Tutorial Exercises: Calculus of Variations

1. The Catenoid
Consider the integrand F (x, y, y 0 ) = y
p
1 + (y 0 )2 in Eq. (1.5) when y is a function of x.
(a) Determine the Lagrange equation.
(b) There is a first integral; write it down and rearrange to make y 0 the subject.
(c) Solve the first-order differential equation by separating variables and integrating.

1. Solution: The Catenoid

(a) The Lagrange equation is

d ∂F ∂F d yy 0 p y 0 y 0 + yy 00 yy 0 y 0 y 00 p
( 0 )− = (p )− 1 + (y 0 )2 = p −p 3 − 1 + (y 0 )2 = 0
dx ∂y ∂y dx 1 + (y 0 )2 1 + (y 0 )2 1 + (y ) 0 2

This can be simplified to

yy 00 − 1 − y 0 y 0
p 3 = 0.
1 + (y 0 )2
Since the denominator cannot be zero this gives the 2nd order ODE
yy 00 − 1 − y 02 = 0.

(b) Since F does not depend explicitly on x there is a first integral

∂F yy 0
y0 0
p
− F = y − y 1 + (y 0 )2 = C
∂y 0
p
1 + (y 0 )2
Rearranging gives the 1st order ODE
r
0 y2 − C 2
y =± .
C2
(c) Separating variables and integrating gives
Z Z
dy dx
p =±
y2 − C 2 C
A good substitution to try is y = C cosh(α) then dy = C sinh(α)dα, thus
Z
C sinh(α)dα x+D
=± .
C sinh(α) C
The l.h.s gives α = cosh−1 (y/C) and therefore
x+D
y = C cosh( )
C
2. Dido’s Problem
Consider the integrand F (s, y, y 0 ) = y
p
1 − (y 0 )2 in Eq. (1.10) when y is a function of s.
(a) Use a first integral to find an expression for y 0 .
(b) Solve the first-order differential equation by separation.
(c) Apply the boundary conditions y = 0 at both s = 0 and s = L to eliminate some constants.
(d) Find an equation for dx/ds and then find x(s).
(e) What geometric curve does this parametric solution represent?

1
2. Solution: Dido’s Problem
(a) Since F does not depend on s a first integral is
∂F −2y 0
y0 0
p
0
− F = y y p − y 1 − (y 0 )2 = C
∂y 2 1 − (y 0 )2
Rearranging gives p
y = −C 1 − (y 0 )2
and solving for y 0 gives the explicit 1st order ODE
p
C 2 − y2
y0 = ± .
C
(b) Separating variables gives Z Z
C dy
p = ± ds.
C 2 − y2
The integral on the left is an inverse trig function so that
y
C sin−1 ( ) = ±(s + D)
C
Thus
s+D
y = ±C sin
C
(c) The boundary conditions give
D
0 = ±C sin
C
and
L+D
0 = ±C sin .
C
All formal solutions are given by D/C = nπ and L/C + nπ = mπ so that C = L/(kπ)
and D = Ln/k for integers n, m, k. But note that D is merely shifting s, so we can choose
D = 0. The simplest solution is k = 1 so that C = Lπ . Choosing |k| > 1 we do not obtain
a positive y for all s ∈ [0, L]. Thus,
L sπ
y= sin
π L
(d) Now
dy sπ
= cos
ds L
Thus from  2
dx dy
+ ( )2 = 1
ds ds
we can determine that
dx sπ
= ± sin
ds L
Thus, by integrating
L sπ
x(s) = ∓
cos + C2
π L
The boundary conditions (and choosing x to be positive) lead to
L sπ L
x(s) = −
cos + .
π L π
There is no endpoint condition on x but instead the x-coordinate of the endpoint of the
optimal curve is determined by x(L) = 2L/π.

2
 (e) This is a semicircle of radius Lπ with centre at ( Lπ , 0). The area of the city is L2 /(2π). Notice
that the first integral can now be retrospectively interpreted as the radius of the city. First
integrals often have interesting interpretations.
3. The Catenary

(a) Determine F (x, y, y 0 ) for the integrand in Eq. (1.12) when y is a function of x.
(b) Solve the corresponding Euler-Lagrange equations or the first integral if there is one.
(c) Apply the boundary conditions y = h at both x = −a and x = a to determine the constants
of integration.

