Next Article in Journal / Special Issue
Singular Bilinear Integrals in Quantum Physics
Previous Article in Journal / Special Issue
Effective Summation and Interpolation of Series by Self-Similar Root Approximants
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

The Schwartz Space: Tools for Quantum Mechanics and Infinite Dimensional Analysis

1
Department of Mathematics, Stephen F. Austin State University, PO Box 13040 SFA Station, Nacogdoches, TX 75962, USA
2
Department of Mathematics, Louisiana State University, Baton Rouge, LA 70803, USA
*
Author to whom correspondence should be addressed.
Submission received: 17 April 2015 / Accepted: 3 June 2015 / Published: 16 June 2015
(This article belongs to the Special Issue Mathematical physics)

Abstract

:
An account of the Schwartz space of rapidly decreasing functions as a topological vector space with additional special structures is presented in a manner that provides all the essential background ideas for some areas of quantum mechanics along with infinite-dimensional distribution theory.

1. Introduction

In 1950–1951, Laurent Schwartz published a two volumes work Théorie des Distributions [1,2], where he provided a convenient formalism for the theory of distributions. The purpose of this paper is to present a self-contained account of the main ideas, results, techniques, and proofs that underlie the approach to distribution theory that is central to aspects of quantum mechanics and infinite dimensional analysis. This approach develops the structure of the space of Schwartz test functions by utilizing the operator
T = d 2 d x 2 + x 2 4 + 1 2 .
This operator arose in quantum mechanics as the Hamiltonian for a harmonic oscillator and, in that context as well as in white noise analysis, the operator N = T − 1 is called the number operator. The physical context provides additional useful mathematical tools such as creation and annihilation operators, which we examine in detail.
In this paper we include under one roof all the essential necessary notions of this approach to the test function space S ( R n ). The relevant notions concerning topological vector spaces are presented so that the reader need not wade through the many voluminous available works on this subject. We also describe in brief the origins of the relevant notions in quantum mechanics.
We present
  • the essential notions and results concerning topological vector spaces;
  • a detailed analysis of the creation operator C, the annihilation operator A, the number operator N and the harmonic oscillator Hamiltonian T;
  • a detailed account of the Schwartz space S ( R d ), and its topology, as a decreasing intersection of subspaces S p ( R d ), for p ∈ {0, 1, 2, }:
    S ( R d ) = p 0 S p ( R d ) S 2 ( R d ) S 1 ( R d ) L 2 ( R d )
  • an exact characterization of the functions in the space S p ( R );
  • summary of notions from spectral theory and quantum mechanics;
Our exposition of the properties of T and of S ( R ) follows Simon’s paper [3], but we provide more detail and our notational conventions are along the lines now standard in infinite-dimensional distribution theory.
The classic work on spaces of smooth functions and their duals is that of Schwartz [1,2]. Our purpose is to present a concise and coherent account of the essential ideas and results of the theory. Of the results that we discuss, many can be found in other works such as [16], which is not meant to be a comprehensive list. We have presented portions of this material previously in [7], but also provide it here for convenience. The approach we take has a direct counterpart in the theory of distributions over infinite dimensional spaces [8,9].

2. Basic Notions and Framework

In this section we summarize the basic notions, notation, and results that we discuss in more detail in later sections. Here, and later in this paper, we work mainly with the case of functions of one variable and then describe the generalization to the multi-dimensional case.
We use the letter W to denote the set of all non-negative integers:
W = { 0 , 1 , 2 , 3 , } .

2.1. The Schwartz Space

The Schwartz space S ( R ) is the linear space of all functions f : RC which have derivatives of all orders and which satisfy the condition
p a , b ( f ) = def sup x R | x a f b ( x ) | <
for all a, bW = {0, 1, 2, }. The finiteness condition for all a ≥ 1 and bW, implies that xafb(x) actually goes to 0 as |x| → , for all a, bW, and so functions of this type are said to be rapidly decreasing.

