8

I'm making a final exam for a first course in calculus at a university. I need some suggestions for a good bonus problem. By "good" I mean an interesting problem that some of the brighter students could solve in about 15 minutes.

Hopefully it should involve limits, derivatives, or integrals.

The problem should NOT involve trig functions.

The problem should NOT be theoretical. These students can't do real analysis proofs, for instance.

10 Answers10

10

Here are a few interesting problems:

  1. Let $f^n(x)$ denote the nth iterate of a function (for example, $f^2(x)=(f\circ f)(x)$). If $f$ has a fixed point at $x=x_0$ (meaning that $f(x_0)=x_0$) then what is $\frac{d}{dx}f^n(x_0)$ in terms of $f'(x)$?

  2. Prove that the intersection of the parabola and the rectangle in the picture is $\frac{2}{3}$ the area of the rectangle:

enter image description here

  1. A triangle starts is an equilateral triangle $ABC$ at $t=0$. However, each second, both $BC$ and the length of the altitude to $BC$ increase by $1$ unit. What does the measure of angle $BAC$ approach as $t$ approaches $\infty$?
Franklin Pezzuti Dyer
  • 39,754
  • 9
  • 73
  • 166
4

Let $L_t$ be the line segment connecting $(t, 0)$ to $(0, 1-t)$, and let $R$ be the union of all these line segments for $t \in [0,1]$. Find the area of $R$.

(Inspired by a book I had as a child which claimed drawing these line segments would trace out a quarter circle -- which turns out to be decisively false. The result just looks vaguely similar to a quarter circle.)

  • Is that really false? I remember looking at it in highschool and I seemed to have convinced myself it was true. I guess if the area of $R$ is anything other than $1 - \frac{\pi}{4}$ it would have to be false! – pjs36 May 03 '17 at 00:00
  • 3
    Trying not to give too much away: intuitively from the picture, you'd expect the midpoint of $L_{1/2}$ to be on the top boundary of the region, and it turns out that's true - but that's $(1/4, 1/4)$ and not $(1 - 1/\sqrt{2}, 1 - 1/\sqrt{2})$. – Daniel Schepler May 03 '17 at 00:05
3
  1. Compute $$ \lim_{x\to0}\frac x{1-e^{x^2}}\int_0^xe^{y^2}\,dy $$

  2. Let $p>1$; prove the following inequality $$ 2^{1-p}\le\frac{x^p+y^p}{(x+y)^p}\le1\;\;\forall x,y>0 $$

  3. Let $f\in\mathcal C^2(\Bbb R)$ s.t. $$ |f(x)|<1\;\;\forall x\in\Bbb R\\ f(0)^2+f'(0)^2=4\;\;. $$ Show that there exists $\xi\in\Bbb R$ such that $f''(\xi)+f(\xi)=0$ (hint: consider the function $g(x):=f(x)^2+f'(x)^2$).

  4. Let $\{a_n\}_n\subset[0,+\infty[$; then prove that $$ \sum_na_n\;\;\mbox{converges iff}\;\;\; \sum_n\frac{a_n}{1+a_n}\;\;\mbox{converges} $$

Joe
  • 11,745
2

My suggestion:

  1. Take any relatively complicated calculus problem.
  2. Describe it entirely in words, with no diagrams, variable names, or math symbols.

And for a specific suggestion:

What is the volume of space enclosed by the outer limits of firing range of a cannon placed on a flat surface with gravitational pull equivalent to Earth's, free rotation in any direction, and a muzzle velocity of 400 meters per second, excluding any consideration of wind, air resistance, or bouncing shots?


Writing the correct applicable equation given a specific verbal description of a problem is a key skill, and one that (in my opinion) quite well distinguishes the brighter students from those who can only solve a ready-made equation. Writing out the right equation is more than half the problem.

Another tricky one:

With a spaceship accelerating at a constant 1 Earth gravity for the entire 384400 km distance to the moon, reversing acceleration direction instantly at the halfway point, how long will it take to complete the journey?

