Diss

Constructive Invariant Theory and the Linearization Problem
Inauguraldissertation
zur Erlangung der Wurde eines Doktors der Philosophie
vorgelegt der Philosophisch-Naturwissenschaftlichen Fakultat der Universitat Basel
Harm Derksen
aus Ven-Zelderheide (Niederlande)
von
Basel, 1997
Genehmigt von der Philosophisch-Naturwissenschaftlichen Fakultat auf Antrag der Herren Prof. Dr. Hanspeter Kraft und Dr. Arno van den Essen. Basel, den 21. Januar 1997 Prof. Dr. Stephen C. Stearns Dekan
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii I. Computation of Reductive Group Invariants . . . . . . . . . . . . . . 1
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. The nullcone . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Finding generators of the zero- ber ideal . . . . . . . . . . . 4. Degree bounds for generating invariants . . . . . . . . . . . . 5. The algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 6. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II. On the hypersurface x + x2 y + z2 + t3 = 0 in C 4 . . . . . . . 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Technical preparation . . . . . . . . . . . . . . . . . . . . . . 3. The proof that x + x2 y + z2 + t3 = 0 is not isomorphic to C 3 III. Nonlinearizable Holomorphic Group Actions . . . . . . . . 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Non-recti able embeddings . . . . . . . . . . . . . . . . . . . 3. Rees spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Non-linearizable group actions . . . . . . . . . . . . . . . . . Glossary of Notations . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lebenslauf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 .4 .5 .6 11 12
15 21
21 22 23 25 15 16 18
32 35 39
Introduction
In this thesis, some problems in the theory of transformation groups will be discussed. The rst chapter is about constructive invariant theory, and the other two chapters are dedicated to the algebraic and holomorphic linearization problem. The three chapters are independent from each other. The rst chapter is related to the fourteenth problem of Hilbert ( 26]). If G is an algebraic group acting linearly on an n-dimensional vector space V (all de ned over some eld k), then G also acts on O(V ) = k X1; : : : ; Xn ], the ring of polynomial functions on V . If O(V )G is the ring of invariant functions, then Hilbert's question can be formulated as: Is the invariant ring O(V )G always nitely generated as a k-algebra? (the original formulation was more general). Nagata gave a counterexample in 1959 ( 50]). Hilbert proved essentially that O(V )G is nitely generated when G is linearly reductive in 1890 ( 24]). This proof was criticized because the new revolutionary abstract methods of Hilbert were not constructive at all. Three years later he published another paper ( 25]) with a di erent, more constructive proof. In this chapter Hilbert's rst proof will be extended to make it constructive too, and an algorithm will be described to compute a set of generators of the invariant ring. Essentially this algorithm consists of a single application of Buchberger's algorithm to compute a Grobner Basis. This algorithm is easy to implement and works for all linearly reductive groups. Using Hilbert's second proof, Popov found an upper bound for the minimal d such that the invariants of degree d generate the invariant ring for semi-simple G in characteristic 0 (see 54] and 55]). Hiss improved this bound and generalized it to arbitrary linearly reductive groups (cf. 27], 11]). Her bound for primary invariants doesn't depend anymore on the dimension of the representation. In special cases, for example tori, nite groups, SL2 , better bounds are known. We will show that there is a relation between some conjectures in commutative algebra, and the problem of nding good degree bounds. If for example the Eisenbud-Goto Conjecture is true (Conjecture I.4.1), then one can proof a good degree bound for generating invariants which would drastically improve the bounds of Popov and Hiss. A classical result of Noether is that for nite groups in characteristic 0, the ring of invariants is generated by invariants of degree jGj, where jGj is the order of the group (cf. 52]). It was conjectured that this is also the case in positive characteristic whenever the characteristic doesn't divide jGj. This can be shown if the Subspaces Conjecture is true (Conjecture I.4.2). The second and the third chapter deal with the linearization problem. Suppose that G is a reductive linear algebraic group over C , acting algebraically on C n . Is every G-action on C n essentially a linear action? Or more precise: Does there exist for every G-action on C n a polynomial automorphism, such that the action will become linear if we conjungate the action with this automorphism. In 1989 the rst counii
Introduction
iii
terexamples were found by Schwarz (see 60]). He used algebraic G-vector bundles over representations to construct counterexamples for SL2 and O2. In fact, Knop showed there are counterexamples for all non-abelian connected reductive groups (cf. 38]). This approach doesn't work for abelian groups, because of a result of Masuda, Moser-Jauslin and Petrie ( 47]), stating that G-vector bundles over representations are trivial whenever G is abelian and reductive. There are no examples known of non-linearizable algebraic actions of abelian reductive groups. For the multiplicative group C ? there are a few positive results known (see 40]). All C ? -actions on C and C 2 are linearizable (see for example 5] and 6]). Kaliman, Koras, Makar-Limanov and Russell are close to proving that all C ? actions on C 3 are linearizable. The C ? linearization problem on C 3 is related to examples of \exotic algebraic structures on C 3 ". There are examples of hypersurfaces in C 4 which are contractible, having a \hardcase" C ? action. By a result of Dimca it follows from contractibility that the hypersurfaces are di eomorphic to R6 (cf. 12]). One can show that if the hypersurface is algebraically isomorphic to C 3 , then the C ? -action cannot be linearizable. The rst examples of contractible hypersurfaces were found by Libgober in a di erent context ( 45]). His list was generalized by Dimca ( 12]), Kaliman ( 33]) and completed by Koras and Russell ( 39]). In order to prove the C ? linearization problem on C 4 one has to proof that all the examples in Koras-Russell's list are not isomorphic to C 3 . The simplest example of the list of Koras and Russell is the hypersurface x + x2 y + z2 + t3 = 0. It has a C ? action given by (x; y; z; t) = ( 6 x;
6 y; 3 z; 2 t):
If x + x2 y + z2 + t3 = 0 is isomorphic to C 3 , then this action cannot be linearizable. However, Makar-Limanov showed in 46] that x + x2 y + z2 + t3 = 0 is not isomorphic to C 3 . He de ned an invariant of rings which distinguishes the rings C x; y; z; t]=(x + x2 y + z 2 + t3 ) and the polynomial ring C x; y; z ]. Kaliman and Makar-Limanov showed that all hypersurfaces in the list of Koras and Russell are not isomorphic to C 3 (cf. 34]). In the second chapter we give a short proof of the fact that x + x2 y + z2 + t3 = 0 is not isomorphic to C 3 . A new invariant is introduced to distinguish the coordinate rings. Until very recently the linearization problem in the holomorphic category was open. Heinzner and Kutzschebauch showed that all holomorphic G-vector bundles over representations are trivial when G is a complex Lie group (see 22]). So the method of Schwarz cannot work in this situation. Kutzschebauch and I adapted an algebraic method of Asanuma (cf. 3], 4]) to the holomorphic category, to obtain non-linearizable holomorphic group actions. Asanuma constructed non-linearizable algebraic R?-actions on R5 and non-linearizable algebraic K ? -actions on K 4 where K is a eld of positive characteristic. For example, a non-linearizable R?-action on R5 can be obtained from non-recti able embeddings of R in R3, in the sense that there doesn't exist any polynomial automorphism of R3 which maps the image of R onto a straight line. Chapter III is basically a paper jointly written with Frank Kutzschebauch. In this chapter will be explained how one can obtain non-linearizable holomorphic C ? -actions using non-recti able proper holomorphic embeddings of C into C n , found by Forstneric, Globevnik, Rosay and Rudin (see 16], 56]). By extending the method of Asanuma, I was able to nd counterexamples for all complex reductive groups.
iv
Introduction
I would like to thank H. Kraft for giving me the opportunity to study at the University of Basel and for his supervising, F. Kutzschebauch for our cooperation and making me acquainted with the holomorphic category, and S. Helmke for writing the TEX-macros. I thank A. van den Essen who has always been very supportive and M. van Hoeij with whom I have discussed a lot of mathematical problems in all those years I have been studying mathematics. I'm grateful to T. Asanuma for giving me his preprint which was essential for our result, and J. Muller for showing us the proof of Lemma III.4.8. I like to thank B. Sturmfels for discussions and helpful suggestions. I'm grateful to the VSB bank and the Swiss National Foundation (Schweizerischer Nationalfonds) for nancial support.
Chapter I
Computation of Reductive Group Invariants
1. Introduction
Let G be a linear algebraic group de ned over the eld k. Let : G ! GL(V ), be a representation of G on a k-vector space V of dimension n < 1. The group G acts linearly on the k-algebra O(V ) = k X1; X2 ; : : : ; Xn] of polynomial functions on V , the coordinate ring of V . The invariant ring O(V )G is the subalgebra of invariant functions. This invariant ring was a major object of research in the last century (see 48]). A natural question to ask is: Is the invariant ring nitely generated as a k-algebra? This question is essentially Hilbert's 14th problem (see 26]), although his formulation was more general. In 1959 Nagata constructed a counterexample (see 50]). For reductive groups G, O(V )G is nitely generated by results of Hilbert, Weyl, Mumford, Nagata, Haboush and others (see 49]). The rst positive result in this direction was proved by Gordan in 1868. He proved in 19] that O(V )SL is nitely generated for every nite dimensional SL2-module V , using a constructive approach known as the symbolic method. Hilbert proved in 1890 that O(V )G is always nitely generated when G is linearly reductive (see 24]). A group is called linearly reductive when every nite dimensional G-module is the direct sum of irreducible modules (see for example 65]). Examples of linearly reductive groups are: reductive groups if char(k) = 0: semisimple groups, torus groups, nite groups, etc.; if char(k) > 0: torus groups, nite groups G such that char(k) doesn't divide the group order jGj. The proof of Hilbert's niteness result was not constructive. When Gordan saw Hilbert's method he made his famous exclamation:\Das ist Theologie und nicht Mathematik".*
2
* This is theology and not mathematics. 1
I. Computation of Reductive Group Invariants
After this criticism Hilbert wrote another paper in 1893 (see 25]) with a more constructive proof. He showed how to construct homogeneous algebraically independent f1; f2 ; : : : ; fr 2 O(V )G such that O(V )G is a nite k f1; f2 ; : : : ; fr ]-module. These f1 ; f2 ; : : : ; fr are called primary invariants and the module generators of O(V )G as a k f1; f2 ; : : : ; fr ]-module are called secondary invariants. This method leads to algorithms as described in Sturmfels' book (see 67]). There are several implementations for computing invariant rings of nite groups. Kemper wrote the invar package for the computer algebra system Maple (cf. 36], 37]). Heydtmann wrote the finvar.lib library in Singular (cf. 23]). There is also an implementation by Gatermann in Reduce (see 18]). In general, Hilbert's method can be complicated because one has to compute the normal closure of k f1; f2 ; : : : ; fr ] within the quotient eld of O(V ) which is a di cult computational problem. We will show that the rst proof of Hilbert can be extended to make it constructive too. This will lead to an algorithm for computing generators of the invariant ring for arbitrary linearly reductive groups G. This algorithm is essentially a single Grobner Basis computation. Suppose now that G is a linearly reductive group. An other interesting problem is to nd upper bounds for the degrees of the generators of the invariant ring. We de ne the following constant:
G (V ) = minf d j O(V )G is generated by invariants of degree
d g:
For some special cases, a good bound is known. Jordan gave a good bound in the case G = SL2 and char(k) = 0, using the methods of Gordan (see 30] and 31]). Let Vd := k X; Y ]d be the vector space of homogeneous polynomials in two variables of degree d. The group SL2 acts by substitution on Vd. These are all irreducible SL2Vdt . Jordan modules. If V is an SL2 -module, then we can write V = Vd Vd proved that where d = maxf2; d1; d2; : : : ; dt g. SL (V ) d6 For nite groups we have the bound of Noether (see 52]): If G is a nite group and char(k) > jGj or char(k) = 0, then
1 2 2
G (V )
jGj:
In special cases, Schmid improved this bound (cf. 58], 59]). If char(k) does not divide jGj then one still has the weaker bound
G (V )
maxfjGj; n(jGj 1)g:
We will show that if the Subspace Conjecture (conjecture 4.2) is true, then Noether's bound holds whenever G is nite and linearly reductive, i.e., if char(k) does not divide the group order. If char(k) is arbitrary, then Noether's bound does not hold in general. Gobel gave in 20] a constructive proof of the bound
G (V )
Smith proved that one still has Noether's bound in this case if G is solvable (cf. 62]).
maxfn; n(n 1)=2g
1. Introduction
if V is a permutation representation. Suppose G := T is a r-dimensional torus acting on V . Let X ? (T ) = Zr be the group of weights. Choose a translation invariant volume form on E := X ? (T ) Z R = Rrsuch that the lattice X ? (T ) E has covolume 1. The action of T on V can be assumed to be diagonal, and it acts with n weights 1; 2 ; : : : ; n 2 X ? (T ). Let CV E be the convex hull of 1 ; : : : ; n . Wehlau proved in 68] the following bound:
T (V )
maxfn r 1; 1g r! vol(CV ):
The rst general bound for semi-simple groups was found by Popov (see 54] and 55]). He combined the second constructive proof of Hilbert with the results of Hochster and Roberts stating that the invariant ring is Cohen-Macaulay (see 28]). In her thesis, Hiss was able to improve this bound and generalize it to arbitrary connected reductive groups, using some ideas of Knop's (see 27], 11]). Her degree bound for primary invariants does not depend on the dimension of the representation anymore. Suppose that G is a m-dimensional connected reductive group, let U G be a maximal unipotent subgroup and let T G be a maximal torus of rank r . Let CV be as before and we de ne NV by
NV := minf ` j X `+1v = 0 for all v 2 V; X 2 Lie(U ) g:

Then we have where
G (V )
n: lcmf1; 2; 3; : : : ; (V )g
and
r 1)! c(G) := 2 (m + )!2 r! : m r ( 2 This bound seems far from sharp. For example if G = SL2 and V = Vd then we get (V ) 96d3 and G(V ) d: lcmf1; 2; : : : ; 96d3 g. This bound is of course much worse than the bound G(V ) d6 of Jordan. Choose some norm k k on E = X ? (T ) Z R and let m (V ) c(G)NV r vol(CV )
LV = maxf k 1k; k 2 k; : : : ; k n k g
where 1 ; : : : ; n are the weights of T appearing in V . We will show that if the Eisenbud-Goto Conjecture is true (conjecture 4.1), then for any connected linearly reductive group there exist constants C; C 0 > 0 depending only on G and k k such that
G (V )
minfC (nLV )m ; C 0Lm +mg: V

2
This would be a great improvement of the bounds of Popov and Hiss. For a more detailed overview about constructive invariant theory, see 11].

