Notes on Basic 3-Manifold Topology

Allen Hatcher

Chapter 1. Canonical Decomposition

1. Prime Decomposition.
2. Torus Decomposition.

Chapter 2. Special Classes of 3-Manifolds

1. Seifert Manifolds.
2. Torus Bundles and Semi-Bundles.

Chapter 3. Homotopy Properties

1. The Loop and Sphere Theorems.

These notes, originally written in the 1980’s, were intended as the beginning of a
book on 3 manifolds, but unfortunately that project has not progressed very far since
then. A few small revisions have been made in 1999 and 2000, but much more remains
to be done, both in improving the existing sections and in adding more topics. The
next topic to be added will probably be Haken manifolds in §3.2. For any subsequent
updates which may be written, the interested reader should check my webpage:
http://www.math.cornell.edu/˜hatcher

The three chapters here are to a certain extent independent of each other. The
main exceptions are that the beginning of Chapter 1 is a prerequisite for almost ev-
erything else, while some of the later parts of Chapter 1 are used in Chapter 2.
§1.1 Prime Decomposition 1

Chapter 1. Canonical Decomposition

This chapter begins with the first general result on 3 manifolds, Kneser’s theorem
that every compact orientable 3 manifold M decomposes uniquely as a connected
sum M = P1 ] ··· ] Pn of 3 manifolds Pi which are prime in the sense that they can
be decomposed as connected sums only in the trivial way Pi = Pi ] S 3 .
After the prime decomposition, we turn in the second section to the canonical
torus decomposition due to Jaco-Shalen and Johannson.
We shall work in the C ∞ category throughout. All 3 manifolds in this chapter
are assumed to be connected, orientable, and compact, possibly with boundary, unless
otherwise stated or constructed.

1. Prime Decomposition
Implicit in the prime decomposition theorem is the fact that S 3 is prime, other-
wise one could only hope for a prime decomposition modulo invertible elements, as
in algebra. This is implied by Alexander’s theorem, our first topic.

Alexander’s Theorem
This quite fundamental result was one of the earliest theorems in the subject:

Theorem 1.1. Every embedded 2 sphere in R3 bounds an embedded 3 ball.

Proof: Let S ⊂ R3 be an embedded closed surface, with h : S →R the height function

given by the z coordinate. After a small isotopy of S we may assume h is a morse
function with all its critical points in distinct levels. Namely, there is a small homotopy
of h to such a map. Keeping the same x and y coordinates for S , this gives a small
homotopy of S in R3 . But embeddings are open in the space of all maps, so if this
homotopy is chosen small enough, it will be an isotopy.
Let a1 < ··· < an be noncritical values of h such that each interval (−∞, a1 ) ,
(a1 , a2 ) , ··· , (an , ∞) contains just one critical value. For each i , h−1 (ai ) consists of
a number of disjoint circles in the level z = ai . By the two-dimensional Schoenflies
Theorem (which can be proved by the same method we are using here) each circle of
h−1 (ai ) bounds a disk in the plane z = ai . Let C be an innermost circle of h−1 (ai ) ,
in the sense that the disk D it bounds in z = ai is disjoint from all the other circles
of h−1 (ai ) . We can use D to surger S along C . This means that for some small ε > 0
we first remove from S the open annulus A consisting of points near C between the
two planes z = ai ± ε , then we cap off the resulting pair of boundary circles of S − A
by adding to S − A the disks in z = ai ± ε which these circles bound. The result of
this surgery is thus a new embedded surface, with perhaps one more component than
S , if C separated S .
This surgery process can now be iterated, taking at each stage an innermost re-
maining circle of h−1 (ai ) , and choosing ε small enough so that the newly introduced
2 Canonical Decomposition §1.1

horizontal cap disks intersect the previously constructed surface only in their bound-
aries. See Figure 1.1. After surgering all the circles of h−1 (ai ) for all i , the original
surface S becomes a disjoint union of closed surfaces Sj , each consisting of a number
of horizontal caps together with a connected subsurface Sj0 of S containing at most
one critical point of h .

Figure 1.1

Lemma 1.2. Each Sj is isotopic to one of seven models: the four shown in Figure 1.2
plus three more obtained by turning these upside down. Hence each Sj bounds a
ball.

Figure 1.2
Proof: Consider the case that Sj has a saddle, say in the level z = a . First isotope
Sj in a neighborhood of this level z = a so that for some δ > 0 the subsurface Sjδ of
Sj lying in a − δ ≤ z ≤ a + δ is vertical, i.e., a union of vertical line segments, except
in a neighborhood N ⊂ int(Sjδ ) of the saddle, where Sj has the standard form of the
saddles in the models. Next, isotope Sj so that its subsurface Sj0 (the complement of
the horizontal caps) lies in Sjδ . This is done by pushing its horizontal caps, innermost
ones first, to lie near z = a , as in Figure 1.3, keeping the caps horizontal throughout
the deformation.

Figure 1.3
After this move Sj is entirely vertical except for the standard saddle and the horizontal
caps. Viewed from above, Sj minus its horizontal caps then looks like two smooth
circles, possibly nested, joined by a 1 handle, as in Figure 1.4.
§1.1 Prime Decomposition 3

Figure 1.4
Since these circles bound disks, they can be isotoped to the standard position of one
of the models, yielding an isotopy of Sj to one of the models.
The remaining cases, when Sj0 has a local maximum or minimum, or no critical
points, are similar but simpler, so we leave them as exercises. u
t

Now we assume the given surface S is a sphere. Each surgery then splits one
sphere into two spheres. Reversing the sequence of surgeries, we start with a collec-
tion of spheres Sj bounding balls. The inductive assertion is that at each stage of the
reversed surgery process we have a collection of spheres each bounding a ball. For
the inductive step we have two balls A and B bounded by the spheres ∂A and ∂B
resulting from a surgery. Letting the ε for the surgery go to 0 isotopes A and B so
that ∂A∩∂B equals the horizontal surgery disk D . There are two cases, up to changes
in notation:
(i) A ∩ B = D , with pre-surgery sphere denoted ∂(A + B)
(ii) B ⊂ A , with pre-surgery sphere denoted ∂(A − B) .
Since B is a ball, the lemma below implies that A and A ± B are diffeomorphic. Since
A is a ball, so is A ± B , and the inductive step is completed. u
t

Lemma 1.3. Given an n manifold M and a ball B n−1 ⊂ ∂M , let the manifold N be
obtained from M by attaching a ball B n via an identification of a ball B n−1 ⊂ ∂B n
with the ball B n−1 ⊂ ∂M . Then M and N are diffeomorphic.

Proof: Any two codimension-zero balls in a connected manifold are isotopic. Ap-
plying this fact to the given inclusion B n−1 ⊂ ∂B n and using isotopy extension, we
conclude that the pair (B n , B n−1 ) is diffeomorphic to the standard pair. So there is an
isotopy of ∂N to ∂M in N , fixed outside B n , pushing ∂N −∂M across B n to ∂M −∂N .
By isotopy extension, M and N are then diffeomorphic. u
t

Existence and Uniqueness of Prime Decompositions

Let M be a 3 manifold and S ⊂ M a surface which is properly embedded, i.e.,
S ∩ ∂M = ∂S , a transverse intersection. We do not assume S is connected. Deleting
a small open tubular neighborhood N(S) of S from M , we obtain a 3 manifold M || S
which we say is obtained from M by splitting along S . The neighborhood N(S) is
4 Canonical Decomposition §1.1

an interval-bundle over S , so if M is orientable, N(S) is a product S × (−ε, ε) iff S is

orientable.
Now suppose that M is connected and S is a sphere such that M || S has two
components, M10 and M20 . Let Mi be obtained from Mi0 by filling in its boundary
sphere corresponding to S with a ball. In this situation we say M is the connected
sum M1 ] M2 . We remark that Mi is uniquely determined by Mi0 since any two ways
of filling in a ball B 3 differ by a diffeomorphism of ∂B 3 , and any diffeomorphism
of ∂B 3 extends to a diffeomorphism of B 3 . This last fact follows from the stronger
assertion that any diffeomorphism of S 2 is isotopic to either the identity or a reflection
(orientation-reversing), and each of these two diffeomorphisms extends over a ball.
See [Cerf].
The connected sum operation is commutative by definition and has S 3 as an
identity since a decomposition M = M ] S 3 is obtained by choosing the sphere S
to bound a ball in M . The connected sum operation is also associative, since in a
sequence of connected sum decompositions, e.g., M1 ] (M2 ] M3 ) , the later splitting
spheres can be pushed off the balls filling in earlier splitting spheres, so one may
assume all the splitting spheres are disjointly embedded in the original manifold M .
Thus M = M1 ] ··· ] Mn means there is a collection S consisting of n − 1 disjoint
spheres such that M || S has n components Mi0 , with Mi obtained from Mi0 by filling
in with balls its boundary spheres corresponding to spheres of S .
A connected 3 manifold M is called prime if M = P ]Q implies P = S 3 or Q = S 3 .
For example, Alexander’s theorem implies that S 3 is prime, since every 2 sphere in S 3
bounds a 3 ball. The latter condition, stronger than primeness, is called irreducibility:
M is irreducible if every 2 sphere S 2 ⊂ M bounds a ball B 3 ⊂ M . The two conditions
are in fact very nearly equivalent:

Proposition 1.4. The only orientable prime 3 manifold which is not irreducible is
1 2
S ×S .

Proof: If M is prime, every 2 sphere in M which separates M into two components

bounds a ball. So if M is prime but not irreducible there must exist a nonseparating
sphere in M . For a nonseparating sphere S in an orientable manifold M the union
of a product neighborhood S × I of S with a tubular neighborhood of an arc joining
S × {0} to S × {1} in the complement of S × I is a manifold diffeomorphic to S 1 × S 2
minus a ball. Thus M has S 1 × S 2 as a connected summand. Assuming M is prime,
then M = S 1 × S 2 .
It remains to show that S 1 × S 2 is prime. Let S ⊂ S 1 × S 2 be a separating sphere,
so S 1 × S 2 || S consists of two compact 3 manifolds V and W each with boundary a
2 sphere. We have Z = π1 (S 1 × S 2 ) ≈ π1 V ∗ π1 W , so either V or W must be simply-
connected, say V is simply-connected. The universal cover of S 1 × S 2 can be identified
e of itself in R3 − {0} . The sphere
with R3 − {0} , and V lifts to a diffeomorphic copy V
§1.1 Prime Decomposition 5

e bounds a ball in R3 by Alexander’s theorem. Since ∂ V

∂V e also bounds V e in R3
e is a ball, hence also V . Thus every separating sphere in S 1 × S 2
we conclude that V
bounds a ball, so S 1 × S 2 is prime. u
t

Theorem 1.5. Let M be compact, connected, and orientable. Then there is a decom-
position M = P1 ] ··· ] Pn with each Pi prime, and this decomposition is unique up
to insertion or deletion of S 3 ’s.

Proof: The existence of prime decompositions is harder, and we tackle this first.
If M contains a nonseparating S 2 , this gives a decomposition M = N ] S 1 × S 2 , as
we saw in the proof of Proposition 1.4. We can repeat this step of splitting off an
S 1 × S 2 summand as long as we have nonseparating spheres, but the process cannot
be repeated indefinitely since each S 1 × S 2 summand gives a Z summand of H1 (M) ,
which is a finitely generated abelian group since M is compact. Thus we are reduced
to proving existence of prime decompositions in the case that each 2 sphere in M
separates. Each 2 sphere component of ∂M corresponds to a B 3 summand of M , so
we may also assume ∂M contains no 2 spheres.
We shall prove the following assertion, which clearly implies the existence of
prime decompositions:

There is a bound on the number of spheres in a system S of disjoint spheres satisfying:

(∗) No component of M || S is a punctured 3 sphere, i.e., a compact manifold obtained

from S 3 by deleting finitely many open balls with disjoint closures.

Before proving this we make a preliminary observation: If S satisfies (∗) and we do

surgery on a sphere Si of S using a disk D ⊂ M with D ∩ S = ∂D ⊂ Si , then at
least one of the systems S 0 , S 00 obtained by replacing Si with the spheres Si0 and Si00
resulting from the surgery satisfies (∗) . To see this, first perturb Si0 and Si00 to be
disjoint from Si and each other, so that Si , Si0 , and Si00 together bound a 3 punctured
sphere P .

Si0
0
B
P Si
00
B
A Si00

Figure 1.5
On the other side of Si from P we have a component A of M || S , while the spheres Si0
and Si00 split the component of M || S containing P into pieces B 0 , B 00 , and P . If both
B 0 and B 00 were punctured spheres, then B 0 ∪ B 00 ∪ P , a component of M || S , would
be a punctured sphere, contrary to hypothesis. So one of B 0 and B 00 , say B 0 , is not a
6 Canonical Decomposition §1.1

punctured sphere. If A ∪ P ∪ B 00 were a punctured sphere, this would force A to be

a punctured sphere, by Alexander’s theorem. This is also contrary to hypothesis. So
we conclude that neither component of M || S 0 adjacent to Si0 is a punctured sphere,
hence the sphere system S 0 satisfies (∗) .
Now we prove the assertion that the number of spheres in a system S satisfying
(∗) is bounded. Choose a smooth triangulation T of M . This has only finitely many
simplices since M is compact. The given system S can be perturbed to be transverse
to all the simplices of T . This perturbation can be done inductively over the skeleta
of T : First make S disjoint from vertices, then transverse to edges, meeting them in
finitely many points, then transverse to 2 simplices, meeting them in finitely many
arcs and circles.
For a 3 simplex τ of T , we can make the components of S ∩ τ all disks, as
follows. Such a component must meet ∂τ by Alexander’s theorem and condition
(∗) . Consider a circle C of S ∩ ∂τ which is innermost in ∂τ . If C bounds a disk
component of S ∩ τ we may isotope this disk to lie near ∂τ and then proceed to a
remaining innermost circle C . If an innermost remaining C does not bound a disk
component of S ∩ τ we may surger S along C using a disk D lying near ∂τ with
D ∩ S = ∂D = C , replacing S by a new system S 0 satisfying (∗) , in which either C
does bound a disk component of S 0 ∩ τ or C is eliminated from S 0 ∩ τ . After finitely
many such steps we arrive at a system S with S ∩ τ consisting of disks, for each τ .
In particular, note that no component of the intersection of S with a 2 simplex of T
can be a circle, since this would bound disks in both adjacent 3 simplices, forming a
sphere of S bounding a ball in the union of these two 3 simplices, contrary to (∗) .
Next, for each 2 simplex σ we eliminate arcs α of S ∩ σ having both endpoints
on the same edge of σ . Such an α cuts off from σ a disk D which meets only one
edge of σ . We may choose α to be ‘edgemost,’ so that D contains no other arcs
of S ∩ σ , and hence D ∩ S = α since circles of S ∩ σ have been eliminated in the
previous step. By an isotopy of S supported near α we then push the intersection arc
α across D , eliminating α from S ∩ σ and decreasing by two the number of points
of intersection of S with the 1 skeleton of T .

Figure 1.6
After such an isotopy decreasing the number of points of intersection of S with
the 1 skeleton of T we repeat the first step of making S intersect all 3 simplices in
disks. This does not increase the number of intersections with the 1 skeleton, so after
finitely many steps, we arrive at the situation where S meets each 2 simplex only in
§1.1 Prime Decomposition 7

arcs connecting adjacent sides, and S meets 3 simplices only in disks.

Now consider the intersection of S with a 2 simplex σ . With at most four ex-
ceptions the complementary regions of S ∩ σ in σ are rectangles with two opposite
sides on ∂σ and the other two opposite sides arcs of S ∩ σ , as in Figure 1.7. Thus if
T has t 2 simplices, then all but at most 4t of the components of M || S meet all the
2 simplices of T only in such rectangles.

Figure 1.7
Let R be a component of M || S meeting all 2 simplices only in rectangles. For a
3 simplex τ , each component of R∩∂τ is an annulus A which is a union of rectangles.
The two circles of ∂A bound disks in τ , and A together with these two disks is a
sphere bounding a ball in τ , a component of R ∩ τ which can be written as D 2 × I
with ∂D 2 × I = A . The I fiberings of all such products D 2 × I may be assumed to
agree on their common intersections, the rectangles, to give R the structure of an
I bundle. Since ∂R consists of sphere components of S , R is either the product
S 2 × I or the twisted I bundle over RP2 . ( R is the mapping cylinder of the associated
∂I subbundle, a union of spheres which is a two-sheeted covering space of a connected
base surface.) The possibility R = S 2 × I is excluded by (∗) . Each I bundle R is thus
the mapping cylinder of the covering space S 2 →RP2 . This is just RP3 minus a ball, so
each I bundle R gives a connected summand RP3 of M , hence a Z2 direct summand
of H1 (M) . Thus the number of such components R of M || S is bounded. Since the
number of other components was bounded by 4t , the number of components of M || S
is bounded. Since every 2 sphere in M separates, the number of components of M || S
is one more than the number of spheres in S . This finishes the proof of the existence
of prime decompositions.
For uniqueness, suppose the nonprime M has two prime decompositions M =
P1 ]···]Pk ]`(S 1 × S 2 ) and M = Q1 ]···]Qm ]n(S 1 × S 2 ) where the Pi ’s and Qi ’s are
irreducible and not S 3 . Let S be a disjoint union of 2 spheres in M reducing M to the
Pi ’s, i.e., the components of M || S are the manifolds P1 , ··· , Pk with punctures, plus
possibly some punctured S 3 ’s. Such a system S exists: Take for example a collection
of spheres defining the given prime decomposition M = P1 ] ··· ] Pk ] `(S 1 × S 2 )
together with a nonseparating S 2 in each S 1 × S 2 . Note that if S reduces M to the
Pi ’s, so does any system S 0 containing S .
Similarly, let T be a system of spheres reducing M to the Qi ’s. If S ∩ T ≠ ∅ ,
we may assume this is a transverse intersection, and consider a circle of S ∩ T which
is innermost in T , bounding a disk D ⊂ T with D ∩ S = ∂D . Using D , surger the
8 Canonical Decomposition §1.1

sphere Sj of S containing ∂D to produce two spheres Sj0 and Sj00 , which we may
take to be disjoint from Sj , so that Sj , Sj0 , and Sj00 together bound a 3 punctured
3 sphere P . By an earlier remark, the enlarged system S ∪ Sj0 ∪ Sj00 reduces M to the
Pi ’s. Deleting Sj from this enlarged system still gives a system reducing M to the
Pi ’s since this affects only one component of M || S ∪ Sj0 ∪ Sj00 , by attaching P to one of
its boundary spheres, which has the net effect of simply adding one more puncture
to this component.
The new system S 0 meets T in one fewer circle, so after finitely many steps of
this type we produce a system S disjoint from T and reducing M to the Pi ’s. Then
S ∪ T is a system reducing M both to the Pi ’s and to the Qi ’s. Hence k = m and the
Pi ’s are just a permutation of the Qi ’s.
Finally, to show ` = n , we have M = N ] `(S 1 × S 2 ) = N ] n(S 1 × S 2 ) , so H1 (M) =
H1 (N) ⊕ Z` = H1 (N) ⊕ Zn , hence ` = n . u
t

The proof of the Prime Decomposition Theorem applies equally well to manifolds
which are not just orientable, but oriented. The advantage of working with oriented
manifolds is that the operation of forming M1 ] M2 from M1 and M2 is well-defined:
Remove an open ball from M1 and M2 and then identify the two resulting boundary
spheres by an orientation-reversing diffeomorphism, so the orientations of M1 and
M2 fit together to give a coherent orientation of M1 ] M2 . The gluing map S 2 →S 2 is
then uniquely determined up to isotopy, as we remarked earlier.
Thus to classify oriented compact 3 manifolds it suffices to classify the irre-
ducible ones. In particular, one must determine whether each orientable irreducible
3 manifold possesses an orientation-reversing self-diffeomorphism.
To obtain a prime decomposition theorem for nonorientable manifolds requires
very little more work. In Proposition 1.4 there are now two prime non-irreducible
manifolds, S 1 × S 2 and S 1 ×
e S 2 , the nonorientable S 2 bundle over S 1 , which can also
arise from a nonseparating 2 sphere. Existence of prime decompositions then works
as in the orientable case. For uniqueness, one observes that N ]S 1 × S 2 = N ]S 1 ×
e S 2 if
N is nonorientable. This is similar to the well-known fact in one lower dimension that
connected sum of a nonorientable surface with the torus and with the Klein bottle
give the same result. Uniqueness of prime decomposition can then be restored by
replacing all the S 1 × S 2 summands in nonorientable manifolds with S 1 ×
e S 2 ’s.
A useful criterion for recognizing irreducible 3 manifolds is the following:

Proposition 1.6. f→M is a covering space and M

If p : M f is irreducible, then so is M .

