1

Appendix 1.
Introduction
to Banach Spaces
A1.1 Linear Vector Spaces and Norms

Definition A1.1 A linear vector space (LVS) X over a field F is a set with two
binary operations: addition (a map of X x X ~ X) and numerical (or scalar)
multiplication (a map of F x X ~ X), which satisfy the commutative, associative,
and distributive properties.
We are primarily interested in spaces of functions and will take F to be either
lR or cc.
Examples A1.2.
(1) X = lRn with the operations (+, .), the usual componentwise operations, and
with F =R
(2) X = C([O, I D, all cqntinuous, complex-valued functions on the interval

[0, I]. Addition is pointwise,thatis, if f, g E X then (j+g)(x) = f(x)+g(x),
and scalar multiplication for F = C is also pointwise.
(3) X = II',
where II' consists of all infinite sequences of complex numbers:
x == C such that L~l IXi I" < 00, for 1 :::: p < 00.
(Xl, ... , x n, ... ), Xi E
Again, addition and scalar multiplication are defined componentwise.
Problem A1.1. Show that II' is an LVS. (Hint: Use the Minkowski inequality:
286 Appendix 1. Introduction to Banach Spaces
Definition A1.3. Let X be an LVS over F, F = lR or <C. A norm II . lion X is a

map II . II : X ~ lR+ U {O} such that
(i) IIx II 2: 0 and if IIx II = 0 then x = 0 (positive definiteness);

(ii) II Ax II = IAlllxll, A E F (homogeneity);
(iii) IIx + yll :::: IIxll + Ilyll, x, y E X (triangle inequality).

Examples AI.4.
(1) X = lRn and define IIxll = (2::;'=1 Ixd2)1/2, xE lRn. This is a norm, called
the Euclidean norm, and has the interpretation of the length of the vector x.
(2) X = C([O, 1]) can be equipped with many norms, for example, the LP-
1 )I/P
norms: Ilfllp == ( fa If(xWdx ,f E x, 1:::: p < oo,andthesup-norm
(p = 00): Ilflloo == sup If(x)l, f E X.
XE[O,IJ
(3) X = IP has a norm given by IIxll == (2::~1 IxilP)I/P, x EX.

Problem A1.2. Show that the maps X ~ lR+ claimed to be norms in Examples
A1.4 are norms.
Definition A1.S. An LVS X with a norm II . II is a normed linear vector space
(NLVS), that is, an NLVS X is a pair (X, II . II) where X is an LVS and II . II is a
norm on X.
Al.2 Elementary Topology in Normed Vector Spaces

Elementary topology studies relations between certain families of subsets of a set
X. A topology for a set X is built out of a distinguished family of subsets, called
the open sets. In an NLVS X, there is a standard construction of these open sets.
Definition A1.6. Let X be an NLVS, and let a E X. An open ball about a of radius
r, denoted Br(a), is defined by
Br(a) == {x E Xlllx - all < r}.
In analogy with the Euclidean metric on lRn (see (1), Examples A1.4), we in-
terpret Ilx - all as "the distance from x to a," and so Br(a) consists of all points
x E X lying within distance r from a. The open balls in X are the fundamental
building blocks of open sets.
Definition AI.7. A subset E C X, X an NLVS, is open iffor each x E E there
exists an r > 0 (depending on x) such that Br(x) C E. A neighborhood ofx EX
is an open set containing x.
Al.2. Elementary Topology in Normed Vector Spaces 287
Example A1.S.
Let X = JRI1 with the Euclidean metric. Then
simply a ball of radius E about the origin. Note that BE(O) does not include its
boundary, that is, {x I (2:::'=1 x?) 1/2 = E}. It is easy to check that BE(O) is an open
set. If we were to include the boundary, the resulting set would not be open.
DefinitionA1.9.Asubset Be X, X anNLVS, isclosedifX\B == {x E Xix rj. B}
(the complement of B in X) is open.
Sequences provide a convenient tool for studying many properties of subsets of
an NLVS X, including open and closed sets.
Definition A1.l0. Let X be an NLVS.
(I) A sequence {XI1 l in X converges to x E X if Ilxn - x II ~ 0 as n ~ 00.
(2) A sequence (x" l in X is Cauchy if Ilxn - Xm I ~ 0 as n, m ~ 00.
Problem Al.3. Prove that
(I) every convergent sequence is a Cauchy sequence,
(2) every convergent sequence is uniformly bounded, that is, if {xn l is conver-
gent, then there exists an M, 0 < M < 00, such that Ilx n I :::: M for all
n = 1,2, ....
(Hints:
(1) Use an E/2-argument and the identity: X il - Xm = Xn - X +X - Xm.
(2) Use the triangle inequality to show that
I Ilxn I - Ilx m I I:::: Ilxn - Xm II·)
A nice feature of NLVS is that sequences can be used to characterize closed

sets.
Proposition Al.ll. A subset E c X, X an NLVS, is closed if for any convergent
sequence (xnl in E, limn->ooxn E E.
Proof.
(1) Suppose E is closed and lim l1 -+ oc Xn =

x rj. E but Xn E E. Since X \ E is
=
open, there is an E > 0 such that BE(x) n E </>. But X I1 ~ x, so for this
E, there is an N such that n > N =? Ilx n - xII < E =? Xn E BE(x). So we
have a contradiction, and x E E.
(2) Conversely, suppose E contains the limit of all its convergent sequences.
Suppose X \ E is not open; then there is a point Y E X \ E such that for
each E > 0, BE(y) n E =I ¢. Let En == lin and choose Yn E BE,Jy) n E.
Then {Yn} is a sequence in E and limn-+oo Yn = y. But y tI E, hence we get
a contradiction (i.e., X \ E is open so E is closed). 0
Definition A1.12. If M c X, X an NLVS, then M, the closure of M in X, is

M U {limits of all nonconstant convergent sequences in M}. The set of points that
are limits in X of nonconstant sequences in M are the accumulation points or
cluster points of M.
Remark Al.13. By definition, M :J M and M is closed. M is, in fact, the smallest
closed set containing M.
Remark A1.14. Exercise (1) in Problem A1.3 is significant because the converse
is not true in every NLVS X. That is, if {xn} is a Cauchy sequence in X, it is not
necessarily true that {xn} is convergent. As an example, consider X = C([O, 1]) in
the II . III" 1:s p < 00, norm (see Example AlA (2)). We construct a sequence
fn E X by
0, O:SX:S~,
f,,(X) = I n(x-~), .l<x<.l+.l
2- -2 n'
1, .l+.l<x<1.
2 n - -
Clearly, fn E C([O, 1]) and, for n ::::: m and p = 1,
Ilfn-fmll! = fo!If,,(X)-fm(X)ldX
h2
1+1
2 11 (n - m)(x - 1/2)dx + h:l
2
1+1-
11
l1l
[1 - m(x - 1/2)]dx
~2 (~-~),
m n
and so {fn} is a Cauchy sequence. It is easy to check that for any p, {fn} is Cauchy.
I
Now for any x E [0, 1], we can compute lim n-+ oo fn(x). The limit function is
0, 0 < x :s ~,
f(x) =
1, ~ < x :s 1.
This limit function f tI C([O, 1]) (i.e., f is discontinuous), and so C([O, 1]) does
not contain the limit of all its Cauchy sequences! Later we will discuss the manner
in which we can extend C([O, 1]) to a larger NLVS (called the completion) to
include the limit of all the Cauchy sequences in C([O, 1]).
AI.3 Banach Spaces

Definition A1.15.An NLVS X is complete if every Cauchy sequence in X converges
to an element of X. A complete NLVS is called a Banach space.
Al.3. Banach Spaces 289
Remark A1.16. Every finite-dimensional NLVS is complete and hence a Banach

space. This follows from the fact that both ffi. and C are complete and from the fact
that every finite-dimensional LVS is isomorphic to ffi.N or C N for some N. For this
reason, the term "Banach space" is usually used to denote an infinite-dimensional
NLVS.
Examples A1.17.
(1) X = ffi.N with the Euclidean norm II . lie is complete.

(2) X = C([O, I)) is complete with the sup-norm II . 1100, but it is not complete
with the p-norm for 1 :s p < 00; see Remark A1.l4.
(3) [P is complete; we prove this in Proposition AU8.
Problem AI.4. Prove statements (l) and (2) in Examples A 1.17. (Hint: For (2),
use standard results on uniform convergence.)
Proposition A1.18. The NLVS IP as defined in (3) of Examples Al.2, with the
I . lip-norm, 1 :s p < 00, is a Banach space.
Proof. Let (x(n)), x(ll) == (xiII), xiII), ... ) E [P, be a Cauchy sequence, that is,
(ALl)
as m, n -+ 00. We must show (1) that (x(Il)} is convergent to some x, and (2) that
x E [P.
(1) Condition (A 1.1) implies that each sequence of real (or complex) numbers
{Xill)}~1 (fixed i) is Cauchy, and from the completeness of ffi. (or q it is
convergent. Hence there is an Xi E ffi. (or C) such that Xi = limll -+ oo x i(Il).
Let x == (XI, X2, .. .). We show that x(ll) -+ x in the [P-norm. From the
definition of (x(Il)), for each E > 0 there is an NE E N such that m, n > NE
implies [2::7=1 Ixt) - xim)IP riP < E. Now we take the limit of the left
side as n -+ 00 (fixed k) and get Sm(k) = [2::7=1 IXi - x;m)IP riP < E.
As a sequence in k, Sm(k) is monotonic (i.e., Sm(k) :s Sm(k') if k < k')
and bounded, and hence limk-+oo Sm (k) exists and is less than E. This is the
[P -norm of (x - x(m), that is, Ilx - x(m) lip < E. Hence, x(m) -+ x.
(2) To prove that x E [P, we use Minkowski's inequality (Problem AU) and
begin by considering a finite sum as above. We write
290 Appendix l. Introduction to Banach Spaces
Since X(Il) E IP and Ilx(ll) - xll p ---+ 0, given E > 0, we can choose n > NE
such that
and hence Ilx lip < 00 (i.e., x E IP). 0
We now introduce some other topological ideas associated with Banach spaces.
Definition A1.19. Let X be an NLVS. A subset A C X is dense in X if each point

x E X can be approximated arbitrarily closely by points in A, namely, for each
x E X there is a sequence {all} c A such that lim n -+ oo an = x. Equivalently, for
°
any x E X and any E > there exists a E A such that Ilx - a II < E.
Examples A1.20.
(1) The rational numbers Q are dense in the real numbers.
(2) The Weierstrass approximation theorem states that the set of all polynomials
in x on [0, 1] (i.e., functions of the form 2:7=1 Ci xi + co) is dense in C([O, 1])
with the sup-norm.
(3) Let X be a Ba~ach space, and let A C X be a subset that is not closed. Then
A is dense in A, its closure.
Definition AI.21. A Banach space X is separable if it contains a countable, dense

subset. (A set is countable ifits elements can be put in one-to-one correspondence
with the set of integers Z.)
Examples A1.22.
(1) The set of polynomials on [0, 1] with rational coefficients is a dense, count-
able set in C([O, 1]) with the sup-norm, so this Banach space is separable.
(2) The set of all elements x E lP with Xi E Q is a dense, countable set in lP,
so IP is separable.
(3) Simply forreference, the space L 00 (lR), the LVS of all (essentially) bounded
functions on lE. with the (ess) sup-norm, is nonseparable.
Remark A1.23. All the spaces used in the chapters are separable, unless specifi-
cally mentioned.
A 1.4. Compactness 291
A 1.4 Compactness
We introduce another important topological notion associated with Banach spaces
(which is, of course, an important notion in its own right).
Definition A1.24. Let X be a (separable) Banach space. A subset Y C X is called
compact (respectively, relatively compact) if any infinite subset of Y contains a
convergent sequence and the limit of this sequence belongs to Y (respectively, to
X).
Note that if Y C X is relatively compact, then Y, the closure of Y in X, is
compact. Moreover, if Y is compact, then Y is a closed subset of X. This follows
since if {Yn} is a Cauchy sequence in Y, it is an infinite subset of Y. Since X is
complete, {Yn} is convergent and the limit y = limn --+ oo Yn E Y by definition of
compactness. Hence, we have the next result.
Proposition A1.25. Let X be a Banach space. If Y C X is compact, then it is
closed and bounded (i.e., 3M > 0 such that Ilyll :::: M Vy E Y).
Proof. It remains to show that Y is bounded. Suppose not; then for each nEZ,
there exists Yn E Y such that IIYnl1 > n. Then {Yn} is an infinite subset of Y. By
compactness, we can choose a subsequence {Yarn)} of {Yn), where a: Z+ --+ Z+
is an order-preserving map, which converges. Then, as Y is compact, 3y E Y such
that I Yarn) II converges to Ily II. But, IIYa(n) I ::: a(n) ::: n --+ 00, and so we get a
contradiction. Hence Y is bounded. D
It is a well-known result that the converse of Proposition A1.25 is not true:
There are many closed and bounded subsets of certain Banach spaces which are
not compact.
Examples A1.26.
(1) The problem just described cannot occur in jRN or (ef: the Heine-Borel
theorem states that a set is compact if and only if it is closed and bounded.
(2) Let X = C([O, 1]), and let Y C X be the subset of all functions in X which,
together with their derivative, are uniformly bounded, that is, Y = {f E
XI Ilflloo < M and Ilf'lloo < M}. The Ascoli-Arzela theorem states that
Y is compact in X.
We now want to relate the definition in A1.24, which depends on the metric
structure of X, to the general topological notion of compactness.
Definition A1.27. Let A C X, X any topological space. A collection of open
subsets {Vi} of X is called an open covering of A if A C Ui Vi.
Theorem Al.28. A subset K C X, X a Banach space, is compact if and only if
any (countable) open covering of K contains afinite subcovering of K.
Proof.
(1) Suppose K C X is compact, and let {Vi} be an open covering of K. Suppose
on the contrary that there exists no finite subcovering. Then K \ U7=1 Vi is
nonempty for any n. Let XII E K \ U7=1 Vi. By the compactness of K, there
exists a convergent subsequence {XII'} and X = limn,--+oo Xn' E K. Since {Vi}
covers K, X E Vj, for some j. Since Vj is open, XII' E Vj for all large n',
say n' > No. Then {VI, ... , VNo' Vj} (where Xn' E Vn) is a finite covering
of K, giving a contradiction.
(2) Suppose K has the finite covering condition but is not compact. Let {xn} be
an infinite (countable) sequence in K with no limit points. The sets Un ==
{Xn, Xn+l, ... } are closed (Un has no limit points) and nnull = nnun = ¢.
Then U,~ is open and {U~} forms an open cover of K. Then there exists a
finite subcovering {U~,} of K with nil' Un' = ¢, which is certainly not true
by construction. D
Appendix 2. The Banach
Spaces LP(IR n ), 1 < p < 00
We present the basic theory of this family of Banach spaces. The reader can find
any of the proofs not presented here in any standard text on real or functional
analysis; see, for example, Reed and Simon, Volume I [RSl], Royden [Ro], or
Rudin [R].
A2.1 The Definition of LP(IR n ), 1 ::::: P < 00
The most important Banach spaces for us are the LP -spaces, 1 S p < 00. The
value p = 2 plays an especially important role, and we will describe its properties
in greater detail later. There are two basic ways to obtain the LP -spaces, and both
are useful. The first, which we will sketch here, uses the idea of the completion
of an NLVS. The second involves Lebesgue integration theory, so we will merely
mention the results.
Definition A2.1. The support of a function f: Q C jRn -+ <C, denoted by supp f,

