0% found this document useful (0 votes)
170 views

Solution To Homework #6: Math220/Cme303: Pdes of Applied Mathematics

This document provides solutions to three problems from homework on partial differential equations (PDEs) of applied mathematics. Problem 1 involves solving two PDEs using the method of characteristics. Problem 2 shows that a proposed function is an entropy solution of the Burgers' equation. Problem 3 computes the entropy solution for the Burgers' equation with piecewise initial data and explains its behavior at different times, suggesting to plot the solution.

Uploaded by

Kash
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
170 views

Solution To Homework #6: Math220/Cme303: Pdes of Applied Mathematics

This document provides solutions to three problems from homework on partial differential equations (PDEs) of applied mathematics. Problem 1 involves solving two PDEs using the method of characteristics. Problem 2 shows that a proposed function is an entropy solution of the Burgers' equation. Problem 3 computes the entropy solution for the Burgers' equation with piecewise initial data and explains its behavior at different times, suggesting to plot the solution.

Uploaded by

Kash
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

MATH220/CME303: PDEs of Applied Mathematics

Solution to Homework #6

1. Problem 1: Solve using characteristics


(a)
∂u ∂u
x1 + x2 = 2u, u(x1 , 1) = g(x1 )
∂x1 ∂x2
(b)
∂u ∂u x1
u + = 1, u(x1 , x1 ) =
∂x1 ∂x2 2

Solution:
(a) For the Cauchy problem
x1 ux1 + x2 ux2 = 2u, u(x1 , 1) = g(x1 )
the surface Γ ⊂ R2 is defined by Γ = {(y1 , y2 ) : y2 = 1}. The characteristic equations become
ξ1′ (s) = ξ1 (s), ξ1 (0) = y1 ,
ξ2′ (s) = ξ2 (s), ξ2 (0) = 1,
U ′ (s) = 2U (s), U (0) = g(y1 )

This leads to the characteristics


ξ1 (s) = y1 es ,
ξ2 (s) = es ,
U (s) = g(y1 ) e2s

Consider the values for s such that a characteristic crosses the point (x1 , x2 ) for x2 ≥ 1. We
obtain
( )
x1
ξ1 (s) = x1 ⇒ s = log ,
y1
x1
ξ2 (s) = x2 ⇒ x2 =
y1
Therefore y1 = x1 /x2 and s = log x2 . Consequently
( )
x1
u(x1 , x2 ) = U (s) = g x22
x2
The solution is undefined for x2 ≤ 0. Note that this is a consequence of the fact that ξ2 (s) > 0
and therefore the characteristics never cross the x1 -axis. The solution only can be extended over
{(x1 , x2 ) : x2 > 0}.

1
(b) For the Cauchy problem
1
uux1 + ux2 = 1, u(x1 , x1 ) = x1
2
the surface Γ ⊂ R2 is defined by Γ = {(y1 , y2 ) : y1 = y2 }. The characteristic equations become

ξ1′ (s) = U (s), ξ1 (0) = y1 ,


ξ2′ (s) = 1, ξ2 (0) = y1 ,
1
U ′ (s) = 1, U (0) = y1
2
The equation for ξ1 can be rewritten as
1
ξ1′′ (s) = 1, ξ1 (0) = y1 , ξ1′ (0) = y1
2
We obtain
1 2 1
ξ1 (s) = s + y1 s + y1 ,
2 2
ξ2 (s) = s + y1 ,
1
U (s) = s + y1
2

Consider the values s for which a characteristic crosses the point (x1 , x2 ). We have

ξ2 (s) = x2 ⇒ s = x2 − y1 ,
1 1
ξ1 (s) = x1 ⇒ x1 = (x2 − y1 )2 + y1 (x2 − y1 ) + y1
2 2
From the second equation we deduce
( )
1 2 1 1 1 1 1 x1 − 12 x22
x1 = x − x2 y1 + y12 + y1 x2 − y12 + y1 = 1 − x2 y1 + x22 ⇒ y1 =
2 2 2 2 2 2 2 1 − 12 x2

We see that for x2 = 2, the value of y1 is undefined. This is due to the fact that, for x2 = 2 we
have s = 2 − y1 and

ξ1′ (2 − y1 ) = 2,
ξ2′ (2 − y1 ) = 2

from which we deduce that the corresponding characteristic curve has to be tangent to Γ. In
other words, the characteristic that crosses (x1 , 2) is precisely a characteristic that is tangent to
Γ. This is the main problem of the solution over these points. We conclude that, for x2 ̸= 2

1 1 x1 − 12 x22 x22 + 2x1 − 4x2


u(x1 , x2 ) = U (s) = x2 − y1 + y1 = x2 − =
2 2 1 − 21 x2 2(x2 − 2)

2
2. Problem 2: Show that
{ ( √ )
− 23 t + 3x + t2 if 4x + t2 > 0
u(t, x) =
0 otherwise
is an entropy solution of the Burgers’ equation
ut + uux = 0