Proof: A sphere S ⊂ M lifts to spheres Se ⊂ M f . Each of these lifts bounds a ball in

f since M
M f is irreducible. Choose a lift Se bounding a ball B in M f such that no other
lifts of S lie in B , i.e., S is an innermost lift. We claim that p : B →p(B) is a covering
e
space. To verify the covering space property, consider first a point x ∈ p(B) − S , with
§1.1 Prime Decomposition 9

U a small ball neighborhood of x disjoint from S . Then p −1 (U ) is a disjoint union

f − p −1 (S) , and the ones of these in B provide a uniform covering of U .
of balls in M
On the other hand, if x ∈ S , choose a small ball neighborhood U of x meeting S in
a disk. Again p −1 (U) is a disjoint union of balls, only one of which, U
e say, meets
B since we chose Se innermost and p is one-to-one on Se . Therefore p restricts to a
homeomorphism of U e ∩B onto a neighborhood of x in p(B) , and the verification that
p : B →p(B) is a covering space is complete. This covering space is single-sheeted on
Se , hence on all of B , so p : B →p(B) is a homeomorphism with image a ball bounded
by S . u
t

The converse of Proposition 1.6 will be proved in §3.1.

By the proposition, manifolds with universal cover S 3 are irreducible. This in-
cludes RP3 , and more generally each 3 dimensional lens space Lp/q , which is the
quotient space of S 3 under the free Zq action generated by the rotation (z1 , z2 ) ,
(e2π i/q z1 , e2pπ i/q z2 ) , where S 3 is viewed as the unit sphere in C2 .
For a product M = S 1 × F 2 , or more generally any surface bundle F 2 →M →S 1 ,
with F 2 a compact connected surface other than S 2 or RP2 , the universal cover of
M − ∂M is R3 , so such an M is irreducible.
Curiously, the analogous covering space assertion with ‘irreducible’ replaced by
‘prime’ is false, since there is a 2 sheeted covering S 1 × S 2 →RP3 ] RP3 . Namely,
RP3 ] RP3 is the quotient of S 1 × S 2 under the identification (x, y) ∼ (ρ(x), −y) with
ρ a reflection of the circle. This quotient can also be described as the quotient of I × S 2
where (x, y) is identified with (x, −y) for x ∈ ∂I . In this description the 2 sphere
giving the decomposition RP3 ] RP3 is {1/2 }× S 2 .

Exercises
1. Prove the (smooth) Schoenflies theorem in R2 : An embedded circle in R2 bounds
an embedded disk.
2. Show that for compact M 3 there is a bound on the number of 2 spheres Si which
can be embedded in M disjointly, with no Si bounding a ball and no two Si ’s bounding
a product S 2 × I .
3. Use the method of proof of Alexander’s theorem to show that every torus T ⊂ S 3
bounds a solid torus S 1 × D 2 ⊂ S 3 on one side or the other. (This result is also due to
Alexander.)
4. Develop an analog of the prime decomposition theorem for splitting a compact
irreducible 3 manifolds along disks rather than spheres. In a similar vein, study the
operation of splitting nonorientable manifolds along RP2 ’s with trivial normal bun-
dles.
5. Show: If M 3 ⊂ R3 is a compact submanifold with H1 (M) = 0 , then π1 (M) = 0 .
10 Canonical Decomposition §1.2

2. Torus Decomposition
Beyond the prime decomposition, there is a further canonical decomposition of
irreducible compact orientable 3 manifolds, splitting along tori rather than spheres.
This was discovered only in the mid 1970’s, by Johannson and Jaco-Shalen, though
in the simplified geometric version given here it could well have been proved in the
1930’s. (A 1967 paper of Waldhausen comes very close to this geometric version.)
Perhaps the explanation for this late discovery lies in the subtlety of the uniqueness
statement. There are counterexamples to a naive uniqueness statement, involving a
class of manifolds studied extensively by Seifert in the 1930’s. The crucial observa-
tion, not made until the 1970’s, was that these Seifert manifolds give rise to the only
counterexamples.

Existence of Torus Decompositions

A properly embedded surface S ⊂ M 3 is called 2-sided or 1-sided according to
whether its normal I bundle is trivial or not. A 2 sided surface without S 2 or D 2
components is called incompressible if for each disk D ⊂ M with D ∩ S = ∂D there
is a disk D 0 ⊂ S with ∂D 0 = ∂D . See Figure 1.8. Thus, surgery on S cannot produce
a simpler surface, but only splits off an S 2 from S .
D0

D D

Figure 1.8
A disk D with D∩S = ∂D will sometimes be called a compressing disk for S , whether
or not a disk D 0 ⊂ S with ∂D 0 = ∂D exists.
As a matter of convenience, we extend the definition of incompressibility to allow
S to have disk components which are not isotopic rel boundary to disks in ∂M .

Here are some preliminary facts about incompressible surfaces:

(1) A surface is incompressible iff each of its components is incompressible. The
‘if’ half of this is easy. For the ‘only if,’ let D ⊂ M be a compressing disk for one
component Si of S . If D meets other components of S , surger D along circles of
D ∩ S , innermost first as usual, to produce a new D with the same boundary circle,
and meeting S only in this boundary circle. Then apply incompressibility of S .
(2) A connected 2 sided surface S which is not a sphere or disk is incompressible
if the map π1 (S)→π1 (M) induced by inclusion is injective. For if D ⊂ M is a com-
pressing disk, then ∂D is nullhomotopic in M , hence also in S by the π1 assumption.
Then it is a standard fact that a nullhomotopic embedded circle in a surface bounds
a disk in the surface. Note that it is enough for the two inclusions of S into M || S to
be injective on π1 .
§1.2 Torus Decomposition 11

The converse is also true, as is shown in Corollary 3.3. For 1 sided surfaces these
two conditions for incompressibility are no longer equivalent, π1 injectivity being
strictly stronger in general; see the exercises. We emphasize, however, that in these
notes we use the term ‘incompressible’ only in reference to 2 sided surfaces.
(3) There are no incompressible surfaces in R3 , or equivalently in S 3 . This is im-
mediate from the converse to (2), but can also be proved directly, as follows. As we
saw in the proof of Alexander’s theorem, there is a sequence of surgeries on S along
horizontal disks in R3 converting S into a disjoint union of spheres. Going through
this sequence of surgeries in turn, either a surgery disk exhibits S as compressible,
or it splits S into two surfaces one of which is a sphere. This sphere bounds balls on
each side in S 3 , and we can use one of these balls to isotope S to the other surface
produced by the surgery. At the end of the sequence of surgeries we have isotoped
S to a collection of spheres, but the definition of incompressibility does not allow
spheres.
(4) A 2 sided torus T in an irreducible M is compressible iff either T bounds a solid
torus S 1 × D 2 ⊂ M or T lies in a ball in M . For if T is compressible there is a surgery
of T along some disk D which turns T into a sphere. This sphere bounds a ball B ⊂ M
by the assumption that M is irreducible. There are now two cases: If B ∩ D = ∅ then
reversing the surgery glues together two disks in ∂B to create a solid torus bounded
by T . The other possibility is that D ⊂ B , and then T ⊂ B . Note that if M = S 3 the
ball B can be chosen disjoint from D , so the alternative D ⊂ B is not needed. Thus
using statement (3) above we obtain the result, due to Alexander, that a torus in S 3
bounds a solid torus on one side or the other.
(5) If S ⊂ M is incompressible, then M is irreducible iff M || S is irreducible. For
suppose M is irreducible. Then a 2 sphere in M || S bounds a ball in M , which must
be disjoint from S by statement (3) above, so M || S is irreducible. Conversely, given an
S 2 ⊂ M , consider a circle of S ∩ S 2 which is innermost in S 2 , bounding a disk D ⊂ S 2
with D ∩ S = ∂D . By incompressibility of S , ∂D bounds a disk D 0 ⊂ S . The sphere
D ∪D 0 bounds a ball B ⊂ M if M || S is irreducible. We must have B ∩S = D 0 , otherwise
the component of S containing D 0 would be contained in B , contrary to statement
(3). Isotoping S 2 by pushing D across B to D 0 , and slightly beyond, eliminates the
circle ∂D from S ∩ S 2 , plus any other circles which happen to lie in D 0 . Repeating
this step, we eventually get S 2 ⊂ M || S , so S 2 bounds a ball and M is irreducible.

Proposition 1.7. For a compact irreducible M there is a bound on the number of

components in a system S = S1 ∪ ··· ∪ Sn of disjoint closed connected incompressible
surfaces Si ⊂ M such that no component of M || S is a product T × I with T a closed
surface.

Proof: This follows the scheme of the proof of existence of prime decompositions.
First, perturb S to be transverse to a triangulation of M and perform the following
12 Canonical Decomposition §1.2

two steps to simplify the intersections of S with 2 simplices σ 2 and 3 simplices σ 3 :

(1) Make all components of S ∩ σ 3 disks. In the proof of prime decomposition, this
was done by surgery, but now the surgeries can be achieved by isotopy. Namely, given
a surgery disk D ⊂ M with D ∩ S = ∂D , incompressibility gives a disk D 0 ⊂ S with
∂D 0 = ∂D . The sphere D ∪ D 0 bounds a ball B ⊂ M since M is irreducible. We have
B ∩ S = D 0 , otherwise a component of S would lie in B . Then isotoping S by pushing
D 0 across B to D and a little beyond replaces S by one of the two surfaces produced
by the surgery.
Note that Step (1) eliminate circles of S ∩ σ 2 , since such a circle would bound
disks in both adjacent σ 3 ’s, producing a sphere component of S .
(2) Eliminate arcs of S ∩ σ 2 with both endpoints on the same edge of σ 2 , by isotopy
of S .
After these simplifications, components of M || S meeting 2 simplices only in rect-
angles are I bundles (disjoint from ∂M ), as before. Trivial I bundles are ruled out by
hypothesis, assuming n ≥ 2 so that M is not a fiber bundle with fiber S = S1 . Non-
trivial I bundles are tubular neighborhoods of 1 sided (hence nonseparating) surfaces
T1 , ··· , Tm , say. We may assume M is connected, and then reorder the components
Si of S so that M || (S1 ∪ ··· ∪ Sk ) is connected and each of Sk+1 , ··· , Sn separates
M || (S1 ∪ ··· ∪ Sk ) . The surfaces S1 , ··· , Sk , T1 , ··· , Tm give linearly independent el-
ements of H2 (M; Z2 ) , for a linear relation among them with Z2 coefficients would
imply that some subcollection of them forms the boundary of a 3 dimensional sub-
manifold of M . (Consider simplicial homology with respect to a triangulation of M in
which all these surfaces are subcomplexes.) This is impossible since the complement
of the collection is connected, hence also the complement of any subcollection.
Thus k + m is bounded by the dimension of H2 (M; Z2 ) . The number of com-
ponents of M || S , which is n − k + 1 , is bounded by m + 4t , t being the number of
2 simplices in the given triangulation. Combining these bounds, we obtain the in-
equalities n + 1 ≤ k + m + 4t ≤ 4t + dim H2 (M; Z2 ) . This gives a bound on n , the
number of Si ’s. u
t

A properly embedded surface S ⊂ M is called ∂ parallel if it is isotopic, fixing

∂S , to a subsurface of ∂M . By isotopy extension this is equivalent to saying that S
splits off a product S × [0, 1] from M with S = S × {0} . An irreducible manifold M is
called atoroidal if every incompressible torus in M is ∂ parallel.

Corollary 1.8. In a compact connected irreducible M there exists a finite collection

T of disjoint incompressible tori such that each component of M || T is atoroidal.

Proof: Construct inductively a sequence of disjoint incompressible tori T1 , T2 , ··· in

M by letting Ti be an incompressible torus in the manifold Mi = M || (T1 ∪ ··· ∪ Ti−1 )
which is not ∂ parallel in Mi , if Mi is not atoroidal. The claim is that this procedure
§1.2 Torus Decomposition 13

must stop at some finite stage. In view of how Mi is constructed from Mi−1 by splitting
along a torus which is not ∂ parallel, there are just two ways that some Mi can have
a component which is a product S × I with S a closed surface: Either i = 1 and
M1 = M = S × I , or i = 2 and M2 = S × I with S a torus. In the latter case M is a torus
bundle with T1 as a fiber. Thus if the process of constructing Ti ’s does not terminate,
we obtain collections T1 ∪ ··· ∪ Ti satisfying the conditions of the theorem but with
arbitrarily large i , a contradiction. u
t

Now we describe an example of an irreducible M where this torus decomposition

into atoroidal pieces is not unique, the components of M || T for the two splittings
being in fact non-homeomorphic.

Example. For i = 1, 2, 3, 4 , let Mi be a solid torus with ∂Mi decomposed as the union
of two annuli Ai and A0i each winding qi > 1 times around the S 1 factor of Mi . The
union of these four solid tori, with each A0i glued to Ai+1 (subscripts mod 4 ), is the
manifold M . This contains two tori T1 = A1 ∪ A3 and T2 = A2 ∪ A4 . The components
of M || T1 are M1 ∪ M2 and M3 ∪ M4 , and the components of M || T2 are M2 ∪ M3
and M4 ∪ M1 . The fundamental group of Mi ∪ Mi+1 has presentation hxi , xi+1 |
q q
xi i = xi+1
i+1
i . The center of this amalgamated free product is cyclic, generated by
q q
the element xi i = xi+1
i+1
. Factoring out the center gives quotient Zqi ∗ Zqi+1 , with
abelianization Zqi ⊕ Zqi+1 . Thus if the qi ’s are for example distinct primes, then no
two of the manifolds Mi ∪ Mi+1 are homeomorphic.
Results from later in this section will imply that M is irreducible, T1 and T2
are incompressible, and the four manifolds Mi ∪ Mi+1 are atoroidal. So the splittings
M || T1 and M || T2 , though quite different, both satisfy the conclusions of the Corollary.

Manifolds like this M which are obtained by gluing together solid tori along non-
contractible annuli in their boundaries belong to a very special class of manifolds
called Seifert manifolds, which we now define. A model Seifert fibering of S 1 × D 2 is
a decomposition of S 1 × D 2 into disjoint circles, called fibers, constructed as follows.
Starting with [0, 1]× D 2 decomposed into the segments [0, 1]× {x} , identify the disks
{0}× D 2 and {1}× D 2 via a 2π p/q rotation, for p/q ∈ Q with p and q relatively
prime. The segment [0, 1]× {0} then becomes a fiber S 1 × {0} , while every other fiber
in S 1 × D 2 is made from q segments [0, 1]× {x} . A Seifert fibering of a 3 manifold
M is a decomposition of M into disjoint circles, the fibers, such that each fiber has a
neighborhood diffeomorphic, preserving fibers, to a neighborhood of a fiber in some
model Seifert fibering of S 1 × D 2 . A Seifert manifold is one which possesses a Seifert
fibering.
Each fiber circle C in a Seifert fibering of a 3 manifold M has a well-defined
multiplicity, the number of times a small disk transverse to C meets each nearby
fiber. For example, in the model Seifert fibering of S 1 × D 2 with 2π p/q twist, the
fiber S 1 × {0} has multiplicity q while all other fibers have multiplicity 1 . Fibers
14 Canonical Decomposition §1.2

of multiplicity 1 are regular fibers, and the other fibers are multiple (or singular, or
exceptional). The multiple fibers are isolated and lie in the interior of M . The quotient
space B of M obtained by identifying each fiber to a point is a surface, compact if
M is compact, as is clear from the model Seifert fiberings. The projection π : M →B
is an ordinary fiber bundle on the complement of the multiple fibers. In particular,
π : ∂M →∂B is a circle bundle, so ∂M consists of tori and Klein bottles, or just tori if
M is orientable.
The somewhat surprising fact is that Seifert manifolds account for all the non-
uniqueness in torus splittings, according to the following theorem, which is the main
result of this section.

Theorem 1.9. For a compact irreducible orientable 3 manifold M there exists a

collection T ⊂ M of disjoint incompressible tori such that each component of M || T is
either atoroidal or a Seifert manifold, and a minimal such collection T is unique up
to isotopy.

Here ‘minimal’ means minimal with respect to inclusions of such collections. Note
the strength of the uniqueness: up to isotopy, not just up to homeomorphism of M ,
for example. The orientability assumption can be dropped if splittings along incom-
pressible Klein bottles are also allowed, and the definition of ‘atoroidal’ is modified
accordingly. For simplicity we shall stick to the orientable case, however.
Before proving the uniqueness statement we need to study Seifert manifolds a
little more. This is done in the following subsection.

Incompressible Surfaces in Seifert Manifolds

There is a relative form of incompressibility which is often very useful: A surface
S ⊂ M is ∂ incompressible if for each disk D ⊂ M such that ∂D decomposes as the
union of two arcs α and β meeting only at their common endpoints, with D ∩ S = α
and D ∩∂M = β (such a D is called a ∂ compressing disk for S ) there is a disk D 0 ⊂ S
with α ⊂ ∂D 0 and ∂D 0 − α ⊂ ∂S . See Figure 1.9.
D0

D D

∂M

Figure 1.9
A surface which is both incompressible and ∂ incompressible we shall call essen-
tial. We leave it as an exercise to show that a surface S is essential iff each com-
ponent of S is essential. Also, as in the absolute case, S is ∂ incompressible if
π1 (S, ∂S)→π1 (M, ∂M) is injective for all choices of basepoint in ∂S .
§1.2 Torus Decomposition 15

Example. Let us show that the only essential surfaces in the manifold M = S 1 × D 2 are
disks isotopic to meridian disks {x}× D 2 . For let S be a connected essential surface
in M . We may isotope S so that all the circles of ∂S are either meridian circles
{x}× ∂D 2 or are transverse to all meridian circles. By a small perturbation S can also
be made transverse to a fixed meridian disk D0 . Circles of S ∩ D0 can be eliminated,
innermost first, by isotopy of S using incompressibility of S and irreducibility of M .
After this has been done, consider an edgemost arc α of S ∩ D0 . This cuts off a
∂ compressing disk D from D0 , so α also cuts off a disk D 0 from S , meeting ∂M
in an arc γ . The existence of D 0 implies that the two ends of γ lie on the same side
of the meridian arc β = D ∩ ∂M in ∂M . But this is impossible since γ is transverse
to all meridians and therefore proceeds monotonically through the meridian circles
of ∂M . Thus we must have S disjoint from D0 , so ∂S consists of meridian circles.
Moreover, S is incompressible in M || D0 , a 3 ball, so S must be a disk since each of
its boundary circles bounds a disk in the boundary of the ball, and pushing such a
disk slightly into the interior of the ball yields a compressing disk for S . It follows
from Alexander’s theorem that any two disks in a ball having the same boundary are
isotopic fixing the boundary, so S is isotopic to a meridian disk in M .

Lemma 1.10. Let S be a connected incompressible surface in the irreducible 3 man-

ifold M , with ∂S contained in torus boundary components of M . Then either S is
essential or it is a ∂ parallel annulus.

Proof: Suppose S is ∂ compressible, with a ∂ compressing disk D meeting S in an

arc α which does not cut off a disk from S . Let β be the arc D ∩ ∂M , lying in a
torus component T of ∂M . The circles of S ∩ T do not bound disks in T , otherwise
incompressibility of S would imply S was a disk, but disks are ∂ incompressible.
Thus β lies in an annulus component A of T || ∂S . If β were trivial in A , cutting off a
disk D 0 , incompressibility applied to the disk D ∪ D 0 would imply that α cuts off a
disk from S , contrary to assumption; see Figure 1.10(a).

D
A
0
D

(a) (b)
Figure 1.10
So β joins the two components of ∂A . If both of these components are the same circle
of ∂S , i.e., if S ∩ T consists of a single circle, then S would be 1 sided. For consider
the normals to S pointing into D along α . At the two points of ∂α these normals
point into β , hence point to opposite sides of the circle S ∩ T .
16 Canonical Decomposition §1.2

Thus the endpoints of β must lie in different circles of ∂S , and we have the
configuration in Figure 1.10(b). Let N be a neighborhood of ∂A∪α in S , a 3 punctured
sphere. The circle ∂N − ∂S bounds an obvious disk in the complement of S , lying
near D ∪ A , so since S is incompressible this boundary circle also bounds a disk in
S . Thus S is an annulus. Surgering the torus S ∪ A via D yields a sphere, which
bounds a ball in M since M is irreducible. Hence S ∪ A bounds a solid torus and S
is ∂ parallel, being isotopic to A rel ∂S . u
t

Proposition 1.11. If M is a connected compact irreducible Seifert-fibered manifold,

then any essential surface in M is isotopic to a surface which is either vertical, i.e., a
union of regular fibers, or horizontal, i.e., transverse to all fibers.