2.2. The Schwartz Topology

The functions pa,b are semi-norms on the vector space S ( R ), in the sense that
p a , b ( f + g ) p a , b ( f ) + p a , b ( g )
and
p a , b ( z f ) = | z | p a , b ( f )
for all f , g S ( R ), and zC. For this semi-norm, an open ball of radius r centered at some f S ( R ) is given by
B p a , b ( f ; r ) = { g S ( R ) : p a , b ( g f ) < r } .
Thus each pa,b specifies a topology τa,b on S ( R ). A set is open according to τa,b if it is a union of open balls.
One way to generate the standard Schwartz topology τ on S ( R ) is to "combine" all the topologies τa,b. We will demonstrate how to generate a "smallest" topology containing all the sets of τa,b for all a, bW. However, there is a different approach to the topology on S ( R ) that is very useful, which we describe in detail below.

2.3. The Operator T

The operator
T = d 2 d x 2 + x 2 4 + 1 2
plays a very useful role in working with the Schwartz space. As we shall see, there is an orthonormal basis { ϕ n } n W of L2(R,dx), consisting of eigenfunctions ϕn of T:
T ϕ n = ( n + 1 ) ϕ n .
The functions ϕn, called the Hermite functions are actually in the Schwartz space S ( R ). Let B be the bounded linear operator on L2(R) given on each fL2(R) by
B f = n W ( n + 1 ) 1 f , ϕ n ϕ n .
It is readily checked that the right side does converge and, in fact,
B f L 2 2 = n W ( n + 1 ) 2 | f , ϕ n | L 2 2 n W | f , ϕ n | L 2 2 = f L 2 2 .
Note that B and T are inverses of each other on the linear span of the vectors ϕn:
T B f = f and B T f = f for all f ,
where
= linear span of the vector ϕ n , for n W .

2.4. The L2 Approach

For any p ≥ 0, the image of Bp consists of all fL2(R) for which
n W ( n + 1 ) 2 p | f , ϕ n | 2 < .
Let
S p ( R ) = B p ( L 2 ( R ) ) .
This is a subspace of L2(R), and on S p ( R ) there is an inner-product 〈·, ·〉p given by
f , g p = def n W ( n + 1 ) 2 p f , ϕ n ϕ n , g = B p f , B p g L 2
which makes it a Hilbert space, having , and hence also S ( R ), a dense subspace. We will see later that functions in S p ( R ) are p-times differentiable.
We will prove that the intersection p W S p ( R ) is exactly equal to S(R). In fact,
S ( R ) = p W S p ( R ) S 2 ( R ) S 1 ( R ) S 0 ( R ) = L 2 ( R ) .
We will also prove that the topology on S ( R ) generated by the norms ║·║p coincides with the standard topology. Furthermore, the elements ( n + 1 ) p ϕ n S p ( R ) form an orthonormal basis of S p ( R ), and
n W ( n + 1 ) ( p + 1 ) ϕ n p 2 = n 1 n 2 p n 2 ( p + 1 ) < ,
showing that the inclusion map S p + 1 ( R ) S p ( R ) is Hilbert-Schmidt.
The topological vector space S ( R ) has topology generated by a complete metric [10], and has a countable dense subset given by all finite linear combinations of the vectors ϕn with rational coefficients.

2.5. Coordinatization as a Sequence Space

All of the results described above follow readily from the identification of S ( R ) with a space of sequences. Let {ϕn}nW be the orthonormal basis of L2(R) mentioned above, where W = {0, 1, 2, }. Then we have the set CW; an element aCW is a map WC : nan. So we shall often write such an element a as (an)nW.
We have then the coordinatizing map
I : L 2 ( R ) C W : f ( f , ϕ n ) n W .
For each pW let Ep be the subset of CW consisting of all (an)nW such that
n W ( n + 1 ) 2 p | a n | 2 < .
On Ep define the inner-product 〈·, ·〉p by
a , b p = n W ( n + 1 ) 2 p a n b ¯ n .
This makes Ep a Hilbert space, essentially the Hilbert space L2(W, µp), where µp is the measure on W given by µp({n}) = (n + 1)2p for all nW.
The definition, Equation (10), for S p ( R ) shows that it is the set of all fL2(R) for which I(f) belongs to Ep.
We will prove in Theorem 16 that I maps S ( R ) exactly onto
E = def p W E p .
This will establish essentially all of the facts mentioned above concerning the spaces S p ( R ).
Note the chain of inclusions:
E = p W E p E 2 E 1 E 0 = L 2 ( W , μ 0 ) .

