Professional Documents
Culture Documents
Geo HW2
Geo HW2
Robbie Lyman
September 22, 2014
~ ) : I} is an
(1) Let 0 = {(U , ) : I} be an atlas for M. Then we know that = {(1 (U ),
atlas for T M, which is smooth because M is. We'd like to show that is an orientation for T M.
~
~ 1
~ 1 (U )1 (U )).
So, given , I, we want to showSthat det( d(
)x ) > 0 for all x (
1
1
1
1
~ ( (U ) (U )) = (UV)Rn ,
Also, note that (U ) (U ) = pUV Tp M so that
n
n
~
~ 1
and we have
: (U V) R (U V) R .
By denition,
1
1
~
~ 1
(x, u) = ( (x), U ,
1
= ( 1
(x), U ,
1
, (x)
1
, (x)
U , ,1
(u))
(x)
1
([U , , u])) = ( 1
(x), d( )x (u)),
since (U , , u) (U , , u) if d( 1
)x (u) = u. So let x1 , . . . , xn be the rst n coordinates
of R2n , u1 , . . . , un be the last n, = (f1 , . . . , fn ), and d( 1
) ({) = (g1 , . . . , gn ). Then
we have
{
( 1
( 1
fi
fi
)i
)i
(x, u) =
(x),
(x, u) =
(x) = 0
xj
xj
uj
uj
" n
#
n
X
2 ( 1
( d( 1
gi
X ( 1
)i
)i
)x )i
(x, u) =
(u) =
(x)uk =
(x)uk ()
xj
xj
xj
xk
xj xk
k=1
k=1
" n
#
( 1
)
gi
X ( 1
i
)i
(x, u) =
(x)uk =
(x).
uj
uj
xk
xj
k=1
So, in block form, if A = aij is the matrix with entries aij given by (), we have
~
~ 1
d(
)(x u) =
d( 1
)x
A
0
.
d( 1
)x
P
Now, the determinant of this matrix, which we'll call B = (bij ) is S2n sgn() 2n
i=1 bi(i) . Note
that if any of these products contains an element in the upper-right-hand corner of the matrix, the
product is zero. This tells us that our nonzero choices are restricted to the upper-left-hand corner for
the rst n terms, and thus the lower-right-hand corner for the last n|i.e.
1
2
~
~ 1
det( d(
)(x u) ) = det( d( )x ) > 0
,
1
(Zero is not a possibility, since the composition of 1
with is the identity, and the
chain rule tells us their Jacobians are inverses and thus their determinants must be nonzero.) This
shows that is an orientation for T M.
(2) (a) As a polynomial, p is a smooth map between the smooth manifolds Rk and R. If a is a regular
value of p, then p1 (a) = Xa will be a k 1-dimensional submanifold of Rk .
1
Now,
dp x =
...
p
x1
p
xn
n
n
X
X
p(tx) txk
p(tx)
p(tx)
mtm1 p(x) = mtm1 p(x) =
=
xk
t
txk
t
txk
k=1
= mp(x) =
n
X
xk
k=1
k=1
p
(x)
xk
p
So suppose p(x) 6= 0, then dpx (x) = xi x
(x) = m p(x) 6= 0. Let R, and t = /mp(x).
i
Then dpx (tx) = t mp(x) = , so that if a 6= 0, a is a regular value of p, as required.
(b) Let a 6= 0 and t = a1/m . Then the mapping from Rk to itself given by x 7 tx is a dieomorphism.
Let f be the restriction of this mapping to Xa . f will be an embedding as soon as we can show
that it is an immersion. But dfx = t Ik , which is injective, so f(Xa ) is a submanifold of Rk . For
x Xa , p(x) = a = p f(x) = p(tx) = tm p(x) = tm a = a/|a| = 1. This tells us that Xa is
dieomorphic to X1 if a > 0, and to X1 if a < 0.
d(det)A =
P
det
x11 (A)
...
det
xnn (A)
where xij (A) = xij Sn sgn() nk=1 ak(k) = Sn (i)=j sgn() nk=1 k6=i ak(k) = Cij ,
where Cij is the i, jthPcofactor of A, i.e. the determinant of the i, jth minor of A times (1)i+j .
2
Now, since det A = ni=1 a1i C1i , we have a vector a in Rn , a = (a11 , . . . , a1n , 0, . . . , 0) for
which d(det)A (a) 6= 0, so by linearity, det is a submersion at A if det(A) 6= 0. In particular,
then, 1 is a regular value of det, as required.
(b) We know that, since 1 is a regular value of det, that TIn SL(n, R) = Ker(d(det)In ). But since
the i, jth minor of In is always just In1 , xij (In ) = (1)i+j , so d(det)In (x) = 0 =
P
i+j
xij = 0. It is clear that this is a linear subspace of Mn (R), since the sum of the
i j (1)
elements of any two matrices whose elements sum to zero (after multiplying by a constant determined by the position of the element) is still zero, as ismultiplying the matrix by a constant.
Further, choosing the rst n2 1 elements, we have one well-dened choice for the last one to
make the sum zero|i.e. this linear subspace has dimension n2 1, as desired.
P
TIn SL(n, R) = {(aij ) Mn (R) : i j (1)i+j aij = 0}.
(c) As we saw in part (a), xij = Cij , the i, jth cofactor of A, and generalising the argument above,
P
we have x TA SLn (R) if i j Cij xij = 0. Therefore,
det
det
det
T SL(n, R) =
X
i,j
Cij bij = 0 ,
~j
~ 1
~j
~ 1
~ 1
With this denition, the transition function
is just
i
i (y, t) = j (i (y), tx) =
1
(j i (y), t), where x is the representative discussed above. The plus or minus comes as follows:
1
i (y) Ui Uj = both xi and xj are not zero, but it could be the case that one is positive while
the other is negative. In such a case, then, our choice of t will dier in sign between the two functions.
Since this sign depends only on i, j and `, and Ui and Uj exclude the crossover region (where xi or
xj could be zero), this last coordinate function is continuous, and therefore smooth, showing that E is
a smooth manifold (of dimension Rn+1 ). So to show that : E M is a vector bundle, we need an
open cover with accompanying local trivialisations that satisfy a transition function condition.
Choose our cover to be {Ui : 1 6 i 6 n + 1}, the domains from the atlas for Pn (R) given in
class, and dene hi : 1 (Ui ) Ui R as hi (`, v) = (`, t), where t is chosen as above. (Given
~ i above, the nal coordinate function ends up being just the identity, so hi is certainly
the chart
a dieomorphism.) For the transition function condition, suppose Ui Uj 6= . Then, similarly as
above, hj h1
i (`, t) = (`, t). The choice of positive or negative, since it depends only on i, j and `
can be thought of as a function from Ui Uj to GL1 R (= R ) is smooth because it is constant on the
connected components of Ui Uj .
,