Ma · Mathematics

Why the Discrete Metric on ℝ Is Not Induced by a Norm

Published By the 9to5Science editorial team

At a glance

  1. 1. Use norm homogeneity

    A norm satisfies ||tx|| = |t| ||x||, so its induced distance must scale when both points are multiplied by t.

  2. 2. Test two points

    The discrete distance from 0 to 1 and from 0 to 2 is 1, but norm homogeneity requires the second distance to be twice the first.

  3. 3. Conclude no norm works

    The scaling contradiction rules out every norm on ℝ; equivalently, every norm on ℝ is a positive multiple of absolute value.

The discrete metric on R\mathbb{R} is defined by assigning distance 00 to equal points and distance 11 to every pair of distinct points. The question is whether this distance can be written as d(x,y)=∥x−y∥d(x,y)=\|x-y\| for some norm on R\mathbb{R}.

It cannot. The quickest proof uses a defining property of every norm: scaling a vector by tt scales its norm by ∣t∣|t|. The discrete metric does not obey that scaling rule.

The scaling property of a norm-induced metric

If a metric is induced by a norm, then

d(x,y)=∥x−y∥.d(x,y)=\|x-y\|.

For any scalar tt and points x,yx,y, norm homogeneity gives

d(tx,ty)=∥tx−ty∥=∥t(x−y)∥=∣t∣ ∥x−y∥=∣t∣d(x,y).\begin{aligned} d(tx,ty) &= \|tx-ty\| \\ &= \|t(x-y)\| \\ &= |t|\,\|x-y\| \\ &= |t|d(x,y). \end{aligned}

Thus multiplying both points by tt must multiply their distance by ∣t∣|t|. This condition holds for every norm-induced metric, regardless of the vector space.

The discrete metric behaves differently. If x≠yx\ne y and t≠0t\ne 0, then tx≠tytx\ne ty, so

d(tx,ty)=1=d(x,y),d(tx,ty)=1=d(x,y),

not ∣t∣d(x,y)|t|d(x,y) in general. Choosing, for example, t=2t=2 contradicts the scaling property. Therefore the discrete metric on R\mathbb{R} cannot be induced by a norm.

A direct contradiction using 0, 1, and 2

Suppose a norm ∥⋅∥\|\cdot\| induced the discrete metric. Since 0≠10\ne 1 and 0≠20\ne 2, the metric would require

∥1∥=d(0,1)=1and∥2∥=d(0,2)=1.\|1\|=d(0,1)=1 \qquad\text{and}\qquad \|2\|=d(0,2)=1.

But 2=2⋅12=2\cdot 1, so homogeneity of the norm gives

∥2∥=∥2⋅1∥=2∥1∥=2,\|2\|=\|2\cdot 1\|=2\|1\|=2,

contradicting ∥2∥=1\|2\|=1. This proves the claim without needing to classify all possible norms on R\mathbb{R}.

What norms on ℝ can induce

There is also a complete description of norms on the real line. For any x∈Rx\in\mathbb{R}, write x=x⋅1x=x\cdot 1. Homogeneity implies

∥x∥=∥x⋅1∥=∣x∣ ∥1∥.\|x\|=\|x\cdot 1\|=|x|\,\|1\|.

Since a norm is positive on nonzero vectors, c=∥1∥c=\|1\| is positive. Hence every norm on R\mathbb{R} has the form

∥x∥=c∣x∣,c>0.\|x\|=c|x|, \qquad c>0.

Its induced metric is therefore d(x,y)=c∣x−y∣d(x,y)=c|x-y|, a constant multiple of the usual distance—not the discrete distance. For example, d(0,1)=cd(0,1)=c and d(0,2)=2cd(0,2)=2c, whereas the discrete metric assigns both distances the value 11.

A topological way to see the difference

The discrete metric makes every singleton open: the open ball of radius 1/21/2 around xx contains only xx. In contrast, for any norm on R\mathbb{R}, an open ball around xx contains other real numbers. Indeed, if the norm is c∣⋅∣c|\cdot|, then every sufficiently small nonzero change hh satisfies ∥h∥=c∣h∣<r\|h\|=c|h|<r for a ball of radius r>0r>0.

So the discrete metric gives R\mathbb{R} the discrete topology, while a norm on R\mathbb{R} gives it the usual topology. This is a useful supporting observation, though the scaling contradiction is the most direct proof.

Common mistakes

  • Checking only translation invariance: Both the discrete metric and norm-induced metrics are translation invariant, so that property alone does not distinguish them.
  • Confusing a metric with a norm: A metric needs to satisfy the metric axioms, but a norm also has homogeneity and compatibility with vector addition. Not every metric on a vector space comes from a norm.
  • Considering only distances equal to 1: The discrete metric assigns 1 to every distinct pair, including pairs whose difference is a scalar multiple of another difference. That clashes with norm homogeneity.
  • Assuming a special choice of norm could fix the issue: On R\mathbb{R}, all norms are positive multiples of absolute value, so changing the norm only changes the scale of the usual metric.

References

  1. Introductory Functional Analysis with Applications — Erwin Kreyszig

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.