Poincare, Beltrami and
Klein, the upper half plane H (or the unit disc D) is a natural model of the
non euclidean hyperbolic geometry discovered independently by Gauss,
Schweikart, Lobscevski and J. Bolyai. Poincare has shown that H is provided
with (natural) Riemann metric with constant negative curvature: this Poincare
dx 4~ dy
2
metric has the simple form ds = 2 , where z = x + iy. The geo-
desic (shortest curves) of this metric arc lines an Euclidean circles orthogonal
to the real axis, y = 0. Therefore Euclid's 5 th axiom is not satisfied:
there are infinitely many "lines" through a point not intersecting a given
line.
We can now better understand the motto of the present Chapter.- the
Riemann surface X is isomorphic to the quotient H/r, where T is a discrete
subgroup of the automorphism group of the non Euclidean plane H.
Let us now give precise definitions:
A transformation group G of the manifold X is a subgroup of Aut (X),
i.e. a set G of automorphisms g : X —> X of X such that
1. (gVg 2 )-*=gi(£2-*), for x<EX, gi,g2^G.
1'. If (gi'g2)'x =g2(gix) G i s t n e "left" transformation group of X).
2. The mapping GXX (g,x)—>g-xE.X is continuous,differentiable,
and holomorphic.
t h e set G-x0 = [g-x0:g GG} is the G-"orbit" of the point x0. If
G-x *= X for any x G X i.e. if each orbit meets every point of X, then G
"acts(transitively" on X. If there is no fixed point of G the group G acts
freelyj o n X-
15

1.18. Definition. The pair (X,G), where G is a transitive transformation!


group of X, is called a homogeneous space. The subgroup GXQ = jg E G :;g#o=!
= # 0 J is the isotropy group at (or of) xQ. In most interesting cases we have
the isomorphism.
1.19. Theorem. Let (X, G) be a homogeneous space then X is isomorphic
to G/GXQ >X = G/GXQ (X = GXQ\G in case of the left transformation group)
where the isomorphism is given by
G/GX0Bg-GX0-+g-x0eX , x0GX.
Therefore, very often homogeneous space is defined as the quotient G/H,
where H is a subgroup of an abstract topological (Lie) group G.
1.20. Examples. 1. The Euclidean space R3 together ^with the group
Tr - of translations, then (i? 3 , Tr) is a homogeneous space.
2. (R3,G) - where G is the group of Euclidean motions: translations
and rotations are plainly a homogeneous space. G0 = 0(3) is the orthogonal
group of (3-dimensional) rotations around the point 0.
3. (S 2 , 0(3))-the sphere S2 with a group of rotations is plainly a homo-
geneous space 0XQ(3) = 0(2) - plane rotations!

The Poincare Plane H as a Homogeneous space.

In the theory of Riemann surfaces, "the uniformisation theory", the


most important role plays the homogeneous space (H,SL(2,R))t where H is
the upper half plane and SL(2,R) is the group — of 2 X 2 — real matrices
( , J with determinant = 1: ad — cb = 1. It is well known that all biholo-
morphic maps of H arc of the form HBz—>(az + b)(cz + d) GH,
ad — cb = 1.
We have the interesting
1.21. Proposition. 1. The group SL(2}R) acts transitively on H. Thq
isotropy group at i is a maximal compact subgroup K — 50(2) is the special
orthogonal group (det& = 1) in two dimensions
H^SO{2)\SL(2yK) = K\G .
2. The group SL(2,R) can be considered as the group of isometrics o j
the Poincare half plane (H,dH).
16

This proposition and the uniformisations Theorem show the fundamental


role of the "symmetry group" SL(2,i?) of H the universal covering surface
of any hyperbolic Riemann surface:
"In the beginning was SYMMETRY" (Heisenberg). We have constructed
the sequence of manifolds and coverings:

