Professional Documents
Culture Documents
MANIFOLDS
AND
mUADRATIC FORMS
F. Hirzebruch and W. D. Neumann
Mathematisches Institut der Universitat Bonn
Bonn, Germany
and
S.S.Koh
Department of Mathematics
West Chester State College
West Chester, Pennsylvania
Appendix 11 by w. SCHARLAU
New York
1971
These Lecture Notes in Mathematics are produced directly from the author's
typewritten notes. They are intended to make available to a wide audience new
developments in mathematical research and teaching that would normally be
restricted to the author's classes and associates.
The publishers feel that this series will provide rapid, wide distribution of
important material at a low price.
70-176304
PREFACE
iii
COliTENTS
1.
Quadratic forms
2.
17
3.
25
4.
32
5.
36
6.
42
7.
46
8.
Plumbing
56
9.
75
Appendix I
References
Appendix lI, Grothendieck and Witt rings, by W. Scharlau
87
92
94
105
108
References
114
Author index
117
Subject index
119
1. QUADRATIC FORMS.
Let
A-lattice
(or A-lattice)
A itself can be
V is again an
V'XV ~ A
de fined by
f= (f,V)
over
to be a symmetric
bilinear map
f: V
where
V is an
XV~
A,
there corresponds a
linear map
f:
--+
V'
f,
given by
x,ye: V.
~
The form
for all
ye:V
=>
x=O.
2
Let
V'
dual base of
and
be a base in
(e i )1=1, _. ,r
and
d ~J
..
<ei,e j > =
characterised by
the
(e:).
1 ,_.,r
~
~=
for each
Lo<
. .e~J
j
'P(e.)
~J
M = M = (O(ij)
f
(e ) ,
i
and its
of
will be called an
rank
(f,V)
form of
= (x1 '
...
,x )
r
(e )
i
V.
and
f
= (Y1'
...
as a
'Yr)
is given by
f(x,y) = x t My =
If
(e.). 1
~
~=
,r
P,
M=
det f
of
to be
obtained from
M of
P~
detM ,
f
domain
then
of a form
Let
and
det f
= (g,W)
ish an isomorphism
x,ye: V.
f = (f,V)
u: V~ W with
Such an isomorphism
(f,V)
and
rela-
is well deA.
f _ g,
if there ex-
f(x,y) = g(u(x),u(y
for all
by
(e )
i
be the matrices of
and
f.
with respect to
3
some bases. Then clearly
LEMMA (1.3):
and
i.e.
and
Let
V1
f1
and
V S V
1
2
M
g
and
f2
(e ,
1
the matrices
f11& f 2
M, M , M ,
1
2
= ptMgP
M
f
IJ
f _ f 1$ f 2 '
decomposes into
,e )
s
( :1
and
then
V1 SV 2 '
are related by
~J
is given for
f, f , f2'
1
of
with
defined by
If a base
If
are equivalent if
Their ~
V2 '
f1
or that
cp :
is invertible, or equivalently,
det f
is a unit in
A.
M of
If
is
a field, the properties of being non-singular and of being non-degenerate coincide. If the restriction of
is non-singular, we say that
on a sublattice
is non-singular on
V
1
V
1
of
Notice
If
which is non-singular on
V1~V2
V ,
1
of
f1
to
4
V1
V1 61V ,
2
where
M1
f.
(e 1 , ,ek ; ek + , ,e )
1
s
is of the form
the matrix
M of
of
is the invertible
matrix
p
(: _~'Lt)
:;
then
ptMP =
r:
_LM- 1L t + N
1
11
A is a
Then
1tf.
The set
Ca). If
0<.,
A'"
fJ
of all units of
and -< + ~
,2
0(
A-m.
A coincides:.with
are non-units in
and
S'l
A then
o(p
are non-units.
over a
local domain
tion,
then
2 s A'"
5
Proof:
Let
~J
with respect to
unary form splits off by (1.4). We can thus assume that no diagonal
entry is in
0<1i e: A*
Since
A*.
for some
det f
is a unit,
OC'11 ~ A*
implies that
i,
(1.6)
o(ii e:
~11~22
has determinant
~12
e: A*
m,
0(12
= 0(21 e: A* ,
2 e: A*
shows that
is invertible, and
IJ
f(a,a) =
0(
e: A* ,
u (x) :=
a
is an involution, that is
is an automorph of
f = (f,V)
u u
a a
2f(x,a)
0<
= id
If
u
ae:V
V--.V
is such
given by
a - x
fixed. Such
Let
2 e: A*.
suppose
f = (f,V)
x,y e: V
If
a= x+y
f1e h -
then
which interchanges
= -u a
e(
e: A*
and
= 2(f(x,x) + f(y,y
y.
e: A*,
a= x-y
or
2 e: A*
Suppose
and
u = u
JJ
of
f(x,x) = f(y,y) =
and
satisfy
f(x-y,x-y) + f(x+y,x+y)
f(a,a) e: A*
the relation
f1 -
with
and let
f1
non-singular. If
f2
f1'
ely. Then
f2
and
W has rank
is an isometry
are defined be
1;
let
u: V1 EPW ~V2ew.
V ,
1
= w
which carries
restriction of
The composite
w
vu
into
to
w
V
1
W respectiv-
and
(f E9h)(w,w)
2
= h(w,w)
V
2
vu: V1 $W
of
f2&h
~V2EPW
It follows that
vu(V )
1
=
h
with
is an isometry
;;;
V
2
so the
f 1 - f2
JI
7
forms. In view of theorem (1.5) we therefore state:
this section
in the rest of
2 e: A*
by
A.
(e.g. if
= e ),
A*/A*2.
In particular, if
A can
A*
A are classified by
of
A*.
A*/A*2
be represented by elements
over
and
A is
f(x,x) ). Let
by the ex-
and
2
+ e:y r+s
f..., g
k$,r.
i f and only i f
In particular if
A = 1R,
we may take
e:"' -1,
m-s = r-k =
o.
This gives us
the number
(called the ~ of
f.
f)
JR.
is determined up to the
DET f e: A*/A*2.
over
A,
det f
Now assume
matr~ces
an d
(10 0)
l'
f..., g
if and only if
is even;
(oe: 0e:)
DET f = DET g
in other words
THEOREM (1.13):
If
m- s
r-k
We have shown:
and
are equivalent if and only if they have the same rank and same DET. 11
Condition (1.12) can be put in the more convenient form:
e:x + e:y
= 1
has a solution in
A.
is a solution of (1.12)',
then
(x,y)
= (a,b)
is a
m=
{O}).
For instance
m,
a).
~,
b).
~,
C,
is a (rational)
prime.
c).
IF
= ~p:iZ,
where
is an odd prime.
:iZ , Q{p) ,
p-adic integers (p
an odd prime).
We may regard
as
Q{p)
b ~ 0 (mod p )
Q,
a/b
E:
:iZ
Q wi th
:iZ
E:
:iZ
p
in
In fact
pt
in
:iZ.
p
The examples
= 1R
(qJp)
A= C
and
= IFp
p-1,
for
(qJp)
E:
A*/A*2 ~ C
2
so
A*
A*
is the
We define the
by
=1
(qJp) = -1
From (1.13) we have:
COROLLARY (1.15)
ratic forms
and only if
!3:.2.!?! :
and
If
g
A = IF
A are equivalent if
:iZ.
p
This
10
follows from the following lemma.
LEMMA (1.16)
in
If
0(
,fS
2
2
e: IF; then o<x + Py = 1
is soluble
[ 0, 1,
of
IF
p
Let
The maps
i, j :
H~
be the subset
IF
, (p-1) /2 }
lF
p
given by
j(y) = 1-Py2
are both injective. Hence their image sets have at least one element
IJ
A = 1lIp
, the ring
Let
f;; (f,V)
where
p is odd or
Proof:
p = 2,
x e: V
then there is an
(O<1}
ges to a. limi t in
111
if
(0(
n+ 1 -
with
if
is odd and
n -+ 0 as
0< )
f(x,x)
p = 2,
since
f(x,'X) = c
n-+
oD
f(x,x) - c '" p u ,
3 if p = 2
Define
~ 1.
1 2t 2
u f(x,x)
= i;P
according as
x e: V with
w = 1 or 3
and
1lIp
Then
(mod pt+1 )
'21
is
11
By this means we construct a sequence
in a sublattice of rank
element
Sf
Let 111-
in
and
in
x= x 'x ,x ,
o 1 2
such that
f(x,x) = c .
p.
converges to an
1I
221.
[ xi
of elements
Then
the
If
JT : 221 --+ IF
p
p
Proof:
that the map
phism. For
(1.17), with
8,
For
0(
ITo(
is in IF
p
221*/221*2--+ IF*/lF*2 ~ C
induced by 1T
p
P
P P
2
p = 2 the same argument gives
2212/221 2 -
We have
implies
is an isomor-
(2Z/82Z)* / (22I/82Z) *2 ~ C xC ,
2
2
modulo
e: 221; ,
If
0(
and
(221/8221) *
where
221/8221
IJ
(o<Jp) := (TTco(Jp)
Then the above argument shows that
Co( Jp) = -1
(o(Jp) = 1
if
is a square and
0(
otherwise.
If
0(
,{1 e: 221;
= 1
is
221
p
is soluble in
the corollary follows by (1. 17)
JI
2Z
by (1.16), so
12
As in the case of IF
COROLLARY (1.20):
of the same rank over
DET f
;; DET g ,
that is,
.-
and
for
0(,
P e:
JJ
p=2,3,5, ,or
sy~
p=oa
is soluble in
._{ .1
otherwise.
-1
(~r2,pc2) p
2.