is the closed set that is the complement of the union of all open sets on which f
vanishes.
Problem A2.1. Prove that supp f = {x E Q I f(x) i OJ.
Definition A2.2. The set of infinitely differentiable functions on jRn with bounded
(and hence compact) supports is denoted by Cgo(jRIJ). IfQ C jRn and is open, then
Cgo(Q) consists of all infinitely differentiable functions with compact supports in
Q.
294 Appendix 2. The Banach Spaces LP(JR"), 1 ::: p < 00
We note that it is easy to see that Co(lR") is an LVS over Cor R We will always
take LVS over C in this section. For each p, I ~ P < 00, and any f E Co(Q),
define the nonnegative quantity
[In
I
Ilfllp == If(x)iP dxJ P , (A2.l)
which exists as a Riemann integral because supp f is compact and because f is

bounded on its support. We see that (Cgo(Q), II . lip) is an NLVS.
Problem A2.2. Prove that for any open Q c JR", (Co(Q), II . lip) is an NLVS for
1 ~ P < 00. (Hint: Use the simple identity, If + glP ~ 2 P (lfl P + IgIP).)
Ourgoalis to associate with X p == (Cgo(Q), I . lip) a Banach space that contains

X p as a dense set. Recall that we have already seen in Remark A 1.14 that X p is
not complete. There are Cauchy sequences in X p whose limits do not exist in X p.
Hence, we want to enlarge X p by adding to it "the limits of all Cauchy sequences
in X p." This will result in a "larger" space, called the completion of X p. It is a
Banach space that contains a copy of X p as a dense subset. We will sketch the
idea of the completion of an NLVS. In this manner, we obtain the Banach spaces
in which we are interested.
Definition A2.3. For any open Q C JR" and 1 ~ p < 00, the Banach space LP(Q)
is the completion of (Co(Q), II . lip).
Let (X, II . II) be anNLVS that is not complete. We show how to obtain a Banach
space X with a dense subset X, such that X is isomorphic to X. That is, the sets
X and X are in one-to-one correspondence, and if i E X corresponds to x EX,
then IIi II = IIx II (the norm on the left is the induced norm on X).
Step 1. Consider the collection of all Cauchy sequences in X. Let a == {x,,} and
f3 == {y,,} denote any two Cauchy sequences. We say that two Cauchy sequences
are equivalent if lim,,--+oo IIx" - y,,11 = O.
Problem A2.3. Show that (i) this defines an equivalence relation on the set of
Cauchy sequences, and (ii) if {x n } is Cauchy, then limn--+oo IIxn II exists.
Let [a] denote the equivalence class of a. We define a set X to be the set of
all equivalence classes of Cauchy sequences in X. The set X has a distinguished
subset X, which consists of all equivalence classes of constant sequences (i.e., if
x E X, then let ax == {x" = x}). We see that [ax] consists of all Cauchy sequences
converging to x. Hence, X is isomorphic with X.
Problem A2.4. Show that X is an LVS, and verify that X is isomorphic to X.

A2.1. The Definition of U(/R n ), 1 :::: p < 00 295
Step 2. We define a norm on X as follows. Let [a] E X, and take {xn} E [a]. Then
define
lI[a]lI x == lim IIxnll.
n-+oo
Problem A2.S. Verify that II . II x is a norm on X. (Be sure to check that it is "well
defined," that is, independent of the representative chosen.)
Note that if [ax] E X, then naturally
so that the isomorphism between Xand X is isometric. We conclude that (X, II . II x)

is an NLVS with a subset X isometric with X.
Step 3. We sketch the proof that (X, II . IIx) is complete. Suppose {[aN]} is a
Cauchy sequence in X. Let {xt'} be a representative. Then for E > 0 there exists
NE such that for N, M > NE , we have
Hence there exists io > 0 such that IIx/" - x,M II < E 12 for i > io. For N > M,
choose an index iN > io for which Ilx~ - x,M1I < E/2. We can choose the map
N --+ iN to be order-preserving (i.e., P > Q => ip > iQ). Since {x,M} is Cauchy,
there exists i 1 > 0 such that j, k > i 1 implies
Choose i M > max(io, i d. We claim {x~} C X is a Cauchy sequence. This follows

since
IIx~ - x~ II :s IIx~ - x~ II + IIx~ - x~ II < E,
by definition of iN. Finally, construct [a] == [{x~}] E X. Then we have for N

sufficiently large,
= lim
k-+oo
xt II
Ilxt' -
< k~~ (IIX~ - xt II + IIxt' - Xi~ II)

3E
<
2
Problem A2.6. Check that the calculations in Step 3 are independent of the rep-
resentative of the equivalence class chosen.
296 Appendix 2. The Banach Spaces LP(JRIl), 1 ::: p < !Xl
Consequently, we know that there exists a Banach space XI' that contains a
dense isometric copy of (COO(S1), I . 111')' A natural question that arises is: What
is the nature of the objects in XI' ? The best way to answer this is by means of the
Lebesgue theory of integration. We wish to avoid this discussion, and so we refer
the reader to any standard text, for example, Royden [Ro]. However, for the reader
familiar with this theory, an element of XI' is an equivalence class of measurable
functions on S1 having the property that If II' is Lebesgue integrable on S1 with
In If II' < 00. Two measurable functions f and g on S1 are equivalent if they
differ on a set of Lebesgue measure zero. Up to this equivalence, then, we can
interpret elements of LI'(S1) as functions such that If II' is finitely integrable over
S1.
A2.2 Important Properties of LP -Spaces

We establish here several basic and important properties of LI' -spaces, which we
use throughout the chapters.
1. Density results
For S1 C ]Rn open, we have seen that the linear vector space offunctions Cg:'(S1) is
dense in LI'(S1). Since LI'(S1) is a Banach space, this means that for any f E LI'(S1)
(where f is understood in the sense described in Section A2.1) and for any E > 0,
we can find g E Cg:'(S1) such that Ilf - gill' < E. Consequently, in most cases, it
suffices to prove a property for smooth functions. The same density result applies
for other classes of functions. We list some of the common classes here:
Co(S1) == all continuous functions on S1 with bounded supports;
Bo(S1) == all bounded functions on S1 with bounded supports.
Problem A2.7. Prove that these two classes of functions are dense in LI'(S1).
When S1 = ]R1l, we define the Schwartz class functions S(]Rn) as follows:
S(]Rll) == {f E Coo(]Rn) I if g is any partial derivative of f, then for any

polynomial P, limllxll-+oo Ig(x)p(x)1 = O}.
The Schwartz functions form a particularly nice class of functions. By the same
argument as in Problem A2.7, these functions form a dense set in LI'(]Rn).
A2.2. Important Properties of U' -Spaces 297
2. The HiJlder Inequality

It is easy to see that the product of two LP -functions need not be in LP. For
example, take the space L 1(JR) and consider the function
IX I- 1, Ixl < 1,
f(x) = ! 0, Ixl ~ 1.
Then, f ELI (JR), but f2 ~ L 1(JR). We do, however, have the possibility that f g
belongs to some LP' -space if f and g belong to some other U - and Lt -spaces.
TheoremA2.4 (The Holder inequality). Let p, q, r E JR+ be such that II p+ llq =

llr.lf f E LP and g E Lq, then fg E L' and
Ilfgllr :s Ilfllp Ilgll q .

Proof.
(I) We first prove the following inequality for any a, b E JR+ and any p, q E JR+
such that 1I p + 1I q = 1:
(A2.2)
Consider the function

a = F(b) = bq - 1
or, equivalently,
= F-1(a) = a P- 1,
b
where we use the fact that (p - 1)(q - 1) = 1. We consider the rectangle

[0, bo] x [0, ao] in the ba-plane, where ao, bo > °
are arbitrary. Then the
curve a = F(b) divides the rectangle into two regions, I and II. The area of
region I, which lies below the curve a = F(b), is
whereas for the complementary region,
It is easy to see that we always have
which is just (A2.2).

298 Appendix 2. The Banach Spaces UJ(JR"), 1 ::c p < 00
(2) We next consider 1I p + 1I q = I I r. Setting a r and b r into (A2.2), we obtain
for lip + l/q = I. Now set p' == rp and q' = rq to obtain
(A2.3)
for 1I p' + 1I q' = 1I r.
(3) If either I or g vanishes almost everywhere, the inequality is trivial. So

suppose I E U and g E Lq are nonzero. Let j == 111/11-;;1 and g
gllgllq-l. Substituting these into (A2.3) and integrating, we obtain
Iljgll; < rp-l Hill: + rq-lllglI~

< r(p-I + q-l) ::c 1.
By the definition of j and g, we see that this is simply the desired inequality.
D
Proposition A2.S. Let I E LP, 1 ::c p < 00. Then
IIIll p = sup {IIIgIIIlIgllq-llgiOand.!.+.!.=I} (A2A)

gELq p q
Proof. The Holder inequality with r = 1 implies that the right side of (A2A) is
bounded above by IIIll p . Assume I i O. Now let g == cIIIP-I. By a simple
calculation,
IIIgllI = cllfllf"
so if we take c == II I 111- P, we get equality in (A2A). Finally, we check that for

such g E Lq,
I
Ilgll q IIII11- p [/ II(x)lq(p-l) dX] q

= IIII11- p IIIIIf,/q = 1
since (p - 1) = plq. D
3. The Minkowski Inequality

We used a basic inequality in Problem A2.2 to show that LP is an LVS. This can
be refined so that the constant does not depend on p.
A2.2. Important Properties of LP -Spaces 299
Theorem A2.6 (The Minkowski inequality). Let 1 :::: p < 00 and f, g E LP.
Then
IIf + gllp :::: Ilfllp + Ilgll p ·
Proof. The case p = 1 is obvious. We use Proposition A2.S to obtain
Ilf +gllp = sup

hELq
{11(f+g)hlllllhll;llhioand~+~=l}
p q
< sup {(llfhlll+lIghlldllhll;l}

hELq
< II flip + Ilgll p .
o
Problem A2.8. Formulate and prove the Minkowski and Holder inequalities for
IP,l :::: p < 00 (see Examples A1.2 and Problem AU).
4. Lebesgue Dominated Convergence

This is one of the most useful results of Lebesgue integration theory. It is, of
course, applicable to Riemann integrable functions. Suppose we have a sequence of
functions fn in LP(lRn) such that fn converges pointwise to f almost everywhere.
We want to know if fn ~ fin LP(]RH).
Theorem A2.7 (Lebesgue dominated convergence). Suppose fn E UJ(]Rn) and

fn ~ f almost everywhere. If there exists g E LP(]Rn) such that Ifnl P :::: IglP
almost everywhere, then f E LP and fn ~ fin LP.
We need one other result about sequences of functions that converge in the
LP-norm.
Theorem A2.8. Suppose that Un} is a sequence offunctions in LI(]Rn) that con-
verges to f in the L I-norm. Then, there is a subsequence Un(k)} that converges
pointwise almost everywhere to f.
We refer the reader to [RS I] or [Ro] for the proof.