Another, although this may require some trig:

An amusement ride centrifuge (in Earth gravity) with a radius (at the base) of 20 feet has its walls tilted outwards at 45 degrees. What speed (in revolutions per minute) must the centrifuge reach before objects inside will slide or roll up the walls?

(I think when I worked this myself I only needed basic geometry, but not 100% sure.)

Wildcard
  • 3,047
  • 3
    I don't think it's fair to assume that calculus students should know terms from physics except those talked about in calculus such as velocity, acceleration, etc. – Ahmed S. Attaalla May 03 '17 at 03:08
  • @AhmedS.Attaalla, it's easy enough to add on an explanatory sentence such as, "gravitational acceleration is approximately 9.8 meters per second per second downward." But for a bonus problem, I don't think it's an outrageous idea at all to reward some extra understanding. I would agree that a standard, typical exam question should be very similar to the problems they've already done on the course. But for a bonus problem? Any high school calculus student with a very good understanding of his material should be able to work the problems I've given, or at least make a very good start. – Wildcard May 03 '17 at 03:13
  • 2
    It's not just a bonus problem. It's a bonus problem in a mathematics* exam*. Explaining any technical term or giving a usable definition with numbers and units for any quantities not familiar to the class (specifically, any that have not been explicitly taught to them) is an absolute necessity. – Nij May 03 '17 at 06:20
2

This might be a tough one but whoever figures it out (without looking here...) is going to deserve the points:

Remind them that the derivative of a differentiable function is zero at a minimizer of that function, and that you can use this condition to locate the minimizer (e.g. finding the vertex of a quadratic).
Ask them to simply write down the above equation for the case where you're trying to find the curve $y(x)$ that has the minimum arc length from $x_1$ to $x_2$.
In other words, what expression do you need to take the derivative of, and with respect to what, in order to represent the fact that $y(x)$ has the minimum possible length from $x_1$ to $x_2$?

Clearly the curve would be a line segment, but that fact shouldn't really help them here. The idea is, if you didn't already know beforehand that such a curve would be linear, how would you start to prove it? Writing down the correct derivative and setting it equal to zero is the first step of that, which is what you're asking them to do.