(o,,B)p
= (~,P) p
= (3,O<)p
(0<,(3) (0<,)$) = (O(,p~)
. p
p
p
(0<,-0<)
Remark:
=1
Property 1
a map
~/~2
F2
- the non-degeneracy is
13
The calculation of the Hilbert symbol is given by the following
properties:
5.
(o<,P)QD
= 1
unless 0<
6.
write
(0<,(3)2
Here
(2\0<)
and
0<
f3 in the form
is of course defined as
Observe that for
then
and
p ~ 2,
or
-1
for
according as
the case
=b =0
if
p ~ 2
if
=2
x e: :lZo2'
(_1)x
is just
7.
If
a, b e: Q*,
then
TT(a,b) = 1
(a, b) p
=1
=2,3,5,
and
,00.
Our next goal is to define the Hasse-Minkowski symbol for a quadratic form over
toC )
LEMMA (1.21)
field
K.
Then
Let
f,g
may
be carried into
is of the form
by successive applica-
tion of binary transformations such that at each stage the form remains
diagonal.
]4
~
Let
and
is by induction on
r.
The case
=2
M=
lation theorem (1.8) and the induction hypothesis we see that for
r> 2
diag(p1,l(2' '~r)
into a form
can be transformed
can be realized by
~i's
/31
= L~'f~1
i 1 1
=(
JD
W1' th
ij )
RtMR -_ H.
S'1nce
~ 0,
U=
carries
M into
goes into
terms.
M'
UtMU =
,,2
0(1 = 0(1)11+
t><.,2
2>21
,..(2 =
~
~3'
~2)
M',
using f;1
'1
1n pace
=1
'>11
.~:tI.e
by permuting
U is non-singular since
,o<'r'
which
~ O.
We can
'21'
0<
l'
p(
2'
Now let
f
define
be
Mf
~.
We may diagonalize
= diag(c(1'
'~r)
We
15
(1.22)
c (f) :=
which is either
on
f,
or
:=
p (0<1' ,0<)
r
-1.
c (f)
Indeed, for
Clearly, for
,Pr)
is another diagonalization of
r =1
we define
c p (f) :=
1
.
c (f) = 1 or -1
r =2
depends only
H
f
f
That is, if
then we claim
according as
has
,0<)
r
is invariant
c (f)
COROLLARY (1.23)
\.
Let
and
f2
ThEm
A quadratic form
over
f1
over
111 clP
p
1>'
since
1lIp
c (f)
p
is also defined.
then
If
f
c (f)
P
=1
11
16
c (f)
Q.
for
p = 2,3,5,
,00
if
c (f) = 1
p
If
ITc
(f)
2,3,5,
,00
p P
where
ranges over
and
=1
11
2.
Let
A the subgroup of
b 1 Eli1b
- b1 - b
2
+,
(b ,b e: B).
1 2
Grothendieck group of
Let
The group
G(B):- F/A
is called the
B.
any semi-
group homomorphism
h: B--;.G
uniquely over
j i
of
factors
commutes.
EXERCISE (2.1)
j:
B~
G(B)
is injective if
B.
17
Kw(X)
18
manifold
X,
(see e.g.
endieck), Bull. Math. Soc. Franc., 86(1958),97- 136). It was this example which first made the importance of the above construction so
apparant. A closely related example is the following.
EXAMPLE (2.2):
Let
the semi-
X,
usually denoted by
KO (X)
(resp.
e.
G(B)
is
..
examples further here (see for instance Atiyah, Hirzebruch; Riemann-Roch theorems for differentiable manifolds, Bull. A.M.S., 65 (1956) ,
276- 281).
Let
and
and
G(~ 0
(A
A.
will be denoted by
G(A)
and
r o (A)
Go (A)
if
A is suit-
ably restricted (see (1.8. Thus by (2.1), one can expect the calculation of
Go(A)
A is a field.
The rank of quadratic forms induces a homomorphism
19
called the augmentation. If
A-lattices
form over
V1
and
V
2
V1 A V2
f1
and
f2
f1f2
to be the
characterised by
G(A)
and
rk: G(A)~~
A.
Go(A).
form
e;(1) = 1 .
rk
G(A)
Clearly
so if
(l(A)
defined by
splits, and
(l (A)
o
G(A) = (l(A)
is the kernel of
Ell ~.
where
Similarly
rk: G (A) ~ ~
0
DET
G (A)
o
DET
since
= (l
to
(A) E9 ~
(l (A)
o
=Im E:
Since
in the representation
20
where
L(A)
is the kernel of
DET
restricted to
(A)
This sequ-
ence does not always split, as is shown by the examples discussed below.
Suppose now that
r o (A),
2 e: A*.
so by (2.1)
Then the
11 (A) eG (A)
0
A*/A*2,
G (A)
o
generates
we have
A*/A *2 eG (A).
Go (A)
generate
EXAMPLES (2.5):
~:
G(m)
a-1 E: ~ (A)
with
and
Furthermore,
~ 1li51li
G (A) ).
o
as a group.
G(e)
~ 1li.
A=m then
(-1) - 1,
in
1li.
For
G(m) ~ 1li
as a group, generated
11
G(~)
a rational prime.
If
(1.18),
1li;/1li;2 ~ C
is in
and
pE:-1,
~/~2 ~ C2 x C2
Indeed, by
~(~)
[ 1,E:}
~/~2
is represented by
e:-1 ,p-1
and
21
= (s-1)(p-1).
For
,p
0<
<=>
(O<,P)p = 1
(o<,P)
=1
E:
~/~2
we have:
= 1 +o<.{$
c<+p
in
G(\)
=1
. has a solution in
\.
that
(0(0
,...~l
(1
0
OO(P
and
Now by property
(e:lp)
= -1
and
modulo 4
(p,p)
Hence in
= (-1Ip) = 1
or -1
2s
s + p ~ 1 + e:p
For
2p ~ 2
in
G(\)
p:: 1 (mod 4)
this gives
2(e:-1)(p-1) = 0 , (s-1)(p-1) ~
( e:-1) - (p-1
G(~)
=2
e:+p ~ 1 + sp
2p = 2
=1 + 1
p = 1 or 3
piii3 (mod 4 )
2s 2
we have
G(\)
p=1 (mod 4 )
where
according as
11
= e:ps
-1 -1
= 1
o.
2(e:-1)
Also
=0
, 2(p-1)
=0
DET(s-1)(p-1) = DETe:p-1)-
111
ED
a(\)
-=
111
E9
111/2111
(s-1 )
E9
111/2111
(p-1)
If)
111/2111
(s-1)(p-1)
where the element under a group shows a generator of that group, and
the last summand
111/2111
is
L(\)
22
p" 3 (mod 4 )
For
G(\)
we have
111
<H\)
111
If)
111/2111
(e-1)
111/4111
(p-1)
Ell!
-2(p-1).
= 1.
Thus
Therefore
~(\)
is generated by
4(p-1) = -2(e-1)(p-1) = 0
Also
The element
2(p-1)
so
C ;
2
C
2
C
2
sent
C ,
2
and
represented by
by
(p-1)
(e-1)(p-1) =
(p-1)
has order
As we know,
{1,e,f,d'}.
respectively modulo
alone.
in G(\)
1112/11122 ~
~~2 ~
Then
(1,e,S,d' ,2,2s,2f,2sS' }.
and
L(\) ~ 1li/21li
so
p = 2.
let it be represented by
and
(e-1)
L(\) ,
generates
whence
ep + p = 1 + s ,
We can repre-
8,
and then
(s,s)2-- 1 , (e'~)2=1 ,
.. _-/
Writing
for
1+1 e G(~2) )
= 1 + ef
= 2
2t = 2
e + e15 = 1+t
e~
+ 1 = 1 + d~
....=
111
1
and
L(~) ~ 111/2111,
the reader.
111/2111 19
111/2111
(f -1)
(t-1)
generated by
2(e-1).
EP
111/4111
(e-1)
23
We conclude this section by giving an invariant definition of the
Hasse-Minkowski symbol. For each finite prime
let
L(\)
Identify
111/2111,
with
c (a)
p
For
p=oo
(i.e.
c'{a) = 0 or 1 ,
so
and define
= (_1)c~{a)
\=m),write
asG{m)
as
a=ot1+/X-{-1) ,
and define
c... (a)
We shall show that if
a = f
= { _1)0(- (o<- -
1)/2
of
c (f)
If
rk a
for
{rk a)1.
Hence
= DETf + DETg -
DETfDETg - 1
By lemma (2.6)
(DETf,DETg)
where
RHS
= {_1)RHS
24
(2.8)
c (f)c (g){DETf,DETg)
p
p
p
c (f)
decomposes,into unary
p=
00
the
proof is similar; alternatively by direct computation from the definition of 1, using property 5 of the Hilbert symbol.
An immediate consequence of the above definition and the exact
rk a
DET a
of
and
G{~)
c (a).
p
~,
finite, is com-
In particular,
p
finite, is
11
Let
AcB
is a form over
A,
denote by
characterized by
write
for
fB
according as
fB
as an
A-module. If
(which
over
We shall simply.
A = 111, Q or
m.
Let
V ~ AB
..rB(XGllb,yGl/c):= bcf(x,y).
~+
0<-
~+
~-
and
of
0(+
+0(-
= rk f ,
the rank of
f.
by
:=
0(+
and
D(
0(
0(
~(f)
ned also for integral and rational forms and forms over
are defi-
Q(p) = Qn 1lIp ,
f (x, x)
lE
0 (mod 2 )
f = (f, V)
for all
E:
over
V,
25
111 or
otherwise
1112 is called ~
f
is called odd.
if
26
Since
(3.1)
f(x+y,x+y)
= f(x,x)
+ f(y,y)+ 2f(x,y) ,
is even if and only if the diagonal entries in its matrix are all
even.
