7

The following square has edges of size $1$ and I'm trying to find the area of the blue region trapped between the parabolic curves created by the straight lines (number of lines is technically infinite).

I assume that the simplest way to go about this is to use integration based on the parabolic equations that the lines create, but is there any more fundamental way of solving this maybe using infinite series and limits? Or at least explaining the integral solution using differentiation and limits.

P.S. You can imagine that a line is drawn at the point $\dfrac 1n$ of one edge to the point $1-\dfrac1n$ of the other edge where $n$ varies from $1$ to infinity.

square with edges of size 1

Kasravnd
  • 377
  • 2
  • 15
  • 1
    Sorry but from the numerical result obtained I've now a doubt that the line you are considering is an astroid, how are you assuming to obtain it? – user May 02 '23 at 20:30
  • Can you wright down formally sentence "parabolic curves created by the straight lines" - what is it? – zkutch May 02 '23 at 21:56
  • Unfortunately, it is not clear how the envelope of the gray lines defines a curve. You say it's a parabola, but there are different parabolic arcs that can be drawn. Moreover, it is not necessarily even a parabola. To uniquely specify the curve in question, you must state how the lines are being drawn. For example, if the lines are drawn with endpoints $(x,0)$ and $(1,x)$, that gives a particular shape to the envelope. Or the lines could be specified such that the total length of the line is equal to the width the square. – heropup May 03 '23 at 00:16
  • @heropup added more explanations on that. – Kasravnd May 03 '23 at 22:18
  • @zkutch It's added to the question now, you can check it out. – Kasravnd May 03 '23 at 22:20
  • 1
    According to your edit, if you draw a line for $n=1$ we obtain $(1,0)$ and $(0,1)$ and for $n=2$ we obtain $(1/2,0)$ and $(0,1/2)$ which doesn't seem in according with your draw. – user May 04 '23 at 09:01
  • I meant I was at a loss how constant length straight segments can make up a parabolic arc as an envelope. On closer look the lengths shrink in fact when slope is $45^{\circ}$ – Narasimham May 04 '23 at 13:06
  • @user edited the question to avoid confusion. Also, note that the number of lines are infinite. – Kasravnd May 04 '23 at 16:18
  • You wrote in your last edit "as mentioned in comments more accurately it would be an astroid", but that is adding even more confusion, in my opinion. If your lines (as I presume) connect point $(x,0)$ with $(0,1-x)$, and so on, then the envelope IS NOT an astroid, but four parabolic arcs. You get an astroid as the envelope of a segment with fixed length sliding on the sides of the square, but it is a completely different curve. – Intelligenti pauca May 04 '23 at 20:32
  • An astroid is neither a parabola nor a hyperbola, it's a different curve: https://en.wikipedia.org/wiki/Astroid – Intelligenti pauca May 05 '23 at 10:37
  • Yeah, it says "a hypocycloid with four cusps". – Kasravnd May 05 '23 at 16:04

3 Answers3

7

Consider a square $ABCD$ and four points $PQRS$ on its sides such that $AP=BQ=CR=DS$ (see figure below). $PQRS$ is a square with the same center $O$ as $ABCD$, hence $O$ lies on the circumcircle of triangles $APQ$, $BQR$, $CRS$, $DSP$.

But the focus of a parabola lies on the circumcircle of the triangle formed by any three of its tangents, hence $PQ$ is tangent to the parabola with focus $O$ and also tangent to $DA$ and $AB$, $QR$ is tangent to the parabola with focus $O$ and also tangent to $AB$ and $BC$, and so on.

enter image description here

Let's consider the parabola with focus $O$, tangent to $DA$ and $AB$. By symmetry, its axis is line $AC$ and its directrix passes through $A$, because two perpendicular tangents to a parabola meet on the directrix. Its vertex $V$ is the midpoint of $OA$ (because $O$ is also the midpoint of the chord of contact $BD$).

If this parabola meets one of the other parabolas at $E$ (see figure below), we can find $OE$ from: $$ OE=EH={EG\over\sqrt2}={1-OE\over\sqrt2} \implies OE=\sqrt2-1. $$ It is then not difficult to find the length of $$ OL=LE={OE\over\sqrt2}=1-{1\over2}\sqrt2 \quad\text{and}\quad LV=OV-OL={3\over4}\sqrt2-1. $$ The area of interest is then the sum of square $EIJK$ and of four parabolic segments whose area can be computed via Archimedes' theorem (area of parabolic segment $IEV$ is ${4\over3}\cdot$ area of triangle $IEV$): $$ area=4OL^2+4\cdot{4\over3}LE\cdot LV={8\sqrt2-10\over3}. $$

enter image description here

3

So, I (very randomly) happen to have been playing with this exact problem a few days ago. Here is my approach. Focusing first on the set of lines from the $y$-axis down to the $x$-axis, we have the following.

