Walter Rudin Real and Complex Analysis Chapter 2 2.14 Riesz representation theorem the last step.

Why did he put the absolute value of $a$ ?

Is not it sufficient to assume $f$ is positive?

Proof. Clearly, it is enough to prove this for real $f$. Also, it is enough to prove the inequality \begin{equation} \tag{16} \Lambda f \leq \int_X f \,\mathrm{d}\mu \end{equation} for every real $f \in C_c(X)$. For once $(16)$ is established, the linearity of $\Lambda$ shows that $$ -\Lambda f = \Lambda(-f) \leq \int_X (-f) \,\mathrm{d}\mu = - \int_X f \,\mathrm{d}\mu, $$ which, together with $(16)$ shows that equality holds in $(16)$.

Let $K$ be the support of the real $f \in C_c(X)$, let $[a,b]$ be an interval which contains the range of $f$ (note the Corollary to Theorem 2.10), choose $\epsilon > 0$, and choose $y_i$, for $i = 0, 1, \dotsc, n$, so that $y_i - y_{i-1} < \epsilon$ and \begin{equation} \tag{17} y_0 < a < y_1 < \dotsb < y_n = b. \end{equation} Put \begin{equation} \tag{18} E_i = \{ x : y_{i-1} < f(x) \leq y_i \} \cap K \qquad (i = 1, \dotsc, n) \end{equation} Since $f$ is continuous, $f$ is Borel measurable, and the sets $E_i$ are therefore disjoint Borel sets whose union is $K$. There are open sets $V_i \supset E_i$ such that \begin{equation} \tag{19} \mu(V_i) < \mu(E_i) + \frac{\epsilon}{n} \qquad (i = 1, \dotsc, n) \end{equation} and such that $f(x) < y_i + \epsilon$ for all $x \in V_i$. By Theorem 2.13, there are functions $h_i \prec V_i$ such that $\sum h_i = 1$ on $K$. Hence $f = \sum h_i f$, and Step II shows that $$ \mu(K) \leq \Lambda\left( \sum h_i \right) = \sum \Lambda h_i. $$ Since $h_i f \leq (y_i + \epsilon) h_i$, and since $y_i - \epsilon < f(x)$ on $E_i$, we have \begin{align*} \Lambda f &= \sum_{i=1}^n \Lambda(h_i f) \leq \sum_{i=1}^n (y_i + \epsilon) \Lambda h_i \\ &= \sum_{i=1}^n (|a| + y_i + \epsilon) \Lambda h_i - |a| \sum_{i=1}^n \Lambda h_i \\ &\leq \sum_{i=1}^n (|a| + y_i + \epsilon)[ \mu(E_i) + \epsilon/n ] - |a| \mu(K) \\ &= \sum_{i=1}^n (y _i - \epsilon) \mu(E_i) + 2 \epsilon \mu(K) + \frac{\epsilon}{n} \sum_{i=1}^n (|a| + y_i + \epsilon) \\ &\leq \int_X f \,\mathrm{d}\mu + \epsilon[ 2\mu(K) + |a| + b + \epsilon ]. \end{align*}

(Original scanned image here.)

  • $\begingroup$ Since every real function $f$ can be written $f=f^+ - f^-$ both of them are positive $\endgroup$ Aug 24, 2018 at 22:50
  • $\begingroup$ To avoid putting the absolute value $a$ $\endgroup$ Aug 24, 2018 at 22:54
  • $\begingroup$ Why can't I apply what I said to this proof ? $\endgroup$ Aug 25, 2018 at 2:28

1 Answer 1


If you use $a$ instead of $|a|$ you cannot go from the second to the third line, since not knowing that sign of $a$ precludes you from knowing if you keep the inequality $\sum_i\Lambda_i\geq\mu(K)$.

And, if you do the proof just for $f\geq0$, you only get the inequality $\Lambda f\leq\int_Xf\,d\mu$, and not equality.

  • $\begingroup$ Thanks Merci Danke $\endgroup$ Aug 25, 2018 at 8:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.