8

Find all $x$ that satisfy the equation $$\tan^{-1}\left(\frac{2x+1}{x+1}\right)+\tan^{-1}\left(\frac{2x-1}{x-1}\right)=2\tan^{-1}\left(1+x\right)$$

Attempt:

Taking tangent on both sides, and using the identities $\displaystyle \tan\left(\alpha+\beta\right)=\frac{\tan\alpha+\tan\beta}{1-\tan\alpha\tan\beta}$ and $\displaystyle \tan\left(2\alpha\right)=\frac{2\tan\alpha}{1-\tan^{2}\alpha}$,

We obtain $$\frac{\frac{\left(2x+1\right)}{x+1}+\frac{\left(2x-1\right)}{x-1}}{1-\frac{\left(4x^{2}-1\right)}{x^{2}-1}}=\frac{2\left(x+1\right)}{1-\left(x+1\right)^{2}}$$ and after removing the root $x=0$, we are left with a simple equation $(x^2-2)(2x+1)=0$. So, the possible roots are $x=0,\pm \sqrt{2}, -\dfrac{1}{2}$ , now my question is, how do I verify which roots satisfy the original equation, because one cannot see directly after plugging, whether it's true or not because it involves arctangents.

To avoid this situation, I re-wrote the original equation as $$\tan^{-1}\left(\frac{2x+1}{x+1}\right)-\tan^{-1}\left(1+x\right)=\tan^{-1}\left(1+x\right)-\tan^{-1}\left(\frac{2x-1}{x-1}\right)$$ and because while using the identity $\tan^{-1}\left(\alpha\right)-\tan^{-1}\left(\beta\right)=\tan^{-1}\left(\frac{\alpha-\beta}{1+\alpha\beta}\right)$, there are no restrictions on $\alpha,\beta$ I went ahead with this, and got the equation $$\frac{\frac{\left(2x+1\right)}{x+1}-\left(x+1\right)}{2\left(x+1\right)}=\frac{\left(x+1\right)-\frac{\left(2x-1\right)}{x-1}}{2x}$$ which simplified to $\left(x+1\right)^{2}\left(x-2\right)+x^{2}\left(x-1\right)=0$ and surprisingly from here, I am not even getting those roots!

I used this method to avoid the extraneous roots, what is going wrong here?

Edit: The immediately above issue has been resolved, however, the first issue remains unresolved.


Note that all calculations have to be performed by hand, since these types of questions are asked in exams in which the allotted time per question averages out to 2-3 minutes.

V.G
  • 4,196
  • 1
    The $2x$ in your last denominator appears to be faulty. It looks like you cancelled $1+x$ with $x-1$. – Gerry Myerson Apr 03 '21 at 10:00
  • @GerryMyerson Oh yes! Thanks for pointing out. In the meantime, I also discovered that with the second approach, one needs to use $\alpha \beta >-1$ which I didn't consider. However, how do I verify the solutions with the first approach? Or How do I get the exact solutions directly from some method... – V.G Apr 03 '21 at 10:03
  • On verifying which roots are correct solutions: tan is a bijection on its domain onto $\mathbb R$, so the only thing you need to check is whether any of said solutions falls outside the domain of the various quantities involved, and then you are good to go. –  Apr 03 '21 at 10:03
  • Two preliminary observations:
    1. " So, the possible roots are $x=0,\pm \sqrt{2}, -\dfrac{1}{2}$ , now my question is, how do I verify which roots satisfy the original equation (because one cannot see directly after plugging, whether it's true or not because it involves arctangents)? "

    A straightforward plug-and-check reveals that $x=0,\sqrt{2},-\dfrac{1}{2}$ satisfy the given equation, whereas $x=-\sqrt{2}$ doesn't. Why was the arctangent function an obstruction?

    – ryang Apr 03 '21 at 10:09
  • " while using the identity $\tan^{-1}\left(\alpha\right)-\tan^{-1}\left(\beta\right)=\tan^{-1}\left(\frac{\alpha-\beta}{1+\alpha\beta}\right)$, there are no restrictions on $\alpha,\beta$ "
  • This identity may not be true when $\alpha\beta\nless 1.$

    – ryang Apr 03 '21 at 10:09
  • @RyanG I agree that I was wrong to say that there are no restrictions, indeed $\alpha \beta$ should be $>-1$. But, how do you plug "straightforward and check" that for example $-1/2 , \sqrt{2}$ satisfy and $-\sqrt{2}$ doesn't? – V.G Apr 03 '21 at 10:12
  • @RyanG Suppose we want to check $-1/2$, on LHS, we have $\tan^{-1} (4/3)$ and while on RHS, it is $2\tan^{-1}(1/2)$ , I don't see immediately how they are equal... and similarly with $\pm \sqrt{2}$.... – V.G Apr 03 '21 at 10:15
  • *correction: I meant to specify the condition $\alpha(-\beta)\nless1.\quad$ 2. re: the plug and check: isn't this verification good enough?
  • – ryang Apr 03 '21 at 10:25
  • @RyanG No, this isn't good for me actually. These types of problems are asked in exams and one needs to do it all by hand and preferably within 3 minutes (since that is the average time per question we get). If I had to use WA, I could have put the whole equation into it in the first place. – V.G Apr 03 '21 at 10:27
  • Cool. You might wish to edit the question to indicate the context (e.g., the calculator restriction). "If I had to use WA, I could have put the whole equation into it in the first place." Well, a regular scientific calculator which is typically permissible in exams can't solve the equation but can verify that said root is extraneous. – ryang Apr 03 '21 at 10:29
  • @LightYagami, Let me know if this doesn't help: https://math.stackexchange.com/questions/1837410/inverse-trigonometric-function-identity-doubt-tan-1x-tan-1y-pi-tan/1837799#1837799 – lab bhattacharjee Apr 04 '21 at 10:57
  • @labbhattacharjee If you thought I wasn't aware of the definition of the sums of arctangents, then you were wrong. But if it can help me in some way, and I am missing something, please can you explain? The issue is ONLY that I need to prove that $-\sqrt{2}$ doesn't satisfy the original equation and that the other roots do as already mentioned at many places here before. – V.G Apr 04 '21 at 11:17
  • @LightYagami, Please find my answer below – lab bhattacharjee Apr 04 '21 at 18:01