2. The nullcone
Let k be a eld, and G be a linearly reductive group over k. Suppose that V is an n-dimensional representation of G. The inclusion O(V )G ,! O(V ) corresponds to a morphism : V ! V= , where V= := Spec O(V )G is the categorical quotient. The =G =G L ring O(V ) = k X1; X2 ; : : : ; Xn ] is graded O(V ) = i 0 O(V )i , and this induces a grading on O(V )G . Let us write m for the maximal homogeneous ideal (X1 ; X2 ; : : : ; Xn ) of O(V ). We de ne IN to be the ideal of O(V ) generated by all homogeneous invariants of positive degree, which we will call the zero- ber ideal. The space Spec(O(V )=IN ) is equal to 1 (0) in the schematic sense. The zero set of IN is the so-called nullcone N . The well-known Finiteness Theorem of Hilbert states that Theorem 2.1. If G is a linearly reductive group, and V is a representation of G, then O(V )G is nitely generated. Proof. The proof of this theorem is based on the following two facts: 1. The polynomial ring over a eld is noetherian: every ideal in the polynomial ring O(V ) is nitely generated (Hilbert's Basissatz). 2. There exists a unique G-invariant linear projection R : O(V ) ! O(V )G (i.e., R(g f ) = R(f ) for all g 2 G, f 2 O(V ) and R(f ) = f for all f 2 O(V )G ), which is called the Reynolds operator (see 65]). It has the following properties: (a) R is a O(V )G-module homomorphism; (b) if W O(V ) is a G-invariant linear subspace, then R(W ) = W G. The ideal IN is generated by invariants, and by the noetherian property we only need nitely many of them to generate IN , say IN = (f1 ; f2 ; : : : ; fr ). Suppose that h is a homogeneous invariant of degree d. Using induction on d, we proof that h 2 k f1; f2 ; : : : ; fr ]. If d = 0, then h 2 k and the statement is clear. Otherwise we can write r X h = ai fi
i=1
with a1 ; a2; : : : ; ar 2 O(V ). Without loss of generality we may assume that ai is homogeneous of degree d deg fi < d for all i. Applying the Reynolds operator yields
h = Rh = R
r X i=1
ai fi =
r X i=1
R(ai )fi :
Because the vector space of homogeneous polynomials of degree deg(ai ) is G-stable, we have that R(ai ) is homogeneous of degree deg(ai ) < d for all i. By induction hypothesis, R(a1 ); : : : ; R(ar ) 2 k f1; f2 ; : : : ; fr ], so h 2 k f1; : : : ; fr ]. The proof of Hilbert was criticized at the end of the nineteenth century because choosing the invariant generators f1; f2 ; : : : ; fr of IN is a non-constructive step. However, we will show that there is a constructive way to obtain these f1 ; f2 ; : : : ; fr . Lemma 2.2. Suppose h1; h2; : : : ; hs 2 O(V ) are homogeneous and generate IN . Then R(h1); R(h2 ); : : : ; R(hs) generate O(V )G as a k-algebra.
3. Finding generators of the zero- ber ideal

Proof. If f1 ; f2 ; : : : ; fr
span the vector space IN =mIN . Therefore, the action of G on IN =mIN is trivial, so the Reynolds operator is equal to the identity on IN =mIN . The image of hi and R(hi ) in IN =mIN are the same for all i. So the images of the R(hi ) span IN =mIN and R(h1); R(h2); : : : R(hs ) generate IN by the homogeneous Nakayama Lemma. Now we follow the proof of Hilbert's Finiteness Theorem to conclude O(V )G = k R(h1 ); R(h2 ); : : : ; R(hs )].
3. Finding generators of the zero-fiber ideal If generators of the IN are known, then generators of the ring of invariants are found
2 O(V )G are generators of IN , then the images of f1 ; : : : ; fr
by applying the Reynolds operator by Lemma 2.2. In this section we deal with the problem of nding generators of IN . Consider the map :G V !V V given by (g; v) = (v; g v). Let B V V be the image of . The ideal b of the Zariski-closure B of B is given by: b = f h 2 O(V V ) j h(v; g v ) = 0 8g 2 G; v 2 V g: We will identify O(V V ) = O(V ) O(V ) with the polynomial ring k X; Y ], where X and Y are abbreviations of X1 ; X2 ; : : : ; Xn and Y1 ; Y2; : : : ; Yn . Theorem 3.1. We have an equality: ((Y1 ; Y2 ; : : : ; Yn) + b) \ k X ] = IN : Proof. \ ": If f (X ) = f (X1 ; : : : ; Xn ) is an invariant of positive degree, then we get f (X ) = (f (X ) f (Y )) + f (Y ). Now f (Y ) 2 (Y1 ; : : : ; Yn) and f (X ) f (Y ) 2 b because (f (X ) f (Y ))(v; g v) = f (v) f (g v) = 0 for all g 2 G, v 2 V . \ ": Conversely, if f (X ) 2 (Y1; : : : ; Yn) + b, then we can write X (3:1) f (X ) = ci(X )fi (Y ) + b(X; Y )
i
where ci(X ) 2 k X ], fi (Y ) 2 (Y1 ; : : : ; Yn ) for all i and b(X; Y ) 2 b. We will view V V as a G-variety, where G acts only on the second factor. The corresponding Reynolds operator: RY : k X; Y ] ! k X; Y ]G = k Y ]G X ] is a k X ]-module homomorphism and its ristriction to k Y ] is equal to the Reynolds operator k Y ] ! k Y ]G because of its uniqueness. We apply RY on (3.1): X (3:2) f (X ) = ci(X )RY (fi (Y )) + RY (b(X; Y )): Now RY (b(X; Y )) 2 b because b is G-stable (we use property (b) of the Reynolds operator). Let : V ,! V V be the diagonal morphism. The corresponding algebra homomorphism is ? : k X; Y ] ! k X ], p(X; Y ) 7! p(X; X ). Applying ? to the equation (3.2) yields X f (X ) = ci(X )R(fi (X )):
i ? (RY (b(X; Y ))) = 0, because RY (b(X; Y )) 2 b ker( ? ). Now the In fact, we have R(fi (X )) are homogeneous invariants of positive degree, and we conclude that f (X ) i
lies in IN .
We can reformulate Theorem 3.1 as: Corollary 3.2. If the ideal b of B is generated by the f1(X; Y ); f2 (X; Y ); : : : ; fr (X; Y ), then f1 (X; 0); f2 (X; 0); : : : ; fr (X; 0) generate IN . The geometric interpretation of Theorem 3.1 is
B \ (V f0g) = N f0g:
It is important to notice that this equality holds in the schematic sense. Remark 3.3. Theorem 3.1 can be generalized. If (h1 (X ); : : : ; hr (X )) G-invariant ideal, then (h1(Y ); : : : ; hr (Y )) + b \ k X ] is the ideal of k X ] generated by (h1(X ); : : : ; hr (X )) \ k X ]G. Let W V is the zero set of h1; : : : ; hr , : V !: V= is the categorical quotient and p1 : V V ! V is the =G projection onto the rst factor. Then the geometric interpretation of this generalization is p1((V W ) \ B) = 1 (W ): A similar statement is true if V is an arbitrary a ne G-variety. Let G (V ) be as in the introduction. Our aim is to give a good upper bound for G (V ). We rst state two conjectures. Conjecture 4.1 (Eisenbud-Goto). If p is a homogeneous prime ideal in the polynomial ring k X1; : : : ; Xn] of degree d, then the module of i-th syzygies can be generated by homogeneous polynomials of degree
4. Degree bounds for generating invariants
k X ] is a
d + i + dimk p1 height p + 1
where p1 is the vector space of linear functions in p (see 14]). Conjecture 4.2 (Subspaces Conjecture). Suppose that W is a k-vector space and W1; W2 ; : : : ; Wd W are subspaces. If I is the ideal of W1 W2 : : : Wd, then I is generated by homogeneous polynomials of degree d. If G is a nite group then G is linearly reductive if and only if char(k) does not divide jGj. Noether's bound G(V ) jGj is known only in the cases char(k) = 0 and char(k) > jGj. Proposition 4.3. Suppose that the Subspaces Conjecture is true, and G is a linearly reductive nite group, i.e., char(k) doesn't divide jGj. Then we have G (V ) jGj. Proof. The image B of : G V ! V V is the union of at most jGj subspaces of dimension n. The ideal b of B = B is generated by polynomials of degree jGj. Using the lemma below we conclude G(V ) d. Lemma 4.4. If b is generated by homogeneous polynomials of degree d then we have G (V ) d.
Proof. If b is generated by f1 (X; Y ); : : : ; fr (X; Y ) 2 k X; Y ], then IN is generated by f1 (X; 0); : : : ; fr (X; 0) by Corollary 3.2 and the ring O(V )G is generated by the invariants R(f1 (X; 0)); : : : ; R(fr (X; 0)) by Lemma 2.2. If b is generated by homogeneous polynomials of degree d, then so is the invariant ring. Suppose that G is a linearly reductive m-dimensional group, and V is an n-dimensional G-module. Choose a maximal torus T G. Now X ? (T ) = hom(T; k? ) = Zr be the group of characters. Choose some norm jj jj on E := X (T ) Z R = Rr. The action of T on V can be assumed to be diagonal, and it acts with n weights, 1 ; 2; : : : ; n 2 X (T ). Let LV = maxfjj 1jj; jj 2jj; : : : ; jj njjg: Theorem 4.5. If the Eisenbud-Goto Conjecture is true, G is a linearly reductive mdimensional group and char(k) = 0 then there exist constants C; C 0 > 0, depending only on the group G and the choice of k k, such that
G (V )
minf C (nLV )m ; C 0Lm +m g: V

2
From now on we suppose char(k) = 0 and k algebraically closed. The variety B V V is a cone. Therefore we can view B as a projective variety in P(V V ) with some degree d. If G is connected, then B is irreducible and b is a prime ideal. If the Eisenbud-Goto Conjecture is true, then G (V ) d by Lemma 4.4. So we need a bound for d. For this we will use the formula of Kazarnovskii. We need some more notation. We x a Borel subgroup B containing T . We denote by 1; 2; : : : ; (m r)=2 , the positive roots. Let W be the Weyl group and let e1 ; e2; : : : ; er be the Coxeter exponents, i.e., e1 + 1; e2 + 1; : : : ; er + 1 are the degrees of the generating invariants of W. We denote by CV E := X ? (T ) Z R the convex hull of the weights appearing in V , 1 ; 2 ; : : : ; n . On E we use the volume form dV given by any isomorphism E = Rr which identi es X ? (T ) with Zr. Finally we x a W-invariant scalar product ( ; ) on E and denote, for any 2 E , by ^ 2 E ? the dual element de ned by ^( ) = 2( ; )=( ; ). For a representation : G ! GL(V ), let G (V ) be the number of solutions g 2 G of (4:1)
S1( (g)) = S2( (g)) = : : : = Sm( (g)) = 0
for general S1; S2; : : : ; Sm 2 End(V )? . Theorem 4.6 (Kazarnovskii, 35]). The number G(V ) is equal to Z m! ^^ ^ 2 (4:2) jWj(e1 !e2 ! er !)2 CV ( 1 2 : : : m r ) :
2
where 1 ; 2 ; : : : ; m r are the positive roots. See 11] for a more algebraic proof and 7] for a generalization of this result. Corollary 4.7. There exists a constant C > 0 such that
2
G (V )
CLm: V
8
Proof. The homogeneous polynomial
f := ( ^1 ^2 : : : ^m r )2
2
has degree m r. So jf (z)j C 0 kzkm r for some C 0 > 0. If we integrate f over a ball of radius LV , then the result will be (2LV )r :C 0Lm r , so V r (V ) jWj(e m! !2 e !)2 C 0Lm = CLm: G V V 1 !e2 r
number of solutions of (4.1) is nite, then this number is still bounded from above by (4.2). b Proof. Let G P(End(V )) be the projective variety given by the cone k ? : (G). The b number of solutions of (4.1) equals the number of solutions of p 2 G of
Lemma 4.8. If S1; S2; : : : ; Sm are not in general position, j 1(k? :id)j < 1 and the
S1(p) = S2(p) = : : : = Sm(p) = 0