2.6. The Multi-Dimensional Setting

In the multidimensional setting, the Schwartz space S ( R d ) consists of all infinitely differentiable functions f on Rd for which
sup x R d | x 1 k 1 x d k d m 1 + + m k f ( x ) x 1 m 1 x d m d | < ,
for all (k1, , kd) ∈ W d and m = (m1, , md) ∈ Wd. For this setting, it is best to use some standard notation:
| k | = k 1 + + k d for k = ( k 1 , , k d ) W d
x k = x k 1 x k d
D m = | m | x 1 m 1 x d m d .
For the multi-dimensional case, we use the indexing set W d whose elements are d–tuples j = (j1, , jd), with j1, , jdW, and counting measure µ0 on Wd. The sequence space is replaced by C W d; a typical element a C W d, is a map
a : W d C : j = ( j 1 , , j d ) a j = a j 1 j d .
The orthonormal basis (ϕn)nW of L2(R) yields an orthonormal basis of L2(Rd) consisting of the vectors
ϕ j = ϕ j 1 ϕ j d : ( x 1 , , x d ) ϕ j 1 ( x 1 ) ϕ j d ( x d ) .
The coordinatizing map I is replaced by the map
I d : L 2 ( R d ) C W d
where
I d ( f ) j = f , ϕ j L 2 ( R d ) .
Replace the operator T by
T d = def T d = ( 2 x d 2 + x d 2 4 + 1 ) ( 2 x 1 2 + x 1 2 4 + 1 ) .
Then
T d ϕ j = ( j 1 + 1 ) ( j d + 1 ) ϕ j
for all jWd.
In place of B, we now have the bounded operator Bd on L2(Rd) given by
B d f j W d [ ( j 1 + 1 ) ( j d + 1 ) ] 1 f , ϕ j ϕ j
Again, Td and Bd are inverses of each other on the linear subspace d of L2(Rd) spanned by the vectors ϕj.
The space Ep is now the subset of C W d consisting of all a C W d for which
a p 2 = def j W d [ ( j 1 + 1 ) ( j d + 1 ) ] 2 p | a j | 2 < .
Thus,
E p = def { a C W d : a p 2 < }
This is a Hilbert space with inner-product
a , b p = j W d [ ( j 1 + 1 ) ( j d + 1 ) ] 2 p a j b ¯ j .
Again we have the chain of spaces
E = def p W E p E 2 E 1 E 0 = L 2 ( W d , μ 0 ) ,
with the inclusion Ep+1Ep being Hilbert-Schmidt.
To go back to functions on Rd, define S p ( R d ) to be the range of Bd. Thus S p ( R d ) is the set of all fL2(Rd) for which
j W d [ ( j 1 + 1 ) ( j d + 1 ) ] 2 p | f , ø j | 2 <
The inner-product 〈·, ·〉p comes back to an inner-product, also denoted 〈·, ·〉p, on S p ( R d ) and is given by
f , g p = B d p f , B d p g L 2 ( R d ) .
The intersection p W S p ( R d ) equals S ( R d ). Moreover, the topology on S ( R d ) is the smallest one generated by the inner-products obtained from 〈·, ·〉p, with p running over W.

3. Topological Vector Spaces

The Schwartz space is a topological vector space, i.e., it is a vector space equipped with a Hausdorff topology with respect to which the vector space operations (addition, and multiplication by scalar) are continuous. In this section we shall go through a few of the basic notions and results for topological vector spaces.
Let V be a real vector space. A vector topology τ on V is a topology such that addition V × VV : (x, y) ⟼ x + y and scalar multiplication R × VV : (t, x) ⟼ tx are continuous. If V is a complex vector space we require that C × VV : (α, x) ⟼ αx be continuous.
It is useful to observe that when V is equipped with a vector topology, the translation maps
t x : V V : y y + x
are continuous, for every xV, and are hence also homeomorphisms since t x 1 = t x.
A topological vector space is a vector space equipped with a Hausdorff vector topology. A local base of a vector topology τ is a family of open sets {Uα}αI containing 0 such that if W is any open set containing 0 then W contains some Uα. If U is any open set and x any point in U then Ux is an open neighborhood of 0 and hence contains some Uα, and so U itself contains a neighborhood x + Uα of x:
I f U i s o p e n a n d x U t h e n x + U α U , f o r s o m e α I .
Doing this for each point x of U, we see that each open set is the union of translates of the local base sets Uα.

