Prove limit comparison test for Integrals

CGandC
Messages
326
Reaction score
34
Homework Statement
Theorem: Given the continuous functions ## f,g : [ 1, +\infty ) \to \mathbb{R} ## that satisfy ##\lim _{x \rightarrow+\infty} \frac{f(x)}{g(x)}=L>0##. Prove the improper integrals ## \int_1^{+\infty} f(x)dx , \int_1^{+\infty} g(x)dx ## converge or diverge together.
Relevant Equations
* ( Integral comparison ) Let ## f,g : [ a,\omega) \to \mathbb{R} ## ( where ## \omega \in \mathbb{R} ## or ## \omega = \infty ## ). Suppose for all ## b < \omega ## it occurs that ## f,g \in \mathbb{R([a,b])} ##. Suppose ## 0 \leq f \leq g ##. If ## \int_a^\omega g < \infty ## then also ## \int_a^\omega f < \infty ##, and if ## \int_a^\omega f = \infty ## then ## \int_a^\omega g = \infty ## .

* ## f \in R([ a,b]) ## means ## f ## is Riemann-Integrable on ## [ a,b] ##
Attempt:
Note we must have that
## f>0 ## and ## g>0 ## from some place
or
## f<0 ## and ## g<0 ## from some place
or
## g ,f ## have the same sign in ## [ 1, +\infty) ##.

Otherwise, we'd have that there are infinitely many ##x's ## where ##g,f ## differ and sign so we can chose a sequence ## 1<x_n \to \infty ## s.t. ## \lim _{n \rightarrow+\infty} \frac{f(x_n)} {g(x_n)}<0## , a contradiction to the given ##\lim _{x \rightarrow+\infty} \frac{f(x)}{g(x)}=L>0 ##.

Suppose ## f>0 ## and ## g>0 ## from some place ( proof for ## f<0 ## and ## g < 0 ## from someplace or proof for the case where ## g,f ## have same sign in ## [1, +\infty ) ## are very similar
).
From the given ## \lim _{x \rightarrow+\infty} \frac{f(x)}{g(x)}=L>0 ## we can say that ## \forall \epsilon>0 . \exists R>0 . \forall x>R . | \frac{f(x)}{g(x)} - L | < \epsilon ##.
Note that ## -\epsilon < \frac{f(x)}{g(x)} - L < \epsilon \longrightarrow
L - \epsilon < \frac{f(x)}{g(x)} < L + \epsilon \longrightarrow (L-\epsilon)g(x) < f(x) ##.

Hence, for ## \epsilon = \frac{L}{2} ## there exists ## R>0 ## s.t. for all ## x>R ## we have that ## (L-\epsilon)g(x) = \frac{L}{2} g(x) < f(x) ##.
Suppose without loss of generality that ## R>1 ##.
Suppose that ## \int_1^{+\infty} f(x)dx ## converges and we'll show that ## \int_1^{+\infty} g(x)dx ## converges ( proving that if ##\int_1^{+\infty} g(x)dx## converges then ## \int_1^{+\infty} f(x)dx ## converges is similar if we look at ## f(x) < ( L + \epsilon) g(x) ## ) .
Under the new assumption, note that ## \int_R^{+\infty} f(x)dx > \frac{L}{2} \int_R^{+\infty} g(x) dx ##, hence from comparison test for integrals ## \int_R^{+\infty} g(x) dx ## exists.
Now, note that ## \int_1^{+\infty} g(x)dx = \int_1^{R} g(x)dx + \int_R^{+\infty} g(x)dx ##, Hence ## \int_R^{+\infty} g(x)dx = \int_1^{+\infty} g(x)dx - \int_1^{R} g(x)dx ## and since ## \int_R^{+\infty} g(x) dx ## exists then these two integrals also exist, hence ## \int_1^{+\infty} g(x)dx ## exists.