Proof: Let C1 , ··· , Cn be fibers of the Seifert fibering which include all the multiple
fibers together with at least one regular fiber if there are no multiple fibers. Let M0
be M with small fibered open tubular neighborhoods of all the Ci ’s deleted. Thus
M0 is a circle bundle M0 →B0 over a compact connected surface B0 with nonempty
boundary. Choose disjoint arcs in B0 whose union splits B0 into a disk, and let A
be the pre-image in M0 of this collection of arcs, a union of disjoint vertical annuli
A1 , ··· , Am in M0 such that the manifold M1 = M0 || A is a solid torus.
The circles of ∂S are nontrivial in ∂M since S is incompressible and M is ir-
reducible. Hence S can be isotoped so that the circles of ∂S are either vertical or
horizontal in each component torus or Klein bottle of ∂M . Vertical circles of S may
be perturbed to be disjoint from A . We may assume S meets the fibers Ci trans-
versely, and hence meets the neighborhoods of these fibers in disks transverse to
fibers. So the surface S0 = S ∩ M0 also has each its boundary circles horizontal or
vertical.
Circles of S ∩ A bounding disks in A can be eliminated by isotopy of S in the
familiar way, using incompressibility of S and irreducibility of M . Arcs of S ∩ A
with both endpoints on the same component of ∂A can be eliminated as follows.
An edgemost such arc α cuts off a disk D from A . If the two endpoints of α lie
in a component of ∂M0 − ∂M , then S can be isotoped across D to eliminate two
intersection points with a fiber Ci . The other possibility, that the two endpoints of α
lie in ∂M , cannot actually occur, for if it did, the disk D would be a ∂ compressing disk
for S in M , a configuration ruled out by the monotonicity argument in the Example
preceding Lemma 1.10, with the role of meridians in that argument now played by
vertical circles.
So we may assume the components of S ∩A are either vertical circles or horizontal
arcs. If we let S1 = S0 || A in M0 || A = M1 , it follows that ∂S1 consists entirely of
horizontal or vertical circles in the torus ∂M1 . We may assume S1 is incompressible
in M1 . For let D ⊂ M1 be a compressing disk for S1 . Since S is incompressible, ∂D
bounds a disk D 0 ⊂ S . If this does not lie in S1 , we can isotope S by pushing D 0
§1.2 Torus Decomposition 17

across the ball bounded by D ∪ D 0 , thereby eliminating some components of S ∩ A .

Since S1 is incompressible, its components are either ∂ parallel annuli or are
essential in the solid torus M1 , hence are isotopic to meridian disks by the Example
before Lemma 1.10. If S1 contains a ∂ parallel annulus with horizontal boundary,
then this annulus has a ∂ compressing disk D with D ∩ ∂M1 a vertical arc in ∂M0 .
As in the earlier step when we eliminated arcs of S ∩ A with endpoints on the same
component of ∂A , this leads to an isotopy of S removing intersection points with a
fiber Ci . So we may assume all components of S1 are either ∂ parallel annuli with
vertical boundary or disks with horizontal boundary.
Since vertical circles in ∂M1 cannot be disjoint from horizontal circles, S1 is either
a union of ∂ parallel annuli with vertical boundary, or a union of disks with horizontal
boundary. In the former case S1 can be isotoped to be vertical, staying fixed on ∂S1
where it is already vertical. This isotopy gives an isotopy of S to a vertical surface.
In the opposite case that S1 consists of disks with horizontal boundary, isotopic to
meridian disks in M1 , we can isotope S1 to be horizontal fixing ∂S1 , and this gives an
isotopy of S to a horizontal surface. u
t

Vertical surfaces are easy to understand: They are circle bundles since they are
disjoint from multiple fibers by definition, hence they are either annuli, tori, or Klein
bottles.
Horizontal surfaces are somewhat more subtle. For a horizontal surface S the
projection π : S →B onto the base surface of the Seifert fibering is a branched cover-
ing, with a branch point of multiplicity q for each intersection of S with a singular
fiber of multiplicity q . (To see this, look in a neighborhood of a fiber, where the map
S →B is equivalent to the projection of a number of meridian disks onto B , clearly
a branched covering.) For this branched covering π : S →B there is a useful formula
relating the Euler characteristics of S and B ,
X
χ (B) − χ (S)/n = (1 − 1/qi )
i

where n is the number of sheets in the branched cover and the multiple fibers of
M have multiplicities q1 , ··· , qm . To verify this formula, triangulate B so that the
images of the multiple fibers are vertices, then lift this to a triangulation of S . Count-
ing simplices would then yield the usual formula χ (S) = n χ (B) for an n sheeted
unbranched cover. In the present case, however, a vertex in B which is the image
of a fiber of multiplicity qi has n/qi pre-images in S , rather than n . This yields a
P
modified formula χ (S) = n χ (B) + i (−n + n/qi ) , which is equivalent to the one
above.
There is further structure associated to a horizontal surface S in a Seifert-fibered
manifold M . Assume S is connected and 2 sided. (If S is 1 sided, it has an I bundle
neighborhood whose boundary is a horizontal 2 sided surface.) Since S →B is onto,
18 Canonical Decomposition §1.2

S meets all fibers of M , and M || S is an I bundle. The local triviality of this I bundle
is clear if one looks in a model-fibered neighborhood of a fiber. The associated
∂I subbundle consists of two copies of S , so the I bundle is the mapping cylinder of
a 2 sheeted covering projection S q S →T for some surface T . There are two cases,
according to whether S separates M or not:
(1) If M || S is connected, so is T , and S q S →T is the trivial covering S q S →S , so
M || S = S × I and hence M is a bundle over S 1 with fiber S . The surface fibers of this
bundle are all horizontal surfaces in the Seifert fibering.
(2) If M || S has two components, each is a twisted I bundle over a component Ti of T ,
the mapping cylinder of a nontrivial 2 sheeted covering S →Ti , i = 1, 2 . The parallel
copies of S in these mapping cylinders, together with T1 and T2 , are the leaves of a
foliation of M . These leaves are the ‘fibers’ of a natural projection p : M →I , with T1
and T2 the pre-images of the endpoints of I . This ‘fiber’ structure on M is not exactly a
fiber bundle, so let us give it a new name: a semi-bundle. Thus a semi-bundle p : M →I
is the union of two twisted I bundles p −1 [0, 1/2 ] and p −1 [1/2 , 1] glued together by a
homeomorphism of the fiber p −1 (1/2 ). For example, in one lower dimension, the Klein
bottle is a semi-bundle with fibers S 1 , since it splits as the union of two Möbius bands.
More generally, one could define semi-bundles with base any manifold with boundary.

The techniques we have been using can also be applied to determine which Seifert
manifolds are irreducible:

Proposition 1.12. A compact connected Seifert-fibered manifold M is irreducible

unless it is S 1 × S 2 , S 1 ×
e S 2 , or RP3 ] RP3 .

Proof: We begin by observing that if M is reducible then there is a horizontal sphere

in M not bounding a ball. This is proved by imitating the argument of the preceding
proposition, with S now a sphere not bounding a ball in M . The only difference is
that when incompressibility was used before, e.g., to eliminate trivial circles of S ∩ A ,
we must now use surgery rather than isotopy. Such surgery replaces S with a pair of
spheres S 0 and S 00 . If both S 0 and S 00 bounded balls, so would S , as we saw in the
proof of Alexander’s theorem, so we may replace S by one of S 0 , S 00 not bounding a
ball. With these modifications in the proof, we eventually get a sphere which is either
horizontal or vertical, but the latter cannot occur since S 2 is not a circle bundle.
If S is a horizontal sphere in M , then as we have seen, M is either a sphere bundle
or a sphere semi-bundle. The only two sphere bundles are S 1 × S 2 and S 1 ×
e S2 . A
sphere semi-bundle is two copies of the twisted I bundle over RP2 glued together via
a diffeomorphism of S 2 . Such a diffeomorphism is isotopic to either the identity or
the antipodal map. The antipodal map extends to a diffeomorphism of the I bundle
RP2 ×
e I , so both gluings produce the same manifold, RP3 ] RP3 . u
t

Note that the three manifolds S 1 × S 2 , S 1 ×

e S 2 , and RP3 ] RP3 do have Seifert
§1.2 Torus Decomposition 19

fiberings. Namely, S 1 ×
e S 2 is S 2 × I with the two ends identified via the antipodal
map, so the I bundle structure on S 2 × I gives S 1 ×
e S 2 a circle bundle structure; and
the I bundle structures on the two halves RP2 ×
e I of RP3 ] RP3 , which are glued
together by the identity, give it a circle bundle structure.
Now we can give a converse to Proposition 1.11:

Proposition 1.13. Let M be a compact irreducible Seifert-fibered 3 manifold. Then

every 2 sided horizontal surface S ⊂ M is essential. The same is true of every con-
nected 2 sided vertical surface except:
(a) a torus bounding a solid torus with a model Seifert fibering, containing at most
one multiple fiber, or
(b) an annulus cutting off from M a solid torus with the product fibering.

Proof: For a 2 sided horizontal surface S we have noted that the Seifert fibering
induces an I bundle structure on M || S , so M || S is the mapping cylinder of a 2 sheeted
covering S q S →T for some surface T . Being a covering space projection, this map
is injective on π1 , so the inclusion of the ∂I subbundle into the I bundle is also
injective on π1 . Therefore S is incompressible. (In case S is a disk, M || S is D 2 × I ,
so S is clearly not ∂ parallel.) Similarly, ∂ incompressibility follows from injectivity
of relative π1 ’s.
Now suppose S is a compressible 2 sided vertical surface, with a compressing
disk D which does not cut off a disk from S . Then D is incompressible in M || S , and
can therefore be isotoped to be horizontal. The Euler characteristic formula in the
P
component of M || S containing D takes the form χ (B) − 1/n = i (1 − 1/qi ) . The
right-hand side is non-negative and ∂B ≠ ∅ , so χ (B) = 1 and B is a disk. Each term
1 − 1/qi is at least 1/2 , so there can be at most one such term, and so at most one
multiple fiber. Therefore this component of M || S is a solid torus with a model Seifert
fibering and S is its torus boundary. (If S were a vertical annulus in its boundary, S
would be incompressible in this solid torus.)
Similarly, if S is a ∂ compressible vertical annulus there is a ∂ compressing disk
D with horizontal boundary, and D may be isotoped to be horizontal in its interior
as well. Again D must be a meridian disk in a solid torus component of M || S with a
model Seifert fibering. In this case there can be no multiple fiber in this solid torus
since ∂D meets S in a single arc. u
t

Note that the argument just given shows that the only Seifert fiberings of S 1 × D 2
are the model Seifert fiberings.
20 Canonical Decomposition §1.2

Uniqueness of Torus Decompositions

We need three preliminary lemmas:

Lemma 1.14. An incompressible, ∂ incompressible annulus in a compact connected

Seifert-fibered M can be isotoped to be vertical, after possibly changing the Seifert
fibering if M = S 1 × S 1 × I , S 1 × S 1 ×
e I (the twisted I bundle over the torus), S 1 ×
e S1 × I
(the Klein bottle cross I ), or S 1 ×
e S1 ×
e I (the twisted I bundle over the Klein bottle).

Proof: Suppose S is a horizontal annulus in M . If S does not separate M then

M || S is the product S × I , and so M is a bundle over S 1 with fiber S , the mapping
torus M × I/{(x, 0) ∼ (ϕ(x), 1)} of a diffeomorphism ϕ : S →S . There are only four
isotopy classes of diffeomorphisms of S 1 × I , obtained as the composition of either
the identity or a reflection in each factor, so ϕ may be taken to preserve the S 1 fibers
of S = S 1 × I . This S 1 fibering of S then induces a circle bundle structure on M in
which S is vertical. The four choices of ϕ give the four exceptional manifolds listed.
If S is separating, M || S is two twisted I bundles over a Möbius band, each ob-
tained from a cube by identifying a pair of opposite faces by a 180 degree twist. Each
twisted I bundle is thus a model Seifert fibering with a multiplicity 2 singular fiber.
All four possible gluings of these two twisted I bundles yield the same manifold M ,
with a Seifert fibering over D 2 having two singular fibers of multiplicity 2 , with S
vertical. This manifold is easily seen to be S 1 ×
e S1 ×
e I. u
t

Lemma 1.15. Let M be a compact connected Seifert manifold with ∂M orientable.

Then the restrictions to ∂M of any two Seifert fiberings of M are isotopic unless M
is S 1 × D 2 or one of the four exceptional manifolds in Lemma 1.14.

Proof: Let M be Seifert-fibered, with ∂M ≠ ∅ . First we note that M contains an

incompressible, ∂ incompressible vertical annulus A unless M = S 1 × D 2 . Namely,
take A = π −1 (α) where α is an arc in the base surface B which is either nonseparating
(if B ≠ D 2 ) or separates the images of multiple fibers (if B = D 2 and there are at least
two multiple fibers). This guarantees incompressibility and ∂ incompressibility of A
by Proposition 1.13. Excluding the exceptional cases in Lemma 1.14, A is then isotopic
to a vertical annulus in any other Seifert fibering of M , so the two Seifert fiberings can
be isotoped to agree on ∂A , hence on the components of ∂M containing ∂A . Since α
could be chosen to meet any component of ∂B , the result follows. u
t

Lemma 1.16. If M is compact, connected, orientable, irreducible, and atoroidal, and

M contains an incompressible, ∂ incompressible annulus meeting only torus compo-
nents of ∂M , then M is a Seifert manifold.

Proof: Let A be an annulus as in the hypothesis. There are three possibilities, indi-
cated in Figure 1.11 below:
§1.2 Torus Decomposition 21

(a) A meets two different tori T1 and T2 in ∂M , and A ∪ T1 ∪ T2 has a neighborhood

N which is a product of a 2 punctured disk with S 1 .
(b) A meets only one torus T1 in ∂M , the union of A with either annulus of T1 || ∂A
is a torus, and A ∪ T1 has a neighborhood N which is a product of a 2 punctured
disk with S 1 .
(c) A meets only one torus T1 in ∂M , the union of A with either annulus of T1 || ∂A
is a Klein bottle, and A ∪ T1 has a neighborhood N which is an S 1 bundle over a
punctured Möbius band.

In all three cases N has the structure of a circle bundle N →B with A vertical.

(a) (b) (c)

Figure 1.11
By hypothesis, the tori of ∂N − ∂M must either be compressible or ∂ parallel
in M . Suppose D is a nontrivial compressing disk for ∂N − ∂M in M , with ∂D a
nontrivial loop in a component torus T of ∂N − ∂M . If D ⊂ N , then N would be a
solid torus S 1 × D 2 by Proposition 1.13, which is impossible since N has more than
one boundary torus. So D∩N = ∂D . Surgering T along D yields a 2 sphere bounding
a ball B 3 ⊂ M . This B 3 lies on the opposite side of T from N , otherwise we would
have N ⊂ B 3 with T the only boundary component of N . Reversing the surgery, B 3
becomes a solid torus outside N , bounded by T .
The other possibility for a component T of ∂N − ∂M is that it is ∂ parallel in
M , cutting off a product T × I from M . This T × I cannot be N since π1 N is non-
abelian, the map π1 N →π1 B induced by the circle bundle N →B being a surjection
to a free group on two generators. So T × I is an external collar on N , and hence can
be absorbed into N .
Thus M is N with solid tori perhaps attached to one or two tori of ∂N − ∂M . The
meridian circles {x}× ∂D 2 in such attached S 1 × D 2 ’s are not isotopic in ∂N to circle
fibers of N , otherwise A would be compressible in M (recall that A is vertical in N) .
Thus the circle fibers wind around the attached S 1 × D 2 ’s a non-zero number of times
in the S 1 direction. Hence the circle bundle structure on N extends to model Seifert
fiberings of these S 1 × D 2 ’s, and so M is Seifert-fibered. u
t

Proof of Theorem 1.9: Only the uniqueness statement remains to be proved. So let
T = T1 ∪ ··· ∪ Tm and T 0 = T10 ∪ ··· ∪ Tn0 be two minimal collections of disjoint
incompressible tori splitting M into manifolds Mj and Mj0 , respectively, which are
either atoroidal or Seifert-fibered. We may suppose T and T 0 are nonempty, otherwise
the theorem is trivial.
22 Canonical Decomposition §1.2

Having perturbed T to meets T 0 transversely, we isotope T and T 0 to eliminate

circles of T ∩ T 0 which bound disks in either T or T 0 , by the usual argument using
incompressibility and irreducibility.
For each Mj the components of T 0 ∩ Mj are then tori or annuli. The annulus
components are incompressible in Mj since they are noncontractible in T 0 and T 0 is
incompressible. Annuli of T 0 ∩ Mj which are ∂ compressible are then ∂ parallel, by
Lemma 1.10, so they can be eliminated by isotopy of T 0 .
A circle C of T ∩ T 0 lies in the boundary of annulus components Aj of T 0 ∩ Mj
and Ak of T 0 ∩ Mk (possibly Aj = Ak or Mj = Mk ) . By Lemma 1.16 Mj and Mk are
Seifert-fibered. If Mj ≠ Mk Lemma 1.14 implies that we can isotope Seifert fiberings
of Mj and Mk so that Aj and Ak are vertical. In particular, the two fiberings of the
torus component Ti of T containing C induced from the Seifert fiberings of Mj and
Mk have a common fiber C . Therefore these two fiberings can be isotoped to agree
on Ti , and so the collection T is not minimal since Ti can be deleted from it.
Essentially the same argument works if Mj = Mk : If we are not in the exceptional
cases in Lemma 1.14, then the circle C is isotopic in Ti to fibers of each of the two
induced fiberings of Ti , so these two fiberings are isotopic, and Ti can be deleted
from T . In the exceptional case Mj = S 1 × S 1 × I , if we have to rechoose the Seifert
fibering to make Aj vertical, then as we saw in the proof of Lemma 1.14, the new
fibering is simply the trivial circle bundle over S 1 × I . The annulus Aj , being vertical,
incompressible, and ∂ incompressible, must then join the two boundary tori of Mj ,
since its projection to the base surface S 1 × I must be an arc joining the two boundary
components of S 1 × I . The two boundary circles of Aj in Ti either coincide or are
disjoint, hence isotopic, so once again the two induced fiberings of Ti are isotopic
and Ti can be deleted from T . The other exceptional cases in Lemma 1.14 cannot
arise since Mj has at least two boundary tori.
Thus T ∩ T 0 = ∅ . If any component Ti of T lies in an atoroidal Mj0 it must be
isotopic to a component Ti0 of T 0 . After an isotopy we then have Ti = Ti0 and M can
be split along this common torus of T and T 0 , and we would be done by induction.
Thus we may assume each Ti lies in a Seifert-fibered Mj0 , and similarly, each Ti0 lies in
a Seifert-fibered Mj . These Seifert-fibered manifolds all have nonempty boundary, so
they contain no horizontal tori. Thus we may assume all the tori Ti ⊂ Mj0 and Ti0 ⊂ Mj
are vertical.
Consider T10 , contained in Mj and abutting Mk0 and M`0 (possibly Mk0 = M`0 ). If
some Ti is contained in Mk0 then Mk0 is Seifert-fibered, by the preceding paragraph.
If no Ti is contained in Mk0 then Mk0 ⊂ Mj , and Mk0 inherits a Seifert fibering from
Mj since ∂Mk0 is vertical in Mj . Thus in any case Mj ∩ Mk0 has two Seifert fiberings:
as a subset of Mj and as a subset of Mk0 . By Lemma 1.15 these two fiberings can be
isotoped to agree on T10 , apart from the following exceptional cases:

— Mj ∩ Mk0 = S 1 × D 2 . This would have T10 as its compressible boundary, so this

§1.2 Torus Decomposition 23

case cannot occur.