3.1. Local Convexity and the Minkowski Functional

A vector topology τ on V is locally convex if for any neighborhood W of 0 there is a convex open set B with 0 ∈ BW. Thus, local convexity means that there is a local base of the topology τ consisting of convex sets. The principal consequence of having a convex local base is the Hahn-Banach theorem which guarantees that continuous linear functionals on subspaces of V extend to continuous linear functionals on all of V. In particular, if V ≠ {0} is locally convex then there exist non-zero continuous linear functionals on V.
Let B be a convex open neighborhood of 0. Continuity of R × VV : (s, x) ⟼ sx at s = 0 shows that for each x the multiple sx lies in B if s is small enough, and so t−1x lies in B if t is large enough. The smallest value of t for which t−1x is just outside B is clearly a measure of how large x is relative to B. The Minkowski functional µB is the function on V given by
μ B ( x ) = inf { t > 0 : t 1 x B } .
Note that 0 ≤ µB(x) < . The definition of µB shows that µB(kx) = B(x) for any k ≥ 0. Convexity of B can be used to show that
μ B ( x + y ) μ B ( x ) + μ B ( y ) .
If B is symmetric, i.e., B = −B, then µB(kx) = |k|µB(x) for all real k. If V is a complex vector space and B is balanced in the sense that αB = B for all complex numbers α with |α| = 1, then µB(kx) = |k|µB(x) for all complex k. Note that in general it could be possible that µB(x) is 0 without x being 0; this would happen if B contains the entire ray {tx : t ≥ 0}.

3.2. Semi-Norms

A typical vector topology on V is specified by a semi-norm on V, i.e., a function µ : VR such that
μ ( x + y ) μ ( x ) + μ ( y ) , μ ( t x ) = | t | μ ( x )
for all x, yV and tR (complex t if V is a complex vector space). Note that then, using t = 0, we have µ(0) = 0 and, using −x for y, we have µ(x) ≥ 0. For such a semi-norm, an open ball around x is the set
B μ ( x ; r ) = { y V : μ ( y x ) < r } ,
and the topology τµ consists of all sets which can be expressed as unions of open balls. These balls are convex and so the topology τµ is locally convex. If µ is actually a norm, i.e., µ(x) is 0 only if x is 0, then τµ is Hausdorff.
A consequence of the triangle inequality Equation (32) is that a semi-norm µ is uniformly continuous with respect to the topology it generates. This follows from the inequality
| μ ( x ) μ ( y ) | μ ( x y ) ,
which implies that µ, as a function on V, is continuous with respect to the topology τµ it generates. Now suppose µ is continuous with respect to a vector topology τ. Then the open balls {yV : µ(yx) < r} are open in the topology τ and so τµτ.

