0387986995Basic TheoryC
.pdf178 VI. BOUNDARY-VALUE PROBLEMS
Remark VI-7-7. If the reflection coefficient r(t) = 0 and if there is no eigenvalue for the problem
- d2y + u(x)y = Ay, y E L2(-oo, +oo),
2
then u(x) = 0 identically for -oo < x < +oo. In fact, f+(x,() is analytic in everywhere in the (-plane. Furthermore, l e-`<= f+(x, () -11 -+ 0 as - oo. Hence, f+(x,() = e'(=. This shows that u(x) = 0.
Remark VI-7-8. If u(x) 0 but u(x) = 0 for }xI _> M, where M is a positive number, then the reflection coefficient r(S) # 0. In fact, if r(C) = 0, then u(x) is analytic for -oo < x < +oo. Hence, u(x) = 0. This is a contradiction. This remark reveals that the problem posed at the beginning of §VI-5 is not as simple as it looks.
VI-8. Construction of a potential for given data
If scattering data {0, (71,... 7x), (cl,... ,cN)} are given, formula (VI.7.3) gives the corresponding reflectionless potential u(x) as it is shown in Examples VI-
7-5 and VI-7-6. However, in these two examples, we assumed implicitly that such a potential exist. Since the existence of u(x) has not been shown yet, it must be proved. To do this, for given data
7i>0, 72>0, ..., 7N>0 and c1>0, c2>0, ..., cN>0,
define g, (j = 1, 2,... , N) by (VI.7.2) and define f+ and u by (VI.7.1) and (VI.7.3).
Hereafter, the first thing that we have to do is to show that y = f+ is one of the
Jost solutions of Ly = (2y.
Observation VI-8-1. Let gi (t = 1, 2,... , N) be determined by (VI.7.2) and u(z) be defined by (VI.7.3). Then,
(VI.8.1) |
gi + 27r9t - ugt = 0 |
(=1,2,... , N). |
||
Proof |
|
|
|
|
Write (VI.7.2) in the form |
|
|
|
|
|
N |
|
|
|
(VI.8.2) |
E ae, (x)9, (x) = 1 |
(t = 1, 2,... , N) |
||
|
3=1 |
|
|
|
and differentiate both sides with reaspect to z. Then, |
|
|||
(VI.8.3) |
Eat, (x)9f (x) + 2t, |
|
9j(x) = 0 |
(t = 1, 2,... , N) |
|
}=1 |
|
|
|
and, hence, |
|
|
|
|
N |
N |
|
g, (x) = 0 |
(t = 1,2,... ,N). |
Eata(x)9j'(x) + 27t 11 - |
1 |
|||
1=1 |
i=1 7e+7) |
|
9. DIFF. EQS. SATISFIED BY REFLECTIONLESS POTENTIALS |
181 |
Observation VI-8-6. Note that f_ (x, ) = a(() f+(x, -{) = |
f+(x, ) for ( = |
|
C real. This means that |
0. Hence, r(C) = 0. Thus, we conclude that u(x) is |
reflectionless.
Remark VI-8-7. For the general r((), the potential u(x) can be constructed by solving the integral equation of Gel'fand-Levitan:
K(x, r) + F(x + r) + |
+00 |
|
+ F(x + r + s)K(x, s)ds = 0, |
||
|
|
0 |
where |
1J + |
|
F(x) = |
2cL` cle-2n,= |
|
|
R |
|
oo |
J=1 |
|
Find K(x, r) by this equation. Then, the potential is given by
u(x) 8x (x,0).
If we set r({) = 0 in this integral equation, we can derive (VI.7.2) and (VI.7.3).
Details are left to the reader as exercises (cf. (Ge1LJ ).
V1-9. Differential equations satisfied by reflectionless potentials
Suppose that u(x) is a reflectionless potential whose associated scattering data
are given by |
0, (r11, r12, ... flN ), |
(cl, c2,... , cN) },where 0 < ril < ]2 < |
||||
< rily. It was proven that one of the Jost solutions is given by |
||||||
|
f+(T,() = e`t= |
1 -Ni 9j (T) |
||||
|
|
|
|
|
|
j_1 (+ |
Furthermore, |
|
|
|
|
|
N |
|
f+(x, -() = e`= 1 - |
|||||
|
g1 (r) |
|||||
|
|
|
|
|
|
-( + trj7 |
is also a solution of |
|
|
|
|
|
|
|
dX2 - (u(x) - A)y = 0, where J = (2. |
|||||
Therefore, |
|
|
|
|
|
|
|
|
N |
|
|
|
|
(VI.9.1) |
P(x, A) = 11 ! 1+11- |
|
f+(T,C)f+(x, -() |
|||
|
|
3=1 |
A ) |
|
|
|
satisfies the differential equation |
|
|
|
|
||
(E) |
3 |
+ 2(u(x) - A) dP + ddx)P = 0. |
||||
|
|
|
|
|
|
|
|
2 dz3 |
|
|
|
|
|
10. PERIODIC POTENTIALS |
183 |
implies that u = 0 if and only if gl = 0 (1 = 1, 2, |
- , N). This, in turn, implies |
|||
that |
|
|
|
|
P(0, A) = 11 C1 + |
A |
|
|
|
and |
|
where |
A=C2. |
|
G(u, A) = f+(x, C)f+(x, |
||||
Hence, ao = 1 and |
N |
+ |
|
|
N |
2 |
|
||
1 + E |
= |
|
|
|
3=1 |
1 =1 |
|
|
|
Example VI-9-2. The function u(x) = -2q2 sech2(17(x /+ p)) is a reflectionless potential corresponding to the data {r7, c}, where p = 2- In 1c) (cf. Example VI-
7-5). In this case, G2(u)+172G1(u) = 0. Also, G3(u)+(n1+7)2 2(u)+rI14Gi(u) _
0 in the case when N = 2 and {r(C) = 0, (i71, n2), (cl, c2)} are the scattering data.