SL(2,R)-?-* H(**SO(2)\SL(l,R)) -*-> X^H/T -?-+ C


which can be drawn as the following figure

G=SL(2,R)

n = K\-

X= H^K\G

X - Universal covering of X
{
X
1.22. Remark. C - the Riemann sphere ^P(C2)^S2.
Ascent. A local holomorphic (algebraic) function / at x€U C C
defines the holomorphic germ fx, which generates by the maximal
analytic continuation the Riemann surface X and the (global) algebraic
function F :X —> C. X has as its universal covering the Poincare halfplane
X —• X s H/Y where F = irl (X) a Deck (X -» X). H- in turn - can be con-
sidered as the quotient (homogeneous) space K\SL(2,R) bz the orthogonal
group K = 50(2) of the symmetry group of H: SL(2,R).
Descent. In the beginning was the symmetry: the group SL(2,R) —
which can be considered as the symmetry group of the non Euclidean plane
H. The first descent is the projection n = K\» of SL(2,R) into the homo-
17

geneous space H = K\G. The second descent is the covering (holomorphic)


projection p :H—> K\G/F given as the space of T — orbits, where T — a
discreet subgroup of SL(2,R) can be considered as the fundamental group of
the Riemann surface X. The third and last descent is the meromorphic func-
tion F : X —> C which gives the covering of the Riemann sphere C, and
which is the global algebraic function, the starting point of the whole dramatic
story.

CHAPTER II

ALGEBRAIC FUNCTION FIELDS AND PLATO'S PARABLE

Function fields on Riemann surfaces.

On each Riemann surface X there exist non constant meromorphic


functions; this profound result was obtained by Tibor Rado!
Meromorphic functions M(X) on a Riemann surface X form a field:
sum difference, product, quotient of meromorphic functions on X are still
meromorphic.
Each function on X can be considered as a function on X which is
T-invariant
f(yz)=f(z) , for 7GDeck(X/X) , z G X .
Meromorphic T-invariant functions on are called automorphic functions,
they form a (function) field M^(X).
2.1. Example. Let X = C, and T-a latine in C i.e. T is generated by
Ti > 72 ^ C and C/T is a torus then M^(C) are elliptic functions. Hence
elliptic functions are meromorphic and double periodic functions on C. One
of the fundamental facts of their vast and important theory is the
2.2. Theorem (Weierstrass). Afp(C) = C(p,p') i.e. any two T-elliptic
functions are algebraically dependent and the field Mp(C) is generated by p
and its derivative p'-, p is the famous Weierstrass elliptic functions de-
fined by
18

2.3. p(z) = p(z,y) :=z"2+ 2 [ ( 2 - 7 ) - 7 " 2 ] . Let us recall now


7Gr-}0}
some elementary notions from the theory of algebraic fields.

Extensions of fields; algebraic function fields. Galois group.

All fields considered here are commutative!

2.4. Examples. Q (rational numbers), R,C are fields. If a field k is a


subfield of / then / is an extension of k and is denoted by l/k (or l-.k).
Plainly / can be considered as a vector space over (the field) k\ the
dimension of this vector field is denoted by dim(//&) and is called dimension
(or degree) of the field extension l/k.

2.5. Example. / = C, k = R, then dim(C/R) = 2.


If a I satisfies a polynomial equation P(a) = 0, where P is a polyno-
mial with coefficients in k then a is algebraic over k\ otherwise a is
transcendent over k. \i A is a subset of iDk, then there exists the smallest
subfield of / which contains A,k , this field is denoted by k(A). Plainly
k(A) is the set of all rational functions generated by elements of k and A.
We have the

2.6. Proposition. If 2 is transcendent over k then k(z) = set of all ra-


tional functions with coefficients in k i.e. the set of quotients p(z)/q(z),
where p,q are polynomials with coefficients in k (i.e. p,q Ek[z]).
If / = k(z), then dim(//&) = 1, and / is a simple extension of / (2 is
called a primitive element).