multiplied by j 1(k? :id)j. By a generalization of the Bezout Theorem by Fulton and McPherson (see 12.3 of 17] and II.2.2 of 61]) this number of solutions is maximal when S1; S2; : : : ; Sm are in general position. We return to the computation of an upper bound for d, the degree of B V V viewed as a projective variety in P(V V ). Observe that B = G (V ) where (V ) is the diagonal of V V and G acts only on the second factor of V V . Now (V ) is the zero set of n linear functions fi = Xi Yi with 1 i n. De ne for i = 1; 2; : : : ; m functions on G (V ? V ? )n+m 1 by
Pi(g; h1; h2 ; : : : ; hn+m 1) = det(g:f1 ; g:f2; : : : ; g:fn; hi ; hi+1; : : : ; hn+i 1)

nite and d. Proof. The dimension of B is equal to n + m because the generic stabilizer is nite. The number of intersection points of B and h1 = h2 = : : : = hn+m 1 = 0 in P(V V ) is exactly d if the hi are general linear functions on V V . So d is equal to the number of g 2 G for which
Lemma 4.9. Suppose that the generic stabilizer of the G-action on V is nite. For hi 2 V ? V ? general enough, xed, the number of common solutions of the Pi is
(4:3)
f1(g:p) = f2 (g:p) = : : : = fn(g:p) = h1(p) = h2(p) = : : : = hn+m 1(p) = 0
has a solution p 2 P(V V ), divided by the order of the generic stabilizer. For every such g and 1 i m we have Pi (g; h1; h2; : : : ; hn+m 1) = 0. So the Pi have at least d common solutions for general h1; : : : ; hn+m 1. We now prove niteness. Let Z G (V ? V ? )n+m 1 be the zero set of P1; P2 ; : : : ; Pm . The projection of Z onto G is surjective. For xed g 2 G we will
compute its ber dimension. Set hi = hi + W 2 (V ? V ? )=W , where W is the span of all g:fi. Then det(g:f1; g:f2; : : : ; g:fn; hi; hi+1; : : : ; hn+i 1) = det(hi; hi+1; : : : ; hn+i 1) The solution space for (h1; : : : ; hn+m 1) for these equations has dimension (n+m)(n 1) by Lemma 4.10 below. For every hi we have an n-dimensional space of representatives hi, so the ber dimension is (n + m 1)n +(n + m)(n 1) = 2n2 +2nm 2n m. The dimension of Z is m+(2n2 +2nm 2n m) = 2n2 +2nm 2n. The generic ber of the projection to (V ? V ? )n+m 1 must have dimension (n + m 1)2n (2n2 +2nm 2n) = 0, so the general ber is nite. Lemma 4.10. The variety
Y := f(v1 ; v2 ; : : : ; vt ) 2 ku t j det(vi ; vi+1; : : : ; vi+u 1) = 0 (1 i t u + 1) g

has dimension (t + 1)(u 1). Proof. The variety Y is given by t u + 1 equations, so the codimension of Y is t u + 1. Let W be the (t u + 1)-dimensional subspace of ku t of all matrices
0 1 B0 B .. B. B @0
0
::: ::: ... 0 ::: 0 :::

2 1
t u+1 t u
t u+1
2 1
::: ::: ... ::: :::
0 0
t u+1 t u
t u+1
0 0 . . . 0
1 C C C C A
Now det(v1; v2 ; : : : ; vu) = 0 gives us n = 0, so 1 = 0. From the second equation 1 det(v2 ; : : : ; vu+1) = 0 follows 2 = 0, etc., so W \ Y = f0g. So Y is a complete intersection and its dimension is tu (t u + 1) = (t + 1)(u 1). Proof of Theorem 4.5. Let
V : G ! GL n(V V
V)
V End n (V
V)
be the representation of G on n(V V Qi 2 End n (V V ) ? is de ned by
V ). Now Pi(g) is equal to Qi ( (g)) where
v1; v2 ; : : : ; vu in V . It follows that L^n V

(
Vn(V V ) ! A(f ^ : : : ^ f ) ^ h ^ : : : ^ h V2n(V V ) = k 1 n i n+i 1 2 V V We Pu Corollary 4.7 for n(V V ). The weights appearing in n(V V ) are apply all with u n and ; : : : ; weights of linear independent T -eigenvectors
A 2 End
j =1 j
1
V)
nLV . The number of solutions of
Q1 ( (g)) = Q2 ( (g)) = : : : = Qm+n 1 ( (g))

is nite by Lemma 4.9. So this number is at most G n(V V ) by Lemma 4.8. On the other hand, from Lemma 4.9 follows that the number of solutions is at least d. So we get G (V ) d C (nLV )m .
10
If the generic stabilizer is not nite, take some xed representation V 0 with trivial generic stabilizer. Then G (V ) G (V V 0 ) C2 ((n + n0 )(maxfLV ; LV 0 g)m C3 (nLV )m where n0 = dim V 0 and C3 > 0 is a large enough constant. If G is not connected, then let G be the connected component of id 2 G, which is a normal subgroup. Now O(V )G is generated by homogeneous invariants of degree C3(nLV )m . From Noether's bound follows that O(V )G = (O(V )G )G=G is generated by homogeneous invariants of degree jG=G jC3(nLV )m C4(nLV )m . Suppose we have a decomposition V = V1a V2a : : : Vtat where the Vi are irreducible and di erent. By Weyl's theorem (see 69]) the generators of invariants of O(V ) can found by polarizing invariants of O(V 0 ) where
1 2
V 0 = V1dim(V ) V2dim(V ) : : : Vtdim(Vt) In this situation we have G (V ) G (V 0 ) C4 (n0 LV )m where n0 = dim V 0 . The number of di erent irreducible representations of highest weight where has length LV is C5Lr . These representations all have dimension C6LVm r = because of V the Weyl formula (see 3.2.6 of 53]):
1 2 ( ) 2
dim V =
1 2
(m r)
i=1
^i( + ) ^i ( )
2
X where = 1 2 i=1 i 2 E
1 2
(m r)
( This implies n0 (C5Lr ) C6LVm r)=2 V
C7Lm , and so we nally deduce V
G (V )
C4(n0 LV )m C4(C7 Lm+1)m = C8Lm(m+1): V V
Example 4.11. Binary forms.
Let G = SL2(k). The set of weights can be identi ed with Z. Let Vd is the vector space of binary forms of degree d. The highest weight of this SL2 -module is d, so we have L = d, m = dim G = 3 and n = dim Vd = d + 1. Suppose that the Eisenbud-Goto Conjecture is true. Because of Theorem 4.5 we get G (Vd ) C ((d + 1)d)3 < C2 d6 for some constant C2 > 0. If V = Vd Vd Vdt with di d for all i, then by Theorem 4.5 we have G (V ) C 0 d3 +3 = C 0 d12 : Indeed, Jordan proved that in this case we even have G (V ) d6 for d 2 (see 30] and 31]).
1 2 2
5. The algorithm
5. The algorithm
11
Suppose that O(G) is isomorphic to k Z1; Z2 ; : : : ; Zs ]=IG, where IG is an ideal of k Z1; Z2; : : : ; Zs] generated by h1; h2; : : : ; hr . The linear action on kn is given by
0 X1g 1 0 a1;1 (g) a1;2 (g) : : : a1;n(g) 1 0 X1 1 B X2g C = B a2;1 (g) a2;2 (g) : : : a2;n(g) C B X2 C B .. C B .. . . CB . C @. A @ . . . A@ . A . . .
g Xn
an;1(g) an;2(g) : : : an;n(g)
Xn
for all g 2 G, where ai;j 2 k Z1; Z2; : : : ; Zs] (ai;j is determined up to elements of IG ). Let G V V be the Zariski-closed subset f (g; x; gx) j g 2 G; x 2 V g. The coordinate ring of is
k Z1; Z2; : : : ; Zs; X1 ; X2 ; : : : ; Xn ; Y1; Y2 ; : : : ; Yn]=I

n a X with i = 1; 2; : : : ; n. where the ideal I is generated by IG and all Yi j =1 i;j j Now B is the closure of the projection of G V V onto V V . The ideal b of B is equal to I \ k X; Y ]. This intersection can be computed using Grobner basis (see 1], 10], or 13] for an introduction to Grobner Basis). Choose an admissible ordering on the monomials of k Z; X; Y ] such that Zi is bigger than any monomial in X and Y . Compute a Grobner basis S of I with respect to this ordering. The set S \ k X; Y ] is a Grobner bases of b = I \ k X; Y ] (with respect to the induced ordering on k X; Y ]). If this Grobner basis is ff1; f2 ; : : : ; fl g k X; Y ], then ff1(X; 0); f2 (X; 0); : : : ; fl (X; 0)g generate IN (by Corollary 3.2). And nally R(f1 (X; 0)); R(f2 (X; 0)); : : : ; R(fl (X; 0)) generate k X ]G (by Lemma 2.2). Application of R is not a trivial computation. It depends on the group G. For example if G is nite then R : k X ] ! k X ]G is given by
1 R( f ) = j G j
g 2G
fg:
For a semi-simple group the map R can be computed using the action of the Lie-algebra on k X ]. We will not discuss this here. The algorithm is as follows: input: s; n 2 N, h1 ; h2 ; : : : ; hr ; ai;j 2 k Z1; Z2 ; : : : ; Zs ] (1 i; j n) Pn a X ; : : : ; Y Pn a X ; h ; : : : ; h g I := fY1 n r j =1 1;j j j =1 n;j j 1 S := Grobner(I ; Z1 > Z2 > : : : > Zs > X1 > : : : > Xn > Y1 > : : : > Yn) b := S \ k X1 ; X2 ; : : : ; Xn ; Y1 ; Y2 ; : : : ; Yn ] IN := f substitute(fY1 = 0; Y2 = 0; : : : ; Yn = 0g; f ) j f 2 b g output: fR(g ) j g 2 IN g
12

6. Examples
In this section we will look at some concrete examples. Example 6.1. Gm-action with weights -5,-3,2,4. We take the multiplicative group Gm with coordinate ring k Z1; Z2]=(Z1 Z2 1). The group acts on the four-dimensional vector-space with weights 5; 3; 2; 4. The matrix of this representation is 0 Z25 0 0 0 1 B 0 Z23 02 0 C @ 0 0 Z1 0 A 4 0 0 0 Z1 We use the algorithm of the previous section to compute k X1; X2 ; X3 ; X4]Gm . Let I k Z; X; Y ] be the ideal of the graph . It is given by
2 4 5 3 I = (Z1 Z2 1; Y1 Z2 X1; Y2 Z2 X2 ; Y3 Z1 X3 ; Y4 Z1 X4 )
Using Grobner basis we can compute b = I \ k X; Y ] which is equal to

b = (X4 Y32
2 X3 Y4; X1 X3 Y2 X2 Y1Y3; X1 X4Y2 Y3 X2 X3Y1Y4 ; 2 2 2 2 2 2 X1X2 X4 Y1Y2Y42; X2 X3 X4 Y22Y3 Y4; X2 X4 Y3 X3 Y22Y42 ; X1 X4 Y22 X2 Y12Y4; 2 3 2 2 2 3 X2 X4 Y1 X1 Y23Y4; X1 X3 X4 Y12Y3 Y42; X1 X2X3 X4 Y1 Y2Y32Y4 ; X2 X3 Y22Y33; 4 2 3 3 2 2 3 X1 X4 Y3 X3Y12 Y43; X2 X3 Y1 X1 Y23Y32; X1 X3 X4 Y12 Y33Y4; X1 X2X3 Y1Y2Y34 ; 3 3 4 2 4 3 2 5 X1 X4 Y2 X2 Y13Y43; X2 X3 Y12 X1 Y24 Y3; X2 X4 Y24Y43 ; X1 X3 Y12Y35; 4 5 3 5 X2 Y13 X1 Y25; X1 X4 Y14 Y45) We nd generators of IN by substituting Y = 0. The null-cone ideal is generated by the elements 4 5 2 5 4 3 4 2 3 X1 X4 ; X1 X3 ; X2 X4 ; X1 X2X3 ; X1 X3 X4 ; 2 2 2 2 2 3 2 X2 X3 ; X1 X2X3 X4; X1 X3 X4 ; X2 X3 X4; X1 X2 X4 We don't have to apply the Reynolds operator here, because these generators are already Gm -invariant. In the torus case the algorithm presented here does the same as Algorithm 1.4.5. in 67]. Example 6.2. The nite group S3 acting on a direct sum of the irreducible 2-dimen-
sional and the signum representation. Let be the third root of unity. As an algebraic set (but not as a group), S3 is isomorphic to the set f1; ; 2g f 1; 1g because they have the same cardinality. A bijection is given by (1 2)j (1 2 3)i 7! ( i ; ( 1)j ). So the coordinate ring of S3 can be 2 3 identi ed with k Z1; Z2 ]=(Z1 1; Z2 1). The representation is given by the matrix
0 (1 + Z )=2 2 @ (1 Z2)=2
0
(1 Z2 )=2 0 Z1 0 0 2 (1 + Z2 )=2 0 A @ 0 Z1 0 A = 0 Z2 0 0 1
10
6. Examples
0 The ideal I de ning is given by
0 Z (1 + Z )=2 Z 2(1 Z )=2 1 2 2 1 @ Z1 (1 Z2)=2 Z12(1 + Z2)=2