Consider a quadratic form
= (f,V)
over
1li or
we: V,
in general not
unique, with
f(x,x)
For if
w
e 1 , ... ,e
= ~ wi e i
f(x,w)
is a base of
'
!I
and we write
xe:V.
x = 2:xiei
'
then
f(x,x)
(mod 2 )
and
f(x,w) = Lf(e.,e'>x.w .
i,j
1
J 1 J
Hence relation (3.2) is equivalent to
= 2:
f(ei,e.)w.
j
J J
and since
of
det f ~O (mod 2 ) ,
modulo 2
Such an element
f.
(mod 2 ) ,
Clearly
e: V
only if
~
for sOlle
z e: V.
= w+ 2z
if ant
27
so by (3.2)
f(w,w)
(mod 8 )
f(w,w)
Let
THIDREM (3.5):
2 1112
111 or
1112
(3.6)
f(w,w)-r-detf +1!! 0
where
= rk f
~:
is the rank of f
Observe that
Sinc e
2212 ,
a form over
1li
det f
modulo 4
=(f, V)
ment, and
a = det f
where
det f
is de-
is well defined
111
2
a:x.
(mod4) ,
w= 1
is odd. Thus
=1
then
is the form
is a characteristic ele-
Suppose now
r = 2.
Then
has
matrix
2a c )
( c 2b
with
w
=0
odd since
4ab + c
Thus
fi
f,
andsince
f(w,w) - r - det f + 1 =
+ 1 ;: 0 (mod 4)
How assume
ran r 1
c 2 ;:1 (mod4),
is odd. Thus
is a characteristic element of
o- 2 -
of
detf
,J
and
for
f = f 61 f2
1
is a decomposition of
respectively. If
i
=1,2,
then
w
i
= -161 w2
into forms of
is a characteristic element
is a characteristic element
28
of
f.
Hence
f(w,w) - r - det f + 1 =
f1(w1,w1)-r1-detf1+1 +
+f2(w2,w2)-r2-detf2+1 +
det f1
and
det f2
are odd, so
= 0
(mod 4)
f.
<r
a e: 111 ,
2
we defined
(_1)a
over
as
or
-1
11
according as
thus define
1112
221
2
~:
over
c (:)
2
If
f1
and
f2
f = f 1 9f.
are forms
Hilbert symbol
Thus by (1.23)
and
c.
29
(3.8) is proved once we prove it for unary and even binary forms.
(3.5).
=a
det f
If
since
e: 1li*2
f
and we choose
is an even
bina~y
2 - 1 ;;; 0 (mod 8 ).
~
congruent over
If
w=1
a = b=0
one of
a,
and
b,
~2
congruent over
say
c2 (f)
so
form we choose
If
is unary then
1 = c (f)
2
w=O
so
=C2 (f).
is
so by property
is
to
so
c (f)
2
= (2a,(4ab- c2 )/2a)2
= (2a,(4ab-c 2 )/2a)2(2a,-2a)2
2
= (2a, c -4ab) 2
It is now a routine matter to check that
0(
c (f) =
2
c 2 (f)
using property
(210<) = 1 or -1
according
11
1" = 't'(f)
THEOREM (3.9):
Q(2) = 1li2 n ~.
Then the
is defined.
For a non-singular quadratic form f
over
Q(2) ,
30
~:
Let
(-1)
1:"
...+
and
=0(+ -
0(-
0(-
= c<+ + 0<-
and clearly
sign det f
In view of theorem
4
_1:"+ r + sign det f - 1
4
2oC- + (-1)
0(-
- 1
4
(mod 2 )
Since
= (_1)0(-(0<- -
c .... (f)
Now suppose
1)/2,
det f =:1:1
c (r) = 1
by lemma (1.25)
c2(f)c~(f) = 1.
THOOREM (3.10):
If
With
so
if
(3.9)
=0
f(w,w) -'t' :; 0
If
11
11
(mod 8 ) .
w= 0
is a characteris-
tic element, so
COROLLARY (3.11):
If
'L(f) :: 0
(mod 8 ) .
11
We remark that this last result is the best possible in the sense
that there actually exists a quadratic form satisfying the hypothesis
with
't" = 8.
Indeed let
E8
31
Construct an integral matrix
M=
(~ij)
J1ij
=2
=1
i f the vertices
f1 ij
=0
otherwise.
.Mii
by the formulae
1~iS8
v. and
~
joined by an edge in E8
v.
are
Thus
2 1 0 0 0 0 0 0
1 2 1 000 0 0
01210000
001 2 1 000
00012101
00001210
00000120
000 0 100 2
1:'
= 8.
E8.
= (f,V)
n.
min f
e , , en
1
min f
=0
we sq f
is a
of the lattice
the form
is given by
f
If
If
has
represents zero.
Wi th respect to a base
f
{If(x,x)11 xsV}
La ..
= . . 1J X.X
1 j
1,J
(e ) of V
i
such that the expression (4.1) is Hermite reduced. This term is defi-
2).
If
= a 11 x 12
a)
1a111 ~ 21 a 1 j 1
b)
1a 11 1 =min
c).
a 11 f =
for
is Hermite reduced.
>1 ,
(~a1ixi)2+ f 1 (x 2 ,
reduced of rank
n-1
32
. ,x )
n
where
is Hermite
33
The following theorem gives a bound for
expression for
in an Hermite reduced
f.
THOOREM (4.2):
n
la11i
If
then
min f <
COROLLARY (4.3):
::::'5
or -
If
2:x~
2:x~. 11
~
repre-
dropped;
THOOREM (4.4) (Meyer):
form of rank
represents zero.
f )
r o (111)
and
-1
~ (111)
defined in 2.
in the semigroup
f.
_x
and
G (111)
o
G (111)
o
will be denoted by
represented by
E:
conventions in 2.
LEMMA (4.5):
Let
1ad-1) 6lg,
=(f, V)
with
non-singular.
34
is a zero form.
f{x,x) = o.
maps
x e: V
Let
Since
ists aye: V
with
z e: V
choose a
with
f{z,z);;; 1 (mod 2)
visible in
by
V,
one of
and
and
fey' ,y')
f{y,y)
f
is
is odd, say
e = y-{m+1)x
2
are indi-
11
If
(8
"t{-1)
By lemma (4.5)
1E9g
e = y-mx
1
'P of
f{y,y)
f{e ,e ) = 1 = -f{e ,e ) .
1 1
2 2
is equivalent to 0(+-1
~:
If
as
THEOREM (4.6):
V'.
in
y'= z+ (1-f{x,zy.
The elements
f{x,y') = 1 .
f{y,y) = 2m+1.
~(x)
<<f'{x),y> = f{x,y) = 1
(-1)E9g
1 $ (-1) g.
Since
11
by a trivial induction.
COROLLARY (4.7):
is equivalent .to
11
G (111)
o
THEOREM (4.8)
~:
For
f e: ~o (111),
either
and
=::r
generate
G (1lI)
1. f
or
(-1). f
f = ~+1 + ~- .(=1)
in
G (111)
o
11
is odd and
111(8111,
11I c lR,
maps
35
Corollary (3.10), which states that
Remark (4.9) :
:: 0 (mod 8)
(4.8).
h: F (?l) ~ 1lI/8?l
o
The map
(mod 8)
h: G
f(w,w)-"t'(f)
defined by
(?l) ~ ?l/81l1.
identically on
generators
'1
(111)
=1
and
of
vanishes
vanishes on the
(111)
two
such forms are equivalent if and only if they have the same rank and
same index. Combining this with
THEOREM (4.10) :
(4.7) we have
equivalent if and only if they have the same rank, index and type
(even or odd).
11
THEOREM (4.11)
rank ~ 8
f' = f ~ 1 .
Then
f' and
E8
g = 1~ 1~
as defined in
$
(9
For example,
3, and let
same rank, signature and type. But they are not equivalent as
f' (x,x) = 1
has only
solutions but
g(x,x) = 1
has
18.
1':
V ~ V'
of
is a direct summand of
non-degenerate form
DETf
by
de't f
Rad(V)
l'
on
V/Rad(V)
DET f
and
V,
for if
of the correlation
and
We shall denote
induces a
det1'
=0
for all
f =
x,ye: V)
and
f
is
then
DETf=1.
THIDREM
over
where
fi
(i = 0, ... ,k)
COROLLARY (5.3):
(p
-t 2)
r i::! 0
lZp
Proofs:
be the matrix of
= 1,2, ,r)
36
is
37
divisible by
in
M = ptM"
write
that
7Zp
where
we say that
M'
= (0<: .)
is divisible by
is not divisible by
~J
We can
Notice
a).
p,
then some of them are units. Applying (1.4), we may split off these
entries.
b).
0<:. ,
~J
p,
then at
minor
where
r.
of the forms
f.
f.
is divisible by
p.
elm -t 7Zp
In
7Zp
for
i~1.
Since
cr:2 S
and hence by
pt7Z
is clearly
r.
-t ct1 s:
f.
The number
rt '
Lr.
giving an
= rankf
p -t 2
(a}
where
of times
f 7Zp /Otj
is isomorphic to
fi
we define
't i
=(det fi I p)
for each
det f i
i
so
is a uni t.
For
38
= DET f i
det f
if
det f. {22I*2
f;
22I*/2Z*2
V \
't i
We can consider
in
V Cl/ \
U = V+/V
f.
and let
V+
be
i. e.
'
V+ = {xe:VCl/\lf(X,y) e: 22Ip
Then
then
t.
(1,-1},
f = (f, V).
the "dual"
with
i
~
lence class of
e: 221*2
P
if
is a finite abelian
for all
ye:V}
p-group, and
induces a bilin-
ear pairing
where
by
(x e: U1pix;o O}.
U.
quotient
W.~ = U./U.
1
~~-
p ,
is of order
IFp
as
t
in-
Therefore
Identify
)f i =
IF */IF *2
p
p
nET f~
(i ~ 1)
with
{1,-1}
Indeed let
so
nET f~
f.=(f.,V.)