2.7. Example. C = R(\/-l).


Galois group of l/k. All automorphisms 7 of the field extension l/k
which leave fixed elements of k (y(a) = a, for aE:k) form a group, the
Galois group Gal (l/k) — or Aut (l/k) — of the field extension l/k. This group
plays the analogue role for the extension l/k as the automorphism group
Deck(F—^->X) (denoted also by Aut(F/X) for the covering p : Y —+X.
Indeed, we have the very important connection between these two
groups. If IT : Y —> X is any (holomorphic) covering then any dectransforma-
tion 7 E D e c k ( F — > X ) defines an automorphism 7 of the field M(Y) as
follows (yf)(y) : = / ( 7 _ 1 ( y ) ) i.e. 7 / : = / ° 7 " \ (7i72>/ = / 0 ( 7 i 7 2 ) _ 1 =
= f° 72 ° T i = 7 i ( / ° 72 1 ) = 71(72/)- m t m s w a y w e obtain a represen-
tation of the group Deck(Y/X) in Aut (M(Y)) — the group of automor-
19

phisms of the field of meromorphic functions M(Y) on Y. On the other


hand the covering ir:Y—>X defines a monomorphism — i.e. a 1-valued
homomorphic map preserving all four field operations
ir*:M(X) > M(Y) , TT*/ = / 0 ^ -
Therefore we can consider the field ir*M(X) CM(Y) as a subfield, whence
M(Y) is an extension of the field TT*M(X). Since IT* is an imbedding we can
identify M(X) with ir*M(X). But y(ir*f) = (TT*/) ° 7 _ 1 = / ° n o y'1 =
= / o TT = TT*(/) for all / E M ( X ) .
Since 7r o 7" 1 = 7r for all 7GDeck(17X). Hence each automorphism
/—» 7 / is an element of the Galois group Gal (M(Y)/ir*M(X)).
If B ={ &!,..., bn\ is such a maximal subset of a field iDk where
blt...,bn are algebraically independent over k, that / = k\b\,..., bn\ then
the number n is called the transcendent degree of / with respect to k and is
denoted by deg txk / (or dim tr^ /), every field extension K of k can be
obtained as an algebraic extension of the transcendent extension k (bt,..., bn)
of k. In this case K is called an algebraic function field of n (transcendent)
variables bx,..., bn.
2.8. Example. The field M^(C) of elliptic functions is a algebraic function
field of one variable (according to the Weierstrass theorem!).
Carl Ludvig Siegel found a powerful method of proving theorems on
algebraic dependence. The most famous is a result anticipated and attacked
without success for many years by Weierstrass and Poincare.
2.9. Theorem (Siegel-Thimm 1939). Let X be a complex compact mani-
fold of dimension n. Then each n meromorphic functions flt... ,fn EM(X)
are algebraicaly dependent and
dim trcM(X)<n .
As a corollary one obtains the famous
2.10. Theorem (Poincare). The field Mr(Cp) where the torus Cp/T is an
algebraic manifold is a finite algebraic extension of the field C(fl}... ,fp)
where f x , . . . ,fp are any p algebraically independent elements of Mr(Cp).
20

Algebraic function fields and Riemann surfaces.

If Y, Y' are isomorphic compact Riemann surfaces i.e. if there exists a


biholomorphic map F: Y —*Y\ then the function fields M(Y) and M(Y')
are isomorphic: the isomorphisms is given by the mapping
F*:M(Y')-+M(Y) , (F*f(y)i = f(F(y)).
But the converse is also true: we have the following
2.11. Theorem. Two Riemann compact surfaces Y,Y' are isomorphic if
and only if the fields M(Y) and M(Y') are isomorphic:
( F ~ Y') *> (M(Y) =M(Y')) .
One can prove the following interesting
2.12. Proposition. If 7r:Y—>X is a (holomorphic) covering with n
sheets (for each x GX ir'1 (x) has n elements) then
dim(M(Y)/Tr*M(X)) = n .
Therefore if z-.Y —> C is a meromorphic function on a compact Riemann
surface Y then any other meromorphic function f on Y satisfies an alge-
braic equation
fn + rx(z)fn-' + ... + r,,-1(2;)/+ rn(z) = 0
where r^, k = l,...,n are rational functions of z. Therefore dim (M(Y)/C(z)) =
= yn. Moreover it can be proved that
M(Y) = C(z,f).
We see that M(Y) is an algebraic function field of one variable.
The narrow connection between the Riemann theory of compact sur-
faces and field extensions can be summarized in the following beautiful
2.13. Theorem. Let k : = M(C) be the field of rational functions (on C)
and let P be an irreducible polynomial of degree n with coefficients in k.
Let z.Y—• C be the global algebraic function defined by P. Let I =M(Y).
Then 1° dim(//&) = n. 2° Each decktransformation.
Deck (Y/X) defines an automorphism of the field extension M(Y)/M(X)
and the representation Deck (Y/X) —• Aut (M( Y)/M(X)) is an isomorphism.
We see the deep analogy between the covering 7r: Y-^X (= C) and
the field extension M(Y)/M(X).
21

