ArticlePDF Available
Invent. math. 123, 233-240 (1996)
Inventiones
mathematicae
9 Springer-Verlag 1996
Non unimodular Hermitian forms
Eva Bayer-Fluckiger, Laura Fainsilber
URA 741 du CNRS, Laboratoire de Mathematiques, Facult~ des Sciences Universit~ de
Franche-Comt6, 16, route de Gray, F-25030 Besan~on, France
Ob[atum 12-XII-1994
Introduction
The classical theory of integral quadratic forms deals not only with unimodular
forms, but also with non unimodular ones. However, during the last 30 years
unimodular forms received much more attention. This is partly due to the
development of the algebraic theory of quadratic forms, which concentrates on
forms over fields, where the interesting forms are indeed unimodular, and to
the fact that forms arising from topology are usually unimodular.
There is now an extensive literature on unimodular hermitian forms, even
over quite general rings (see for instance [6], [11]). On the other hand, com-
paratively little is known about non unimodular forms. The purpose of the
present note is to show that the classification of non unimodular forms can
be reduced to a module theoretic question and the classification of unimodular
forms over a different ring, and to give some applications of this method.
The main result is proved in Sect. 4 and 5. In Sect. 6, we illustrate our
technique in the case of principal ideal domains. We then consider algebras
which are free modules of finite type over complete discrete valuation rings.
The above results are applied in Sect. 7 to prove a Witt type cancellation theo-
rem. This was previously known for unimodular forms (Quebbemann, Scharlau,
and Schulte [7]), as well as for some non unimodular forms over hereditary
orders (Riehm [10]). Finally, Sect. 8 contains an analogue of a theorem of
Springer for hermitian forms.
1. Hermitian forms over rings
Let A be a ring, and let -:A ~ A be an involution (that is, an anti-
automorphism of order 2 of A). Let M be a left A-module of finite type.
Let e = il. An
e-hermitian
form over A is a biadditive map h : M x M --~ A
234 E. Bayer-Fluckiger, L. Fainsilber
such that
h(ax, by) = ah(x, y)b
and that
h(y,x) = eh(x,y)
for all
x, y E M
and
all
a,b E A.
An isomorphism ~: : (M,h)---+
(M',h')
is an isomorphism of left
A-modules such that
h'(4)(x), (~(y)) = h(x, y)
for all x, y C M.
Let M* = Hom(M,A). Then M* has a structure of left A-module, given
by
a.f(x) = f(x)~
for all a E A, f E M* and x E M. An e-hermitian form h
induces a morphism of left A-modules M ~ M*, which we also denote by h.
The form
(M,h)
is said to be
unimodular
if and only if h : M ---+ M* is an
isomorphism.
A left A-module M is said to be
reflexive
if the morphism e~t : M ---+ M**,
given by
eM(m)(f) = f(m),
is an isomorphism. For instance, every projective
module is reflexive.
The
orthogonal sum (M,h)|
of two forms
(M,h)
and
(N,g)
is
given by the form (M O N, h | g), where
(h 0 g)(m 9 n, m ~ 9 n ~) = h(m, m') +
g(n,n').
2. Hermitian categories
Let ~ be an additive category. Let ,:c~ __, ~ be a duality functor, i.e. an
additive contravariant functor with a natural isomorphism
(Ec)cc~ :id--~**
such that
E~Ec. = idc*
for all C C ~.
A sesquilinear form
in the category ~ is a pair
(C,H),
where C is an
object of cg and H is a morphism H : C --~ C*. Let ~: = +1. A sesquilinear
form
(M,h)
is said to be
~:-hermitian
if H =
cH*EM.
A form
(C,H)
is said
to be
unimodular
if H : C --~ C* is an isomorphism.
Let
(C,H)
and
(C,H ~)
be two e-hermitian forms. A morphism q5 : (C,H)
(C~,H ~)
is an
isomorphism dp : C ---+ C ~
of the category ~ such that H =
~*H'4,.
We denote by H':(~) the category of e-hermitian forms on ~, and by
H~(~) the full subcategory of unimodular e-hermitian forms on ~.
Example.
Let Jg be the category of finitely generated reflexive left A-modules.