~
nET f:
and
lFp
;t 1
is defined.
We claim that
he as in (5.2). Then
39
f.
is just
't i
the
reduced modulo p
and for
and
det f
to
it follows since
Observe that
The invariance of
DETf:
;0
to
ll'1' t2 ,
mined by
'6.
so
i~1
to(f) = ~1(Pf)
f~
and
r.
is deter-
= rk1:,
~
giving
r .
~
The integers
THEOREM (5.4):
and
'6i
(i
~O)
constitute a
complete set of invariants for the equivalence classes of non-degenerate quadratic forms over
2Z
(p 12).
11
The following lemma is not hard to prove (c.f. Jones [11J p91).
For
pl2
over
2Z
p
For
p ;i 2
f
f
and
and
\.
For
=2
if they have the same type (even or odd) and are equivalent over
~2'
11
is equivalent to
THOOREM (5.6):
over
2Zp
for
p = 2,3,
,DO
,00
2Zp
for
and
p = 3,5,
40
Proof:
g
~2.
and
detf:; detg
Thus by (2.9)
c. (f)
2
over
=c 2 (g)
lZ.oo
and
rk f = rk g
implies
and hence
c 2 (f)
=c (g)
2
ilt
over
THEOREM (5.7):
11
DETf = DETg
If
and
det f = det g ,
Remark (5.8):
denote the invariants
r.
VI/t
and let
lZ
p
V+
U: = V+ IV
ilt2
11
by
of
and
Namely, consider
be the "dual" of
VCVI/t
V+:= (xe:Viltlf(x,y)e:lZ
Then
over
and
1.
f = (f,V)
Let
f- g
Since
as a form defined on
given by
for all
ye:V}
I det f I ,
and
p,
U~p)
1.
= (u e: U
I piu = O}
and
W~p)
1.
U~p)
/U~p)
1.
1.-1
then
41
L
1f~p)
~
and
M be a
V:= H
2k (M,lZ)/Torsion
is a
form
SM
~(SM)
~(M)
for t::..tch
H2k(M,~)
In particular for
2k (M,~ M;~)
=4k
, i
=2k,
which correspond to
42
we have
respectively.
43
Since
o(v~
we may consider vp
e: H4k (M,dMjlZ) = 7Z ,
intersection number of
and
SM
into
SM
by cup product.
7Z
Show that if
Let
M over
7Z/22Z
7Z/2lZ.
H*(Mj7Z/27Z)
The quadratic
S(x,y)
=xvy
w = w (M)
2
2
x e: H2 (M;7Z/27Z). We
H2(M;2Z/22Z)
with
xvx
= xvw2
for each
w
2
is actually the
M.
=0
If
M is a 4-manifold with
.)M = t;1
and
then
-r(M)
!!
(mod 16 )
44
Let
11":
H2(Mi7Z)
~ H2(M;7Z/27Z)
If
then
quadratic form of
.!:!:.2..21.:
where
SM
is such that
is the integral
istic element of
11'.
de: H2(M;7Z)
~,
is a character-
11
-r:(M);;; 0 (mod 8)
is of course
~ 7Z ~ lZ --+ 7Z/2lZ ~ 0
of
i
H (Mj7Z)
2.
If
have no elements
de: H2 (Mj7Z)
so there exists a
d,
considered as elements of
w
2
=0
with
1Td .. w
2
2
H (M;7Z)/Torsion,
SM.
We conclude that if
SM
M has no
is even. Thus we
45
THIDREM (6.2a):
BM
is even, then
If
11
Ea
~ith
given in 3 cannot
no 2-torsion.
then
If
fold
that is
G -t 0 ,
with
II X = t;1
H. (X;G) 'it G
~
this implies
i = O,k,
dim X
=k
and
H. (X;G) = 0 otherwise. I f
~
LEMMA (7.1)
with
a).
H1 (Y;7Z)
b).
By
has no 2-torsion;
be a
be as in lemma
(7.1). Define
- 1:'(y)
}'(X) := ~
that is
reduc ed modulo
THIDREM (7.2)
morphism type of
~(X)
E:
QV7Z ,
1.
X.
46
the
l!-invariant of
X.
This
47
is a special case of the ft-invariants studied by Eells and Kuiper in
If
is a
a).
Hi (X;7Z)
b)
Xis a
~ -homology sphere;
c).
is a
i #'O,k ;
7Z2 de-
2
~ H. 1 (X;7Z/27Z) ~ H. (X;7Z) ~H. (X;7Z) ~
~+
If
i #,O,k
then
H. (X;?l) ~ H. (X;?l)
~
is an epimorphism, so since
the groups involved are finitely generated, a) follows. b) is an obvious consequence of a), so it remains to prove c).
7Z2
Hi (X; 7Z )
2
m
Hi (X; 7Z 2 ) ':' ~ 7Z.(a k
(7.5)
where
cr1 ~ ()(2 ~
~ am c 7Z2
are ideals in
7Z2
2 7Z
j
j
7Z/2 7Z ':' 7Z/2 7Z.
2
7Z by
48
Hi (Xj7Z 2 )
Hi (Xj2Z 2 )
Hi (Xj7Z 2 ) = 0 for
is epic for
i;l O,k.
I t follows that
11
i;l O,k.
Y.
X.
Here
Let
Y1
and
M=Y 1 U-Y2
-Y 2
is indepen-
j4.(X)
Y2 be two 4-manifolds as
pasted together along the com-
of course means
Y2 with orientation
M has, after smoothing if necessary, a differentiable
[17], [18]
We claim that
1).
M has no 2-torsionj
2)
SM
3).
S~ EI1-S~ = S~
where S~ is defined as in 3.
Y1
Y2
M
Before we prove these properties, observe that they suffice to
is even;
M = Y1 U -Y 2
~H.
to obtain
(X)
~H.
(Y 1 ) $H. (Y2)
~
~H.
(M)
~H.
~-
1(X)
X is a
~-homology
sphere, so (7.6)
gives
0~H2(Y1;~)
It
H2(Y2j~) ~H2(Mj~) ~O
49
Taking orientations into account, it is not hard to see that
proving property 3).
Next we take integral coefficients in (7.6) to obtain
HO(X,7Z) = 0
where
Y1
and
Y
2
H (Xj7Z)
1
follows that
H1 (Mj7Z)
groups (see for instance [23] p245, Satz m) now proves property 1).
Finally, we take coefficients
7Z2
obtain
Tor(odd torsion,7Z ) = 0 ,
2
H (Y j7Z )
Recall that
and
s'f2 is defined on
~
SM
11
e2
Let
and
is
50
(n, q)
=1
for each
S3.
~ Z/nZ.
Then
Z/n:1Z
on
by
is,
S3.
Z/nZ
Since
L(n,q)
has
S3
formations,
L(n,q)
at dimension
are
1,0,0,1
If
Clearly
L(n,q)
0 < q < n.
LEMMA (7.9)
modulo
n,
Furthermore
Proof:
S3
carries
L(n,q)
onto -L(n,n-q) ,
since
11
Since clearly
X,
}L(-X)
= -fl(X)
n/q> 1
L(n,q)
where
takes even
51
n/q = b
-~1_ _ __
2
-
bs
(notation)
where each
bi
that if
is odd and
is an integer with
bi
~2
Ib.1
~
It is easy to prove
even. Deno te by
b. 's
~
and
-=
in this unique
expansion.
RECIPE (7.10) :
( (n,q
,.ML
=p
+ -p
16
= [2,2,2,2,2,2]
7/2 = [4,2]
Therefore,
~(L(7,6
= 3/8
~(L(7,2
= 1/8
j-<-(L(7,1
= -r(L(7,6 = -3/8
= 5/8
(mod 1 ).
L(n,q)
n.
L(7,1)
or
L(n,q')
-qq'
have the
is a quadratic
11
and
L(7,2)
qq'
and
and
52
Now let
call that
X1
X1
and
and
X
2
X
2
X1
and
X
2
W such that
~W
= X1 U -X
(disjoint union)
THEOREM (Z.12):
Let
X
1
and
X
2
W.
Let
= ~Y.
X.
(i
= 1,2)
I
\
\
is a deformation retract of
Y ,
2
SN = SY2
so
(Z.1) with
~N
=X1
W,
1:(N)
It follows that
L(Z,1)
= - 't'(1Yi
be
is homotopy equivalent
is a manifold as in
and
(mod 1 ) == r(X 2 )
L(Z,2)
and
X
2
11
Hence
= - 1'b (mod 1 )
n-manifolds
are as in lemma
() (D
to
X2
-X2 X2
X
2
Y.
pasted along
Since
where the
of two
*,
03
is a commutative
53
semigroup with identity. One easily proves the following
THIDREM (Z.13):
As a map
If3 ~ CV1li,
)I..
is a semigroup
The operation
r3
11
is even a homomorphism of
)I..
03
into
Q,/7Z.
JR4
Let
then )A(X) = 0
JR4.
We now give the promised proof of lemma (Z.1). Recall the well-known fact that every 3-manifold is parallelizable (Stiefel [26]).
Therefore, by Milnor [1Z] (see also M.W. Hirsch [10J), every 3-manifold wi th empty boundary bounds a simply connected 1I-manifold. Let
X
be a
fold wi th
Let
~Y::
X.
since
M be the double of
X.
Y,
We claim
that is
w (M)
2
M = Y U -Y
=O.
Indeed
pasted along
w (Y)
2
=0
54
Mayer-Vietoris sequence with coefficients in
~/2~
6, that
BM
By
is even, as was to be
11
shown.
=X
be an oriented
and assume
and
2k
and this is
2).
