Integration - Find all functions f(x)

Join the discussion
Ask a follow-up here, or get your own question answered by working scientists, mathematicians and engineers — people, not an autocomplete.
Real named experts · corrections over time · the nuance an AI answer skips
11 replies · 3K views
Dethrone
Messages
716
Reaction score
0
Find all functions $f(x)$ so that $\left(\int \frac{dx}{f(x)}\right)\left(\int f(x) \,dx\right)=c$, constant.

The question says "no guessing". I looked at families of functions, starting with $f(x)=a$, $f(x)=x^n$, and $f(x)=\sin\left({x}\right)$, but they all fail. Any hints? (Wondering)
 
Physics news on Phys.org
Rido12 said:
Find all functions $f(x)$ so that $\left(\int \frac{dx}{f(x)}\right)\left(\int f(x) \,dx\right)=c$, constant.

The question says "no guessing". I looked at families of functions, starting with $f(x)=a$, $f(x)=x^n$, and $f(x)=\sin\left({x}\right)$, but they all fail. Any hints? (Wondering)

Hmm... how about $f(x)=e^{ax}$? (Wondering)
 
That works! :D I mainly posted my question to see if there was a better method of solving the problem instead of going through all elementary functions, but yeah. The question says: "Note: simply guessing an $f(x)$ that satisfies the equation is worth only 1 mark.", so I wasn't sure whether guessing families of $f(x)$ and trying them is considered "guessing". What do you think? (Wondering)
 
Rido12 said:
That works! :D I mainly posted my question to see if there was a better method of solving the problem instead of going through all elementary functions, but yeah. The question says: "Note: simply guessing an $f(x)$ that satisfies the equation is worth only 1 mark.", so I wasn't sure whether guessing families of $f(x)$ and trying them is considered "guessing". What do you think? (Wondering)

I'm not sure. It seems as if we should do something more. (Thinking)

We can at least generalize to $f(x)=ae^{bx}$, which also satisfies the equation with a condition on $b$ that is related to $c$. So at least we're not guessing a single $f(x)$.

The suggestion is there though, that we're supposed to come up with a smarter method to find all f(x), or otherwise that we should prove that we've found all of them.
 
Given $\displaystyle \int\frac{1}{f(x)}dx\cdot \int f(x)dx =c.........(1)$

Now Differentiate both side w. r to $x$, We Get

$\displaystyle \int \frac{1}{f(x)}dx\cdot f(x)+\frac{1}{f(x)}\cdot \int f(x)dx = 0.............(2)$

Now from equation $(1)$, We get $\displaystyle \int\frac{1}{f(x)}dx = \frac{c}{\int f(x)dx}$

and put into equation $(2)$, We get $\displaystyle \frac{c\cdot f(x)}{\int f(x)dx}+\frac{\int f(x)dx}{f(x)} = 0$

So $\displaystyle \left(\int f(x)dx\right)^2 = -c\cdot \left(f(x)\right)^2$

Now Let $-c=k^2\;,$ Then $\displaystyle \left(\int f(x)dx\right)^2 = k^2\cdot \left(f(x)\right)^2$

So $\displaystyle \int f(x)dx = \pm k\cdot f(x)$
 
Last edited:
Very neat! (Wondering) The answer does not give us a specific type of function, but rather, the relationship that said function must satisfy with its antiderivative. :D
I can only imagine how evil my professor will be...(Dull) for this question is a typical exam question...
 
Rido12 said:
Very neat! (Wondering) The answer does not give us a specific type of function, but rather, the relationship that said function must satisfy with its antiderivative. :D

Can you solve it?

Hint: take the derivative.
 
Can we not simply say...

$$f(x)=\frac{\int f(x) \,dx}{\pm k}$$

but:

$$f(x)=\pm kf'(x)$$

I hope this is it, because it seems like it could involve differential equations, which I haven't really studied yet. :(
 
Rido12 said:
Can we not simply say...

$$f(x)=\frac{\int f(x) \,dx}{\pm k}$$

but:

$$f(x)=\pm kf'(x)$$

I hope this is it, because it seems like it could involve differential equations, which I haven't really studied yet. :(

Ah.
Yes, sorry, it does involve a differential equation. (Doh)If you're interested, I'll show you.

If we write $y=f(x)$ and $\d y x = f'(x)$, we get:
\begin{array}{lcl}
y&=&\pm k\d y x \\
\frac {dy}{y} &=& \pm \frac{dx}{k} \\
\int \frac {dy}{y} &=& \int \pm \frac{dx}{k} \\
\ln |y| &= & \pm\frac x k + C_1 \\
|y| &=& e^{\pm x/ k + C_1} \\
y &=& \pm e^{C_1} e^{\pm x/ k} \\
y &=& C_2 e^{\pm x/ k} \\
y &=& C_2 e^{\pm x/ \sqrt{-c}} \\
\end{array}

And this is what we already found by guessing.
Substituing it in the original equation, shows that this entire family of solutions is indeed the solution. (Mmm)
 
Alternatively, without resorting to differential equations (well... sort of), we can do:

\begin{array}{lcll}
f(x) &=& \pm k f′(x) \\
\frac{f'(x)}{f(x)} &=& \pm \frac 1 k & \text{This is the derivative of the next line.}\\
\ln(f(x)) &=& \pm \frac x k + C_1 & \text{Actually, I'm leaving out a $\pm$ here, that should be included.}\\
f(x) &=& e^{\pm x/k + C_1}
\end{array}
 
Yep, but as you have said, I do smell the differential equation in that one, too. (Wasntme)

This course was meant for students to have already learned differential equations, so I'm sure this is the right way! :D (Yes)