— Mj ∩ Mk0 = S 1 × S 1 × I . One boundary component of this is T10 and the other must
be a Ti . (If it were a Ti0 , then T 0 would not be minimal unless Ti0 = T10 , in which
case T = ∅ , contrary to hypothesis.) Then T10 can be isotoped to Ti and we
would be done by induction.
— Mj ∩ Mk0 = S 1 × e I . This has only one boundary component, so Mk0 ⊂ Mj and
e S1 ×
we can change the Seifert fibering of Mk0 to be the restriction of the Seifert fibering
of Mj .
Thus we may assume the fibering of T10 coming from Mk0 agrees with the one coming
from Mj . The same argument applies with M`0 in place of Mk0 . So the Seifert fiberings
of Mk0 and M`0 agree on T10 , and T10 can be deleted from T 0 . u
t

Exercises
1. Show: If S ⊂ M is a 1 sided connected surface, then π1 S →π1 M is injective iff
∂N(S) is incompressible, where N(S) is a twisted I bundle neighborhood of S in M .
2. Call a 1 sided surface S ⊂ M geometrically incompressible if for each disk D ⊂ M
with D ∩ S = ∂D there is a disk D 0 ⊂ S with ∂D 0 = ∂D . Show that if H2 M = 0 but
H2 (M; Z2 ) ≠ 0 then M contains a 1 sided geometrically incompressible surface which
is nonzero in H2 (M; Z2 ) . This applies for example if M is a lens space Lp/2q . Note
that if q > 1 , the resulting geometrically incompressible surface S ⊂ Lp/2q cannot be
S 2 or RP2 , so the map π1 S →π1 Lp/2q is not injective. (See [Frohman] for a study of
geometrically incompressible surfaces in Seifert manifolds.)
3. Develop a canonical torus and Klein bottle decomposition of irreducible nonori-
entable 3 manifolds.
24 Special Classes of 3-Manifolds §2.1

Chapter 2. Special Classes of 3-Manifolds

In this chapter we study prime 3 manifolds whose topology is dominated, in one
way or another, by embedded tori. This can be regarded as refining the results of the
preceding chapter on the canonical torus decomposition

1. Seifert Manifolds
Seifert manifolds, introduced in the last chapter where they play a special role in
the torus decomposition, are among the best-understood 3 manifolds. In this section
our goal is the classification of orientable Seifert manifolds up to diffeomorphism.
We begin with the classification up to fiber-preserving diffeomorphism, which is fairly
straightforward. Then we show that in most cases the Seifert fibering is unique up to
isotopy, so in these cases the diffeomorphism and fiber-preserving diffeomorphism
classifications coincide. But there are a few smaller Seifert manifolds, including some
with non-unique fiberings, which must be treated by special techniques. Among these
are the lens spaces, which we classify later in this section.
The most troublesome Seifert fiberings are those with base surface S 2 and three
multiple fibers. These manifolds are too large for the lens space method to work
and too small for the techniques of the general case. They have been classified by a
study of the algebra of their fundamental groups, but a good geometric classification
has yet to be found, so we shall not prove the classification theorem for these Seifert
manifolds.
All 3 manifolds in this chapter are assumed to be orientable, compact, and con-
nected. The nonorientable case is similar, but as usual we restrict to orientable man-
ifolds for simplicity.

Classification of Seifert Fiberings

We begin with an explicit construction of Seifert fiberings. Let B be a compact
connected surface, not necessarily orientable. Choose disjoint disks D1 , ··· , Dk in the
interior of B , and let B 0 be B with the interiors of these disks deleted. Let M 0 →B 0
be the circle bundle with M 0 orientable. Thus if B 0 is orientable M 0 is the product
B 0 × S 1 , and if B 0 is nonorientable, M 0 is the twisted product in which circles in B 0
are covered either by tori or Klein bottles in M 0 according to whether these circles are
orientation-preserving or orientation-reversing in B 0 . Explicitly, B 0 can be constructed
by identifying pairs of disjoint arcs ai and bi in the boundary of a disk D 2 , and then
we can form M 0 from D 2 × S 1 by identifying ai × S 1 with bi × S 1 via the product of
the given identification of ai and bi with either the identity or a reflection in the S 1
factor, whichever makes M 0 orientable.
Let s : B 0 →M 0 be a cross section of M 0 →B 0 . For example, we can regard M 0 as the
double of an I bundle, that is, two copies of the I bundle with their sub ∂I bundles
§2.1 Seifert Manifolds 25

identified by the identity map, and then we can choose a cross section in one of the
I bundles. The cross section s together with a choice of orientation for the manifold
M 0 allows us to speak unambiguously of slopes of nontrivial circles in the tori of
∂M 0 . Namely, we can choose a diffeomorphism ϕ of each component of ∂M 0 with
S 1 × S 1 taking the cross section to S 1 × {y} (slope 0 ) and a fiber to {x}× S 1 (slope
∞ ). An orientation of M 0 induces an orientation of ∂M 0 , and this determines ϕ up
to simultaneous reflections of the two S 1 factors, which doesn’t affect slopes. We are
assuming the standard fact that each nontrivial circle in S 1 × S 1 is isotopic to a unique
‘linear’ circle which lifts to the line y = (p/q)x of slope p/q in the universal cover
R2 .
From M 0 we construct a manifold M by attaching k solid tori D 2 × S 1 to the torus
components Ti of ∂M 0 lying over ∂Di ⊂ ∂B 0 , attaching by diffeomorphisms taking a
meridian circle ∂D 2 × {y} of ∂D 2 × S 1 to a circle of some finite slope αi /βi ∈ Q in
Ti . The k slopes αi /βi determine M uniquely, since once the meridian disk D 2 × {y}
is attached to M 0 there is only one way to fill in a ball to complete the attaching of
D 2 × S 1 . The circle fibering of M 0 extends naturally to a Seifert fibering of M via a
model Seifert fibering on each attached D 2 × S 1 , since the fibers of M 0 in Ti are not
isotopic to meridian circles of the attached D 2 × S 1 . Namely, fibers have slope ∞ ,
meridians have slope αi /βi ≠ ∞ . Note that the singular fiber in the i th D 2 × S 1 has
multiplicity βi since the meridian disk of D 2 × S 1 is attached to a slope αi /βi circle
and hence meets each fiber of ∂M 0 βi times. Recall that the multiplicity of a singular
fiber is the number of times a transverse disk meets nearby regular fibers.
We use the notation M(±g, b; α1 /β1 , ··· , αk /βk ) for this Seifert-fibered manifold
M , where g is the genus of B , with the sign + if B is orientable and − if B is
nonorientable, and b is the number of boundary components of B . Here ‘genus’
for nonorientable surfaces means the number of RP2 connected summands. Revers-
ing the orientation of M(±g, b; α1 /β1 , ··· , αk /βk ) has the effect of changing it to
M(±g, b; −α1 /β1 , ··· , −αk /βk ) .
We say two Seifert fiberings are isomorphic if there is a diffeomorphism carrying
fibers of the first fibering to fibers of the second fibering.

Proposition 2.1. Every orientable Seifert-fibered manifold is isomorphic to one of the

models M(±g, b; α1 /β1 , ··· , αk /βk ) . Seifert fiberings M(±g, b; α1 /β1 , ··· , αk /βk )
and M(±g, b; α01 /β01 , ··· , α0k /β0k ) are isomorphic by an orientation-preserving dif-
feomorphism iff, after possibly permuting indices, αi /βi ≡ α0i /β0i mod 1 for each i
P P
and, if b = 0 , i αi /βi = i α0i /β0i .

This gives the complete isomorphism classification of Seifert fiberings since the
numbers ±g and b are determined by the isomorphism class of a fibering, which
determines the base surface B , and the Seifert fiberings M(±g, b; α1 /β1 , ··· , αk /βk )
and M(±g, b; α1 /β1 , ··· , αk /βk , 0) are the same.
26 Special Classes of 3-Manifolds §2.1

Proof: Given an oriented Seifert-fibered manifold M , let M 0 be the complement of

open solid torus model-fibered neighborhoods of fibers C1 , ··· , Ck including all multi-
ple fibers. Choose a section s of the circle bundle M 0 →B 0 . As before, this determines
slopes for circles in ∂M 0 , and we see that M has the form M(±g, b; α1 /β1 , ··· , αk /βk ) .
It remains to see the effect on the αi /βi ’s of choosing a different section s .
Let a be an arc in B 0 with endpoints in ∂B 0 . Above this in M 0 lies an annulus
A . We can rechoose s near A so that instead of simply crossing A transversely, it
winds m times around A as it crosses. See Figure 2.1. The effect of this change in s
is to add m to all slopes in the boundary torus of M 0 at one edge of A and subtract
m from all slopes in the boundary torus at the other edge of A . In particular, if both
ends of A lie in the same boundary torus there is no change in boundary slopes.

Figure 2.1
Thus if b ≠ 0 we can choose A connecting the boundary torus near the fiber Ci
with a torus in ∂M , and then change αi /βi by any integer, keeping all other αj /βj ’s
fixed. Similarly, if b = 0 we can add and subtract an integer m from any two αi /βi ’s,
so we can change the αi /βi ’s to any fractions which are congruent mod1 , subject
P
only to the constraint that i αi /βi stays constant.
We claim that any two choices of the section s are related by a sequence of ‘twist’
modifications near vertical annuli A as above, together with homotopies through sec-
tions, which have no effect on slopes. To see this, take disjoint vertical annuli Aj
splitting M 0 into a solid torus. Any two sections can be homotoped, through sec-
tions, to coincide outside a neighborhood of the Aj ’s. Then it is clear that near the
Aj ’s the two sections can be homotoped to coincide, after inserting the appropriate
number of twists. u
t

In the case of Seifert fiberings M(±g, 0; α1 /β1 , ··· , αk /βk ) of closed manifolds,
P
the invariant i αi /βi is called the Euler number of the fibering. When there are no
multiple fibers we can take k = 1 and then the Euler number, which is an integer, is
the obstruction to the existence of a section B →M , i.e., the Euler number vanishes
iff such a section exists. (Exercise.) More generally:

Proposition 2.2. Let M be an orientable Seifert-fibered manifold.

(a) If ∂M ≠ ∅ , horizontal surfaces exist in M .
(b) If ∂M = ∅ , horizontal surfaces exist iff the Euler number of the fibering is zero.
§2.1 Seifert Manifolds 27

Proof: In (a), view M as a circle bundle M0 with model-fibered solid tori Mi attached,
each Mi attaching along an annulus in its boundary. Namely, these annuli in M project
to arcs in the base surface cutting off disks each containing the image of one multiple
fiber. Given a positive integer n there is a horizontal surface S0 ⊂ M0 meeting each
fiber in n points. To see this, we can regard M0 as a quotient of a trivially fibered
solid torus S 1 × D 2 in which certain vertical annuli in S 1 × ∂D 2 are identified. Each
identification can be chosen to be either the identity or a fixed reflection in the S 1
factor. Taking n points xj ∈ S 1 which are invariant (as a set) under the reflection,
the n meridian disks {xj }× D 2 in S 1 × D 2 give the desired surface S0 in the quotient
M0 .
Now let n be a common multiple of the multiplicities qi of the multiple fibers in
the solid tori Mi attached to M0 . In Mi let Si be the union of n/qi meridian disks,
so Si meets each regular fiber in n points. We can isotope Si through horizontal
surfaces so that its n arcs of intersection with the vertical annulus M0 ∩ Mi match up
with the n arcs of S0 in M0 ∩ Mi . Then the union of S0 with the Si ’s is a horizontal
surface in M .
For (b), let M = M(±g, 0; α1 /β1 , ··· , αk /βk ) , with section s : B 0 →M 0 as before.
Let M0 be M with a fibered solid torus neighborhood of a regular fiber in M 0 deleted.
Suppose S0 is a horizontal surface in M0 .

Claim. The circles of ∂S0 in ∂M0 have slope equal to e(M) , the Euler number of M .
This easily implies (b): By (a), such a surface S0 exists. If e(M) = 0 , S0 extends via
meridian disks in M − M0 to a horizontal surface S ⊂ M . Conversely, if a horizontal
surface S ⊂ M exists, the surface S0 = S ∩ M0 must have its boundary circles of slope
0 since these circles bound disks in M − M0 . So e(M) = 0 .
To prove the Claim, let M00 = M0 ∩ M 0 and S00 = S0 ∩ M 0 . The circles of ∂S00
in ∂M 0 have slopes αi /βi , and we must check that the circles of ∂S00 in ∂M0 have
P
slope i αi /βi . This we do by counting intersections of these circles with fibers and
with the section s . Since S00 is horizontal, it meets all fibers in the same number of
points, say n . Intersections with s we count with signs, according to whether the
slope of ∂S00 near such an intersection point is positive or negative. The total number
of intersections of ∂S00 with s is zero because the two intersection points at the end
of an arc of s ∩ S00 have opposite sign. Thus the number of intersections of ∂S00 with
P
s in ∂M0 equals the number of intersections in ∂M 0 . The latter number is i nαi /βi
since the slope of ∂S00 near the i th deleted fiber is αi /βi , which must equal the ratio
of intersection number with s to intersection number with fiber; the denominator of
this ratio is n , so the numerator must be nαi /βi .
P P
Thus the slope of ∂S0 is ( i nαi /βi )/n = i αi /βi = e(M) . u
t
28 Special Classes of 3-Manifolds §2.1

Classification of Seifert Manifolds

Here is a statement of the main result:

Theorem 2.3. Seifert fiberings of orientable Seifert manifolds are unique up to iso-
morphism, with the exception of the following fiberings:
(a) M(0, 1; α/β) , the various model Seifert fiberings of S 1 × D 2 .
(b) M(0, 1; 1/2 , 1/2 ) = M(−1, 1; ), two fiberings of S 1 ×
e S1 ×
e I.
(c) M(0, 0; α1 /β1 , α2 /β2 ) , various fiberings of S 3 , S 1 × S 2 , and lens spaces.
(d) M(0, 0; 1/2 , −1/2 , α/β) = M(−1, 0; β/α) for α, β ≠ 0 .
(e) M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 ) = M(−2, 0; ), two fiberings of S 1 ×
e S1 ×
e S1 .

The two Seifert fiberings of S 1 ×

e S1 ×
e I in (b) are easy to see if we view S 1 ×
e S1 ×
eI
as obtained from S 1 × I × I by identifying S 1 × I × {0} with S 1 × I × {1} via the diffeo-
morphism ϕ which reflects both the S 1 and I factors. One fibering of S 1 ×
e S1 ×
e I then
comes from the fibers S 1 × {y}× {z} of S 1 × I × I and the other comes from the fibers
{x}× {y}× I ; in the latter case the two fixed points of ϕ give multiplicity-two fibers.

Figure 2.2
Note that the examples in (a) and (b) generate the remaining examples: The fiber-
ings in (c) are obtained by gluing together two model fiberings from (a); (d) is obtained
by gluing a model fibering into each of the two fiberings in (b); and (e) is simply the
double of (b).
In most cases the Theorem is a consequence of the following:

Proposition 2.4. If M1 and M2 are irreducible orientable Seifert-fibered manifolds

which are diffeomorphic, then there is a fiber-preserving diffeomorphism provided
that M1 contains vertical incompressible, ∂ incompressible annuli or tori, and M2
contains no horizontal incompressible, ∂ incompressible annuli or tori.

Proof: First we do the case of closed manifolds. In the base surface B1 of M1 choose
two transversely intersecting systems C and C 0 of disjoint 2 sided circles not passing
through singular points (projections of singular fibers), such that:

(1) No circle of C or C 0 bounds a disk containing at most one singular point.

(2) The components of B1 || (C ∪ C 0 ) are disks containing at most one singular point.
§2.1 Seifert Manifolds 29

(3) No component of B1 || (C ∪ C 0 ) is a disk bounded by a single arc of C and a single

arc of C 0 , and containing no singular point.

For example, one can choose C to be a single circle, then construct C 0 from suitably
chosen arcs in B1 || C , matching their ends across C . (Details left as an exercise.) Let
T and T 0 be the collections of incompressible vertical tori in M1 lying over C and C 0 .
Let f : M1 →M2 be a diffeomorphism. Since M2 contains no incompressible hor-
izontal tori, we may isotope f to make f (T ) vertical in M2 by Proposition 1.11. The
circles of T ∩ T 0 are nontrivial in T , so we may isotope f to make the circles of
f (T ∩ T 0 ) vertical or horizontal in each torus of f (T ) . By condition (3) the annuli of
T 0 || (T ∩ T 0 ) are incompressible and ∂ incompressible in M1 || T , so we may isotope f ,
staying fixed on T , to make the annuli of f (T 0 ) || f (T ∩ T 0 ) vertical or horizontal in
M2 . If any of these were horizontal, they would be part of horizontal tori in M2 , so we
now have f (T ∪ T 0 ) vertical in M2 . Since f (T ∩ T 0 ) is vertical, we can can isotope f
to be fiber-preserving on T ∪ T 0 , and then make f fiber-preserving in a neighborhood
M10 of T ∪ T 0 .
By condition (2) the components of M1 − M10 are solid tori, so the same is true for
M2 −M20 , where M20 = f (M10 ) . Choose an orientation for M1 and a section for M10 →B10 .
Via f these choices determine an orientation for M2 and a section for M20 →B20 . Note
that f induces a diffeomorphism of B10 onto B20 , so the closed surfaces B1 and B2
are diffeomorphic. The fractions αi /βi for corresponding solid tori of M1 − M10 and
M2 = M20 must be the same since these are the slopes of boundary circles of meridian
disks, and f takes meridian disks to meridian disks (up to isotopy). Thus M1 and
M2 have the same form M(±g, 0; α1 /β1 , ··· , αk /βk ) , completing the proof for closed
manifolds.
For non-empty boundary the proof is similar but easier. Let T be a collection of
incompressible, ∂ incompressible vertical annuli splitting M1 into solid tori. Isotope
f first to make f (T ) vertical, then to make f fiber-preserving in a neighborhood of
T ∪ ∂M1 . The rest of the argument now proceeds as in the closed case. u
t

Let us see how close Proposition 2.4 comes to proving the Theorem. Consider
first the case of irreducible manifolds with non-empty boundary. Then vertical incom-
pressible, ∂ incompressible annuli exist except in the model fiberings M(0, 1; α/β) .
To see when horizontal annuli exist we apply the Euler characteristic formula χ (B) −
P
χ (S)/n = i (1 − 1/βi ) . In the present case S is an annulus, so we have χ (B) =
P 2
i (1 − 1/βi ) ≥ 0 , so B is a disk, annulus, or Möbius band. If B = D , we have
P
1 = i (1 − 1/βi ) , a sum of terms 1/2 , 2/3 , 3/4 , ···, so the only possibility is 1 = 1/2 + 1/2
P
and the fibering M(0, 1; 1/2 , 1/2 ). If B = S 1 × I , we have 0 = i (1 − 1/βi ) so there are
no multiple fibers and we have a product fibering M(0, 2; ) of S 1 × S 1 × I . Similarly,
if B = S 1 ×
e I we have the fibering M(−1, 1; ) of S 1 ×
e S1 ×
e I . The manifolds S 1 × S 1 × I
and S 1 ×
e S1 ×
e I are not diffeomorphic since they deformation retract onto a torus and
30 Special Classes of 3-Manifolds §2.1

a Klein bottle, respectively, and so are of distinct homotopy types. Thus the Theorem
follows in the case of irreducible manifolds with non-empty boundary.
For closed irreducible manifolds the analysis is similar but more complicated.
Incompressible vertical tori exist unless the base surface B is S 2 and there are at
most three multiple fibers, or B is RP2 and there is at most one multiple fiber. If
horizontal tori exist we must have χ (B) ≥ 0 , so there are the following cases:
P
— B = S 2 and 2 = i (1 − 1/βi ) . Since 1/2 ≤ 1 − 1/βi < 1 there must be either
three or four multiple fibers. With four multiple fibers, the multiplicities must
all be 2 , so we have the fibering M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 ) since the Euler number
must be zero. With three multiple fibers we have 1/β1 + 1/β2 + 1/β3 = 1 so
(β1 , β2 , β3 ) = (2, 3, 6) , (2, 4, 4) , or (3, 3, 3) , up to permutations. We leave it for
the reader to check that the fibering must be isomorphic to M(0, 0; 1/2 , −1/3 , −1/6 ),
M(0, 0; 1/2 , −1/4 , −1/4 ), or M(0, 0; 2/3 , −1/3 , −1/3 ), allowing orientation-reversing iso-
morphisms.
P
— B = RP2 and 1 = i (1 − 1/βi ) , implying just two multiple fibers, of multiplicity
2 . Since the Euler number is zero the fibering must be M(−1, 0; 1/2 , −1/2 ).
— B = S 1 × S 1 or S 1 ×
e S 1 with no multiple fibers and Euler number zero, the fiberings
M(1, 0; ) and M(−2, 0; ) .
Thus we have the following six manifolds:

M1 = M(1, 0; ) = S 1 × S 1 × S 1
M2 = M(−2, 0; ) = M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 ) = S 1 ×
e S1 ×
e S1
M3 = M(0, 0; 2/3 , −1/3 , −1/3 )
M4 = M(0, 0; 1/2 , −1/4 , −1/4 )
M5 = M(0, 0; 1/2 , −1/3 , −1/6 )
M6 = M(−1, 0; 1/2 , −1/2 )

Each of these seven fiberings has Euler number zero, hence does in fact contain a
horizontal surface S . By the Euler characteristic formula, χ (S) = 0 , so S is either a
torus or Klein bottle. In the latter case S is one-sided, and the boundary of a tubular
neighborhood of S is a horizontal torus. So horizontal tori exist in all seven Seifert
fiberings. As we saw in §1.2 following Proposition 1.11, this implies the manifolds
M1 M6 are torus bundles or torus semi-bundles.
There remains the case of reducible Seifert manifolds. As shown in the proof
of Proposition 1.13 there must exist a horizontal sphere in this case, which implies
that the manifold is closed and the Euler number of the Seifert fibering is zero. By
the Euler characteristic formula the base surface B must be S 2 or RP2 , with at most
three multiple fibers in the first case and at most one multiple fiber in the second
case. In the latter case there is in fact no multiple fiber since the Euler number is
zero, so we have the Seifert fibering M(−1, 0; ) , and it is not hard to see that this
§2.1 Seifert Manifolds 31

manifold is RP3 ] RP3 . In the case B = S 2 it is an exercise with fractions to rule

out the possibility of three multiple fibers, using the Euler characteristic formula and
the fact that the Euler number is zero. Then if there are at most two multiple fibers,
we are in the exceptional case (c) of the classification theorem, with a Seifert fibering
M(0, 0; α/β, −α/β) of S 1 × S 2 .