3.3. Topologies Generated by Families of Topologies

Let {τα}αI be a collection of topologies on a space. It is natural and useful to consider the the least upper bound topology τ, i.e., the smallest topology containing all sets of ∪αIτα. In our setting, we work with each τα a vector topology on a vector space V.
Theorem 1. The least upper bound topology τ of a collection {τα}αI of vector topologies is again a vector topology. If { W α , i } i I α is a local base for τα then a local base for τ is obtained by taking all finite intersections of the form W α 1 , i 1 W α n , i n.
Proof. Let B be the collection of all sets which are of the form W α 1 , i 1 W α n , i n.
Let τ′ be the collection of all sets which are unions of translates of sets in B (including the empty union). Our first objective is to show that τ′ is a topology on V. It is clear that τ′ is closed under unions and contains the empty set. We have to show that the intersection of two sets in τ′ is in τ′. To this end, it will suffice to prove the following:
If C 1 and C 2 are sets in B , and x is a point in the intersection of the translates a + C 1 and b + C 2 , then x + C ( a + C 1 ) ( b + C 2 ) for some C in B .
Clearly, it suffices to consider finitely many topologies τα. Thus, consider vector topologies τ1, , τn on V.
Let B n be the collection of all sets of the form B1 ∩ Bn with Bi in a local base for τi, for each i ∈ {1, , n}. We can check that if D, D B n then there is an E B n with ED ∩ D′.
Working with Bi drawn from a given local base for τi, let z be a point in the intersection B1∩Bn. Then there exist sets B i, with each B i being in the local base for τi, such that z + B i B i (this follows from our earlier observation Equation (31)). Consequently,
z + i = 1 n B i i = 1 n B i .
Now consider sets C1 an C2, both in B n. Consider a, bV and suppose x ∈ (a + C1) (b + C2). Then since xaC1 there is a set C 1 B n with x a + C 1 C 1; similarly, there is a C 2 B n with x b + C 2 C 2. So x + C 1 a + C 1 and x + C 1 a + C 1. So
x + C ( a + C 1 ) ( b + C 2 ) ,
where C B n satisfies CC1 ∩ C2. This establishes Equation (35), and shows that the intersection of two sets in τ′ is in τ′.
Thus τ′ is a topology. The definition of τ′ makes it clear that τ′ contains each τα. Furthermore, if any topology σ contains each τα then all the sets of τ′ are also open relative to σ. Thus τ′ = τ, the topology generated by the topologies τα.
Observe that we have shown that if Wτ contains 0 then WB for some B B.
Next we have to show that τ is a vector topology. The definition of τ shows that τ is translation invariant, i.e., translations are homeomorphisms. So, for addition, it will suffice to show that addition V × VV : (x, y) ⟼x + y is continuous at (0, 0). Let Wτ contain 0. Then there is a B B with 0 ∈ BW. Suppose B = B1 ∩ Bn, where each Bi is in the given local base for τi. Since τi is a vector topology, there are open sets Di, D i τ i, both containing 0, with D i + D i B i. Then choose Ci, C i in the local base for τi with CiDi and C i D i. Then C i + C i B i. Now let C = C1∩Cn, and C = C 1 C n. Then C, C B and C + C B. Thus, addition is continuous at (0, 0).
Now consider the multiplication map R × VV : (t, x) ⟼tx. Let (s, y), (t, x) ∈ R × V. Then
s y t x = ( s t ) x + t ( y x ) + ( s t ) ( y x ) .
Suppose Fτ contains tx. Then
F t x + W ,
for some W B. Using continuity of the addition map
V × V × V V : ( a , b , c ) a + b + c
at (0, 0, 0), we can choose W 1 , W 2 , W 3 B with W1 + W2 + W3W′. Then we can choose W B, such that
W W 1 W 2 W 3 .
Then W B and
W + W + W W .
Suppose W = B1 ∩ Bn, where each Bi is in the given local base for the vector topology τi. Then for s close enough to t, we have (st)xBi for each i, and hence (st)xW. Similarly, if y is τ–close enough to x then t(yx) ∈ W. Lastly, if st is close enough to 0 and y is close enough to x then (st)(yx) ∈ W. So sytxW′, and so syF, when s is close enough to t and y is τ–close enough to x. □
The above result makes it clear that if each τα has a convex local base then so is τ. Note also that if at least one τα is Hausdorff then so is τ.
A family of topologies {τα}αI is directed if for any α, βI there is a γI such that τατβτγ. In this case every open neighborhood of 0 in the generated topology contains an open neighborhood in one of the topologies τγ.

3.4. Topologies Generated by Families of Semi-Norms

We are concerned mainly with the topology τ generated by a family of semi-norms {µα}αI; this is the smallest topology containing all sets of α I τ μ α. An open set in this topology is a union of translates of finite intersections of balls of the form B µ i ( 0 ; r i ). Thus, any open neighborhood of f contains a set of the form
B µ 1 ( f ; r 1 ) B µ n ( f ; r n ) .
This topology is Hausdorff if for any non-zero xV there is some norm µα for which µα(x) is not zero.
The description of the neighborhoods in the topology τ shows that a sequence fn converges to f with respect to τ if and only if µα(fnf) → 0, as n, for all αI.
We will need to examine when two families of semi-norms give rise to the same topology:
Theorem 2. Let τ be the topology on V generated by a family of semi-norms = {µi}iI, and τ′ the topology generated by a family of semi-norms = { µ j } j J. Suppose each µi is bounded above by a linear combination of the µ j. Then ττ′.
Proof. Let µM. Then there exist µ j , , , and real numbers c1, , cn > 0, such that
μ c 1 μ 1 + + c n μ n .
Now consider any x, yV. Then
| μ ( x ) μ ( y ) | μ ( x y ) i = 1 n | c i | μ i ( x y ) .
So µ is continuous with respect to the topology generated by µ 1 , , µ n. Thus, τµτ′. Since this is true for all µ, we have ττ′. □