lines from (t,0) to (0,1-t)

Each line that goes from $(t,0)$ to $(0,1-t)$ has the equation

$$ y = \frac{t-1}{t} x + 1-t = x - \frac{1}{t}x +1 - t$$

for $t\in [0,1]$. Then we want to know what line (i.e. which $t$), for a given $x$ has a maximum $y$ value. Thus, we need to optimize with respect to $t$.

a red line from the x axis up vertical to the highest value that the other lines attain.

$$\frac{dy}{dt} = \frac{1}{t^2}x-1$$

Setting this equal to zero, to find critical points, we get $t = \sqrt{x}$, and this is a maximum! Thus,

$$ y = x - \frac{1}{\sqrt{x}}x +1 - \sqrt{x} = x-2\sqrt{x} +1$$

is the curve that is the maximized line at each $x$ value. (As is stated in other answers, this is equivalent to $\sqrt{x}+\sqrt{y}=1$)

A red curve that is tangent to all the curves that we drew originally.

Now that we have the curve, we just need to integrate properly. Noticing that the curves in your original image intersect at $x, y = \frac{1}{2}$. And symmetry tells us that the $x$ value where $y$ is $\frac{1}{2}$ is the same as the $y$ value where $x$ is $\frac{1}{2}$, which is $\left(1-\sqrt{\frac{1}{2}} \right)^2$.

So to find the area of the shaded region inside all four of the transformed versions of this curve, we can find the area of the purple region in the image below and multiply by four.

A shaded region from the center point (0.5, 0.5) to the graph straight in the x and y direction.

This will be done by solving (with wolfram alpha, since this is not the interesting part) $$\int_{\left(1-\sqrt{\frac{1}{2}} \right)^2}^\frac{1}{2} \frac{1}{2} - \left( x-2\sqrt{x} +1\right) dx = \frac{1}{6} (4 \sqrt{2}-5)$$

And so the total area is $\frac{2}{3} (4 \sqrt{2}-5)\approx 0.4379$. This is also what Intelligenti pauca got, when they used only geometry, beautifully.

N. Owad
  • 6,972
-1

Background:

The astroid is an envelope enclosed between a vertical wall and floor made by a sliding ladder of constant length L. $a=1.$ $$ (x/a,y/a) = (\cos^3 t, \sin ^3 t); x^ \frac 23 + y^ \frac 23 = a^ \frac 23; \tag 1 $$

HINT:

For Parabolic arc

A parabolic arc envelope occurs by variable ladder length when maximum length L is at $(0,90) $ deg and minimum length $ L/\sqrt {2}$ occurs at $45^\circ $ to floor.

We may say that the parabolic arc has a property of constant $ (x , y) $ intercepts sum at least between tangency points at $x,y$ axes. ( Because the parabola expands after tangential contact ). The straight line length varies.

Shifting of axes as given enables finding area in first quadrant. Other three quadrants have same area by symmetry of axes. The magenta parabolic arc has parametrization

$$ (x(t)/a,y(t)/a) = (\frac12- \cos^4 t,\frac12- \sin ^4 t); \tag 2$$

The domain for t is $( t= 0,\pi/2)$

enter image description here

The independent parameters $(t_1,t_2) $ are given from the above parametrization to be

$$ \left(\cos^{-1}\frac{1}{2^ {\frac14}}, \sin^{-1}\frac{1}{2^ {\frac14}}\right) \approx (0.571859,0.998937) \tag 3 $$

and four times the shaded area is

$$ 4 \int_{t1} ^{t2} y(t) x'(t) dt $$

Plug in from (2) and trig substitutions.

EDIT 1:

Derivation of parabolic arc.

${u}$ is shortening in y- axis segment= offset on x-axis. $$ \frac{x}{u}+ \frac{y}{a-u} = 1 \tag {1a}$$ $$ \frac{x}{u/a}+ \frac{y}{1-u/a} = a \tag {1b}$$ Partial differentiate (1 a) w.r.t. $u$ $$ \frac{-x}{u^2}+\frac{y}{(a-u)^2}=0 \tag 2 $$ $$ \frac{u}{a}= \frac{\sqrt x }{\sqrt x + \sqrt y }~; 1- \frac{u}{a}= \frac{ \sqrt y }{\sqrt x + \sqrt y }~; \tag 3 $$ Plug into (1b) and simplify $$ x+y + 2 \sqrt {xy} =a \tag 4 $$

Take square root

$$\sqrt{x}+\sqrt{y}=\sqrt{a} \tag 5 $$

Narasimham
  • 42,260