To complete the main classification theorem, there remain three things to do:
(1) Show the manifolds M1 M6 above are all distinct, and have only the seven Seifert
fiberings listed.
(2) Classify the Seifert manifolds which possess fiberings over S 2 with at most two
multiple fibers.
(3) Show that the different Seifert fiberings over S 2 with three multiple fibers are all
distinct manifolds, distinct also from the manifolds in (2).
We shall do (2) in the following subsection, and in §2.2 we shall do (1) as part of a
general classification of torus bundles and torus semi-bundles. As for (3), the only
way which seems to be known for doing this is to show these manifolds are all distin-
guished by their fundamental groups, apart from a few special cases where geometric
techniques are available. We shall not cover this in the present version of these notes;
see [Orlik] for a proof.

Classification of Lens Spaces

Let Lp/q be the manifold obtained by attaching two solid tori S 1 × D 2 together by
a diffeomorphism ϕ : S 1 × ∂D 2 →S 1 × ∂D 2 sending a meridian {x}× ∂D 2 to a circle of
slope p/q , where we use the convention that a meridian has slope ∞ and a longitude
S 1 × {y} has slope 0 . The fraction p/q determines Lp/q completely since circles in
S 1 × ∂D 2 are determined up to isotopy by their slopes, and once the meridian disk
of the second S 1 × D 2 has been attached to the first S 1 × D 2 there is only one way
to attach the remaining 3 ball. Note that L1/0 = S 1 × S 2 , and L0/1 = S 3 since S 3 =
∂(D 2 × D 2 ) = ∂D 2 × D 2 ∪ D 2 × ∂D 2 .
The manifolds Lp/q are Seifert-fibered in many ways, by taking model fiberings
on the two S 1 × D 2 ’s whose restrictions to S 1 × ∂D 2 correspond under ϕ ∈ SL2 (Z) ; a
model fibering on the second S 1 × D 2 with boundary fibers of slope u/v extends to
a model fibering on the first S 1 × D 2 provided ϕ(u/v) ≠ 1/0 . These Seifert fiberings
have base surface S 2 and at most two multiple fibers. Conversely, such a Seifert-
fibered manifold is the union of two solid tori with boundaries identified, hence is an
Lp/q .
We shall exclude the special case L1/0 = S 1 × S 2 from now on.
The manifold Lp/q can also be obtained as the quotient space S 3 /Zq where Zq is
the group of rotations of S 3 generated by ρ(z1 , z2 ) = (e2π i/q z1 , e2π ip/q z2 ) , where S 3
is regarded as the unit sphere in C2 . This action can be pictured in R3 , with the point
32 Special Classes of 3-Manifolds §2.1

at infinity added to get S 3 = S 1 × D 2 ∪ D 2 × S 1 , as a 1/q rotation of the first factor

and a p/q rotation of the second factor.

p/ q

1/q

Figure 2.3
Each quotient S × D /Zq and D × S /Zq is a solid torus, so the quotient S 3 /Zq has
1 2 2 1

the form Lp0 /q0 . To see that p 0 /q0 = p/q , note that slope p circles on the common
boundary torus S 1 × S 1 are invariant under ρ , hence become longitudes in S 1 × D 2 /Zq .
The boundary circle of a meridian disk of D 2 × S 1 /Zq then intersects the longitude of
S 1 × D 2 /Zq p times and the meridian q times, hence has slope p/q , up to a sign
determined by orientations.
In particular π1 Lp/q ≈ Zq . This can also be seen directly from the definition of
Lp/q since the meridian disk of the second solid torus is attached to S 1 × D 2 along a
circle which wraps around the S 1 factor q times, and the subsequent attaching of a
3 ball has no effect on π1 . Since π1 Lp/q ≈ Zq , the number q is uniquely determined
by Lp/q . (We take q non-negative always.)
For p there is some ambiguity, however. First, we can rechoose the longitude in
S × D 2 by adding any number of twists around the meridian. This changes slopes by
1

an arbitrary integer, so Lp0 /q = Lp/q if p 0 ≡ p mod q . Also, reversing the orientation

of S 1 × D 2 changes the sign of slopes, so L−p/q = Lp/q . (We are not specifying an ori-
1 2
 two S × D halves of L
entation of Lp/q .) Finally, the roles of the  p/q can
 be reversed,
p −q
replacing ϕ by ϕ−1 . Taking ϕ = s p ∈ SL2 (Z) , then ϕ−1 = −s p , so we obtain
r q

L−r /q = Lp/q , where r ≡ p −1 mod q since pr − qs = 1 . Summarizing, Lp0 /q = Lp/q if

p 0 ≡ ±p ±1 mod q .

Theorem 2.5. Lp0 /q is diffeomorphic to Lp/q iff p 0 ≡ ±p ±1 mod q .

Proof: We shall show that there is, up to isotopy, only one torus T in Lp/q bounding
a solid torus on each side. This implies the theorem since such a T yields a decompo-
sition Lp/q = S 1 × D 2 ∪ϕ S 1 × D 2 , and the only ambiguities in the slope p 0 /q0 defining
such a representation of Lp/q are the ones considered in the paragraph preceding the
theorem: choice of longitude and orientation in the boundary of the first solid torus,
§2.1 Seifert Manifolds 33

and switching the two solid tori.

Let T be the standard torus bounding solid tori on each side, from the definition
of Lp/q . Let Σ be the core circle S 1 × {0} of the first solid torus S 1 × D 2 bounded by
T . The union of a meridian disk D of the second S 1 × D 2 with the radial segments in
slices {x}× D 2 of the first S 1 × D 2 joining points of ∂D to Σ is a 2 complex ∆ ⊂ Lp/q .
Abstractly, ∆ is obtained from the circle Σ by attaching a disk D 2 by a q to 1 covering
map ∂D 2 →Σ .
Let T 0 be another torus in Lp/q bounding solid tori on both sides. The concentric
copies of T 0 in these two solid tori, together with the two core circles, define a singular
foliation F of Lp/q , whose leaves can also be described as the level sets of a function
f : Lp/q →[0, 1] .
The first thing we do is put ∆ in good position with respect to F , as follows.
Perturb Σ to be disjoint from the two singular leaves (circles) of F , and so that f || Σ is
a morse function with all critical points in distinct levels. Near local maxima of f || Σ ,
perturb the q sheets of ∆ to lie ‘above’ Σ , as in Figure 2.4. Likewise, at local minima
of f || Σ make ∆ lie ‘below’ Σ .

Figure 2.4
Now, keeping ∆ fixed in a small neighborhood of Σ , perturb ∆ to be transverse to the
two singular leaves of F and so that away from these two leaves f is a morse function
on ∆ − Σ with all saddles in levels distinct from each other and from the levels of the
critical points of f || Σ .
The foliation F then induces a singular foliation on ∆ , hence also on D 2 via the
quotient map D 2 →∆ . The singularities of the foliation of D 2 are of the following
types:

— ‘centers’ in the interior of D 2 , where f || (∆ − Σ) has local maxima or minima, in

particular where ∆ meets the two singular leaves of F .
— saddles in the interior of D 2 .
— ‘half-saddles’ in ∂D 2 , at local maxima or minima of f || Σ .

Figure 2.5 shows the various possible configurations for a singular leaf in D 2 contain-
ing a saddle or half-saddle.

(a) (b) (c) (d) (e) (f)

Figure 2.5
Recall that all saddles and half-saddles on ∆ lie in distinct levels. However, each local
34 Special Classes of 3-Manifolds §2.1

maximum or minimum of f || Σ gives q half-saddles in D 2 in the same level, and some

of these may be contained in the same singular leaf, as indicated by the dashed line
in case (e). The saddles and half-saddles in cases (a) and (e) are called essential, all
others inessential. In (a) the singular leaf divides D 2 into four quarter disks, and in
(e) the singular leaf cuts off two or more half disks from D 2 .
Let D be a quarter or half disk containing no other quarter or half disks, and let α
be the arc D ∩ ∂D 2 . The two endpoints of α are part of the same singular leaf, and as
we move into the interior of α we have pairs of points joined by a nonsingular leaf in
D 2 . Continuing into the interior of α , this pairing of points of α by nonsingular leaves
in D 2 continues until we reach an inessential saddle of type (b) or an inessential half-
saddle, since D was chosen as a smallest quarter or half disk, and the singular leaves
in cases (c) and (d) do not meet ∂D 2 . An inessential saddle of type (b) presents no
obstacle to continuing the pairing of points of α , which can therefore continue until
we reach an inessential half-saddle. At this point, however, the two paired points of
α coalesce into one point, the half-saddle point itself. We conclude from this that D
contains exactly one inessential half-saddle.
In particular, note that the projection α→Σ must be an embedding on the interior
of α since f || Σ has at least two critical points. The two endpoints of α might have
the same image in Σ , however. If this happens, D must be a half disk and the two
endpoints of α must be half-saddles at the same critical point of f || Σ . In this case
isotoping α rel ∂α across D to ∂D − α induces an isotopy of Σ so that it lies in a
leaf of F . This is a desirable outcome, as we shall see below, and our next goal is to
reduce to this case.
So suppose the disk D is a half disk and α has a half-saddle at only one of its
endpoints. The configuration is as shown in Figure 2.6. Here we can isotope Σ , and
hence ∆ , via isotopy extension, by dragging α across D to a new position, shown in
dashed lines, with f || Σ having two fewer critical points.
Σ
D1
x0

Figure 2.6
The other possibility is that D is a quarter disk, with the configuration in Fig-
ure 2.7 below. In this case we again isotope Σ by pushing α across D , dragging the
other q − 1 sheets of ∆ along behind, enlarging them by parallel copies of D . This
does not change the number of critical points of f || Σ , while the number of essential
saddles decreases, since the essential saddle in ∂D becomes q half-saddles and no
new essential saddles are created. (The inessential half-saddle in α becomes q − 1
§2.1 Seifert Manifolds 35

inessential saddles. Inessential saddles in the interior of D are also replicated. In

addition, other essential saddles in D 2 − D may become inessential.)

x0
Σ

Figure 2.7
Thus after finitely many steps we are reduced to the earlier case when Σ is iso-
toped into a leaf of F , which we may assume is the torus T 0 . The circle Σ is nontrivial
in T 0 since it generates π1 Lp/q ≈ Zq and we may assume q > 1 . The torus T is the
boundary of a solid torus neighborhood of Σ . Shrinking this neighborhood if neces-
sary, we may assume T ∩ T 0 consists of two circles, parallel copies of Σ in T 0 , so T
intersects each of the solid tori X 0 and Y 0 bounded by T 0 in an annulus, and likewise
T 0 intersects each of the solid tori X and Y bounded by T in an annulus.
We now appeal to the following fact: An annulus A ⊂ S 1 × D 2 with ∂A consisting
of nontrivial, nonmeridional circles in S 1 × ∂D 2 must be ∂ parallel. For such an A
must be incompressible since π1 A→π1 (S 1 × D 2 ) is injective, hence must be either
∂ incompressible or ∂ parallel. In the former case A could be isotoped to be vertical
in the product Seifert fibering of S 1 × D 2 , but then A would again be ∂ parallel.
This applies to each of the four annuli of T − T 0 and T 0 − T Note that meridional
boundary circles cannot occur since Σ is not nullhomotopic in Lp/q . We conclude that
the two annuli of T − T 0 can be isotoped to the two annuli of T 0 − T , as indicated in
Figure 2.8, so T is isotopic to T 0 . u
t

T0
Σ

Figure 2.8

Exercises
1. What happens if αi /βi = 1/0 in the manifold M(±g, b; α1 /β1 , ··· , αk /βk ) ?
36 Special Classes of 3-Manifolds §2.2

2. Torus Bundles and Semi-Bundles

In the preceding section we encountered a few Seifert manifolds which are bundles
over the circle with torus fibers. Most torus bundles are not Seifert-fibered, however,
though they share one essential feature with most Seifert manifolds: Their topology is
dominated by incompressible tori. This makes them easy to classify, and Theorem 2.6
below reduces this classification to understanding conjugacy classes in GL2 (Z) , an
algebraic problem which has a nice geometric solution described later in this section.
After the case of torus bundles we move on to a closely-related class of manifolds,
torus semi-bundles, which always have a torus bundle as a 2 sheeted covering space.
The torus bundles and torus semi-bundles which are not Seifert-fibered have a
special interest because they are the manifolds in one of Thurston’s eight geometries,
called solvgeometry.

The Classification of Torus Bundles

A torus bundle has the form of a mapping torus, the quotient Mϕ = T × I/(x, 0) ∼
(ϕ(x), 1) for some diffeomorphism ϕ ∈ GL2 (Z) of the torus T . This is because
every diffeomorphism f : T →T is isotopic to a linear diffeomorphism ϕ ∈ GL2 (Z) .
Moreover, ϕ is uniquely determined by f since it is essentially the map on H1 (T ) ≈ Z2
induced by f . By Proposition 1.6, Mϕ is irreducible since its universal cover is R3 .
Once again we restrict attention to orientable 3 manifolds for simplicity. For Mϕ
this means restricting ϕ to be in SL2 (Z) .

Theorem 2.6. Mϕ is diffeomorphic to Mψ iff ϕ is conjugate to ψ±1 in GL2 (Z) .

Proof: The ‘if’ half is easy: Applying a diffeomorphism f : T →T in each slice T × {y}
of T × I has the effect of conjugating ϕ by f when we form the quotient Mϕ =
T × I/(x, 0) ∼ (ϕ(x), 1) . And by switching the two ends of T × I we see that Mϕ−1 is
diffeomorphic to Mϕ .
The main step in the converse is the following:

Lemma 2.7. An incompressible

  surface in Mϕ is isotopic to a union of torus fibers,
1 0
unless ϕ is conjugate to ± n 1 .

Proof: Let S ⊂ Mϕ be incompressible. We may suppose S has transverse intersection

with a fiber T0 of Mϕ . After we eliminate trivial circles of S ∩ T0 by isotopy of S in
the usual way, the surface S 0 = S || T0 is incompressible in Mϕ || T0 = T × I . (Given
a compressing disk D for S 0 in T × I , then ∂D bounds a disk in S , which must lie
in S 0 since any intersections with T0 would be trivial circles of S ∩ T0 .) If S 0 is
∂ compressible, it must have ∂ parallel annulus components. These can be eliminated
by pushing them across T0 , decreasing the number of circles of S ∩ T0 . So we may
assume S 0 is incompressible and ∂ incompressible in T × I .
§2.2 Torus Bundles and Semi-Bundles 37

Thinking of T × I as a trivial circle bundle, S 0 can be isotoped to be either hori-

zontal or vertical, hence to consist either of vertical tori parallel to T0 , i.e., fibers of
Mϕ , or annuli whose boundary circles have the same slope in both ends T × {0} and
T × {1} . In the latter case, ϕ must preserve this slope in order for S 0 to glue together
2
to form  original surface S . This means ϕ has an eigenvector in Z , so is conjugate
 the
1 0
to ± n 1 . u
t

Continuing the proof of the theorem, suppose f : Mϕ→Mψ is a diffeomorphism.

1 0
The lemma implies that if ψ is not conjugate to ± n 1 , we can isotope f to take
a fiber T of Mϕ to a fiber of Mψ . We may assume that T is the image of T × ∂I
in the quotient mapping torus T × I/(x, 0) ∼ (ϕ(x), 1) , and similarly that f (T ) is
the image of T × ∂I in T × I/(x, 0) ∼ (ψ(x), 1) . Then f is the quotient of a diffeo-
morphism F : T × I →T × I . We may assume F takes each of T × {0} and T × {1} to
themselves, rather than interchanging them, by replacing ψ by ψ−1 if necessary,
switching the ends of T × I . Let f0 and f1 be the restrictions of F to T × {0} and
T × {1} . Since (x, 0) is identified with (ϕ(x), 1) , it must be that (f0 (x), 0) is identi-
fied with (f1 ϕ(x), 1) . But (f0 (x), 0) is identified with (ψf0 (x), 1) , so we conclude
that f1 ϕ = ψf0 . This equation implies that ϕ and ψ are conjugate in GL2 (Z) since
f0 and f1 are isotopic to the same diffeomorphism in GL2 (Z) , as they induce the
same map on H1 (T ) , namely the map induced by F on H1 (T × I) .

 There
 remains the case that ψ is conjugate, hence we may assume equal, to
1 0
± n 1 . Since ψ preserves slope ∞ , Mψ is a circle bundle with fibers the slope ∞
circles in the torus fibers of Mψ . It is not hard to see that, with a suitable choice of
orientation for Mψ , this circle bundle, as a Seifert manifold, is M(ε, 0;n) with ε equal
1 0
to either +1 or −2 , the sign of ε being the same as the sign of ± n 1 . The bun-
dle projection Mψ →S 1 factors as a composition of two circle bundles Mψ →B →S 1 ,
where B is the base surface of the Seifert fibering.
We may isotope a diffeomorphism f : Mϕ →Mψ so that the incompressible torus
f (T ) , with T a fiber of Mϕ , is either horizontal or vertical in Mψ . If it is vertical,
we may rechoose the torus fibering of Mψ so that f (T ) is a fiber, by rechoosing the
fibering B →S 1 so that the image circle of f (T ) is a fiber. (In the case that B is a Klein
bottle, this circle is 2 sided and nonseparating in B , so its complement
 must be an
1 0
annulus, not a Möbius band.) In the new fibering, the new ψ is still ± n 1 since this
matrix was determined by the Seifert fibering M(ε, 0; n) , as we saw. Now we have f
taking T to a torus fiber of Mψ , and the argument can be completed as before.
If f (T ) is horizontal we must have the Euler number n equal to zero, so ψ =
±I . We cannot have ψ = −I since then the base surface of M(−2, 0; 0) is nonori-
entable, so the circle fibers of M(−2, 0; 0) cannot be coherently oriented, which means
that the horizontal surface f (T ) is separating, splitting M(−2, 0; 0) into two twisted
I bundles. If ψ = I , then Mψ is the 3 torus, and after composing f with a diffeomor-
38 Special Classes of 3-Manifolds §2.2

phism in GL3 (Z) we may assume the map on π1 induced by f takes the subgroup
π1 T into the Z2 subgroup corresponding to the torus fiber of Mψ , with cyclic image
in π1 B . Then when we isotope f to make f (T ) horizontal or vertical, f (T ) cannot
be horizontal since for a horizontal surface the projection to B is a finite-sheeted
covering space, with π1 image of finite index in π1 B , hence non-cyclic. This reduces
us to the previous case that f (T ) was vertical. u
t

We now determine which Seifert-fibered manifolds are torus bundles Mϕ . If the

fiber torus T is isotopic to a vertical surface, then since its complement is T × I , which
has only the product Seifert fibering, up to isomorphism, the Seifert fibering
  of Mϕ
1 0
must be of the form M(1, 0; n) or M(−2, 0; n) , with ϕ conjugate to ± n 1 . If T is
isotopic to a horizontal surface, Mϕ must be one of the manifolds M1 M5 from §2.1:

M1 = M(1, 0; ) = S 1 × S 1 × S 1 , with ϕ = I
M2 = M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 ) = S1 ×
e S1 ×
e S 1, with ϕ = −I
 
−1 −1
M3 = M(0, 0; 2/3 , −1/3 , −1/3 ), with ϕ = 1 0
 
M4 = M(0, 0; 1/2 , −1/4 , −1/4 ), with ϕ = 01 −1 0
 
M5 = M(0, 0; 1/2 , −1/3 , −1/6 ), with ϕ = 01 −1 1

For M2 and M4 , ϕ is rotation by 180 and 90 degrees, respectively. For M3 and M5 , if

we view Z2 not as the usual square lattice in R2 but as the regular hexagonal lattice,
ϕ is rotation by 120 and 60 degrees, respectively. So for M1 M5 , ϕ is periodic, of
period 1, 2, 3, 4 , and 6 , respectively. We leave it for the reader to verify that with these
ϕ ’s, Mϕ is the given Seifert fibering, with circle fibers the images of the segments
{x}× I ⊂ T × I in the mapping torus Mϕ . This just amounts to checking that the
multiple fibers have the multiplicities given, since the Euler number is necessarily
zero; see §2.1.
In particular, the theorem implies that none of these Seifert manifolds M1 M5
are diffeomorphic to each other or to the manifolds M(1, 0; n) or M(−2, 0; n) with
n ≠ 0 , since their ϕ ’s are not conjugate, being of different orders. This completes
the classification of Seifert-fibered torus bundles, left over from §2.1.

