Thread View: sci.math.research
2 messages
2 total messages
Started by Bjorn Poonen
Thu, 30 Jul 1992 14:45
paths and multisets
Author: Bjorn Poonen
Date: Thu, 30 Jul 1992 14:45
Date: Thu, 30 Jul 1992 14:45
38 lines
2634 bytes
2634 bytes
QUESTION: Suppose you have an n-element multiset S deforming into another n-element multiset T. Can you find a path from a given element of S to some element of T? More precisely, let M be a topological space and give M^n the product topology. Let the symmetric group S_n act on M^n by permuting the coordinates, and give M^n/S_n the quotient topology. (M^n/S_n is the space of n-element multisets.) Suppose w : [0,1] -> M^n/S_n is a continous path, and P is an element of the multiset w(0). Can one find a path w_1 : [0,1] -> M such that w_1(0)=P and w_1(t) is an element of w(t) for all t in [0,1]? Assume that M has nice properties, if it helps. I need the case M=R^2 (i.e., the plane). (The application is the following: take all polynomials 1 + e_1 z + e_2 z^2 + ... + e_n z^n where each e_i is 0 or 1, and n is arbitrary. Let W be the set of all complex zeros of such polynomials. I have proved that the closure of W is connected. If the question above has a positive answer, I can prove that it is path connected as well.) More generally one can ask whether the path w : [0,1] -> M^n/S_n can always be decomposed into n paths w_1,w_2,...,w_n in M. This is equivalent to asking whether the path can be lifted to a path in M^n. (This would certainly imply a positive answer to the first question.) Here is what I know so far. If M=R this general result is true. (Let w_k(t) be the k^th largest element of w(t).) If M=R^2 and n=2, the result is true. (The map M^n -> M^n/S_n is a covering projection if one deletes the places where two elements of the multiset coincide, so one can lift the path on each (open) connected component of the set of t for which w(t) has distinct elements. For n=2, this suffices, since for the remaining values of t, there is only one possible lift of w(t).) Please e-mail solutions, useful references, or comments to me at poonen@research.att.com before August 24 poonen@math.berkeley.edu after August 24 Bjorn Poonen
Re: paths and multisets
Author: David desJardins
Date: Mon, 03 Aug 1992 05:00
Date: Mon, 03 Aug 1992 05:00
65 lines
3973 bytes
3973 bytes
Bjorn Poonen <poonen@digel.Berkeley.EDU> writes:
>QUESTION: Suppose you have an n-element multiset S deforming into
>another n-element multiset T. Can you find a path from a given element
>of S to some element of T?
I think the following proves the stronger version of Bjorn's conjecture,
that a path in M^n/S_n can be lifted to M^n.
Theorem. Let I be the unit interval [0,1], and let M be an arbitrary
metric space. Let S_n act on M^n by permuting the coordinates, and let
p: M^n -> M^n/S_n be the projection map. Let f be a continuous map f: I
-> M^n/S_n. Then there exists g: I -> M^n such that f = p o g.
Proof of Theorem 1. The proof goes by induction on n. Clearly the
result is trivial for n = 1, because the projection map p is a bijection
and there is a unique possible lift. So, we assume the result for every
n' < n, and proceed to prove it for n.
Without loss of generality, we can assume the special case that for
every 0 < x < 1, f(x) is a multiset which contains at least two distinct
values. Suppose that the theorem is true in this special case. Then,
in the general case, the points x where f(x) has n repeated values
partition the remaining points in [0,1] into disjoint subintervals. On
each such subinterval, we can apply the theorem in the special case to
obtain a lift. At the points where there are n repeated values, there
is a unique lift. Combining all of these lifts gives a lift on all of
[0,1].
Let 0 < x < 1. Consider the multiset f(x). In the special case, it
contains k distinct values, z_1,...,z_k, where k \ge 2. Let the minimum
distance in M between any two z_i be d. By continuity, there is a open
interval N containing x such that, for x' in N, f(x') is within distance
d/2 of f(x). This implies that we can partition all of the elements of
M which occur in multisets f(x'), for x' \in N, into points which are
within distance d/2 of z_1, points which are within distance d/2 of z_2,
and so on.
For each k, let n_k be the cardinality of z_k in f(x). Construct
f_k: N -> M^{n_k}/S_{n_k}, where f_k(x') consists of the n_k elements of
f(x') which are within distance d/2 of z_k. Now, n_k < n, so by the
induction hypothesis on Theorem 1, there is a lift g_k: N -> M^{n_k}.
The product of all the g_k is a lift of f on N. Now, let x = 1/2.
Consider the union of all open intervals containing 1/2 on which a lift
of f exists. This is the open interval (a,b), with a < 1/2 < b. Choose
a lift of f on (a,b). By continuity, this extends to a lift of f on
[a,b].
Suppose b < 1. Then, apply the previous argument again with x = b,
to construct a lift of f on an open interval containing b. By
restriction, this gives a lift of f on the half-open interval [b,c) for
some c > b. And, we can act on this lift by an element of S_n to make
it agree with the previous lift at b. Combining the lifts gives a lift
on (a,c), which contradicts the hypothesis that (a,b) is the maximal
open interval containing 1/2 on which a lift exists.
Thus b must be equal to 1. The same argument shows that a must be
equal to 0. Thus we have constructed a lift of f on [0,1] in the
special case, and the previous argument shows that this implies the
general case of the theorem. QED.
David desJardins
Thread Navigation
This is a paginated view of messages in the thread with full content displayed inline.
Messages are displayed in chronological order, with the original post highlighted in green.
Use pagination controls to navigate through all messages in large threads.
Back to All Threads