Mathematical Proceedings of the Cambridge

Philosophical Society
http://journals.cambridge.org/PSP

Additional services for Mathematical

Proceedings of the
Cambridge Philosophical Society:

Email alerts: Click here

Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here

Symmetric products of the circle

H. R. Morton

Mathematical Proceedings of the Cambridge Philosophical Society / Volume 63 / Issue 02 / April 1967, pp 349
- 352
DOI: 10.1017/S0305004100041256, Published online: 24 October 2008

Link to this article: http://journals.cambridge.org/abstract_S0305004100041256

How to cite this article:

H. R. Morton (1967). Symmetric products of the circle. Mathematical Proceedings of the
Cambridge Philosophical Society, 63, pp 349-352 doi:10.1017/S0305004100041256

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/PSP, IP address: 149.156.68.132 on 29 Apr 2014

Proc. Camb. Phil. Soc. (1967), 63, 349 349
POPS 63-42
Printed in Great Britain

Symmetric products of the circle

BY H. R. MORTON
University of Warwick, Coventry

(Received 3 June 1966)

The nth. symmetric product of a topological space, X, is defined to be the quotient

of the Cartesian product Xn by the action of the symmetric group which permutes the
factors. Even if X is a manifold, this product is, in general, not a manifold. The purpose
of this note is to determine these products when X is the circle, S1, and to show that
they are manifolds with boundary.
When X is the sphere, S2, the symmetric products of X are well known to be mani-
folds. The proof of this gives an interesting interpretation for the products of S1.
Regard S2 as the complex projective line. The Cartesian product (S2)n represents
ordered w-tuples of complex homogeneous parameters X1... Ara. The homogeneous
polynomial of degree n having these as roots determines, by its coefficients, a point in
%-dimensional complex projective space, Pn(C). This gives a continuous map
(S2)n - Pn(C). Permuting the roots \1...An does not alter the polynomial, so this map
factors through the nth symmetric product of S2. The map from this symmetric
product to Pn(C) is certainly 1-1, and since every such complex polynomial has n com-
plex roots, it is onto as well. The facts that (S2)n is compact, and Pn(C) is Hausdorff now
show that this symmetric product is homeomorphic to Pn(G).
A similar argument, regarding S1 as the real projective line, gives an embedding of
its nth symmetric product in Pn(R), w-dimensional real projective space, which is not
onto. This embedding gives a homeomorphism of this symmetric product with the
subspace of real homogeneous polynomials of degree n which have all their roots real.
Using this interpretation it is easy to determine the first couple of symmetric
products from projective geometry, but for the general case a different approach is
better.
The Cartesian product (Sx)n is covered by Rn. The group Z + Z + ...+Z = Zn of
covering translations, and the group Sn, acting on Rn by permuting the coordinates,
both lie in the group of afiine motions of Rn. Let G be the subgroup of this group
generated by Sn and Zn. Then Zn is normal in G with quotient Sn. Hence RnjG is the
quotient of Rn first by Zn and then by Sn, and so it is the required symmetric product
of AS1. The approach is now to break up this action of G in a different way.
Regard Rn as the product R11-1 x R1, where R1 is the diagonal, (x, x,..., x), x e R, and
-R"1 is the hyperplane 'Exi = 0. Then the action of G preserves this product structure.
Let Go be the subgroup of G which stabilizes the first factor. Then Go is normal in G
and G/GQ s Z. Hence Rn/G s [Rn-XIGO x R>)\Z. RnjG is then a bundle over S1 with
fibre Rn-lIGQ. This bundle is determined by the homeomorphism of Rn~1IG0 on itself
nduced by the operation of the generator of Z.
350 H. K. MORTON
Remark. Other products can be denned for any space X by factoring Xn with some
subgroup of Sn. For the circle, the preceding argument will still go through, replacing
G by the appropriate subgroup F, to give a bundle over S1 with fibre Rn-1jY0.
These products, however, are not all manifolds with boundary. Let Z3 <= Se act on
(iS1)6 by cycling the three factors S1 x S1. Then the corresponding product is not a
manifold. For consider a neighbourhood of a point on the diagonal S1. This will be
(R1)6, and the neighbourhood of this point in the quotient will be the similar Z3
quotient of R1.
Regarding R6 as C x C x G, the Zz action cycles the factors. Decompose as before
into the diagonal and the hyperplane z1 + zi + z3 = 0. Then the action takes place
entirely in the hyperplane. Here it is the linear extension of a fixed-point free Z3 action
on the unit sphere S3. Hence the quotient is the cone on a closed 3-manifold M, with
7T1(ikf) = Z3. Thus the quotient of R6 is the cone on the double suspension of M. This
double suspension E2M has H3(E3M) = H^M) = Z3 and so is not the sphere S5. Hence
the quotient of R6 is not RG again, and so a point on the diagonal S1 of (^S1)6 does not
have a Euclidean neighbourhood in the quotient.
Returning to the case of the symmetric product, we shall work out the quotient
Rn~xIO0, by finding a subspace of Rn~x which is homeomorphic to it. Such a subspace
exists if the conditions of the following lemma are satisfied.
LEMMA 1. Let the finite group F act continuously on a space X, and let F be a closed
subspace of X such that:
(i) no distinct points of F belong to the same orbit;
(ii) the translates of F cover X.
Then F is homeomorphic to XjY.
Proof. Consider the continuous map 6: F <= X-^* X/T. Then, by (i), 6 is 1-1, and by
(ii), it is onto. It remains to show that d is open.
Let U be any open set in F. Then V = (X - F) u U is open in X. By (i), U Ug <= V.
Hence U P g c f | Vg, which is also open in X, since F is finite,
r r
Now {r\Vg)nF<= U = {UUg)nF.
r r
So by (ii), f\Vg <= \J Ug. Hence U Ug is open in X, and p{ U Ug) = dU is open in XjV.
v r
We can relax the condition that the group F be finite to include a group such as Go
which is the extension of a group which acts without fixed points by a finite group, in
this case &,.J m
n
LEMMA 2. The region F, xx < x2 < ... < xn < x1 + 1, forms a suitable subspace of
n 1
R ~ , Y*xi = 0, for the action of Go.
Proof. Any point of Rn~x can be translated by Go to lie in F by a standard procedure
to be defined, which takes the whole orbit of a point to one point of F. This procedure
maps F to itself by the identity. The conditions of Lemma 1 will then be satisfied.
Any given point f' of R11'1 can be translated uniquely to a point | of the half-open
unit cube 1 < xt < 0 in Rn under the action of Zn. If 7/' = gr'g' is any other point in
the orbit of ,', then T} = g%, for some geG. Since Zn is normal in G, we can write
 Symmetric products of the circle 351
n n n
G = Sn. Z , so g = p. z. But Z maps the half-open cube in R outside itself, so z = e,
and g = p, some permutation.
Apply a permutation cr to to give = <r\ satisfying