The Classification of Torus Semi-bundles

Let N = S 1 ×
e S1 ×
e I , the twisted I bundle over the Klein bottle. Two copies of
N glued together by a diffeomorphism ϕ of the torus ∂N form a closed orientable
manifold Nϕ , which we called a torus semi-bundle in §2.1: Nϕ is foliated by tori
parallel to ∂N , together with a Klein bottle at the core of each copy of N . We can
identify ϕ , or rather its isotopy class, with an element of GL2 (Z) via a choice of
coordinates in ∂N , which we can get by thinking of N as the quotient of S 1 × S 1 × I
obtained by identifying (x, y, z) with (−x, ρ(y), ρ(z)) , where the ρ ’s are reflections.
§2.2 Torus Bundles and Semi-Bundles 39

Then the x and y coordinates define ‘horizontal’ and ‘vertical’ in ∂N , and an arbitrary
choice of orientations of these directions completes the choice of coordinates for ∂N .
If we compose ϕ on the left or the right with a diffeomorphism of ∂N which
extends to a diffeomorphism of N , then we get the same manifold Nϕ . For example,
we can compose ϕ with diffeomorphisms
  reflecting either the horizontal or vertical
±1 0
directions in ∂N , with matrices 0 ±1 , since the reflection ρ acting in either S 1 factor
of S 1 × S 1 × I passes down to a diffeomorphism of the quotient N . As we shall see
below, these four diffeomorphisms are, up to isotopy, the only diffeomorphisms of
∂N which extend over N .
Replacing ϕ by ϕ−1 obviously produces the same manifold Nϕ , by interchanging
the roles of the two copies of N in Nϕ . Thus we have proved the easier ‘if’ half of the
following:
   
±1 0 ±1 0
Theorem 2.8. Nϕ is diffeomorphic to Nψ iff ϕ = 0 ±1 ψ±1 0 ±1 in GL2 (Z) , with
independent choices of signs understood.

Proof: First, consider an incompressible, ∂ incompressible surface S ⊂ N . Give N

the Seifert fibering M(−1, 1; ) as a circle bundle over the Möbius band, and isotope S
to be either horizontal or vertical. If S is vertical, it consists either of tori parallel to
∂N , or nonseparating annuli with boundaries of slope ∞ in ∂N , with respect to the
coordinates in ∂N chosen earlier. If S is horizontal it consists of separating annuli
with boundaries of slope 0 . This can be seen as follows. Orient ∂S by the positive
direction of the x coordinate in N . Let s be a section of the circle bundle N , with
∂s of slope 0 . The arcs of S ∩ s pair off the points of ∂S ∩ ∂s into pairs of opposite
sign, so S and s have algebraic intersection number 0 . Since all the components of
∂S have parallel orientations, each component must have intersection number 0 with
∂s , hence must have slope 0 .
Now let S be an incompressible surface in Nψ . As in the proof of Lemma 2.7,
we may isotope S so that for each of the two copies N1 and N2 of N whose union is
Nψ , S ∩ Ni is an incompressible, ∂ incompressible surface Si in Ni . By the preceding
paragraph, either S consists of parallel copies of the torus T splitting Nψ into N1
and N2 , or ψ takes slope 0 or ∞ to slope 0 or ∞ . We shall see later that this means
Nψ is Seifert-fibered.
Suppose f : Nϕ →Nψ is a diffeomorphism, and let T be the torus in Nϕ splitting
Nϕ into two copies of N . Applying the preceding remarks to S = f (T ) , there are two
possibilities:
(1) We can isotope f so that f (T ) is the torus splitting Nψ into two copies of N . Then
ϕ must be obtained from ψ by composing on the left and right by diffeomorphisms
of ∂N which extend to diffeomorphisms of N . Such diffeomorphisms must preserve
slope 0 and slope ∞ , since these are the unique slopes of the boundaries of incom-
pressible, ∂ incompressible separating and nonseparating annuli in N , respectively,
40 Special Classes of 3-Manifolds §2.2

as noted earlier. The theorem is thus proved in this case.

(2) We can isotope f so that f (T ) meets each of the two copies of N in incom-
pressible, ∂ incompressible annuli. These must be separating, since T is separat-
ing, so they must have boundaries
  of slope 0 . Thus ψ preserves slope 0 , having,
1 n
we may assume, the form 0 1 . In this case if we take all the slope 0 circles in
the torus and Klein bottle fibers of the torus semi-bundle Nψ we obtain the Seifert
fibering M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 , n), with Euler number n , as the reader can verify by
inspection. The torus f (T ) can be isotoped to be vertical in this Seifert fibering,
since horizontal surfaces in M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 , n) are nonseparating (because the
base surface of this Seifert fibering is orientable, hence the fiber circles are coher-
ently orientable, making any horizontal surface the fiber of a surface bundle). We
can then rechoose the torus semi-bundle structure on Nψ so that f (T ) becomes a
fiber torus. (The other torus and Klein bottle fibers are also unions of circle fibers of
M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 , n).) The ψ
 for
 the new torus semi-bundle structure is clearly
1 n
the same as the old one, namely 0 1 . So this reduces us to case (1). u
t

Which Seifert fiberings are torus semi-bundles Nϕ ? If the torus fiber is vertical
in the Seifert fibering, it splits Nϕ into two Seifert-fibered copies of N . Since N has
just the two Seifert fiberings M(−1, 1; ) and M(0, 1; 1/2 , 1/2 ), the possibilities are:
 
1 0
(1) M(−2, 0; n) , with ϕ = n 1
   
−1 0 1 n 1
(2) M(−1, 0; /2 , /2 , n), with ϕ = 1 n , or equivalently 1 0
1
 
1 n
(3) M(0, 0; 1/2 , 1/2 , −1/2 , −1/2 , n), with ϕ = 0 1

On the other hand, if the torus fiber of Nϕ is horizontal, then as we saw in §2.1, the
Seifert fibering is either the one in (1) or (2) with n = 0 .
The theorem implies that all the manifolds listed in (1)-(3) are distinct, modulo
changing the sign of n and the coincidence of (1) and (3) when n = 0 . In most cases
this was already proved in §2.1. The one exception is the manifold M(−1, 0; 1/2 , −1/2 ),
labelled M6 in §2.1, which we did not then distinguish from the other manifolds in
(1)-(3).
For each Nϕ there is a torus bundle Mψ which is a natural 2 sheeted covering
space of Nϕ . Over each of the two copies of N in Nϕ , Mψ consists of the double cover
S 1 × S 1 × I →N by which N was defined. The first copy of S 1 × S 1 × I is glued to the
second by ϕ along S 1 × S 1 × {0} and by τϕτ along S 1 × S 1 × {1} , where τ(x, y) =
(−x, ρ(y)) , as in the definition of N . Thus ψ = τϕ−1 τϕ . Since the 2 sheeted
covering Mψ →Nϕ takes fibers to fibers, it fits into a commutative Mψ −
−
−−
−→ Nϕ
−

−
→

diagram as at the right, where the two vertical maps are the bundle
and semi-bundle projections, and the map S 1 →I is the ‘folding’ map S
1
−
−−
−−
→I
which factors out by the reflection ρ . The bundle Mψ →S 1
can be regarded as the
pullback of the semi-bundle Nϕ →I via the folding map S 1
→I . There are similar
§2.2 Torus Bundles and Semi-Bundles 41

n to– 1 folding maps I →I , folding up I like an n segmented carpenter’s ruler, and

Nϕ →I can be pulled back to other torus semi-bundles via these folding maps.

The Geometry of SL 2 ( Z )
 
A matrix
a b
c d ∈ SL2 (Z) gives rise to a linear fractional transformation z ,
(az + b)/(cz + d) of the Riemann sphere C ∪ {∞} . Since this has real coefficients
it takes R ∪ {∞} to itself, preserving orientation since the determinant is positive.
Hence it takes the upper half plane to itself since all linear fractional transformations
preserve orientation of C ∪ {∞}. Let H denote the upper half plane, including its
boundary points R ∪ {∞} . The action of SL2 (Z) on H extends to an action of GL2 (Z)
on H if we identify points in the upper and lower half planes via complex conjugation,
since a linear fractional transformation with real coefficients takes conjugate points
to conjugate points. Then an element of GL2 (Z) preserves or reverses orientation of
H according to whether  determinant is +1 or −1 .
 its
∈ GL2 (Z) let `A be the semicircle in H orthogonal to ∂ H
a b
For a matrix A = c d
with endpoints at the rational numbers c/a and d/b in ∂ H . We orient `A from c/a
to d/b . In case a or b is zero, `A is a vertical line in H with lower endpoint at d/b
or c/a , respectively, and other ‘endpoint’ at ∞ = ±1/0 . Note that there are exactly
four A ’s with the same oriented `A , obtained by multiplying either column of A by
−1 . When the interior of H is taken as a model for the hyperbolic plane, the `A ’s are
geodesics, so we shall refer to them as ‘lines.’
The collection of all the lines `A is invariant under the action of GL2 (Z) on H .
Namely, if A, B ∈ GL2 (Z) then A(`B ) = `AB , as the reader can readily verify using the
fact that linear fractional transformations preserve angles and take circles and lines
in C to circles or lines. Note that this action is transitive: `B = BA−1 (`A ) .

Proposition 2.9. The collection of all lines `A forms a tessellation T of H by ideal

triangles: triangles with their vertices on ∂ H .

See Figure 2.9 for two pictures of this tesselation T , the second picture being a
redrawing of the first to make H a disk.
4/ 1/ 3/
3/ 3 1 4
2 2/
3
5/ 3/
3 5
2/ 1/
1 2
5/ 2/5
2
3/1 1/3

4 /1 1/4

1 /0 0/
1
- 4/ - 1/
1 4
- 3/ - 1/
1 3
- 1/ - 1/ 0/ 1/ 1/
1 2 1 2 1 - 5/ - 2/
2 5
- 2/ - 1/ 1/3 2/
3 3 3 - 2/ - 1/
1 2
- 5/ - 3/
3 5
- 3/ - 2/
2 - 4/ - 3
3 1/1 - 3 /4

Figure 2.9
42 Special Classes of 3-Manifolds §2.2

Proof: First we check that no two lines `A cross. Since the action of GL2 (Z) on the
a b
`A ’s is transitive, it suffices to consider an `A crossing `I . If A = c d and `A is
oriented from left to right (which can be achieved by switching the columns of A if
necessary), then c/a < 0 < d/b . We may assume c is negative and a , b , and d are
positive. But then ad − bc > 1 , which is impossible since elements of GL2 (Z) have
determinant ±1 .
Next we observe that any two rational numbers are joined by an edgepath con-
sisting of finitely many edges `A , each two successive edges in the edgepath having
a common vertex.Forgiven c/a ∈ Q with (a, c) = 1 , there exist integers b, d with
a b
ad − bc = ±1 , so c d ∈ GL2 (Z) . Thus each rational is the endpoint of a line `A . To
 
get an edgepath joining c/a to 0/1 , use the fact that ac db can be reduced to a diago-
nal matrix by a sequence of elementary column operations, each operation consisting
of adding an integer multiple of one column to the other column. (This is essentially
the Euclidean algorithm in Z .) This produces a finite sequence of matrices in GL2 (Z) ,
each pair of successive matrices having a column in common, so the corresponding
sequence of `A ’s forms an edgepath. The final edge has endpoints 0 = 0/±1 and
∞ = ±1/0 , since the diagonal matrix has ±1 ’s down the diagonal.
The three `A ’s with endpoints 0/1 , 1/0 , and 1/1 form one triangle in H . Let
U be the union of all the images of this triangle under the action of GL2 (Z) , with all
vertices in ∂ H deleted. To finish the proof we need to show that U = int(H) . We do
this by showing that U is both open and closed in int(H) . Openness is clear, since
GL2 (Z) acts transitively on the `A ’s, and the edge `I is shared by the two triangles
whose third vertices are +1 and −1 . If U were not closed, there would be an infinite
sequence zi ∈ U converging to a point z ∈ int(H) − U . We may assume the zi ’s
lie in distinct lines `Ai . Since no `Ai ’s cross, there would have to be a limit line `
containing z . Since z ∉ U , ` is not an `A . Neither endpoint of ` can be rational
since every line through a rational point crosses lines `A (true for 1/0 , hence true for
all rationals by transitivity) and if ` crossed an `A so would some `Ai . On the other
hand, if both endpoints of ` are irrational, then again ` cannot cross any `A ’s, and
in this case it would not be possible to join rationals on one side of ` to rationals on
the other side by edgepaths consisting of `A ’s, contradicting an earlier observation.
Thus U must be closed in int(H) . u
t

For the action of GL2 (Z) on H , only ±I act as the identity. So we have an injection
of the quotient P GL2 (Z) = GL2 (Z)/±I into the symmetry group of the tesselation T .
This is in fact an isomorphism, since P GL2 (Z) acts transitively on oriented edges, and
also contains orientation-reversing maps of H . The index-two subgroup P SL2 (Z) is
the group of orientation-preserving symmetries of T .
 
a b
An element A = c d ∈ SL2 (Z) not equal to ±I has one or two fixed points in
H , the roots of the quadratic z = (az + b)/(cz + d) . There are three possibilities:
§2.2 Torus Bundles and Semi-Bundles 43

(1) There are two distinct real irrational roots r1 and r2 . Then the line (semicircle)
in H joining r1 and r2 is invariant under A . The infinite sequence of triangles of T
meeting this line is arranged in a periodic pattern, invariant under A , which acts as a
translation of this strip, as in Figure 2.10, with ai triangles having a common vertex
on one side of the strip, followed by ai+1 triangles with a common vertex on the other
side, etc.
a1 a1

a2 a2n a2

Figure 2.10
The cycle (a1 , ··· , a2n ) is a complete invariant of the conjugacy class of A in P GL2 (Z) ,
by our earlier comment that P GL2 (Z) is the full symmetry group of T . The fixed points
r1 and r2 are slopes of eigenvectors of A . The two corresponding eigenvalues have
the same sign since A ∈ SL2 (Z) , and this sign together with the cycle (a1 , ··· , a2n ) is
a complete invariant of the conjugacy class of A in GL2 (Z) . A specific representative
of this conjugacy class is
! ! ! !
1 0 1 a2 1 0 1 a2n
± ···
a1 1 0 1 a3 1 0 1

The conjugacy class of A−1 is represented by the inverse cycle (a2n , ··· , a1 ) with the
same sign of the eigenvalues. Geometrically, we can think of the conjugacy class of
the pair {A, A−1 } as the triangulated cylinder obtained from the strip in Figure 5.2
by factoring out by the translation A , together with a sign. Diffeomorphism classes
of torus bundles Mϕ with ϕ having distinct real eigenvalues correspond bijectively
with such triangulated, signed cylinders, by Theorem 2.6.

(2) There is only one rational root r . In this case the triangles having r as a vertex
form a degenerate infinite strip, with quotient a cone divided into some number n
of triangles. The number n together with the sign of the eigenvalue ±1 of A is a
complete invariant of theconjugacy
 class of {A, A−1 } in GL2 (Z) . A representative of
1 n
this conjugacy class is ± 0 1 .

(3) There are two complex conjugate roots. The root in int(H) gives a fixed point of
A in int(H) , so A acts either by rotating a triangle 120 degrees about its center or
by rotating an edge 180 degrees about its center. Since P GL2 (Z) acts transitively on
triangles and edges, these rotations of H fall into two conjugacy classes in P GL2 (Z) ,
according to the angle of rotation. In GL2 (Z) we thus have just the conjugacy classes
of the matrices of orders 2, 3, 4 , and 6 discussed earlier.
44 Special Classes of 3-Manifolds §2.2

Now let us use the tesselation T to make explicit the equivalence relation on ele-
ments of GL2 (Z) corresponding to diffeomorphism
 of torus semi-bundles, described
a b
in Theorem 2.8. An element A = c d ∈ GL2 (Z) determines an oriented edge `A in
T , as above. The other
 three A ’s with the same oriented `A differ by multiplication on
±1 0
the right by 0 ±1 , so give the same torus semi-bundle NA . Multiplication on the left
 
by ±1 0 ±1 corresponds either to the identity of H or a reflection of H across the edge
0

`I , so we may normalize A by choosing it so that `A lies in the right half of H . Let

T (A) be the smallest connected union of triangles of T containing `I and `A . This
is a finite triangulated strip with two distinguished oriented edges in its boundary, `I
and `A , as in Figure 2.11.
1/0 c/a

`I `A

0/1 d/b

Figure 2.11
In degenerate cases the top or bottom edges of the strip can reduce to a single point.
We leave it as an exercise for the reader to check that T (A−1 ) is obtained from T (A)
by simply interchanging the labelling of the two oriented edges `I and `A . Thus dif-
feomorphism classes of torus semi-bundles correspond bijectively with isomorphism
classes of triangulated strips with oriented left and right edges.
As another exercise, the reader can show that the torus bundle double covering
NA , as described earlier in this section, corresponds to the cylinder (or cone) obtained
by gluing two copies of T (A) together by the identity map of their boundary, with
the sign +1 if the two ends of T (A) have parallel orientation and the sign −1 in the
opposite case; this assumes we are not in the degenerate case that T (A) is just a
single edge, with no triangles.
§3.1 The Loop and Sphere Theorems 45

Chapter 3. Homotopy Properties

1. The Loop and Sphere Theorems
These are the basic results relating homotopy theory, specifically π1 and π2 , to
more geometric properties of 3 manifolds.

The Loop Theorem

Let M be a 3 manifold-with-boundary, not necessarily compact or orientable.

Theorem 3.1. If there is a map f : (D 2 , ∂D 2 )→(M, ∂M) with f || ∂D 2 not nullhomo-

topic in ∂M , then there is an embedding with the same property.

Proof: The first half of the proof consists of covering space arguments which reduce
the problem to finitely many applications of the relatively simple special case that f
is at most two-to-one. The second half of the proof then takes care of this special
case. It is a curious feature of the proof that at the end of the first step one discards
the original f completely and starts afresh with a new and nicer f .
It seems more convenient to carry out the first half of the proof in the piecewise
linear category. Choose a triangulation of M and apply the simplicial approximation
theorem (for maps of pairs) to homotope the given f to a map f0 which is simplicial
in some triangulation of the domain D 2 and which still Dn > Vn > M n
−−−−−
satisfies the same hypotheses as f . We now construct →
→
−−−− pn
a diagram as at the right. To obtain the bottom row, let
−−−−

...
−−−−−
D0 be f0 (D 2 ) , a finite subcomplex of M , and let V0 be a fn →
−−−−

p2
neighborhood of D0 in M that is a compact 3 manifold
D1 > V1 > M1
−−−−

f1 −−−−−
−−→
deformation retracting onto D0 . The classical construc- →
tion of such a V0 is to take the union of all simplices in −
−−
f
p 1

D −−−
−−→ D0 > V0 > M
2 0
the second barycentric subdivision of M that meet D0 .
In particular, V0 is connected since D0 is connected. If V0 has a connected 2 sheeted
cover p1 : M1 →V0 , then f0 lifts to a map f1 : D 2 →M1 since D 2 is simply-connected.
Let D1 = f1 (D 2 ) and let V1 be a neighborhood of D1 in M1 chosen as before. This
gives the second row of the diagram. If V1 has a connected 2 sheeted cover we can
repeat the process to construct a third row, and so on up the tower.
To see that we must eventually reach a stage where the tower cannot be continued
further, consider the covering pi−1 (Di−1 )→Di−1 . Both these spaces have natural sim-
plicial structures with simplices lifting the simplices of D0 . The nontrivial deck trans-
formation τi of the covering space pi−1 (Di−1 )→Di−1 is a simplicial homeomorphism.
Since Mi is connected, so is pi−1 (Di−1 ) since Mi deformation retracts to pi−1 (Di−1 )
by lifting a deformation retraction of Vi−1 to Di−1 . The set pi−1 (Di−1 ) = Di ∪ τi (Di )
is connected, so Di ∩ τi (Di ) must be nonempty. This means that τi must take some
46 Homotopy Properties §3.1

simplex of Di to another simplex of Di , distinct from the first simplex since τi has
no fixed points. These two simplices of Di are then identified in Di−1 , so Di−1 is a
quotient of Di with fewer simplices than Di . Since the number of simplices of Di is
bounded by the number of simplices in the original triangulation of the source D 2 , it
follows that the height of the tower is bounded.
Thus we arrive at a Vn having no connected 2 sheeted cover. This says that π1 Vn
has no subgroup of index two, so there is no nontrivial homomorphism π1 Vn →Z2 ,
and hence H 1 (Vn ; Z2 ) = Hom(H1 (Vn ), Z2 ) = 0 . Consider now the following three
terms of the long exact sequence of homology groups for the pair (Vn , ∂Vn ) :

