Minimal Surface of Revolution
Calculus of variations can be used to find the curve from a point (x_1,y_1) to a point (x_2,y_2) which, when revolved around the x-axis, yields a surface of smallest surface area A (i.e., the minimal surface). This is equivalent to finding the minimal surface passing through two circular wire frames. The area element is
| dA=2piyds=2piysqrt(1+y^('2))dx, |
(1)
|
so the surface area is
and the quantity we are minimizing is
| f=ysqrt(1+y^'^2). |
(3)
|
This equation has f_x=0, so we can use the Beltrami identity
to obtain
| y(1+y^('2))-yy^('2)=asqrt(1+y^('2)) |
(6)
|
| y=asqrt(1+y^('2)) |
(7)
|
| [画像: y/(sqrt(1+y^('2)))=a ] |
(8)
|
| [画像: (y^2)/(a^2)-1=y^('2) ] |
(9)
|
| [画像: y=acosh((x-b)/a), ] |
(12)
|
which is called a catenary, and the surface generated by rotating it is called a catenoid. The two constants a and b are determined from the two implicit equations
which cannot be solved analytically.
The general case is somewhat more complicated than this solution suggests. To see this, consider the minimal surface between two rings of equal radius y_0. Without loss of generality, take the origin at the midpoint of the two rings. Then the two endpoints are located at (-x_0,y_0) and (x_0,y_0), and
But cosh(-x)=cosh(x), so
Inverting each side
| -x_0-b=-x_0+b, |
(17)
|
so b=0 (as it must by symmetry, since we have chosen the origin between the two rings), and the equation of the minimal surface reduces to
| [画像: y=acosh(x/a). ] |
(18)
|
At the endpoints
| [画像: y_0=acosh((x_0)/a), ] |
(19)
|
but for certain values of x_0 and y_0, this equation has no solutions. The physical interpretation of this fact is that the surface breaks and forms circular disks in each ring to minimize area. Calculus of variations cannot be used to find such discontinuous solutions (known in this case as Goldschmidt solutions). The minimal surfaces for several choices of endpoints are shown above. The first two cases are catenoids, while the third case is a Goldschmidt solution.
To find the maximum value of x_0/y_0 at which catenary solutions can be obtained, let p=1/a. Then (17) gives
| y_0p=cosh(px_0). |
(20)
|
Now, denote the maximum value of x_0 as x_0^*. Then it will be true that dx_0/dp=0. Take d/dp of (20),
Now set dx_0/dp=0
| y_0=x_0sinh(px_0^*). |
(22)
|
From (20),
| py_0^*=cosh(px_0^*). |
(23)
|
| px_0^*=coth(px_0^*). |
(24)
|
Defining u=px_0^*,
| u=cothu. |
(25)
|
This has solution u=1.1996789403.... From (22), y_0p=coshu. Divide this by (25) to obtain y_0/x_0=sinhu, so the maximum possible value of x_0/y_0 is
Therefore, only Goldschmidt ring solutions exist for x_0/y_0>0.6627....
The surface area of the minimal catenoid surface is given by
but since
Some caution is needed in solving (◇) for a. If we take x_0=1/2 and y_0=1 then (◇) becomes
| [画像: 1=acosh(1/(2a)), ] |
(37)
|
which has two solutions: a_1=0.2350... ("deep"), and a_2=0.8483... ("flat"). However, upon plugging these into (◇) with x_0=1/2, we find A_1=6.8456... and A_2=5.9917.... So A_1 is not, in fact, a local minimum, and A_2 is the only true minimal solution.
The surface area of the catenoid solution equals that of the Goldschmidt solution when (◇) equals the area of two disks,
Plugging in
| [画像: (y_0)/a=cosh((x_0)/a) ] |
(41)
|
Defining
| u=(y_0)/a |
(43)
|
gives
| usqrt(u^2-1)+cosh^(-1)u-u^2=0. |
(44)
|
This has a solution u=1.2113614259. The value of x_0/y_0 for which
| A_(catenary)=A_(2 disks) |
(45)
|
is therefore
| [画像: (x_0)/(y_0)=((x_0)/a)/((y_0)/a)=(cosh^(-1)((y_0)/a))/((y_0)/a)=(cosh^(-1)u)/u=0.5276973967. ] |
(46)
|
For x_0/y_0 in (0.52770,0.6627), the catenary solution has larger area than the two disks, so it exists only as a local minimum.
