3

In an attempt to find a way to express the n-th derivative of $y=\frac{1}{(1+x^2)^2}$ using the binomial theorem I got stuck in the last computation.

I'll explain what I did:

$$f(x)=\frac{1}{(1+x^2)^2}=\frac{1}{(x+i)^2(x-i)^2}=\\ \frac{i}{4}\left( \frac{1}{x+i} \right)-\frac{1}{4}\left( \frac{1}{(x+i)^2} \right)-\frac{i}{4}\left( \frac{1}{x-i} \right)-\frac{1}{4}\left( \frac{1}{(x-i)^2} \right)$$

Then I used $y=\frac{1}{ax+b} \Rightarrow y_{n}=\frac{(-1)^nn!a^n}{(ax+b)^{n+1}}$ and $y=\frac{1}{(ax+b)^2} \Rightarrow y_{n}=\frac{(-1)^n(n+1)!a^n}{(ax+b)^{n+2}}$ to get:

$$\frac{(-1)^n}{4}\left( in!\left( \frac{(x-i)^{n+1}-(x+i)^{n+1}}{(x^2+1)^{n+1}} \right)-(n+1)!\left( \frac{(x-i)^{n+2}+(x+i)^{n+2}}{(x^2+1)^{n+2}} \right) \right)$$

Then I stopped for a second and I analized $(x-i)^{n+1}-(x+i)^{n+1}$ and $(x-i)^{n+2}+(x+i)^{n+2}$ using the binomial theorem:

-

$(x-i)^{n+1}-(x+i)^{n+1}$ = $$\sum_{k=0}^{n+1}\binom{n+1}{k}x^{n+1-k}\cdot i^k((-1)^k-1)$$

$i^k((-1)^k-1) =0$ for $k=2n$ ,whereas for $k=2n+1$ $\longrightarrow$$-2i(-1)^{\frac{k-1}{2}}$

-

$(x-i)^{n+2}+(x+i)^{n+2}$ = $$\sum_{k=0}^{n+2}\binom{n+2}{k}x^{n+2-k}\cdot i^k((-1)^k+1)$$

$i^k((-1)^k+1) =0$ for $k=2n+1$ , whereas for $k=2n$ $\longrightarrow$$2(-1)^{\frac{k}{2}}$.

And now I got stuck sorting out the indeces $n,k$ of summations; what I obtain is:

$$\frac{d^n}{dx^n}(f(x))=\\ \frac{(-1)^nn!}{2(x^2+1)^{n+1}}\cdot \sum_{k=1}^{n+1}\binom{n+1}{k}x^{n+1-k}(-1)^{\frac{k-1}{2}} - \frac{(-1)^n(n+1)!}{2(x^2+1)^{n+2}}\cdot \sum_{k=2}^{n+2}\binom{n+2}{k}x^{n+2-k}(-1)^{\frac{k}{2}}$$

But testing it I can see it’s wrong. Can anyone help me please sorting the $n,k$ indices and perhaps have a look at my computations also? Thank you

DavidM.
  • 150
  • 2
    Another approach is to start from the observation that the $n$th derivative is always of the form $f^{(n)}(x)=p_{n}(x)/(1+x^2)^{n+2}$ where $p_n(x)$ is some degree-$n$ polynomial. These satisfy the recurrence relation $p_{n+1}(x)=(1+x^2)p'_n(x)-2x(n+2) p_n(x)$ with $p_0(x)=1$. That at least makes it convenient to compute the numerators. – Semiclassical Aug 25 '21 at 23:19

4 Answers4

2

Update (28/08/2021): I eventually managed to solve my problem. This is the (tested) result obtained: $$\frac{d^n}{dx^n}\left(\frac{1}{(1+x^2)^2} \right)=$$ $$ \fbox{$\begin{array}{rcl}\displaystyle \frac{(-1)^{n+1}n!}{2(1+x^2)^{n+2}}\left\{ (n+1)\sum_{k=0/even}^{n+2}\binom{n+2}{k}x^{n+2-k}(-1)^{\frac{k}{2}}-(1+x^2)\sum_{k=0/odd}^{n+1}\binom{n+1}{k}x^{n+1-k}(-1)^{\frac{k-1}{2}} \right\}\end{array}$} $$

Thanks a lot to everyone who invested time and effort into answering me.

DavidM.
  • 150
1

I will check later about your indexes and the calculations.

I know you asked explicitly for a Binomial Series solution, yet I just wanted to point out that the $n$-th derivative of the function above can be easily found with

$$\dfrac{\text{d}^n}{\text{d}x^n} \dfrac{1}{(1+x^2)^2} = \dfrac{n!}{4}\left[-\dfrac{2 + n + ix}{(-x + i)^{n+2}} - \dfrac{2+n - ix}{(-x-i)^{n+2}}\right]$$

Enrico M.
  • 26,114
1

The analysis and the result stated in OPs answer are sound (+1). Since I also did already the calculation, here it is together with a few hints.