user541686
  • 13,772
  • Is this calculus? Or calculus of variations? – Jason DeVito - on hiatus May 03 '17 at 16:42
  • 1
    @JasonDeVito: I don't know what your criterion for differentiating (heh) the two is, but you don't need Euler-Lagrange or anything like that from calculus of variations, if that's what you mean (even though proofs commonly use that). It's literally just a derivative with respect to a scalar that you need to set equal to zero. – user541686 May 03 '17 at 17:15
  • Then I think I'm missing something. I'm assuming we are trying to minimize $\int_{x_1}^{x_2} \sqrt{1 + (y')^2}; dx$ among all $C^1$ curves $y$ with $y(x_1) = y_1$ and $y(x_2) = y_2$. What am I thinking wrong? – Jason DeVito - on hiatus May 03 '17 at 17:42
  • @JasonDeVito: Haha, I guess the question is not easy ;) what you have is correct, just unfinished. Let $\hat{y}$ be the shortest curve and rewrite your $y$ and your expression in terms of that. What do you get? Can you now take a derivative and set it to zero? – user541686 May 03 '17 at 17:45
  • I feel like I'm still being dense. Write $y = \hat{y} + t h$ for some arbitrary $C^1$ curve $h$. Then, if I'm not screwing stuff up, $0=\frac{d}{dt}|{t=0} \int{x_1}^{x_2} \sqrt{1 + (\hat{y}' + th')^2}; dx = \int_{x_1}^{x_2} \frac{\hat{y}' h'}{\sqrt{1+ (\hat{y}')^2}}$. Could you give me more of a hint? – Jason DeVito - on hiatus May 03 '17 at 17:57
  • Aren't you done already? Though $h$ is not an arbitrary $C^1$ curve, it still has to satisfy $h(x_1) = h(x_2) = 0$. All you needed to do was introduce $t$ and differentiate with respect to it, which you did. That's all the question was asking for. – user541686 May 03 '17 at 17:59
  • I'm done? That integral must be $0$ for all $h$ meeting the condition you listed. This shoud imply $\hat{y}' = 0$, but I'm just not seeing it. (Am I working it the way you intended?) – Jason DeVito - on hiatus May 03 '17 at 18:03
  • @JasonDeVito: No, that last reasoning is not correct. You have $h'$ in the last integral, not $h$. The fact that it's $h'$ rather than $h$ means it's not forcing $\hat{y}'$ to be zero. Integrate by parts to get it in terms of $h$. Then you'll get $\hat{y}'' = 0$ which is the correct condition. (And yes, you're doing it the way I intended, but the solution I've written was for arbitrary curves and not just functions, so it was parametric and used norms and stuff. Similar procedure though.) – user541686 May 03 '17 at 18:07
  • Ha, of course we're not trying to get $\hat{y}' = 0$, woops. I got as far as integration by parts before my last comment, but the derivative of $\frac{\hat{y}}{\sqrt{1+(\hat{y}')^2}}$ was nasty enough to scare me off course. Incidentally, usually the arclength formula I used above is derived by knowing straight lines minimize distance., so this is somewhat circular. I guess this still proves uniqueness. Anyway, I get enough of the gist now to back to my work that actually matters ;-) Thanks for putting up with me. – Jason DeVito - on hiatus May 03 '17 at 18:52
  • @JasonDeVito: Happy to! :) It's one of my favorite problems haha. – user541686 May 03 '17 at 19:36
2

"By considering the turning point(s) of the function $f(x) = x^{1 \over x}$ on $(0, \infty)$ prove that $e^{\pi} > \pi^e$."

Differentiating reveals a global maximum at $x = e$ implying $f(e) > f(\pi)$ and the result follows after raising each side to ${\pi}e$.

  • It is a fun problem. I tend to think of it in terms of the function $f(x) = \ln(x)/x$. For a bonus question it is probably best to give no hint at all. A Calc III version of the problem is to give a heuristic solution by considering the linearization of the function $f(x,y) = x^y$ at the point $(3,3)$ – John Coleman May 03 '17 at 12:31
  • A simpler way to obtain the desired inequlity is to apply $e^x\geq 1+x$ for a smart choice of $x$. – mickep May 03 '17 at 17:18
  • Problem one, more like lots of fun. – mtheorylord May 24 '17 at 19:40
1

The following is a nice result of Euler:

Let $P(x)$ be a polynomial of degree $r<n$. Then $$ \sum_{k=0}^{n} (-1)^k \binom{n}{k} P(k) = 0. $$

There's a nice proof that simply applies $P(x \frac{d}{dx})$ to $(1-x)^{n}$. You can build up to it by asking for the proof for a monomial first, or start right at the beginning with $x\frac{d}{dx}$ acting on $x^k$, then calculating $x\frac{d}{dx}$ acting on the equality $(1-x)^n = \sum_{k=0}^{n} (-1)^k \binom{n}{k} x^k$, and so on. This was used on an entrance exam at my university a couple of years ago, with some guidance like this: the exact text of the question is as follows:

An operator $D$ is defined, for any function $f$, by $$ Df(x) = x\frac{df(x)}{dx}. $$ The notation $D^n$ means that $D$ is applied $n$ times; for example $$ D^2 f(x) = x\frac{d}{dx} \left( x \frac{df(x)}{dx} \right). $$ Show that for any constant $a$, $D^2x^a = a^2x^a$.

