5

If $d_1, d_2$ are metrics of $X$, is it true that $d_1 +d_2 $, $d_1 - d_2$, $d_1\cdot d_2$, $\sqrt d_1$ are metrics on $X$?

Here is my attempt:

If we take $d_1 = d_2 $ = standard metric on the real line, then $d_1\cdot d_2 = d_1^2$ is not a metric.

$d_1 - d_2$ may not be metric because it may not even be always non-negative.

But I am not sure about others. I need help.

Thanks for giving me time.

MJD
  • 65,394
  • 39
  • 298
  • 580
Srijan
  • 12,518
  • 10
  • 73
  • 115

1 Answers1

5

To prove that $d_1+d_2$ is a metric, just check that it has each of the properties of a metric. The only one that takes any work is the triangle inequality, and it’s not hard, using the fact that $d_1$ and $d_2$ both satisfy the triangle inequality. That leaves only $\sqrt{d_1}$, and it’s clear that the only question is whether it satisfies the triangle inequality.

In other words, must it always be true that $$\sqrt{d_1(x,z)}\le\sqrt{d_1(x,y)}+\sqrt{d_1(y,z)}\;?\tag{1}$$

Since both sides of $(1)$ are non-negative, $(1)$ holds iff the inequality that you get by squaring both sides of $(1)$ holds; does it?

Brian M. Scott
  • 616,228
  • 1
    Sir, based on your Idea I have done like this: $d_1(x,z)\leq d_1(x,y)+d_1(y,z)$ and $d_2(x,z)\leq d_2(x,y)+d_2(y,z)$ . Adding both will give required result. Thinking about second one. – Srijan Jun 04 '12 at 10:11
  • 1
    I understand now. Thank you very much.:) – Srijan Jun 04 '12 at 10:13