ERRATA FOR INTRODUCTION TO SYMPLECTIC TOPOLOGY

DUSA MCDUFF AND DIETMAR A. SALAMON

Abstract. This note corrects some typos and some errors in Introduction
to Symplectic Topology (2nd edition, OUP 1998). In particular, in the latter
book the statements of Theorem 6.36 (about Hamiltonian bundles) and Ex-
ercise 10.28 (about the structure of the group of symplectomorphisms of an
open Riemann surface) need some modication.
We thank everyone who pointed out these errors, and in particular Kotschick
and Morita who noticed the problem with Exercise 10.28.
p 32/3: In our discussion of the symplectic camel problem we are tacitly assuming
that n 2.
p 60: The formula for VolE should read
Vol E =
_
E

n
0
n!
=

n
n!
n

j=1
r
2
j
.
p 64: In line 3 of the proof we should have J R
2n2n
.
p 72: The formula for the unitary basis s
k
is incorrect. One should dene s
k
for
k n as s

k
/[s

k
[ where:
s

k
= s
k

k1

j=1
s
j
, s
k
) s
j

k1

j=1
(s
j
, s
k
) Js
j
.
p 74: The additivity axiom in Theorem 2.69 is stated correctly only when E is
a line bundle. For general vector bundles E
i
of rank n
j
, where j = 1, 2, it
should read:
c
1
(E
1
E
2
) = c
1
(E
1
) + c
1
(E
2
), c
1
(E
1
E
2
) = n
2
c
1
(E
1
) + n
1
c
1
(E
2
).
p 77, middle: This should read: Additivity follows from the identities:
det(U
1
U
2
) = det(U
1
) det(U
2
), det(U
1
U
2
) =
_
det(U
1
)
_
n2
_
det(U
2
)
_
n1
for unitary matrices U
j
U(n
j
).
p 82, line -4: The reference should be to Proposition 3.6 not Proposition 3.2.
p 87, line -7: It is clearer to write L
q
= (T
q
S)

instead of L
q
= T
q
S

.
p 109, line -4: Besides the stated condition a contact isotopy
t
should satisfy

0
= id.
Date: 14 December 2004.
2000 Mathematics Subject Classication. 53C15.
Key words and phrases. symplectic geometry.
The rst author is partly supported by the NSF grant DMS 0305939.
1
p 132/3: It is easy to deduce from de Rhams theorem and Hodge theory that the
Poincare duality homomorphism PD : H
nk
(M; R) H
k
(M; R) is well dened
and surjective. If it is not surjective then there is a nonzero harmonic k-form

k
(M) that is L
2
-orthogonal to its image; hence
_
N
= 0 for every closed
oriented (n k)-submanifold N M and so, by deRhams theorem, is exact;
since is harmonic this implies = 0. Injectivity now follows for dimensional
reasons.
p 167, line 14: H

is an element of C

(C
n
) not C

(R
n
). The moment map
: C
n
u(n) is related to H

by the formula H

(z) = tr((z)

) and so is given
by:
(z) =
i
2
zz

p 169, Ex 5.26: In line 5 the formula should be

g g R : (,

) , [,

]).
p 170, line -3 and -11: Replace Lie(G) by Lie(().
p 184: Delete the phrase regular value in the assertion of Lemma 5.51.
p 186, line 6: Replace p by x.
p 188, line -10: The sentence should start as follows: To see this let := (x)
(M) and approximate x by a sequence ...
p 204, line 5: B S
2k
should be B S
2
.
p 216, line 3: Integration over the ber gives a map H
j
(M) H
j2k
(B).
p 216, line -7:
t
(x) not
t
(s).
p 220, line -10: The reference should be to (6.2) not (6.1).
p 221, line 9: Replace by .
p 222: The identity in the assertion of Lemma 6.26 should read
([v

