Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

star shaped open sets

16 views
Skip to first unread message

Italo

unread,
Jul 4, 2009, 9:13:27 AM7/4/09
to
Hi to all!
I have a question, i'm trying to prove that every
star-shaped open subset of R^n is diffeomorphic to
R^n.
I found this sketch of proof:

Let U be an open and star-shaped region of R^n. Wlog let 0 be a
star-center of U. There is a smooth function f:R^n -> R such
that f>0 on U and f=0 outside U.
Such a function certanly exists for open balls, and any open set
of R^n is a locally finite union of open balls, so one just has
to sum up all the funcion of the open balls in such a covering.
Define g:U -> R by g(x) = integral_0^1 dt/f(tx).
Define h:U -> R^n by h(x) = g(x)x.
One can prove the following:
1 is one-to-one
2. h is onto
3. h is smooth
4. the Jacobian of h is invertible
Then h is a smooth diffeomorphism by the inverse function theorem

how can i prove 1,2,4?

If anyone knows another way to prove that every open star convex subset
of R^n is diffeomorphic to R^n tell me please.

Thank you in advance

W^3

unread,
Jul 5, 2009, 1:35:43 AM7/5/09
to
In article
<569058.72611.12467132...@nitrogen.mathforum.org>,
Italo <shin...@tin.it> wrote:

> Hi to all!
> I have a question, i'm trying to prove that every
> star-shaped open subset of R^n is diffeomorphic to
> R^n.
> I found this sketch of proof:
>
> Let U be an open and star-shaped region of R^n. Wlog let 0 be a
> star-center of U. There is a smooth function f:R^n -> R such
> that f>0 on U and f=0 outside U.
> Such a function certanly exists for open balls, and any open set
> of R^n is a locally finite union of open balls, so one just has
> to sum up all the funcion of the open balls in such a covering.
> Define g:U -> R by g(x) = integral_0^1 dt/f(tx).
> Define h:U -> R^n by h(x) = g(x)x.
> One can prove the following:
> 1 is one-to-one
> 2. h is onto
> 3. h is smooth
> 4. the Jacobian of h is invertible
> Then h is a smooth diffeomorphism by the inverse function theorem
>
> how can i prove 1,2,4?

For each unit vector v, let R(v) = sup {rv : r >= 0, rv in U}. Because
of the rapid vanishing of f at the boundary of U, for all v we have
g(rv) -> oo as r -> R(v) from below. Thus h maps the U-ray [0, R(v))*v
onto the full ray [0, oo)*v. This shows h is onto. To see h is 1-1,
note |h(rv)| = int_0^r dt/f(tv), which implies |h(rv)| is strictly
increasing as a function of r.

For the Jacobian, let D_j denote the jth partial derivative and let
e_j be the usual basis vector. Then D_jh(x) = g(x)*e_j + D_jg(x)*x.
Thus the linear transformation Dh(x) equals g(x)*I + T_x, where I is
the indentity and T_x is a linear transformation whose range is
contained in the span of the vector x. Clearly Dh(0) equals g(0)*I,
which is invertible. If x is nonzero, then the formula for |h(rv)|
shows Dh(x) applied to x itself is a positive multiple of x. If w is a
vector perpendicular to x, then Dh(x)(w) = g(x)*w + scalar multiple of
x. It follows that the range of Dh(x) contains x and its orthogonal
complement, which implies Dh(x) is invertible.

Italo

unread,
Jul 5, 2009, 8:16:37 AM7/5/09
to
Many many thanks!

Bacle

unread,
Jul 7, 2009, 2:06:13 PM7/7/09
to

I have a different idea that I think may work:
We can show that every set of this type is diffeo-
morphic to an open metric ball in R^n , i.e.,
a ball B={ x in R^n, ||x||<r } , for r>0.
You can ( in the case the set is bounded )
contain the star set S in an open n-ball centered
at the origin, and then draw a line from the
origin passing thru a point in S and ending up
at a point in the n-ball. This gives you a smooth
bijection between the n-ball and S .
And then it is relatively easy to show that
an open metric ball in R^n is diffeomorphic to
R^n. Use, e.g., some variant of the arctan
function used for R^1.

Think this should help.
join

W^3

unread,
Jul 7, 2009, 6:07:18 PM7/7/09
to
In article
<17069003.86560.1246990...@nitrogen.mathforum.org>
,
Bacle <ba...@yahoo.com> wrote:

That looks doubtful to me. Can you supply some details?

0 new messages