Documente Academic
Documente Profesional
Documente Cultură
Dan Crytser
August 6, 2012
Problem 20.1a
To show that the Manhattan metric d(x, y) = |x1 y1 | + . . . + |xn yn | induces the
standard topology on Rn , we show that it induces the same topology as the square metric.
Let (x, y) denote the square distance between x and y in Rn , i.e. (x, y) = max |xi yi |.
First we prove that d(x, y) = |x1 y1 | + . . . + |xn yn | defines a metric. As it is a sum
of non-negative numbers for any x, y, d(x, y) is always non-negative. If d(x, y) = 0, then
0 |xk yk | d(x, y) = 0 for any k. Thus xk = yk for all k, and x = y. The symmetry
follows from |xk yk | = |yk xk |. Let x, y, z Rn . The triangle inequality follows from
|xk yk | |xk zk | + |zk yk |, and then taking the sum over all k.
Suppose that x, y Rn . Claim: (x, y) d(x, y). For (x, y) = |xi yi | for some index
i, so we have (x, y) = |xi yi | |x1 y1 | + . . . + |xn yn | = d(x, y). The inequality comes
from adding non-negative numbers to |xi yi |.
Claim: d(x, y) n(x, y). Let k beP
an index. Then |xk yk | |xi yi | = (x, y). Then
summing over k = 1, . . . , n, we obtain nk=1 |xk yk | |xi yi | + |xi yi | + . . . + |xi yi | =
n|xi yi |.
Now we can show that the topologies generated by the square metric and the Manhattan
metric are the same. Let > 0 and let x Rn . We show that Bd (x, ) is open in the square
metric. Claim: B (x, /n) Bd (x, ). For if (x, y) < /n, then d(x, y) n(x, y) < .
Now we show that the topology induced by the Manhattan metric contains the topology
induced by the square metric. Let x Rn and > 0. We show that B (x, ) is open in the
Manhattan metric. Claim: Bd (x, ) B (x, ). For if d(x, y) < , then (x, y) d(x, y) < .
Problem 20.3 Let X be a metric space with metric d.
(a) We show that the metric map d : X X R is continuous. Let x x0 X X. Set
C = d(x, x0 ), and let (C , C + ) be a basis element of the standard topology which
contains C. Let U = B(x, 2 ) and let V = B(x0 , 2 ). Then U V is a nbhd of x x0 in
X X.
Claim: d(U V ) (C , C + ). For if u U and v V , then
d(u, v) d(u, x) + d(x, v) d(u, x) + d(x, y) + d(y, v) <
+ C + = C + .
2
2
So d(u, v) < C + . But also d(x, y) d(x, u) + d(u, v) + d(v, y) < d(u, v) + . So we also
have C < d(u, v). Thus u v U V implies that d(u, v) (C , C + ). Thus d
is continuous.
(b) Suppose that d : X 0 X 0 R is continous. We show that the topology from X 0 contains
the metric topology. Let x X, let B(x, ) be an open ball for some > 0. We prove
that there is an open set U in X 0 such that x U B(x, ). Note that d(x, x) (, ),
so that x x is in the set d1 (, ). As d is continous when the domain is X 0 X 0 ,
this preimage must be a nbhd of x x. Thus we can take a basis element in the product
topology U V such that x x U V d1 (, ). Then U is an open subset
of X 0 which contains x. We prove that U B(x, ). Suppose that y U . Then
y x U V , hence d(y, x) (, ). But this means y B(x, ). So x U B(x, ),
and the topology on X 0 is at least as fine as the metric topology.
Problem 20.9 Show that the Euclidean metric on Rn is a metirc.
(a)
x (y + z) =
n
X
xi (yi + zi ) =
i=1
n
X
(xi yi + xi zi ) = x y + x z.
i=1
(b) If either x or y is 0, then |x y| = 0, so the claim holds. Suppose that both are nonzero.
y
x
+ ||y||
, which must be greater than or equal to 0.
We compute the square norm of ||x||
||
x
y 2
xx
xy
yy
+
|| =
+2
+
2
||x|| ||y||
||x||
||x||||y|| ||y||2
xy
)0
= 2(1 +
||x||||y||
x
||x||
y
||y||
xy
||x||||y||
and we obtain:
x
y 2
xx
xy
yy
) =
2
+
2
||x|| ||y||
||x||
||x||||y|| ||y||2
xy
= 2(1
)
||x||||y||
0
xy
xy
Now we divide by 2 and add ||x||||y||
to both sides of the inequality to obtain 1 ||x||||y||
.