1
, v

2
]
Ver
)
bre
= (v

2
)(v

1
)d d(v

2
)(v

1
);
the rst identity in the proof of Lemma 6.26 should read:
d(v

1
, v

2
, Y ) = ([v

1
, v

2
], Y ) + ([v

2
, Y ], v

1
) + ([Y, v

1
], v

2
)
+ L
Y
((v

1
, v

2
)) +L
v

1
((v

2
, Y )) +L
v

2
((Y, v

1
));
the last displayed inequality in the proof should read
([v

1
, v

2
], Y ) +L
Y
((v

1
, v

2
)) = ([v

1
, v

2
]
Ver
, Y ) + (Y )d (v

2
)(v

1
).
p 228-229: Theorem 6.36 is wrong as it stands. A counterexample is the quotient
M :=
S
2
T
2
Z
2
where we think of S
2
R
3
as the unit sphere and of T
2
= R
2
/Z
2
as the standard
torus; the nontrivial element of Z
2
acts by the involution
(x, y) (x, y + (1/2, 0)), x S
2
, y T
2
.
The closed 2-form
= dy
1
dy
2

2
(S
2
T
2
)
descends to a closed connection 2-form on M; its holonomy around each contractible
loop in RP
2
is the identity and around each noncontractible loop is the symplecto-
morphism (y
1
, y
2
) (y
1
+1/2, y
2
). Thus M satises condition (iii) in Theorem 6.36
and the restriction of M to every loop in RP
2
admits a symplectic trivialization.
However, this bration does not satisfy (i): one can show that the holonomy of every
closed connection 2-form on M over every noncontractible loop in RP
2
diers from
the half rotation (y
1
, y
2
) (y
1
+ 1/2, y
2
) by a Hamiltonian symplectomorphism
and so cannot itself be Hamiltonian. Another way to see that (i) does not hold
is to observe that the classifying map RP
2
BSymp
0
(T
2
) is not null homotopic,
while Ham(T
2
) is contractible.
The mistake in the proof of Theorem 6.36 is the assertion on page 229 that the
homomorphism f

:
1
(B) H
1
(F; R)/

given by the ux of the holonomy of

around the loops in B lifts to a homomorphism

:
1
(B) H
1
(F; R).
If this lift exists then the rest of the proof that M B has a Hamiltonian structure
goes through. But there is a lift if and only if the ux of the holonomy of every
torsion element in
1
(B) is zero (in H
1
(F; R)/

). In the above example

1
(B) is
torsion and f

is nonzero. Hence the lift

f

does not exist.

We claim that the existence of the lift

f

does not depend on the choice of the

closed connection form . To see this, we x a symplectic trivialization : B
1
F
M over the 1-skeleton and a homology class a H
2
(M; R) such that

b
a = [
b
] for
some (and hence every) b B. Each such pair (a, ) denes a homomorphism

a
: H
1
(B
1
; Z) H
1
(F; R)
via

a
([]), []
_
:=
_
S
1
S
1
( )

a
for every pair of based loops : S
1
B
1
and : S
1
F. If is a closed connection
form in class a, then

-parallel translation around the loop in B

1
F gives a
path g
t

0t1
in Symp(F) that starts at the identity and is such that

a
([]) = Flux
_
g
t

_
.
Hence

a
does lift the pullback
H
1
(B
1
; Z) H
1
(B; Z)
f
H
1
(F; R)/

of f

to the free group H

1
(B
1
; Z). But we need a lift

f

that is dened on H
1
(B; Z),
and this exists if and only if

a
([])

for every loop : S

1
B
1
that
represents a torsion class in H
1
(B; Z).
It remains to prove that the latter condition is independent of a. Let : S
1
=
R/Z B
1
represent a torsion class in H
1
(B; Z). Then there is a positive integer k
such that the loop s (ks) is homologous to zero in B. This implies that the k-
fold multiple of the cycle () : T
2
M is homologous to a cycle v : T
2
F
b0
in the ber. To be explicit, note that there is an oriented Riemann surface
with two boundary components
0
and
1
, dieomorphisms
0
: S
1