Two questions naturally arise at this point: I. Is any abstract algebraic


function field (of one variable!) isomorphic to the concrete function field
M(Y) of meromorphic functions on some compact Riemann surface M}\
This question is quite analogue to the fundamental question of the theory
of Riemann surfaces:
II. Is any abstract compact Riemann surface M the covering of the
sphere C ? In other words-does there exists on Y a non constant meromor-
phic function? As we know the positive answer to II gives:
2.14. Theorem of Rado. On any Riemann surface (compact or not!) there
exists a non constant meromorphic function.
Question I was answered affirmatively by the fundamental, classic paper
(1882) of Richard Dedekind and Heinrich Weber. This paper was a true eye
opener: it marks the beginning of the theory of function fields.
2.15. Theorem (Dedekind-Weber). Let / be an algebraic function field
over complex numbers C. And let dim l/C(z). Then there exists a concrete
compact Riemann surface Y which is an w-sheeted cover of C, such that
l^M(Y).
Moreover the three automorphism groups Aut (//C(z)), Aut (M(Y)/M(C))y
Aut (Y/C) are isomorphic. The representation Deck (Y/C) —• Gal (M(Y)/C(z))
— given by y —> yf — is an isomorphism. The following table gives this
astrounding analogy between function fields and Riemann surfaces

Fields extensions Riemann surfaces

field extension l/k covering IT : Y—>X

Galois group Gal (l/k) = Aut (l/k) Deck (Y/X) = Aut (Y/X)

algebraic function field / the field M(Y)

dim (l/C(z)) = n 7T: Y —• C in w-sheeted

C(z,f) as a model of / concrete Riemann surface z : Y —>• C

The content of this chapter could be described as another example of


our general principle of development in modern mathematics: which in turn
is a beautiful illustration of Plato's Parable:
Ascent. (The field M(Y) of meromorphic functions on a compact of
Riemann surface) / (An abstract algebraic function field).
22

Descent. (An abstract function field /) \ (The construction of a compact


Riemann surface Y such that l=M(Y)).

Riemann surface Y as an algebraic variety.

We have mentioned in Chapter I the famous Kodaira embedding theorem,


which is a magnificient generalisation of the projective embedding of compact
Riemann surfaces. Let X = H/T be a Riemann surface each meromorphic
function on X can be considered as a T-invariant meromorphic function on
//-hence a r-automorphic function. M(X)=Mp(H), Poincare has constructed
for each X, r-automorphic functions f\,...;fN such that
HBx-* (/i(*),... ,fN+1(x)) =:F(x) GPN(C)
is an imbedding. Since F(yx) = F for 7 G T = Deck (Z//X), x H we obtain
the map Fp :X —> PN. Hence X is an analytic subvariety of a projective
space PN, and by a famous theorem of Chow is an algebraic variety. We have
thus the beautiful concretisation of the Kodaira's theorem due to Poincare.
2.16. Theorem. Every compact Riemann surface can be embedded by
automorphic functions into a projective space, thus it can be considered as an
algebraic surface (even in P 3 - cf. the next Chapter).
Remark. We have considered the most important case: Riemann surfaces
of genus g > 1. The case of the genus g = 1, hence a torus is much easier: in
that case X = C and the r-automorphic functions are simply elliptic func-
tions. Hence we have a
2.17. Proposition (Weierstrass). Each complex torus can be imbedded
projectively by elliptic (Weierstrass) functions:
C32-^(l,pr(p),^(2))GP2
thus
C/T-^P2.
The circle is closed: the whole complex analysis began with the fantastic
discovery of elliptic functions by the greatest Norwegian: Niels-Hendrik Abel
(1821). He had discovered that some interesting integrals of algebraic func-
tions when considered in complex domain possess as inverse functions two
periods: the theory of elliptic function was born.
Let us close with a brilliant remark of a great master of our science

i
23

Martin Eichler: "... mathematics obeys the law of the biological theory of
evolution: all individuals follow the general development of the species... The
immense formative force of mathematics on the human mind assures particu-
lar adherence to this law".

