Math 121A: Homework 11 solutions
1. For F ( x, y, y0 ) = y02 + y2 the Euler Lagrange equation is
d ∂F ∂F d 0
− 2y = 2(y00 − y).
0= 0
− = 2y
dx ∂y ∂y dx
This is a second-order linear differential equation. Substituting y = emx , it can be
seen that m2 = 1 and thus m = ±1. The general solution is therefore
y( x ) = Ae x + Be− x
for some constants A and B.
2. (a) If r 0 = dr/dθ, the given functional can be written as
p
F (θ, r, r 0 ) = r2 + r 02
and hence the Euler equation gives
d ∂F ∂F
0 = −
dθ ∂r 0 ∂r
d
r 0
r
= √ −√
dθ r 2 + r 02 r 2 + r 02
r 00 r (r + r 0 )
0 r
= √ − 2 − √ .
r2 + r 02 (r + r 02 )3/2 r 2 + r 02
Multiplying through by (r2 + r 02 )3/2 gives
0 = r 00 (r2 + r 02 ) − r 0 (rr 0 + r 0 r 00 ) − r (r2 + r 02 )
= r 00 r2 + r 00 r 02 − rr 02 − r 00 r 02 − r3 − rr 02
= r 00 r2 − 2rr 02 − r3
and by removing the common factor of r, the equation
r 00 r − 2r 02 − r2 = 0 (1)
is obtained. An alternative equation can be obtained by using the Beltrami
identity,
∂F p r 02 r2
C = F − r 0 0 = r 2 + r 02 − √ =√ (2)
∂r r 2 + r 02 r 2 + r 02
for some constant C.
1
(b) The straight line from (1, −1) to (1, 1) can be expressed in polar coordinates as
1
r (θ ) = = sec θ
cos θ
for the range − π4 ≤ θ ≤ π
4. The first and second derivatives are given by
0 sin θ 00 1 2 sin2 θ
r (θ ) = , r (θ ) = + .
cos2 θ cos θ cos3 θ
Substituting into Eq. 1 gives
1 2 sin2 θ sin2 θ 1
r 00 r − 2r 02 − r2 = + − 2 − =0
cos2 θ cos4 θ cos4 θ cos2 θ
and the equation is satisfied. Substituting into Eq. 2 gives
r2 sec2 θ 1
√ =q =p =1
r 2 + r 02 1
+ sin2 θ cos2 θ + sin2 θ
cos2 θ cos4 θ
and the equation is also satisfied.
3. (a) For F ( x, y, y0 ) = g( x ) 1 + y02 , the Euler equation gives
p
!
g( x )y0
d ∂F d
0= =
dx ∂y0
p
dx 1 + y 02
and hence
g( x )y0
p =C
1 + y 02
for some constant C. Therefore
[ g( x )]2 y02 = C2 + y02 C2
and
C2
y 02 = .
[ g( x )]2 − C2
Hence
C
y0 = p .
[ g( x )]2 − C2
(b) For g( x ) = 1, the equation becomes
C
y0 = √
1 − C2
2
which is equivalent to y0 = D for some constant D, and hence y( x ) = Dx + E
for some constant E. This will satisfy the boundary conditions for
y( x ) = 2x − 2.
√
If g( x ) = x, the equation is
C
y0 = √
x − C2
and hence
C
Z p
y( x ) = √ dx = 2C x − C2 + D.
x − C2
The boundary conditions give
p p
0 = y(1) = 2C 1 − C2 + D, 2 = y(2) = 2C 2 − C2 + D
If the hint is used, and D is assumed to be zero, then it can be seen that C = 1
is the only solution, and thus
√
y( x ) = 2 x − 1.
Without the hint, the solution can be found by eliminating D to obtain
p p
2C 2 − C2 = 2 + 2C 1 − C2 . (3)
Dividing by two and squaring both sides gives
p
C2 (2 − C2 ) = 1 + C2 (1 − C2 ) + 2C 1 − C2
and hence p
C2 − 1 = 2C 1 − C2 .
Squaring both sides again gives
C4 − 2C2 + 1 = 4C2 (1 − C2 ),
which simplifies to
5C4 − 6C2 + 1 = 0.
