Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.


The integral is a time-like curve between $x_1$ and $x_2$ and at imagine fgf(x1) is a lower left corner of the rectangle and fgf(x2) is the upper right corner and $x_2-x_1$ is the length of the base of the rectangle. The geodesic is the shortest length curve parametrized by this on the condition that the area = c. The solution isn't always a strait line but could be some curved function. How do I find the function ? I know this is a variational problem involving a length integral and an area integral.

Do I need to find the 2x2 metric tensor for the 2-d parameterization curve and solve a pair of differential equations like here ?

share|improve this question
Are you trying to solve $\inf \{ l(f) | \int f = c \}$ (where $l(f) = \int \sqrt{1+(f')^2}$)? –  copper.hat Jul 29 '12 at 5:50
yea...trying to find f –  mygeodesic Jul 29 '12 at 6:44
Are there other constraints, because the solution to that will be $f(x) = \frac{c}{x_2-x_1}$? –  copper.hat Jul 29 '12 at 6:46
this is a horizontal line which gives the areas and is a minimal length, but does not ajoin the endpoints. A simple case would be if the area parameter were (x2-x1)*f(x1)+(x2-x1)*(f(x2)-f(x1))/2 and then geodesic would be of the form y=mx+b –  mygeodesic Jul 29 '12 at 7:11

1 Answer 1

You can incorporate the area constraint with the Lagrange multiplier method; for example, the book of John Oprea Differential geometry and its applications has an accessible treatment of such variational problems.

However I recommend to look for a curve in parametric form $x=f(t), y=g(t)$ rather than in the graph form $y=f(x)$. The reason is as follows. The free part of the boundary (the curve you are looking for) wants to have constant curvature. Indeed, if a curve has different curvature $\kappa_1>\kappa_2$ at two points $p_1,p_2$, then we can push a little at the point $p_1$ to decrease the curvature, and push from the opposite side at $p_2$ to increase it. This perturbation will decrease the length without changing the area. A planar curve of constant curvature is a circular arc. So, the boundary wants to be a circular arc with given endpoints, bounding the given area. When such an arc is a graph $y=f(x)$, it is the solution. But in some cases (depending on your parameters) the extremal circular arc is not a graph, because it intersects some vertical line (either $x=x_1$ or $x=x_2$) twice. If you insist on the graph form of the solution, you create an obstacle problem, which makes the problem substantially more difficult.

On the geometric grounds (without doing the computations) I expect that when the obstacle comes into effect, there will not be a minimizer satisfying the given boundary conditions. Instead, a minimizing sequence will converge to a function which violates one or both boundary conditions, and whose graph is a circular arc tangent to the effective obstacle(s). For a simple example, take $x_1=-1, x_2=1$, prescribe the boundary conditions $f(-1)=f(1)=0$ and demand the area to be $10+\pi/2$. The minimizing functions will converge to $f(x)=5+\sqrt{1-x^2}$, which fails the boundary conditions. To recover the existence of minimizer, you can relax the problem by adding the length of vertical segments to the length of the proper graph: that is, the new variational problem is to minimize $|f(x_1)-y_1|+|f(x_2)-y_2|+\int_{x_1}^{x_2}\sqrt{1+f\,'(x)^2}\,dx$ subject to $\int_{x_1}^{x_2} f(x)\,dx=c$, and no boundary conditions. Still, an additional complication arises when the prescribed area $c$ is so small that $f$ wants to be negative.

share|improve this answer
thanks for your answer. i spent all day looking for pdf files looking for examples similar to my problem but only found one here: ocw.mit.edu/courses/mathematics/… solving for the parameters(m,c,d) of the ellipse doesn't see to be possible. i imagine this problem is too difficult to be included in most texts –  mygeodesic Jul 30 '12 at 3:19
@mygeodesic It's not going to be an ellipse, because ellipses do not have constant curvature. –  user31373 Jul 30 '12 at 4:00

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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