CHAPTER III

IMBEDDINGS THEOREMS AND PLATO'S CURVE PARABLE

The impulse for the creation of new mathematical concepts, theories and
disciplines is always given by the desire to solve concrete problems.
The deepest and boldest mathematical ideas of Riemann's theory were
"dictated" by the desire to understand algebraic functions and their integrals.
Complex manifolds and complex spaces arose in a "natural" and convincing
way from the idea of analytic continuation of germs of holomorphic func-
tions. The set of these germs form an "abstract" topological manifold, which
is in a natural way endowed with a complex structure, but which is not
situated in C or Rk. It was Riemann again, who stressed the need of multi-
dimensional geometry and who has given the notion of Riemann space
(manifold) — as an "abstract" object i.e. as an mathematical entity not im-
bedded in any Rk. It is interesting that in Riemann, concepts (and of his
genial follower and English translator William Kingdon Clifford (1845-1879))
physical considerations played an important role. There is no a priori reason
that the real 4-dimensional space-time should be imbedded in an Euclidean
space (of many dimensions). Of course a surface or a manifold imbedded in
EN inherit from EN its natural Riemann metric. The natural question arises:
can these magnificent and indispensable mathematical objects: differentiable
manifolds, Riemann spaces, Riemann surfaces and their natural many dimen-
sional analogues — Stein manifolds — be isomorphic ally imbedded in RN or
C N ? By "isomorphically" we always understand an imbedding which pre-
serves the structure of the corresponding object: in the case of Riemann surface
and complex manifold, a biholomorphic maping; in the case of Riemann (analytic)
manifold, an isometric differentiable (or analytic) imbedding. Solutions of
24

important and natural problems are very difficult and therefore require the
greatest invention and labour of several generations of mathematicians.
Let us now collect some of these magnificiant theorems:
A. Theorem of Whitney (1936). Let M be an ^-dimensional differentiable,
second countable manifold. Then there exists a proper C°° imbedding
F:M->R2m+1.
Remark. The map F is proper if F'1 (K) is compact for each compact K.
B. The problem of isometric imbeddings of a Riemann manifold (M,d) is
plainly much more difficult although this problem is much older then the A
(B was considered already by Schlaefli in 1873). The problem was completely
solved in the magnificiently difficult labour of John Nash (1954).
3.1. Theorem of J. Nash. Every m-dimensional Riemann manifold with
C -metric, where 3 < & < ° ° , can have a Ck-isometric embedding into RN,
k

where N= 111 m3 4- 7m 2 + 1 1 / 2 m.
The work of Nash uses very sharp and difficult estimates from the theory
of elliptic differential equations.
C. Embedding of Stein manifolds. The theory of functions of several complex
variables encounters quite different phenomena then the complex analysis of
1 dimension. For instance every open subset 0 C C is a natural domain of
some holomorphic function f-.f cannot be extended holomorphically to a
greater domain. As was remarked by Fritz Hartogs (1905) the situation in C 2 ,
is quite different: there are such domains O C C 2 that every holomorphic
function in 0 can be extended to a greater domain.
Holomorphy domains in CN, N> 1 could be characterized by some
sort of convexity. This lead Karl Stein (1951) to consider a very natural class
of complex manifolds which are a generalization of holomorphy domains in
CN. These spaces were investigated in the fundamental papers of Hans Grauert
(a pupil of Stein and Behnke) which considerably simplified the axiomes of
Stein:
3.2. Definition. An w-dimensional complex manifold X is Stein manifold if
a) y for every pair of different points x,y E X there exists such fGA(X),
that f(x)¥=f(y) (one says that holomorphic functions A(X) on X sep-
arate points of X).
b) For each zGX there exist n functions flt ...,/„ £A(X) which form a
coordinate system in z.
25