( After proving that if ##\int_1^{+\infty} g(x)dx## converges then ## \int_1^{+\infty} f(x)dx ## converges, we proved that ##\int_1^{+\infty} g(x)dx ~ \text{ converges} \iff \int_1^{+\infty} f(x)dx ~ \text{converges} ##. Hence we also proved ##\int_1^{+\infty} g(x)dx ~\text{ diverges} \iff \int_1^{+\infty} f(x)dx ~ \text{diverges} ## )
## \square ##There are couple of things that I have questions about:
1. Where have I used the fact that ##f,g ## are continuous? It seems like I proved the theorem without using this given.
( I know if a function is continuous on a bounded interval the it is Riemann Integrable on that interval, but It seems to me that under the assumption that ## \int_1^{+\infty} f(x)dx ## converges, the assumption that ##f,g ## are continuous is necessary in-order to say that ##g ## is Riemann Integrable on ## [1,\infty) ## , hence I could perform the integrable ##\int_1^{+\infty} g(x)dx ## in the inequality ## \int_R^{+\infty} f(x)dx > \frac{L}{2} \int_R^{+\infty} g(x) dx ## )
2. Was I right that the cases " ## f,g> 0 ## from some place or ##f,g< 0 ## from some place or ## g,f ## have the same sign in ## [1,+\infty) ## " are the all the possible cases?
3. I think that the proofs for the other cases ( besides " ## f,g> 0 ## from some place " ) are very similar for all the cases I have written, is that right?
4. what do you think about the proof? Is it specious? if so, then why? if not, then what could be done better?
 
Physics news on Phys.org
The proof can be shortened considerably.

We know, because f(x)/g(x) \to L &gt; 0, that there exists R \geq 1 such that if x &gt; R then <br /> \left| \frac{f(x)}{g(x)} - L \right| &lt; \frac L2 so that <br /> \frac L2 &lt; \frac{f(x)}{g(x)} &lt; \frac{3L}2. Continuity of f and g rules out the possibility that they can change sign without passing through zero, and if f/g is to be positive in (R, \infty) then zeros of f and g in (R, \infty) must coincide. But then f/g would be undefined at such a zero, which contradicts the requirement that f(x)/g(x) \in (\frac{L}2,\frac{3L}2). So f and g do not change sign in (R, \infty). It follows that \int_1^\infty f(x)\,dx converges if and only if \int_R^\infty |f(x)|\,dx converges since
  • for every S &gt; R we have \int_R^S |f(x)|\,dx = \pm \int_R^S f(x)\,dx, and
  • \int_1^\infty f(x)\,dx = \int_1^R f(x)\,dx + \int_R^\infty f(x)\,dx converges if and only if \int_R^\infty f(x)\,dx converges
and the same hold for g.

Now we have that <br /> \frac L2 |g(x)| &lt; |f(x)| &lt; \frac {3L}2|g(x)| and integrating from R to S &gt; R yields <br /> \frac L2\int_R^S |g(x)|\,dx &lt; \int_R^S |f(x)|\,dx &lt; \frac{3L}{2}\int_R^S g(x)\,dx. Taking the limit S \to \infty and applying the integral comparison test to the above shows that
  • if \int_R^\infty |g(x)|\,dx converges, then by the second inequality so does \int_R^\infty |f(x)|\,dx, and conversely
  • if \int_R^\infty |f(x)|\,dx converges, then by the first inequality so does \int_R^\infty |g(x)|\,dx
as required.
 
Thank you so much for the time and effort you put into helping me, now I understand what I've wanted. Your proof is much more elegant and understandable.
 
There are two things I don't understand about this problem. First, when finding the nth root of a number, there should in theory be n solutions. However, the formula produces n+1 roots. Here is how. The first root is simply ##\left(r\right)^{\left(\frac{1}{n}\right)}##. Then you multiply this first root by n additional expressions given by the formula, as you go through k=0,1,...n-1. So you end up with n+1 roots, which cannot be correct. Let me illustrate what I mean. For this...

Similar threads

Back
Top