A problem in class today with a neat, non-standard solution that I wanted to share.
In terms of difficulty, it needs some A level FP1 content, but the level of imagination required for the solution is something I could see in a TMUA paper (WE all flopped TMUA).
the problem
Show that
$$ \frac {\mathrm{d}^{2n}} {\mathrm{d}x^{2n}} x^2 \sin(2x) = f(x) \cos(2x) + g(x) \sin(2x) $$
for $n > 1$, where $f(x)$ and $g(x)$ are polynomials to be found.
hints
- Use Leibniz's theorem - how many terms do you need?
- How does the parity of $n$ change your result for $\frac {\mathrm{d}^{2n}} {\mathrm{d}x^{2n}} \sin(2x)$?
my solution
(I say "my solution", but this is very similar to the solution we walked through right after in class.)
Leibniz's theorem is really the only roadblock in mathematical knowledge here. The way we write it out in class is as a table of terms:
$$ \begin{array}{ c c c c } r & {}^{2n}C_r & f^{(r)} & g^{(2n-r)} \\ \hline \end{array} $$
Because we're differentiating $2n$ times, the binomial coefficients are ${}^{2n}C_r$. From there, the obvious split is $f(x) = x^2$ and $g(x) = \sin(2x)$. The key thing that makes this problem actually doable is that the $x^2$ term disappears after we differentiate 3 times:
$$ \begin{array}{ c c c c } r & {}^{2n}C_r & f^{(r)} & g^{(2n-r)} \\ \hline 0 & & x^2 & \\ 1 & & 2x & \\ 2 & & 2 & \\ 3 & & 0 & \\ \vdots & & \vdots & \end{array} $$
Which means that we only have to worry about the first 4 terms, up to $r=3$. That's why we choose $x^2$ as our $f$. Think of it like the ILATE rule for integration by parts. After that, we can quickly complete the ${}^{2n}C_r$ column:
$$ \begin{array}{ c c c c } r & {}^{2n}C_r & f^{(r)} & g^{(2n-r)} \\ \hline 0 & 1 & x^2 & \\ 1 & 2n & 2x & \\ 2 & n(2n-1) & 2 & \end{array} $$
And then the last column is the interesting part.
A good starting point is to begin by looking at what happens when you differentiate $\sin(2x)$.
$$ \begin{array}{ c c } g^{(0)} & \sin(2x) \\ g^{(1)} & 2\cos(2x) \\ g^{(2)} & -4\sin(2x) \\ g^{(3)} & -8\cos(2x) \\ g^{(4)} & 16\sin(2x) \end{array} $$
From this we can figure out a few things:
- The actual number in front of $\sin$ or $\cos$ (the absolute value of the coefficient) is just $2^{2n-r}$, since it spits out a 2 every time we differentiate.
- When $2n-r$ is even, we end up with a $\sin(2x)$, and when it's odd, we end up with a $\cos(2x)$.
- The sign of the result alternates every 2 derivatives.
Now we have to work backwards, all the way from $\frac {\mathrm{d}^{2n}} {\mathrm{d}x^{2n}} \sin(2x)$. We know:
- $r=0$, so the number in front of it is going to be $2^{2n}$.
- $2n-r=2n$ is even, so we're going to end up with a $\sin(2x)$.
But what's the sign of the first term? We've differentiated $2n$ times - you might notice that if $2n$ is a multiple of 4 (like in $g^{(0)}$ and $g^{(4)}$), the result is positive; but otherwise (like in $g^{(2)}$), the result is negative. In other words, if $n$ is odd, we multiply by -1, and otherwise, we multiply by 1.
Of course, a neat little way of writing that is $(-1)^n$! After that, we can figure out the other two terms by integrating back from our final result without much work.
Finally, if we put it allll together, we end up with this:
$$ \begin{array}{ c c c c } r & {}^{2n}C_r & f^{(r)} & g^{(2n-r)} \\ \hline 0 & 1 & x^2 & (-1)^n \cdot 2^{2n} \sin(2x) \\ 1 & 2n & 2x & -(-1)^n \cdot 2^{2n-1} \cos(2x) \\ 2 & n(2n-1) & 2 & -(-1)^n \cdot 2^{2n-2} \sin(2x) \end{array} $$
These are all of the terms of the Leibniz expansion (or whatever) of the derivative, so we can sum them all up.
$$ \begin{aligned} = \left(-1\right)^n \left(2^{2n}x^2 - 2^{2n-1}n\left(2n-1\right)\right) \sin\left(2x\right) \\ -\left(-1\right)^n \left(2^{2n+1}nx\right) \cos\left(2x\right) \\ = \boxed{\left(-1\right)^n 2^{2n-1} \left(\left(2x^2 - n\left(2n-1\right)\right) \sin\left(2x\right) - 4nx \cos\left(2x\right)\right)} \end{aligned} $$
Ugly, isn't it? I still think it's a really nice result with an interesting method.
This problem was cool, but the main reason I wrote this up was to give me practise writing in $\TeX$ (I love that it does that) and make sure that MathJax was working properly. I think that mission was certainly accomplished.
I don't promise to make this a series, but I might be back soon!