Seventy deadly sines

Calculus Level 3

f ( x ) = sin sin . . . sin 70 sines ( x ) . f(x)=\overbrace{\sin\sin ... \sin}^{70\text{ sines}} (x).

Calculate the seventh derivative of f f evaluated at x = 0. x=0.

Note: Using computer software to answer this question is an even deadlier sin than nesting 70 sines inside each other, so don't do it.


The answer is -19463360.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Mark Hennings
Jul 14, 2018

Consider the Maclaurin series sin ( x + a x 3 + b x 5 + c x 7 ) = x + ( a 1 6 ) x 3 + ( b 1 2 a + 1 120 ) x 5 + ( c 1 2 a 2 1 2 b + 1 24 a 1 5040 ) x 7 + \sin(x + ax^3 + bx^5 + cx^7) \; = \; x + \big(a - \tfrac16)x^3 + \big(b - \tfrac12a + \tfrac{1}{120}\big)x^5 + \big(c - \tfrac12a^2 - \tfrac12b + \tfrac{1}{24}a - \tfrac{1}{5040}\big)x^7 + \cdots Suppose that f n ( x ) f_n(x) is the n n -fold composition of sin \sin with itself, and suppose that f n f_n has the Maclaurin expansion f n ( x ) = x + a n x 3 + b n x 5 + c n x 7 + f_n(x) \; = \; x + a_nx^3 + b_nx^5 + c_nx^7 + \cdots Then we deduce that a n + 1 = a n 1 6 b n + 1 = b n 1 2 a n + 1 120 c n + 1 = c n 1 2 a n 2 1 2 b n + 1 24 a n 1 5040 \begin{aligned} a_{n+1} & = \; a_n - \tfrac16 \\ b_{n+1} & = \; b_n - \tfrac12a_n + \tfrac{1}{120} \\ c_{n+1} & = \; c_n - \tfrac12a_n^2 - \tfrac12b_n + \tfrac{1}{24}a_n - \tfrac{1}{5040} \end{aligned} with a 1 = 1 6 a_1 = -\tfrac16 , b 1 = 1 120 b_1 = \tfrac{1}{120} and c 1 = 1 5040 c_1 = -\tfrac{1}{5040} . Solving these recurrence relations in turn, we have a n = 1 6 n b n = 1 120 n ( 5 n 4 ) c n = 1 15120 n ( 175 n 2 336 n + 164 ) \begin{aligned} a_n & = \; -\tfrac16n \\ b_n & = \; \tfrac{1}{120}n(5n-4) \\ c_n & = \; -\tfrac{1}{15120}n(175n^2 - 336n + 164) \end{aligned} and we calculate f 70 ( 7 ) ( 0 ) = 7 ! × c 70 = 19463360 f_{70}^{(7)}(0) = 7! \times c_{70} \; = \; \boxed{-19463360} .

My solution is not materially different, only choosing the form f n ( x ) = x + a n 3 ! x 3 + b n 5 ! x 5 + c n 7 ! x 7 + f_n(x) = x + \frac{a_n}{3!}x^3 + \frac{b_n}{5!}x^5 + \frac{c_n}{7!}x^7 + \cdots to clear all the denominators.

Still, nice writeup, and you beat me to it as well ^_^

Brian Moehring - 2 years, 11 months ago

Hey Mark, can you please explain why have you taken only terms up to x^7 in the first line? I would have assumed that higher terms would influence on the Maclaurin expansion. Thanks

Omer Lichtenstein - 2 years, 10 months ago

Log in to reply

The higher order terms will affect the whole Maclaurin series, but will not affect the terms up to x 7 x^7 . If you consider sin f ( x ) = f ( x ) 1 3 ! f ( x ) 3 + 1 5 ! f ( x ) 5 1 7 ! f ( x ) 7 + 1 9 ! f ( x ) 9 + \sin f(x) \; = \; f(x) - \tfrac{1}{3!}f(x)^3 + \tfrac{1}{5!}f(x)^5 - \tfrac{1}{7!}f(x)^7 + \tfrac{1}{9!}f(x)^9 + \cdots where f ( x ) f(x) has a Maclaurin expansion with f ( 0 ) = 0 f(0) = 0 , then the smallest order term in the expansion of f ( x ) n f(x)^n will be the x n x^n term. Thus, if we are only interested in the terms of the Maclaurin expansion of sin f ( x ) \sin f(x) up to x 7 x^7 , we do not need to worry about the terms f ( x ) 9 f(x)^9 and beyond.