3. Solution: The Catenary

(a) The integral is
Z B Z B
λ
(ρgy + λ)ds = ρg (y + )ds
A A ρg
λ
but since λ is arbitrary there is not much point in writing ρg instead of writing just λ so
one can do the calculation using (y + λ)ds.
When y is a function of x this gives
p
F (x, y, y 0 ) = (y + λ) 1 + y 02 .

(b) Since F does not depend on x the appropriate first integral is

∂F (y + λ)y 0 p
y0 0 − F = y0 p − (y + λ) 1 + y 02 = C
∂y 1 + y 02
or
(y + λ)
−p = C.
1 + y 02
Thus,
(y + λ)2 − C 2
y 02 = .
C2
Solving by separation gives
Z Z
dy dx
p =±
(y + λ)2 − C 2 C
and a good substitution is y + λ = C cosh α or dy = C sinh α dα and so
x 
α=± +D .
C
The solution is y = C cosh[(x/C) + D] − λ.
(c) The boundary conditions x = −a and x = a can be used to eliminated D and λ. Even
though λ was introduced to control L at this stage we may use it to satisfy the other
boundary conditions. Hence we get
hxi hai
y = C cosh − C cosh + h.
C C
The final remaining unknown C is determined by the length of the catenary
Z Z a r Z a
x x h x ia a
L = ds = 1 + sinh2 dx = cosh dx = C sinh = 2C sinh .
x=−a C x=−a C C x=−a C
Unfortunately, the final result cannot be put in a form more explicit than that.

3
4. Calculating ds in a different coordinate system
Cylindrical polar coordinates are defined by

x = ρ cos φ
y = ρ sin φ
z = z

(a) Confirm that dx = dρ cos φ − ρ sin φdφ.

(b) Calculate a similar expression for dy.
(c) Starting from ds2 = dx2 + dy 2 + dz 2 show that ds2 = dρ2 + ρ2 dφ2 + dz 2 .
(d) Having warmed up with that calculation, repeat with spherical polar coordinates which are
defined by

x = r sin θ cos φ
y = r sin θ sin φ
z = r cos θ

and show that ds2 = dr2 + r2 dθ2 + r2 sin2 θdφ2 .

Hint: The spherical result is easier to get starting from the cylindrical result and using
ρ = r sin θ.

4. Solution: Calculating ds in a different coordinate system

(a) This is a simple application of the product rule dx = dρ cos φ − ρ sin φdφ.
(b) dy = dρ sin φ + ρ cos φdφ.
(c) Now
dx2 + dy 2 = (dρ cos φ − ρ sin φdφ)2 + (dρ sin φ + ρ cos φdφ)2
The cross terms cancel so

dx2 + dy 2 = dρ2 (cos2 φ + sin2 φ) + ρ2 dφ2 (sin2 φ + cos2 φ) = dρ2 + ρ2 dφ2

Adding dz 2 gives the desired result.

(d) Start with ds2 = dρ2 + ρ2 dφ2 + dz 2 . Now ρ2 = r2 sin2 θ and the product rule applied to dρ
gives
dρ = dr sin θ + r cos θdθ
and also
dz = dr cos θ − r sin θdθ
Thus substituting into ds2 from above gives

ds2 = (dr sin θ + r cos θdθ)2 + r2 sin2 θdφ2 + (dr cos θ − r sin θdθ)2

Once again the cross-terms cancel leaving

ds2 = dr2 (sin2 θ + cos2 θ) + r2 (cos2 θ + sin2 θ)dθ2 + r2 sin2 θdφ2

which simplifies to the desired result ds2 = dr2 + r2 dθ2 + r2 sin2 θdφ2 .

4
5. Geodesics on the Sphere
The equation of a sphere in spherical polar coordinates is particularly simple: it is r = a, where
a is a constant.

(a) Starting with ds in spherical polar coordinates, write down the simplified form of ds when
r = a is a constant.
(b) Use this expression for ds to write down an integral that represents the distance between
two points connected by a path that lies on the surface of a sphere. Write the integral in
the form where φ is a function of θ.
(c) Write down a first integral for this integrand.
(d) Show that
φ − φ0 = sin−1 [α cot θ]
satisfies the first integral, where φ0 and α are two independent constants.
(e) The equation of a plane through the origin is Ax + By + Cz = 0. Rewrite this equation in
spherical polar coordinates. Rearrange the equation to make it look like the solution above
and find α and φ0 in terms of A, B and C.
(f) Thus give a simple geometric description and method of finding geodesics on a sphere.