Then ~//g is an additive category. The functor * : M --+ M which sends M to M*
is a duality functor on o~, with EM = e~t (cf. Sect. 1 ). The category of e-
hermitian forms over the ring A is equivalent to H~:(~/). The equivalence is
obtained by sending a form h to the corresponding morphism h : M ~ M*.
3. The category of morphisms of ~/~
A morphism of ~' is a triple
(MbM2,f),
where Ml and M2 are finitely
generated reflexive left A-modules and f : M1 --* M2 is a morphism of left A-
modules. Direct sums are defined in the obvious way. Let us call ~ the category
of morphisms of ~. A
~-morphism (M1,M2,f) ~ (N1,N2,g)
is a pair q5 =
(~bl,q52) of morphisms of left A-modules qSi :
Mi ~ Ni
such that 4~2f = g~bl.
Let tlM*2,M*l,jc*~1 be the
dual
of
(Mb M2, f ).
Set
E(M,,M:, f) = ( eM, , eM, ).
This
defines a duality functor on the category ~?.
Non unimodular Hermitian forms 235
4. Non unimodular hermitian forms
Let H~:(.~) be the category of unimodular e-hermitian forms over 2 (cf. Sect.
2).
We define a functor F from the category H':(.<H) of e-hermitian forms over
the ring A to the category //~,(~), as follows. Let
(M,h)
be an e-hermitian
form, h :M --~ M*. Then
(M,M*,h)
is an object of 2. It is easy to check that
the pair
(eM,~:idM*)
defines a unimodular e-hermitian form over
(M,M*,h).
Therefore,
F(M,h)= ((M,M*,h),(eM,~:idM.))
is an object of Hi(z? ).
Theorem
1.
The functor F is an equivalence of categories between the cate-
gory H~(~) of e-hermitian .forms over A and H[~(~).
Proof
As already pointed out, Y sends an e-hermitian form to an object of
H~(22). In order to prove that .~- is a functor, it remains to be checked that
sends morphisms to morphisms.
Let q~ :
(M,h) --, (N,g)
be a morphism of e-hermitian forms. Then we have
the commutative diagram
M J~ M*
N ~ N*.
Therefore,
(fl/),~p*-l) : (M,M*,h) ---+ (N,N*,g)
is a morphism in 2. Moreover, as ~b**eM = ex~b by naturality of (eM)ME //, it
is straightforward to see that it is also a morphism of H~(~).
Let us define a functor N from H~"(22) to the category of e-hermitian forms
over A. Let (Q,~) E H~:(2), with Q =
(N,N*,g)
and ~ = (~1,~2). Set .C6(Q, ~) =
(N, ~2g). This is an e-hermitian form over A, which is unimodular if and only
if g is bijective.
Let us check that N sends morphisms to morphisms. Indeed, let q5 =
(~)l,q~2)
be a morphism of H~(~ ~b: (P,q) -~ (Q,~), where (P,I/) =
((M,M*,h),(~II,I12)),
(Q,~)= ((N,N*,y),(~I,~2)).
In particular, q5 is a mor-
phism of 22, therefore 9q5~ = q~2h. Moreover, q~ is also a morphism of H,;:(A),
therefore 172 q~T~2~b2.
This implies that
r/2h *~ -I ,
: = (/)lg2q)2~b 2 ,q~bl = ~)l(k_2,q)q~l,
so (bl is a morphism of H%~H).
Finally, it is easy to verify that ~ o ~N is the identity of
H%/#)
and that
,r o (q is isomorphic to the identity of H~(~). Therefore, .~- : H,':(2) -~ H%/I/)
is an equivalence of categories. This completes the proof of the theorem.
5. Hermitian forms corresponding to a given object of oQ
If (M,h) is an e-hermitian form over A, we denote by
q(M,h)
the object
(M,M*,h)
of 2.
236 E. Bayer-Fluckiger, L. Fainsilber
We describe the set of isomorphism classes of ~-hermitian forms corre-
sponding to a given object of ~? in terms of rank one hermitian forms over the
ring of endomorphisms of this object. Let us fix an e,-hermitian form
(Mo, ho),
and let E be the ring of endomorphisms of 22o =
q(Mo, ho)
in ~. Let us also fix
a unimodular e,-hermitian form ~/0 over Q0. Such forms exist, for instance we
~
can take r/0 = (eMo,
~idM,; ).