Appendix 3. Linear Operators
on Banach Spaces
A3.1 Linear Operators

Definition A3.1. Let X and Y be LVS. A linear operator A from X to Y is a linear
map defined on a linear subspace D(A) C X, called the domain of A, into Y.
We write A : DCA) -+ Y. A linear map has the property that for A E C (or]R)
and f, g E D(A),
A(A.f + g) = AAf + Ag.
Note that D(A) is a linear space by definition, so f, g E D(A) implies that f +Ag E
D(A).
Examples A3.2.
(1) Let X = Y = ]Rn. Any linear transformation A : ]Rn -+ ]Rn (i.e., D(A) = ]Rn)
can be written in the form (AX)i =L:J=l AijXj, for some n 2 quantities
Aij E R This matrix representation of A depends on the choice of basis for
]Rn. The numbers Xi are the coefficients of X relative to this basis.
(2) Let X = Y = C([O, 1]), and let a E X. Define A: X -+ X by (Af)(x) =

a(x)f(x), f E X. Then A is a linear operator with D(A) = X. The oper-
ator A is "multiplication by the function a" and is called a multiplication
operator.
(3) Let X = Y = C([O, 1]) and let D(A) = C1([O, 1]), the LVS of continuous
functions that are differentiable and whose first derivatives are continuous.
Note that D(A) =I X. Define an operator A, for f E D(A), by (Af)(x) =
(dfldx)(x). Then A: C1([O, 1]) -+ C([O, 1]) is a linear operator.
302 Appendix 3. Linear Operators on Banach Spaces
(4) Let X = Y = C([O, I]) and let K(s, t) be continuous in sand ton [0, I] x
[0, IJ. Let f EX, and define (Af)(t) = JdK (t, s) f(s )ds.1t is easy to check
that this integral defines an element of X for each f E X. Hence, we may
take D(A) = X, and A is a linear operator.
(5) Let X = C([O, I]) and Y = R For each f E X, define Af = fol f(s)ds.
Then A: X --+ Y with D(A) = X is a linear operator.
(6) Let X = Y = C([O, I]). The map (Af)(x) = f 2 (x), defined on all of X, is
not linear. This is an example of a nonlinear operator.
Problem A3.1. Verify all the statements in Examples A3.2.
Definition A3.3. A linear operator A : D(A) C X --+ Y is densely defined if

D(A) is dense in X.
Remark A3.4. Note that the definition of a linear operator includes both the do-
main and the action of the operator on elements in the domain. If the domain is
changed, the operator is also changed. All the operators in Examples A3.2 are
densely defined. A linear operator that is not densely defined has many undesir-
able properties (for example, we shall see that it does not have a uniquely defined
adjoint). For this reason, we will work only with densely defined operators. More-
over, we will assume that X is an NLVS. If D(A) C X is not dense, we will replace
X by D(A) == XI, the closure of D(A) in X, which is a Banach space. We then
consider the operator A with dense domain D(A) in X I.
It is possible to define algebraic operations on the family of linear operators
from X to Y. Assume X and Y have the same field of scalars F. In particular, we
have
(1) Numerical multiplication: If A E F, and x E D(A), we define AA by

(AA)(x) = A(Ax). Thus, if A =I 0, D(AA) = D(A), and if A = 0, AA extends
to the zero operator on X.
(2) Addition: Let A, B be two linear operators on X into Y, and suppose f E

D(A) n D(B). We define A + B on f by (A + B)f = Af + Bf. In some
cases it may happen that D(A)n D(B) = {O}, so we cannot define A + B. In
general, D(A) n D(B) is a nonempty subset of D(A) and D(B) (although
it may not be dense). For example, consider the operator A defined in part
(3) of Examples A3.2. Let B == d 2 jdx 2 , with D(B) = C 2 ([0, 1]). Then,
D(A + B) = D(B) c D(A) and is dense.
(3) Multiplication: Let A, B be linear operators from X into Y, and let x E D(B)
be such that Bx E D(A). Then we can define AB on the set of all such x
by (AB)x = A(Bx). Note that if A and B are everywhere defined operators
from X into X, then AB and BA are everywhere defined on X.
A3.2. Continuity and Boundedness of Linear Operators 303
A3.2 Continuity and Boundedness of Linear Operators

We will now explicitly assume X and Yare NLVS, and we will write A: X ~ Y
even when A is defined only on D(A) C X. Moreover, by operator we will always
mean linear operator.
Definition A3.5. An operator A : X ~ Y is continuous iffor each convergent
sequence {x,,} C D(A) with lim,,--+oox n = x E D(A), we have lim,,-+oo AXil =
A(lim,,-+oo x ll ) = Ax.
Problem A3.2. Show that A is continuous if for each sequence {x,,} C D(A) with
lim n --+ oo Xn = 0, we have limn-+oo Ax" = 0.
Example A3.6. With reference to part (2) of Example A3.2, the operator A, mul-
tiplication by a E C([O, 1D, is continuous on X = C([O, 1]) with any norm on X
as given in Example A 1.4 (2). This follows from the fact that a is bounded: Let
sUPSE[O.ljla(s)1 = Iialloo. Then, it follows that
IIAflloo :S Iialloollfiloo
and
IIAfilp :s Iialloollfll p , 1:S p < 00.
So if fn ~ °
in any of these norms, Af" ~ 0.
Definition A3.7. An operator A : X ~ Y is bounded if there exists a K > Osuch
thatforany f E D(A),
IIAflly :S Kllfllx.
Remarks A3.8.
(1) If should be noted that both continuity and boundedness of an operator

depend on the norms of X and Y.
(2) Ifboth X and Yare finite-dimensional, any operator A : X ~ Y is bounded

and continuous.
Proposition A3.9. Let X be an NLVS.
(1) Any bounded operator on X is continuous.
(2) Any bounded operator A : X ~ Y, Y a Banach space, can be extended to

all of X, that is, there is a unique bounded operator A with D(A) = X such
that Af = Af, f E D(A).
Proof.
(1) Let {x,,} C D(A) be a sequence such that x" ~ 0. Since A is bounded,
IIAx n II :S K Ilx n II ~ 0, so AXn ~ 0.
(2) Let x E X \ D(A). Since D(A) is dense, there is a sequence {xnl C D(A)
such that Xn ~ x. Because A is bounded, it follows that {Ax nl is Cauchy in
Y. Since Y is a Banach space, the sequence {Ax n l converges to some Y E Y.
Now define A On X by Ax == Ax, x E D(A), and Ax == limn--->oo Axn , for
x = limn---> 00 Xn tI- D(A) and {xnl C D(A). D
Problem A3.3. Complete the proof of Proposition A3.9 by showing that A is well
defined (i.e., independent of the sequence), that A is linear, that A is unique, and
that A is bounded.
Theorem A3.IO. Let A: X ~ Y be a linear operator. Then the following two
statements are equivalent:
(1) A is continuous;
(2) A is bounded.
Proof.
(2) => (I) If A is bounded, we observed that A is continuous; in fact, we can

extend A to a bounded operator on X and the extension (which is bounded)
is continuous.
(I) => (2) Suppose A is continuous but unbounded. This means the exists a
sequence {xnl in D(A), with IIxnll < M, such that IIAxnl1 2: nllx,& Let
Yn == Xn IIAx n II-I. Then {Ynl C D(A), llYn II :s n- I , so Yn ~ 0, but
IIAYn II = 1; that is, lim n-+ oo AYn =I 0 but Yn ~ O. This contradicts the
continuity of A, so A must be bounded. D
Definition A3.11. Let A: X ~ Y be bounded. Then IIAII, the norm of A, is

defined by
IIAII == inf{K 2: 0 IIIAxily :s Kllxllx, for all X EX},
which is equivalent to
IIAII == sup Ilxllx:IIIAxIlY.

XEX
IlxlliO
Problem A3.4. Prove the equivalence of these two definitions.

Note that we have assumed that if A is bounded, then D(A) = X, as follows
from Proposition A3.9. Moreover, we have
IIAxl1 :s IIAII IlxlI, x EX.
Examples A3.12 (with reference to Examples A3.2).

A3.2. Continuity and Boundedness of Linear Operators 305
o
FIGURE A3.1 The functions fn.
(1) All linear operators A : IR n -+ IRn are bounded. Note that there are actually
many norms on the set of n x n matrices. The most common are II A 1100 =
sUPi.j laij I, and IIAll lr =Tr(A* A)1/2. The norm defined in Definition A3.11
is the operator norm. This norm is distinguished by the fact that it satisfies
IIA* A II = IIA 11 2, where A * = AT and AT is the transpose of A.
(2) Let Af = af, with a, f E X == C([O, 1]). The multiplication operator
A is bounded in all norms on C([O, 1]) (although this space is a Banach
space only in the sup-norm). In the Banach space (C([O, 1]), II . 1100), we
have IIAII = lIalloo· To see this, note that for any f E X, IIAflloo =
lIafiloo :'S lIallooll flloo, and so IIAII :'S Iiall oo . Upon choosing f = 1 E X,
we have IIAfiloo = lIall oo , so lIalloo :'S IIAII.
(3) Let Af = f', f E C1([0, 1]), and X = C([O, 1]). Then A is unbounded. To
see this, we can, for example, note by Theorem A3.9 that if A was bounded
we could extend it to all X, but there are plenty of functions in X for which
l' does not exist! Alternately, consider a sequence of functions fn whose
graphs are given in Figure A3.1. Then each fn E C1([0, 1]), IIfn II = 1, but
II f~ II 00 -+ 00 as n -+ 00. Consequently, there exists no finite K > 0 such
that II Ain II :'S K II fn II 00 = K, for all n.
(4) Let Af(t) = J~ K(t, s)f(s)ds, where X = C([O, 1]) and K is continuous
on [0,1] x [0,1]. In the sup-norm on X, IIAII = SUPSE[O, 1] Jd IK(s, t)ldt.
Problem A3.5. Compute the bound on A given in Example A3.12 (4).
We now return to the algebraic structure on operators introduced in Section
A3.1. There, this structure was complicated by the fact that the domains depended
on the operators. In light of Proposition A3.9, we do not encounter this problem
if we restrict ourselves to bounded operators since we can then take the domain to
be all of X.
Definition A3.13. Let X and Y be two Banach spaces. We denote by L(X, Y) the
family of all bounded linear operators from X to Y.
Let us make a few remarks. First, in this definition, the space X need not be a
Banach space; an NLVS will do. Second, we will always assume that a bounded
operator on X has been extended so that its domain is all of X. Finally, if Y = X,
we write £(X) for £(X, X)).
Proposition A3.14. If A, B E £(X, Y) and c E C (or JR.), then cA and A + B E

£(X, Y) and
ileA II = IcIIiAII,
IIA + BII < IIAII + IIBII·
Moreover, if A E £(X, Y) and B E £(Y, Z), then BA E £(X, Y) and
IIABII :s IIAII IIBII·
Problem A3.6. Prove Proposition A3.14.
Corollary A3.15. £(X, Y) is an LVS over C (or JR.).

It follows from the last two results that II . II, as defined in Definition A3 .11, is a
norm on £(X, Y). This follows from Proposition A3.14 and the facts that IIAII :::: 0
and IIAII = 0 if and only if A = O. This shows that £(X, Y) is an NLVS. Hence
we can ask if £(X, Y) is complete in this norm.
Theorem A3.16. Let X be an NLVS and Y be a Banach space. Then £(X, Y) is a

Banach space with the norm II . II given in Definition A3. I I.
Proof. Let {An} be a Cauchy sequence in £(X, Y). We must show that there exist
A E £(X, Y) such that A = limn---+oo An. Take any x E X and consider the
sequence {Anx}. Then we have
by linearity and boundedness. Hence, {Anx} is a Cauchy sequence in Y, and as Y

is a Banach space, it converges. Let y = limn---+oo Anx, and define A : X --+ Y by
Ax =y.
Problem A3.7. Continue the proof of Theorem A3.16. Show that A is well de-
fined on X, linear, and bounded. (Hint: Use the fact that III An II - II Am III :s
II An - Am II, so { II An II } is Cauchy and hence converges. For boundedness of
A, use
II Ax II :s n--+oo
lim IIAn x II :s ( lim IIAn II)
n---+oo
IIx II·)
Completion of the proof. We show that limn---+oo An =A:

IIA - An II = suPII(An - A)xllllxll-',
x;<D
A3.3. The Graph of an Operator and Closure 307
and
II(An - A)xll IIAnx - Axil < lim IIAnx - Amxll

m---+oo
< (lim IIAn - Amll)

m---+oo
Ilxll.
So, by choosing n suitably large and using the Cauchy property of {An}, we have
lim ll --7oo II An - A II = O. 0
A3.3 The Graph of an Operator and Closure

We introduce a tool useful in the study of operators.
Definition A3.17. Let A: X --+ Y with domain D(A) C X. The graph of A,

Y == {(x,y)lx E X,y E Y}givenby
denotedr(A), is the subset of X x
rcA) == { (x, Ax) Ix E D(A) }.
Problem A3.8. Show that reX x Y is a linear subspace (we assume that X and
Y have the same field of scalars). Note that X x Y is an LVS under componentwise
addition and scalar multiplication.
We denote by {x} x Y == {(x, y)ly E Y) c X x Y. One may ask when a subset

reX x Y is the graph of some linear operator A: X --+ Y. Of course, r must
be a subspace of X x Y. A convenient condition on r is the following.
Proposition A3.18. A linear subspace reX x Y is the graph of some linear

operator A if and only if
({O) x Y) nr = {CO, O)}. (A3.1)
Proof.
(1) If there exists a linear operator A such that r = rCA) and (0, y) E r, then
y = A(O) = 0, so condition (A3.1) holds.
(2) Assume (A3.l) holds. Let D(A) = {x I (x, y) E r, for some y E Y}.Foreach
x E D(A), define Ax = y, where (x, y) E r. We want to show that A is well
defined and linear. Suppose there exist y, y' E Y such that (x, y) and (x, y')
are in r. Since r is linear, (x, y) - (x, y') = (0, y - y') E r and by (A3.1),
Y = yr. Hence, A is well defined. Second, if Xl, X2 E D(A) and A E C, then
(X] + X2, AXI + AX2) E r. This implies that (Xl + X2, A(XI + X2» E r, so
A(x] + X2) = AX2 + AX2. Since A(XJ, AXJ) = (Ax I , AAx]), A(AxI) = AAxl.
Hence, A is linear. D
Definition A3.19. Let X, Y be NLVS. For each (x, y) E X x Y, define
II(x, y)11 = Ilxlix + lIylly·
Lemma A3.20. (X x Y, II . II) is an NLVS.