Now multiply through by the product of the norms to obtain x y ||x||||y||. As |x y|
is either x y or x y, we have that |x y| ||x||||y||.
1 + d(x, z) + d(z, y)
d(z, y)
d(x, z)
+
=
1 + d(x, z) + d(z, y) 1 + d(x, z) + d(z, y)
d(x, z)
d(z, y)
+
1 + d(x, z) 1 + d(z, y)
= d0 (x, z) + d0 (z, y)
d0 (x, y) =
The first inequality comes from the triangle inequality for d and the increase of the
x
function 1+x
on [0, ).
Now we check that d0 and d induce the same topology on X. Let x X and let
> 0. We show that Bd (x, ) is open in the topology induced by d0 . Let = 1+
. Claim:
Bd0 (x, ) = Bd (x, ).
Note about increasing functions: Let f : R R be increasing. If f (x) < f (y), then
x < y. For we cannot have x = y, and if x > y we would have f (x) > f (y).
Suppose that y Bd0 (x, ). Then d0 (x, y) < = 1+
. As d0 increases as a function of d,
d(x,y)
we must have that d(x, y) < . Similarly, if d(x, y) < , then d0 (x, y) = 1+d(x,y)
< 1+
= .
Thus the open balls are equal.
Before we prove the second part, we note that algebra can be used to solve d(x, y) =
d0 (x,y)
= 1d01(x,y) 1. This gives d as an increasing function of d0 .
1d0 (x,y)
Now we show that the ball Bd0 (x, ) is open in the topology generated by d. Let =
1
1. Claim: Bd (x, ) = Bd0 (x, ). Suppose y Bd (x, ). Then d(x, y) < and so
1
d(x,y)
d0 (x,y)
0
d (x, y) = 1+d(x,y)
< . Suppose instead that d0 (x, y) < . Then d(x, y) = 1d
0 (x,y) < 1 =
(here we are assuming that < 1, but this is fine as long as balls of radius < 1/2 are open,
we are done). Thus y Bd (x, ). So the two open balls are equal.
Thus every open ball in each topology is an open ball in the other topology. The bases
are equal, so the topologies are the same.
Problem 21.2 Suppose that f : X Y is an isometric embedding (with an e).
First, note that if f (x) = f (x0 ), then dX (x, x0 ) = dY (f (x), f (x0 )) = 0, so that x = x0 .
Thus f is injective. Let x X and let BY (f (x), ) be a basis element which contains f (x).
Claim: f (B(x, )) B(f (x), ). For if x0 B(x, ), then dY (f (x), f (x0 )) = dX (x, x0 ) < .
Thus f (B(x, )) B(f (x), ). So f is continuous. If we take a basis element for f (X) in the
subspace topology, it looks like B = BY (f (x), ) f (X). The image of this under the map
f 1 is just BX (x, ), which is an open set in X.
21.3a
Let Xk be a metric space for k = 1, . . . , n with dk the respective metric. Show that
(x, y) = maxk dk (x, y) is a metric which induces the product topology on X1 . . . Xn .
Let X = X1 . . . Xn . Suppose that x, x0 X. Then (x, x0 ) = maxk dk (xk , x0k ) 0,
as each of the dk (xk , x0k ) 0. If x = x0 then xk = x0k for all k, and hence (x, x0 ) = 0. If
(x, x0 ) = 0, then xk = x0k for all k and hence x = x0 . The symmetry follows from the fact that
dk (xk , x0k ) = dk (x0k , xk ) for all k. Let x, y, z X. For each k, dk (x, y) dk (x, z) + dk (z, y).
Note that dk (x, z) (x, z) and dk (z, y) (z, y) for all k. Thus we can replace the dk (x, z)
and dk (z, y) on the right with (x, z) and (z, y) to obtain dk (x, y) (x, z) + (z, y). But
taking the maximum over all k on the left we obtain (x, y) (x, z) + (z, y). Thus is a
metric.
We show that the topology induced by is the product topology on X = X1 . . . Xn .
First, we show that the open balls relative to are open sets in the product topology. Let
x X. Then
B (x, ) = {y X : (x, y) < } = Bd1 (x1 , ) . . . Bdn (xn , );
for if (x, y) < , we must have dk (xk , yk ) < for all k, and conversely.
4