This can be factorized as
(5C2 − 1)(C2 − 1) = 0
√
and thus C = ±1, ± 1/5. Since squaring both sides may introduce more solu-
tions, these possible values for C can then be checked that they indeed satisfy
Eq. 3. It can be verified that C = 1 is the only valid solution.
3
2
1.5
1
y
0.5
Path for g( x ) =
√1
Path for g( x ) = x
0
1 1.2 1.4 1.6 1.8 2
x
Figure 1: Extremal paths for the weighted geodesics considered in question 3.
√
(c) The two solutions are plotted in Fig. 1. It can be seen that the path for g( x ) = x
√ near x = 1 before leveling off near x = 2. This should
increases more quickly
be expected since x is smaller near x = 1, and thus y is able to grow more
quickly in this region without incurring a large penalty in the integral.
4. (a) To derive the more general Euler equation, assume first that y( x ) is an extremal
path, and consider Y ( x ) = y( x ) + eη ( x ) where e is a small parameter and η ( x )
is a variation that is twice-differentiable. Assume further that η ( x1 ) = η ( x2 ) =
0 and that η 0 ( x1 ) = η 0 ( x2 ) = 0. Then
Z x2 Z x2
0 00
I [Y ] = F ( x, Y, Y , Y )dx = F ( x, y + eη, y0 + eη 0 , y00 + eη 00 )dx
x1 x1
and Z x2
dI ∂F 0 ∂F 00 ∂F
0= = η +η 0 +η dx. (4)
de e=0 x1 ∂y ∂y ∂y00
Consider the second and third terms in this integral. Applying integration by
parts to the second term gives
∂F x2
Z x2 Z x2
0 ∂F d ∂F
η 0 dx = η 0 − η dx
x1 ∂y ∂y x1 x1 dx ∂y0
Z x2
d ∂F
= − η dx.
x1 dx ∂y0
4
Applying integration by parts twice to the third term gives
Z x2 x2 Z x
0 d
2
00 ∂F 0 ∂F ∂F
η dx = η 00 − η dx
x1 ∂y00 ∂y x1 x1 dx ∂y00
x2 Z x
d2
d ∂F 2 ∂F
= − η + η 2 dx
dx ∂y00 x1 x1 dx ∂y00
d2
Z x2
∂F
= η 2 dx.
x1 dx ∂y00
Therefore Eq. 4 can be written as
d2
Z x2
∂F d ∂F ∂F
0= η − + 2 dx.
x1 ∂y dx ∂y0 dx ∂y00
For this to be true for all possible variations η ( x ), it follows that
d2 ∂F d ∂F ∂F
2 00
− 0
+ = 0.
dx ∂y dx ∂y ∂y
(y00 )2
(b) For the case of F ( x, y, y0 , y00 ) = 2 , the equation from part (a) can be used to
obtain
d2 00
y =0
dx2
and thus
d4 y
= 0.
dx4
By integrating four times, it can be seen that the general solution is a cubic,
y ( x ) = a3 x 3 + a2 x 2 + a1 x + a0 .
To satisfy the boundary conditions y(0) = 1 and y0 (0) = 0 it can be seen that
a1 = 0 and a0 = 1. The condition y(1) = −1 gives
−1 = a3 + a2 + 1
and the condition y0 (1) = 0 gives
0 = 3a3 + 2a2 ,
and hence a3 = −4 and a2 = −6, so the solution is
y( x ) = 1 − 6x2 + 4x3 .
It can be seen that
y0 ( x ) = −12x + 12x2 , y00 ( x ) = −12 + 24x
5
1 y( x ) = 1 − 6x2 + 4x3
y∗ ( x ) = cos πx
0.5
-0.5
-1
0 0.2 0.4 0.6 0.8 1
x
Figure 2: Graph of the two functions y and y∗ considered in question 4.
and hence
1 1
Z
I [y] = (−12 + 24x )2 dx
2 0
144 1
Z
= (1 − 4x + 4x2 )dx
2 0
4
= 72 1 − 2 +
3
72
= = 24.
3
(c) If y∗ ( x ) = cos πx then y0∗ ( x ) = −π sin πx and hence
y∗ (0) = cos 0 = 1, y∗ (0) cos π = −1,
y0∗ (0) = −π sin 0 = 0, y0∗ (0) = −π sin π = 0
so the function satisfies the given boundary conditions. It can be seen that
6
y00∗ ( x ) = −π 2 cos πx and hence
1 1 00 2
Z
I [y∗ ] = (y ) dx
2 0 ∗
1 1 4
Z
= π cos2 πx dx
2 0
π4 1 1
Z
= (1 + cos 2πx ) dx
2 0 2
π4 sin 2πx 1
= x+
4 2π 0
π 4
=
4
= 24.35227 . . . .