Solution: Let F (u) = 12 u2 . We first check that the proposed function u defined as
{ ( √ )
− 23 t + 3x + t2 , 4x + t2 > 0,
u(t, x) =
0, o.w.
is a pointwise solution of Burguers’ equation away from its discontinuities. This is trivially true for
4x + t2 < 0. For 4x + t2 > 0 we have
1
ux (t, x) = −√ ,
3x + t2
2 2 t
ut (t, x) = − − √
3 3 3x + t2
2 2
= − + tux (t, x),
3 3
2 2√
u(t, x)ux (t, x) = − tux (t, x) − 3x + t2 ux (t, x)
3 3
Therefore
2 2 2 2√
ut + uux = − + tux − tux (t, x) − 3x + t2 ux (t, x) = 0
3 3 3 3
We need to verify Rankine-Hugoniot’s condition for the discontinuity points of u. It satisfies 4x(t)+t2 =
0, from which we deduce ẋ(t) = − 12 t. Remark that
u+ (t) = lim u(t, x)
x↓x(t)
( √ )
2 3 2
= − t+ − t +t 2
3 4
= −t,
u− (t) = lim u(t, x)
x↑x(t)
= 0
so that Rankine-Hugoniot’s condition in our case becomes
F (u+ (t)) − F (u− (t)) t −0
1 2
1
ẋ = = 2 =− t
u+ (t) − u− (t) −t − 0 2

So u is indeed a solution of Burgers’ equation. To prove it is an entropy solution, it suffices to verify


that u+ < ul , which is clear from our previous calculations.

3
3. Problem 3: Compute explicitly the entropy solution for the Burger’s equation with the initial data


 1 if x < −1

0 if −1 < x < 0
g(x) =

 2 if 0 < x < 1


0 if 1 < x

and explain what happens at various times t > 0. Plot the graph of u(t, x) at various times when it
behaves differently.

Solution: We denote as usual F (u) = 12 u2 . Qualitatively the leftmost wave will propagate as a shock-
wave. We denote this discontinuity as x1 . The second wave will develop a rarefaction wave on its left
and a shockwave on its right. We denote the rarefaction front as xf and its shockwave as x2 . We
construct each stage of the solution analyzing the point at each one of them begins to evolve.

Stage 1 (begins at t = 0): The Rankine-Hugoniot condition for the shockwaves is


F (1) − F (0)
ẋ1 =
1−0
1
= ,
2
F (2) − F (0)
ẋ2 =
2−0
= 1

The rarefaction wave is of the form x/t and therefore xr (t) = 2t. It is not difficult to see that


 1, x < −1 + 21 t,




0, −1 + 2 t < x < 0,
1

x
u1 (t, x) = t , 0 < x < 2t,



 2, 2t < x < 1 + t,


0, 1 + t < x

This expression shows that the left shockwave will hit the rarefaction wave at t such that −1 + 21 t = 0,
i.e. t = 2 whereas the rarefaction front will hit the right shockwave at t such that 2t = 1 + t, i.e. t = 1.
Therefore, the next step is to analyze the evolution of the solution for t > 1.

Stage 2 (begins at t = 1): In this case the rarefaction front hits the right shockwave. The result is
a rarefaction wave with shockwave x2 of height α2 = x2 /t. The Rankine-Hugoniot condition for this
case is
F (α2 ) − F (0)
ẋ2 =
α2 − 0
1
= α2
2

4
1
= x2
2

Additionally, x2 (1) = 2. This implies √


x2 (t) = 2 t

Therefore, the solution at the second stage is




 1, x < −1 + 21 t,

0, −1 + 21 t < x < 0,
u2 (t, x) = x √


t,

0 < x < 2 t,

0, 2 t<x

We remark that the left shockwave will hit the rarefaction wave at t = 2, which is the next stage of
the solution.

Stage 3 (begins at t = 2): In this case the left shockwave hits the rarefaction wave. The left
shockwave will consequently change its propagation speed. The Rankine-Hugoniot condition is
F (1) − F (x1 )
ẋ1 =
1 − x1
1
2 − 1 2
2t2 x1
=
1 − 1t x1
( )
1 1
= 1 + x1
2 t

Remark that ∫ ( )1/2 ( )1/2 ( )1/2


t
ds t ∫t t ∫t 2
, e− 2
ds ds
= log ⇒ e 2 2s = 2s =
2 2s 2 2 t
and therefore ( )1/2 ∫ t ( )1/2 ∫ t
t 1 2 ds
x1 (t) = ds = t1/2 1/2
2 2 2 s 2 2s

The solution is then √


x1 (t) = t − 2t
and the solution in this stage is consequently
 √

1, x <√ t − 2t,

u3 (t, x) = xt , t − 2t < x < 2 t,

 √
0, 2 t < x
√ √ √
Finally, note that both shockwaves will merge at t such that t − 2t = 2 t, i.e. t = (2 + 2)2 . This
is the last stage of the solution.