0

(orientation reversing) and

1
: S
1

1
(orientation preserving), and a smooth
map u : B such that u
1
(s) = (ks) and u
0
(s) = b
0
for s S
1
. Since
the bration M admits a (symplectic) trivialization over every loop in B, the given
trivialization of M over extends to a symplectic trivialization : F M
over u such that (
1
(s), q) = (ks, q) for s S
1
and q F. Hence the cycle
T
2
M : (s, t) ((ks), (t)) = (
1
(s), (t))
is homologous to the cycle
T
2
F
b0
: (s, t) v(s, t) := (
0
(s), (t)).
It follows that

a
([]), []
_
:=
1
k
_
T
2
v

b0
.
Since the map u and the trivialization along u are independent of a so is the
cycle v and the cohomology class

a
([]) H
1
(F; Z).
Theorem 6.36 becomes correct if we add to (iii) the existence of the lift

f

and to (iv) the condition that the homomorphism

a
: H
1
(B
1
; Z) H
1
(F; R)
takes the elements of H
1
(B
1
; Z) that descend to torsion classes in H
1
(B; Z) to the
subgroup

H
1
(F; R), for some (and hence every) symplectic trivialization
over the 1-skeleton. For further discussion of these questions see [2].
p 236, line -6: The homology group is H
2
(X; Z).
p 240: The rst displayed formula should read: F(z) = f([z[)z/[z[. The second
displayed formula should read:
() =
i
2
_

([z[
2
+
2
log([z[
2
)
_
=
i
2
_
dz d z +
2
dz d z
[z[
2

2
z dz z d z
[z[
4
_
,
p 291: In Figure 9.3 the second point should be x
N
not x
1
.
p 330, line 10: The calculation should start with
_
G
n
not
_
G
n
.
p 332: We state in Exercise 10.28 that if (M, ) is a connected symplectic manifold
such that = d then the direct sum Flux CAL denes a homomorphism from
the identity component Symp
c
0
M of the group of compactly supported symplecto-
morphisms to the sum H
1
c
(M; R) R. As pointed out by KotschickMorita in [1]
this is correct when dimM = 2n 4 but may fail when dimM = 2. They observe
that CAL extends to a map
CAL : Symp
c
0
M R,
1
n + 1
_
M

()
n1
,
that satises the identity
CAL() = CAL() + CAL() +
1
n + 1
_
M
Flux() Flux()
n1
.
Hence this map is a group homomorphism only if the cup product pairing
H
1
c
(M; R) H
1
c
(M; R) R, (, )
_
M

n1
vanishes. This pairing does vanish when n > 1 because is exact. However, it
need not vanish when n = 1. In this case, the correct target group is the central
extension of H
1
c
(M; R) by R that is dened by this pairing. For more detail see [1,
Prop 11].
p 334, lines -1 and -7: Replace
t
by
t
.
p 342: In Example 11.3 the second function should be
H(x, y) = sin 2x + sin2y.
p 350, line -5 : The homology exact sequence of the triple (N
b
, N
a
, L) shows
that
d
ab
k1
+ d
ab
k
= c
b
k
c
a
k
b
b
k
+ b
a
k
.
p 374, line 7: Replace by the monotonicity axiom by by the monotonicity and
conformality axioms.
p 387/8: The dieomorphism f whose description begins at the bottom of p 387
takes a neighbourhood of the arc (

2r, r) into a neighbourhood of U

1
L.
p 399, line -9: The equation (X) = should be replaced by L
X
= .
References
[1] D. Kotschick and S. Morita, Signatures of foliated surface bundles and the symplectomor-
phism groups of surfaces, SG/0305182, to appear in Topology.
[2] D. McDu, Enlarging the Hamiltonian group (2004).
Dept Math, Stony Brook University, Stony Brook, NY 11794-3651, USA
E-mail address: dusa@math.sunysb.edu
URL: http://www.math.sunysb.edu/~dusa
ETH-Z urich
E-mail address: salamon@math.ethz.ch
URL: http://www.math.ethz.ch/~salamon