This form induces an involution 9 E ---+ E, defined
by
j~ --I o
= r/0 f r/0,
for all f E E, where fo denotes the dual of f in 2. Let E be the group of
units of E. Set
E + = {f E E ? = f},
and define an equivalence relation on E + by
f =-- f' ~ 3gEE• gfg= f'.
Note that
E+/:-
is the set of isomorphism classes of rank one unimodular
hermitian forms over (E,").
Theorem
2.
The set of isomorphism classes of e-hermitian Jbrms (M, h) such
that q(M, h) = Qo is in bijection with E+/=_.
Proof
Let r/=
(eM,~idM*).
The bijection is given by
(M,h), ~
~o~.
Indeed, it is straightforward to check that r/olr/E E +. Moreover, if
(N,g)
is
another e-hermitian form with
q(N,
g) = Q0, then
(M,h ) ~- (N,g) ~=~ ~loLtl = qol ~,
where ~ =
(ex, eidx, ).
This completes the proof of the theorem.
Note that the results of the last two sections indeed reduce the classification
of arbitrary hermitian forms to a module theoretic question and to the classifi-
cation of unimodular forms over the ring of endomorphisms, as announced in
the introduction.
6. An example: symmetric bilinear forms of given determinant over a prin-
cipal ideal domain
In this section, we assume that A is a principal ideal domain. Let K be the
field of fractions of A. The aim of this section is to illustrate the methods of
Sects. 4 and 5 in this case.
o, C K. Let a denote
Let al,...,an be non-zero elements of A and set
ai, j = a--~
the n-tuple (al ..... an) and let E(a) be the subring of
Mn(A)
defined by
E(a) =
{(ai, jxi, j)i,j E Mn(A)lxi, j E A, Vi,j}.
Non unimodular Hermitian forms 237
Let us define an involution a : E(a) --+ E(a) by
a((a,,j,xi,j),,j) = (x/,i)i,1.
Note that
xj, i
= ai.j(aj.ix/.t),
therefore
(xj,~)i.j C
E(a).
As in Sect. 5, we use the notations
E(a) + =
{fE E(a)• = f},
and
f --= f' ~=~ 3g E E•
a(g)fg = f'.
For any non-zero element d E A, we denote by
A(n,d)
the set of n-tuples
(al ..... an) of non-zero elements of A such that
ai
divides a,+j for all i =
1
..... n - 1 and d = al ...an. If a--
(al ..... an) E A(n,d),
let us denote by aA
the n-tuple of ideals
(alA ..... anA)
generated by aj ..... a~. Let
I(n,d)
be the
set of n-tuples of ideals aA with aE
A(n,d).
Proposition
1.
Let n be a positive integer and let d be a non-zero element
of A. The set of isomorphism classes of symmetric bilinear .forms over A of
rank n and determinant d is in bijection with the set
{(aA, f) I a
E
A(n,d), f C
E(a)+/~}.
Proof
Let M be a free A-module of rank n and let h : M x M --~A be a
symmetric bilinear form of determinant d. Let h : M --~ M* be the associated
morphism. As in Sect. 5, we denote by
q(M,h)= (M,M*,h)
the object of
associated to
(M,h).
By the theorem of elementary divisors, there exists
a = (al ..... an) C A(n,d)
such that
q(M,h) ~- (An,A",diag(a)),
where diag(a) :
A n -~ A n is the diagonal map determined by a = (at ..... an). The isomorphism
class of
q(M,h)
is determined by the n-tuple of ideals aA C
I(n,d).
The ring of endomorphisms of
q(M,h)
is
E = {(e, fl) C
Mn(A)
M,(A) I a diag(a) = diag(a)fl}.
Let t/=
(eM, idM* ).
The involution of E induced by t/ (cf. Sect. 5) is (e, fl) =
(fit, at). A computation shows that
E = {((a,,jx,,i),,j, (x~,j)i,j) E
E(a) E(a)}.
Therefore (E,-) _~ (E(a),a). By Ths. 1 and 2, this proves the proposition.
7. Witt's cancellation theorem for E-hermitian forms
Let R be a complete discrete valuation ring. Let n be a uniformiser of R.