0
0 0A Z2
13
3 2 2 I = (Z1 1; Z2 1; Y1 X1Z1(1 + Z2)=2 X2 Z1 (1 Z2)=2; 2 Y2 X1 Z1(1 Z2 )=2 X2 Z1 (1 + Z2)=2; Y3 X3 Z2) Eliminating Z1 and Z2 with Buchbergers Algorithm yields 2 3 3 b = (X3 Y32 ; X1 X2 Y1 Y2 ; X1 + X2 Y13 Y23 ; 2 2 3 X3 Y13 X3Y23 + 2X2 Y3 Y13Y3 Y23Y3; X2 X3 Y12 X1 X3 Y2 + X2 Y12Y3 X1 Y2Y3; 2 2 2 2 X1 X3Y12 X2 X3 Y2 X1Y12 Y3 + X2 Y2 Y3; X2 X3Y1 X1 X3 Y22 + X2 Y1 Y3 X1Y22 Y3; 3 3 2 2 X1 X3Y1 X2 X3 Y22 X1 Y1 Y3 + X2Y22 Y3; X2 X3 X3 Y23 + X2 Y3 Y23Y3; 4 2 X2 X2Y13 + X1 Y1Y2 X2 Y23) We obtain the zero-cone ideal by substituting Y = 0: 2 3 3 3 4 IN = (X3 ; X1 X2; X1 + X2 ; X2 X3 ; X2 ) 4 4 3 3 2 Note that we don't need X2 because X2 = X2 (X1 + X2 ) X1 (X1 X2). The rst three 3 X3 ) = 1 (X 3 X3 X 3 X3 ). So we nally get generators are already invariant and R(X2 1 2 2 2 3 3 3 3 k X1; X2; X3 ]S = k X3 ; X1X2 ; X1 + X2 ; X2 X3 X1 X3] Remark 6.3. See 36] and 18] for other methods to compute invariants of nite group actions. Kemper implemented his method in the Invar package in Maple. In 37] he deals with the case where the ground eld has positive characteristic. Example 6.4. The group SL2 acting on W W S 2W where W is the standard
3
2-dimensional representation. We take the group SL2 whose coordinate ring is k Z1; Z2 ; Z3; Z4 ]=(Z1Z4 Z2 Z3 1). The standard representation on the 2-dimensional space W is given by the matrix
Z1 Z2 Z3 Z4 The representation on S 2W is given by the matrix 0 Z2 2 1 2Z2Z1 Z2 1 @ Z3 Z1 Z4Z1 + Z3 Z2 Z4Z2 A 2 2 Z3 2Z4Z3 Z4

Now we will take the 7-dimensional vector space V := W W S 2W and compute the invariants k X1; X2 ; : : : ; X7]SL . Let I k Z; X; Y ] be the ideal which describes the graph G V V . We have
2
I = (Z1 X1 + Z2X2 Y1; Z3 X1 + Z4 X2 Y2; Z1 X3 + Z2X4 Y3; Z3X3 + Z4X4 Y4;
14
2 2 Z1 X5 + 2Z1 Z2X6 + Z2 X7 Y5; Z1 Z3X5 + Z2 Z3X6 + Z1Z4X6 + Z2Z4 X7 Y6; 2 2 Z3 X5 + 2Z3Z4 X6 + Z4 X7 Y7; Z2Z3 + Z1 Z4 1) We choose an ordering on the monomials in Z1; : : : ; Z4; X1 ; : : : ; X7; Y1 ; : : : ; Y7 such that Zi > every monomial in X and Y and we compute the reduced Grobner basis. In a computation in Singular this basis had 66 elements of which 18 lie in the ring k X; Y ]. These 18 polynomials generate the ideal b. The ideal b has 9 minimal generators, namely 2 b = X1 X7 2 2X1X2 X6 + X2 X5 Y12Y7 + 2Y1Y2Y6 Y22Y5;
X1 X3X7 2X2X3 X6+X2 X4X5 +X6Y2 Y3 X6Y1Y4 Y1Y3Y7+Y2Y3 Y6+Y1Y4Y6 Y2Y4 Y5; 2 2 X3 X7 2X3X4 X6 + X4 X5 Y32Y7 + 2Y3Y4 Y6 Y42Y5; X3X7 Y1 X4 X6Y1 X1 X7 Y3 + X2X6 Y3 + X4 Y1Y6 X2 Y3Y6 X4 Y2Y5 + X2 Y4Y5; X3X6 Y1 X4 X5Y1 X1 X6 Y3 + X2X5 Y3 + X3 Y1Y6 X1 Y3Y6 X3 Y2Y5 + X1 Y4Y5; X3X7 Y2 X4 X6Y2 X1 X7 Y4 + X2X6 Y4 + X4 Y1Y7 X2 Y3Y7 X4 Y2Y6 + X2 Y4Y6; X3X6 Y2 X4 X5Y2 X1 X6 Y4 + X2X5 Y4 + X3 Y1Y7 X1 Y3Y7 X3 Y2Y6 + X1 Y4Y6; 2 X2X3 X1X4 Y2Y3 + Y1 Y4; X6 X5 X7 Y62 + Y5Y7 Substituting Y = 0 yields
2 2 IN = (X1 X7 2X1X2 X6 + X2 X5; X1 X3 X7 2X2 X3X6 + X2X4 X5 ; 2 2 2 X3 X7 2X3X4 X6 + X4 X5 ; X2X3 X1X4 ; X6 X5X7 ) These are all invariants except X1 X3X7 2X2X3 X6 + X2X4 X5. If we apply the Reynolds-operator then we obtain
R(X1 X3 X7 2X2 X3X6 + X2X4 X5) = X1 X3 X7 X2 X3X6 X1X4 X6 + X2 X4 X5

The ring of invariants is therefore generated by
2 2 X1 X7 2X1 X2X6 + X2 X5 ; X1 X3X7 X2X3 X6 X1 X4 X6 + X2 X4X5 ; 2 2 2 X3 X7 2X3 X4 X6 + X4 X5; X2 X3 X1X4 ; X6 X5 X7 Remark 6.5. The algorithm presented here is short and easy to implement. The computation of the Grobner basis of I takes a lot of time, so it is recommended to use a fast computer algebra system. The algorithm is implemented in the computer algebra system Singular. The computation of each example took less than a second on a Sparcstation 20.
Chapter II
On the hypersurface 2 2 3 4 x + x y + z + t = 0 in C
The hypersurface Y given by x + x2y + z2 + t3 = 0 has been studied because it is closely related to the linearization conjecture for C ? (see 40]): Conjecture 1.1 (Linearization Conjecture for C ? -actions). Every algebraic C ? -action on C n is a linear action up to conjungation with an automorphism of C n . This conjecture is true for n = 2 (see for example 5] and 6]) and open for n 3. However, Kaliman, Koras, Makar-Limanov and Russell are close to proving that C ? -actions on C 3 are linearizable. In the attempt to prove this conjecture, they encountered examples of \exotic algebraic structures on C 3 ". These examples are smooth contractible hypersurfaces in C 4 with a \hardcase" C ? -action. The hypersurfaces are di eomorphic to R6 by a result of Dimca (see 12]). If such a hypersurface is algebraically isomorphic to C 3 , then the C ? -action cannot be linearizable. The rst examples were found by Libgober in a di erent context (cf. 45]). His list was generalized by Dimca ( 12]), Kaliman ( 33]) and completed by Koras and Russell ( 39]). In order to prove the C ? -linearization conjecture on C 3 , one has to prove that these examples are not isomorphic to C 3 . The simplest example of the list of Koras and Russell is the hypersurface Y : x + x2 y + z 2 + t3 = 0 in C 4 . The C ? -action is given by (x; y; z; t) = ( 6 x; 6 y; 3 z; 2 t): If the hypersurface Y would be isomorphic to C 3 then this C ? -action would not be linearizable, and moreover, the function f = x + x2 y + z2 + t3 on C 4 would be a counterexample to the Sathaye Conjecture (cf. 57]): Conjecture 1.2 (Sathaye Conjecture for embeddings of C n 1 into C n ). If f 1 (0) is isomorphic to C n 1 , for a polynomial map f : C n ! C , then any ber of f is isomorphic to C n 1 . However Makar-Limanov showed that Proposition 1.3. The hypersurface Y : x + x2 y + z2 + t3 = 0 is not isomorphic to C 3. 15
1. Introduction
16
II. On the hypersurface x + x2 y + z2 + t3 = 0 in C 4
Let R be a commutative C -algebra. For a derivation D : R ! R we will de ne RD as the subring of constants ff 2 R j Df = 0g. Makar-Limanov de ned the following invariant of rings: \ ML(R) = RD where D runs trough all locally nilpotent C -derivations R ! R (see section 2 for the de nition and properties of a locally nilpotent derivation). Then ML(C x; y; z]) = C and Makar-Limanov proved that ML C x; y; z; t]=(x + x2 y + z2 + t3) 6= C : It follows that the rings C x; y; z] and C x; y; z; t]=(x + x2 y + z2 + t3) are not isomorphic, or equivalently, Y is not isomorphic to C 3 . In this chapter, we will give an other proof of the fact that Y is not isomorphic to 3 . We will de ne a new invariant C (R) as the algebra generated by all RD where C D runs through all non-trivial locally nilpotent derivations. Then we will prove that C (C x; y; z ]) = C x; y; z] and
D
C C x; y; z; t]=(x + x2 y + z2 + t3) 6= C x; y; z; t]=(x + x2 y + z2 + t3):

And it follows that Y is not isomorphic to C 3 .
2. Technical preparation
In this section rings are always commutative C -algebras. Filtrations. Suppose that RT a ring with an ascending ltration by subvector spaces is S R = 1 1 F iR satisfying 1 1 F iR = f0g. The associated graded ring will be i= i= denoted by 1 M i i1 Gr R = F R=F R: There is a canonical map gr : R ! Gr R, de ned by gr(f ) = f + F i 1 R 2 F i R=F i 1R for f 2 F i R n F i 1R and gr(0) = 0. Observe that if f 6= 0, then gr(f ) 6= 0. Degree functions. A degree function on a ring R is a map deg : R ! Z f 1g satisfying 1. deg(f ) = 1 if and only if f = 0, 2. deg(fg) = deg(f ) + deg(g) for all f; g 2 R, 3. deg(f + g) maxfdeg(f ); deg(g)g. A degree function induces a natural ltration, namely
i= 1
F iR := f f 2 R j deg(f ) i g
a 1{1 correspondence between locally nilpotent derivations and regular Ga-actions on Spec R, where Ga is the additive group (C ; +). A Ga-action on Spec R corresponds to an algebra homomorphism ' : R ! O(Ga ) R = R t] satisfying certain properties. If we de ne D : R ! R by d Df = dt '(f ) t=0 then D is a locally nilpotent derivation on R. On the other hand, if D is a locally nilpotent C -derivation then 1 X tiDi f f 7! exp(tD)f = i=0 i! de nes an algebra homomorphism R ! R t] which corresponds to a morphism Ga Spec R ! Spec R which is in fact a Ga -action. If R is a domain and the Ga-action is non-trivial, or equivalently D 6= 0, then it is well-known that dim RD = dim RGa = dim R 1. Lemma 2.1. Suppose that R is a domain and that D is a locally nilpotent derivation on R. If f; g 2 R n f0g and D(fg) = 0, then Df = Dg = 0. It follows that D(K ) = f0g for any sub eld K R. Proof. If D(fg ) = 0 then fg = exp(tD)(fg ) = (exp(tD)f )(exp(tD)g ). Now exp(tD)f and exp(tD)g must be constant polynomials in t. So we have Df = Dg = 0. For any c 2 C n f0g we have Dc = 0 because D(c 1 ) = D1 = 0. c Lemma 2.2. If Ga acts non-trivial on a connected a ne variety Z of dimension 1, then Z must be isomorphic to C . Proof. Because Z has a non-trivial orbit, it must contain C as a subvariety. If Z is a ne then we must have Z = C . Inheritance of locally nilpotent derivations. Suppose that R = S1 1 F i R is a i= T ring with an ascending ltration such that 1 1 F iR = f0g and D is a (non-trivial) i= locally nilpotent derivation on R. Suppose moreover, that DF i R F i+k R for some integer k. This condition is satis ed if Gr R is nitely generated. We may assume that k is as small as possible. Now we de ne gr D : Gr R ! Gr R by (gr D) gr f = Df + F i+k 1R 2 F i+k R=F i+k 1 R for f 2 F i R n F i 1 R: It is easy to check that gr D is a locally nilpotent derivation on Gr R. De nition. We de ne the ring invariant C (R) as the algebra generated by all RD , with D a non-trivial locally nilpotent derivation. @ @ @ For example, if R = C x; y; z ] then the kernels of @x , @y and @z are C y; z ], C x; z] and C x; y ] respectively. Therefore we get C (R) = C x; y; z ] = R.
Locally nilpotent derivations. A C -derivation D on R is called locally nilpotent if for every f 2 R there exists a natural number N such that DN f = 0. There is
T and we have 1 1 F iR = f0g. i=
2. Technical preparation
17
18

