Translation Invariant Measure on Euclidean Space is Multiple of Lebesgue Measure
Theorem
Let $\mu$ be a measure on $\R^n$ equipped with the Borel $\sigma$-algebra $\map \BB {\R^n}$.
Suppose that $\mu$ is translation invariant.
Also, suppose that $\kappa := \map \mu {\hointr 0 1^n} < +\infty$.
Then $\mu = \kappa \lambda^n$, where $\lambda^n$ is the $n$-dimensional Lebesgue measure.
Proof
From Characterization of Euclidean Borel Sigma-Algebra, we have:
- $\map \BB {\R^n} = \map \sigma {\JJ^n_{ho, \text {rat} } }$
where $\JJ^n_{ho, \text {rat} }$ denotes the collection of half-open $n$-rectangles with rational endpoints.
So let $J = \horectr {\mathbf a} {\mathbf b} \in \JJ^n_{ho, \text {rat} }$.
Let $M \in \N$ be a common denominator of the $a_i, b_i$ (which are rational by assumption).
We may then cover $J$ by finitely many pairwise disjoint half-open $n$-rectangles $\hointr 0 {\dfrac 1 M}^n$, in that:
- $\ds J = \bigcup_{i \mathop = 1}^{\map k J} \mathbf x_i + \hointr 0 {\dfrac 1 M}^n$
for some $\map k J \in \N$ and suitable $\mathbf x_i$, where:
- $\mathbf x_i + \hointr 0 {\dfrac 1 M}^n := \horectr {\mathbf x_i} {\mathbf x_i + \dfrac 1 M}$
Using that $\mu$ is translation invariant, this means:
- $\map \mu J = \map k J \, \map \mu {\hointr 0 {\dfrac 1 M}^n}$
Also, by Lebesgue Measure is Translation Invariant:
- $\map {\lambda^n} J = \map k J \, \map {\lambda^n} {\hointr 0 {\dfrac 1 M}^n}$
A moment's thought shows us that the half-open $n$-rectangle $I = \hointr 0 1^n$ may be covered by $M^n$ copies of $\hointr 0 {\dfrac 1 M}^n$, so that:
- $\map k I = M^n$
For brevity, write $I / M$ for $\hointr 0 {\dfrac 1 M}^n$.
Now, using that:
- $\ds \map {\lambda^n} {I / M} = \prod_{i \mathop = 1}^n \frac 1 M = \frac 1 {M^n}$
and $\map \mu I = \kappa$, compute:
\(\ds \map \mu J\) | \(=\) | \(\ds \map k J \map \mu {I / M}\) | Definition of $\map k J$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \frac {\map k J} {M^n} \paren {M^n \map \mu {I / M} }\) | $M^n / M^n = 1$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \frac {\map k J} {M^n} \map \mu I\) | $M^n = \map k I$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \frac {\kappa \, \map k J} {M^n}\) | Definition of $\kappa$ | |||||||||||
\(\ds \) | \(=\) | \(\ds \kappa \, \map k J \, \map {\lambda^n} {I / M}\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \kappa \, \map {\lambda^n} J\) | Definition of $\map k J$ |
Therefore, $\map \mu J = \kappa \, \map {\lambda^n} J$ for all $J \in \JJ^n_{ho, \text {rat} }$.
Let us quickly verify the other conditions for Uniqueness of Measures.
For $(1)$, we have Half-Open Rectangles Closed under Intersection.
For $(2)$, observe the exhausting sequence $\hointr {-k} k^n \mathop \uparrow \R^n$
Finally, for $(4)$, we recall that $\kappa < +\infty$.
Thus, by Uniqueness of Measures, $\mu = \kappa \lambda^n$.
$\blacksquare$
Sources
- 2005: René L. Schilling: Measures, Integrals and Martingales ... (previous) ... (next): $5.8 \ \text{(ii)}$