5

Stage 4 (begins t = (2 + 2)2 ): In this case, there is only a single shockwave that propagates. The
Rankine-Hugoniot condition is simply
F (1) − F (0)
ẋ1 =
1−0
1
=
2

The last stage of the solution is


{ √
1, x < 12 (2 + 2)2 + 12 t,
u4 (t, x) = √
0, 21 (2 + 2)2 + 12 t < x

In summary, the solution is (see attached diagram at the end for a plot of the different stages)


 u1 (t, x), 0 < t < 1,

u (t, x), 1 < t < 2,
u(t, x) =
2


 u3 (t, x), 2 < t < (2 + 2)2 ,

 √ 2
u4 (t, x), t < (2 + 2)

4. Problem 4: Consider the viscous Burgers’ equation

ut + uux = εuxx

where ε > 0 is a constant.


(a) Find all solutions of this equation that have the form u(t, x) = V (x − ct), where the function V (y)
has the limits
lim V (y) = 1, lim V (y) = 0
y→−∞ y→∞

In particular, find all c for which such solution exists.


(b) Let ε → 0, and show that the limit u(t, x) from the previous item is an entropy solution of the
Burgers’ equation.

Solution:
(a) We search a solution of the form u(t, x) = V (x − ct). Remark that ut (t, x) = −cV ′ (x − ct),
ux (t, x) = V ′ (x − ct) and uxx (t, x) = V ′′ (x − ct). Let z = x − ct. Then the PDE becomes

−cV ′ (z) + V V ′ (z) = εV ′′ (z)

which is simply an ODE. We assume that V is regular at infinity and therefore that V ′ (z) → 0
as |z| → ∞. In this case we can integrate from z to ∞ to obtain
1
cV (z) − V 2 (z) = −εV ′ (z)
2

6
Using the regularity of V ′ at infinity and the condition V (z) → 1 as z → −∞ we obtain c − 12 = 0.
Therefore, the only admissible speed of propagation of a stationary wave satisfying the imposed
conditions is c = 21 . Finally, we obtain the nonlinear ODE
1 1
V ′ (z) = V (z)2 − V (z)
2ε 2ε
which is the Riccati equation (this is also a form of the Bernoulli equation). Let W be such that
V (z) = −2εW ′ (z)W (z)−1 . Then
( ′ )2
W ′′ (z) W (z) W ′′ (z) 1
V ′ (z) = −2ε + 2ε = −2ε + V (z)2
W (z) W (z) W (z) 2ε
from which we deduce
W ′′ (z) 1 1 W ′ (z)
−2ε = V ′ (z) − V (z)2 = − V (z) =
W (z) 2ε 2ε W (z)
Therefore
1 ′
W ′′ (z) + W (z) = 0

and, for constants A, B ∈ R, we deduce that
W (z) = A e− 2ε z + B
1

so that
W ′ (z) A 1
V (z) = −2ε = 1 = 1
W (z) A + B e 2ε z 1 + C e 2ε z
with C = A−1 B. Remark that V (z) → 0 as z → ∞ and V (z) → 1 as z → −∞ for any choice of
constant C ̸= 0. If C < 0, V has a singularity at z = −2ε log(|C|) and therefore is not a solution
of the viscous Burgers’ equation. Consequently for C > 0 the function V (z) defined is the desired
solution. Note that, for these cases, C acts as a translation parameter. In fact, let z0 = log C.
Then
1
V (z) = 1
1 + e 2ε z+z0
for any z0 ∈ R.
(b) Define
1
Vε (z) = 1
1 + C e 2ε z
We prove that Vε (z) → Θ(z) as ε ↓ 0, where Θ(z) = χ(−∞,0] . Fix z > 0 and let ε be small enough
so that C −1 e− 2ε z < 1/2. Then
1

1 1
= C −1 e− 2ε z ≤ 2C −1 e− 2ε z → 0
1 1
|Vε (z) − Θ(z)| ≤
C −1 e− 2ε z
1 1
−1 + C e 2ε z 1−
1
as ε → 0. For z < 0, let ε be such that C e 2ε z
< 1/2

C e 2ε
1
z 1
C e 2ε z
1
|Vε (z) − Θ(z)| = ≤ ≤ 2C e 2ε z → 0
1 + C e 2ε z 1 − C e 2ε
1 1
z

7
as ε ↓ 0. Consequently Vε → Θ for z ̸= 0 pointwise. Therefore, define uε (t, x) = Vε (x − 12 t). We
conclude ( )
1
lim uε (t, x) = u(t, x) = Θ x − t
ε↓0 2
We now prove that this is indeed an entropy solution for Burgers’ equation with initial condition
u(0, x) = Θ(x). The function u(t, x) satisfies the initial conditions and it is a pointwise solution
for x < 12 t and x > 12 t. Additionally, the discontinuity point propagates at speed c = 12 , which
satisfies Rankine-Hugoniot’s condition. Finally, it is not difficult to see that the characteristics
satisfy

ξt (s) = s,
{
y + s, y < 0,
ξx (s) =
y, y>0

Therefore, for any (t, x), if x − 12 t < 0, we choose the characteristic ξx (s) = x − (t − s). Clearly
ξx (s)− 21 ξt (s) = x− 21 t− 12 (t−s) < x− 12 t < 0, so this characteristic does not cross the discontinuity
curve nor another characteristic. If x − 12 t > 0, then we simply choose ξx (s) = x. Therefore, u is
an entropy solution of Burgers’ equation, as desired.

You might also like