3. The proof that x + x2 y + z 2 + t3 = 0 is not isomorphic to C 3
Take R1 = C x; y; z; t]=(x + x2 y + z2 + t3). We can view R1 as a subring of C z; t; x; x 1 ] if we identify y with x 2(x + z2 + t3). We de ne a degree function on R1 by deg1(f ) = vx(f ) for all f 2 R1 where vx is the valuation in x on the ring C z; t; x; x 1 ]. For example deg1(z) = 0, deg1 (t) = 0, deg1(x) = 1 and deg1 (y) = 2. Proposition 3.1. Let D be a non-trivial locally nilpotent derivation on R1 and f 2 R1 a non-zero element. If Df = 0, then deg1(f ) 0. Proof of Proposition 1.3. The ring R1 is not isomorphic to C x; y; z ] because C (R1) 6= R1 by Proposition 3.1. Proof of Proposition 3.1. Let us assume that there is an element f1 2 R1 with deg1 f1 > 0 and a locally nilpotent derivation D1 : R1 ! R1 such that D1f1 = 0. The degree function deg1 de nes a ltration on R1, and we put R2 := Gr R1, f2 := gr f1 and D2 = gr D1. Observe that D2 f2 = gr(D1 f1) = 0. Lemma 3.2. If we just write x; y; z; t instead of gr x; gr y; gr z; gr t, then we get
R2 = C x; y; z; t]=(x2 y + z2 + t3):
Proof of the lemma. Using the relation x2 y = x
f 2 R1 in the form
z2 t3 we can write every
f = a(x; z; t) + yb(y; z; t) + xyc(y; z; t)
for unique polynomials a; b; c
For gr f there are 3 possibilities: 1. gr f = (gr x)r p(gr z; gr t) for some r 0, p a polynomial, 2. gr f = (gr y)r p(gr z; gr t) for some r > 0, p a polynomial, 3. gr f = gr x(gr y)r p(gr z; gr t) for some r > 0, p a polynomial. So gr x; gr y; gr z; gr t are generators of the ring R2. Because
x2 y + z2 + t3 = x 2 F 1 R1
we get (gr x)2 gr y + (gr z)2 + (gr t)3 = 0 More relations between gr x, gr y, gr z and gr t would imply that dim R2 2, because x2 y + z2 + t3 is an irreducible polynomial. But we have dim R2 3, because gr x, gr z and gr t are algebraically independent. So we can conclude
R2 = C x; y; z; t]=(x2 y + z2 + t3):
3. The proof that x + x2 y + z2 + t3 = 0 is not isomorphic to C 3
19
Because deg1 f1 > 0 there are only two possibilities for f2 = gr f1 6= 0: 1. f2 = xyr p(z; t) for some r > 0, p a polynomial, 2. f2 = yr p(z; t) for some r > 0, p a polynomial. case 1: f2 = xyr p(z; t). Because D2 f2 = 0, we have D2 x = D2y = 0 (see Lemma 2.1). We have a morphism: : Spec R2 = f (x; y; z; t) 2 C 4 j x2 y + z2 + t3 = 0 g ! C 2 (x; y; z; t) = (x; y) We look at the Ga-action corresponding to D2 . Now x and y are Ga-invariant functions, therefore Ga stabilizes the bers of . The ber of ( ; ) 2 C 2 is
1(
de ned by
; ) = f ( ; ; z; t) j z2 + t3 +
= 0 g:
For ; 6= 0 this is an elliptic curve, hence not isomorphic to C . So the action on these bers must be trivial by Lemma 2.2. Because these bers form a dense subset, the Ga-action must be trivial, or equivalently D2 = 0. This is a contradiction. So we must have: case 2: f2 = yr p(z; t). Suppose that D is constant. Because dim R2 = 3, we have p D = 2. So there is a g 2 R which is algebraically independent of f2 . dim R2 2 We may assume that g is homogeneous with respect to deg2. For some k we have k k deg2 (f2 g) = k deg2 (f2 ) + deg2(g) > 0. If 0we replace f2 by f2 g then we still have f2 6= 0, D2f2 = 0 and f2 is of the form yr p0 (z; t). But now p0 (z; t) is not constant anymore. So without loss of generality we may assume that p is not constant. We de ne a new degree function deg2 on R2. We put deg2(x) = 0, deg2(z) = 3 and deg2(t) = 2 (deg2 (y) = 6). We let R3 = Gr R2, f3 = gr f2 and D3 = gr D2 . We just write x; y; z; t instead of gr x; gr y; gr z; gr t. Now
2 2
R3 = C x; y; z; t]=(x2 y + z2 + t3)
because deg2 does not only de ne a ltration, but even grading on R2, and so R2 and R3 are canonically isomorphic. Now f3 is of the form
w Y
f3 = yr gr p = yr zutv
i=1
( i z2 + i t3)
r > 0, u; v; w 0 and i ; i 2 C n f0g
Because p is not constant we have either u > 0, v > 0 or w > 0. case 2a: u > 0. Now D3 y = D3z = 0. The general ber of the map (x; y; z; t) 2 f (x; y; z; t) j x2 y + z2 + t3 = 0g ! (y; z) is again an elliptic curve. We get a similar contradiction as in case 1. case 2b: v > 0. Now D3y = D3 t = 0. The general ber of the map (x; y; z; t) 2 f (x; y; z; t) j x2 y + z2 + t3 = 0g ! (y; t)
20
is isomorphic to C x; z]=(x2 + z2 1) which doesn't allow a non-trivial Ga-action. Contradiction. case 2c: w > 0. Now D3 y = D3 (z2 + t3) = 0 ( = 1 = 1). If = 1, then D2 (x2 y) = 0, so D2 x = D2y = 0 and we get a contradiction like in case 1. So assume 6= 1. Let Z ; be the ber 1 ( ; ) where is de ned by : (x; y; z; t) 2 f (x; y; z; t) j x2 y + z2 + t3 = 0g ! (y; z2 + t3) The functions x; t de ne a map Z ; ! C 2 . The image is one-dimensional and contained in f (x; t) j x2 + (1 )t3 + = 0 g which is an elliptic curve for ; 6= 0. For ; 6= 0, Z ; allows a dominant map to an elliptic curve, so it cannot be rational. But then Z ; doesn't allow a Ga-action for ; 6= 0 which yields a contradiction. So the assumption in the beginning that deg(f1 ) > 0 leads to a contradiction and we conclude deg(f1 ) 0.
Chapter III
Nonlinearizable Holomorphic Group Actions
1. Introduction
The aim of this chapter is to give a negative solution to the following long standing question. Holomorphic Linearization Problem. Let G ,! Authol (C n ) be a complex reductive subgroup of the group Authol (C n ) of holomorphic automorphisms of C n . Can one conjugate this subgroup by a single automorphism into the general linear group GLn(C ) Authol (C n ), i.e., is every action of a complex reductive group on C n linearizable? The complex algebraic analogue has been studied quite well. For an overview on this subject see the article by Kraft ( 42]). The rst non-linearizable complex algebraic actions were found by Schwarz in 1989 (see 60]), using non-trivial algebraic G-vector bundles over representations. The problem is still open for abelian groups. For the case C ? , see the survey by Kraft ( 40]). Recently Asanuma ( 4]) has constructed non-linearizable multiplicative group actions in the real algebraic category and also in positive characteristic. Concerning holomorphic linearization several positive results are known: First Suzuki proved that holomorphic C -actions on C 2 are linearizable (see 64]). His result was generalized by Jiang (see 29]) who proved using the methods of Kraft and Schwarz (see 44]) that actions of reductive groups on C n with one-dimensional quotient are linearizable. Surprisingly the method of constructing counterexamples to the algebraic linearization problem does not work in the holomorphic setting. It follows from the equivariant Oka-principle proved by Heinzner and Kutzschebauch (see 22]) that holomorphic actions arising from equivariant vector bundles over representation spaces are linearizable. Recently Ahern and Rudin proved linearization for nite cyclic groups acting on C 2 by automorphisms lying in a dense subgroup of Authol (C 2 ) the group of overshears (see 2]). This result was generalized by Kraft and Kutzschebauch to arbitrary complex reductive groups (see 43]). In this chapter we construct non-linearizable holomorphic actions on C n for all reductive groups. Our main result is (see Theorem 4.10) 21
22
III. Nonlinearizable Holomorphic Group Actions
exists a natural number NG such that for all l NG there exists an e ective nonlinearizable holomorphic action of G on C l . Our construction is an adaptation of a method of Asanuma to the holomorphic setting. In 4] he uses exotic algebraic embeddings of R in R3 to construct a non-linearizable algebraic R?-action on R5. For a eld K of positive characteristic, he constructs a non-linearizable K ? action on K 4 using an exotic embedding K ,! K 2. Our construction depends on the existence of non-recti able holomorphic embeddings of C into C n for n 2. We collected the facts we need about these embeddings for our construction in the second paragraph. In the third paragraph we de ne the Rees spaces, a holomorphic analogue of the Rees algebra, and prove some of their properties. The main idea of the construction is explained in the rst part of paragraph 4, where we give counterexamples for actions of C . In the remaining part of the chapter we give similar constructions and use some representation theory to obtain counterexamples for all reductive groups. I wouldn't have found the results of this chapter without the many discussions with Frank Kutzschebauch, who explained the theory of non-recti able embeddings to me. He also helped me to carry out the ideas of Asanuma in the holomorphic category to obtain non-linearizable actions for C ? . The generalization to arbitrary groups is based on my ideas.
Theorem. For every complex reductive Lie group G (except the trivial group) there
In this section we explain how one constructs holomorphic embeddings of C k into C n (0 < k < n) which are not equivalent to the standard embedding. Such embeddings were found by Forstneric, Globevnik and Rosay in 16]. Since we need some special version of their result (see Corollary 2.4) we recall the main ideas. A holomorphic map F : C n ! C n is called nondegenerate if JF (p) 6= 0 for some p 2 C n , where JF is the determinant of the Jacobi matrix. Theorem 2.1 (Rudin, Rosay). If n > 1 then there exists a discrete subset E C n such that all nondegenerate F : C n ! C n satisfy F (C n ) \ E 6= ; (cf. 56]). Such discrete subsets E C n are called unavoidable. Theorem 2.2 (Forstneric, Globevnik, Rosay). For any discrete subset E C 2 there exists a proper holomorphic embedding ' : C ,! C 2 such that '(C ) E (cf. 16]). De nition. A proper holomorphic embedding ' : C k ! C n is called recti able if there exists a 2 Authol (C n ) such that
2. Non-rectifiable embeddings
'(z1; : : : ; zk ) = (z1 ; : : : ; zk ; 0; : : : ; 0):