c) X is holomorphically convex: for each compact KCX the set


£=i2eX:|/(2)|<sup|/(/C)| forall fGA(X)\
is compact.
3.3. Remarks. 1. Grauert proved that every Stein space is second countable:
has a countable basis of neighbourhoods.
2. If one takes in 3.2 c) only linear functions / i n X C CN then X is
convex. Therefore the holomorphic convexity is a natural extension of the
notion of convexity!
3. Plainly no compact complex manifold is Stein since only constant
functions are holomorphic and a) is not valid.
4. A fundamental theorem of Behnke and Stein asserts that open (i.e.
non compact) Riemann surfaces are Stein.
5. By 4. Behnke and Stein solved a famous
3.4. Conjecture of Caratheodory, that on every open Riemann surface X
there exist non constant holomorphic functions f.X—> C. Therefore each
open Riemann surface X is a covering of an open subset of (C.
6. (OCC N -is Stein) -o- (0 is domain of holomorphy). These remarks
show the importance of the
3.5. Theorem (Remmert-Bishop-Narasimhan). Every ^-dimensional Stein
manifold X can be imbedded properly and holomorphically into C2n+1.
Hence {X is Stein) <& (X is holomorphic submanifold of C 2w+1 ).
D. The problem of analytic imbedding of real analytic manifolds is even more
difficult than A and B.
It is of philosophic interest that the solution of D was obtained (by
Hans Grauert) only by the methods of complex analysis:
3.6. Theorem (Grauert 1958). Every real analytic manifold of n dimen-
sions can be analytically and properly imbedded into R2m+1. Hence
(X is real analytic manifold) -& (X is a w-dimensional
analytic surface in R2m+1).
E. Projective imbeddings. The great mathematician and historitian of mathe-
matics Salomon Bochner writes: "The advent of mathematical space (and mod-
ern mathematics K.M.) was heralded by the emergence of linear perspective in
26

painting in the work of Alberti (1404-1472) (which was followed by the begin-
nings of projective and descriptive geometry in the 16 th century) and by the
emergence of a mathematically conceived spatial structure of the universe in
the theology of Nicholas of Cusa (1401-1464). Problems of persective ("infi-
nite points") led to the notion of the projective space P3- first real and soon
afterwards to the complex projective space P(Cn)".
The most important geometric objects are algebraic curves and surfaces:
conic sections and quadrics were known in hellenistic geometry. A natural and
fundamental notion is that of the algebraic set (or variety): the locus given by
zeros of polynomials. The geometry of the XIX century was the geometry of
algebraic varietes and the mathematics of the XX century is concerned to a
large extent with algebraic geometry.
The projective space P(E) where E is a complex (real) vector space can
be defined in three ways as:
i) the set of 1-dimensional complex subspaces (complex lines) of the vector
space C w+1 ;
ii) the set of the set of («4-1)-tuples a0,... ,an of complex numbers not
all zero modulo the equivalence relation
{aQ9...,an)«»(zaQt...,zan) , zGC-{0|;

iii) the unit sphere \a G Cn+1: \\a\\2 = l ) in Cn+1 modulo the equivalence
relation
(a0,...,an)^(eita0,...,eitan) , 0<t<2ir.

The (m 4- l)-tuple a = (a0,..., an) defining a point x of Pn is set of homo-


geneous coordinates of x Pn can be viewed as Cn U $ a "hyperplane at
infinity".
Let Hk = \ct E Cn : Uk = 0}, then Pn -Hk isomorphic to Cn by the
map
/ a0 ak_x an \
(a0,... ,an) —> ( — , ... , , ... , — I .
a
\ *k k #k /
Since by iii) P is compact and H^ is isomorphic to Pn~1 (by disregarding
n

the kth coordinate), hence Pn can be regarded as the result of compactify-


ing Cn by suitably glueing Pn~x at infinity.
A closed algebraic set in Pn is a subset of the form V{fx,... ,fr) =
= {x £P M : if (a0,..., an) are homogeneous coordinates of x, then fx (a0,...,
an) = ... =fr(a0)... ,an) = 0J, where fi,...,fr are homogeneous polynomials
27

