Supremum Norm on Vector Space of Real Matrices is Norm
Theorem
Supremum Norm forms a norm on the vector space of real matrices.
Proof
Let $M \in \R^{m \times n}: m, n \in \N_{>0}$ be a real matrix.
Denote the $\paren {i, j}$-th entry of $M$ by $a_{i j}$.
Note that the set of matrix elements of $M$ is a finite set of real numbers.
We have that:
Therefore, $M$ has the greatest element.
Norm Axiom $\text N 1$: Positive Definiteness
\(\ds \norm M_\infty\) | \(=\) | \(\ds \max_{\begin {split}
1 \mathop \le i \mathop \le m \\ 1 \mathop \le j \mathop \le n \end {split} } \size {a_{i j} }\) |
Greatest Element is Supremum, Definition of Max Operation | |||||||||||
\(\ds \) | \(\ge\) | \(\ds 0\) |
Equality is obtained for $M$ being a zero matrix.
Suppose $\norm M_\infty = 0$.
Then:
- $\ds \forall i, j: 1 \le i \le m, 1 \le j \le n: \size {a_{i j} } = 0$
In other words, $M$ is a zero matrix.
$\Box$
Norm Axiom $\text N 2$: Positive Homogeneity
\(\ds \norm {\alpha \cdot M}_\infty\) | \(=\) | \(\ds \max_{\begin {split}
& 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size {\alpha m_{i j} }\) |
Greatest Element is Supremum | |||||||||||
\(\ds \) | \(=\) | \(\ds \max_{\begin {split}
& 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size \alpha \size {a_{i j} }\) |
Absolute Value Function is Completely Multiplicative | |||||||||||
\(\ds \) | \(=\) | \(\ds \size \alpha \max_{\begin {split}
& 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size {a_{i j} }\) |
||||||||||||
\(\ds \) | \(=\) | \(\ds \size \alpha \norm M_\infty\) | Greatest Element is Supremum |
$\Box$
Norm Axiom $\text N 3$: Triangle Inequality
Let $P, Q \in \R^{m \times n}$.
Denote their $\paren {i, j}$-th matrix elements as $p_{i j}$ and $q_{i j}$ respectively.
Fix $i, j \in \N: 1 \le i \le m, 1 \le j \le n$.
We have that:
\(\ds \size {p_{i j} + q_{i j} }\) | \(\le\) | \(\ds \size {p_{i j} } + \size {q_{i j} }\) | Triangle Inequality for Real Numbers | |||||||||||
\(\ds \) | \(\le\) | \(\ds \max_{\begin {split}
& 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size {p_{i j} } + \max_{\begin {split} & 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size {q_{i j} }\) |
Definition of Max Operation | |||||||||||
\(\ds \) | \(=\) | \(\ds \norm P_\infty + \norm Q_\infty\) | Greatest Element is Supremum, Definition of Supremum Norm |
This holds for any $i, j$.
Hence:
\(\ds \norm {P + Q}_\infty\) | \(=\) | \(\ds \max_{\begin {split}
& 1 \mathop \le i \mathop \le m\\ & 1 \mathop \le j \mathop \le n \end {split} } \size {p_{i j} + q_{i j} }\) |
Greatest Element is Supremum | |||||||||||
\(\ds \) | \(\le\) | \(\ds \norm P_\infty + \norm Q_\infty\) |
$\Box$
All norm axioms are seen to be satisfied.
Hence the result.
$\blacksquare$
Sources
- 2017: Amol Sasane: A Friendly Approach to Functional Analysis ... (previous) ... (next): $\S 1.2$: Normed and Banach spaces. Normed spaces