Proof. We have seen that X x Y is an LVS. For the norm, we must check the
conditions of Definition A 1.3:
(i) II(x, y)11 ~ 0 since Ilxllx ~ 0, Ilylly ~ 0 and it follows that II(x, y)11 =0
if and only if x = 0 = y.
(ii) IIA(X, y)11 = II(Ax, Ay)1I = IAI(llxllx + IIYlly) = IAIII(x, Y)II·

(iii) II(x,y)+(x',y')11 = lI(x+x',y+y')11
= IIx +x'llx + Ily + illy
:s (1lxllx + IIYlly) + (lIx'llx + Ililly)
II(x, y)11 + II(x', i)ll. D
Problem A3.9. Prove that the map X x Y -+ lR defined by
II(x, y)IIE == [ IIxll~ + IIYII~]1

is a norm on X x Y. Show that the topology on X x Y determined by II . II E is
equivalent to the topology induced by the norm in Definition A3.19. This means
that an open set in one topology contains an open set in the other.
Problem A3.10. Show that if A E L(X, Y), then rCA) c X x Y is closed.
(We will see later that the converse is not true). (Hint: rCA) is closed if for any
convergent sequence {(x n , Axn )} in rCA), limn--+oo(xn , Axn ) = (x, y) E rCA),
that is, x E D(A) and Ax = y.)
Definition A3.21. Let A be an operator on X with D(A). An operator A with
domain D(A) is called the closure of A ifr(A) = rCA). An operator A is said to
be closable if it has a closure.
Remark A3.22. Suppose A is a linear operator with graph rCA). It may happen
(and does!) that rCA) is not the graph of any operator (precisely how this may
happen is given in Proposition A3.18). In this case A is not closable. Suppose,
however, that A is closable. Then the closure A is unique and satisfies Ax =
Ax, x E D(A). This follows directly from Definition A3.21.
An important connection between everywhere defined operators (i.e., D(A) =
X) and boundedness is given by the next result.
Theorem A3.23. (Closed graph theorem). Let X, Y be Banach spaces. Let A be

defined on X (i.e., D(A) = X). If A is closed (i.e., A = A), then A is bounded.
The proof of this is given in Reed and Simon, Volume I [RS I]. As the proof
requires more machinery, the Baire category theorem, we will not give it here.
A3.4. Inverses of Linear Operators 309
Note that there are everywhere defined, unbounded operators. By Theorem A3.23,
these will not be closed!
Examples A3.24.
(1) Let X = (C([O, 1]), 11·1100) and A = dldx with D(A) = CI([O, 1]), as in Ex-
ample A3.2 (3). We claim that (A, D(A» is closed. To see this, let (fn, Afn)
be a convergent sequence in rcA). Then fn ~ f in X and Afn = f~ ~ gin
X. By a standard result on uniform convergence, f E CI([O, 1]) and f' = g.
Consequently, f E D(A) and (f, Af) E rCA). We can modify this example
slightly to obtain a closable operator whose closure is (A, D(A». Simply
take Al = dldx and D(Ad = C 2([0, 1]).
ProblemA3.11.Provethat(AI, D(A I»is closable and that its closure is (A, D(A».
(2) Here is an example of an operator that is not closable. Let X = L 2(IR),

and choose some fo E X with lIfoll = 1. Define B, on D(B) = Cgo(IR),
by Bf = f(O)fo. Then (B, D(B» is a densely defined operator and is
not closable. Consider, for example, a function h E CO' ([ -1, 1]), h 2: 0,
h(O) = 1, and J h = 1. Let hn(x) == h(nx). Then IIh n II = lin ~ 0 as
n ~ 00. However, Bh n = hn(O)fo = fo, so IIBhnll = 1. This shows that
(0, fo) E reB), and so it cannot be the graph of an operator, by Proposition
A3.18.
A3.4 Inverses of Linear Operators

Let X and Y be Banach spaces (over <C for simplicity), and let A be a linear operator
from X to Y. We associate with A three linear subspaces:
D(A) c X, domain of A;
Ran(A) c Y, range of A, Ran(A) == {y E Y I y = Ax, some x E D(A)};
ker A C X, kernel ofker A == {x E D(A) I Ax = OJ.
Definition A3.25. An operator A -I : Ran(A) ~ D(A) is called the inverse of A

if A-I A = idD(A) (the identity map on D(A») and AA- I = idRan(A).
Theorem A3.26. An operator A has an inverse if and only if
ker A = {OJ.
Proof.
(1) {= Ifker A = {O}, then for any y E Ran(A), there exists a unique x E D(A)
such that Ax = y, for suppose XI, X2 E D(A) and AXI = Y = AX2. Then,
by linearity, AXI - AX2 = A(xi - X2) = 0, so XI - X2 E ker A. But

ker A = to}, whence XI = X2. Because of this we can define an operator
A-I: Ran(A) -+ D(A) by A-Iy == X, where Ax = y. The operator A-I
is well defined and linear. Moreover, A -I Y = A -I (Ax) = x, so A -I A =
idD(A), and for y E Ran(A), AA-Iy = Ax = y, so AA- I = idRan(A).
(2) ::::} Suppose A-I: Ran(A) -+ D(A) exists. If x E ker A, then x = A-I Ax =
0, soker A = {OJ. D
As this theorem shows, the condition for the existence of A-I is simply that
ker A = (OJ. This guarantees that the inverse map is a function. However, we have
little information about A -I. In particular, we would like A -I to be bounded on
all of Y.
Definition A3.27. A linear operator A is invertible if A has a bounded inverse
defined on all of Y.
Problem A3.12. Prove that if A is invertible, then A -1 is unique.
Example A3.28. Let X =C([O, 1]), and define A by
(Af)(t) = fat f(s)ds, f E X.
This operator A has an inverse by the fundamental theorem of calculus: A -I _

d / dt. But A is not invertible, since A -I is not bounded!
Warning: According to our definition, A may have an inverse but not be invertible.
Theorem A3.29. Let A and B be bounded, invertible operators. Then AB is
invertible and (ABr l = B- 1A -I.
Problem A3.13. Prove Theorem A3.29.

Theorem A3.30. Let T be a bounded operator with II T II < 1. Then 1 - T is
invertible, and the inverse is given by an absolutely convergent Neumann series:
00
(1- T)-I = L:T n ,

n=O
that is, limN->oo L~=o Tk converges in norm to (1 - T)-I.
Proof. Since L~=o IITkll < Lf:o IITll k , and the series Lk IITilk converges (as
II TIl < 1, it is simply a geometric series), the sequence of bounded operators
L~=O Tk is norm Cauchy. Since Y is a Banach space, this Cauchy sequence con-
verges to a bounded operator. Now we compute
00 00
(1 - T) L:Tk = L:(T k - Tk+I) = 1,

k=O k=O
A3.4. Inverses of Linear Operators 311
where the manipulations are justified by the norm convergence of the power series.
Similarly,
(fk=O
Tk) (1 - T) = 1,
so the series defines (1 - T)-I. 0
Theorem A3.31. Let A be invertible, and let B be bounded with II B II <
IIA-Ill- i . Then A + B is invertible and
(A+B)-I =(1+A- I B)-IA- I =A- I (l+BA- I )-I.
Proof. Write
A + B = A(l + A-I B),
and note that IIA -I BII :s IIA -III IIBII < I, so by Theorem A3.30, 1 + A -I B
is invertible (take T = -A -I B). By Theorem A3.29, the product A(l + A -I B) is
invertible and the first formula is obtained. By writing
A+B = (1 + BA-I)A,
we obtain the second formula. o
Example A3.32. Let X = Y = C([O, 1]), and let K(t, s) be a continuous function
on [0, 1] x [0, 1] (see Example A3.2 (4». We define a bounded operator K by
(Kf)(t) =lot K(t, s)f(s)ds.

Consider the equation on C([O, 1]):
f -)'Kf= fo,
where fo E
r
C([O, 1]) is fixed. We claim that the equation has a unique solution
for any fo E C([O, 1]) if 1).1 < IIKII- I = (sUPO:,:t:'01 f~ IK(t, s)lds l . Indeed, in
that case II)'K II < I, so 1 - )'K is invertible by Theorem A3.30 and we have
f =(1 - )'K)-I fo,
which is easily seen to be a solution. It is unique since (1 - )'K) is invertible.

Moreover, the Neumann expansion for (1 - )'K)-I gives
00
f= L).iKifo,
i=O
which is useful since it provides a norm-approximate solution
N
fN = L).i Ki fo.
i=O
This problem can also be solved by applying the Fredholm alternative, Theorem
9.12.
A3.5 Different Topologies on LeX)

Let X be a Banach space. We recall that L(X), the LVS of all bounded operators
from X into itself, is a Banach space with the norm defined in Definition A3.ll.
According to section 2 of Appendix I, the norm induces a topology on L(X), called
the norm or uniform topology. There are, however, other important topologies on
L(X), and we discuss the weak and strong topologies here. As these topologies are
not induced by a metric, they are not completely described by the convergences of
sequences. However, it is this property that will be important for us, so we simply
give the conditions for a sequence in L(X) to converge weakly or strongly.
Definition A3.33. Let {Bn} be a sequence of bounded operators (i.e., Bn E L(X)).
(I) Bn converges uniformly or in norm to B E L(X) if II Bn - B I -;. 0, as

n -;. 00.
(2) Bn converges strongly to B E L(X) iffor all x E X, II Bllx - Bx II -;. 0, as

n -;. 00.
We note that since L(X) is a Banach space, if {Bn} converges uniformly, then
the limit always exists and belongs to L(X). To describe the third notion of con-
vergence, weak convergence, we must introduce the LVS of linear functionals on
a Banach space X.
Definition A3.34. A map I: X -;. ~ is called a bounded linear functional if I is
linear on X, that is, for any x, y E X and A E ~ (or C)
I(Ax + y) = Al(x) + l(y),
and 3Kt > ° such that for all x E X, II(x)1 :::: Ktllx II. The set of all bounded
linear functionals on X is denoted by X*, the dual of X.
It is an elementary fact that X* is a Banach space. Here, we are interested in the
following definition.
Definition A3.35. A sequence {Bn} in L(X) converges weakly to B E L(X) iffor
all I E X* and x EX, [1(Bnx) -1(Bx)] = I(Bnx - Bx) -;. 0, as n -;. 00.
Problem A3.14. Prove that uniform convergence =} strong convergence =} weak
convergence.
Appendix 4. The Fourier Transform,
Sobolev Spaces, and Convolutions
A4.1 Fourier Transform

A very important transformation on L2(JR n ) is the Fourier transform. We develop
the main points of the theory here.
Definition A4.1. For any f E S(JR n ), define the Fourier transform j(k) by
j(k) == (2n) 2" f e- ik .x f(x)dx, (A4.1)
where k· x = I:7=1 kiXi·

Problem A4.1. Show that j E S(JR n ) and that f -+ j is a linear transformation
on S(JRn ). Verify that (A4.l) can be extended to L 1(JR n ).
Since S is dense in L 2(JR), we would like to extend the Fourier transform to all
of L 2(JRn). By Appendix 3, we can do this provided the transformation is bounded
on a dense set in L 2(JRn). Consider the set E of all finite linear combinations of
functions of the form p(x )e- a (x- fJ )2 /2, for a > 0, f3 E JR, and p any polynomial
in x. Such functions are in S(JRn). We compute the Fourier transform of a typical
one. Without loss of generality, we can take a = 1 and f3 = 0. Then, setting
g(x) = p(x)e- x2 / 2 , we have
314 Appendix 4. The Fourier Transform, Sobolev Spaces, and Convolutions
Due to the simple fact that
(J/Jki)e- ik .x = -ixie-ik.x, (A4.2)
we have
where D stands for any partial derivative in ki. By the fast convergence of the
integral, we can write
The resulting integral can be computed explicitly by contour deformation.
Problem A4.2. Prove the identity
where J~oo e- x2 dx = rr 1/2.

Consequently, we obtain
We want to compute the L 2- norm of g:
IIgll 2 = f ig(k)i 2 dk
= f Ip(iD)e-~f dk
(t, p(iD)p(iD)t) ,
where we have integrated by parts and write t(k) == e- k2 /2. Retracing the above
steps, the inner product can be written as
f dk((2rr)-~ f e-ik.Xe-4dx)* (P(iD)p(iD)e-~)

= f dk ( (2rr)-211 p(i D) f·e-1k-x e - .z\ dx
2 ) * ( p(iD)e- zk 2 )
= f dk (2rr)-~ f e- ik .x g(X)dX)* (p(iD)e-~)

f dx g(x)* . (2rr)-~
IIgll~ .
A4.1. Fourier Transform 315
The reader can verify that the integration by parts and the interchange of the order
of integration are all valid due to the absolute convergence of the integral and the
superexponential vanishing of the integrand as Ilx II or Ilk II -+ 00. Because of the
linearity of the map fEE -+ J E E, we have shown for any fEE,
IIfl12 = IIJI12.
Now E is dense in L2(]Rn), since the Hermite functions form a basis for L2(]Rn).
Hence, we have proven the next result.
Theorem A4.2. The map f E S(]Rn) -+ J E S(]Rn), defined by the absolutely

convergent integral (A4.1), extends to a bounded, linear transformation F on
L 2(]Rn) with
IIFfl12 = Ilf112, (A4.3)
for any f E L 2(]Rn). Furthermore, F is invertible with inverse F- 1 defined on E
by
f
so that
g(x) = (271T~ eik .x g(k)dk. (A4.4)
Remarks A4.3.
(I) The identity (A4.3) is referred to as the Plancherel theorem.
(2) Formula (A4.4) is called the Fourier inversion formula. Formulas (A4.1) and
(A4.4) are not defined for arbitrary f E L2(]Rn). It can be shown, however,
that the functions defined for R > 0 and f E L2(]Rn) by
actually converge strongly to Ff as R -+ 00.
(3) Because F is an invertible isometry, it is a unitary transformation on L 2(]Rn).