3.5. Completeness

A sequence (xn)n∈N in a topological vector space V is Cauchy if for any neighborhood U of 0 in V, the difference xnxm lies in U when n and m are large enough. The topological vector space V is complete if every Cauchy sequence converges.
Theorem 3. Let {τα}αI be a directed family of Hausdorff vector topologies on V, and τ the generated topology. If each τα is complete then so is τ.
Proof. Let (xn)n≥1 be a sequence in V, which is Cauchy with respect to τ. Then clearly it is Cauchy with respect to each τα. Let xα = limn xn, relative to τα. If τατγ then the sequence (xn)n≥1 also converges to xγ relative to the topology τα, and so xγ = xα. Consider α, βI, and choose γI such that τατβτγ. This shows that xα = xγ = xβ, i.e., all the limits are equal to each other. Let x denote the common value of this limit. We have to show that xnx in the topology τ. Let Wτ contain x. Since the family {τα}αI which generates τ is directed, it follows that there is a βI and a Bβτβ with xBβW. Since (xn)n≥1 converges to x with respect to τβ, it follows xnBβ for large n. So xnx with respect to τ. □

3.6. Metrizability

Suppose the topology τ on the topological vector space V is generated by a countable family of semi-norms µ1, µ2, . For any x, yV define
d ( x , y ) = n 1 2 n d n ( x , y )
where
d n ( x , y ) = min { 1 , μ n ( x y ) } .
Then d is a metric, it is translation invariant, and generates the topology τ [10].

4. The Schwartz Space S ( R )

Our objective in this section is to show that the Schwartz space is complete, in the sense that every Cauchy sequence converges. Recall that S ( R ) is the set of all C functions f on R for which
p a , b ( f ) = def f a , b sup x R | x a D b f ( x | <
for all a, bW = {0, 1, 2, }. The functions pa,b are semi-norms, with ║ · ║0,0, being just the sup-norm. Thus the family of semi-norms given above specify a Hausdorff vector topology on S ( R ). We will call this the Schwartz topology on S ( R ).
Theorem 4. The topology on S ( R ) generated by the family of semi-norms ║·║a,b for all a, b ∈ {0, 1, 2, }, is complete.
Proof. Let (fn)n≥1 be a Cauchy sequence on S ( R ). Then this sequence is Cauchy in each of the semi-norms ║·║a,b, and so each sequence of functions xaDbfn(x) is uniformly convergent. Let
g b ( x ) = lim n D b f n ( x ) .
Let f = g0. Using a Taylor theorem argument it follows that gb is Dbf. For instance, for b = 1, observe first that
f n ( y ) = f n ( x ) + 0 1 d f n ( ( 1 t ) x + t y ) ) d t d t = f n ( x ) + 0 1 f n ( ( 1 t ) x + t y ) ( y x ) d t ,
and so, letting n, we have
f ( y ) = f ( x ) + 0 1 g 1 ( ( 1 t ) x + t y ) ( y x ) d t ,
which implies that f′(x) exists and equals g1(x).
In this way, we have xaDbfn(x) → xaDbf(x) pointwise. Note that our Cauchy hypothesis implies that the sequence of functions xaDbfn(x) is Cauchy in sup-norm, and so the convergence
x a D b f n ( x ) x a D b f ( x )
is uniform. In particular, the sup-norm of xaDbf(x) is finite, since it is the limit of a uniformly convergent sequence of bounded functions. Thus f S ( R ).
Finally, we have to check that fn converges to f in the topology of S ( R ). We have noted above that xaDbfn(x) → xaDbf(x) uniformly. Thus fnf relative to the semi-norm ║·║a,b. Since this holds for every a, b ∈ {0, 1, 2, 3, }, we have fnf in the topology of S ( R ). □
Now let’s take a quick look at the Schwartz space S ( R d ). First some notation. A multi-index a is an element of {0, 1, 2, }d, i.e., it is a mapping
a : { 1 , , d } { 0 , 1 , 2 , } : j a j .
If a is a multi-index, we write |a| to mean the sum a1+⋯+ad, xa to mean the product x 1 a 1 x d a d, and Da to mean the differential operator x x 1 a 1 x x d a d. The space S ( R d ) consists of all C functions f on Rd such that each function xaDbf(x) is bounded. On S ( R d ) we have the semi-norms
f a , b = sup x R d | x a D b f ( x ) |
for each pair of multi-indices a and b. The Schwartz topology on S ( R d ) is the smallest topology making each semi-norm ║·║a,b continuous. This makes S ( R d ) a topological vector space.
The argument for the proof of the preceding theorem goes through with minor alterations and shows that:
Theorem 5. The topology on S ( R d ) generated by the family of semi-norms ║·║a,b for all a, b ∈ {0, 1, 2, }d, is complete.