(coefficients in ~ ). Let
of
where
is defined on
V = H (y) ,
2k
V.
correlation for
group, so
H2k(y,X)
By
By,
and
ye: V}.
cokern(f) ~ H _ (X)
2k 1
is non-degenerate. Furthermore
is essentially
(5.8),
Define, as in remark
V+
and
H2k-1(X)
Now
cP
= V+/V
By
,
is the
is a finite torsion
Hom(H
2k
(y) ,~)
55
U.
5y
L: U xU~
CV7Z
~~p)
~
and
r~p)
~
of
5y ,
p an odd prime. However, as the informed reader may already have observed,
(see
L ,
Y1
and
X.
Y2
and
52
of
equivalent over
JY1
and
JY
Y1
and
1li ....
= lR ,
Y2
then
51
and
52
8. PLUMBING.
S2.
SO(n) ,
Sn
SO(n)
bundles over
1L _ SO(n)
n 1
Sn = (x E: :nf+1 I Ixl = 1 }
mn+1 Sn is the union of the two discs Dn = (x E: snl
xn+1 ~O }
+
n
} , which intersect in their common boundary
Dn
_ =(
X E: S I xn+ 1 ~ 0
Consider
n l xn+ 1 = 0 }
Sn-1 = (
X E: S
Sn,
If
is a principal
SO(n)
in
and
bundle over
let
be trivialisations of
map
is of the form
( t, x) -
f: Sn-1 ~ SO(n).
( t, !( t) x)
where!
56
[f]
is a map
E:TC _ S0(n)
n 1
is the class-
57
Let
'1
= (E1'P1'S~)
Sn.
= (E2'P2'S~)
and
;2
Let
D~ c S~
~
be two oloien:;ed
j1
and
~2
E.ID~
~
and
f (y,x)
2
for
E1
i = 1,2
and
E2
for each
n
(x,y) e: D XDn
x e: f 1 ( Sn-1 XSn-1)
lR.
'
where
lR
+
the transformation
26+1t'
2B+1I:
(r cos(t}) ,r sin(e t---+ (r cos -3,r sin -3-)
=0
and
E2
is
58
j.~
n-disc bundle
1't'n_1S0 (n)
v.
To each vertex
We then plumb
joined by an edge of
Vi '
of
Vi
v.
T we choose an
of
ji
~j
whenever
I f several edges of
v.
of
Vj
are
and
m.
n
D
in
S~
~
to
gives:
P(T,m.)
~
59
From now on we restrict
Y
= P(T,mi )
to be even, say
=2k.
Let
(T,m ) ,
i
and
and let
= aY
2k
v VS
2k
of
times).
copies of
Ho (Y;2Z) ~ 2Z
S2k,
H (Y;2Z) ~ 2Z$ $ 2Z
2k
and
where
has a basis
Ji
m
i
sn
by
is the
e(m ) ,
i
n2k_1SO(2k).
since it
There
=2Z
-P.,
is just
where
Milnor [15] or theorem 35.12 in Steenrod [25]). The "classical" definition is essentially that
e(m.)
S2k C E.
a , ,as
1
of
H (Y)
2k
Sy
is given by
with
1
if
i!j
and
v.
and
connected by an edge in
v. are
J
T ,
0(
~J
0
e(m. )
~
if
i!j
if
i =j
of
otherwise,
11
60
Since the above quadratic form depends only on the tree
(T,e(mi
(T,e(mi
weighted in
lZ,
X. JY.
Namely let
H.~ (XjlZ) =0
M above.
for
i~0,2k-1,2k,4k-1,
and
Ker f
~H.(y,X)~Hi 1(X)~H.
i ~ 2k,4k.
Hence
1(Y)~
By Poincare-Lefschetz dual.ity,
has no torsion,
~-
H.(X)
= 0
~
for
and since
i ~ O,2k-1,2k,4k-1.
In the
11
SY'
61
det M ~ 0
In particular if
I det M I
finite of order
Thus if
then
H2k (X)
det M
det M =:1:1
is odd,
then
is a
is a 2Z -homology
X is simply
det M
= 1
implies that
k = 1,
dim Y
that is
is
is even a
=4,
this is
The principal
'"'1S0(2) :;; 2Z,
generator in
S2.
SO(2)
bundles over
S2
are classified by
S3~S2 corresponding to a
SO(2)
with
S1).
By the remarks
e: "'1SO(2)
of a principal
2Z,
SO(2)
bundle over
with
S2
2Z
2Z,
(T,m )
i
weighted
Y '" p(T,m )
i
is equal
(T,m.)
~
62
when weighted by
-2,
v~----'v~~t---v"'; - - - - - - -
--V":
v~------------v:
'.
is weighted
-2
TABLE
T
H1 (ap(T); 221)
22I/(s+1) 221
Ds
1't'1 (dP(T
{ ",/2",. "'/2"'
221/4221
.... }
, s odd
=: FT
s+1
D'
s-2
221/3221
T'
E7
221/2221
w'
E8
I'
FT
s/16 (s even)
E6
The groups
)o({Clp( T) )
6/16
8/16
S3 (. unit
Cn ,
63
Dn (not to be confused with the tree
octahedral and icosahedral groups
double
uoverin~
subgroups of
SO(3~,
of
S3
en
~nat
D~
of
Du' T ,Wand
SO(3)
H1 (JP(T
(s~,
a
T = E6
T = E8
and
~P(T)
is
is
negative definite and even for each of the trees listed and is equal
to the quadratic form of
~1dP(T)
P(T).
relations, however these groups can be obtained much more easily from
the general results of
briefly.
An oriented 3-manifold will be called a Seifert manifold if it
X,
S2
with fibre
S1
of integers
for each
0<./(0<. _ A.)
~
'~i
i=1, , r .
~i/(~i - ~i)
~i
r~
"l (1) ]
si
64
and let
Then
= ~P(T).
of lens spaces via plumbing, which we shall study later in this section.
Remark:
folds and lens spaces are opposite to ours, so his weighted trees are
the negatives of those here. We discuss various orientation conventions
occurring in the literature in the appendix.
of
s3,
S3/F
is a Seifert
S3/I'
S /FT
S1
oP(T)
For instance,
= 5/4
3/(3-1)
= 3/2 =
[2,2J
2/(2-1)
= 2/1
= [2,2,2,2J
r .. 3
of
and
65
By the above, we get the star-shaped tree:
-2
-2
-2
-2
-2
-2
-2
L2
But this is
Ea
hedral space if
I'.
~P(Ea)
in other words
Ea
is weighted by
-2.
n1~P(Ea)
In particular
The other cases in the table may be checked in the same way.
To investigate plumbing in more detail, we need a precise description of ~he bundles involved. We shall identify
so that the operation in
S1
by
m = (X m,p m,s2)
represented by a map
~~,
S2
s()
Tr S1
with
is written additively.
S1 bundle over
S1
of degree
-m.
Hence
as follows:
X
where
= D2 X S 1 U D2 X S1
f
f: d(D2XS1)=S1XS1~S1XS1= ~(D2XS1)
is the map
D2XS1
-x
mayalso
(-1 0)
-m
66
space
L(-m,1) ,
denotes
so
Xm = L(-m,1).
-L(m,1) = L(m,m-1)
S3~S2,
L(-m,1)
is by observing that
(z1,z2)
~[z1,z2]
S1 bundle
S3
when
q = 1.
Using the
The principal
is negative,
X = L(-m,1)
m
-m
in standard notation.
If
j(m)
~(-m).
is just
dP(T)
rather than
called plumbing;
operator
1).
If
~P: T~ ~P(T)
m2
will also be
Let
be the tree
Let
itself, where
m,
define
= Xm
~P(T)
2).
P.
P(T)
2
) = (X ,p,S)
m2
D2
be 2-discs
2
j(m1) = (X ,p,S)
m1
and
X I Di
mi
ffii
and
D1
X I (S2 - Int D. )
mi
~
XmilDi
Denote by
posite map
f:
where
,)X~1
f- 1
~(Xm1ID1)~)(D1XS1)~d(D2XS1)~dX~2
t: d(D1XS1)=S1XS1~S1XS1 = J{D2XS1)
is defined by
67
t(x,y) = (y,x)
3).
Then
peT)
= X' U x'
after smoothing.
m1 f m2
The map
Denote by
f),.
[0,1]
81
x'
m1
m2
where
1
1
= D2 X 8 U AX 8
f1
l::.x 8
1
f.: 8 1 X 8 1 ~ 8 X 8 1
1 U D2 X 8 1
f2
(-1 0)
-m
i
for
i = 1,2
Thus
~P(T)
1
1
= D2 X 8 UAX81U~X81 U D2 X 8
t
f1
f2
1
1
= D2 X 8 U D2 X 8
f
pt
where
f = f2ot of1
(-1
-m 2 1
68
More generally let us consider the weighted tree
..m1
m2
(As,m ) 1 ~ i ~ s
i
..
ms _1
ms
----~.~ ------------.~----
where the
m
i
S1 XS1~S1 XS1
f.
the map
with matrix
(-1 0)
-m. 1
~
'"
U 6x S1 U D2 X S 1
f
t
s
(8.3)
where
f=fot-f
S
s-1
1
U D2 X S
f
D2 X S1
...
J.(-1 0)
et
we see that
-m.
-mi
( -1
has matrix
1)
0
has matrix
-q p.)
(n q
:=
-m
101)
-m
( -1
0). (-mS_ 1 1)
(-1
s
-1
'"
(-m 1 1)
-1 0
0; (8.4)
69
so we have proveu
THIDREM (8.6)
~P{A
Then
(8.5).