Assume that A is an R-algebra of finite rank over R, and that there exists an
element a of the center of A with a + ~ = 1. For instance, this condition is
fulfilled if 2 is invertible in R.
238 E. Bayer-Fluckiger, L. Fainsilber
Theorem
3. Let (M,h), (Ml,hl) and
(m2,h2)
be e-hermitkm forms over A.
Assume that
(Ml,hl ) 9 (m,h ) ~- (m2,h2)
@
(m,h ).
Then we have
(Ml,h~ ) ~- (M2,h2).
In other words, Witt's cancellation theorem holds for arbitrary e-hermitian
forms over A.
As in Sect. 4, let us denote by o) the category of morphisms of reflexive
modules over A. In Sect. 4, we have seen that the category of e-hermitian
forms over A is equivalent to the category H~'(~) of unimodular e-hermitian
forms over the category ~. Hence, it suffices to show that Witt cancellation
holds in the category H,':(~). This is the purpose of the next proposition.
Proposition
2. Let (Q,q), (Ql,~l ) and
(Q2,q2)
be objects of H~:(2?). Assume
that
(Ql,qi)O(Q,q) ~ (Q2,r/2) | (Q,~/).
Then
(Ql,ql) ~ (Qz,r/2).
Proof We deduce the proposition from some results of Quebbemann, Scharlau
and Schulte [7]. Let us start by checking that the category ~ satisfies the
hypotheses (i), (ii) and (iii) of [7].
(i) All idempotents in ~ split. This is clear: if z is an idemponent of Q,
then it is easy to check that ~(Q) is a direct summand of Q.
(ii) Every object Q of ~ has a decomposition Q _~ Ql 0... O Q,., with Q,
indecomposable and End(Q,) local. This follows from [9], Sect. 2, (8), (10)
and (11).
(iii) For every object Q of ~, the ring of endomorphisms End(Q) is J(Q)-
adically complete, where J(Q) is the radical of End(Q). Indeed, there exists
a positive integer k such that j(Q)k C 7zEnd(Q) C J(Q) (cf. [2], 5.22), hence
the J(Q)-adic topology of End (Q) is equivalent to the ~z-adic topology.
Moreover, we are assuming that the center of A contains an element a
such that a + 6 = 1. This element a belongs to the endomorphism ring of any
object of 2?. Therefore every e-hermitian form over 2~ is even. By [7], 3.4, we
obtain the desired cancellation result.
8. Springer's theorem for E-hermitian forms
Let K be a local field, CK its ring of integers, and AK an t~x-algebra of finite
rank with an t~x-linear involution -. Let L be a finite extension of K, Cr its
ring of integers, and AL = AK @e~ ~L with the 6.'L-linear involution that extends
-. As in Sect. 4, let 27K (resp. 27t) be denote the category of morphisms of re-
flexive modules over AK (resp. AL). If Q = (M,N, f) is an object of 27K, we set
QL = (ML,Nt,fL) where Mt = M | CL, NL = N | CL, and fL extends f.
Non unimodular Hermitian forms 239
Lemma 1.
Let Q = (M,N,f) and Q'= (M',N',f') be two object in )QK.
Then QL is isomorphic to Q~ in ~z if and only if Q is isomorphic' to Q'
in dx.
Proof
Let (qS,~b)' QL---+ Q~ be a ~L-isomorphism. Then q~ and ~b are AL-
linear and commute with fL and
f'L.
Let n =
[L'K],
and note that 6L is
a free 6)K-module of rank n. We can also see QL and Q~ as objects in
~X :ME, NL, M~, N[ are A~:-modules, isomorphic to
M '~, N ~, M", N ~"
re-
spectively, and
fc, f~
are A~c-linear. The ~c-isomorphism (~b,~b) is also a
~K-isomorphism from QL --~ Q" to Q~ ~- Q~". So, by the Krull-Schmidt theo-
rem (see for instance [2] or [9]), the summands Q and Q~ are ~K-isomorphic.
A theorem of Springer states that if two quadratic forms over a field k
become isomorphic over an odd degree extension of k, then they are already
isomorphic over k (see for instance [11], 2.5.4). Th. 4. below is an analogue
of this result of Springer.
Theorem 4.