(with complex coefficients). If an algebraic set is a complex manifold then it


is called an algebraic manifold. Algebraic manifolds are compact (as closed
subsets of the compact set Pn) hence they are never Stein! One of the most
important facts of algebraic geometry is the
3.7. Theorem of Chow (1949), (X can be embedded holomorphically in
n
P ) <* (X is an algebraic manifold).
This fundamental result of Chow reduces many problems of algebraic
geometry to problems of analysis: algebraic manifolds are holomorphic
subsets of P(Cn+1). We see here a vast gulf between holomorphic submani-
folds of Cn - they are (non compact!) Stein manifolds and holomorphic sub-
manifolds of Pn =P(Cn+1): they are (compact!) algebraic sets!!!
Chow's theorem could be considered as a generalization of the classic
elementary
3.8. Theorem (Riemann). Every meromorphic function on P 1 ( = C ) is
rational function.
We have already mentioned that each compact Riemann surface can be
holomorphically imbedded into some Pn, hence by Chow's theorem is an
algebraic surface. The next big step came with the theorems on projective
inbeddings of certain complex tori Ck/T, called Abelian varieties (manifolds).
The crowning result in this long and dramatic story is the famous result of
Kunihiko Kodaira (1954), who proved that algebraic manifolds can be defined
intrinsically: they are precisely complex manifolds provided with a hermitian
metric of a special type the so called Hodge manifolds.
3.9. Theorem of Kodaira. (X is algebraic manifold) <=> (X is Hodge mani-
fold) O (F : X - ^ > Pn).
This theorem says that a compact complex manifold provided with
special Riemann metric — the Hodge metric — can be projectively imbedded
and hence is algebraic. ^
H. Grauert extended (in 1958) the Kodaira theorem for an important
class of complex spaces and gave an intrinsic characterisation of algebraic
varieties (algebraic sets witn "reasonable" singularities).
The question arises of what is the dimension n of the imbedding projec-
tive space Pn ? The answer is given by
3.10. Theorem. If dim c X = k (that is the real dimension of the manifold
X is 2k) then there exists a projective imbedding into p2k+1.
We have therefore the important
28

Corollary. Every Riemann surface can be holomorphically imbedded


into the 3-dimensional complex projective space P3 = P((C4).

CHAPTER IV

COMMUTATIVE BANACH ALGEBRAS. THE SPECTRAL THEOREM

In this last Chapter we shall give another important example of our


principle: it is the beautiful Gelfand spectral theory of commutative Banach
algebras i.e. algebras, which are Banach spaces. This theory allowed I.M.
Gelfand to unify the Hilbert-v. Neumann spectral theory and harmonic
analysis. If is an indispensable tool of the theory of group representation and
of some parts of mathematical physics. The beginnings of this theory are to
be found in the v. Neumann theory of operator algebras in Hilbert space. John
v. Neumann was led to his theory by his attempts to axiomatise quantum
mechanics: we have the following line of development:

Hilbert —• Heisenberg —> v. Neumann —> Weyl —• Gelfand.


If we denote by L (H) the set of all bounded linear operators in a Hilbert
space H and if \\A\\ denotes the norm of the operator A EL(H) and A*
the adjoint of A (defined by (Af\g) = (f\Ag), identically for / , # £ / / ,
where (•!•) denotes the scalar product in H), then L(H) is a Banach
algebra with involution *. Plainly L (H) has the following properties
1° (A*)*=A ; 2° (A+B)* =A* +B* ;
3° (zA)* =zA* forall zE(L ; 4° (AS)* = B*A*
5° ||,41|2 < ||AM ||
A s u b s e t ^ of L(H)7 such that
6° 1H G si, and 7° AB = BA for all A,BGs/
is called a commutative algebra with unit. If moreover 8° the normed space
(j/, II II, *) is complete is called a commutative C*-algebra (of operators in
a Hilbert space H).
29

