If $f$ and $g$ are nonnegative Lebesgue measurable functions, then we know that $\int (f+g) d\lambda = \int f d \lambda + \int g d \lambda $. Given the difinition of integral of an arbitrary Lebesgue measurable function, that is $ \int f d\lambda = \int f^+ d\lambda - \int f^- d\lambda$, how do you prove $\int (f_1 - f_2) d\lambda = \int f_1 d\lambda - \int f_2 d\lambda $ for nonnegative Lebesgue integrable functions $f_1$ and $f_2$ ?

That $f_1 - f_2$ is Lebesgue integrable is easily seen. Now, I can show that $\int (-f) d\lambda = -\int f d\lambda$, as $\int (-f) d\lambda = \int (-f)^+ d\lambda - \int (-f)^- d\lambda = \int f^- d\lambda - \int f^+ d\lambda = -\int fd\lambda$. Any ideas how to proceed?


So I came up with a way to do it:

The idea is to split the domain of integration into part on which $f_1 - f_2$ is positive and part on which it is negative, and use the identity

$$ \int |f_1 - f_2| = \int (f_1 -f_2)^+ + \int (f_1-f_2)^- $$

More precisely we let $E = \{x: f_1(x) \ge f_2(x) \}$, so $\int_E |f_1 - f_2| = \int_E (f_1 - f_2) $. Now since $f_1 - f_2$ is nonnegative we can apply the argument of drhap to deduce that $\int_E (f_1 - f_2) = \int_E f_1 - \int_E f_2$. Similarly, $\int_{E^c} |f_1 - f_2| = \int_{E^c} -(f_1-f_2) = \int_{E^c} f_2 -\int_{E^c}f_1$. From what I already proved, the negative sign can be taken out of the second integral. Now if we combine these we'll get the result.

  • $\begingroup$ Looks okay to me. $\endgroup$ – drhab Apr 28 '14 at 17:32

Let $h=f+g$ and notice that $$h^+-h^-=f^+-f^-+g^+-g^-.$$ This means $$h^++f^-+g^- = h^-+f^++g^+,$$ and since both sides have constant sign, $$\int h^++ \int f^-+ \int g^- = \int h^-+ \int f^++ \int g^+.$$ Now collect terms in a suitable way.

This proof is taken from W. Rudin's Real and complex analysis.

  • $\begingroup$ That didn't work. The last line give me $\int h =\int f + \int g$, not what I want to prove. $\endgroup$ – ywx Apr 28 '14 at 15:02
  • $\begingroup$ @ywx On base of $\int\left(f+g\right)=\int f+\int g$ you can finish the job by noting that $\int\left(f_{1}-f_{2}\right)+\int f_{2}=\int\left[\left(f_{1}-f_{2}\right)+f_{2}\right]=\int f_{1}$ and consequently $\int\left(f_{1}-f_{2}\right)=\int f_{1}-\int f_{2}$ $\endgroup$ – drhab Apr 28 '14 at 15:36
  • $\begingroup$ As I am following the book, the property $\int (f + g) = \int f +\int g$ hasn't yet been generalized to arbitrary functions, I can only assume it's true for nonnegative functions. We can try splitting the domain of integration to positive and negative part of $f_1 - f_2$ $\endgroup$ – ywx Apr 28 '14 at 16:13
  • $\begingroup$ I think I've got an answer however SE won't let me answer (<10 rep) That's frustrating $\endgroup$ – ywx Apr 28 '14 at 16:50
  • $\begingroup$ Now give your answer. (>10 rep) $\endgroup$ – drhab Apr 28 '14 at 17:00

Use the usual trick... split up the cases where $f_1-f_2$ is positive and negative and use your original result, then combine terms properly.

Namely, consider $(f_1-f_2)^+$ and $(f_1-f_2)^-$ now what exactly is

$$\int (f_1-f_2)^+ - \int(f_1-f_2)^-$$


Now use linearity etc.

  • $\begingroup$ "Now use linearity etc." I can't, this is what I want to prove. $\endgroup$ – ywx Apr 28 '14 at 15:04

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.