Let
(A ,m.)
s ~
,m.) = L{n,q),
~
where
and
11
This shows
np + qq' 1
In particular
qq'
is
1 (mod n ) .
and
L{n,q'),
so
L{n,q) = L{n,q')
= q,
~P{A )
and
0 < q < n,
q,
= L{n,q)
AO
=n
AO
A1
= a1~1 = a 2A2 -
A2
o ~)..2 <).,1
A,3
o ~A3 <'\2
n/q
= [a 1 ,a 2 ,
n/q
70
Using (8.7) we get by induction
(a 1)
s
-1 0
(8.8)
(a. 1)
-1 0
so
(:~ ~) (:~~) =
Comparing this with theorem (8.6) and the remarks following this
theorem we obtain
THEOREM (8.9) :
where the
4-manifold
tree
(A ,m.)
s
If
As
a.
L(n,q)
(A ,m.)
s
with
m. = -a .
~
As the integers
a~
when the
n
q
~-in-
""1 = q
(8.10)
Ib s 1>0
where each
b
i
= (A ,n.)
s ~
with weights
n.~
= -b.
71
As above, one can easily show that
JP(A ) ~ L(n,q) ,
s
As = -1.
L(n,q)
bounds a 4-manifold
P(A )
s
and
p-
In this construction
quadratic form
Sy
= p- -
r(Y)
Sy
is
-"t'(Y)/16 (mod 1)
can be calculated as
even though
p+.
where
respectively
(8.11)
-b
Mc
1
1 -b
I > 1.
Then
Proof:
Let
1:"(M)
= p- -
If
c
i
I b. I > 1
~
diag(-c ,
1
,-c)
s
is congruent over
where
= b. _ _1!.-_ _ _.,....
~
b i + 1 - _..:!.bs
for all
then clearly
sign c
an integer prime to
= sign b i
which
11
lR
Suppose that
be an odd integer
is
72
the prime decomposition of
via the Legendre symbol
It is clear that if
(qIPi)
and
n.
q'
(qln)
is defined
by the formula
(that is,
qq' = x
qq'
is a
has a solution in
then
(qln) = (q'ln)
(qln)
is an orientation preserving
L{n,q)
with odd
n.
Notice
spaces.
THEOREM (8.14):
16jd:L{n,q
If
is odd and
=- Q (mod 4)
+ (n-1)
coprime to
then
and
Proof:
even. Then
(8.15)
and
L{n, q) = - L{n,n-q)
the theorem also follows for
to the case that
Expand
n/q
is even.
accordingto{8.1Q),thatis
We have shown that
L{n,q)
Ib l>1
i
and
bi
bounds a 4-manifold
Y peA )
s
i=1, ,s.
whose
73
quadratic form
Sy
= -Sy
then
s
2
Lc.y.
i=1
~ ~
c.
j-tL(n,q) = -"t'(Sy)/16
1:'(s)/16
(mod 1)
p.
assume
is a prime factor of
= :tn
we have that
then
n,
=1
p =Pi
c (S)
p
say
'
where
Since
(q,n)
=
=1
in the form
and
=:tq,
n.
by (1.24). We hence
ptq.
where
det S2
Write
det S
Hence if
c (S)
so since
cp(S)
It follows that
ptq ,
Also, since
S1
is a
Since
and
(n, q)
=1
we may write
and by prop-
74
fl
Hence
-1--1.
cp(S)
p odd pr~me
t
IT
i::1
fJ
(S.
(qlp.) ~(signdetS2Ip.) ~
~
n/ q > 0 ,
so
Applying lemma
(1.25), we get
(8.16)
c (S)c",,(S)
2
Combining (8.15), (8.16) and (8.17) now easily proves the theorem.
11
M = M(n)
plex analytic subset is said to have maximal dimension if it has complex co dimension
N is given locally
phic function
ponent of
defined on
p,
otherwise
in
and a holomor-
such that
U of
can be chosen as a
The point
pe:N
is
function
coo~dinate
is called singular.
N is called irreducible if it is
on
=~
"'"n.N.
~ ~
75
fi
{fi}i e: I
is defined an an open
76
subset
U
i
of
(a)
{uilieI
(b)
fi
(c)
fi/fj
and
is an open covering of
M,
Ui '
and
uinu j
We express this by
{f }
i
is holomor-
[D]
over
=(f)
and
D _ {fi}ieI
ND
D _ {f}.
= If =0 I
For such a
c ([D])
1
c 1 (D)
=0
H.
H.
is a theory with
ports (see for instance Borel Moore [5]. A singular theory of this
type, defined on the category of locally compact spaces and proper
maps, is given by allowing infinite, but locally finite chains). The
main property of
K.
77
duality isomorphism
M without
boundary.
Using this, one can generalize the intersection numbers of cycles
to the case where one cycle may have non-compact support. Namely if
x e:
K . (Mj2Z)
and
m-~
y e:
H~
(M;2Z) ,
x.y
define
= 6(x) (y)
y.
A(x)
on the
homology classes they represent. One can check that this gives the
usual definition in the compact case.
Returning to the complex manifold
written as
= ~niNi'
if
D is a divisor
Ni
has
h(D) e:
H2n _2 (Mj2Z).
Borel and
Haeflinger proved:
LEMMA (9.1):
h(D) e:
If
:a2(Mj2Z).
h(D)
= (t).
= 0 , so certainly h(D). x
considering
D as a cycle on
= 0
for any
M we have shown:
= c 1 (D)
x e: H (Mj 2Z).
2
= 0 ,
Hence
78
If
COROLLARY (9.2):
D.x
D = (f),
Now let
M(n).
11
The com-
in
in
complex manifold
M.
"'pM;
M by
Cp(n-1)
to
obtain a new
of
in
centered at
given by
(Le.
z(p) =0).
r
be the graph of
f,
c U X Cp(n-1)
and
K : = {p} X CP (n-1) c U X O:P (n-1)
p
Then
let
cover
U X Cp(n-1).
U X Cp(n-1)
given by
By normalizing so that
w.
=1
Cp(n-1) ,
The
we can
consider
to be local coordinates in
N
is given by
Vi
79
SO
N n Vi
in
r;;;
N- K
p
d' M
p
we identify
M - {p}
and
depends only on
U- {p}
to obtain
M and
with
<5 M
P
and not
which maps
M-
onto
and maps
biholomorphically onto
ifM-K
{p}
Complex Dimension 2
In the following
Let
on a neighbourhood
of
in
z1' z2
M.
(centred at
p)
2.
defined
u,v
and
u,~
6 M
P
such
that
=0
and
=0
of
compact support in if M
cycle with
80
LEMMA (9.5) :
Proof:
ing,
is a function defined on
a function on
~;1 (U)
given in the
(u,v)-chart by
z1
on
K oK
P P
n- 1 (U)
p
-1.
(Strictly speak-
is
and in the
z10TCp
as
which is
(u,v)-chart by llV
1 z1 = 01
in 1"C- 1 (U)
P
1z1 = 01 = K + lu= 01
p
is clearly
+1
consists
and
I'u = 01
gives
0= Iz 1 =01oK = K oK + lu=ol"K = K oK +1
p
P P
P
P P
This clearly proves the lemma.
Remark (9.6):
M=
and consider
The
blowin~up
11
as a
C~-manifold,
then
K1
and
K2
M and consider
then
M is diffeomorphic to
p= P1
be a point in
K
TC;~ (p) and blow
P1
by K. ;
Denote
~i
:re
:=Tt" on:
: O'.p p M~ M maps K1 U K2 onto p. We now olow
P1 P 2
P2 P1
1 2
up at a point P3 e: K1 U K , and so on. We finally reach a complex
2
manifold
1t
= cfP1 ... P
K1 U U Ks.
81
We can construct a weighted tree
v l'
K.~
by joining
t Vs
K.
and
v.
and
vertices
by an edge in
v.
J
K.~
in
T
v.
i f and only if
-intersection number of
of
with
The self-intersec-
K.
1i- 1 (u)
where
is a coordinate neighbourhood
M.
ities in the
8.
Example (9.7):
ction
Let
M = (1:2
and let
defined on
M.
singularity at
= or -1( P1)
M up at
= z31 + z42
P1
of
or:
where
M.
has a non-uniformizable
Rf~M
is the projection
Iw= 01
f
occurs. Blow up
u,v
q'P1M
as a function on
lu 3 (1+uv 4 ) =01
=01
is
P1
given by (9.4). We consider
1~3('f+lr3)
f;
P1 = (0,0)
The origin
Iw=OI
in c1'p1M
and
M at
u,~
(see remark
is clearly
82
1u
=01
and
K1
Iv=OI
3,
in the point
We represent
3 __
-------1-(~IV-+-U-3-=
__
0_1_
- K1
----
and
1 2
f = w / (considered as a function on
where
(u,v) = (0,0).
M) is at the point P2
P1
Since this point does not appear in the (u,v)6
u= u 1
satisfying
w to the (u,v)-
v 3 (v + u3 )
take local coordinates
and in
,v= u 1v
1
I w = 01
or in a diagram:
is given by
Iv 1 + u 1 = 01
Next let
be a point where
(u 2 ,v 2 )
and
(u 1 'v 1) = (0,0).
(u2 'v2 )
in
Blow up at
6p1P~3M
P3
satis-
83
is given by
Iw= 01
and
respectively.
IU 2 + v 2
Finally we blow up at
manifold
by
IU
M = 6'
P1 P 2 P 3P4
12 3(
v3 v3 + 1) = 01
3
=01
P4
where
(u ,v )
2 2
= (0,0)
Iw = 01
8 12
11r3 ~ 3 (1 + 1r ) = 0 I
3
and
to obtain a
is expressed
In a diagram this
is
1
The function
f
= w1/2
K1
in
and
L .. I v3 + 1 = 01
K4'
K4
Rf
of the function
K;
covers
K1
once and
K1
K;
K4
and
and
f
L'
on
M,
K4
sticking
L' covers
84
divisor
D =
I f= 01
Rf
in
1
I
11
1
1
1
1
K'4
I
1
L'
where
and
K;
I
I
K'
cover
K
2
AJ.l
K; U UKG
is just
E6
KioKi
=1,
,6).