As a consequence, we have
(Ff, Fg) = (f, g),
Finally, we note that the simple identity (A4.2) and the related identity
( a)
-
aXi
e -ik·x = - I·kie -ik·x (A4.S)
show the utility of the Fourier transform for solving constant coefficient partial dif-
ferential equations on 1K1l. In particular, if g E S(lKll ), then denoting the Laplacian
by -~,
we obtain
(A4.6)
Using the inversion formula, we can write the left side of (A4.6) as
or
(A4.7)
valid on the domain of the Laplacian in L2(1K1l). This identity will be useful in our
study of Sobolev spaces in the next section.
A4.2 Sobolev Spaces

We collect here basic results on the Sobolev spaces HS,P(lK ll ) associated with
1K1l. There are two approaches to this material: distributional derivatives and the
Fourier transform. To introduce the notion of a weak or distributional derivative,
let uS first consider f E Cg"(lKll) and suppose g is simply in CI(lKll). Then the
integration-by-parts formula is
f f(X)(::)(X)=- f (:~)(X)g(X). (A4.8)
Note that the right side of (A4.8) makes sense even if g is not differentiable.
Definition A4.4. Let g E LP(lK ll ). We say that g has a weak or distributional

derivative with respect to Xi in U' if there exists </> E LP(lK ll ) such that for any
f E Cg"(lKll),
f f(x)</>(x)dx =- f (:~) (x)g(x)dx.
By Holder's inequality, Theorem A2.4, both sides converge. We call </> the LP-
derivative of g with respect to Xi. In a similar manner, we can define higher-order,
weak LP -derivatives.
We can now define Sobolev spaces as the completion of Cg"(lKll) with respect
to the following norm. Let a = (aI, ... , all) denote an n-tuple of positive integers,
A4.3. Convolutions 317
and let la I == L:7=1 ai ~ O. For 1 :::: p < 00 and sEN, we define for any
f E cgoORn),
IIfIIs,p = [
nil aai f II P]
I~s ~ ax~i P
~ (A4.9)
Problem A4.3. Show that (Cg"(JR n), II . IIs,p) is an NLVS.
Definition A4.S. The pth Sobolev space of order s, H s, p (JR n), is the Banach space
obtained as the completion of (Cg"(JR n ), II . IIs,p).
As in the case of LP(JRn), it can be shown that HS,P(JR n ) is the set of all func-
tions f E LP that have distributional LP -derivatives up to and including order s.
Additionally, when p = 2, we write H s (JR n ) for the Sobolev space of order s.
Problem A4.4. Show that HS(JR n ) is a Hilbert space.
The Fourier transform and formulas (A4.2) and (A4.5) allow us to characterize
HS(JRn ) as follows. For f E S(JRn ), we define
(A4.1O)
These two norms (A4.9) and (A4.1O) are equivalent in that they have the same
Cauchy sequences (see Problem A3.9). Consequently, the completion yields the
same Banach space HS(JR n ).
Problem A4.S. Verify that the norms (A4.9) and (A4.1O) on S(JRn ) are equiva-
lent.
Functions in the Sobolev spaces S(JRn ) become more regular with increasing
index s. We will state one version of a theorem of this type. This is an example
of a body of results called the Sobolev embedding theorems. These theorems are
used to prove the regularity of solutions to elliptic differential equations.
Theorem A4.6. Any function f E Hs+k(JR n), for s > nl2 and k ~ 0, can be
represented by afunction in Ck(JRn ).
A4.3 Convolutions
We recall that for two functions f, g E Cg"(JRn ), the convolution of f and g,
denoted by f * g, is defined by
(f * g)(x) = f I(x - y)g(y)dy.
To extend this definition to LP -spaces, we can use Young's inequality.
Theorem A4.7. (Young's inequality). Let p, q, and r be nonnegative real num-

bers, and suppose lip + llq = 1+ llr. Then if I E LP(lRn ) and g E U(JR n ),
the convolution I * g E U (JR n ) and satisfies
(A4.11)
Convolutions of functions in S(JRn ) behave nicely under the Fourier transform.

For I, g E S(JRn ), we have
fg(k) = f * g(k). (A4.12)
Problem A4.6. Verify the identity (A4.12).

References
[AF] Agler, J., and Froese, R.: Existence of Stark Ladder Resonances, Commun.
Math. Phys. 100, 161-171 (1985).
[AC] Aguilar, J., and Combes, 1. M.: A Class of Analytic Perturbations for
One-Body Schrodinger Hamiltonians, Commun. Math. Phys. 22, 269-279
(1971).
[Agl] Agmon, S.: Lectures on Exponential Decay of Solutions of Second-Order

Elliptic Equations, Princeton University Press, Princeton, NJ, 1982.
[Ag2] Agmon, S.: Lectures on Elliptic Boundary Value Problems, Van Nostrand,
Princeton, NJ, 1965.
[AgHeSk] Agmon, S., Herbst, I. w., and Skibsted, E.: Perturbation of Embedded
Eigenvalues in the Generalized N-Body Problem, Commun. Math. Phys.
122,411-438 (1989).
[Ah] Ahia, F.: Lower Bounds on Width of Stark Resonances in One Dimension,
Letts. Math. Phys. 24,21-29 (1992).
[AG] Akhiezer, N. I., and Glazman, I. M.: The Theory of Linear Operators in
Hilbert Space, Vols. 1 & 2, Frederick Ungar, New York, 1963.
[AJS] Amrein, W. 0., Jauch, J. M., and Sinha, K. B.: Scattering Theory in Quan-
tum Mechanics, Lecture Notes and Supplements in Physics, Vol. 16, W. A.
Benjamin, Inc., London, 1977.
320 References
[Ar] Arsen'ey, A. A.: On the Singularities of Analytic Continuations and Reso-

nances Properties of a Solution of the Scattering Problem for the Helmholtz
Equation, Zh. Vyohisl. Mat. i Mat. Fiz. 12, 112-138 (1972).
[AH] Ashbaugh, M., and Harrell, E. M.: Perturbation Theory for Shape Reso-
nances and Large Barrier Potentials, Commun. Math. Phys. 83, 151-170
(1982).
[Ay] AYron, J. E.: Bender-Wu Formulas for the Zeeman Effect in Hydrogen, Ann.
Phys. 131,73-94 (1981).
[AyHe] AYron, J. E., and Herbst, I. W.: Spectral and Scattering Theory of
Schrodinger Operators Related to the Stark Effect, Commun. Math. Phys.
52, 239-254 (1977).
[AHS] AYron, J. E., Herbst, I. w., and Simon, B.: Schrodinger Operators with
Magnetic Fields, I. General Interactions, Duke Math. 1. 45,847 (1978).
[BFS] Bach, v., Frohlich, 1., and Sigal, I.M.: Mathematical Theory of Nonrela-
tiyistic Matter and Radiation, Letters in Math. Phys.34, 183-201 (1995).
[B 1] Balsley, E.: Analytic Scattering Theory for 2-body Schodinger operators, 1.

Funct. Anal. 29,375-396 (1978).
[B2] Balsley, E.: Local Spectral Deformation Technique for Schrodinger Opera-
tors, 1. Funct. Anal. 58, 79-105 (1984).
[B3] Balsley, E.: Analyticity Properties of Eigenfunctions and the Scattering Ma-
trix, Commun. Math. Phys. 114,599 (1988).
[BS1] Balsley, E., and Skibsted, E.: Resonance Theory of Two-Body Schrodinger
Operators, Ann. Inst. Henri Poincare 51,2 (1989).
[BS2] Balsley, E., and Skibsted, E.: Asymptotic and Analyticity Properties of
Resonance Functions, Ann. Inst. Henri Poincare 53, 123-137 (1990).
[BB] Babbitt, D., and Balsley, E.: Local Distortion Techniques and Unitarity of
the S-Matrix for the 2-Body Problem, 1. Math. Anal. Appl. 54, 316-347
(1976).
[BC] Balsley, E., and Combes, 1. M.: Spectral Properties of Many Body
Schodinger Operators with Dilation Analytic Interactions, Commun. Math.
Phys. 22,280-294 (1971).
[Be] Beale, J. T.: Scattering Frequencies of Resonators, Commun. Pure Appl.

Math. 26,549-563 (1973).
[Ben] Bentosela, F.: Bloch Electrons ill Constant Electric Field, Commun.
Math. Phys. 68, 173-182 (1979).
References 321
[BCDSSW] Bentosela, F., Carmona, R., Duclos, P., Simon, B., Soulliard, B., and
Weder, R.: SchrOdinger Operators with an Electric Field and Random or
Deterministic Potentials, Commun. Math. Phys. 88,387-397 (1983).
[BG] Bentosela, F., and Grecchi, Y.: Stark Wannier Ladders, Commun. Math.
Phys. 142, 169-192 (1991).
[Br] Briet, P.: Bender Wu Formula for the Zeeman Effect, 1. Math. Phys.36, 3871
(1995).
[BCD1] Briet, P., Combes, J. M., and Duclos, P.: On the Location of Resonances
for Schrodinger Operators in the Semiclassical Limit, I. Resonance Free
Domains, 1. Math. Anal. Appl. 126,90-99 (1987).
[BCD2] Briet, P., Combes, J. M., and Duclos, P.: On the Location of Resonances
for Schrodinger Operators, II. Barrier Top Resonances, Commun. Partial
Diff. Equations 12, 201-222 (1987).
[BCD3] Briet, P., Combes, J. M., and Duclos, P.: Spectral Stability under Tunnel-
ing, Commun. Math. Phys. 126, 133-156 (1989).
[BHM1] Brown, R. M., Hislop, P. D., and Martinez, A.: Lower Bounds on the
Interaction between Cavities Connected by a Thin Tube, Duke Math. 1. 73,
163-176, (1994).
[BD] Buslaev, Y. S., and Dmitrieva, L. A.: Spectral Properties of the Bloch Elec-
trons in External Fields, Berkeley preprint PAM-477, University of Califor-
nia, 1989.
[CS] Carmona, R., and Simon, B.: Pointwise Bounds On Eigenfunctions and Wave
Packets in N-Body Quantum Systems, Commun. Math. Phys. 80, 59-98
(1981).
[CDS] Combes, 1. M., Duclos, P., and Seiler, R.: Krein's Formula and One-
Dimensional Multiple Wells, 1. Funct. Anal. 52,257-301 (1983).
[CDKS] Combes, J. M., Duclos, P., Klein, M., and Seiler, R.: The Shape Reso-
nance, Commun. Math. Phys. 110,215-236 (1987).
[CH] Combes, J. M., and Hislop, P. D.: Stark Ladder Resonances for Small Elec-
tric Fields, Commun. Math. Phys. 140,291-320 (1991).
[CHI] Combes, J. M., and Hislop, P. D.: Localization for Some Continuous, Ran-
dom Hamiltonians in d-Dimensions, 1. Funct. Anal. 124, 149-180 (1994).
[CT] Combes, J. M., and Thomas, L.: Asymptotic Behavior of Eigenfunctions

for Multiparticle Schrodinger Operators, Commun. Math. Phys. 34, 251-
276 (1973).
322 References
[Cy] Cycon, H. L.: Resonances Defined by Modified Dilations, Helv. Phys. Acta
58, 969-981 (1985).
[CFKS] Cycon, H. L., Froese, R. G., Kirsch, w., and Simon, B.: Schrodinger
Operators, with Application to Quantum Mechanics and Global Geometry,
Springer-Verlag, New York, 1986.
[Da] Davies, E. B.: Metastable States of Molecules, Commun. Math. Phys. 75,
263-283 (1980).
[DeBH] DeBievre, S., and Hislop, P. D.: Spectral Resonances for the Laplace-
Beltrami Operator, Ann. Inst. Henri Poincare 48, 105-146 (1988).
[DeBHS] DeBievre, S., Hislop, P. D., and Sigal, I. M.: Scattering Theory for
the Wave Equation on Non-Compact Manifolds, Reviews in Mathematical
Physics 4,575-618 (1992).
[DHSV] Deift, P., Hunziker, w., Simon, B., and Vock, E.: Pointwise Bounds on
Eigenfunctions and Wave Packets in N-Body Quantum Systems IV, Com-
mun. Math. Phys. 64, 1-34 (1978).
[EG] Evans, L. c., and Gariepy, R. F.: Measure Theory and Fine Properties of
Functions, CRC Press: Boca Raton, FL, 1991.
[Fef] Fefferman, C. L.: The N -Body Problem in Quantum Mechanics, Commun.

Pure App!. Math. 39 S, S67-S109 (1986).
[Fe] Fernandez, C. A.: Resonances in Scattering by a Resonator, Indiana Univ.

Math. 1. 34,115-125 (1985).
[FL] Fernandez, C. A., and Lavine, R.: Lower Bounds for Resonance Widths
in Potential and Obstacle Scattering, Commun. Math. Phys. 128, 263-284
(1990).
[FrHe] Froese, R. G., and Herbst, I. W.: Exponential Bounds and Absence of
Positive Eigenvalues for N-Body SchrOdinger Operators, Commun. Math.
Phys. 87,429-447 (1982).
[FHHO 1] Froese, R., Herbst, I., Hoffmann-Ostenhof, M., Hoffmann-Ostenhof, T.:

On the Absence of Positive Eigenvalues for One-Body Schrodinger Opera-
tors. 1. d'Ana!. Math. 41, 272-284 (1982).
[FHH02] Froese, R., Herbst, I., Hoffmann-Ostenhof, M., and Hoffmann-

Ostenhof, T.: L 2-Exponential Lower Bounds to Solutions of the Schrodinger
Equation, Commun. Math. Phys. 87,265-286 (1982).
[FrHi] Froese, R. G., and Hislop, P. D.: Spectral Analysis of Second Order Elliptic
Operators on Noncompact Manifolds, Duke Math. 1. 58, 103-129 (1989).
References 323
[FrSp] Frohlich, J., and Spencer, T.: Absence of Diffusion in the Anderson Tight
Binding Model for Large Disorder or Low Energy, Commun. Math. Phys.
88,151-184 (1983).
[Gadl] Gadyl'shin, R. R.: On Acoustic Helmholtz Resonator and on Its Electro-

magnetic Analogy, 1. Math. Phys. 35, 3464-3481 (1994).
[Gad2] Gadyl'shin, R. R.: Asymptotics of Scattering Frequencies with Small

Imaginary Parts for an Acoustic Resonator, Math. Model. Numer. Anal.
28, 761-780 (1994).
[Ga] Gamow, G.: Zur Quantentheorie des Atomkemes, z. Phys. 51, 204-212
(1928).
[G 1] Gerard, C.: Resonances in Atom-Surface Scattering, Commun. Math. Phys.