Mark Hennings - 2 years, 10 months ago

Log in to reply

Nice. Thank you very much

Omer Lichtenstein - 2 years, 10 months ago

Did the same, I found b n b_n but I made mistake(s) while solving the recurrence relation for c n c_n (the recurrence relation was right) that I was not able to found...My expression of c n c_n gives me 19294310 -19294310 so close !! :'(

Théo Leblanc - 2 years, 10 months ago

I guess I'm in the wrong neighborhood.

Ichan Tongohan - 2 years, 10 months ago

I wasn't up to this. My brain can't even comprehend the simpler solution below.

Ichan Tongohan - 2 years, 10 months ago

This is a very neat solution! Can anyone please explain why a1 is equal to = -1/6 ? I don't get it. Where does that come from? I mean setting a,b,c to be all equal to 0 will yield the normal sin(x) expansion with the coefficient for x^3 to be -1/6, but why does a1 = -1/6 ? Does it come from a0 or something?

Thanachote Katanyutapant - 2 years, 9 months ago

You need to justify some things. Like sin(sin(x)) can be obtained by substituting the power series for sin(x) in for x in the power series of sin(x). We are dealing with infinite series. Can we do all this.

Srikanth Tupurani - 2 years, 6 months ago

Log in to reply

We are dealing with absolutely convergent infinite series. Yes we can do all this!

Mark Hennings - 2 years, 6 months ago

Log in to reply

nice solution. yes. just i wrote this as there are many problems of this type in brilliant.org. we don't have patience to write full details. but we need to justify things. these are not some algebraic expressions. we are dealing with infinite series.

Srikanth Tupurani - 2 years, 6 months ago

Log in to reply

@Srikanth Tupurani there is a nicebook on infinite series by bromwich. this is an amazing book. we can see some mysterious infinite series in this book.

Srikanth Tupurani - 2 years, 6 months ago
Jeremy Galvagni
Jul 23, 2018

How to solve by sinning (WolframAlpha):

I didn't dare jump to 70 nested sines and seventh derivatives so I found smaller cases and looked for a pattern.

First, any even depth derivative with any number of nested sines will give 0 at x=0

First derivatives for 1,2,3... nested sines are 1, 1, 1, 1... ok, nice pattern.

Third derivatives go -1, -2, -3, -4... also nice and simple.

Fifth derivatives go 1, 12, 33, 64, 105... messier, but it turns out to be quadratic. 5 n 2 4 n 5n^{2}-4n

I've got enough to guess at a pattern here. These are all polynomials with increasing degree. I'm guessing the next will fit a cubic.

Seventh derivatives go -1,-128, -731, -2160, -4765... It does indeed fit the cubic 1 3 ( 175 n 3 336 n 2 + 164 n ) -\frac{1}{3}(175n^{3}-336n^{2}+164n)

Plug 70 into this to get the answer -19463360.

I've only shared this to show how one might approach a problem of this type if all else fails. My hat is off to those who used the Maclaurin series.

[Please don't Upvote this solution, it isn't worthy. Mark's is.]

Frankly speaking, I was working on the similar lines, I first found the script for Wolfram Alpha "seventh derivative nest(sin,x,n) at x=0" for finding the series from (n=1,8) {-1, -128, -731, -2160, -4765, -8896, -14903, -23136). The wolfram alpha was showing time out beyond n=8.

I just fell short of typing "best fit (-1, -128, -731, -2160, -4765)" in wolfram alpha and I also would have got, best Cubic fit answer as -(58+1/3)n^3+112n^2-(54+2/3)n, identical to your answer.

Wolfram Alpha is tailor made app for solving this problem.

Vinod Kumar - 2 years, 10 months ago

Log in to reply

Wow, I could have cheated even more than I did? I didn't try pushing wolfram alpha hard enough.

Jeremy Galvagni - 2 years, 10 months ago

Up voting this instead of Mark's because it's iterative from an ordinary plage, instead of starting off at a series I've never heard of.

Omar Balbuena - 2 years, 10 months ago

Thumbs up for the approach! I did exactly the same!

Carlos Higuera - 2 years, 10 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...