  1. Show that if $P(x)$ is a polynomial of degree $r$ (where $r \geqslant 1$) then, for any positive integer $n$, $D^n P(x)$ is also a polynomial of degree $r$.
  2. Show that if $n$ and $m$ are positive integers with $n<m$, then $D^n(1 − x)^m$ is divisible by $(1 − x)^{m−n}$.
  3. Deduce that, if $m$ and $n$ are positive integers with $n<m$, then $$ \sum_{r=0}^m (-1)^r \binom{m}{r} r^n = 0. $$

So it does the monomial case. $P(D)$ would then be a generalisation of this notation: if $P(x) = \sum_{k=0}^{n} a_k x^k$, then $P(D) = \sum_{k=0}^{n} a_k D^k$; this is an easy extension of the monomial case since $D$ is linear.

Chappers
  • 67,606
  • Hmm... my solution to that would be simply $LHS = (\Delta^n P)(0) = 0$. – Daniel Schepler May 03 '17 at 00:00
  • In my experience, finite differences aren't taught to that level of sophistication these days, at least in first courses in calculus. – Chappers May 03 '17 at 00:01
  • Could you elaborate on the proof? I'm having a bit of trouble following your explanation. More specifically, I don't fully understand what the notation $P(x \frac{d}{dx})$ signifies in your answer. – S.C.B. May 03 '17 at 11:47
  • 1
    @S.C.B. I'll dig up the specific text of the question; that should make it clearer. – Chappers May 03 '17 at 12:09
  • @S.C.B. Updated with more details. – Chappers May 03 '17 at 12:22
1

I suggest this terse (Euclidean) geometry problem, which can be solved using basic calculus

$A$, $B$, $C$, $D$ are distinct points, with $A$, $B$, $C$ collinear,
such that $|AB|=|BD|=|CD|=1$ and $|AC|=|AD|$.

Give the full set of possible $|AC|$ values.

Solution (hover mouse to reveal spoiler):

$\left\{\,{\sqrt 5-1\over2}\,,\,{\sqrt 5+1\over2}\,\right\}$

Illustration (spoiler).


As pointed in that other answer, writing the correct applicable equation given a specific verbal description of a problem is a key skill, thus alternatively the question might be

Alice, Bob and Carl stand on a straight line.

Alice is one furlong (220 yards) away from Bob.

Dana stands one furlong away from both Bob and Carl.

Carl is as far from Alice as Alice is from Dana.

How far can Alice be from Carl?

Answer using a well-formed sentence, with distance in furlong (exact, or with at least three significant digits).

fgrieu
  • 1,758
0

Here was one I gave to my students in first-semester calculus as a bonus problem on a regular midterm (not a final exam). It's taken from one of the usual college textbooks on calculus. You can always modify it a bit. A figure goes with it.

enter image description here

Define $F$ by $F(x) = \int_{0}^{x} f(t) \, dt$, where the graph of $f$ is shown.

  1. Compute: $F(0)$, $F(3)$, $F(6)$.
  2. State the critical numbers of $F$ that are in the open interval $(0, 7)$.
  3. For each critical number found in the previous part, determine if $F$ attains a local maximum value, a local minimum value, or neither at that critical number, and explain why.

By the way, you should check out some of the past exams given out at, say, MIT and Stanford. Some of the problems are very good, having many parts (part a, part b, etc.) and involving many key topics.

Mark Twain
  • 2,157
  • 4
    Why would this be a bonus problem? Everything here is very basic: high school students taking AP calculus must do this sort of thing on the AP exam (in the United States). – symplectomorphic May 03 '17 at 00:10
  • True, but it wasn't worth many points at all (and it wasn't on a final exam). It was mainly so my students would be exposed to it while they prepared for their final exam. – Mark Twain May 03 '17 at 00:13
  • 2
    Fair enough, but then this probably isn't an answer to the OP's question. – symplectomorphic May 03 '17 at 00:18
-3

Browse old problems from the Putnam exam. http://kskedlaya.org/putnam-archive/

keej
  • 1,213