\[\Large\lVert{Ax}\rVert\leq\lambda\lVert{x}\rVert, \forall x \in H.\] The norm of a bounded operator of \(A\) is defined as the infimum of all scalars \(\lambda\) that satisfy the preceding inequality (i.e., nonnegativity, homogeneity, triangle inequality, etc.) Prove that the concept of norm above is really a norm on the space of operators on \(H\). Prove that \(\lVert{\cdot}\rVert\) is a norm, i.e. it satisfies the norm axioms.
Proof:
Assume that \(A\) is a bounded operator on a Hilbert space \(H\). As a bounded operator, the norm of \(A\) would be defined as follows: \[\Large\lVert{A}\rVert = \inf \{\lambda : \lambda \in \mathbb{R} \ and \ \lVert{Ax}\rVert \leq \lambda\lVert{x}\rVert, \forall x \in H\}\]
It follows that \[\Large A(0) = A(0*x) = 0 * Ax = 0\]
and, as a norm in a Hilbert space,
\[\Large \lVert{A}\rVert \geq 0, \lVert{A}\rVert = 0 \iff Ax = 0, \forall x \in H.\]
Then if, \[\Large\lVert{A}\rVert = 0, \ and \ \lVert{x}\rVert > 0 \implies \exists x > 0 \in H\]
such that \[\Large Ax = 0 \ and \ A = 0\]
\(\Large \blacksquare\)
Proof:
First, let \(\Large x \in H.\) Then recall that by definition \[\Large \lVert{AB}\rVert = \sup_{\lVert{x}\rVert=1} \lVert{ABx}\rVert.\]
Then, by the homogeneity property of norms, \[\Large \lVert{ABx}\rVert \leq \lVert{A}\rVert\lVert{Bx}\rVert\] and, \[\Large \lVert{A}\rVert\lVert{Bx}\rVert \leq \lVert{A}\rVert\lVert{B}\rVert\lVert{x}\rVert.\] Since \(\Large \lVert{x}\rVert = 1\),
\[\Large \lVert{ABx}\rVert\ \leq \lVert{A}\rVert\lVert{B}\rVert \implies \sup_{\lVert{x}\rVert = 1}\lVert{ABx}\rVert \leq \lVert{A}\rVert\lVert{B}\rVert.\] And because \(\Large \lVert{x}\rVert = 1\) we can say, \[\Large \lVert{AB}\rVert \leq \lVert{A}\rVert\lVert{B}\rVert.\]
\(\Large \blacksquare\)
Proof:
Assume, for the sake of contradiction, that there is a sequence \(\Large (x_{n})_{n \in \mathbb{N}}\) in the unit sphere of \(H\).
Then we have, \[\Large\lVert{x_{n}}\rVert = 1, \forall n \ and \ \lVert{Ax_{n}}\rVert \to \infty \ as \ n \to \infty.\]
If \(\Large\lVert{Ax_{n}}\rVert \to \infty\,\) then \(\Large\exists \ M > 0\) such that \(\Large\lVert{Ax_{n}}\rVert > M, \forall n.\)
A linear operator must satisfy the following property: \(\Large A(cx) = cAx, \forall x \in H \ and \ \forall c \in \mathbb{C}\)
Applying this property, we can assign \[\Large y_{n} = \frac{1}{\lVert{Ax_{n}}\rVert}x_{n}.\]
Then, if we apply A, we get: \[\Large Ay_{n} = A(\frac{1}{\lVert{Ax_{n}}\rVert}x_{n}) = \frac{1}{\lVert{Ax_{n}}\rVert}Ax_{n}\]
Since \(\Large \lVert{Ax_{n}}\rVert \to \infty \ as \ n \to \infty\), we know that \(\Large \frac{1}{\lVert{Ax_{n}}\rVert} \to 0 \ as \ n \to \infty.\)
Then, \[\Large \lVert{Ay_{n}}\rVert = \frac{\lVert{Ax_{n}}\rVert}{\lVert{Ax_{n}}\rVert^{2}} = \frac{1}{\lVert{Ax_{n}}\rVert} \to 0 \ as \ n \to \infty.\]
Given that \(\Large \lVert{Ay_{n}}\rVert \to 0\), the sequence \((Ay_{n})\) is bounded in \(H\). This contradicts the original assertion that \(\Large \lVert{Ax_{n}}\rVert \to \infty.\)
\(\Large \blacksquare\)