Asymptotic Methods: Example Sheet 3

Send corrections to David Stuart [email protected]

One more Steepest-Descent Example.

(1)
1. The Hankel function H ν (x) is defined by
Copyright © 2017 University of Cambridge. Not to be quoted or reproduced without permission.

1 i∞ − π i (νt − x sin t)
Z
H (1)ν (x) = − e dt ,
π − i∞
where the path of integration satisfies − π ≤ Re(t) < 0 . Use the method of
steepest descent to show that, as ν −→ + ∞ , one has
 
(1) ν 1
− 1 iπ
H ν ∼ 2 2 πν tan α 2 eiν(tan α − α) e− 4 ,
cos α
where 0 < α < π/2 . (This is the limit in which ν −→ ∞ , with x/ν a positive
constant greater than 1.) Find an equation for the path along which you integrate,
sketch it and justify that it is indeed the path of steepest descent.
Liouville-Green and WKB approximation.

2. (a) For the equation ε2 y ′′ = q(x)y consider solutions given formally1 as

 
1 2
y = exp S0 + S1 + εS2 + ε S2 + · · · .
ε
for small ǫ. Show that the functions {Sn }∞
n=0 are determined iteratively by

(S0′ )2 = q , 2S0′ S1′ + S0′′ = 0 ,

and
n−1
X
2S0′ Sn′ + ′′
Sn−1 + Sj′ Sn−j
′
= 0.
j=1

Find expressions for S0 , S1 and S2 in terms of q (taking care over the sign choice in
S0 ).
(b) Now try to apply the above solution method to the Airy equation y ′′ = xy , as
follows: let2 ε = 1, q(x) = x and find a relation which determines Sn (x) in terms of
1
This means they should satisfy the equation to each order in ǫ, i.e. as formal power series in ǫ.
2
There are two ways you can think of the procedure of introducing the small parameter ǫ and then putting
it equal to one. Either ǫ arises as a small parameter by first rescaling x = µX with µ → +∞ as in lectures,
deriving the asymptotic relations for the {Sj (X)}, and noticing that they are in fact scale invariant so one
can rescale back. Alternatively, just think of it as a formal device to obtain the system for the {Sj } which is
formally completely equivalent to the original equation - in this case the asymptotic content arises in checking
that the solutions of the system form an asymptotic sequence, as you are asked to do here.
 n−1
the functions {Sj (x)}j=1 . Show inductively that this defines a sequence of functions
3
of the form Sj (x) = cj x 2 (1−j) for j ≥ 2 and some constant cj . Deduce that Ai might
reasonably be expected to admit an asymptotic expansion of the form3
  
− 41 2 3 5 −3 1
Ai(x) ∼ Ax exp − x 2 1 − x 2 + O( 3 ) , (x −→ ∞) .
3 48 x

3. Show that, for large positive x , the equation

w′′ − x3 w′ + x−2 w = 0
has two independent solutions with asymptotic behaviour
 4
x
Copyright © 2017 University of Cambridge. Not to be quoted or reproduced without permission.

−4

−3


w ∼ 1+O x and w ∼ x exp 1 + O x−4 .
4

4. (a) Consider the differential equation

z w′′ − z w′ − w = 0 (1)
in a neighbourhood of the point at infinity, z = ∞ . Classify this point.
(b) Show that one solution of (1) is w(z) = zez , and hence find an expression for a
second independent solution as an integral.
(c) Find µ(z) and q(z) such that W (z) = µ(z)−1 w(z) satisfies an equation W ′′ −
q(z) W = 0 . Write down expressions for the two independent Liouville–Green so-
pof q(z) . Calculate the first two terms of the expan-
lutions of this equation in terms
sion in inverse powers of z of q(z) about z = ∞ , and hence determine from the
Liouville–Green solutions the leading–order asymptotic form for w as z −→ ∞ .
Compare these with the two exact solutions obtained previously.
5. The wavefunction ψ obeys the equation
d2 ψ s(s + 1)
− 2 − ψ = Eψ ,
dx cosh2 x
with s > 0 . Use the WKB approximation to find approximate values for the bound-
state energies (eigenvalues E for which ψ → 0 as |x| → ∞)
1
q
q

−E = s s+1 − n − .
2
What are the allowed values of n? Compare with the exact bound state energies
E = − (s − n)2 , where n = 0 , 1 , 2 , . . . and n < s.
[Hint: You may find the substitution sinh x = z useful, and note that if a, b are
real, Z a √ 2
a − z2 √
2 + z2
dz = π( 1 + b−2 a2 − 1) ,
−a b
which can be derived via contour integration.]
3
You are not expected to prove that this procedure gives a complete asymptotic expansion. The integral
formulation for the Airy functions provides an efficient basis for a rigorous approach to their asymptotics - in
addition to what we will do in lectures see Section 7.6 in Hormander’s “Analysis of Linear Partial Differential
Operators I” or the appendix to “Complex Analysis” by Stein and Shakarchi for some proofs.