In particular, G3 (U) + 13G2(u) +36G1(u) = 0 if i1 = 2 and '12 = 3.
Remark VI-9-3. The materials of §§VI-5-VI-9 are also found in (TDJ.
VI-10. Periodic potentials
In this section, we consider the differential equation
(VI.10.1) |
2 + (A - u(x))y = 0 |
under the assumption that
(I)u(x) is continuous for -co < x < +oo,
(II)u(x) is periodic of period 1, i.e., u(x + e) = u(x) for -oo < x < +oo.
The period a is a positive number and A is a real parameter. Denote by 01(x, A) and 02(x, A) the two linearly independent solutions of (VI.10.1) such that
(VI.10.2) 01(0,A) = 1, |
|
1(o |
A) = 0, |
02 (0, A) = 0, |
d2(0,A) = 1. |
|
Set |
|
|
01(e, A) |
02(t, A) |
|
|
|
|
|
|
|
||
.D(A) |
= |
O1 (t"\) |
(e A) |
|
|
|
The two eigenvalues Z1(A) and Z2 (A) of the matrix |
are the multipliers of the |
|||||
periodic system |
|
|
|
|
|
|
d |
|
|
|
|
|
|
(VI.10.3) |
[Y2J = [u(--O-A 0, [y21 |
|
||||
dx |
|
186 |
VI. BOUNDARY-VALUE PROBLEMS |
Theorem VI-10-6. Assuming that u(x) is continuous and periodic of period t > 0 on the entire real line R, consider three boundary-value problems:
(A) |
2 + (A - u(x)) y = 0, y(0) = 0, |
y(t) = 0, |
||
($)2 + (A - u(x)) y = 0, |
W) = y(0), |
dx (t) _ |
y (0), |
|
(C) |
d2 + (A - u(x)) y = 0, |
y(t) = -y(0), |
dY(t) _ |
L(0). |
Then, each of these three problems has infinitely many eigenvalues:
|
µI |
<µ2</A3...<An<..., |
|
(VI.10.9) |
A0<\1<A2<A3<a4 <... |
<A2n_I<\2n <..., |
|
|
V1 <v2 <v3 :5 L14 <... <V2n-1 <V2n <... , |
||
respectively. Furthermore, if we denote by an(x), |
and 7n(x) the eigenfunc- |
tions associated with the eigenvalues µn, An, and vn, respectively, then
(1)A0 < v1,
(2)v2n-1 :5 92n-I 1'2n < A2n-1 !5 µ2n < A2n < V2n+1 < µ2n+1 < i/2(n+l) for
n = 1,2,...,
(3)02n_1(x) and /32n(x) are linearly independent if A2,,-1 = A2n,
(4)72n-1(x) and 72n (x) are linearly independent if v2n-1 = v2n,
(5)3D(x) does not have any zero on the interval 0 < x < t,
(6)02._1(x) and /32,,(x) have exactly 2n zeros on the interval 0< x < t,
(7)72n-1(x) and 72n (x) have exactly 2n - 1 zeros on the interval 0 < x < t.
Proof.
We prove this theorem in five steps. Note first that (VI.10.9) is obtained from
Figure 1.
Step 1. If ¢(x, A) is a solution of differential equation (VI.10.1), then |
(x, A) is |
a solution of the initial-value problem |
|
(VI.10.10) |
|
d2w + (A - u(x)) w + O(x, A) = 0, |
w(0, A) = |
(0, )L), |
i i(0,,\) = |
|
|
dx2 |
TA_ (dx |
|
|
(cf. §lI-2). Therefore, using the variation of parameters method, we obtain
(VI.10.11) |
|
_ |
-1(x, A) _ (x, A) fT 01(t, A)02(t, A)dt |
¢2(x, A) |
f01(t"\)2dt' |
d z (x,,\) = 01(x"\) 102(t, A)2dt - 02(X,,\) f -01(t, A) 02(t, A)dt