Note that if
K:
and
(i ~ j)
K'
j
K L' = +1.
K:.K'. = +1
~
J
= 6 + 3K;'K;
o = K4D = 3 + 1 + 4 + 4 + 6K4K4,
= K;.D
K;.K; = K4oK4
showing that
K:K: = -2
~
= -2.
-2;
= K; U
Rf
AJ.so
Therefore
UK6
is
(E6 ;-2)
V is a tubular neighbourhood of
of (the modified)
E6
then
in
V is diffeo-
85
morphic to
P(E ;-2)
6
~V ~ s3/T,
particu1ar
T'
where
(1:2
onto
and let
C2
1
f- (O) = q
R
f
Then
of
is a
R
f
(I:
xC =
a: 3
The boun-
~V ~ S3/T,.
( z1,z2,z3) e:
on
satisfying
f,
that
3
4
2
z1 + z2 = z3
z; +
z~
, /
z1 (z~ +
z~)
q= 0
M = (1:2.
of this surface in
Show that for
(n ~ 2) , / z; -
z~
(n ~ 2) ,
f ..
a:3
/z1(z~+ z~)
-2.
K ,K ,
1 2
pair
(di~Ki = 2)
,Kn
i,j
(i
j)
the curves
K.~
and
0(. .
~J
We have.:
M = (~ij) ,
86
THEOREM (9.9)
definite.
This is essentially a classical theorem (see Du Val [27J). The
proof we are about to give is due to D. MUmford [19J.
Proof: Let
1 = f - f(p)
where
11=01
of the curve
in
i=1, ,n
2:.o(~ .x. Xj
S' _
~J
~,J
Clearly
if and only if
(9.10)
if
i~j
(9.11)
(j
= 1, ,n) ,
and
LO<:
j
i
~
(9.12)
Write
< 0
for some
j.
L<><i'.x.x.=
. j
~,
J ~ J
L(Lo(j)X~- LO<i'(x.-x.)2
j
'<j
S is negative
x. =0
J
for some
= x n .11
boundary. Let
2
H (Mj2Z)
re:
and let
2
de:H (Mj?li)
duality
~ H2 (Mj?li!2?li)
1rd=W (M).
2
corresponds to an element
Under Poincare
be: H (Mj?li).
2
Theorem (6.5)
states that
THEOREM:
ding
f:
If
S2~M
is congruent to
Proof:
f(S2)
-bundle
in
~(M)
bob = (d"d)[M]
modulo 16
bb = -1.
tubular neighbourhood
normal
S 1-bundle of
M1 = (M - Int A) U e 4
is attached to
~A
whose boundary
f(S2)
t
where
= S3
~A t
f(S2)
M a
S3 ~ S2.
87
SO(2)-
considered as the
has in
Then
fJe 4
Put
= S3
88
Clearly
-reM)
w2(~1)
=0
and hence
1,
by (6.2). Since
particular case.
bob = s
_CP(2)
M if necessary. Let
0,
P1' ,P s+1
by reversing
be
s+1
and let
Let
where
denotes a generator of
cc
= bob
H (P ;221)
s+1
+ Lg.og.
i=1 ~ ~
Then
= -1
embedding of
S2
in
S2
in
M'.
can be represen-
Since
is
shows that
-r(M') - -1
Since
we deduce that
1:"(M) ;:
as was to be proved.
11
(mod16)
(mod 16 )
89
EXAMPLE (10.1):
Let
M = S2 X S2
2
H (M;:iZ) = 2Z19:iZ
SM
of
and
D(
,p e: H2 (M;2Z) be the
and let
if and only if
of an element
m
and
m= 2m' , n = 2n'
mC<+ np e: H2(M;2Z)
is equal to
mc<+ np
can be represented
2<>(+ 213
is not representable
that is
m'n'
is even. In particular
Let
M = G:p(2)
M.
and let
g e: H2 (M;:iZ)
be the
is obviously rep-
nge: H2(M;:iZ)
k 5 0 or 3 (mod
(2k+1)2
=1
(mod16);
4)
Cp(2) ,
and let
zo,z1,z2
P(zo,z1,z2)
ng
M=
be homogeneous coordinates in
90
degree
G:p(2)
in
the function
~o
consider
Zo
P = O.
If
Q=
, so
and
for
ng.
of degree
P=0
is dual
In particular let
The curve
P=O
(1,0,0).
Using
n-1 ,
in the 4-manifold
Cp(2)
P=O,
-1)/2
P P
m (m
of the
maps
c:J. 1 ) into
P= 0
ng,
n~
1 ,
by a
91
n
=0
and for
n ::; -1
X8
combinatorial embedding of
REFERENCES
[1J
[2J
OJ
m,
[4J
------
[5J
[6J
[7J
[8J
[9J
[10J
[11J
[12J
[13J
[14J
ditto
Am.
[15J
[16J
92
93
[17J
------
[18J
------
[19J
[2OJ
R. von Randow, Zur Topologie von dreidimensionalen Baummannigfaltigkeiten, Bonner Math. Schriften 14, Bonn, 1962.
[21J
V.A. Rohlin, A new result in the theory of 4-dimensional manifolds (russian), Dokl. Akad. Nauk. SSSR 84 (1952) 221224.
[22J
[23J
[24J
[25J
[26J
E. Stiefel, Richtungsfelder und Fernparallelismus in Mannigfaltigkeiten, Comm. Math. Helv. ~ (1936) 3-51.
[27J
P. Du Val, On absolute and non-absolute singularities of algebraic surfaces, Revue de la Faculte de Science de l'Universite d'Istanbul (A) 91 (1944) 159-215.
[28J
APPENDIX I.
1. Definition of
~adratic
Forms.
V is
case for all rings considered in these notes, then finitely generated
projective A -modules are free, so this is no change.
If
quadratic form
on
is defined as a map
q:
such that
is bilinear;
(0{
E:
V~A
and
A)
is non-singular or non-degener-
94
95
form.
If
is a unit in
A,
(1)
b(x,y)
determines a symmetric
= b q (x,y)
q(x) = ~1 b(x,x)
(2)
However if
2 t A*
If
= 221
or
2212
2Z and
F2 '
we
G (A)
defined in 2 would
The analogously
GU (A)
o
the associated
96
symmetric bilinear form
leads to a map
GU (A) ~ KU (A) ,
which is an isomorphism if
The quotient groups of
2 e: A*
KU (A)
and
obtained by setting
hyperbolic forms (see Appendix lI) equal to zero are called the Witt
groups and denoted by
B(A)
and
W(A)
Grothendieck and Witt groups has taken enormous strides since these
notes were first written. In particular many connections with other
fields such as algebraic number theory and algebraic K-theory have
developed. We refer the reader to Appendix II
more information.
3. Integral Forms.
Niemeyer [22J.
4. Symbols.
G(~).
is a field, a symbol on
is a
97
bimultiplicative symmetric map
F*
into an abelian group
for
1.
Hasse - Vlitt
xF*~
C
2
satisfying
(a ,b) "' 1 ,
c(f)
or
f
by
in 2,
so if this
K (F) ,
2
K2
is generally difficult to
F, given by setting
(a, b),
(a,b)
[27]
98
oriented manifolds
M1
and
M2
then
in the equi varian t case) can be found in Atiyah Singer [3 J , see also
Jani.ch [14J.
r- invar-
r- invariant
to detecting h-cobord-
ism type of lens spaces in special cases (7) has been superceded by
a complete h-cobordism classification (Atiyah Bott [ 1 J p479). It turns
out to be the same as the diffeomorphism classification. A general
reference for this sort of classification problem is C.T.C. Wall's
book [3OJ
= P(T,m.)
~
99
homotopy sphere if and only if the intersection form
equal to the form of the weighted tree
%1.
(T, e(m.
~
SY'
which is
has determinant
k-spheres for
Sk
#=.
Those
subgroup
2
for
for
k;;; 0 (mod 2 ),
or
k _ 1 (mod 4).
Clearly, if one only plumbs stably trivial bundles, then
P(T,m.)
~
is stably parallelizable, so
an element of
bP
2n
Sy
generator of
bP
if a sphere, represents
(n even) or
dY,
(n ~ 3)
Y =
dY
E8
represents a
7.
In
we described
~-invariant
with invariants
100
is a
modulo
on the
if one simul-
X by plumbing. In particular if
X as the boundary
X is a
X = dY
of a manifold
m
i
Y = P(T,m.)
~
z~ + z~
"nasty" points (points where the "Riemann surface" has a non-uniformizable singularity) until none are left - will work for instance for
functions of the form
f(z1,z2) with
(z1~-q)1/n
=1).
101
more, it turns out that essentially the above "naive method" can be
used to reduce the general problem to this latter case, giving a general method of uniformizing such singularities. For details see
Hirzebruch [9 J.
As we indicated in
broid function
z~ + z~
z~ + z~ = z; in (:3.
clearly isomorphic to
This variety is
s =
and
Us..
'"
whose intersection
(T,-bi):
, b s J
J
102
where the
f3 k
are given by
0 <fk < ~
~3.
and
ns5
where
S5
is the unit
are
1
(t e: S ) ,
c~
a.
mi
is a :iZ - homology
(T,m.)
~
mi
bundles with odd euler characteristic for this, which only exist in
dimensions
2, 4 and 8
103
necessarily pairwise coprime) one can "replace" the singular point by
~(a1,a2,a3)/s1,
n
z3
= z1z2n-q
cribed at the beginning of this section. One thus also gets a star-shaped dual tree, but the central curve usually has higher genus and
the tree more branches than in the coprime case. The details of this
method are a bi t messy and Orlik and W.agreich (loc. ci t.) use a
similar but more economical method to resolve in fact all 2-dimensional
normal singularities which have an effective
C*-action.