Suppose that L/K is' an odd degree extension, and that the resid-
ual extension is separable. Let (M,h) and (M~,h ~) be two a-hermitian forms
over AK. If (ML,hL) and (M[,h2) are isomorphic over AL then (M,h) and
(M',h ~) are isomorphic over AK.
Proof
Consider the morphisms
q(ML, hL), q(M[,h~),
and
q(M,h), q(M',h').
The first two are ~?L-isomorphic, so by the lemma above, the last two are
~K-isomorphic. Let Q =
q(M,h)
and let E be the ring of ~
of Q. By th. 2, the set of isomorphism classes of e-hermitian forms (M',h')
such that
q(M',h') ~- Q
is in bijection with the set of isomorphism classes of
rank one unimodular hermitian forms over (E, - ). So we have two unimodular
hermitian forms r/, r over E that extend to isomorphic forms over EL = E Qe~
[~'L. By Springer's theorem for unimodular hermitian forms over E (cf. [3]),
we have ~/-~ r/. Hence
(M,h) ~_ (M',h r)
over
AK.
This concludes the proof of
the theorem.
Acknowledgements.
We thank the Swiss National Foundation for Scientific Research for
partial support during the preparation of this paper, and the University of Geneva for its
hospitality.
References
1. E. Bayer-Fluckiger, H.W. Lenstra, Jr.: Forms in odd degree extensions and self-dual
normal bases. Am. J. Math. 112 (1990) 359-373
2. C.W. Curtis, 1. Reiner: Methods of representation theory, with applications to finite
groups and orders. Vol I, Wiley, New York (1981)
3. L. Fainsilber: Formes hennitiennes sur des alg6bres sur des anneaux locaux. Publica-
tions Math6matiques de Besancon (1994)
4. L. Fainsilber: Formes hermitiennes sur les alg6bres p-adiques. Th+se, Universit6 de
Franche-Comb6 (1994)
5. T.Y. Lam: The algebraic theory of quadratic forms. W.A. Benjamin, Addison-Wesley,
Reading, Mass. 1973
240 E. Bayer-Fluckiger, L. Fainsilber
6. M. Knus: Quadratic and hermitian forms over rings. Grundlehren Math. Wiss. 294,
Springer, Berlin, Heidelberg, New York (1991)
7. H.-G. Quebbemann, W. Scharlau, M. Schulte: Quadratic and hermitian forms in additive
and abelian categories. J. Algebra 59 (1979) 264-289
8. H.-G. Quebbemann, R. Scharlau, W. Scharlau, M. Schulte: Qudratische Fomlen in
additiven Kategorien. Bull. Soc. Math. France, M6moire 48 (1976) 93-101
9. I. Reiner, K.W. Roggenkamp: Integral representations. Lecture notes in mathematics
744, Springer-Verlag (1985)
10. C. Riehm: Hermitian forms over local hereditary orders. Am. J. Math. 106 (1984)
780-800
11. W. Scharlau: Quadratic and hermitian forms. Grundlehren Math. Wiss. 270, Springer-
Verlag (1985)
... (4) If dim(Q/X) = dim(Q ′ /X) then [Q] = [Q ′ ] in the Grothendieck ring of varieties K 0 (Var/k). (5) If the base scheme X is integral the classes of general fibers Q K(X) and Q ′ K(X) in the Witt group of quadratic forms over the field of rational functions K(X) on X are equal. If, moreover, dim(Q/X) = dim(Q ′ /X) then Q K(X) ∼ = Q ′ K(X) and Q is birational to Q ′ over X. ...
... As it is explained in the survey [4], modifying the standard duality operation on the category of vector bundles on X one can define the Witt group W(X, L) that classifies classes of line bundle valued non-degenerate quadratic forms q : Sym 2 E → L ∨ . Moreover, a trick described in [5] allows one to define the Witt group W nu (X, L) of non-unimodular quadratic forms (i.e., forms that are allowed to be degenerate) as the usual Witt group of the category of morphisms of vector bundle. Thus, quadric bundles (1.1) define elements of W nu (X, L). ...