Remark 2.3. Every automorphism of C k
f0g C n extends to an automorphism of n . It follows that a proper holomorphic embedding ' : C k ! C n is recti able if and C only if there exists a 2 Authol (C n ) such that ('(C k )) = C k f0g C n .
Take an unavoidable subset E C 2 and let ' : C ! C 2 be a proper holomorphic embedding such that '(C ) E . De ne 'k : C C k ! C 2 C k by 'k (z; y) = ('(z); y).
3. Rees spaces
23
Ck
is not biholomorphic to C 2+k . So the embedding 'k is not recti able. Proof. Suppose : C k+2 ! (C 2 n '(C )) C k is the universal covering and let p : (C 2 n'(C )) C k ! C 2 n'(C ) be the projection onto the rst factor. The composition p : C k+2 ! C 2 n '(C ) is a submersion. The restriction of p to a generic linear two-dimensional subspace is nondegenerate. This contradicts the unavoidability of E '(C ). The following lemma shows that every embedding of C k into C n whose image lies in a linear subspace of codimension at least k is recti able, so adding some additional dimensions to C n one can rectify every embedding inside the higher-dimensional space. We need this fact at the end of paragraph 3 to show that the product of some Rees spaces with C l (l 0) are a ne spaces. Lemma 2.5. Let X be a Stein space. If '1 : X ! C n and '2 : X ! C m are two proper holomorphic embeddings, then there exists a 2 Authol (C n C m ) such that ('1 0) = 0 '2. (compare with Lemma 3.2 in 4]). Proof. By a Corollary of Cartan's Theorem B the holomorphic map '1 : X ! C n can be extended to a holomorphic map 1 from C m '2(X ) to C n (so 1 '2 = '1). Likewise there exists a holomorphic map 2 : C n ! C m such that 2 '1 = '2. De ne 1 ; 2 2 Authol (C n C m ) by 1 (x; y ) = (x; y + 2 (x)) and 2 (x; y ) = (x + 1 (y ); y ). Now we have 1 ('1 0) = '1 '2 = 2 (0 '2 ): So take = 2 1 1.
Corollary 2.4. The universal covering of C 2+k n 'k (C 1+k ) = (C 2 n '(C ))
In this section we introduce the Rees spaces. Important for the next paragraph are especially Example 3.5 and Corollary 3.7. Let X , Y and Z be Stein spaces such that X Y and let f 2 O(Z ) be a holomorphic function on Z . Suppose that there are nitely many holomorphic functions h1; h2; : : : ; hr on Y generating the ideal IX (Y ) of X . This is for instance true if X and Y are manifolds (see 15], Satz 5.5). Consider the map: : Y (Z n ff = 0g) ! Y Z C r (y; z) = y; z; h1((zy)) ; h2((zy)) ; : : : ; hr((zy)) : f f f of the image of within Y Z C r . Remark 3.1. The holomorphic closure of Im( ) within Y Z C r equals the closure in the C -topology. Remark 3.2. The de nition of the Rees space doesn't depend on the choice of generators h1; h2; : : : ; hr . Suppose that g1; g2 ; : : : ; gs also generate IX (Y ), and let 0 be the corresponding map Y (Z n ff = 0g) ! Y Z C s . It is easy to see that there exist holomorphic maps ' : Y Z C r ! Y Z C s and '0 : Y Z C s ! Y Z C r such
3. Rees spaces
De nition. We de ne the Rees space R(X; Y; Z; f ) as Im( ), the holomorphic closure
24
=
that 0 = ' and = '0 0 . Now '0 ' = id on Im( ), therefore '0 ' = id on Im( ) and likewise ' '0 = id on Im( 0 ). So the restriction of ' de nes a biholomorphic map Im( ) ! Im( 0 ). Remark 3.3. If we take Z := C and f := z , then the Rees space is just the holomorphic analogue of the Rees algebra. We will use the symbol = to denote biholomorphic. Lemma 3.4. Rees spaces have the following properties:
1. R(X; Y; Z; f ) = R('(X ); Y; Z; f ) for ' 2 Authol (Y ), 2. R(X W; Y W; Z; f ) = R(X; Y; Z; f ) W for any Stein space W , 3. R(X f0g; Y C m ; Z; f ) = R(X; Y; Z; f ) C m . Proof. The rst two properties are trivial. We show the third property. Suppose IX (Y ) is generated by h1; h2; : : : ; hr 2 O(Y ). Then IX f0g (Y C m ) is generated by h1; : : : ; hr ; u1; : : : ; um where the ui are the coordinate functions of C m . Consider the map : Y C m (Z n ff = 0g) ! Y C m Z C r C m (y; u; z) = (y; u; z; f (z) 1 h; f (z) 1 u); where h = (h1; : : : ; hr ) and u = (u1; : : : ; um). De ne ' 2 Authol (Y C m Z C r C m ) by '(y; u; z; v; w) = (y; u f (z):w; z; v; w). Now we have ' (y; u; z) = (y; 0; z; f (z) 1 :h; f (z) 1 :u); and we get R(X f0g; Y C m ; Z; f ) = Im( ) = Im(' ) = R(X; Y; Z; f ) C m :
Example 3.5. Suppose X
R(X; Y; Z; f ) = Im( ) = f(y; z; w) 2 Y Z C j f (z )w = h(y )g: Lemma 3.6. If '1 : X ! C n and '2 : X ! C m are proper holomorphic embeddings,
Let : Y get
then
Y is given by h = 0 with h 2 O(Y ) (IX (Y ) = O(Y )h). (Z n ff = 0g) ! Y Z C given by (y; z) = (y; z; h(y)=f (z)). Then we
(compare with 4], Corollary 4.2). Proof. From Lemma 2.5 and Lemma 3.4 it follows that R('1 (X ); C n ; Z; f ) C m = R('1 (X ) f0g; C n C m ; Z; f ) = = R(f0g '2(X ); C n C m ; Z; f ) = R('2 (X ); C m ; Z; f ) C n :
R('1 (X ); C n ; Z; f )
Cm
= R('2 (X ); C m ; Z; f )
Cn
Corollary 3.7. For an embedding ' : C k ! C n and f 2 O(C l ) we have

R('(C k ); C n ; C l ; f )
Ck
= R(C k ; C k ; C l ; f )
Cn
= C k+l+n :
4. Non-linearizable group actions

25
The goal of this section is to nd for every non-trivial reductive group a non-linearizable action on C n for some n. As usually we denote by X G the xed point set of an action of G on X . Several times we will use the notion of categorical quotient X : X ! X= for =G a holomorphic action of a reductive group G on a Stein space X . Recall that this is the quotient of X with respect to the equivalence relation which is given by the invariant holomorphic functions and that X= carries a complex structure such that X is a =G holomorphic map. Since G-invariant holomorphic functions on a closed G-invariant subspace Y of a holomorphic Stein G-space X extend to G-invariant holomorphic functions on X , the categorical quotient Y ! Y= is just given by restricting X to =G Y . Beside some easy examples of quotients for linear actions this is the only fact we use. So for further details on categorical quotients we refer to the papers of Snow ( 63]) and Heinzner ( 21]). Let ' = '0 : C ! C 2 be a proper holomorphic embedding like in Corollary 2.4. By the solution of the second Cousin problem we nd a function h 2 O(C 2 ) which vanishes precisely on '(C ) of order 1.
Non-linearizable C ? -actions Consider the Rees space X = R('(C ); C 2 ; C ; z). Like in Example 3.5 we have
X = f(x; y; z; w) 2 C 4 j h(x; y) = zwg:

By Corollary 3.7 there exists a biholomorphism k : X C k ! C k+3 for all k 1. Now X C k C 4 C k is stable under the linear C ? -action on C 4 C k given by
=
C?
C4
Ck
! C4
Ck
(x; y; z; w; t1 ; : : : ; tk ) = (x; y; z; 1 w; t1 ; : : : ; tk ): The restriction of this action to X C k induces via k a holomorphic C ? -action k : C ? C k+3 ! C k+3 . Proposition 4.1. The action k is not linearizable (k 1). So for all l 4 there exists a non-linearizable C ? action on C l . Proof. Suppose k : X C k ! C 3+k is a biholomorphic C ? -equivariant map, where C ? acts linearly on C 3+k . This representation of C ? on C k+3 must be isomorphic to the representation of C ? on the tangent space of some xed point of X C k . With respect to some coordinates, this action is given by
=
(z; w; u1 ; : : : ; uk+1) = ( z; The categorical quotient of X

Ck
1 w; u1 ; : : : ; uk+1 ): 1
is given by
:X
2
Ck
! C 2+k ,
1 (x; y; z; w; t1 ; : : : ; tk ) = (x; y; t1 ; : : : ; tk )
and the categorical quotient of C 3+k is given by
: C 3+k ! C 2+k ,
2 (z; w; u1 ; : : : ; uk+1 ) = (zw; u1 ; : : : ; uk+1 ):
26
The xed point set (X C k )C is f(x; y; z; w; t1 ; : : : ; tk ) 2 X C k j h(x; y) = z = w = 0g: Its image under 1 is '(C ) C k C 2 C k . On the other hand (C 3+k )C is f(z; w; u1 ; : : : ; uk+1) 2 C 3+k j z = w = 0g and its image under 2 is f0g C 1+k C 2+k . Now k : X C k ! C 3+k induces a biholomorphism : C 2+k ! C 2+k such that ('(C ) C k ) = f0g C k+1 . This contradicts Corollary 2.4.
= =
Suppose G = Z=nZ (n 2). De ne a holomorphic function f 2 O(C n 1 ) by f (z1 ; z2; : : : ; zn 1) = z1 z2 zn 1. Consider the Rees space X := R('(C ); C 2 ; C n 1 ; f ) which is given by (see Example 3.5) X = f(x; y; z1 ; z2 ; : : : ; zn) 2 C 2+n j h(x; y) = z1 z2 zng: From Corollary 3.7 we get X C k = C 1+n+k for k 1. Let G act on C 2 C n by g (x; y; z1 ; z2; : : : ; zn) = (x; y; z2 ; z3 ; : : : ; zn ; z1) where g is a generator of G. This induces by restriction an action of G on X . If we let G act trivially on C k then we get an action of G on C 1+n+k = X C k (k 1) which we denote by k . The xed point set of G on X C k is biholomorphic to Yn C k where Yn = f(x; y; z) j h(x; y) = zng: We de ne a G action on Yn by g (x; y; z) = (x; y; z) where is a primitive n-th root of unity. If we let G act trivially on C k then we get an action of G on Yn C k which we will denote by k . Proposition 4.2. For k 1 at least one of the following statements is true: 1. The action k on C n+k+1 is not linearizable. 2. Yn C l = C 2+l and the action l of G on Yn C l is not linearizable for all l k. Proof. If the action k is linearizable, then its xed point set Yn C k must be a vector space, so Yn C l = C 2+l for all l k. The xed point set of G on Yn C l is isomorphic to W C l where W is given by W = f(x; y; z) 2 C 3 j h(x; y) = z = 0g If l is linearizable, then Yn C l n W C l = (Yn n W ) C l is biholomorphic to C ? C l+1 . But this contradicts Lemma 4.3 below. Lemma 4.3. For k 0 the space (Yn n W ) C k is not biholomorphic to C ? C 1+k . Proof. The map (x; y; z ) 2 (Yn n W ) ! (x; y ) 2 C 2 n '(C ) is an n-sheeted unrami ed covering of C 2 n '(C ). So (Yn n W ) C k and (C 2 n '(C )) C k have the same universal covering. But by Corollary 2.4 this universal covering is not biholomorphic to C 2+k .
Non-linearizable actions for finite cyclic groups
Corollary 4.4. For l n + 2 there exists a non-linearizable action of Z=nZon C l .
4. Non-linearizable group actions De ne

Nonlinearizable actions for simple non-abelian groups
27
Xn := f(x; y; z1 ; : : : ; zn ; w1; : : : ; wn) 2 C 2n+2 j h(x; y) = z1 w1 + : : : + znwng:

is biholomorphic to C 3 2n 2 . Proof. We prove the lemma by induction. For n = 0 this is trivial. Suppose that Xn C 3 2n 2n 3 = C 3 2n 2. Observe that Xn+1 = R(Xn ; C 2n+2 ; C ; z) (see Example 3.5). By Corollary 3.7 we have
Lemma 4.5. The space Xn
C3 2
n 2n 3
Xn+1 R(Xn
C3 2
n+1 2(n+1) 3 n n = R(Xn ; C 2n+2 ; C ; z) C 3 2 2n 3 C 3 2 2 = n+1 2 n n n : C 3 2 2n 3 ; C 2n+2 C 3 2 2n 3 ; C ; z ) C 3 2 2 = C 3 2
Suppose that there exists a subgroup H of G, an n-dimensional representation V of G and a homogeneous G-invariant polynomial a on V of degree 2 such that: V G = f0g, dim V H = 1 and (*) H. a is not identical to 0 on V Then we de ne U1 := f(x; y; v) 2 C 2 V j h(x; y) = a(v)g U2 := f(x; y; v; w) 2 C 2 V V j h(x; y) = a(v) + a(w)g: The spaces U1 and U2 carry G-actions de ned by
g (x; y; v) = (x; y; g v) (x; y; v) 2 U1 C 2 V; g 2 G g (x; y; v; w) = (x; y; v; g w) (x; y; v; w) 2 U2 C 2 V V; g 2 G:
Proposition 4.6. For k 3 2n 2n 3 we have U2

one of the following statements is true:
Ck
= C k+2n+1 and at least

Cl
1. The action of G on U2 C k is not linearizable. 2. U1 C l = C l+n+1 for l k and the action of G on U1 Proof. After a change of coordinates we have
2 2 2 a(z) = z1 + z2 + : : : + zr
is not linearizable.
for z = (z1 ; z2 ; : : : ; zn ) with respect to some coordinates on V . So

2 2 2 2 2 2 a(z) + a(w) = z1 + z2 + : : : + zr + w1 + w2 + : : : + wr
where w = (w1 ; w2; : : : ; wn). Again after a linear change of coordinates in V get 0 0 0 0 0 0 a(z) + a(w) = z1w1 + z2 w2 + : : : + zr wr :
V we
28
So U2 = Xr C 2n 2r and it follows from Corollary 3.7 that U2 C k = C k+2n+1 for all k 3 2n 2n 3. The xed point set of G on U2 C k is U1 C k . So if the action of G on U2 C k is linearizable then U1 C l = C l+n+1 for all l k. Take p 2 V such that V H = C p. Then we have (U1
C l )H
= f(x; y; z) 2 C 2
p j h(x; y) = a(z)g
Cl
Cl
Because a(p) 6= 0, this is biholomorphic to
f(x; y; z) 2 C 3 j h(x; y) = z2 g
On the other hand (U1
C l )G
= Y2 =W
Cl
: :
is the subset given by