5. Hermite Polynomials, Creation and Annihilation Operators

We shall summarize the definition and basic properties of Hermite polynomials (our approach is essentially that of Hermite’s original [11]). We repeat for convenience of reference much of the presentation in Section 2.1 of [7].
A central role is played by the Gaussian kernel
p ( x ) = 1 2 π e x 2 / 2 .
Properties of translates of p are obtained from
e x y y 2 2 = p ( x y ) p ( x ) .
Expanding the right side in a Taylor series we have
e x y y 2 2 = p ( x y ) p ( x ) = n = 0 1 n ! H n ( x ) y n ,
where the Taylor coefficients, denoted Hn(x), are
H n ( x ) = 1 p ( x ) ( d d x ) n p ( x ) .
This is the n–th Hermite polynomial and is indeed an n–th degree polynomial in which xn has coefficient 1, facts which may be checked by induction.
Observe the following
R p ( x y ) p ( x ) p ( x z ) p ( x ) p ( x ) d x = e y 2 + z 2 2 R e x ( y + z ) p ( x ) d x = e y 2 + z 2 2 + ( y + z ) 2 2 = e y z .
Going over to the Taylor series and comparing the appropriate Taylor coefficients (differentiation with respect to y and z can be carried out under the integral) we have
H n , H m L 2 ( p ( x ) d x ) = n ! δ n m .
Thus an orthonormal set of functions is given by
h n ( x ) = 1 n ! H n ( x ) .
Because these are orthogonal polynomials, the n–th one being exactly of degree n, their span contains all polynomials. It can be shown that the span is in fact dense in L2(p(x)dx). Thus the polynomials above constitute an orthonormal basis of L2(p(x)dx).
Next, consider the derivative of Hn:
H n ( x ) = ( 1 ) n p ( x ) 1 p ( n + 1 ) ( x ) ( 1 ) n p ( x ) 1 p ( x ) p ( x ) 1 p n ( x ) = H n + 1 ( x ) + x H n ( x ) .
So
( d d x + x ) h n ( x ) = n + 1 h n + 1 ( x ) .
The operator
( d d x + x )
is called the creation operator in L2(R;p(x)dx).
Officially, we can take the creation operator to have domain consisting of all functions f which can be expanded in L2(p(x)dx) as ∑n≥0 anhn, with each an a complex number, and satisfying the condition ∑n≥0(n + 1)|an|2 < ; the action of the operator on f yields the function n 0 n + 1 a n h n + 1. This makes the creation operator unitarily equivalent to a multiplication operator (in the sense discussed later in subsection A.5) and hence a closed operator (see A.1 for definition). For the type of smooth functions f we will mostly work with, the effect of the operator on f will in fact be given by application of ( d d x + x ) to f.
Next, from the fundamental generating relation Equation (41) we have :
y e x y y 2 / 2 = lim ϵ 0 n 0 1 n ! 1 ϵ [ H n ( x + ϵ ) H n ( x ) ] y n .
Using Equation (41) again on the left we have
n 1 1 ( n 1 ) ! H n 1 ( x ) y n = lim ϵ 0 n 0 1 n ! 1 ϵ [ H n ( x + ϵ ) H n ( x ) ] y n .
Letting y = 0 allows us to equate the n = 0 terms, and then, successively, the higher order terms. From this we see that
H n ( x ) = n H n 1 ( x )
where H−1 = 0. Thus:
d d x h n ( x ) = n h n 1 ( x ) .
The operator
d d x
is the annihilation operator in L2(R;p(x)dx). As with the creation operator, we may define it in a more specific way, as a closed operator on a specified domain.