Figure 2 shows plots of y and y∗ . It can be seen that the curves are similar, and
thus it should be expected that I [y∗ ] is close to I [y] in value. It can also be seen
that I [y∗ ] > I [y] as would be expected since y is an extremal path, and in this
case a minimal path.
5. Assume that the curve is between the two points ( x, y) = (± a, 0), and that its shape
is given by y( x ), so that y(± a) = 0. The problem involves maximizing
Z a q
I [y] = 2πy 1 + y02 dx
−a
subject to the constraint Z a q
J [y] = 1 + y02 dx = l.
−a
This can be carried out by introducing a Lagrange multiplier λ and considering
K [y, λ] = I [y] + λ( J [y] − l )
Z a q
= −λl + (2πy + λ) 1 + y02 dx
Z a −a
= F (λ, x, y, y0 )dx
−a
where F (λ, x, y, y0 ) = (2πy + λ) 1 + y02 . Since there is no explicit x dependence in
p
F, the Beltrami identity can be used: if C is a constant then
∂F
C = F − y0
∂y0
(2πy + λ)y02
q
= (2πy + λ) 1 + y02 − p
1 + y 02
2πy + λ)
= p .
1 + y 02
7
Hence
C2 (1 + y02 ) = (2πy + λ)2
which can be rearranged to give
dy
q
C = (2πy + λ)2 − C2 .
dx
Then
dx dy
Z Z
= p
C (2πy + λ)2 − C2
and by making the substitution 2πy + λ = C cosh u, so that 2πdy = C sinh u du,
x − x0 1 C sinh u 1 u
Z Z
= du = du =
C 2π C sinh u 2π 2π
and hence
−λ + C cosh 2π (xC−x0 )
y( x ) =
2π
where x0 is a constant. From the conditions y(± a) = 0, it follows that x0 = 0 and
the solution is even. Furthermore
−λ + C cosh 2πa
C
0 = y( a) =
2π
and hence
2πa
λ = C cosh .
C
To determine C, the constraint can be considered, which gives
Z a q Z a r
2πx
l = 1 + y02 dx = 1 + sinh2 dx
−a −a C
2πx a
Z a
2πx C C 2πa
= cosh dx = sinh = sinh .
−a C 2π C −a π C
Writing α = 2πa/C, this can written as
l sinh α
= .
2a α
While it is not possible to write down an analytic expression for α there will be a
solution to this equation for l/2a ≥ 1, which should be expected, since the straight-
line distance between the two endpoints is 2a. Once α is determined, then an explicit
value for C can be determined. Hence the solution is
C 2πx 2πa
y( x ) = cosh − cosh .
2π C C
where C is a function of l and a.
8
6. (a) The kinetic energy and potential energy are given by
ma2 θ̇ 2
K= , V = −mga cos θ
2
respectively, where g is the gravitational acceleration. Hence the Lagrangian is
ma2 θ̇ 2
L(t, θ, θ̇ ) = K − V = + mga cos θ.
2
The Euler–Lagrange equation therefore gives
d ∂L ∂L d 2
0= − = ma θ̇ − (−mga sin θ ) = ma2 θ̈ + mga sin θ
dt ∂θ̇ ∂θ dt
and hence
aθ̈ + g sin θ = 0.
(b) If θ is small then sin θ ≈ θ and the equation becomes
aθ̈ + gθ = 0.
This is a linear second-order differential equation with a solution
θ (t) = A cos λt + B sin λt
p
where λ = g/a. Hence the frequency of oscillations is given by
r
λ 1 g
= .
2π 2π a
(c) For the case where L has no explicit t dependence, the Beltrami identity is
∂L
L − θ̇ =C
∂θ̇
for some constant C. In this case
ma2 θ̇ 2
2
+ mga cos θ − θ̇ ma θ̇ = C
2
and hence
ma2 θ̇ 2
− mga cos θ = −C.
2
This is equivalent to
K + V = const.
expressing conservation of energy.
