Help in understanding the proof of a claim made in the proof of the Beppi Levi Theorem in Schilling.

71 Views Asked by At

In the proof of the Beppo Levi theorem in Schilling the following claim is made: $$f\leq u\Rightarrow I_\mu(f)\leq\sup_{j\in\mathbb{N}}\int u_j d\mu.$$

The proof of this claim is not hard but there is one point that I have a hard time seeing of why it is done, which is why is the parameter $\alpha$ being introduced.

Let $(X,\mathcal{A},\mu)$ be a measure space, $f=\sum\limits_{k=0}^My_k\mathbf{1}_{A_k}$ a standard representation of a non-negative measurable simple function, $u_i:X\rightarrow [0,\infty]$ an increasing sequence of measurable functions and $u=\sup\limits_{i\in\mathbb{N}}u_i$. Define,

$I_\mu(f):=\sum\limits_{k=0}^My_k\mu(A_k)$ and for any non-negative measurable function $w$,

$\int wd\mu:=\sup\{I_\mu(g):g\leq w$ & $g\geq 0$ measurable simple function$\}.$

The proof of the claim from Schilling is as follows. For a simple non-negative $f$ such that $f\leq u$, we can see that $$(\forall\alpha\in(0,1))(\forall x\in X)(\exists N(\alpha,x))(\forall j\in \mathbb{N})(j\geq N(\alpha,x)\Rightarrow \alpha f(x)\leq u_j(x))\space\space \space -(\dagger)$$

Thus the (measurable) sets $B_j:=\{\alpha f \leq u_j\}\uparrow X$ as $j\uparrow\infty$. Also we clearly have, $$\alpha f\mathbf{1}_{B_j}f\leq \mathbf{1}_{B_j}u_j\leq u_j.$$ So then, $$\sum_{k=0}^M\alpha y_k\mu(A_k\cap B_j)=I_\mu\bigg(\sum_{k=0}^M\alpha y_k\mathbf{1}_{A_k}\mathbf{1}_{B_j}\bigg)=I_\mu(\alpha\mathbf{1}_{B_j}f)\leq\int u_jd\mu\leq\sup_{j\in\mathbb{N}}\int u_jd\mu.$$

Then by $\sigma$-additivity of $\mu$, $\mu(A_k\cap B_j)\uparrow \mu(A_k)$ as $B_j\uparrow X$, $j\uparrow \infty$.

Thus $$\alpha I_\mu(f)=\alpha\sum\limits_{k=0}^My_k\mu(A_k)\leq\sup_{j\in\mathbb{N}}\int u_jd\mu$$ and letting $\alpha\rightarrow 1$, the claim follows.

Now my problem is with $(\dagger)$ and why the $\alpha$ is needed? Now as $\alpha\in(0,1)$ and $f\geq 0$, we obviously have $\alpha f\leq f$. So if $f\leq u$ we then have $\alpha f\leq sup_{i\in\mathbb{N}}u_i$, and consequently for a $x\in X$ there is a $N(\alpha,x)\in\mathbb{N}$, such that $\alpha f(x)\leq u_{N(\alpha,c)}(x)$ and as the $u_i$'s are increasing we have that $\alpha f(x)\leq u_j(x)$, for all $j\geq N(\alpha,x)$.

But from this reasoning and the one above it will seem to work if the $\alpha$ was never introduced, so I would appreciate any help to explain to me why I am wrong in thinking the $\alpha$ is not needed.

Thanks in advance.