126,263-290 (1989).
[G2] Gerard, C.: Resonance Theory for Periodic SchrMinger Operators, Bull.
Soc. Math. Fr. 118,27-54 (1990).
[G3] Gerard, C.: Semiclassical Resolvent Estimates for Two- and Three-Body
Schrodinger Operators, Commun. Partial Diff. Eq. 15, 1161-1178 (1990).
[GG] Gerard, C., and Grigis, A.: Precise Estimates of Tunneling and Eigenvalues
Near a Potential Barrier, J. Diff. Eq. 72, 149-177 (1988).
[GM1] Gerard, C., and Martinez, A.: Principe d' Absorption Limite pour les
Operateurs de Schrodinger a Longue Portee, C. R. Acad. Sci. Paris 306,
121-123 (1988).
[GM2] Gerard, c., and Martinez, A.: Prolongement Meromorphic de la Matrice

de Scattering pour les Problemes a Deux Corps a Longue Portee, Ann. Inst.
Henri Poincare 31,81-110 (1989).
[GMR] Gerard, C., Martinez, A., and Robert, D.: Breit-Wigner Formulas for the
Scattering Phase and the Total Cross-Section in the Semiclassical Limit,
Commun. Math. Phys. 121, 323-336 (1989).
[GS] Gerard, C., and Sigal, I. M.: Space-Time Picture of Semiclassical Reso-
nances, Commun. Math. Phys. 145,281-328 (1992).
[GSj] Gerard, c., and Sjostrand, J.: Semiclassical Resonances Generated by a

Closed Trajectory of Hyperbolic Type, Commun. Math. Phys. 108, 391-
421 (1987).
[Gr] Graf, G. M.: The Mourre Estimate in the Semiclassical Limit, Lett. Math.
Phys. 20,47-54 (1990).
[GrGr] Graffi, S., and Grecchi, Y.: Resonances in the Stark Effect and Perturbation
Theory, Commun. Math. Phys. 62, 83-96 (1978)
324 References
[GC] Gurney, R. w., and Condon, E. U.: Nature 122439 (1928).

[Ha] Hagedorn, G.: A Link between Scattering Resonances and Dilation-
Analytic Resonances in Few-Body Quantum Mechanics, Commun. Math.
Phys. 65, 181-201 (1979).
[Hal] Halmos, P.: Introduction to Hilbert Space and the Theory of Spectral Mul-
tiplicity, Chelsea Publishing Company, New York, 1957.
[HI] Harrell, E. M.: Double Wells, Commun. Math. Phys. 75, 239-261 (1980).
[H2] Harrell, E. M.: General Lower Bounds for Resonances in One Dimension,
Commun. Math. Phys. 86, 221-225 (1982).
[HaSi] Harrell, E. M., and Simon, B.: Mathematical Theory of Resonances Whose
Widths Are Exponentially Small, Duke Math. J. 47, 845 (1980).
[Hel] Helffer, B.: Semiclassical Analysis of Schrodinger Operators and Applica-

tions, Springer Lecture Notes in Mathematics No. 1336, Springer-Verlag,
Berlin, 1988.
[HeM] Helffer, B., and Martinez, A.: Comparaison entre les Diverses Notions de
Resonances, Helv. Phys. Acta 60, 992-1003 (1987).
[HR] Helffer, B., and Robert, D.: Puits de Potentiel Generalises et Asymptotique
Semi-classique, Ann. Inst. Henri Poincare 41,291-331 (1984).
[HSjl] Helffer, B., and Sjostrand, J.: Multiple Wells in the Semi-classical Limit
I. Commun. in P.D.E. 9,337-408 (1984).
[HSj2] Helffer, B., and Sjostrand, J.: Puits Multiples en Limite Semi-classique,
II. Interaction de Moleculaire, Symetries, Perturbation. Ann. Inst. Henri
Poincare 42, 127-212 (1985).
[HSj3] Helffer, B., and Sjostrand, J.: Multiple Wells in the Semi-classical Limit,
III. Interaction through Non-resonant Wells, Math. Nachrichte 124, 263-
313 (1985).
[HSj4] Helffer, B., and Sjostrand, J.: Resonances en Limite Semi-classique. Mem-
oire de la SMF, Nos. 24-25, 114 (1986).
[Hel] Herbst, I. W.: Schrodinger Operators with External Homogeneous Electric

and Magnetic Fields, in Rigorous Atomic and Molecular Physics, Proc.
fourth Int. School of Math. Phys., Enrice, G. Velo and A. S. Wightman,
eds., Plenum, New York, 1981.
[He2] Herbst, I. W.: Perturbation Theory for the Decay Rate of Eigenfunctions
in the Generalized N-Body Problem, Commun. Math. Phys. 158,517-536
(1993).
References 325
[He3] Herbst, I. w.: Dilation Analyticity in Constant Electric Field, I. The Two-
Body Problem, Commun. Math. Phys. 64, 279-298 (1979).
[He4] Herbst, I. w.: Exponential Decay in the Stark Effect, Commun. Math. Phys.
75, 197-205 (1980).
[HH] Herbst, I. W., and Howland, J.: The Stark Ladder and Other One-
Dimensional External Field Problems, Commun. Math. Phys. 80, 23-42
(1981).
[HiSm] Hirsch, M., and Smale, S.: Differential Equations, Dynamical Systems,
and Linear Algebra, Academic Press, Boston, 1974.
[H] Hislop, P. D.: Singular Perturbations of Dirichlet and Neumann Domains

and Resonances for Obstacle Scattering, Asterisque 210, 197-216 (1992).
[His] Hislop, P. D.: The Geometry and Spectra of Hyperbolic Manifolds, Proceed-
ings of the Indian Academy of Sciences; Math. Sci. 104,715-776 (1994).
[HM] Hislop, P. D., and Martinez, A.: Scattering Resonances of a Helmholtz

Resonator, Indiana Univ. Math. 1. 40,767-788 (1991).
lHN] Hislop, P. D., and Nakamura, S.: Semiclassical Resolvent Estimates, Ann.
Inst. Henri Poincare 51, 187-198 (1989).
[HSl] Hislop, P. D., and Sigal, I. M.: Shape Resonances in Quantum Mechanics,
in Differential equations and Mathematical Physics, I. Knowles, ed., Lect.
Notes in Math. Vol. 1285, Springer-Verlag, New York, 1986.
[HS2] Hislop. P. D., and Sigal, I. M.: Semiclassical Theory of Shape Resonances
in Quantum Mechanics, Memoirs of the AMS 399,1-123 (1989).
[HOAM] Hoffmann-Ostenhof, M., Hoffmann-Ostenhof, T., Ahlichs, R., and

Morgan, 1.: On Exponential Fall-Off of Wave Functions and Electron Den-
sities, in Mathematical Problems in Theoretical Physics, Lecture Notes in
Physics 116, K. Osterwalder, Ed., Springer-Verlag, Berlin, 1980.
[HoI] Howland, 1.: Puiseux Series for Resonances at an Embedded Eigenvalue,

Pacific 1. Math. 55, 157-176 (1974).
[Ho2] Howland, J. S.: Imaginary Part of a Resonance in Barrier Penetration, 1.

Math. Anal. Appl. 86,507-517 (1982).
[Hu 1] Hunziker, W.: Schrodinger Operators with Electric or Magnetic Fields,

in Mathematical Problems in Theoretical Physics, Lecture Notes in Physics,
Vol. 116, K. Osterwalder, ed., Berlin, Springer-Verlag, 1979.
[Hu2] Hunziker, W.: Distortion Analyticity and Molecular Resonance Curves,

Ann. Inst. Henri Poincare 45,339-358 (1986).
326 References
[Hu3] Hunziker, W.: Resonances, Metastable States, and Exponential Decay Laws
in Perturbation Theory, Commun. Math. Phys. 132, 177-188 (1990).
[Hu4] Hunziker, W.: Notes on Asymptotic Perturbation Theory for Schrodinger

Eigenvalue Problems, Helv. Phys. Acta 61, 257-304 (1988).
[Hu5] Hunziker, w., and Pillet, C. A.: Degemerate Asymptotic Perturbation The-
ory, Commun. Math. Phys. 90,219-233 (1983).
[JP] Jaksi6, v., and Pillet, C. A.: On a Model for Quantum Friction I, Fermi's
Golden Rule and Dynamics at Zero Temperature, Ann. Inst. Henri Poincare
62,47-68 (1995).
[11] Jensen, A.: Local Distortion Technique, Resonances and Poles of the S-
Matrix, J. Math. Anal. Appl. 59,505-513 (1977).
[J2] Jensen, A.: Resonances in an Abstract Analytic Scattering Theory, Ann.

Inst. Henri Poincare 33,209-223 (1980).
[13] Jensen, A.: High Energy Resolvent Estimates for Generalized N-Body
SchrMinger Operators, Publ. RIMS, Kyoto Univ. 25, 155-167 (1989).
[J4] Jensen, A.: Bounds on Resonances for Stark-Wannier and Related Hamil-
tonians,1. Oper. Th. 19,69-80 (1988).
[J5] Jensen, A.: Perturbation results for Stark Effect Resonances, J. Reine Angew.
Math. 394,168-179 (1989).
[JLMSl] Jona-Lasinio, G., Martinelli, F., and Scoppola, E.: New Approach in
the Semi-classical Limit of Quantum Mechanics, I. Multiple Tunnelings in
One-Dimension, Commun. Math. Phys. 80,223 (1981).
[JLMS2] Jona-Lasinio, G., Martinelli, F., and Scoppola, E.: Multiple Tunnelings
in d-Dimensions: A Quantum Particle in a Hierarchical Model, Ann. Inst.
Henri Poincare 43,2(1985).
[KR] Kaidi, N., and Rouleux, M.: Multiple Resonances in the Semi-Classical
Limit, Commun. Math. Phys. 133,617-634 (1990).
[K] Kato, T.: Perturbation Theory for Linear Operators, second edition,
Springer-Verlag, Berlin, 1980.
[KS1] Kirsch, W., and Simon, B.: Universal Lower Bounds on Eigenvalue Split-
ting for One-Dimensional SchrMinger Operators, Commun. Math. Phys.
97,453-460 (1985).
[KS2] Kirsch, w., and Simon, B.: Comparison Theorems for the Gap of
SchrMinger Operators, J. Funct. Anal. 75, 396-410 (1987).
References 327
[KIl] Klein, M.: On the Absence of Resonances for SchrOdinger Operators with
Nontrapping Potentials in the Classical Limit, Commun. Math. Phys. 106,
485-494 (1986).
[KI2] Klein, M.: On the Mathematical Theory of Predissociation, Ann. of Physics

178,48-73 (1987).
[KIp] Klopp, F.: Resonances for Perturbations of a Semi-classical Periodic

SchrOdinger Operator, Arkiv for Math. 32, 323-371 (1994).
[Kn] Knopp, A: Theory ofAnalytic Functions, Dover Publications, Mineola, NY,

1966.
[LL] Landau, L. D., and Lifshitz, E. M.: Quantum Mechanics, Pergamon Press,
Oxford, 1977.
[La] Lavine, R.: Spectral Density and Sojourn Times, in Atomic Scattering The-
ory, J. Nuttall, ed., University of Western Ontario Press, London, Ontario,
1978.
[LP] Lax, P. D., and Phillips, R. S.: Scattering Theory, revised edition, Academic
Press, Inc., Boston, 1989.
[Li] Lieb, E. H.: Stability of Matter, Review of Modern Physics 48, 553-569
(1976).
[Ma] Marsden, J. E.: Basic Complex Analysis, W. H. Freeman and Company, San
Francisco, 1973.
[Ml] Martinez, A.: Estimations de l'Effet Tunnel pour Ie Double Puits, I. J. Math.
Pures Appl. 66, 195-215 (1987).
[M2] Martinez, A: Estimations de l'Effet Tunnel pour Ie Double Puits, II. Etats
Hautement Exites, Bull. Soc. Math. Fr. 116, 199-229 (1988).
[M3] Martinez, A: Resonances dans I' Approximation de Born-Oppenheimer I,

J. Diff. Eq. 91,204-234 (1991).
[M4] Martinez, A: Resonances dans l' Approximation de Born-Oppenheimer II.

Largeur des Resonances, Commun. Math. Phys. 135,517-530 (1991).
[MR] Martinez, A, and Rouleux, M.: Effet Tunnel entre Puits Degeneres, Comm.
Partial Differential Equations 13, 1157-1187 (1988).
[Mas] Maslov, V. P., and Nasaikinskii, V. E.: Asymptotics of Operator and Pseudo-
Differential Equations, Contemporary Soviet Mathematics, Consultants Bu-
reau, New York, 1988.
[Mol] Morawetz, C.: Exponential Decay for Solutions of the Wave Equation,
Commun. Pure Appl. Math. 19,439-444 (1966).
328 References
[Mo2] Morawetz, c.: Notes on Time Decay and Scattering for Some Hyperbolic
Problems, SIAM, Philadelphia, PA, 1975.
[Nl] Nakamura, S.: A Remark on Eigenvalue Splittings for One-Dimensional

Double-Well Hamiltonians, Letters in Math. Phys. 11,337-340 (1986).
[N2] Nakamura, S.: Scattering Theory for the Shape Resonance Model I, Ann.
Inst. Henri Poincare 50,115-131 (1989).
[N3] Nakamura, S.: Scattering Theory for the Shape Resonance Model II, Ann.
Inst. Henri Poincare 50, 132-142 (1989).
[N4] Nakamura, S.: Shape Resonances for Distortion Analytic Schrodinger Op-
erators, Commun. Part. Diff Eqns. 14, 1385-1419 (1989).
[N5] Nakamura, S.: Distortion Analyticity for Two-Body Schrodinger Operators,

Ann. Inst. Henri Poincare 53, 149-157 (1990).
[N6] Nakamura, S.: A Note on the Absence of Resonances for Schrodinger Op-
erators, Letters in Math. Phys. 16,217-233 (1988).
[NN1] Nenciu, A., and Nenciu, G.: Dynamics of Bloch Electrons in External
Electric Fields, I. Bounds for Interband Transitions and Effective Wannier
Hamiltonians, 1. Phys. A14, 2817-2827 (1981).
[NN2] Nenciu, A., and Nenciu, G.: Dynamics of Bloch Electrons in External
Electric Fields, II. The Existence of Stark Wannier Ladder Resonances, 1. Phys.
A15, 3313-3328 (1982).
[Ne] Newton, R. G.: Scattering Theory of Waves and Particles, second edition,
[OC] O'Connor, T.: Exponential Decay of Bound State Wave Functions, Commun.
Math. Phys. 32, 319-340 (1973).
[OY] Okamoto, T., and Yajima, K.: Complex Scaling Technique in Nonrelativistic
QED, Ann. Inst. Henri Poincare 42,311-327 (1985).
[Op] Oppenheimer, R. 1.: Three Notes on the Quantum Theory of Aperiodic

Effects, Phys. Rev. 31,66-81 (1928).
[Or] Orth, A.: Quantum Mechanical Resonance and Limiting Absorption: The
Many Body Problem, Commun. Math. Phys. 126,559-573 (1990).
[Ou] Outassourt, A.: Analyse Semi-c1assique pour l'Operateur de Schrodinger a

Potential Periodique, 1. Funct. Anal. 72, 1 (1987).
[Per] Persson, A.: Bounds for the Discrete Part of the Spectrum of a Semi-bounded
Schrodinger Operator, Math. Scand. 8, 143-153 (1960).
References 329
[PS] Petkov, v., and Stoyanov, L.: Geometry ofReflecting Rays and Inverse Spec-
tral Problems, John Wiley & Sons, Chichester, 1992.
[RS1] Reed, M., and Simon, B.: Methods of Modern Mathematical Physics, I.
Functional Analysis, second edition, Academic Press, New York, 1980.
[RS2] Reed, M., and Simon, B.: Methods of Modern Mathematical Physics, Il.
Fourier Analysis, Self-Adjointness, Academic Press, New York, 1975.
[RS3] Reed, M., and Simon, B.: Methods of Modern Mathematical Physics, Ill.
Scattering Theory, Academic Press, New York, 1979.
[RS4] Reed, M., and Simon, B.: Methods of Modern Mathematical Physics, IV.
Analysis of Operators, Academic Press, New York, 1978.
[R1] Robert, D.: Autour de ['Approximation Semiclassique, Birkhauser, Basel,

1983.
[RTl] Robert, D., and Tamura, H.: Semiclassical Bounds for Resolvents of
Schrodinger Operators and Asymptotics for Scattering Phases, Commun.
Partial Diff. Eq. 9,1017-1058 (1984).
[RT2] Robert, D., and Tamura, H.: Semiclassical Estimates for Resolvents and
Asymptotics for Scattering Phases, Ann. Inst. Henri Poincare 46, 415-442
(1987).
[Ro] Royden, H.: RealAnalysis, second edition, MacMillan Publishing Co., New
York,1968.
[R] Rudin, w.: Real and Complex Analysis, McGraw-Hill Book Co., New York,
1966.
[Sch] SchrOdinger, E.: Collected Papers on Wave Mechanics, London and Glas-
gow, Blackie and Son Limited, 1928.
[ST] Segur, H., and Tanveer, S., eds.: Asymptotics BeyondAll Orders, Birkhauser,
Boston, 1991.
[Sig] Sigal, I. M.: Scattering Theory for Many-Body Quantum Mechanical

Systems-Rigorous Results, Lecture Notes in Mathematics Vol. 1011,
[Sill Sigal, I. M.: Geometric Methods in the Quantum Many-Body Problem.

Non-existence of Very Negative Ions. Commun. Math. Phys. 85, 309-324
(1982).
[Si2] Sigal, I. M.: Complex Transformation Method and Resonances in One-

Body Quantum Systems. Ann. Inst. Henri Poincare 41, 103-114 (1984);
and Addendum 41, 333 (1984).
330 References
[Si3] Sigal, I. M.: A Generalized Weyl Theorem and the LP -Spectra of Schodinger
Operators, 1. Operator Theory 13, 119-129 (1985).
[Si4] Sigal, I. M.: General Characteristics of Non-Linear Dynamics, in Spectral

and Scattering Theory, M. Ikawa, ed., Marcel Dekker, Inc., New York, 1994.
[Si5] Sigal, I. M.: Sharp Exponential Bounds on Resonance States and Width of
Resonances, Advances in Math. 9, 127-166 (1988).
[SimI] Simon, B.: Pointwise Bounds on Eigenfunctions and Wave Packets in N-

Body Quantum Systems I, Proc. Arner. Math. Soc. 42,395-401 (1974).
[Sim2] Simon, B.: Pointwise Bounds on Eigenfunctions and Wave Packets in N-

Body Quantum Systems II, Proc. Arner. Math. Soc. 45,454-456 (1974).
[Sim3] Simon, B.: Pointwise Bounds on Eigenfunctions and Wave Packets in N-

Body Quantum Systems III, Trans. Arner. Math. Soc. 208,317-329 (1975).
[Sim4] Simon, B.: Trace Ideals and Their Applications, Cambridge University
Press, Cambridge, England, 1979.
[Sim5] Simon, B.: Semiclassical Analysis of Low Lying Eigenvalues, I. Non-

degenerate Minima: Asymptotic Expansions, Ann. Inst. Henri Poincare 38,
296-307 (1983).
[Sim6] Simon, B.: Semiclassical Analysis of Low Lying Eigenvalues, II. Tunnel-
ing,Ann. Math. 120,89-118 (1984).
[Sim7] Simon, B.: Semiclassical Analysis of Low Lying Eigenvalues, III. Width
of the Ground State Band in Strongly Coupled Solids, Ann. of Phys. 158,
415-442 (1984).
[Sim8] Simon, B.: Semiclassical Analysis of Low Lying Eigenvalues, IV. The Flea
on the Elephant, 1. Funct. Anal. 63, 123 (1984).
[Sim9] Simon, B.: The Theory of Resonances for Dilation Analytic Potentials and
the Foundations of Time Dependent Perturbation Theory, Ann. Math. 97,
247-274 (1973).
[SimlO] Simon, B.: Resonances and Complex Scaling: A Rigorous Overview, Int.
1. Quant. Chern. 14,529-542 (1978).
[Simll] Simon, B.: The Definition of Molecular Resonance Curves by the Method
of Exterior Complex Scaling. Phys. Letts. 71A, 211-214 (1979).
[Sim12] Simon, B.: Large Orders and Summability of Eigenvalue Perturbation

Theory: A Mathematical Overview, Int. 1. Quant. Chern. 21, 3-25 (1982).
[SimI3] Simon, B.: Schrodinger Semigroups, Bull. Arn. Math. Soc. 7, 447-526
(1982).
References 331
[SWYY] Singer, I. M., Wong, B., Yau, S. T., and Yau, S. S. T.: An Estimate of the
Gap of the First Two Eigenvalues for a Schrodinger Operator, Ann. Scuola
Norm. Sup. Pisa Cl. Sci. 12, 319-333 (1985).
[SkI] Skibsted, E.: Truncated Gamow Functions, a-Decay, and the Exponential
Law, Commun. Math. Phys. 104,591-604 (1986).
[Sk2] Skibsted, E.: Truncated Gamow Function and the Exponential Law, Ann.
[nst. Henri Poincare 46,131-153 (1987).
[Sk3] Skibsted, E.: On the Evolution of Resonance States, 1. Math. Anal. Appl.
141, 27-48 (1989).
[St] Stone, M. H.: Linear Transformations in Hilbert Space and Their Applica-
tions to Analysis, Amer. Math. Soc., Providence, RI, 1963.
[Th3] Thirring, W.: A Course in Mathematical Physics 3, Quantum Mechanics
of Atoms and Molecules, E. M. Harrell, trans., Springer-Verlag, New York,
1979.
[Th4] Thirring, W.: A Course in Mathematical Physics 4, Quantum Mechanics of
Large Systems, E. M. Harrell, trans., Springer-Verlag, New York, 1983.
[Tl] Titchmarsh, E. c.: The Theory of Functions, second edition, Cambridge
University Press, Cambridge, England, 1939.
[T2] Titchmarsh, E. c.: Eigenfunction Expansions Associated with Second-Order
Differential Equations, Part I, second edition, Part II, Clarendon Press, Ox-
ford, England, 1962 and 1958.
[VH] Vock, E., and Hunziker, W.: Stability of SchrOdinger Eigenvalue Problems,
Commun. Math. Phys. 38, 281-302 (1982).
[vN] von Neumann, J.: The Mathematical Foundations of Quantum Mechanics,
Princeton University Press, Princeton, NJ, 1955.
[WI] Wang, X. P.: Time-Decay of Scattering Solutions and Resolvent Estimates
for Semiclassical Schrodinger Operators, 1. Diff. Eq. 71,348-395 (1988).
[W2] Wang, X. P.: Barrier Resonances in Strong Magnetic Fields, Commun. Par-
tial Diff. Eq. 17, 1539-1566 (1992).
[Wx] Waxler, R.: The Time Evolution of a Class of Metastable States, to appear in
Commun. Math. Phys. 1995.
[Y] Yafaev, D. R.: The Eikonal Approximation and the Asymptotics of the Total
Cross-Section for the Schrodinger Equation, Ann. [nst. Henri Poincare 44,
397-425 (1986).
[Z] Zhislin, G.: Discussion of the Spectrum of the Schrodinger Operator for
Systems of Several Particles, Tr. Mosk. Mat. Obs. 9,81-128 (1960).
Index
O-notation, 219 approximate eigenvalue, 54

approximate Hamiltonian, 269
absence of eigenvalues, 236 approximate identity, 83
absence of positive eigenvalues, 163 approximate wave function, 115
absolute value of an operator, 92 approximation by finite-rank operators,
absolutely continuous functions, 50 96
accumulation point of a set, 288 Ascoli-ArzeUl theorem, 291
adjoint of a product, 42 asymptotic expansion, 117
adjoint of a sum, 42 asymptotic series, 273
adjoint of an inverse, 43
adjoint of an operator, 40 Banach space, 288
adjoints and closures, 43 Banach space: separable, 290
adjoints and spectra, 45 Banach spaces LP, 293
admissible contour, 59 Bessel's inequality, 22, 70
admissible potential, 192 bound state, 161
Agmon metric, 30 bounded operator, 303
Aguilar-Balslev-Combes theorem, 164
algebraic multiplicity, 11,64, 150 canonical form of a compact operator,
algebraic properties of operators, 302 96
almost everywhere pointwise convergence, Cauchy sequences, 287
299 classical limit, 109
analytic families of type-A, 153 classical nontrapping condition, 238
analytic family of operators, 152 classical turning surfaces, 217
analytic operator-valued function, 10, 152, classically forbidden region, 28,161,217
165, 199 classification of Hilbert spaces, 24
analytic vectors, 165, 182 closed operator, 40
approximate eigenfunctions, 70 closed set, 287
334 Index
closed subspace, 17 equivalent topologies, 308

closure of a set, 288 essential spectrum, 11, 69
closure of an operator, 40, 308 essentially self-adjoint, 80
commutator, 32, 102 Euclidean norm, 286
commutator estimate, 236 exponential decay, 27
compact operator, 89 exponential decay: isotropic, 37
compactness, 291 exponential decay: L 2, 29
completeness of an NLVS, 288 exponential decay: pointwise, 34
completion of an NLVS, 294 extension, 49
condition L, 183 exterior dilation analyticity, 186
conjugation of an operator, 45 exterior dilations, 179
continuous family of operators, 202 exterior region, 161
continuous linear operator, 303
convergence of operators, 97 Faris-Lavine theorem, 265
convolution, 317 Fermi's golden rule, 252
core, 80, 202 finite-rank operator, 91
Coulomb potential, 186 first resolvent identity, 12
countable set, 290 Fourier inversion formula, 315
Fourier transform, 46, 313
degenerate inner product, 30 Fourier transform and PDEs, 46
dense sets in NLVS, 290 Fredholm alternative, 93, 142,311
dense subset, 20
densely defined operator, 302 gauge transformation, 32
density of smooth functions, 296 Gaussian wave packets, 74
diffeomorphism, 178 generalized eigenvector, 64, 150
diffeomorphism: one-parameter family, generalized virial, 237, 239
178 geodesic, 30
dilation analyticity, 166 geodesic bisector, 124
dilation analyticity: exterior, 186 geometric eigenspace, 11
dilation group, Ill, 181,218 geometric multiplicity, 11,64, 150
direct sum, 20 geometric perturbation theory, 164,224
discrete spectrum, 11 geometric resolvent equation, 271, 279
distance to a subspace, 18 geometric resolvent formula, 212, 226,
distributional derivative, 82, 316 257
domain of an operator, 301, 309 geometric spectral analysis, 104,210
dual of a Banach space, 312 geometric spectral theory, 140
global flow, 179
eigenspace, 59, 67 graph norm, 40
eigenvalue, 11 graph of an operator, 40, 307
eigenvalue expansions, 156 ground state, 122
eigenvalue perturbation: degenerate case, ground state energy, 122
155
eigenvalue perturbation: nondegenerate Hamiltonian: classical, 238
case, 155 Hamiltonian: quantum, 80
eigenvector, 11 harmonic oscillator Hamiltonian, 110
elliptic regularity, 34, 75 harmonic oscillator potential, 86
embedded eigenvalue, 67, 270 Heine-Borel theorem, 291
energy functional, 30 Helmholtz resonator, 275
equivalent Cauchy sequences, 294 Hilbert space, 14
Index 335
Hilbert-Schmidt operators, 94 Minkowski's inequality, 298

HOlder's inequality, 297 multiplication operator, 78, 301
ideal of compact operators, 92 neighborhood, 286

imaginary part of the resonance, 252 Nelson's theorem, 182
implicit function theorem, 156 Neumann series for the inverse, 310
IMS localization formula, 211 nondegeneracy of V, 219
inner product, 13 nonempty spectrum, 12
inner product: degenerate, 30 nonseparable space, 290
inner product space, 13 nontrapping condition, 235
integral operator, 90, 94 nontrapping potential, 221
integral operator for the resolvent, 47 nontrapping: classical, 238
inverse of an operator, 309 nontrapping: definition, 237
inverse of the adjoint, 43 norm from an inner product, 14
invertible, 9 norm of an operator, 304
invertibility: lack of, 11 norm-resolvent convergence, 200, 201
invertible operator, 310 norm topology on operators, 312
isometric isomorphism, 24 norm: LP, 286
isometry, 45 norm: Euclidean, 286
norm: supremum, 286
Jacobian determinant, 178 normed linear vector space (NLVS), 286
Jacobian matrix, 178 null space of an LF, 20
numerical range, 203
Kato's inequality: general case, 84
Kato's inequality: smooth case, 81 open ball, 286
Kato potential, 143 open covering, 291
Kato-Rellich potentials, 136 open set, 286
Kato-Rellich theorem, 131 operator norm, 39
ker-Ran decomposition for a closed op- Oppenheimer's formula, 264
erator,43 order notation, 113
kernel of an operator, 309 orthogonal complement, 18
orthogonal projection, 56
Laplacian, 46, 73, 78, 80, 316 orthonormal basis, 22
Laplacian: local compactness, 99 orthonormal set, 22
Laplacian: resolvent kernel, 47, 164
Lebesgue dominated convergence, 299 parallelogram law, 14
length functional, 30 partial isometry, 92
lifetime, 282 partition of unity, 210, 211
limiting absorption principle, 164 perturbation, 81
linear functional, 312 perturbation theory, 113, 139, 149
linear functional (LF), 20 perturbation theory: analytic, 149
linear operator, 301 perturbation theory: asymptotic, 197
linear vector space (LVS), 285 Plancherel's theorem, 315
local compactness, 99 Planck's constant, 109
local operator, 103 Poincare's inequality, 276
localization formulas, 224 polar decomposition of an operator, 92
positive eigenvalues, 163
matrix element, 125 positive operator, 55, 92
metric on a manifold, 30 potential well, 161, 217
336 Index
projection operator, 56 singular values of a compact operator,

projection theorem, 19 97
projections: difference of, 151 Sobolev embedding theorems, 34, 317
Puiseux expansion, 155 Sobolev space, 14,316
Pythagorean theorem, 22 Sobolev trace theorem, 262
spectral deformation, 177
quantum resonance, 162, 165, 193 spectral deformation family, 163, 165
quantum tunneling effect, 28, 162 spectral properties of locally compact
quasiresonance, 174 operators, 10 1
spectral radius, 55
range of an operator, 309 spectral stability, 139
Rayleigh-Schrodinger series, 156,264 spectral stability: discrete spectrum, 149
reference Hamiltonian, 219 spectral stability: essential spectrum, 141
relative bound, 132 spectrum, 9
relatively bounded operator, 132 spectrum: deformed Schrodinger oper-
relatively bounded perturbations, 131 ator, 193
relatively compact perturbations, 139 spectrum: deformed Laplacian, 190
residual spectrum, 11 square root of a positive operator, 92
residual spectrum of self-adjoint opera- stability estimate, 223
tors, 51 stability: criteria for, analytic case, 151
resolvent, 10 stability: criteria for, general case, 207
resolvent bound for a self-adjoint oper- stability of eigenvalues, 151, 198
ator,53 Stark Hamiltonian, 195
resolvent set, 10 Stark ladder resonances, 264, 269
resonance-free domains, 249 Stone's theorem, 182
restriction of an operator, 49 strong analyticity, 152
Riesz's integral, 61 strong convergence, 15
Riesz lemma or representation theorem, strong convergence in the generalized
21 sense, 201
Riesz projections, 59 strong resolvent convergence, 201
Riesz-Schauder theorem, 93 strong topology on operators, 312
subspace, 17
scattering matrix, 174 summable, 23
scattering resonances, 276 sup-norm, 14,286
Schrodinger operator, 27, 80 support of a function, 293
SchrOdinger operators: deformed, 187 symmetric operator, 77
Schwartz class functions, 296 symmetric or Hermitian operator, 49
Schwarz inequality, 15
second resolvent identity, 13 tangent space, 30
self-adjoint, 50 topology of NLVS, 286
self-adjoint: fundamental criteria, 79 translation analyticity, 195
semi bounded operator, 136 translation group, 181
semiclassical regime, 109, 162, 171,218 truncated cone, 192
semiclassical resolvent estimates, 249
separable, 23 unitary group, 181
sequences in NLVS, 287 unitary group representation, III
shape resonance model, 216 unitary operator, 44, 45
similar operators, III
single-pole approximation, 173 variational inequalities, 119
Index 337
vector field, 178, 179 Weyl's theorem, 142

vector field exterior to a surface, 221 Weyl's theorem:
virial condition, 236 closed operators, 193
virial theorem, 235 WKB method, 29
weak analyticity, 152 Young's inequality, 318

weak convergence, 15
weak topology for operators, 312 Zhislin sequence, 102, 147
Weierstrass approximation theorem, 290 Zhislin sequence for closed operators,
Weyl sequences for closed operators, 105 106
Weyl's criterion, 54, 70 Zhislin spectrum, 102
Applied Mathematical Sciences
(continued from page ii)
61. SattingerlWeaver: Lie Groups and Algebras with 87. Weder: Spectral and Scattering Theory for Wave
Applications to Physics, Geometry, and Propagation in Perturbed Stratified Media.
Mechanics. 88. BogaevskilPovzner: Algebraic Methods in
62. LaSalle: The Stability and Control of Discrete Nonlinear Perturbation Theory.
Processes. 89. O'Malley: Singular Perturbation Methods for
63. Grasman: Asymptotic Methods of Relaxation Ordinary Differential Equations.
Oscillations and Applications. 90. MeyerlHall: Introduction to Hamiltonian
64. Hsu: Cell-to-Cell Mapping: A Method of Global Dynamical Systems and the N-body Problem.
Analysis for Nonlinear Systems. 91. Straughan: The Energy Method, Stability, and
65. Rand/Armbruster: Perturbation Methods, Nonlinear Convection.
Bifurcation Theory and Computer Algebra. 92. Naber: The Geometry of Minkowski Spacetime.
66. HlavaceklHaslingerlNecasllLovlsek: Solution of 93. ColtoniKress: Inverse Acoustic and
Variational Inequalities in Mechanics. Electromagnetic Scattering Theory.
67. Cercignani: The Boltzmann Equation and Its 94. Hoppensteadt: Analysis and Simulation of
Applications. Chaotic Systems.
68. Temam: Infinite Dimensional Dynamical 95. Hackbusch: Iterative Solution of Large Sparse
Systems in Mechanics and Physics. Systems of Equations.
69. GolubitskylStewartiSchaeffer: Singularities and 96. MarchiorolPulvirenti: Mathematical Theory of
Groups in Bifurcation Theory, Vol. II. Incompressible Nonviscous Fluids.
70. ConstantiniFoiaslNicolaenkolTemam: Integral 97. Lasota/Mackey: Chaos, Fractals, and Noise:
Manifolds and Inertial Manifolds for Dissipative Stochastic Aspects of Dynamics, 2nd ed.
Partial Differential Equations. 98. de BoorlHlJlliglRiemenschneider: Box Splines.
71. Catlin: Estimation, Control, and the Discrete 99. HaleiLunel: Introduction to Functional
Kalman Filter. Differential Equations.
72. LochaklMeunier: Multiphase Averaging for 100. Sirovich (ed): Trends and Perspectives in
Classical Systems. Applied Mathematics.
73. Wiggins: Global Bifurcations and Chaos. 101. NusseIYorke: Dynamics: Numerical
74. MawhinIWillem: Critical Point Theory and Explorations.
Hamiltonian Systems. 102. Chossatl/ooss: The Couette-Taylor Problem.
75. Abraham/Marsden/Ratiu: Manifolds, Tensor 103. Chorin: Vorticity and Turbulence.
Analysis, and Applications, 2nd ed. 104. Farkas: Periodic Motions.
76. Lagerstrom: Matched Asymptotic Expansions: 105. Wiggins: Normally Hyperbolic Invariant
Ideas and Techniques. Manifolds in Dynamical Systems.
77. Aldous: Probability Approximations via the 106. CercignanilIllnerlPulvirenti: The Mathematical
Poisson Clumping Heuristic. Theory of Dilute Gases.
78. Dacorogna: Direct Methods in the Calculus of 107. Antman: Nonlinear Problems of Elasticity.
Variations. 108. Zeidler: Applied Functional Analysis:
79. Hernandez-Lerma: Adaptive Markov Processes. Applications to Mathematical Physics.
80. Lawden: Elliptic Functions and Applications. 109. Zeidler: Applied Functional Analysis: Main
81. BlumaniKumei: Symmetries and Differential Principles and Their Applications.
Equations. llO. Diekmannlvan Gi/slVerduyn LunellWalther:
82. Kress: Linear Integral Equations. Delay Equations: Functional-, Complex-, and
83. BeberneslEberly: Mathematical Problems from Nonlinear Analysis.
Combustion Theory. lll. Visintin: Differential Models of Hysteresis.
84. Joseph: Fluid Dynamics of Viscoelastic Fluids. 112. Kuznetsov: Elements of Applied Bifurcation
85. Yang: Wave Packets and Their Bifurcations in Theory.
Geophysical Fluid Dynamics. 113. HisloplSigal: Introduction to Spectral Theory:
86. DendrinoslSonis: Chaos and Socio-Spatial With Applications to Schriidinger Operators.
Dynamics.

1

Uploaded by

Copyright:

Available Formats

1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1

Uploaded by

Copyright:

Available Formats

Appendix 1.

A1.1 Linear Vector Spaces and Norms

(2) X = C([O, I D, all cqntinuous, complex-valued functions on the interval

Definition A1.3. Let X be an LVS over F, F = lR or <C. A norm II . lion X is a

(i) IIx II 2: 0 and if IIx II = 0 then x = 0 (positive definiteness);

(iii) IIx + yll :::: IIxll + Ilyll, x, y E X (triangle inequality).

(3) X = IP has a norm given by IIxll == (2::~1 IxilP)I/P, x EX.

Al.2 Elementary Topology in Normed Vector Spaces

Br(a) == {x E Xlllx - all < r}.

(I) A sequence {XI1 l in X converges to x E X if Ilxn - x II ~ 0 as n ~ 00.

(2) A sequence (x" l in X is Cauchy if Ilxn - Xm I ~ 0 as n, m ~ 00.

Problem Al.3. Prove that

(I) every convergent sequence is a Cauchy sequence,

(1) Use an E/2-argument and the identity: X il - Xm = Xn - X +X - Xm.

(2) Use the triangle inequality to show that

I Ilxn I - Ilx m I I:::: Ilxn - Xm II·)

A nice feature of NLVS is that sequences can be used to characterize closed

(1) Suppose E is closed and lim l1 -+ oc Xn =

Definition A1.12. If M c X, X an NLVS, then M, the closure of M in X, is

Clearly, fn E C([O, 1]) and, for n ::::: m and p = 1,

AI.3 Banach Spaces

Remark A1.16. Every finite-dimensional NLVS is complete and hence a Banach

(1) X = ffi.N with the Euclidean norm II . lie is complete.

(3) [P is complete; we prove this in Proposition AU8.

and hence Ilx lip < 00 (i.e., x E IP). 0

Definition A1.19. Let X be an NLVS. A subset A C X is dense in X if each point

(1) The rational numbers Q are dense in the real numbers.

Definition AI.21. A Banach space X is separable if it contains a countable, dense

A2.1 The Definition of LP(IR n ), 1 ::::: P < 00

Definition A2.1. The support of a function f: Q C jRn -+ <C, denoted by supp f,

Problem A2.1. Prove that supp f = {x E Q I f(x) i OJ.

Ilfllp == If(x)iP dxJ P , (A2.l)

which exists as a Riemann integral because supp f is compact and because f is

Ourgoalis to associate with X p == (Cgo(Q), I . lip) a Banach space that contains

Problem A2.4. Show that X is an LVS, and verify that X is isomorphic to X.

Note that if [ax] E X, then naturally

so that the isomorphism between Xand X is isometric. We conclude that (X, II . II x)

Choose i M > max(io, i d. We claim {x~} C X is a Cauchy sequence. This follows

by definition of iN. Finally, construct [a] == [{x~}] E X. Then we have for N

< k~~ (IIX~ - xt II + IIxt' - Xi~ II)

A2.2 Important Properties of LP -Spaces

Co(S1) == all continuous functions on S1 with bounded supports;

Bo(S1) == all bounded functions on S1 with bounded supports.

When S1 = ]R1l, we define the Schwartz class functions S(]Rn) as follows:

S(]Rll) == {f E Coo(]Rn) I if g is any partial derivative of f, then for any

2. The HiJlder Inequality

TheoremA2.4 (The Holder inequality). Let p, q, r E JR+ be such that II p+ llq =

Ilfgllr :s Ilfllp Ilgll q .

Consider the function

where we use the fact that (p - 1)(q - 1) = 1. We consider the rectangle

whereas for the complementary region,

It is easy to see that we always have

which is just (A2.2).

(2) We next consider 1I p + 1I q = I I r. Setting a r and b r into (A2.2), we obtain

for lip + l/q = I. Now set p' == rp and q' = rq to obtain

for 1I p' + 1I q' = 1I r.

(3) If either I or g vanishes almost everywhere, the inequality is trivial. So

Iljgll; < rp-l Hill: + rq-lllglI~

Proposition A2.S. Let I E LP, 1 ::c p < 00. Then

IIIll p = sup {IIIgIIIlIgllq-llgiOand.!.+.!.=I} (A2A)