There also exist solutions with a disk (of radius r) between the rings supported by two catenoids of revolution. The area is larger than that for a simple catenoid, but it is a local minimum. The equation of the positive half of this curve is
| [画像: y=c_1cosh(x/(c_1)+c_3). ] |
(47)
|
At (0,r),
| r=c_1cosh(c_3). |
(48)
|
At (x_0,y_0),
The area of the two catenoids is
Now let u=x/c_1+c_3, so du=dx/c_1
The area of the central disk is
| A_(disk)=pir^2=pic_1^2cosh^2c_3, |
(58)
|
so the total area is
| A=pic_1^2{sinh[2((x_0)/(c_1)+c_3)]+[cosh^2c_3-sinh(2c_3)]+(2x_0)/(c_1)}. |
(59)
|
By Plateau's laws, the catenoids meet at an angle of 120 degrees, so
and
| [画像: c_3=sinh^(-1)(1/(sqrt(3))). ] |
(63)
|
This means that
so
Now examine x_0/y_0,
| [画像: (x_0)/(y_0)=((x_0)/(c_1))/((y_0)/(c_1))=((x_0)/(c_1))/(cosh((x_0)/(c_1)+c_3))=usech(u+c_3), ] |
(68)
|
where u=x_0/c_1. Finding the maximum ratio of x_0/y_0 gives
| utanh(u+c_3)=1, |
(70)
|
with c_3=sinh^(-1)(1/sqrt(3)) as given above. The solution is u=1.0799632187, so the maximum value of x_0/y_0 for two catenoids with a central disk is y_0=0.4078241702.
If we are interested instead in finding the curve from a point (x_1,y_1) to a point (x_2,y_2) which, when revolved around the y-axis (as opposed to the x-axis), yields a surface of smallest surface area A, we proceed as above. Note that the solution is physically equivalent to that for rotation about the x-axis, but takes on a different mathematical form. The area element is
| dA=2pixds=2pixsqrt(1+y^('2))dx |
(71)
|
| [画像: A=2piintxsqrt(1+y^('2))dx, ] |
(72)
|
and the quantity we are minimizing is
| f=xsqrt(1+y^'^2). |
(73)
|
Taking the derivatives gives
so the Euler-Lagrange differential equation becomes
| [画像: (xy^')/(sqrt(1+y^('2)))=a ] |
(77)
|
| x^2y^('2)=a^2(1+y^('2)) |
(78)
|
| y^('2)(x^2-a^2)=a^2 |
(79)
|
| [画像: (dy)/(dx)=a/(sqrt(x^2-a^2)) ] |
(80)
|
Solving for x then gives
| [画像: x=acosh((y-b)/a), ] |
(82)
|
which is the equation for a catenary. The surface area of the catenoid product by rotation is
Isenberg (1992, p. 80) discusses finding the minimal surface passing through two rings with axes offset from each other.
See also
Minimal Surface, Surface of RevolutionExplore with Wolfram|Alpha
More things to try:
References
Arfken, G. Mathematical Methods for Physicists, 3rd ed. Orlando, FL: Academic Press, pp. 931-937, 1985.Goldstein, H. Classical Mechanics, 2nd ed. Reading, MA: Addison-Wesley, p. 42, 1980.Isenberg, C. The Science of Soap Films and Soap Bubbles. New York: Dover, pp. 79-80 and Appendix III, 1992.Referenced on Wolfram|Alpha
Minimal Surface of RevolutionCite this as:
Weisstein, Eric W. "Minimal Surface of Revolution." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/MinimalSurfaceofRevolution.html