关于Rudin《实分析与复分析》5.22节抽象泊松积分的泛函有界性疑问

Hey JohnNash, let's walk through this clearly— I remember wrestling with this exact Rudin section too, so I get the confusion!

First, let's recap what we need to prove for a bounded linear functional:

• Linearity: The mapping $\Lambda_x(f) = f(x)$ has to satisfy $\Lambda_x(af + bg) = a\Lambda_x(f) + b\Lambda_x(g)$ for scalars $a,b$ and $f,g \in M$. That's straightforward here, since evaluating a function at a point preserves linear combinations.
• Boundedness: We need a constant $C \geq 0$ such that $|\Lambda_x(f)| \leq C|f|$ for all $f \in M$. The norm of the functional is the smallest such $C$, i.e., $|\Lambda_x| = \sup_{|f| \leq 1} |\Lambda_x(f)|$.

Now, inequality (3) from the section is almost certainly the estimate $|f(x)| \leq |f|$ for every $f \in M$. Let's tie that directly to our functional:

• This inequality tells us exactly that $|\Lambda_x(f)| = |f(x)| \leq 1 \cdot |f|$. So right away, we know $\Lambda_x$ is bounded, and its norm is at most 1 ($|\Lambda_x| \leq 1$).

Next, we need to show the norm is exactly 1, not just less than or equal to it. To do this, we just need to find a function in $M$ that hits this upper bound:

• In the abstract Poisson integral setup, $M$ almost always includes constant functions. Take the constant function $f \equiv 1$: its norm $|f| = 1$ (assuming the norm on $M$ is consistent with this— which it is in Rudin's context), and $\Lambda_x(f) = f(x) = 1$. That means $\sup_{|f| \leq 1} |\Lambda_x(f)| \geq 1$.

Putting those two pieces together: $|\Lambda_x| \leq 1$ and $|\Lambda_x| \geq 1$, so $|\Lambda_x| = 1$.

To sum it up: Inequality (3) gives the boundedness directly, and finding a function that achieves the bound proves the norm is exactly 1.

备注：内容来源于stack exchange，提问作者JohnNash