We obtain \begin{align*} &\color{blue}{D^n}\color{blue}{f(x)}\\ &=\frac{i}{4}(-1)^nn!\frac{(x-i)^{n+1}-(x+i)^{n+1}}{\left(x^2+1\right)^{n+1}}\\ &\qquad-\frac{1}{4}(-1)^n(n+1)!\frac{(x-i)^{n+2}+(x+i)^{n+2}}{\left(x^2+1\right)^{n+2}}\tag{1}\\ &=\frac{i(-1)^nn!}{4\left(x^2+1\right)^{n+1}}\sum_{k=0}^{n+1}\binom{n+1}{k}x^{n+1-k}i^k\left((-1)^k-1\right)\\ &\qquad-\frac{i(-1)^n(n+1)!}{4\left(x^2+1\right)^{n+2}}\sum_{k=0}^{n+2}\binom{n+2}{k}x^{n+2-k}i^k\left((-1)^k+1\right)\tag{2}\\ &=-\frac{(-1)^{n}n!}{4\left(x^2+1\right)^{n+1}}\sum_{k\geq 0}\binom{n+1}{2k+1}x^{n-2k}(-1)^k(-2)\\ &\qquad-\frac{(-1)^n(n+1)!}{4\left(x^2+1\right)^{n+2}}\sum_{k\geq 0}\binom{n+2}{2k}x^{n+2-2k}(-1)^k\cdot 2\tag{3}\\ &\,\,\color{blue}{=\frac{(-1)^nn!}{2\left(x^2+1\right)^{n+2}}\left(\left(x^2+1\right)\sum_{k\geq 0}\binom{n+1}{2k+1}(-1)^kx^{n-2k}\right.}\\ &\qquad\qquad\qquad\qquad\color{blue}{-\left.(n+1)\sum_{k\geq 0}\binom{n+2}{2k}(-1)^kx^{n+2-2k}\right)}\tag{4} \end{align*}

Comment:

  • In (2) we apply the binomial theorem and collect terms with equal power.

  • In (3) we observe that in the left-hand sum terms with even powers vanish. We take odd indices $k\to 2k+1$ only. Note we take the index range $k\geq 0$ which is admissible, since $\binom{n}{k}=0$ if $k>n$. In the right-hand sum terms with odd powers vanish and we take even indices $k\to 2k$ only.

  • In (4) we simplify further by factoring out the common denominator $\left(x^2+1\right)^{n+2}$.

Note: The answer stated by @Turing is nice and correct. We observe there is just one common denominator $(x-i)^{n+2}(x+i)^{n+2}=\left(x^2+1\right)^{n+2}$. We could think to also transform (1) this way to get another binomial expansion which is somewhat different to (4).

Markus Scheuer
  • 108,315
0

I am sorry, i haven't comprehend everything in detail, but you have to be very careful, when you would like to use $(a\cdot b)^n=a^n\cdot b^n$ for complex numbers $a,b$ instead of positive real numbers. There is a popular example here Why $\sqrt{-1 \cdot {-1}} \neq \sqrt{-1}^2$?

However, you can find the power series. We know $\frac{1}{1-t}=\sum_{k=0}^\infty t^k$ for $|t|<1$. If we calculate the derivative, we get $$ \frac{1}{(1-t)^2}=\sum_{k=0}^\infty kt^{k-1}=\sum_{k=0}^\infty (k+1)t^{k}. $$ Now, let $t=-x^2$, then \begin{align} \frac{1}{(1+x^2)^2} &=\sum_{k=0}^\infty (k+1)(-x^2)^{k}\\ &=\sum_{k=0}^\infty (-1)^k(k+1)x^{2k}\\ &=\sum_{2\mid n}^\infty (-1)^{n/2}\left(\frac{n}{2}+1\right)x^{n}\\ &=\sum_{2\mid n}^\infty \frac{n!(-1)^{n/2}(\frac{n}{2}+1)}{n!}x^{n}. \end{align} Hence by taylor's formula the $n$-th derivative of $f$ is $$f^{(n)}(0)=\begin{cases}0&2\nmid n\\ n!\cdot (-1)^{n/2}\cdot \left(\frac{n}{2}+1\right)&2\mid n \end{cases}.$$

Jochen
  • 2,260
  • I am sorry but I don’t understand how this could help me (?). I’d like to express the nth derivative of $y=\frac{1}{(1+x^2)^2}$ using the binomial theorem and I was asking if you can give it a check :) – DavidM. Aug 25 '21 at 21:48
  • Hello :) Ah ok, you aren't interested in an explicite expression? Yes, i know, you want to check, if your result is correct. If it is correct, all derivatives of odd order have to be constantly zero. Maybe, you can check if your expression is zero for $n=1$ or $n=3$. – Jochen Aug 25 '21 at 21:51
  • Are you sure, that you have to get a result using the binomial theorem? – Jochen Aug 25 '21 at 21:56
  • What do you mean by “derivatives of odd order have to be zero”? The third derivative of the function isn’t zero. Yes, I’d like to have a result expressed with the Binomial theorem – DavidM. Aug 25 '21 at 21:59
  • i haven't used any trigonometric functions :) But you're correct, i haven't used the binomial theorem as well. – Jochen Aug 25 '21 at 22:00
  • This only gives the $n$ evaluated at $0$. That's not what the OP is asking. – jjagmath Aug 27 '21 at 14:57
  • I know now, I am sorry, should i remove it? – Jochen Aug 27 '21 at 15:15