This point will be uniquely determined, although the permutation a may not be,
if some of the ,t are equal. Since rj = p | , the point rj will also give rise to the point
under this treatment.
We now have with E ^ = r for some integer r, 0 ^ r < n. Add 1 to each of the
first r coordinates of , and then cycle them r times to return them to ascending order.
This gives the required point E,eF, with 2 ^ = 0, and 2 ^ 2 < ... < E,n 4, x + 1 .
Consider this process for a point ,' e F. Choose r so that exactly the last r coordinates
of ' are > 0. Subtract 1 from each of the coordinates to give f. Permuting these new
coordinates into ascending order is done by cycling them r times. Obviously 2 ^ = r,
so we must add 1 to the first r coordinates. But these are just the original last r
coordinates minus 1. We now have the original coordinates, so placing them in
ascending order returns us to the point E,'.
Thus
W-7GO ^ F.
Now a bounded subspace of Rn~1 defined by n linear inequalities, and containing
points at which all the inequalities are strict, is an (n l)-simplex. But F is bounded by
|x^| ^ 1 for all i, and contains interior points, so it is an (n l)-simplex.
So the symmetric product is an (n 1 )-disc bundle over the circle. It only remains to
determine whether it is orientable or non-orientable.
Regard Bn/G0 as F x R1, where this is given as x ^ ... ^ xn ^ xx + 1. The product
structure is determined by F x 0 ^ F x a, where the isomorphism is given by adding
a to each coordinate of the original F, which is taken as F x 0.
The action of Z on this product can be described by adding 1 to the first coordinate,
and then acting by Go to return each fibre to the standard form. This gives a map
F x 0 - F x 1/n. In this case, having added 1 to the first coordinate, it is only necessary
to cycle the coordinates once in order to return them to the standard form
xx *S ... < xn < x1+l

in F x I/n. So the map of F to itself induced by this action has its orientation deter-
mined by this cycling of the coordinates. The map is thus orientation reversing or
preserving as n is even or odd respectively.
This gives rise to the following description.

THEOREM. {Sx)nISn^ S1xDn-1 = ihe non-orientable bundle (n even), ^/S'xZ) 11 - 1

(n odd).
The symmetric product is thus homeomorphic to the tubular neighbourhood of a
projective line in Pn{R)-
So, returning to the representation of the symmetric product as a subset of Pn(R),
it would be nice to find an actual projective line lying in this subset.
352 H. R. MORTON
Consider the cross-section of the disc-bundle above consisting of the barycentre of
each simplex. These barycentres represent n points equally spaced around the circle,
and, in Pn(B), polynomials having these points as roots, where the point a corresponds
to the non-homogeneous root tan \OL.
The locus of such polynomials in Pn(R) turns out to be the projective line
XQ = x2 = #4 = . . . , x = x3 = x5 =

Thus the symmetric product can be thought of as a tubular neighbourhood of this

projective line in Pn(R).

I would like to thank R.L. E.Schwarzenberger for mentioning this problem, and
the Ministry of Education of Northern Ireland for theirfinancialsupport.