... For the second invariant, assume X is smooth, connected and proper k-scheme, n = dim(X) is even, and L ⊗ ω X ∼ = M 2 for a line bundle M on X, where ω X is the canonical line bundle of X. Then we define the bilinear form (1. 5) H n/2 (X, E ⊗ M) ⊗ H n/2 (X, E ⊗ M) q − − → H n (X, L ∨ ⊗ M ⊗ M) ∼ = H n (X, ω X ) = k on the cohomology group H n/2 (X, E ⊗ M) which we denote H n/2 (q) or H n/2 (Q). This form, of course, depends on the choice of the line bundle M (if Pic(X) has 2-torsion, there may be several choices), but we suppress this in the notation. ...
Preprint
We introduce the notion of hyperbolic equivalence for quadric bundles and show that hyperbolic equivalent quadric bundles share many important properties: they have the same Brauer data; moreover, if they have the same dimension over the base, they are birational over the base and have equal classes in the Grothendieck ring of varieties. Furthermore, when the base is a projective space we show that two quadric bundles are hyperbolic equivalent if and only if their cokernel sheaves are isomorphic up to twist, their fibers over a fixed point of the base are Witt equivalent, and, in some cases, certain quadratic forms on intermediate cohomology groups of the underlying vector bundles are Witt equivalent. For this we show that any quadric bundle over Pn\mathbb{P}^n is hyperbolic equivalent to a quadric bundle whose underlying vector bundle has many cohomology vanishings; this class of bundles, called VLC bundles in the paper, is interesting by itself.
... The aim of this section is to prove that the category of sesquilinear forms S R (A, σ) is equivalent to the hermitian category H(M R ) defined in §3. We will use the methods of [3]. ...
... Proof. As in [3], proposition 2, we check that the category M R satisfies the conditions (i), (ii) and (iii) above. It follows from theorem 6.2 that Witt's cancellation theorem is true in the category H(M R ). §7. ...
... This follows from theorem 6.2, observing that the category M R satisfies conditions (i), (ii), (iii) (see e.g. [3], proposition 2). ...
Article
Full-text available
Many classical results concerning quadratic forms have been extended to Hermitian forms over algebras with involution. However, not much is known in the case of sesquilinear forms without any symmetry property. The present paper will establish a Witt cancellation result, an analogue of Springer's theorem, as well as some local-global and finiteness results in this context.
... Our aim is to put Theorem 0.3 in a different perspective, and to study how far one can generalize it. Our point of view is inspired by the treatment of non-unimodular forms in [3], [6] and [5]. Roughly speaking, these works reduce the treatment of (systems of) non-unimodular forms to (single) unimodular forms over a different base ring. ...
... Then (P F , f F ) ∼ = (P ′ F , f ′ F ) implies (P, f ) ∼ = (P ′ , f ′ ). This theorem is derived from the following theorem about unimodular hermitian spaces, using results from [3]. ...
... Theorem 3.2 is the subject matter of section 3. In section 4, we prove Theorem 4.3, applying Theorem 2.1 and some results and methods of [3] and [5]. Finally, section 5 contains the above mentioned application to quadratic forms invariant by a finite group (Theorem 5.2). ...
Article
Let R be a semilocal Dedekind domain. Under certain assumptions, we show that two (not necessarily unimodular) hermitian forms over an involutary R-algebra, which are rationally isometric and have isomorphic semisimple coradicals, are in fact isometric. The same result is also obtained for quadratic forms equipped with an action of a finite group. The results have cohomological restatements that resemble the Grothendieck-Serre conjecture, except the group schemes involved are not reductive. We conjecture that this type of result holds for a larger family of group schemes over semilocal Dedekind domains that we describe.
... When 2 ∈ R × , we extend the previous results to systems of sesquilinear forms and non-unimodular hermitian forms, using results from [6], [8] and [7]. Specifically, we show that: ...
... In this final section, we extend Theorems 5.1 to systems of sesquilinear forms and Theorem 6.7 to non-unimodular hermitian forms. This will be done using results from [6], [8] and [7]. As an application, we prove that Witt's Cancellation Theorem and Springer's Theorem hold for hermitian forms over hereditary orders. ...
... it is enough to show that cancellation holds for (arbitrary) u-hermitian forms over finiteR p -algebras, and this was shown in [6,Th. 3] (see also [7, §5A] for further results of this kind). ...