4.1. Fundamental Example. Let X be a compact space and C(X) denotes


the space of complex continuous functions on X. Then if we put
||/||:=max|/(X)i , /*(*): = /(*) ,
1-the constant function x-+l, then plainly C(X) is a commutative C*-al-
gebra with unit.
We define now an ideal (i.e. "abstract") object
Definition. A Banach algebra with involution satisfying l°-8° is called
a commutative C*-algebra with unit. During world war II I.M. Gelfand had
built one of the most perfect theories of functional analysis: the so called
"Gelfand spectral theory" or "Gelfand's theory of maximal ideals": He proved
that maximal ideals mx of a commutative Banach algebra Cwith unit) $4 are
in 1-1-correspondence with continuous homomorphisms of x : A —• (C:
mx = kerx = {^4 Gstf: x(A) = 0} .
Such homomorphisms are also called linear and multiplicative functionals.
The set Horn (A, (C) of such homomorphisms is of paramount importance
and is called the spectrum of the algebra A and is denoted by spec A . A
fundamental theorem of Geldand asserts:
4.2. Theorem (Gelfand). If A is a commutative Banach algebra with unit
then the spectrum spec^l of A is a compact space.there is a norm decreasing
There is a norm decreasing homomorphism srf —> C(X) defined by
(4.3) s/3A^fA(-)<EC(A) , fA(x) = x{A),
where x £ Horn (A, (C). The map (4.3) is often called the Gelfand map.
Remark. The "Gelfand map" was discovered in another context by
Richard DEDEKIND in his famous Commentary to Dirichlets "Vorlesungen
uber Zahlen Theorie", where the notion of an ideal is also introduced!
We have considered several C*-algebras, the famous theorem ,: Gelfand
and Najmark asserts that the 4.1. Example embraces all commutative
C* -algebras:
4.4. Theorem (Gelfand-Najmark). Every (abstract) commutative C*-alge-
bra s/ with unit is isometrically isomorphic to the algebra C(X) where the
compact space is the set spec^, the set of all maximal ideals of the alge-
bra J/.
30

This theorem is a beautiful illustration of the platonic cave parable. Such


a deep result must have important applications. We will give here only one, the
famous
4.5. Spectral theorem (Hilbert-v. Neumann). Let A =A* GL(H) be a
selfadjoint bounded operator in a Hilbert space H. Then there is a spectral
measure Ex on the spectrum X of A such that for each Borel function (f
on X we have

(4.6) <p(A) = I y(x)dEx , especially

(4.7) A = I xdEx
Jx
The proof is now quite simple: One extends the Gelfand (-Najmark)
isomorphism (4.3), where $4(A) is now the C*-algebra generated by the
operator A (and 1H), to all Borel functions on Spec$/(/4). Then to the
characteristic function %s of a Borel subset S CX corresponds precisely
the projective operator Es. Then the fundamental t formula (4.6) of the
spectral theorem is the well known fact: each Borel function <p on X can
be approximated by step functions.
Remark. Formula (4.6) is often used to define functions of an self-
adjoint (or normal) operator A.

Concluding Remark. The history of mathematics is a history of con-


tinual ascents from Plato's cave to the world of ideas. These ascents are
accomplished mostly in several stages.
Descents are even more difficult then the ascents. But this spiritual
movement is progressive rather than circular: the cave becomes brighter
and brighter.
31

BIBLIOGRAPHY

[1] Eichler, M., Introduction to the theory of algebraic numbers and functions. N.Y.
and London 1966.
[2] Maurin, K., Analysis //, Warsaw 1980.
[3] , Globale Analysis, Heidelberg 1976.
[4] , Mathematic als Sprache and Kunst in u Offene Systeme, Hsg. Maurin. Michelski,
Rudolph, Stuttgart 1981.
[5] Shafarevich, I., Basic Algebraic Geometry, Springer 1977.
[6] Weyl, H., Die Idee der Riemannschen Flache (1913, 1955).

KRZYSZTOF MAURIN, Universita di Varsavia, Polonia.

Lavoro pervenuto in redazione il 26/11/1982

Potrebbero piacerti anche