13
$\begingroup$

I discovered that if we want an arc of catenary in the interval $[a,b]$ we solve $$\int_a^b \sqrt{\cosh '(x)^2+1} \, dx=\int_a^b \cosh x \, dx$$ which means that the "result" of the length is equal to the result of the area in the same interval, though in different units.

So I asked myself if there are any other curve with the same property.

I set $$y=\sqrt{y'^2+1}\to y^2=y'^2+1; y(0)=1$$ then $$y'=\sqrt{y^2-1}\to dx=\frac{dy}{\sqrt{y^2-1}}\to x=\cosh^{-1} \,y$$ hence, arbitrary constant is zero, $y=\cosh x$

But I am not sure how to deal with the other solution $y'=-\sqrt{y^2-1}$ even if Mathematica gives the same result $y=\cosh x$

I'd like someone checking this proof, you know: I am not a pro, I am just an (almost) retired high school teacher :)

Update 9/1/2020. Now I am officially retired :)

$\endgroup$
3
  • 2
    $\begingroup$ It should just be $$-\int\frac{dy}{\sqrt{y^2-1}}=-\cosh^{-1}(y)$$ $\endgroup$ Jul 17, 2017 at 16:35
  • 1
    $\begingroup$ Related (johndcook.com/blog/2017/05/15/when-length-equals-area) $\endgroup$
    – Jean Marie
    Jul 17, 2017 at 20:01
  • 4
    $\begingroup$ Differentiating $y^2=(y')^2+1$, we obtain $ 2yy'=2y'y'',$ so $y=y'' $ whenever $y'\ne 0.$ Since $y(0)=1$, this implies that $y(x)=\cosh x +K\sinh x$ for all $ x $, for some constant $ K.$ Substituting this into $y^2=(y')^2+1$ we find that $K=0.$ $\endgroup$ Jul 18, 2017 at 4:02

2 Answers 2

1
$\begingroup$

From

$$\frac{y'}{\sqrt{y^2-1}}=\pm1$$ you draw

$$\text{arcosh}(y)=c\pm x$$

and

$$y=\cosh(c\pm x).$$

With the initial condition $y(0)=1$,

$$y=\cosh(\pm x)$$ which is $$y=\cosh(x).$$

$\endgroup$
1
$\begingroup$

EDIT1:

I understood your question in this way:

How is it that the area under a Catenary is proportional to the arc length? i.e., how is $$ c=\dfrac{A}{L}$$ valid for some constant of proportionality $c$?

At first about the sign in front of radical sign in DE

Let us at the outset consider very familiar similar situations:

If two DEs are given as $$ y'= + \sqrt {1-y^2},\; y'= - \sqrt {1-y^2} $$

we have in either case by squaring $$ y^{'2} = (1-y^2) $$

Differentiate

$$ 2 y' y^{''}= -2 y y',\to y^{''}+y =0 $$ which is the differential equation of a sine curve.

With BC $ x=0,y=1,y'=0 \to y= \cos x $ in either case

Similarly if two DEs are given as

$$ y'= + \sqrt {y^2-1},\; y'= - \sqrt {y^2-1} $$

we have in either case

$$y^{'2}= (y^2-1)$$

Differentiating

$$ 2 y' y^{''}= 2 y y',\to y^{''}-y =0 $$ which is the differential equation of a Catenary. With BC $ x=0,y=1,y'=0 \to y= \cosh x $ in either case.

However, if you do not wish to square thereby losing its sign but wish to directly integrate the two BCs, the following:

$$ y'= + \sqrt {1-y^2},\; y'= - \sqrt {1-y^2} $$

we get

$$ \sin^{-1}y= x +c_1, \sin^{-1}y=- x-c_2 $$

$$y= \sin (x+c_1),y= -\sin (x+c_2)$$

For (an even ) a symmetric solution $ x=0, y=1 $ we have respectively

$$c_1=\pi/2, c_2= 3 \pi/2$$

both yielding the same solution

$$ y = \cos x $$

When we have here our actual case

$$ y'= + \sqrt {1+y^2},\; y'= - \sqrt {1+y^2} $$

we get

$$ \cosh^{-1}y= x +c_1, \cosh^{-1}y=- x-c_2 $$ $$y= \cosh (x+c_1),y= \cosh (x+c_2)$$

For even symmetric solution $ x=0, y=1 $ we have respectively

$$c_1= c_2= 0 $$

yielding both the same solution

$$ y = \cosh x $$

So we can say in conclusion that in front of any (square root) radical sign we have $\pm$ and both signs are equally applicable for first order DE. It is only by a convention that we put in a positive sign implying the unsaid negative. They result in the same differential equation and hence also the same integrand for given boundary conditions in this particular case.

Geometrically a negative or positive sign of derivative relates to different slopes of the curve in different portions of the curve.

Next to answer what I considered to be your main question let us set up its DE which uniquely defines the curve.

To get a physical/geometrical idea a length dimension quantity $c$ is introduced as the quotient of covered area $A$ to the length of its curved "roof".

$$c=\dfrac {\int y \; dx}{\int\sqrt{1+y'^2}dx}$$

Using Quotient Rule differentiate to simplify

$$c=\dfrac{ y} {\sqrt{1+y'^2}}= \to y' = \dfrac{\sqrt{y^2-c^2}}{c} $$

which is the differential equation of the unique curve being sought.

Integrating with boundary condition $ y(0)=c ,y'(0)=0,$ one obtains the equation of the only curve that satisfies the required property.

$$ \dfrac{y}{c}= \cosh\dfrac{x}{c}$$

which is recognized as a catenary as stated. And in association this property is recognized as well... that $c$ is the constant of proportionality which is the minimum distance of the catenary to the x-axis.

$$ c=\dfrac{A}{L}$$

as also shown here graphically.

enter image description here

$\endgroup$
5
  • $\begingroup$ I am happily officially retired since last September 1st :) $\endgroup$
    – Raffaele
    Oct 6, 2020 at 11:19
  • $\begingroup$ Hope you would have stimulating experience in teaching and also in this site. $\endgroup$
    – Narasimham
    Oct 6, 2020 at 12:23
  • $\begingroup$ I don't think that you address the OP's question, which is about negative $y'$. (And how you solve the equation is preyty mysterious.) :-) $\endgroup$
    – user65203
    Nov 17, 2020 at 9:53
  • $\begingroup$ @ Raffaele. In my answer before edit I had assumed that the differential equations of a parabola, sine curve etc. were understood. So did not go into their detail. Hope it is now fully clear. $\endgroup$
    – Narasimham
    Nov 17, 2020 at 20:38
  • $\begingroup$ @ Yves Daoust: What the OP said " I asked myself if there is any other curve with the same property " had me answer that way. $\endgroup$
    – Narasimham
    Nov 17, 2020 at 20:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .