Embed
Email

Analytical Methods Week 2[390]

Document Sample
Analytical Methods Week 2[390]
Analytical Methods: Solutions 2

1. ut + uux = 0.

We scale s = εa t, z = εb x and v = εc u. The equation becomes

εa−c vs + εb−2c vvz = 0,

so we have a balance if c = b − a. Then since the quantities

t−b/a x t−c/a u = t(a−b)/a u

are invariant, we can pose a solution (with b = ma)

u = t(b−a)/a f (ξ) = tm−1 f (ξ) ξ = t−b/a x = t−m x.

The resultant ODE is

[f (ξ) − mξ]f ′ (ξ) + (m − 1)f (ξ) = 0.

If we are given the inital condition

x + (x2 − 1)1/2

u(x, 1) =

2

then that fixes the function

ξ + (ξ 2 − 1)1/2

f (ξ) =

2

and we determine m from the ODE: m = 1/2. Then our solution is

xt−1/2 + (x2 t−1 − 1)1/2 x/t + [(x/t)2 − t−1 ]1/2

u = t−1/2 = .

2 2



Using the method of characteristics, the characteristic curves are given by

dx

=u t=r+1 x = ur + x0 .

dt

with implicit solution

u = F (x + u − ut).

The initial condition

x + (x2 − 1)1/2 x + (x2 − 1)1/2

u(x, 1) = gives F (x) =

2 2

with implicit solution

2u = (x + u − ut) + ((x + u − ut)2 − 1)1/2

which rearranges to



2 x±x2 − t

−4tu + 4xu − 1 = 0 u= .

2t

Taking the positive root in order to match the initial condition, we obtain

the same solution as before:



x + x2 − t x/t + [(x/t)2 − t−1 ]1/2

u= = .

2t 2

2. εx3 + x2 + (2 − ε)x + 1 = 0.

Scale x ∼ δ and note that the εx term is always smaller than the 2x term.

Further, if x2 ≫ 1 then x2 ≫ 2x; if x2 ≪ 1 then 2x ≪ 1 so the 2x term

only dominates if two other terms balance. We are now comparing the

following terms:



[A] εδ 3 [B] δ 2 [C] 1.



For small δ, [C] dominates. [B] catches up first at δ = 1. Then [A]

catches up with [B] when εδ 3 = δ 2 , δ = ε−1 . The distinguished scalings

are x ∼ 1 and x ∼ ε−1 .

We solve first for the regular root(s): x = x0 + εx1 + ε2 x2 + · · ·

Substiting in gives



x20 + 2x0 + 1 = 0

εx3

0 + 2εx0 x1 + 2εx1 − εx0 = 0

2 2

3ε x0 x1 + 2ε2 x0 x2 + 2ε2 x2

1 + 2ε2 x2 − ε2 x1 = 0



The leading order term gives x0 = −1. At order ε we have



−1 − 2x1 + 2x1 + 1 = 0



which is automatically satisfied. At order ε2 we obtain



3x1 − 2x2 + 2x2 + 2x2 − x1 = 0

1 2x1 (1 + x1 ) = 0



which is satisfied either by x1 = −1 or x1 = 0. In fact x = −1 is an exact

root:

εx3 + x2 + (2 − ε)x + 1 = (x + 1)(εx2 + (1 − ε)x + 1)

so no further terms are available for the other root.

Looking for the singular root, we pose: x = ε−1 x−1 + x0 + · · ·

and substituting gives



ε−2 x3

−1 + ε−2 x2

−1 = 0

3ε−1 x2 x0

−1 + 2ε−1 x−1 x0 + 2ε−1 x−1 = 0



At order ε−2 we obtain x−1 = −1 (recall the leading x term is strictly

order 1 so 0 is not a valid solution); at order ε−1 we get x0 = 2. Thus the

three roots are x = −1 − 2ε + O(ε2 ); x = −1; and x = −ε−1 + 2 + O(ε).

3. εx4 − x2 − x + 2 = 0.

First we look for scalings. One of x2 or 2 is always at least as large as x

so we only consider



[A] εδ 4 [B] δ 2 [C] 1



At small δ [C] is largest, and it is equalled first by [B] when δ = 1. For

larger δ, [A] reaches [B] when εδ 4 = δ 2 i.e. δ = ε−1/2 .

Look at the regular root(s) first (and assume a regular expansion):



x = x0 + εx1 + · · ·

gives

− x2

0 − x0 + 2 = 0

εx4

0 − 2εx0 x1 − εx1 = 0

At order 1 we have



x2 + x0 − 2 = 0

0 (x0 + 2)(x0 − 1) = 0 x0 = 1 or x0 = −2.



If x0 = 1 the next order gives 1 − 3x1 = 0 so x ∼ 1 + ε/3.

If x0 = −2 the next order gives 16 + 3x1 = 0 so x ∼ −2 − 16ε/3.

Now we move on to the singular roots and the scaling suggests an expan-

sion in ε1/2 :

x = ε−1/2 x0 + x1 + ε1/2 x2 + · · ·

Substituting gives



ε−1 x40 − ε−1 x20 = 0

4ε−1/2 x3 x1

0 − 2ε−1/2 x0 x1 − ε−1/2 x0 = 0



At leading order we have



x4 − x2 = 0

0 0 x2 (x0 + 1)(x0 − 1) = 0

0 x0 = 1 or x0 = −1.



If x0 = 1 then the next order gives 2x1 − 1 = 0 so x ∼ ε−1/2 + 1/2.

If x0 = −1 then the next order gives −2x1 + 1 = 0 so x ∼ −ε−1/2 + 1/2.



4. xe−x = ε. Define f (x) = xe−x . This function is positive for x > 0; zero

at both x = 0 and x → ∞; and f (1) = e−1 ≫ ε so we expect two roots,

one in 0 1 depends more strongly on the exponential than on the

x term, so we try a logarithmic scaling. Let us try the values of f (x) when

x = x0 ln (1/ε):



• If x0 = 1 then f (x) = ε ln (1/ε) ≫ ε.

• If x0 = 2 then f (x) = 2ε2 ln (1/ε) ≪ ε.

These two points bracket the root so we know the scaling is correct. We

begin our expansion

x = x0 ln (1/ε) + δ1 x1 + · · ·

and substitute it in, using L1 = ln (1/ε), to obtain:

ε = εx0 (x0 L1 + δ1 x1 + · · · ) exp [δ1 x1 + · · · ] = εx0 x0 L1 exp [δ1 x1 ] + · · ·

To match the powers of ε we need x0 = 0; then to make the logarithm

terms work we need

δ1 x1 = −L2 δ1 = L2 , x1 = −1.

in which we have used L2 = ln L1 .

The beginning of the expansion is

x ∼ ln (1/ε) − ln (ln (1/ε)) + · · ·

t x+c(t−t′ )

1

5. u = F (x′ , t′ ) dx′ dt′ .

2c 0 x−c(t−t′ )

The easy part is the derivatives wrt x: for ease, split the integral in two:

t x+c(t−t′ ) t x−c(t−t′ )

1 ′ ′ ′ 1′

u= F (x , t ) dx dt − F (x′ , t′ ) dx′ dt′

2c 0 0 2c 0 0



Then

t t

∂u 1

= F (x + c(t − t′ ), t′ ) dt′ − F (x − c(t − t′ ), t′ ) dt′

∂x 2c 0 0

2 t t

∂ u 1

= Fx (x + c(t − t′ ), t′ ) dt′ − Fx (x − c(t − t′ ), t′ ) dt′

∂x2 2c 0 0

Partial derivatives wrt t must be taken with more care, as both inner and

outer integrals have limits which depend on t:

x+c(t−t) x−c(t−t)

∂u 1 1

= F (x′ , t) dx′ − F (x′ , t) dx′

∂t 2c 0 2c 0

t t

1 1

+ cF (x + c(t − t′ ), t′ ) dt′ − (−c)F (x − c(t − t′ ), t′ ) dt′

2c 0 2c 0



The first two terms here cancel and we are left with

t

∂u 1

= F (x + c(t − t′ ), t′ ) + F (x − c(t − t′ ), t′ ) dt′

∂t 2 0

2

∂ u 1

= (F (x + c(t − t), t) + F (x − c(t − t), t))

∂t2 2

1 t

+ cFx (x + c(t − t′ ), t′ ) − cFx (x − c(t − t′ ), t′ ) dt′

2 0

Putting these together gives the required result:

∂2u ∂2u

2

− c2 2 = F (x, t).

∂t ∂x

6. ∂ 2 u/∂t2 − e2x ∂ 2 u/∂x2 . Here c(x) = ex so the characteristics are given by

dt

= ±ex t = ±ex + α

dx

Through the point x = 0 , t = 1, the two equations become

t = ex and t = 2 − ex .

∂2f ∂2f ∂f

7. − − ε cos xf = x with f (x, 0) = (x, 0) = 0.

∂t2 ∂x2 ∂t

Set f = f0 + εf1 + · · · .

∂ 2 f0 ∂ 2 f0

− = x

∂t2 ∂x2

2 2

∂ f1 ∂ f1

ε 2 − ε 2 − ε cos x f0 = 0

∂t ∂x

It is useful to note that the solution of the inhomogeneous wave equation

with c = 1:

∂2f ∂2f

− = F (x, t)

∂t2 ∂x2

is ′

1 t x+t−t

f (x, t) = p(x + t) + q(x − t) + F (x′ , t′ ) dx′ dt′

2 0 x−t+t′

and applying the boundary conditions f (x, 0) = ∂f /∂t(x, 0) = 0 to this

form gives



1 t x+t−t

f (x, t) = F (x′ , t′ ) dx′ dt′ .

2 0 x−t+t′

Order 1. The inhomogeneous wave equation

∂ 2 f0 ∂ 2 f0 ∂f0

− =x f0 (x, 0) = (x, 0) = 0

∂t2 ∂x2 ∂t

has the solution

t x+t−t′ t x+t−t′

1 ′ ′ 1 ′ 1 ′2

f0 (x, t) = x dx dt = x dt′

2 0 x−t+t′ 2 0 2 x−t+t′

t

t

= 1

x(t − t′ ) dt′ = x(tt′ − 2 t′2 ) 0

= 2 xt2 .

1

0



f0 (x, t) = 1 xt2 .

2



Order ε. We are solving

∂ 2 f1 ∂ 2 f1

− = f0 cos x = 2 x cos xt2

1

∂t2 ∂x2

with the same zero boundary conditions as before. The solution is

t x+t−t′

1

f1 (x, t) = x′ cos x′ t′2 dx′ dt′

4 0 x−t+t′



which becomes, after tedious but straightforward work,



f1 (x, t) = 2 t2 (x cos x − 2 sin x) − x cos x + 4 sin x

1



1

+ 1 (x + t) cos (x + t) + 2 (x − t) cos (x − t) − 2 sin (x + t) − 2 sin (x − t).

2


Other docs by DHarperii
RCGM_Minutes_2_080701.doc
Views: 14  |  Downloads: 0
Redeployment Policy summary
Views: 5  |  Downloads: 0
MSc Crime Science
Views: 11  |  Downloads: 0
CURRICULUM%20VITAE%202008
Views: 58  |  Downloads: 0
Mathematical Methods 4 Fourier Transforms[450]
Views: 12  |  Downloads: 0
Matrix-vector systems
Views: 3  |  Downloads: 1
Operator form of the Laplacian
Views: 7  |  Downloads: 0
Applied Mathematics Particle dynamics
Views: 6  |  Downloads: 0
Differentiation Part 2 (in PDF format)
Views: 19  |  Downloads: 1
By registering with docstoc.com you agree to our
privacy policy

You are almost ready to download!

You are almost ready to download!