H2 ( Vn , ∂ Vn ; Z 2 ) −
−−
−−
→ H1 ( ∂Vn ; Z 2 ) −
−−
−−
→ H1 ( Vn ; Z 2 )
≈

≈
H ( Vn ; Z 2 ) H ( Vn ; Z 2 )
1 1

The isomorphism on the left comes from Poincaré duality and the one on the right
comes from the universal coefficient theorem. Since the two outer terms of the exact
sequence vanish, so does H1 (∂Vn ; Z2 ) . This implies that all the components of the
compact surface ∂Vn are 2 spheres.
In the component of ∂Vi containing fi (∂D 2 ) , let Fi = (p1 ◦ ··· ◦ pi )−1 (∂M) , a
compact surface if the neighborhoods Vi are chosen reasonably. Let Ni ⊂ π1 Fi be
the kernel of (p1 ◦ ··· ◦ pi )∗ : π1 Fi →π1 (∂M) . Note that [fi || ∂D 2 ] ∉ Ni by the initial
hypothesis on f , so Ni is a proper normal subgroup of π1 Fi .
At the top of the tower the surface Fn is a planar surface, being a subsurface of
a sphere. Hence π1 Fn is normally generated by the circles of ∂Fn . Since Nn ≠ π1 Fn ,
some circle of ∂Fn must represent an element of π1 Fn − Nn . This circle bounds a
disk in ∂Vn . Let gn : D 2 →Vn be this embedding of a disk, with its interior pushed
into Vn so as to be a proper embedding. We have [gn |∂D 2 ] ∉ Nn . We can take gn to
be a smooth embedding and work from now on in the smooth category.
The rest of the proof consists of descending the tower step by step, producing
embeddings gi : D 2 →Vi with [gi |∂D 2 ] ∉ Ni . In the end we will have an embedding
g0 : D 2 →V0 ⊂ M with [g0 |∂D 2 ] ∉ N0 , which says that g0 |∂D 2 is not nullhomotopic
in ∂M , and the proof will be complete.
Consider the inductive step of producing gi−1 from gi . Since pi is a 2 sheeted
cover, we can perturb gi so that the immersion pi gi has only simple double curves,
either circles or arcs, where two sheets of pi gi (D 2 ) cross transversely. Our aim is to
modify gi so as to eliminate each of these double curves in turn.
Suppose C is a double circle. A neighborhood N(C) of C in pi gi (D 2 ) is a bundle
over C with fiber the letter X . Thus N(C) is obtained from X × [0, 1] by identifying
X × {0} with X × {1} by some homeomorphism given by a symmetry of X . A 90 degree
rotational symmetry is impossible since pi is two-to-one. A 180 degree rotational
symmetry is also impossible since it would force the domain D 2 to contain a Möbius
band. For the same reason a symmetry which reflects the X across one of its crossbars
§3.1 The Loop and Sphere Theorems 47

is impossible. The two remaining possibilities are the identity symmetry, with N(C) =
X × S 1 , and a reflection of X across a horizontal or vertical line. Note that a reflection
can occur only if M is nonorientable.
In the case that N(C) = X × S 1 there are two circles in D 2 mapping to C . These
circles bound disks D1 and D2 in D 2 . If D1 and D2 are nested, say D1 ⊂ D2 , then we
redefine gi on D2 to be gi |D1 , smoothing the resulting corner along C . This has the
effect of eliminating C from the self-intersections of pi gi (D 2 ) , along with any self-
intersections of the immersed annulus pi gi (D2 − D1 ) . If D1 and D2 are disjoint we
modify gi by interchanging its values on the disks D1 and D2 , then smoothing the re-
sulting corners along C , as shown in Figure 3.1(a). Thus we eliminate the circle C from
the self-intersections of pi gi (D 2 ) without introducing any new self-intersections.

(a) (b)
Figure 3.1
In the opposite case that N(C) is not X × S 1 we modify the immersed disk pi gi (D 2 )
in the way shown in Figure 3.1(b), replacing each cross-sectional X of N(C) by two
disjoint arcs. This replaces an immersed annulus by an embedded annulus with the
same boundary, so the result is again an immersed disk with fewer double circles.
Iterating these steps, all double circles can be eliminated without changing gi |∂D 2 .
Now consider the case of a double arc C . The two possibilities for how the two
corresponding arcs in D 2 are identified to C are shown in Figure 3.2.
β β

γ α γ α

δ δ
(a) (b)
Figure 3.2
In either case, Figure 3.3 shows two ways to modify gi to produce new immersions
gi0 and gi00 eliminating the double arc C .

g i0 g i00
Figure 3.3
The claim is that either [gi0 || ∂D 2 ] ∉ Ni or [gi00 || ∂D 2 ] ∉ Ni . To see this, we break the
2
loop gi |∂D up into four pieces α , β , γ , δ as indicated in Figure 3.2. Then in case
48 Homotopy Properties §3.1

(a) we have
αβγδ = (αγ)δ−1 (αβ−1 γδ−1 )−1 (αγ)δ
|| || ||
gi0 || ∂D 2 gi00 || ∂D 2 gi0 || ∂D 2
and in case (b) we have
αβγδ = (αγ −1 )(γδ)−1 (αγ −1 )−1 (αδγβ)(γδ)
|| || ||
gi0 || ∂D 2 gi0 || ∂D 2 gi00 || ∂D 2
So in either case, if gi0 |∂D 2 and gi00 |∂D 2 were both in the normal subgroup Ni , so
would gi |∂D 2 be in Ni .
We choose the new gi to be either gi0 or gi00 , whichever one preserves the condi-
tion [gi |∂D 2 ] ∉ Ni . Repeating this step, we eventually get gi with pi gi an embedding
gi−1 , with [gi−1 |∂D 2 ] ∉ Ni−1 . u
t

Corollary 3.2 (Dehn’s Lemma). If an embedded circle in ∂M is nullhomotopic in M ,

it bounds a disk in M .

Proof: First delete from ∂M everything but a neighborhood of the given circle C ,
producing a new manifold M 0 with ∂M 0 an annulus. (If C had a Möbius band neigh-
borhood in ∂M , it would be an orientation-reversing loop in M , which is impossible
since it is nullhomotopic.) Apply the loop theorem to M 0 to obtain a disk D 2 ⊂ M 0
with ∂D 2 nontrivial in ∂M 0 , hence isotopic to C . u
t

Corollary 3.3. Let S ⊂ M be a 2 sided surface. If the induced map π1 S →π1 M is

not injective for some choice of basepoint in some component of S , then there is a
disk D 2 ⊂ M with D 2 ∩ S = ∂D 2 a nontrivial circle in S .

Proof: Let f : D 2 →M be a nullhomotopy of a nontrivial loop in S . We may assume f

is transverse to S , so f −1 (S) consists of ∂D 2 plus possibly some circles in D 2 − ∂D 2 .
If such a circle in D 2 − ∂D 2 gives a trivial loop in the component of S it maps to, then
we can redefine f so as to eliminate this circle (and any other circles inside it) from
f −1 (S) . So we may assume no circles of f −1 (S) are trivial in S . Then the restriction
of f to the disk bounded by an innermost circle of f −1 (S) (this circle may be ∂D 2
itself) gives a nullhomotopy in M − S of a nontrivial loop in S . Now apply the loop
theorem to (M − ∂M) || (S − ∂S) . u
t

Here is another application of the loop theorem:

Proposition 3.4. A compact connected orientable prime 3 manifold M with π1 M ≈

Z is either S × S or S 1 × D 2 .
1 2

Proof: First let us do the case ∂M ≠ ∅ . Then ∂M can contain no spheres, otherwise
primeness would imply M = B 3 , contradicting π1 M ≈ Z . To restrict the possibilities
for ∂M further we need the following basic result.
§3.1 The Loop and Sphere Theorems 49

Lemma 3.5. If M is a compact orientable 3 manifold then the image of the boundary
map H2 (M, ∂M)→H1 (∂M) has rank equal to one half the rank of H1 (∂M) .

Here ‘rank’ means the number of Z summands in a direct sum splitting into cyclic
groups. Equivalently, we can use homology with Q coefficients and replace ‘rank’ by
‘dimension.’

Proof: With Q coefficients understood, consider the commutative diagram

H2 ( M , ∂ M ) −
−−
−−
→ H1 ( ∂ M ) −−−−∗−→ H1 ( M )
∂ i

≈
∗
1
H (M ) −−−−
i
−→ H 1( ∂ M ) −
−−
−−
δ
→ H 2( M , ∂M )
where the vertical isomorphisms are given by Poincaré duality. Since we have coeffi-
cients in a field, the map i∗ is the dual of i∗ , obtained by applying Hom(−, Q) . Linear
algebra then implies that dim Ker i∗ = dim Coker i∗ . Hence dim Im ∂ = dim Ker i∗ =
dim Coker i∗ = dim Coker ∂ which implies the result. u
t

Continuing with the proof of the Proposition, H1 (M) has rank one, hence so does
1
H (M) ≈ H2 (M, ∂M) and we conclude that ∂M must be a single torus. The map
π1 ∂M →π1 M is not injective, so the loop theorem gives a disk D 2 ⊂ M with ∂D 2
nontrivial in ∂M . Splitting M along this D 2 gives a manifold with boundary S 2 , and
this leads to a connected sum decomposition M = N ] S 1 × D 2 . Since M is prime, we
conclude that M = S 1 × D 2 .
When ∂M = ∅ we use another basic fact about 3 manifolds:

Lemma 3.6. For a 3 manifold M , every class in H2 (M) is represented by an em-

bedded closed orientable surface S ⊂ M , and similarly every class in H2 (M, ∂M) is
represented by an embedded compact orientable surface (S, ∂S) ⊂ (M, ∂M) . If M is
orientable, S may be chosen in both cases so that the inclusion of each component of
S into M induces an injective map on π1 .

Proof: Taking simplicial homology with respect to some triangulation of M , repre-

sent a class in H2 (M) by a cycle z , a sum of the oriented 2 simplices in M with
integer coefficients. Reversing orientations of some simplices, we may assume all
coefficients are non-negative. To desingularize this 2 cycle z into a surface, first re-
place each 2 simplex σ with coefficient nσ by nσ nearby parallel copies of itself,
all with the same boundary ∂σ . Next, look in a small disk transverse to a 1 simplex
of M . The copies of 2 simplices incident to this 1 simplex have induced normal ori-
entations, which we can think of as either clockwise or counterclockwise around the
1 simplex. Since z is a cycle, the number of incident 2 simplices with clockwise ori-
entation equals the number with counterclockwise orientation. Thus there must be
two consecutive 2 simplices with opposite orientation. The edges of these along the
given 1 simplex can be joined together and pushed off the 1 simplex. Repeat this
50 Homotopy Properties §3.1

step until z is ‘resolved’ to a surface whose only singularities are at vertices of M . To

eliminate these isolated singularities, consider a small ball about a vertex. The surface
meets the boundary of this ball in disjoint embedded circles, so we can cap off these
circles by disjoint embedded disks in the interior of the ball. Doing this for all vertices
produces an embedded oriented surface S , with a natural map to the 2 skeleton of
M taking the fundamental class for S to z . The case of relative homology is treated
in similar fashion.
To obtain the final statement of the lemma, suppose the map π1 S →π1 M is not
injective for some component of S . Then Corollary 3.3 gives a disk D ⊂ M realizing
this non-injectivity. Surgering S along D produces a new surface which is obviously
in the same homology class. Since S is compact, surgery simplifies S , either splitting
one component into two components of lower genus, in case ∂D separates S , or in
the opposite case, reducing the genus of a component of S . So after finitely many
steps we reach the situation that each component of S is π1 injective. u
t

Returning to the proof of Proposition 3.4 in the case of closed M , we have Z ≈

H1 (M) ≈ H 1 (M) ≈ H2 (M) . By the lemma we may represent a nontrivial class in
H2 (M) by a closed embedded surface S with each component π1 injective. Since
π1 M ≈ Z , each component of S must then be a 2 sphere. The separating spheres are
trivial in homology, so can be discarded, leaving at least one nonseparating sphere.
As in Proposition 1.4 this gives a splitting M = N ] S 1 × S 2 , finishing the proof. u
t

Dropping the hypothesis that M is prime, the proof shows that M = N ] S 1 × S 2

or N ] S 1 × D 2 . Then π1 M ≈ π1 N ∗ Z , hence N is simply-connected, since a nontrivial
free product is necessarily nonabelian. This implies that ∂N , hence also ∂M , consists
of spheres, since as we showed in the proof of the loop theorem, H1 (N; Z2 ) = 0 is
sufficient to obtain this conclusion. Let N 0 be obtained from N by capping off its
boundary spheres by balls. Thus N 0 is a closed simply-connected 3 manifold. The
next proposition describes the homotopy type of such an N 0 .

Proposition 3.7. A closed simply-connected 3 manifold is a homotopy sphere, i.e.,

homotopy equivalent to S 3 . A compact 3 manifold is contractible iff it is simply-
connected and has boundary a 2 sphere.

Proof: Let P be a closed simply-connected 3 manifold. It is orientable since π1 P = 0 .

By Poincaré duality and the Hurewicz theorem we have π2 P ≈ H2 (P ) ≈ H 1 (P ) ≈
H1 (P ) = 0 . Then π3 P ≈ H3 (M) ≈ Z , and we have a degree one map S 3 →P . This
induces an isomorphism on all homology groups. Since both S 3 and P are simply-
connected, Whitehead’s theorem implies that the map S 3 →P is a homotopy equiva-
lence.
The second assertion can be proved in similar fashion. u
t
§3.1 The Loop and Sphere Theorems 51

The Poincaré Conjecture (still unproved) is the assertion that S 3 is the only ho-
motopy 3 sphere, or equivalently, B 3 is the only compact contractible 3 manifold.

The Sphere Theorem

Here is the general statement:

Theorem 3.8. Let M be a connected 3 manifold. If π2 M ≠ 0 , then either

(a) there is an embedded S 2 in M representing a nontrivial element in π2 M , or
(b) there is an embedded 2 sided RP2 in M such that the composition of the cover
S 2 →RP2 with the inclusion RP2 > M represents a nontrivial element of π2 M .

A 2 sided RP2 ⊂ M can exist only if M is nonorientable, so for orientable M

the theorem asserts that there is an S 2 ⊂ M which is nontrivial in π2 M , if π2 M is
nonzero. An example of a nonorientable manifold where RP2 ’s are needed to repre-
sent a nontrivial element of π2 is S 1 × RP2 . This has S 1 × S 2 as a double cover, so π2
is Z . However, S 1 × RP2 is irreducible, by Proposition 1.12, so any embedded S 2 in
S 1 × RP2 is nullhomotopic.
We remark that it is not true in general for orientable M that every element of
π2 M is represented by an embedded sphere. For example, π2 (S 1 × S 2 ) ≈ Z , but as
we saw in Proposition 1.4, a separating S 2 in S 1 × S 2 bounds a ball since S 1 × S 2 is
prime, while a nonseparating S 2 is a slice {x}× S 2 in a product structure S 1 × S 2
hence represents a generator of π2 .

Corollary 3.9. Let M be a compact connected orientable irreducible 3 manifold with

f.
universal cover M
(a) If π1 M is infinite, and in particular if ∂M ≠ ∅ , then M is a K(π , 1) , i.e., πi M = 0
for all i > 1 , or equivalently, Mf is contractible.
f is a homotopy 3 sphere.
(b) If π1 M is finite, then either M = B 3 or M is closed and M

Proof: Since M is irreducible, the Sphere Theorem implies that π2 M is zero. It fol-
lows that π3 M ≈ π3 Mf ≈ H3 Mf by the Hurewicz theorem. If π1 M is infinite, Mf is
noncompact, so H3 Mf = 0 . Thus π3 M = 0 . By Hurewicz again, π4 M ≈ π4 M
f ≈ H4 M f.
Since H4 of a 3 manifold is zero, we get π4 M = 0 . Similarly, all higher homotopy
groups of M are zero. Now assume π1 M is finite. If ∂M ≠ ∅ then by Lemma 3.5,
f is
∂M must consist of spheres and irreducibility implies M = B 3 . If ∂M = ∅ , then M
also closed since the covering M →M is finite-sheeted, and so by Proposition 3.7 M
f f
is a homotopy sphere. u
t

Proposition 3.10. An embedded sphere S 2 ⊂ M is zero in π2 M iff it bounds a

compact contractible submanifold of M . A 2 sided RP2 ⊂ M is always nontrivial in
π2 M .

Proof: Let Se be any lift of the nullhomotopic sphere S ⊂ M to the universal cover
f . Via the isomorphism π2 M ≈ π2 M
M f , Se is nullhomotopic in M
f , hence it is also
52 Homotopy Properties §3.1

f , bounding a compact submanifold N

homologically trivial in M f . Rechoosing Se
e of M
e e
to be an innermost lift of S in N , we may assume N contains no other lifts of S . Then
f→M to N
the restriction of the covering projection M e →N , as
e is a covering space N
in the proof of Proposition 1.6. This covering is 1 sheeted, i.e., a homeomorphism,
since it is 1 sheeted over S . By the van Kampen theorem, N e is simply-connected
f is. By Proposition 3.7, N is contractible.
since M
f→M ,
For the case of a 2 sided RP2 ⊂ M , consider the orientable double cover M
corresponding to the subgroup of π1 M represented by orientation-preserving loops
in M . The given RP2 lifts to a unique S 2 ⊂ Mf . If this is nullhomotopic, the first
2
half of the proposition implies that this S bounds a contractible submanifold N e of
f . The nontrivial deck transformation of M
M f takes Ne to itself since the given RP2
is 2 sided in M . So the restriction of the covering projection to Ne is a double cover
e →N , with ∂N = RP2 . This is impossible by the following lemma. (Alternatively, the
N
e →N
Lefschetz fixed point theorem implies that the nontrivial deck transformation N e
e is contractible.)
must have a fixed point since N u
t

Lemma 3.11. If N is a compact 3 manifold, then the Euler characteristic χ (∂N) is

even.

Proof: By the proof of Lemma 3.5 with Z2 coefficients now instead of Q coefficients,
we see that H1 (∂N; Z2 ) is even-dimensional, from which the result follows. u
t

The converse statement, that a closed surface of even Euler characteristic is the
boundary of some compact 3 manifold, is easily seen to be true also.
Here is another consequence of the Sphere Theorem:

Proposition 3.12. In a compact orientable 3 manifold M there exists a finite collec-

tion of embedded 2 spheres generating π2 M as a π1 M module.

Proof: By the prime decomposition there is a finite collection S of disjoint spheres

in M such that each component of M || S is an irreducible manifold with punctures.
We claim that the spheres in S , together with all the spheres in ∂M , generate π2 M
as a π1 M module. For let f : S 2 →M be given. Perturb f to be transverse to S , so
that f −1 (S) is a collection of circles in S 2 . Since the components of S are simply-
connected, we may homotope f , staying transverse to S , so that f takes each circle
in f −1 (S) to a point. This means that f is homotopic to a linear combination, with
coefficients in π1 M , of maps fj : S 2 →M − S . It suffices now to show that if P is
irreducible-with-punctures then π2 P is generated as a π1 P module by the spheres
in ∂P . This is equivalent to saying that if Pe→P is the universal cover, then π2 Pe is
generated by the spheres in ∂ Pe . To show this, let Pe0 be Pe with its boundary spheres
filled in with balls B` . We have a Mayer-Vietoris sequence:
§3.1 The Loop and Sphere Theorems 53

=
∼ ∼
H2 ( ∪` ∂ B` ) −
−−
−→ H2 ( ∪` B` ) ⊕ H2 ( P ) −
−−−
→ H2 ( P 0 )

≈
≈
∼ ∼
π2 ( ∪` ∂ B` ) −
−−
−−
−−
−−
−−
−→ π2 ( P ) −
−−−
−−
−−
−−→ π 2 ( P 0)
Now Pe0 is the universal cover of the irreducible manifold P 0 obtained by filling in
the boundary spheres of P , so π2 Pe0 ≈ π2 P 0 = 0 by the Sphere Theorem. Since
S S
H2 ( ` B` ) = 0 , we conclude that π2 ( ` ∂B` )→π2 (Pe) is surjective, i.e., π2 Pe is gener-
ated by the spheres in its boundary. u
t

Proof of the Sphere Theorem: f→M be the universal cover. By Lemma 3.6
Let p : M
we may represent a nonzero class in π2 M ≈ π2 M f ≈ H2 M f by an embedded closed
f , each component of which is π1 injective, hence a sphere
orientable surface S ⊂ M
f = 0 . Some component will be nontrivial in H2 M
since π1 M f , hence in π2 M
f . Call this
component S0 .
The rest of the proof will be to show that the sphere S0 can be chosen so that the
union of all the images τ(S0 ) under the deck transformations τ of M f is a collection
of disjoint embedded spheres. Then the restriction p : S0 →p(S0 ) is a covering space,
so its image p(S0 ) is an embedded S 2 or RP2 in M representing a nontrivial element
of π2 M . If it is a 1 sided RP2 , then the boundary of an I bundle neighborhood of this
RP2 is an S 2 with the inclusion S 2 > M homotopic to the composition S0 →RP2 > M ,
hence nontrivial in π2 M by the choice of S0 . Thus the proof will be finished.

