In the catenary variational problem we have to find the curve that results from letting a weighted string with fixed length hang between two endpoints, and waiting until it minimizes its potential energy:

enter image description here

The constrained lagrangian for this problem is: $$L(y(x),y'(x))=(y(x) - \lambda)\cdot \sqrt{1+y'(x)^2}$$

What is wrong with my approach?

We have to solve the Euler-lagrange Equation, which is:

$$L_y=\frac d{dx}L_{y'}$$

So we find the derivatives of $L$: $$L_y=\sqrt{1+y'(x)^2}$$ $$L_{y'}= (y(x)-\lambda)\frac {y'(x)}{\sqrt{1+y'(x)^2}}$$ $$\frac d {dx} L_{y'}= \frac {y'(x)^2}{\sqrt{1+y'(x)^2}}+ y''(x)\cdot (y(x)-\lambda)\frac {1}{({1+y'(x)^2})^{3/2}}$$

This results in the Euler-Lagrange equation: $$\frac {y'(x)^2}{\sqrt{1+y'(x)^2}}+ y''(x)\cdot (y(x)-\lambda)\frac {1}{({1+y'(x)^2})^{3/2}}=\sqrt{1+y'(x)^2}$$

What surprised me is that according to this website, the Euler-lagrange equation looks very different:

enter image description here


There is not even a $y''(x)$ in their Euler-Lagrange equation. What am I doing wrong?

ps. here is another question that I posted where I seem to be making a very similar mistake, so I must fundamentally misunderstand something about the calculus of variations and the Euler-Lagrange Equation.

EDIT: (based on @user121049's comment): I've tried doing what you said, and it gets me to the same result, so this solves my question. (But I'm still not sure about the final solution) :

$$y''\frac {(y-\lambda)} {1+y'^2}=1 \implies \int y'' \cdot \frac {y'}{1+y'^2}dx=\int \frac {y'}{y-\lambda}dx$$ Which implies $$\int \cdot \frac {y'}{1+y'^2}d(y')=\int \frac {1}{y-\lambda}dy\implies \ln|1+y'^2|=2 \ln(|y-\lambda|^2)+C\implies$$


This Answers my question.

Now to solve it further: substitute $z(x)=y(x)-\lambda$ to get $z^2=c_1(1+z')\implies z'=\sqrt{\frac {z^2 -c_1}{c_1}}\implies c_1\int \frac 1 {\sqrt {z^2-a}}dz=x+c_2\implies \ln|\sqrt {z^2-c_1}+x|=\frac x c_1 +c_2\implies$ $$y(x)=z(x)+\lambda =\sqrt {\frac {c_2}{x^2}e^{\frac x {c_1} } +c_1}+\lambda$$

Is this correct? It doesn't seem like it describes the problem.

  • $\begingroup$ They have already integrated once. Your equations look right. Multiple through by sqrt term to simplify. You can then rejig it to look like $y'' f(y')=y'/(y-\lambda)$. This you can integrate. $\endgroup$ – user121049 Aug 30 '17 at 19:39

The clue is that your Lagrangian is independent of $x$. For any Lagrangian function $L\equiv L(\,y(x),y'(x)\,)$:

$$\frac{dL}{dx} \,=\, \frac{\partial L}{\partial y}\,y' + \frac{\partial L}{\partial y'} y'' \quad\Longrightarrow\quad \frac{\partial L}{\partial y}\,y' \,=\, \frac{dL}{dx} - \frac{\partial L}{\partial y'} y''. \qquad\qquad(*)$$

From the Euler-Lagrange equations:

$$\begin{aligned} \frac{\partial L}{\partial y} - \frac{d}{dx}\left(\frac{\partial L}{\partial y'}\right) \;=\; 0\quad&\Longrightarrow \quad y'\frac{\partial L}{\partial y} - y'\frac{d}{dx}\left(\frac{\partial L}{\partial y'}\right) \;=\; 0 \\[0.2cm] \quad&\Longrightarrow\quad \frac{dL}{dx} - \frac{\partial L}{\partial y'} y'' - y'\frac{d}{dx}\left(\frac{\partial L}{\partial y'}\right) \;=\; 0 \\[0.2cm] \quad&\Longrightarrow\quad \frac{dL}{dx} - \frac{d}{dx}\left[y'\frac{\partial L}{\partial y'}\right] \;=\; 0\\[0.2cm] \quad &\Longrightarrow\quad \frac{d}{dx}\left[ L - y'\frac{\partial L}{\partial y'}\right] \;=\; 0 \end{aligned}$$

where on the second line we substituted in $(*)$. Integrating this, as @user121049 already mentioned, gives you

$$L - y'\frac{\partial L}{\partial y'} \;=\; a$$

for some constant $a$. This is the Beltrami identity; you should be able to use this derivation to help with your problem.

  • $\begingroup$ What do you mean by $\frac{\partial L}{dy}$ in the Euler-Lagrange equation? shouldn't this be $\frac{\partial L}{\partial y'}$? $\endgroup$ – user56834 Aug 31 '17 at 5:08
  • $\begingroup$ To be clear, I actually have never before seen someone write a partial operator $\partial$ in the numerator but a $d$ operator in the denominator. Does this have a specific meaning or is it just a typo? $\endgroup$ – user56834 Aug 31 '17 at 5:44
  • $\begingroup$ Yes - you're right on both counts! The primes have been added as has $d\rightarrow\partial$, so thanks for the corrections. My math gets sloppy late at night! $\endgroup$ – AloneAndConfused Aug 31 '17 at 6:37
  • 1
    $\begingroup$ I now understand it, thank you :) $\endgroup$ – user56834 Aug 31 '17 at 7:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.