Cl Cl
f(x; y; z) 2 C 3 j h(x; y) = z = 0g
The space (U1 C l )H n (U1 C l )G = (Y2 n W ) C l is not biholomorphic to C ? C l+1 by Lemma 4.3. Therefore the G-action on U1 C l cannot be linearizable. In the remaining part of this paragraph we show in a case by case study that for all simple non-abelian groups the setting (*) can be achieved thus by Proposition 4.6 constructing non-linearizable actions. First suppose G is a positive dimensional simple Lie group. Let V := g be the Lie algebra of G. Consider the adjoint action of G on g. Fix a maximal torus T G. The xed point set of T is t , the Lie algebra of T . Now the Weyl group W = NG (T )=T acts on t. Suppose that s1 ; s2 ; : : : ; sr are the simple generating re ections of W corresponding with the simple roots 1; : : : ; r . The Killing form h ; i given by (a; b) := Trace(ad(a)ad(b)) is a G-invariant bilinear form on g. The restriction to t is nondegenerate, so we can identify t with its dual. Let W0 be the subgroup of W generated by s1; s2 ; : : : ; sr 1 . The xed point set of W 0 is one dimensional and spanned by the fundamental weight r . Let H be the group containing T such that H=T = W0 . If we de ne a(v) := (v; v) then the restriction of a to V H is not identically 0 because a( r ) = ( r ; r ) > 0. So by Proposition 4.6 we obtain: Corollary 4.7. If G is a positive dimensional simple Lie group then for all l 3 2dim G 2 there exist a non-linearizable action of G on C l . Second we consider the case where G is a non-abelian nite simple group. The proof of the following lemma was shown to us by Jurgen Muller: Lemma 4.8. For any non-abelian nite simple group G, there exist a subgroup H and a real representation VR such that VRG = f0g and VRH is one-dimensional. Proof. We use the classi cation of nite simple groups. Alternating groups: For G = An (n 5) , take
VR = f(x1 ; x2 ; : : : ; xn ) 2 Rn j x1 + x2 + : : : + xn = 0g
and let G act on it by permuting the coordinates. Take H = An 1, the alternating subgroup which permutes only the rst n 1 coordinates. It is clear that VRH is one-dimensional, spanned by the vector (1; 1; : : : ; 1; 1 n).
29
For the groups of Lie type and the sporadic ones we will look for a non-trivial irreducible real character such that h ; (1H )G i = 1 , where (1H )G is the character of G induced by the trivial character 1H of H . Now (1H )G is the character of the real representation R G=H ]. Let VR be the isotypic component of in R G=H ]. The multiplicity of in VR is h ; (1H )G i = 1. So VR is a real representation with character . The dimension of VRH is h H ; 1H i which is equal to h ; (1H )G i = 1 by Frobenius reciprocity. Finite simple groups of Lie type: Take H := B , the Borel subgroup. Let I be the set of simple re ections. For every subset J I we have a parabolic subgroup PJ . Let (1PJ )G be the character of G induced by the trivial character 1PJ of PJ . De ne a function St on G by X St = ( 1)jJ j (1PJ )G
J I
This turns out to be an irreducible character and it has the following properties (see Chapter 6 of 8]): hSt; (1B )G i = 1, St(1) = jU j where U is the maximal unipotent subgroup. Now St is clearly a real character, and it belongs to a real representation VR of dimension jU j satisfying dim VRH = 1. P Sporadic groups: If (1H )G = r=1 ai i where a1; a2 ; : : : ; ar 2 Z and 1; 2 ; : : : ; r i di erent irreducible characters, then ai = h i ; (1H )Gi. For most of the sporadic groups and some of its large subgroups the decomposition of (1H )G can be found in the Atlas of Finite Groups (see 9]). The table below shows the possibilities for G, H and as they can be found in the Atlas. The notation of the Atlas is used. For example, if G = J3 (the third Janko group) then 323a stands for the rst 323-dimensional irreducible representation of J3 as listed in the character table of J3 in the Atlas.
30 group G M11 M12 M22 M23 M24 J1 J2 J3 Suz HS McL Co3 Co2 He Fi22 Fi23 Fi024 O'N Ly Ru
III. Nonlinearizable Holomorphic Group Actions subgroup H M10 M11 L3(4) M22 M23 L2(11) U3(3) L2(16) : 2 G2(4) M22 U4(3) McL : 2 U6(2) : 2 S4(4) : 2 2 U6(2) 2 Fi22 Fi23 L3(7) : 2 G2(5) 2 F4 (2) character 10a 11a 21a 22a 23a 56a 36a 323a 780a 22a 22a 275a 275a 680a 429a 782a 57477a 10944a 45694a 783a
For the other sporadic groups the computer algebra system GAP can be used to nd: group G subgroup H character J4 211 : M24 889111a Co1 Co2 299a HN A12 133a 3 D4 (2) : 3 Th 4123a B Th 96255a M 2 B 196883a For all sporadic groups we took for H the maximal subgroup of G, except in the cases G = B and G = M where the maximal subgroups aren't classi ed yet. Now we realize the setting (*). Take H and VR as in Lemma 4.8. Put V := VR R C . Let p 2 VRH n f0g. Choose a real linear function b 2 VR? such that b(p) 6= 0. De ne P the G-invariant polynomial a = g2G (g b)2 on V . Now V G = f0g, V H = C p and a(p) > 0. From Proposition 4.6 we obtain: Corollary 4.9. If G is a nite simple group, then there exists a non-linearizable action of G on C l for all l 3 2d 2 where d = n 1 if G = An , d = St(1) = jU j if G is of Lie type, d is like in the tables of the proof of Lemma 4.8 if G is sporadic.

Effective non-linearizable actions for all reductive groups
31
there exists a natural number NG such that for all l NG there exists an e ective non-linearizable holomorphic action of G on C l . Proof. Let H be the maximal closed normal subgroup of G. For all l NG=H there exists an e ective non-linearizable action of G=H on C l by Proposition 4.1, Corollary 4.4, Corollary 4.7 and Corollary 4.9. Take a G-representation V on which H acts e ectively. Let W = V=V H . Now H acts on W e ectively and W H = f0g. The action of G on C l W is e ective. If this action were linearizable, then the action of G on (C l W )H = C l f0g would be linearizable, but this is not the case.
Theorem 4.10. For every complex reductive Lie group G (except the trivial group)
Glossary of Notations
k G
O(V ) O(V )G
char(k) jGj G (V ) Vd T X ? (T ) E 1; : : : ; n vol(CV ) m r NV (V )

CV
k k
m IN
LV V= =G
N R
B B
RY
p1
?
p p1
base eld, 1 linear algebraic group, 1 representation of G, 1 n-dimensional G-module, 1 coordinate ring of V , 1 invariant ring, 1 characteristic of the eld k, 1 group order, 1 smallest upper bound for generating invariants, 2 binary forms of degree d, 2 Torus, 3 weights of T , 3 X ? (T ) Z R, 3 weights appearing in V , 3 convex hull of 1; : : : ; n, 3 volume of CV , 3 dimension of G, 3 rank of T , 3 nilpotency degree, 3 bound for primary invariants, 3 norm on E , 3 maxf k 1 k; k 2k; : : : ; k nk g, 3 categorical quotient, 4 maximal homogeneous ideal, 4 zero- ber ideal, 4 nullcone, 4 Reynolds operator, 4 map G V ! V V de ned by (g; v) 7! (v; g:v), 5 image of , 5 Zariski closure of B, 5 ideal in O(V ) of B V V , 5 Reynolds operator k X; Y ] ! k X; Y ]G, 5 diagonal map V ! V V , 5 algebra homomorphism k X; Y ] ! k X ] corresponding to , 5 projection V V ! V onto the rst factor, 6 homogeneous prime ideal, 6 linear part of p, 6 32
Chapter I
Glossary of Notations
P(
33
V V) 1 ; : : : ; (m
r)=2
e1 ; : : : ; er (; ) ^ G (V ) Pin V G IG I
(V
V)
Gm
C?
projective space of V V , 7 positive roots, 7 Weyl group, 7 Coxeter exponents, 7 W-invariant scalar product, 7 dual element of 2 E , 7 number of g 2 G such that (g) lies in a given generic m-dimensional subspace of End(V ), 7 polynomial on G (V ? V ? )n+m 1 , 8 n-th antisymmetric power of V V , 9 connected component of id 2 G, 10 ideal de ning the group G as an a ne variety, 11 graph of the action, 11 ideal de ning , 11 multiplicative group, 12 Filtration of a ring R, 16 associated graded ring, 16 map R ! Gr R, 16 degree function, 16 locally nilpotent derivation, 17 additive group (C ; +), 17 transcendence degree of the quotient eld of R over C , 17 an invariant of rings, 17 valuation, 18 group of holomorphic automorphisms of C n , 21 complex reductive group, 22 discrete subset C n , 22 non-recti able proper holomorphic embedding C ! C 2 , 22 non recti able proper holomorphic embedding C k+1 ! C k+2 , 22 holomorphic functions on a Stein space Z , 23 Ideal of all functions in O(Y ) vanishing on X , 23 map needed for de nition of Rees space, 23 Rees space, 23 xed point set, 25 categorical quotient, 25 holomorphic function on C 2 vanishing on '(C ), 25 subgroup of G, 27 representation of G, 27 homogeneous G-invariant on V , 27 Lie algebra of G, 28 maximal torus, 28 Lie algebra of T , 28
F R Gr R gr deg D
dim R C (R) vx Authol (C n ) G E ' 'k O(Z ) IX ( Y )
Ga
Chapter II multiplicative group C n f0g, 15
Chapter III
g t
X= =G h(x; y) H V a
R(X; Y; Z; f ) XG
34
Glossary of Notations simple re ections, 28 simple roots, 28 Killing form, 28 Weyl groups, 28 real representation, 28 Alternating groups, 28 character, 29 scalar product for characters, 29 trivial character of H , 29 character of G induced by 1H , 29 R-space spanned by G=H seen as a G-module, 29 restriction of to H , 29 Borel subgroup, 29 parabolic subgroup, 29 Steinberg character, 29 maximal unipotent subgroup, 29 323 dimensional irreducible representation, 29 Mathieu groups, 30 Janko groups, 30 Suzuki group, 30 Higman-Sims group, 30 McLaughlin group, 30 Conway groups, 30 Held group, 30 Fischer groups, 30 O'Nan group, 30 Lyons group, 30 Rudvalis group, 30 Harada-Norton group, 30 Thompson group, 30 baby monster, 30 monster, 30 for more group notation see the Atlas of nite groups ( 9]), 30
s1 ; : : : ; sr 1; : : : ; r (; ) W; W0 VR An
h; i 1H 1G H R G=H ]
B PJ St U 323a Mi J1; J2; J3
H
Suz HS McL Co1; Co2; Co3 He Fi22; Fi23; Fi024 O'N Ly Ru HN Th B M 3 D4 (2) : 3; : : :
References
1] W. W. Adams, P. Loustaunau, An Introduction to Grobner Bases, AMS, Graduate Studies in Math. III, 1994. 2] P. Ahern, W. Rudin, Periodic automorphisms of C n , Indiana Univ. Math. J. 44 (1995), 287{303. 3] T. Asanuma, Non-linearizable algebraic group actions on A n , J. Algebra 166 (1994), 72-79. 4] T. Asanuma, Non-linearizable algebraic k?-actions on a ne spaces, Preprint, 1996. 5] A. Bialynicki-Birula, Remarks on the action of an algebraic torus on kn, I, Bull. Acad. Polon. Sci. Ser. Sci. Math. 14 (1966), 177{181. 6] A. Bialynicki-Birula, Remarks on the action of an algebraic torus on kn, II, Bull. Acad. Polon. Sci. Ser. Sci. Math. 15 (1967), 123{125. 7] M. Brion, Groupe de Picard et nombres characteristiques des varietes speriques, Duke Math. J. 58 (1989), 397{424. 8] R. W. Carter, Finite Groups of Lie Type { conjugacy classes and complex characters, John Wiley & Sons Ltd., 1985. 9] J. H. Conway, R. T. Curtis, S. P. Norton, R. A. Parker, R. A. Wilson, Atlas of Finite Groups, Clarendon Press, Oxford, 1985. 10] D. Cox, J. Little, D. O'Shea, Ideals, Varieties and Algorithms, Springer-Verlag, New York, 1992. 11] H. Derksen, H. Kraft, Constructive invariant theory, to appear in: Colloques et Seminaires (S.M.F.). 12] A. Dimca, Hypersurfaces in C n di eomorphic to R2n 2, Max-Planck Institut Bonn, Preprint, 1990. 13] D. Eisenbud, Introduction to Commutative Algebra with a View Towards Algebraic Geometry, Graduate Texts in Mathematics, Springer-Verlag, New York, 1995. 14] D. Eisenbud, S. Goto, Linear Free Resolutions and Minimal Multiplicity, J. of Algebra 88 (1984), 89{133. 15] O. Forster, Zur Theorie der Steinschen Algebren und Moduln, Math. Z. 97 (1967), 376{405. 16] F. Forstneric, J. Globevnik, J.-P. Rosay, Non straightenable complex lines in C 2 , Ark. Mat. 34 (1996), 97{101. 17] W. Fulton, Intersection Theory, Springer-Verlag, 1984. 18] K. Gatermann, Semi-invariants, equivariants and algorithms, Preprint, 1995. 35
36
References
19] P. Gordan, Beweis dass jede Covariante und Invariante einer binaren Form eine ganze Function mit numerischen Coe cienten solcher Formen ist, J. fur reine u. angew. Math. 69 (1868), 323{354. 20] Computing bases for rings of permutation-invariant polynomials, J. Symbolic Comput 19 (1995), no. 4, 285{291. 21] P. Heinzner, Geometric Invariant Theory on Stein spaces, Math. Ann. 289 (1991), 631{662. 22] P. Heinzner, F. Kutzschebauch, An equivariant version of Grauert's Oka principle, Invent. Math. 119 (1996), 317{346. 23] A. E. Heydtmann, Generating Invariant Rings of Finite Groups, Diploma Thesis, Saarbrucken, 1996. 24] D. Hilbert, Uber die Theorie der algebraischen Formen, Math. Ann. 36 (1890), 473{534. 25] D. Hilbert, Uber die vollen Invariantensysteme Math. Ann. 42 (1893), 313{373. 26] D. Hilbert, Mathematische Probleme, Archiv fur Math. und Physik 1 (1901), 44{ 63, 213{237; In: Gesammelte Abhandlungen Band III, Springer Verlag, Berlin{ Heidelberg{New York, 1970, 290{329. 27] K. Hiss, Constructive invariant theory for reductive algebraic groups, Preprint, 1993. 28] M. Hochster, J. Roberts, Rings of invariants of reductive groups acting on regular rings are Cohen-Macaulay, Adv. Math. 13 (1974), 115{175. 29] M. Jiang, Thesis, Brandeis (1992). 30] C. Jordan, Memoire sur les covariants des formes binaires, J. de Math. (3) 2 (1876), 177{232. 31] C. Jordan, Sur les covariants des formes binaires, J. de Math. (3) 5 (1879), 345{378. 32] S. Kaliman, Smooth contrible hypersurfaces in C n and exotic algebraic structures on C 3 , Math. Zeitschrift 214 (1993), 499{510. 33] S. Kaliman, Exotic structures on C n and C ? -actions on C 3 , In: Proceedings on \Complex Analysis and Geometry" (Trento, 1993), Marcel Dekker Inc. New York, 1996. 34] S. Kaliman, L. Makar-Limanov, On Russell's contractible threefolds, Preprint 1995. 35] B. Kazarnovskii, Newton polyhedra and the Bezout formula for matrix-valued functions of nite-dimensional representations, Functional Analysis and its Applications 21(4) (1987), 73{74. 36] G. Kemper, The INVAR package for calculating rings of invariants, IWR Preprint 93-94 (1993), University of Heidelberg. 37] G. Kemper, Calculating invariant rings of nite groups over arbitrary elds, J. Symbolic Computation 21 (1996), 351{366. 38] F. Knop, Nichtlinearisierbare Operationen halbeinfacher Gruppen auf a nen Raumen, Invent. Math. 105 (1991), 217-220.
References
37
39] M. Koras, P. Russell, Contractible threefolds and C ? -actions on C 3 , CICMA preprint series, Concordia Laval McGill bf 1995-04. 40] H. Kraft, C ? -Actions on a ne space, In: Operator Algebras, Unitary Representations, Enveloping Algebras and Invariant Theory, edited by A. Connes et al., Progress in Mathematics 92 (1990), 561{579. 41] H. Kraft, Geometrische Methoden in der Invariantentheorie, Vieweg-Verlag, 1984. 42] H. Kraft, Challenging problems on a ne n-space, Seminaire BOURBAKI, 47eme annee 802 (1995). 43] H. Kraft, F. Kutzschebauch, Equivariant a ne line bundles and linearization, Math. Res. Letters (1996), to appear. 44] H. Kraft, G. Schwarz, Reductive group actions with one dimensional quotient, Publ. Math. IHES 76 (1992), 1{97. 45] A. Libgober, A geometrical procedure for killing the middle dimensional homology groups of algebraic hypersurfaces, Proc. Amer. Math. Soc. 63 (1977), 198{202. 46] L. Makar-Limanov, On the hypersurface x + x2 y + z2 + t3 = 0 in C 4 , Preprint 1994. 47] M. Masuda, L. Moser, T. Petrie, The equivariant Serre Problem for abelian groups Topology 35 (1996), no. 2, 329{334. 48] F. Meyer, Invariantentheorie, Encyklopadie der math. Wissenschaften IB2 (1899), 345{378. 49] D. Mumford, J. Forgarty, F. Kirwan, Geometric Invariant Theory, SpringerVerlag, 1994. 50] M. Nagata, On the 14th problem of Hilbert, Amer. J. Math. 81 (1959), 766{772. 51] M. Nagata, Invariants of a group in an a ne ring, J. Math. Kyoto Univ. 3 (1963/64), 369{377. 52] E. Noether, Der Endlichkeitssatz der Invarianten endlicher Gruppen, Math. Ann. 77 (1916), 89{92. 53] A. L. Onishchik, E. B. Vinberg, editors, Lie Groups and Lie Algebras III, Encyclopaedia of Mathematical Sciences 41, Springer-Verlag, 1994. 54] V. Popov, Constructive invariant theory, Asterique 87{88 (1981), 303{334. 55] V. Popov, The constructive theory of invariants, Math. USSR Izvest. 10 (1982), 359{376. 56] J.-P. Rosay, W. Rudin, Holomorphic maps from C n to C n , Trans. Amer. Math. Soc. 310 (1988), 47{86. 57] A. Sathaye, On linear planes, Proc. Amer. Math. Soc. 56 (1976), 1{7. 58] B. Schmid, Generating invariants of nite groups, C. R. Acad. Sci. Paris 308 Serie I (1989), 1{6. 59] B. Schmid, Finite groups and invariant theory, In: P. Dubreil, M.-P. Malliavin, editors, Seminaire d'Algebre, Lecture Notes in Math. 1478, Springer-Verlag, Berlin{ Heidelberg{New York, 1991. 60] G. Schwarz, Exotic algebraic group actions, C.R. Acad. Sci. Paris 309, Serie I (1989), 89{94.
38
References
61] I. R. Shafarevich, editor, Algebraic Geometry I, Encyclopaedia of Mathematical Sciences 23, Springer-Verlag, 1994. 62] L. Smith, E. Noether's bound in the invariant theory of nite groups, Arch. Math. (Basel) 66 (1996), no. 2, 89{91. 63] D. Snow, Reductive group actions on Stein spaces, Math. Ann. 259 (1982), 79{97. 64] M. Suzuki, Sur les operationes holomorphes du group additif complexe sur l'espace de deux variables complexes, Ann. Scient. Ec. Norm. Sup. 4e serie 10 (1977), 517{ 546. 65] T. A. Springer, Aktionen Reduktiver Gruppen auf Varietaten, In: H. Kraft, P. Slodowy, T. A. Springer, editors, Agebraische Transformationsgruppen und Invariantentheorie, Birkhauser Verlag, 1989. 66] T.A. Springer, Linear Algebraic Groups, Birkhauser Verlag, 1991. 67] B. Sturmfels, Algorithms in Invariant Theory, Springer-Verlag, Wien{New York, 1993. 68] D. Wehlau, Constructive invariant theory for tori, Ann. Inst. Fourier 43, 4, 1993, 1055{1066. 69] H. Weyl, The Classical Groups, their Invariants and Representations, Princeton Mathematical Series 1, Princeton Univ. Press, 1946.
Lebenslauf
Am 10. Juli 1970 wurde ich, Harm Derksen, als Sohn von Wim und Joke Derksen in Ven-Zelderheide, Niederlande, geboren. 1976{1982 1982{1988 Juni 1988 1988{1993 Besuch der Grundschule St. Antonius Abt, Ven-Zelderheide Besuch der Gesamtschule Elzendaalcollege Boxmeer (VWO) Abitur Studium der Mathematik an der Katholischen Universitat Nijmegen Mai 1993 Diplom in Mathematik an der Katholischen Universitat Nijmegen (Cum Laude) unter Anleitung von Dr. Van den Essen. Diplomthema war die Beziehung zwischen lokal endliche Derivationen und Gruppenwirkungen und Anwendungen. 1993{1994 AIO (Assistent in Ausbildung) am Mathematischen Institut der Katholischen Universitat Nijmegen unter Anleitung von Dr. Van den Essen und Prof. Dr. Levelt Februar{Mai 1994 Besuch der Universitat Basel mit einem Stipendium der VSB Bank Ab Oktober 1994 Assistent am Mathematischen Institut der Universitat Basel bei Prof. Dr. Kraft, unterstutzt durch den Schweizerischen Nationalfonds
39

Diss

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Diss

Uploaded by

Copyright:

Available Formats

Constructive Invariant Theory and the Linearization Problem

vorgelegt der Philosophisch-Naturwissenschaftlichen Fakultat der Universitat Basel

* This is theology and not mathematics. 1

I. Computation of Reductive Group Invariants

maxfjGj; n(jGj 1)g:

maxfn; n(n 1)=2g

NV := minf ` j X `+1v = 0 for all v 2 V; X 2 Lie(U ) g:

minfC (nLV )m ; C 0Lm +mg: V

I. Computation of Reductive Group Invariants

3. Finding generators of the zero- ber ideal

2 O(V )G are generators of IN , then the images of f1 ; : : : ; fr

I. Computation of Reductive Group Invariants

4. Degree bounds for generating invariants

minf C (nLV )m ; C 0Lm +m g: V

S1( (g)) = S2( (g)) = : : : = Sm( (g)) = 0

I. Computation of Reductive Group Invariants

S1(p) = S2(p) = : : : = Sm(p) = 0

Pi(g; h1; h2 ; : : : ; hn+m 1) = det(g:f1 ; g:f2; : : : ; g:fn; hi ; hi+1; : : : ; hn+i 1)

f1(g:p) = f2 (g:p) = : : : = fn(g:p) = h1(p) = h2(p) = : : : = hn+m 1(p) = 0

4. Degree bounds for generating invariants

Y := f(v1 ; v2 ; : : : ; vt ) 2 ku t j det(vi ; vi+1; : : : ; vi+u 1) = 0 (1 i t u + 1) g

::: ::: ... 0 ::: 0 :::

::: ::: ... ::: :::

be the representation of G on n(V V Qi 2 End n (V V ) ? is de ned by

V ). Now Pi(g) is equal to Qi ( (g)) where

v1; v2 ; : : : ; vu in V . It follows that L^n V

nLV . The number of solutions of

Q1 ( (g)) = Q2 ( (g)) = : : : = Qm+n 1 ( (g))

I. Computation of Reductive Group Invariants

( This implies n0 (C5Lr ) C6LVm r)=2 V

C7Lm , and so we nally deduce V

C4(n0 LV )m C4(C7 Lm+1)m = C8Lm(m+1): V V

Example 4.11. Binary forms.

an;1(g) an;2(g) : : : an;n(g)

k Z1; Z2; : : : ; Zs; X1 ; X2 ; : : : ; Xn ; Y1; Y2 ; : : : ; Yn]=I

I. Computation of Reductive Group Invariants

Using Grobner basis we can compute b = I \ k X; Y ] which is equal to

0 The ideal I de ning is given by

0 Z (1 + Z )=2 Z 2(1 Z )=2 1 2 2 1 @ Z1 (1 Z2)=2 Z12(1 + Z2)=2

Z1 Z2 Z3 Z4 The representation on S 2W is given by the matrix 0 Z2 2 1 2Z2Z1 Z2 1 @ Z3 Z1 Z4Z1 + Z3 Z2 Z4Z2 A 2 2 Z3 2Z4Z3 Z4

I = (Z1 X1 + Z2X2 Y1; Z3 X1 + Z4 X2 Y2; Z1 X3 + Z2X4 Y3; Z3X3 + Z4X4 Y4;

I. Computation of Reductive Group Invariants

R(X1 X3 X7 2X2 X3X6 + X2X4 X5) = X1 X3 X7 X2 X3X6 X1X4 X6 + X2 X4 X5

II. On the hypersurface x + x2 y + z2 + t3 = 0 in C 4

C C x; y; z; t]=(x + x2 y + z2 + t3) 6= C x; y; z; t]=(x + x2 y + z2 + t3):

T and we have 1 1 F iR = f0g. i=

II. On the hypersurface x + x2 y + z2 + t3 = 0 in C 4

z2 t3 we can write every

f = a(x; z; t) + yb(y; z; t) + xyc(y; z; t)

for unique polynomials a; b; c

3. The proof that x + x2 y + z2 + t3 = 0 is not isomorphic to C 3

r > 0, u; v; w 0 and i ; i 2 C n f0g

II. On the hypersurface x + x2 y + z2 + t3 = 0 in C 4

III. Nonlinearizable Holomorphic Group Actions

'(z1; : : : ; zk ) = (z1 ; : : : ; zk ; 0; : : : ; 0):

Corollary 2.4. The universal covering of C 2+k n 'k (C 1+k ) = (C 2 n '(C ))

De nition. We de ne the Rees space R(X; Y; Z; f ) as Im( ), the holomorphic closure

III. Nonlinearizable Holomorphic Group Actions

Corollary 3.7. For an embedding ' : C k ! C n and f 2 O(C l ) we have

4. Non-linearizable group actions

X = f(x; y; z; w) 2 C 4 j h(x; y) = zwg:

(z; w; u1 ; : : : ; uk+1) = ( z; The categorical quotient of X

and the categorical quotient of C 3+k is given by