We now introduce the key technical ideas which will be used. Let N be a triangu-
lated 3 manifold, with triangulation T . Then a surface S ⊂ N is said to be in normal
form with respect to the triangulation if it intersects each 3 simplex ∆3 of T in a
finite collection of disks, each disk being bounded by a triangle or square of one of
the types shown in Figure 3.4.

Figure 3.4
There are four types of triangles in each ∆3 separating one vertex from the other
three, and three types of squares separating opposite pairs of edges of ∆3 . Squares
of different types necessarily intersect each other, so an embedded surface in normal
form can contain squares of only one type in each ∆3 . The edges of the squares and
triangles are assumed to be straight line segments, so ‘triangles’ are literally triangles,
54 Homotopy Properties §3.1

but ‘squares’ do not have to lie in a single plane or have all four sides of equal length.
‘Quadrilateral’ would thus be a more accurate term.
For a compact surface S which meets the 1 skeleton of the triangulation trans-
versely we define the weight w(S) to be the total number of points of intersection of
S with all the edges of the triangulation.

Lemma 3.13. If N contains an embedded sphere which is nontrivial in π2 N , then it

contains such a sphere in normal form, and we may take this sphere to have minimal
weight among all spheres which are nontrivial in π2 N .

Proof: As in the proof of existence of prime decompositions, we begin by perturbing

a nontrivial sphere S ⊂ N to be transverse to all simplices of T , and then we perform
surgeries to produce an S meeting each 3 simplex only in disks. A surgery on S
produces a pair of spheres S 0 and S 00 , and in π2 N , S is the sum of S 0 and S 00 , with
respect to suitably chosen orientations and paths to a basepoint. So at least one of
S 0 and S 00 must be nontrivial in π2 N if S is, and we replace S by this nontrivial S 0
or S 00 . Then we repeat the process until we obtain a nontrivial sphere S meeting all
3 simplices in disks. Note that replacing S by S 0 or S 00 does not yield a sphere of
larger weight since w(S) = w(S 0 ) + w(S 00 ) .
Next, we isotope S so that for each 3 simplex ∆3 , the boundary ∂D of each disk
component D of S ∩ ∆3 meets each edge of ∆3 at most once. For if such a circle ∂D
meets an edge ∆1 more than once, consider two points of ∂D ∩ ∆1 . After rechoosing
these two points if necessary, we may assume the segment α of ∆1 between these
points meets D only in the two points of ∂α . The arc α lies in one of the two disks
of ∂∆3 bounded by ∂D , and separates this disk into two subdisks. Let E be one of
these subdisks. We may perturb E so that it lies entirely in the interior of ∆3 except
for the arc α in ∂E , which we do not move, and so that the rest of ∂E , an arc β , stays
in D . Now we can perform an isotopy of S which removes the two points of ∂α = ∂β
from the intersection of S with the edge ∆1 by pushing β across the repositioned
E . It may be that before this isotopy, E intersects S in other curves besides β , but
this is not a real problem as we can just push these curves along ahead of the moving
β . The key property of this isotopy is that it decreases the weight w(S) by at least
two, since it eliminates the two intersection points of ∂α without introducing any new
intersections of S with edges of the triangulation.
Since this isotopy decreases w(S) , finitely many such isotopies, interspersed with
the former surgery steps to make S meet each 3 simplex in disks, suffice to produce a
sphere S meeting 3 simplices only in disks, with the boundaries of these disks meet-
ing each 1 simplex at most once. We can then easily isotopy S to meet all 2 simplices
in straight line segments, without changing the intersections with edges. We leave it as
an exercise for the reader to check that the condition that the boundary circles of the
disks of S ∩∆3 meet each 1 simplex at most once implies that each of these boundary
§3.1 The Loop and Sphere Theorems 55

circles consists of either three or four segments, forming a triangle or square as in

Figure 3.4.
Thus we produce a sphere S in normal form which is still nontrivial in π2 N .
Since the procedure for doing this did not increase the weight of S , it follows that the
minimum weight of all homotopically nontrivial spheres in N is achieved by a normal
form nontrivial sphere. u
t

Now we return to the proof of the Sphere Theorem. Choose a triangulation of

f . By the preceding lemma we may take S0
M and lift this to a triangulation T of M
to be in normal form with respect to T and of minimum weight among all nontrivial
spheres in Mf . The immersed sphere p(S0 ) ⊂ M is then composed of triangles and
squares in all the 3 simplices of M , and by small perturbations of the vertices of
these triangles and squares we can put their boundary edges in general position with
respect to each other. This means than any two edges of triangles or squares meet
transversely in at most one point lying in the interior of a 2 simplex, and no three
edges have a common point. We may also assume the interiors of the triangles and
f , this means that the
squares intersect each other transversely. Lifting back up to M
collection Σ of all the translates τ(S0 ) of S0 under deck transformations τ ∈ π1 M
consists of minimum weight nontrivial spheres in normal form, having transverse
intersections disjoint from the 1 skeleton of T .
Define compact subsets Bi ⊂ M f inductively, letting B0 be any 3 simplex of T
and then letting Bi be the union of Bi−1 with all 3 simplices which meet Bi−1 . So
S
B0 ⊂ B1 ⊂ ··· ⊂ f
i Bi = M . Let Σi be the collection of all the spheres in Σ which meet
Bi . This is a finite collection since only finitely many τ(S0 ) ’s meet each 3 simplex of
T , and Bi is a finite union of 3 simplices.
We describe now a procedure for modifying the spheres in Σi to make them
all disjoint. Suppose S ∩ S 0 ∩ T2 ≠ ∅ for some S, S 0 ∈ Σi , where T2 denotes the
2 skeleton of T . As a first step, circles of S ∩ S 0 in the interiors of 3 simplices of
T can be eliminated by isotopy of S without affecting S ∩ T2 . Namely, since S and
S 0 are in normal form, such a circle in a simplex ∆3 bounds disks D ⊂ S ∩ ∆3 and
D 0 ⊂ S 0 ∩ ∆3 . We may choose D to be innermost, so that D ∩ S 0 = ∂D . Then D ∪ D 0
is a sphere, which bounds a ball in ∆3 by Alexander’s theorem, and we can isotope S 0
by pushing D 0 across this ball to eliminate the circle ∂D = ∂D 0 from S ∩ S 0 .
Next we will appeal to the following crucial lemma:

Lemma 3.14. Let S and S 0 be minimum weight nontrivial spheres in normal form,
intersecting transversely and with S ∩ S 0 disjoint from 1 simplices. Assume that
S ∩ S 0 ≠ ∅ , and that in each 3 simplex, S ∩ S 0 consists only of arcs, no circles. Then
there is a circle C of S ∩ S 0 which is innermost in both S and S 0 , i.e., C bounds disks
D ⊂ S and D 0 ⊂ S 0 with D ∩ S 0 = ∂D and D 0 ∩ S = ∂D 0 . Furthermore we may choose
D and D 0 so that the sphere D ∪ D 0 is trivial in π2 M .
56 Homotopy Properties §3.1

Let us postpone the proof of this lemma, which is rather lengthy, and continue
with the proof of the Sphere Theorem. Having a circle C as in this lemma, we replace
D by D 0 in S to obtain a new sphere S , and similarly modify S 0 by replacing D 0
by D . Since D ∪ D 0 is trivial, the new spheres S and S 0 are homotopic to the old
ones, hence are nontrivial. They are also still of minimum weight since the total
weight w(S) + w(S 0 ) is unchanged, hence both terms are unchanged since S and S 0
each have minimal weight. With a small perturbation of S and S 0 we can eliminate
the circle C from S ∩ S 0 . The new spheres S and S 0 are not in normal form, but
can easily be made so: First straighten their arcs of intersection with 2 simplices
without changing the endpoints of these arcs; note that these arcs must join distinct
edges of each 2 simplex, otherwise S or S 0 would not be of minimum weight. Having
straightened the intersection of S and S 0 with 2 simplices, the intersections of S and
S 0 with 3 simplices must still be disks, otherwise the procedure for putting them into
normal form would decrease their weights, which is not possible. The new S and S 0
together have no more points of intersection in 2 simplices with other spheres in Σi
than the old S and S 0 did, since these other spheres are in normal form, so when
we straightened the intersections of the new S and S 0 with 2 simplices this will not
produce new intersections with the other spheres.
Thus by repeating this process for the other circles of S ∩ S 0 and then for all pairs
S , S 0 , in Σi we eventually obtain a collection Σi of spheres which are disjoint in all
2 simplices. They can easily be made to be disjoint in the interior of each 3 simplex
∆3 as well. For the new spheres of Σi meet ∂∆3 in a collection of disjoint circles, so
we need only choose disjoint disks in ∆3 bounded by these circles, then isotope the
disks of Σi ∩ ∆3 to these disjoint disks, rel Σi ∩ ∂∆3 .
Suppose we perform this disjunction procedure on the spheres in Σ0 , Σ1 , Σ2 , ···
in turn. We claim that no sphere is modified infinitely often. To see this, observe first
that each sphere of Σi is contained in Bi+w where w is the common weight of all the
spheres of Σ . This is because the cell structure on each sphere S in Σ consisting of
the triangles and squares given by normal form has w vertices, so a maximal tree in
its 1 skeleton has w − 1 edges and any two vertices can be joined by an edgepath
with w − 1 edges, so if S meets Bi , S must be contained in Bi+w . So a sphere in Σi
will not be modified at any stage after the spheres which meet Bi+w are modified, i.e.,
the spheres in Σi+w .
So the infinite sequence of modifications produces a collection Σ of disjoint non-
trivial spheres. Let S denote the union of these spheres. This is a surface in normal
form, perhaps noncompact. We claim that the intersection of S with the 2 skeleton
T2 is invariant under all the deck transformation. This is certainly true for the inter-
section with the 1 skeleton T1 , which was unchanged during the whole modification
procedure. Observe that for a surface T in normal form, T ∩ T1 determines T ∩ T2
uniquely, since for each simplex ∆2 , the weights a , b , c of T at the three edges of
§3.1 The Loop and Sphere Theorems 57

∆2 determine the numbers x , y , z of arcs of T ∩ ∆2 joining each pair of edges, as

indicated in Figure 3.8.

x+y = c x = 1
(
2 b + c - a)
1
c x b x+z = b y = 2(
a + c - b)
y z y+z = a z = 1 a
2(
+ b - c)

Figure 3.8
So for each deck transformation τ , the normal form surfaces S and τ(S) have the
same intersection with T1 hence also with T2 , as claimed.
Finally, to make S itself invariant under deck transformations we need only
choose the disks of S − T2 to be invariant. This can be done by fixing a choice of
these disks in a particular 3 simplex of T lying over each 3 simplex in M , then trans-
lating these disks to all other 3 simplices of T via the deck transformations. u
t

Proof of Lemma 3.14: Consider disks D ⊂ S and D 0 ⊂ S 0 bounded by circles of

S ∩ S 0 . Such disks may or may not be innermost, i.e., meeting S ∩ S 0 only in their
boundary. Here are some facts about such disks.

(1) The minimum weight of disks in S and S 0 is the same. For consider a disk of
minimum weight in both S and S 0 , say D ⊂ S . We may assume D is innermost. Let
D 0 and E 0 be the disks in S 0 bounded by ∂D . At least one of the spheres D ∪ D 0 ,
D ∪ E 0 is non-trivial, say D ∪ E 0 . Then w(D ∪ E 0 ) ≥ w(S 0 ) = w(D 0 ∪ E 0 ) since S 0
has minimal weight. Weights being additive, we get w(D) ≥ w(D 0 ) . This implies
w(D) = w(D 0 ) by the choice of D , so D 0 also achieves the minimum weight of disks
in S and S 0 .

(2) If D ⊂ S is an innermost minimum weight disk, separating S 0 into disks D 0 and

E 0 with D ∪ E 0 nontrivial, then D 0 is a minimum weight disk and D ∪ E 0 is a minimum
weight nontrivial sphere. This was shown in (1) except for the observation that D ∪ E 0
is minimum weight, which holds because w(D ∪ E 0 ) = w(D 0 ∪ E 0 ) since w(D) =
w(D 0 ) .

(3) Each disk D has w(D) > 0 . To verify this we may assume D is innermost. The
normal form sphere S is decomposed into triangles and squares by the 2 skeleton
T2 . If w(D) = 0 , D is disjoint from the corners of these triangles and squares, and
so meets them in regions of the following types:

Figure 3.5
58 Homotopy Properties §3.1

Construct a graph G ⊂ D by placing a vertex in the interior of each of these regions,

together with an edge joining this vertex to the midpoint of each of the arcs of inter-
section of the region with the sides of the triangle or square. Clearly D deformation
retracts onto G , so G is a tree, but this contradicts the fact that each vertex of G is
incident to at least two edges of G .

(4) By (3), a minimum weight disk D which is not innermost can contain only one
innermost disk D0 , and the annuli in D − D0 bounded by circles of S ∩ S 0 must have
weight zero. By the Euler characteristic argument in (3), these annuli must consist
entirely of the rectangular shaded regions in Figure 3.5.

(5) For a minimum weight disk D , there must exist in some 2 simplex ∆2 of T an
arc component α of D ∩ ∆2 having one end on ∂∆2 and the other in the interior of
∆2 . This is because D meets T1 (since w(D) > 0) , and if a component of D ∩ T2
containing a point of D ∩ T1 consisted entirely of arc components of D ∩ ∆2 with both
endpoints on ∂∆2 , for each ∆2 , then D would be a closed surface.

(6) An arc α as in (5) is divided into segments by the points of α ∩ S 0 . Each such
segment, except the one meeting ∂∆2 , must connect the two boundary components
of an annulus of D − S 0 , as in (4). The segment of α meeting ∂∆2 must lie in the
unique innermost disk D0 ⊂ D .

Having these various facts at our disposal, we now begin the proof proper.

Case I. Suppose we have a minimum weight disk D and an arc α of D ∩ ∆2 as in

(5) with the additional property that the endpoint of α in the interior of ∆2 is the
endpoint of an arc α0 of S 0 ∩ ∆2 whose other endpoint lies on the same edge of ∆2
as the other endpoint of α ; see Fig. 3.6a. The two arcs α and α0 cut off a triangle T
from ∆2 , and the following argument will involve replacing T by subtriangles until
the situation of Fig. 3.6c is achieved, with α ∩ S 0 = α0 ∩ S = α ∩ α0 .

α α0 α α0 α α0

(a) (b) (c)

Figure 3.6
After replacing D by its innermost subdisk we may assume by (6) that α∩S 0 = α∩α0 ,
as in Fig. 3.6b. Near α ∩ α0 , α0 lies in a disk D 0 ⊂ S 0 . If all of α0 is not contained
in D 0 , then the circle ∂D = ∂D 0 meets α0 at another point in the interior of α0 , and
we can rechoose α to have this point as an endpoint. So we may assume α0 ⊂ D 0 . If
D ∪ D 0 is nontrivial, relabel D 0 as E 0 in the notation of (1) and (2) above. But then the
minimum weight nontrivial sphere D ∪ E 0 can obviously be isotoped to decrease its
§3.1 The Loop and Sphere Theorems 59

weight, a contradiction. So D ∪ D 0 is trivial, and by (1) D 0 is also a minimum weight

disk. We can then repeat the preceding steps with the roles of D and D 0 reversed,
obtaining in the end the desired situation of Fig. 3.6c, with both D and D 0 innermost
and minimum weight, and D ∪ D 0 trivial. This finishes the proof in Case I.

Case II. Suppose we have a minimum weight disk D such that D ∩∆2 has a component
which is an arc β in the interior of ∆2 , for some 2 simplex ∆2 of T . The arc β is
divided into segments by the points of S 0 ∩ D . One of these segments must lie in
the innermost disk D0 in D , otherwise these segments would all be arcs crossing the
annuli of D − S 0 as we saw in (4), an impossibility since both ends of β are on ∂D .
Replace D by its innermost disk, with the new β a subsegment of the old one. The
two endpoints of β are joined to ∂∆2 by arcs α1 and α2 of S 0 ∩ ∆2 , which may be
chosen to have their outer endpoints on the same edge of ∆2 , as in Figure 3.7a.

β α0

α1 α2 α

(a) (b)

Figure 3.7
2
If there are other arcs of D ∩ ∆ meeting α1 or α2 , these arcs cannot cross α1 or α2
since D is innermost. Furthermore, these arcs must lie on the same side of α1 and α2
as β since D lies on one side of S 0 near ∂D . Assuming we are not in Case I, these arcs
then join α1 to α2 , and we can replace β by an edgemost such arc joining α1 to α2 .
So we may assume α1 and α2 meet D only at ∂β . Then α1 and α2 lie in a disk D 0 of
S 0 . If the sphere D ∪ D 0 is nontrivial, it must be minimum weight, by (2). But there is
an obvious isotopy of D ∪ D 0 decreasing its weight by two. So D ∪ D 0 is trivial and D 0
is a minimum weight disk, by (2) again. If D 0 is innermost, we are done. Otherwise,
replace D 0 by its innermost subdisk, α1 and α2 being replaced by subarcs which still
meet ∂∆2 by (6). If we are not in Case I, the endpoints of α1 and α2 in the interior
of ∆2 are then joined by a new arc β in some disk D of S with ∂D = ∂D 0 . By (2) the
sphere D ∪ D 0 cannot be nontrivial, since there is an obvious isotopy decreasing its
weight. So D ∪ D 0 is trivial and D is minimum weight. If D is not innermost we can
go back to the beginning of Case II and repeat the argument with a smaller rectangle
bounded by α1 ∪ β ∪ α2 in ∆2 . So we may assume D is innermost, finishing the proof
in Case II.

Case III. Considering again an arc α as in (5), there remains the possibility that the
component A0 of S 0 ∩ ∆2 containing the endpoint of α in the interior of ∆2 has both
its endpoints on edges of ∆2 not containing the other endpoint of α . Thus we have
the configuration shown in Figure 3.7b. The arc α lies in a minimum weight disk D
60 Homotopy Properties §3.1

which we may assume is innermost. In particular this means that no arcs of S 0 ∩ ∆2

below A0 join the left and right edges of ∆2 . By (2) we have ∂D = ∂D 0 for some
minimum weight disk D 0 in S 0 . Assuming Cases I and II are not applicable, there is
an arc α0 in D 0 as shown in the figure, with one endpoint at the upper endpoint of α
and the other endpoint in the left edge of ∆2 . If D 0 is not innermost we may replace
it by an innermost disk, choosing a new α to the left of the current one. Thus we
may assume both D and D 0 are innermost. If D ∪ D 0 is nontrivial then by (2) we can
rechoose D 0 on the other side of α , but this configuration would be covered by Case
I or II. So we may assume D ∪ D 0 is trivial, and Case III is finished. u
t

Now we can prove a converse to Proposition 1.6:

Theorem 3.15. f→M is a covering space with M irreducible and orientable,

If M
f is irreducible.
then M

Proof: This follows quite closely the scheme of proof of the Sphere Theorem. The
main thing to observe is that Lemmas 3.13 and 3.14 remain valid if ‘triviality’ for
embedded spheres is redefined to mean ‘bounding a ball.’ The two properties of trivial
spheres which were used were: (1) Surgery on a nontrivial sphere cannot produce two
trivial spheres, and (2) Triviality is invariant under isotopy.
Suppose M f is reducible, hence contains a nontrivial sphere S0 . The proof of the
Sphere Theorem gives a new nontrivial sphere S ⊂ M f which covers either a sphere
or an RP2 in M . In the latter case the RP2 is 1 sided since we assume M orientable,
and an I bundle neighborhood of this RP2 is RP3 minus a ball, so irreducibility of
f = S 3 , which is irreducible, a contradiction. In case the
M implies M = RP3 , hence M
sphere S ⊂ M f covers a sphere in M , the latter sphere bounds a ball in M , which lifts
f bounded by S , a contradiction again.
to a ball in M u
t

Recall the example S 1 × S 2 →S 1 × RP2 of a covering of an irreducible manifold

by a reducible manifold, which shows the necessity of the orientability hypothesis in
Theorem 3.15. To extend the theorem to the nonorientable case one needs to restrict
M to be P 2 irreducible, that is, irreducible and containing no 2 sided RP2 ’s.

Exercises
1. Show that a closed connected 3 manifold with π1 free is a connected sum of
S 1 × S 2 ’s, S 1 ×
e S 2 ’s, and possibly a homotopy sphere.
2. Extend Proposition 3.4 to the nonorientable case.
3. Show that every closed surface of even Euler characteristic is the boundary of some
compact 3 manifold.
4. Show that if M is simply-connected, then every circle in ∂M separates ∂M . In
particular, components of ∂M which are closed must be spheres.