Article
Full-text available
Let R be a semilocal principal ideal domain. Two algebraic objects over R in which scalar extension makes sense (e.g. quadratic spaces) are said to be of the same genus if they become isomorphic after extending scalars to all completions of R and its fraction field. We prove that the number of isomorphism classes in the genus of unimodular quadratic spaces over (non necessarily commutative) R-orders is always a finite power of 2, and under further assumptions, this number is 1. The same result is also shown for related objects, e.g. systems of sesquilinear forms. A key ingredient in the proof is a weak approximation theorem for groups of isometries, which is valid over any (topological) base field, and even over semilocal base rings.
... (Note that the base ring in Keller's counterexample [11, §2] is semilocal but not semiperfect.) This in turn leads to other cancellation theorems as follows: In [4], [7] and [6], it was shown that systems of (not-necessarily unimodular) sesquilinear forms can be treated as (single) unimodular hermitian forms over a different base ring. Thus, cancellation holds when this base ring is semiperfect. ...
... 6.2] with Theorem 4.10 (or Corollary 4.12), we see that condition (C) can indeed be dropped. We have therefore obtained: We now combine Corollary 4.15 with results from [4], [7] and [6] to obtain cancellation results for systems of (not-necessarily unimodular) sesquilinear forms. Henceforth, let R be a commutative ring, and let C be an R-category equipped with R-linear hermitian structures { * i , ω i } i∈I ; see [6, §2.4, ...
... Finally, we use [4] to show that non-unimodular hermitian forms over nonhenselian valuation rings cancel from orthogonal sums. Here, the prefix "non-" stands for "not-necessarily" and not for absolute negation. ...
Article
We prove that every isometry of between (not-necessarily orthogonal) summands of a unimodular quadratic space over a semiperfect ring can be extended an isometry of the whole quadratic space. The same result was proved by Reiter for semilocal rings, but with certain restrictions on the base modules, which cannot be removed in general. Our result implies that unimodular quadratic spaces over semiperfect rings cancel from orthogonal sums. Combining this with other results implies more general cancellation theorems (e.g. for systems of sesquilinear forms). In addition, we also determine the group generated by the reflections of a unimodular quadratic space over a semiperfect ring.
... Two elements equivalent in the above sense will be called congruent. Many classification problems in the theory of quadratic and hermitian forms can be reduced to determining the congruence classes in a suitable algebra with involution [3,6,8,9,11,12,13,15]. Note that H(Λ) is also the cohomology set H 1 (C 2 , Λ × ) in non-abelian cohomology, where the non-trivial element in C 2 acts via λ →λ −1 . ...
... The ring Λ is actually the endomorphism ring of the arrow (1, π 2 ) : O 2 → O 2 associated with the quadratic form < 1, π 2 > (see [3]), and H(Λ) corresponds to the set of (non-unimodular) forms {< 1, π 2 >, < ε, επ 2 >, < 1, επ 2 >, < ε, π 2 >}. The defect in injectivity reflects the fact that the first two forms and the last two become isometric over K. ...
Article
Let Λ be a ring endowed with an involution a → ã. We say that two units a and b of Λ fixed under the involution are congruent if there exists an element u ∈ Λx such that a = ubũ. We denote by H(Λ) the set of congruence classes. In this paper we consider the case where Λ is an order with involution in a semisimple algebra A over a local field and study the question of whether the natural map H(Λ) → H(A) induced by inclusion is injective. We give sufficient conditions on the order Λ for this map to be injective and give applications to hermitian forms over group rings.
... The bar over Hom means that we consider the set of all antihomomorphisms, compare [11], §1.4. Note that M can be identified with the ring of germs of meromorphic function f : (−ǫ, ǫ) → C. The canonical isomorphism s ′ : X → e ′ e ′ (X) is given by formula (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13) in [11] (evaluation and conjugation). ...
Preprint
This paper continues math.DG/9903140. Here we construct a linking form on the torsion part of middle dimensional extended L^2 homology and cohomology of odd-dimensional manifolds. We give a geometric necessary condition when this linking form is hyperbolic. We compute this linking form in case when the manifold bounds. We introduce and study new numerical invariants of the linking form: the Novikov - Shubin signature and the torsion signature. We compute these invariants explicitly for manifolds with π1=Z\pi_1 = Z in terms of the Blanchfield form. We develop a notion of excess for extensions of torsion modules and show how this concept can be used to guarantee vanishing of the torsion signature.