'~+1)
X(a 1 ,
in
a: n+1
givenby
z~1+ +z::'1 1 =o
(ai >1),
= X(a 1 ,
,an +1 ) ns2n+1
bP2n
Ki
collapsing
is negative definite;
K
M(2)
such that
K is
M' obtained by
normal singularity at this point. Thus the classification of 2-dimensional normal singularities is essentially reduced to the determination
of the dual graphs which can occur, together with the possible complex
structures in a neighbourhood of such a union of curves
K.
This is
104
a very difficult problem and is not solved in general. A connected
account will appear in the notes by Henry Laufer [17J ; see also
Brieskorn [5 J for a discussion of the case of rational si ngul ari ties.
9. Orientation Conventions.
As mentioned in 8, von
Seifert spaces and lens spaces are opposite to those adopted here, so
his weighted trees are actually the negatives of the trees we gave when
quoting his results on Seifert spaces and lens spaces.
In these notes we have tried to follow Raymond's orientation
conventions of [26J and [23J. This is also the convention adopted in
[21J. In the original mimeographed version of these notes (apart from
a sign error at one point) and in Orlik Wagreich [24J the opposite
convention has been adopted for Seifert spaces (but not for lens
spaces). This leads to compatible results on obtaining the manifolds
~(a1,a2,a3)
105
only i f it is dual to
coprime. For
classes of
2
2
S XS
~p(2)
me( + nf3
(notation of (10.1
classes in T(n(M2n)
can be so represented.
m and
0, g, 2g
In dimensions
with
H2 (M) =
~2(M)
REFERENCES
[1J
M.F. Atiyah, R. Bott, A Lefschetz fixed point formula for elliptic' complexes, IT Applications, Ann. of Math. 88 (1968)
451-491.
-
[2J
[3J
[4J
E. Brieskorn, Beispiele zur Differentialtopologie von Singularitaten, Inv. Math. ~ (1966) 1-14.
Rationale Singularitaten komplexer Flachen, Inv. Math.
~ (1968) 336-358.
[6J
[7J
[8J
[9J
F. Hirzebruch, Uber vierdimensionale Riemannsche Flachen mehrdeutiger analytischer Funktionen von zwei komplexen
Veranderlichen, Math. Ann. 126 (1953) 1-22.
[10J
------
106
[11J
F. Hirzebruch, K. Janich, Involutions and singularities, Algebraic Geometry, Papers presented at the Bombay Coll. 1968
Oxford University Press, 1969, 219-240.
[12J
[13J
[14J
[15J
[16J
[17J
[18J
[19J
[20J
[21J
\-l.D.
[22J
[23J
[24J
P. Orlik, P. Wagreich, Isolated singularities of algebraic surfaces with C* action, Ann. of Math. 22 (1971) 205-228.
[25J
[26J
F. Raymond, Classification of actions of the circle on 3-manifolds, Trans. A.M.S. 121 (1968) 51-78.
[27J
.2
Neumann, S 1-actions and the 0( - invariant of their involutions, Dissertation Bonn, 1969, to appear as Bonner
Math. Schriften 44.
-,
- .. ~'Y_I"
C
__d,
_ .............
C',,. '....,
................
..., _
lite 'It ..................1'1 .. .....
DI1 ..... ..-a.
~I"_"'''''''''''''''''''''
APPENDIX II
W. Scharlau
Mathematisches Institut,
Universitat Munster, Munster, Germany.
1. Generalities.
Let
space over
projective
R-module
b:
MxM~
is a pair
(M,b)
KUo(R)
q:
M~
is a pair
R-module
(M,q)
M and a quad-
con-
b (x,y):=
q
is a unit in
GUo(R) ~ KUo(R)
GUo(R)
which is
R.
108
109
follows.
Let
b:MXM~R
q:M~R
is defined as follows:
m{M,b)
MED M*
(x, f)
q{x) + f{x)
or
Both forms are non-singular. The hyperbolic forms are the special case
~ m{M,o)
E,lJH{M,-b)
m{M,q) EPJH{M,-q)
~ m{M,o)
E9JH{M,-q)
and
so the subgroup of
m{M,b) or
]H{M,o)
KU (R)
m{M, q)
or
The quotient group by this subgroup is called the Witt group and
denoted by
B{R)
and
Remark: KU (R)
o
W{R)
and
respectively.
B{R)
110
and
GU (R)
is in a natural. way a
W(R)
B(R) - module.
2. Dedekind Domains.
From now on let
(Tr) ==
'1
char (F)
over
be a
T'C
12.
by non-singular
with
ai,b
in the form
e: R - , .
The form
'd 2 (b):
B(R/f)
= diag(b 1 ,
,bn)
gives a
sequences
O~
B(R)
B(F)
~B(R/l)-+O
and
0 - + KU (R)
~KU
(F) LB(R/1O) - + 0
0 0 1 '
d2
class form.
The element
d (b) : = diag(a 1 ,
B(R/1)
,~)
is al.so well
class form.
THIDREM:
If
is complete and
canonical. isomorphisms
char(R/1)
111
B(R) ';;: B(R/~)
Furthermore the above exact sequences split by means of the first
residue class form, so one has direct sum representations
Li terature:
char(F) ~ 2
and
THIDREM:
1J E:
B(R/f)
Max{R)
Frohlich OJ.
THEOREM:
where
GU (R)
o
assigns to an ideal
ot
of
the element
C/C2~
JH(ot) - E(R)
of
GU (R).
o
Proof:
112
3. Algebraic Number Fields.
The calculation of the Witt group has been done so far only for a
few fields. One of the main results is
THEOREM (Hasse - Minkowski)
Let
is injective, where
places.
The image of this map can also be determined.
Li terature: O. T. 0 I Meara [13J, Wi tt [26J.
Let
Then the first exact sequence of the preceding page can be extended to
an exact sequence
1E: Max(R)
B(R!y) ~ C/C2~O
the infinite primes and the quadratic reciprocity law, can be found
in [8J.
Example:
O~
B(2Z)
B(Q)
.LL
B(2Z/p2Z)
~O
p=2, 3, ...
On the other hand
so
B(~)
113
of the notes).
FUrther Literature: Frohlich [3J, Milnor [12J, Scharlau [aDJ,
C.T.C. Wall [25J.
4. FUnction fields.
A similar situation to the case
function fields
F = k(t)
exact sequence
B(k[tJ/p) --+0
p irred.
B(k[tJ)
Since
B(k)
B(k)cB(k[tJ) ,
= B(k[tJ)
it follows that
then
5. Characteristic Classes.
One can define invariants for quadratic forms which are in a
114
certain way analogous to characteristic classes for vector bundles.
This can for instance be done by means of Galois cohomology in a very
similar way to how it is done for vector bundles, see Scharlau [17J,
Springer [23J. The definitions vary a little according to whether one
works with the Witt group or the Grothendieck group.
Literature: Witt [26J, O'Meara [13J, Scharlau [17J,[18J, Milnor
[11J.
6. Pfister Theory.
Let
be a field of characteristic
2.
The theorems of
B(F)
= W(F).
W-z:(F)
of
W(F)
is a 2-group
where
W(F.. )
~ 2Z
(observe that
W(F)
is not
is a 2-group.
REFERENCES
[1J
[2J
115
[3J
[4J
W.D. Geyer, G. Harder, M. Knebusch, W. Scharlau, Ein Residuensatz fUr symmetrische Bilinearformen, Inv. Math. 11
(1970) 319-328.
--
[5J
[6J
[7J
[8J
[9J
[10J
[11J
[12J
[13J
(1970) 318-344.
1971.
[14J
[15J
[16J
[17J
[18J
[19J
[2OJ
116
[21J
[22J
[23J
[24J
[25J
[26J
AUTHOR INDEX
Numbers in square brackets are reference numbers and indicate
that an author's work is referred to although his name is not
cited in the text. Numbers that are underlined show the page
on which a complete reference is listed.
A
Atiyah, M.F.,
.1Q2
~,
1Q2
Cassels, J.W.S.,
1Q2, 106
~,
74, 92
Du Val,
P., 86,
22
ill
117
118
106
o
O'Meara, O.T., 112, 114, 112
Orlik, P., 101, 103, 104[231,
104, 106
22, 96,
122
69, 93
~,
100, 106
116
SUBJECT INDEX
Equivalence of forms, 2
Euler number, 59, 61
Even form, 25, 30
F
Base of a lattice, 1
Blowing up, 78
Binary form, 2
Binary polyhedral groups, 63
Brauer group, 97
Genus of a form, 39
Grothendieck group, 17, 95, 108
of analytic sheaves, 18
of vector bundles, 18
Grothendieck ring, 19
Icosahedral group, 63
Indefinite form, 33
Index of a form, 8
Integral form, 25, 32, 96
Intersection number of cycles,
42, 77
Irreducible variety, 75
Isometry, 2
119
120
Jacobi symbol, 72
<
K-group, algebraic, 97
Kronecker product, 1
K-theory, 18
K-theory, algebraic, 108
L
Manifold, 42
Matrix of a form, 2
Minkowski invariant, 38
J{- invariant, 46, 62, 70, 97
N
Non-degenerate form, 1
Non-singular form, 3
o
Octahedral group, 63
Odd form, 25, 33
Orientation conventions, 104
P
Rank 0 f a form, 2
of a lattice, 1
Rational form, 25
Real form, 25
Reflection, 6
Regular point of a complex
variety, 75
Residue class form, 110
Riemann surface, 85, 100
S
TYPe of a form, 35
Tree, 58
Tetrahedral group, 63
U
Unary form, 2
Unimodular form, 30, 32
v-
Weighted tree, 58
Witt group, 95, 109
Zero form, 32