5. Solution: Geodesics on the Sphere

(a) If r = a is a constant then
ds2 = a2 dθ2 + a2 sin2 θ dφ2 .
(b) The integral is
Z Z θB q
I= ds = a 1 + sin2 θ φ0 2 dθ .
θA
p
Thus F (θ, φ, φ0 ) = 1 + sin2 θ φ0 2 .
(c) Since ∂F/∂φ = 0 a first integral is
sin2 θ φ0
p =C
1 + sin2 θ φ0 2
or !
0 C
φ =± p .
sin θ sin2 θ − C 2
(d) Direct differentiation and some algebra, yields the result.
(e) In spherical coordinates this becomes
Ar sin θ cos φ + Br sin θ sin φ + Cr cos θ = 0
The r cancels, the cos θ can be moved to the other side and both sides divided by sin θ to
give
A cos φ + B sin φ = −C cot θ
Trig identities can be used to rewrite the left hand side as
p
A2 + B 2 sin(φ − φ0 ) = −C cot θ
√
where φ0 = − tan−1 (A/B) and α = −C/ A2 + B 2 .
(f) In other words, the curve with the shortest distance lies simultaneously on the surface of a
sphere AND on a plane through the origin. The intersection of such a plane and a sphere
is called a great circle.

5
6. Geodesics in three dimensions

(a) Determine ds2 for Cartesian coordinates x, y, z in three dimensions.

(b) Write the integrand for the geodesic problem when both y and z are functions of x.
(c) Write down all the first integrals. (There are three.)
(d) Pick two of the first integrals (presumably the nicest two) and solve the system of differential
equations.
(e) Interpret the solution geometrically.

6. Solution: Geodesics in three dimensions

(a)
ds2 = dx2 + dy 2 + dz 2

(b) p
F (x, y, z, y 0 , z 0 ) = 1 + y 02 + z 02 .

(c) The integrand is independent of y thus

y0
p = C1 .
1 + y 02 + z 02

The integrand is independent of z thus

z0
p = C2 .
1 + y 02 + z 02

The integrand is independent of t thus

y 02 z 02 p
p +p − 1 + y 02 + z 02 = C3
1 + y 02 + z 02 1 + y 02 + z 02

which is the same as

1
−p = C3
1 + y 02 + z 02
(d) These can be rearranged to give y 0 = A a constant, and z 0 = B a constant. Thus y = Ax+C
and z = Bx + D.
(e) This is a straight line in 3 dimensions.

7. The Brachistochrone in Parametric Form

(a) Determine the integrand F (t, x, y, ẋ, ẏ) for Eq. (1.6) when both x and y are functions of
some parameter t.
(b) Write down all the first integrals. Are any surprising?
(c) Show that the parametric form of the cycloid given in the text solves these equations.
(d) Find a different pair of functions x(t) and y(t) that also solve the same equations.
(e) Why should this be regarded as the same solution?

6
7. Solution: The Brachistochrone in Parametric Form

(a) p
ẋ2 + ẏ 2
F (t, x, y, ẋ, ẏ) = √
h−y
(b) The integrand is independent of x so

∂F ẋ
=√ p =C
∂ ẋ h − y ẋ2 + ẏ 2

The integrand is independent of t so

p
∂F ∂F ẋ2 ẏ 2 ẋ2 + ẏ 2
ẋ + ẏ −F = √ p +√ p − √ = 0 = a constant
∂ ẋ ∂ ẏ h − y ẋ2 + ẏ 2 h − y ẋ2 + ẏ 2 h−y

The information that 0 is a constant is true but not very useful!

(c) Let the parameter t be the same as θ in the text. Then ẋ = α2 − α2 cos(2t) and ẏ =
−α2 sin(2t) and substituting this into the first integral for x gives

α2 − α2 cos(2t)
p =C
α sin(t) 2α4 − 2α4 cos(2t)

which simplifies to
1 − cos(2t)
p =C
α sin(t) 2 − 2 cos(2t)
or
1
√ =C
α 2
(d) You can let θ equal any function of t and the equations will still be satisfied. This is because
there are an equal number of dots in the top and bottom of the LHS of the first integral.
(e) The parametric form of a curve is not unique, but they all represent the same curve.