... In a more general setting Bayer-Fluckiger and Fainsilber ( [1]) have considered the general problem of equivalence of hermitian or skew hermitian forms over arbitrary rings and are able to prove some quite general reduction theorems for this problem. ...
Article
Full-text available
We study the classification problem of possibly degenerate hermitian and skew hermitian bilinear forms over local rings where 2 is a unit.
Preprint
Torsion objects of von Neumann categories describe the phenomen "spectrum near zero" discovered by S. Novikov and M. Shubin. In this paper we classify Hermitian forms on torsion objects of a finite von Neumann category. We prove that any such form can be represented as a discriminant form of a degenerate Hermitian form on a projective module. We also find a relation between the Hermitian forms on projective modules which holds if and only if their discriminant forms are congruent. A notion of superfinite von Neumann category is introduced. It is proven that the classification of torsion Hermitian forms in a superfinite category can be completely reduced to the isomorphisn types of their positive and negative parts.
Preprint
Full-text available
We give necessary and sufficient conditions for an integer to be the signature of a (4q-1)-knot in the (4q+1)-sphere with a given square-free Alexander polynomial.
Book
For a long time - at least from Fermat to Minkowski - the theory of quadratic forms was a part of number theory. Much of the best work of the great number theorists of the eighteenth and nineteenth century was concerned with problems about quadratic forms. On the basis of their work, Minkowski, Siegel, Hasse, Eichler and many others crea­ ted the impressive "arithmetic" theory of quadratic forms, which has been the object of the well-known books by Bachmann (1898/1923), Eichler (1952), and O'Meara (1963). Parallel to this development the ideas of abstract algebra and abstract linear algebra introduced by Dedekind, Frobenius, E. Noether and Artin led to today's structural mathematics with its emphasis on classification problems and general structure theorems. On the basis of both - the number theory of quadratic forms and the ideas of modern algebra - Witt opened, in 1937, a new chapter in the theory of quadratic forms. His most fruitful idea was to consider not single "individual" quadratic forms but rather the entity of all forms over a fixed ground field and to construct from this an algebra­ ic object. This object - the Witt ring - then became the principal object of the entire theory. Thirty years later Pfister demonstrated the significance of this approach by his celebrated structure theorems.
Article
Let X be a topological space and , be the ring of continuous k-valued functions on X. We give algebraic conditions for a subring of C(X, k) to have, up to isomorphism, the same quadratic or λ-hermitian forms.
Article
During the last few years several papers concerned with the foundations of the theory of quadratic forms over arbitrary rings with involution have appeared. It is not necessary to give detailed references, in particular one thinks of the well known work of Bak [l], Bass [3], Karoubi, Knebusch [ll, 121, Ranicki, Vaserstein, and C. T. C. Wall. During the same period a number of problems quite similar to those occuring in the theory of quadratic forms were discussed, which, however, did not fit in the formalism developed so far. For example, one thinks of problems like the classification of pairs of forms, of sesquilinear forms, isometries , quadratic spaces with systems of subspaces, and also of quadratic forms over schemes, see e.g. [12, 13, 20, 22, 231. This situation called for a more general foundation of the theory of quadratic and hermitian forms. In this paper we try to give this foundation. Our basic object is an additive category &! together wit a duality functor *: & + A. In this situation one can define the most important notions of the theory of quadratic forms. Under suitable finiteness conditions one can prove a Krull-Schmidt theorem which is a sharpening of the classical Witt theorem. This result is basic for applications to the problems mentioned above. A preliminary version of this material is contained in [ 171. As just one application we discuss the classification of quadratic spaces with four subspaces. We hope that this disucssion will show clearly how one can solve a number of important classification problems of linear algebra. More applications can be found in [15, 16, 17, 22, 241. In the second part of the paper we discuss hermitian (not quadratic) forms in an abelian category. In an abelian category one has more structure, in particular one can introduce the notion of orthogonality. This allows one to introduce Grothendieck and Witt groups analogous to the G-groups in linear algebraic Ktheory which are obtained by factoring out exact sequences. As a basic result a Jordan-Holder theorem is proved for categories where all objects are of finite length. Using this theorem the computation of the Grothendieck group is 264