2
 6. Show that the large eigenvalues of the system
d2 w
x4 + x2 − λ 2 w ,


= w(− ∞) = w(∞) = 0 ,
dx2
are given by
  − 43 
1 1 3  23  1
λ ∼ 2 π Γ
3 3n + + O n3 ,
4 2
where n is a positive integer.
7. (∗) Quantum tunneling through a potential barrier in the stationary approach to scat-
Copyright © 2017 University of Cambridge. Not to be quoted or reproduced without permission.

tering theory is described by the time-independent Schrödinger equation

~2 d 2 ψ
− + V (x)ψ = Eψ
2m dx2
where V ≥ 0 and V → 0 faster than 1/|x| as |x| → ∞. Consider a particle incident
2 k2
from x = −∞ with an energy E = ~2m , and assume that E < V on an interval (a, b)
and E > V on (−∞, a) and (b, ∞). This means that for x → −∞ the wave function
should approach eikx +re−ikx while for x → +∞ the wave function should approach
teikx , where the constants r and t are known as the reflection and transmission coef-
ficients of the potential. (The term eikx corresponds to the incident wave, travelling
to the right from x = −∞ in the corresponding time-dependent description.)
Write down Liouville-Green approximations to the wave functions in the three re-
gions (a, b), (−∞, a) and (b, ∞), consistent with the set-up just described and indi-
cating clearly which are oscillatory and which are real exponential solutions. Show
using the connection formulae of the WKB approximation that the transmission prob-
ability |t|2 is approximately

2 bp
 Z 
exp − 2m(V (x) − E) dx ,
~ a
when the integral is large compared with ~.
[Note: For a careful treatment see Landau and Lifshitz, Quantum Mechanics, 3rd
ed., p.178, or Bender and Orszag p.524.]
Further examples of Stokes’ phenomenon and recap on asymptotics of integrals.

8. The analytic function f (z) is defined by

Z z
exp s2 ds .


f (z) =
i∞

Use the method of integration by parts to obtain an asymptotic expansion of the form
 
2

1 1 3 15
f (z) ∼ exp z + + + + ... ,
2z 4z 3 8z 5 16z 7
in the sector 0 ≤ arg z ≤ π , including an expression for the remainder after a finite
number of terms and a verification of the asymptotic property.

3
 To deal with the case − π ≤ arg z ≤ 0 , convert f (z) to the form
Z z
1
exp s2 ds − i π 2 ,


f (z) =
−i∞

and, without making

 detailed1calculations,
 obtain an asymptotic expansion valid in
this domain for f (z) + i π 2 . Find the Stokes lines for these asymptotic expan-
sions of f as |z| → ∞.
9. (∗) In discussion of the methods of stationary phase and steepest descent we have
considered the integral Z 1
exp z t3 dt


I(z) =
Copyright © 2017 University of Cambridge. Not to be quoted or reproduced without permission.

0
in the case that z = ix is purely imaginary. This question is concerned with the
general case z ∈ C.
(a) In the case that ℜz > 0 show that
Γ 34 ez X Γ n + 32

∞

1
I(z) + 1 ∼ 2

(†)
z3 3 n=0 Γ 3 z n+1

as z −→ + ∞ in the sector Arg z ∈ (− π2 + δ, π2 − δ) for small positive δ, and

obtain a similar relation for the sector Arg z ∈ ( π2 + δ, 3π
2
− δ).
(b) Generalize the application of the method of steepest descent in the notes to de-
velop the asymptotic expansion

iπ
ez X Γ(n + 32 ) 1 Γ 34 e 3
∞
I(z) ∼ + ,
Γ( 23 ) z n+1
1
3 n=0 z3
as |z| → ∞ in the sector Arg z ∈ (δ, π − δ). Find a similar relation for the sector
Arg z ∈ (π + δ, 2π − δ). Discuss the relation between the different expansions in
regions of overlap.
R[You may use4 Watson’s Lemma with complex parameter: Consider I(z) =
∞ α
0
t g(t)e−zt dt, ℜz > 0, where g is a locally integrable function, and
(i) α > −1 ;
(ii) g has an asymptotic expansion g(t) ∼ ∞ j
(t → 0+ ) ,;
P
j=0 aj t ,

(iii) there exist positive constants K, β such that |g(t)| ≤ Keβt for t ≥ 0 .
Then I(z) has an asymptotic expansion
∞
X Γ(α + j + 1)
I(z) ∼ aj ,
j=0
z α+j+1

as |z| → +∞ in the sector Arg z ∈ (− π2 + δ, π

2
− δ) for arbitrary small positive δ. ]

4
See also Chapter 6 of Ablowitz and Fokas for a further discussion of Watson’s Lemma in the complex
plane