6. Hermite Functions, Creation and Annihilation Operators

In the preceding section we studied Hermite polynomials in the setting of the Gaussian space L2(R;p(x)dx). Let us translate the concepts and results back to the usual space L2(R; dx).
To this end, consider the isomorphism:
U : L 2 ( R , p ( x ) d x ) L 2 ( R , d x ) : f p f .
Then the orthonormal basis polynomials hn go over to the functions ϕn given by
ϕ n ( x ) = ( 1 ) n 1 n ! ( 2 π ) 1 / 4 e x 2 / 4 d n e x 2 / 2 d x n .
The family {ϕn}n≥0 forms an orthonormal basis for L2(R, dx).
We now determine the annihilation and creation operators on L2(R, dx). If fL2(R, dx) is differentiable and has derivative f′ also in L2(R, dx), we have:
( U d d x U 1 ) f ( x ) = p ( x ) d d x [ p ( x ) 1 / 2 f ( x ) ] = f ( x ) + p ( x ) 1 / 2 ( 1 / 2 ) p ( x ) 3 / 2 p ( x ) f ( x ) = f ( x ) + 1 2 x f ( x ) .
So, on L2(R, dx), the annihilator operator is
A = d d x + 1 2 x
which will satisfy
A ϕ n = n ϕ n 1
where ϕ−1 = 0. For the moment, we proceed by taking the domain of A to be the Schwartz space S ( R ).
Next,
( U ( d d x + x ) U 1 ) f ( x ) = f ( x ) + x f ( x ) 1 2 x f ( x ) = ( d d x + 1 2 x ) f ( x )
Thus the creation operator is
C = A * = d d x + 1 2 x .
The reason we have written A* is that, as is readily checked, we have the adjoint relation
A f , g = f , ( d d x + 1 2 x ) g
with the inner-product being the usual one on L2(R, dx). Again, for the moment, we take the domain of C to be the Schwartz space S ( R ) (though, technically, in that case we should not write C as A, since the latter, if viewed as the L2–adjoint operator, has a larger domain).
For this we have
C ϕ n = n + 1 ϕ n + 1 .
Observe also that
A C = 1 4 x 2 d 2 d x 2 + 1 2 I and C A = 1 4 x 2 d 2 d x 2 1 2 I
which imply:
[ A , C ] = A C C A = I , the identity .
Next observe that
C A ϕ n = n n ϕ n = n ϕ n
and so CA is called the number operator N:
N = A * A = C A = d 2 d x 2 + x 2 4 1 2 the n u m b e r o p e r a t e r .
As noted above in Equation (59), the number operator N has the eigenfunctions ϕn:
N ϕ n = n ϕ n .
Integration by parts (see Lemma 10) shows that
f , g = f , g
for every f , g S ( R ), and so also
f , g = f , g .
It follows that the operator N satisfies
N f , g = f , N g
for every f , g S ( R ).
Now consider the case of Rd. For each j ∈ {1, , d}, there are creation, annihilation, and number operators:
A j = x j + 1 2 x j , C j = x j + 1 2 x j , N j = C j A j .
These map S ( R d ) into itself and, as is readily verified, satisfy the commutation relations
[ A j , C k ] = δ j k I , [ N j , A k ] = δ j k A j , [ N j , C k ] = δ j k C j .
Now let us be more specific about the precise definition of the creation and annihilation operators. The basis {ϕn}n≥0 for L2(R) yields an orthonormal basis { ϕ m } m W d of L2(Rd) given by ϕ m = ϕ m 1 ( x 1 ) ϕ m d ( x d ). For convenience we say ϕm=0 if some mj <0. Given its effect on the orthonormal basis { ϕ m } m W d the operator Ck has the form:
ϕ m m k + 1 ϕ m ,
where m i = m i for all i ∈ {1, , d} except when i = k, in which case m k = m k + 1. The domain of Ck is the set D ( C k ) given by
D ( C k ) = { f L 2 ( R ) | m W d ( m k + 1 ) | a m | 2 < where a m = f , ϕ m } .
The operator Ck is then officially defined by specifying its action on a typical element of its domain:
C k ( m W d a m ϕ m ) = m W d a m m m + 1 ϕ m ,
where m′ is as before. The operator Ck is essentially the composite of a multiplication operator and a bounded linear map taking ϕmϕm where m′ is as defined above. (See subsection A.5 for precise formulation of a multiplication operator.) Noting this, it can be readily checked that Ck is a closed operator using the following argument: Let T be a bounded linear operator and Mh a multiplication operator (any closed operator will do); we show that the composite MhT is a closed operator. Suppose xnx. Since T is a bounded linear operator, TxnTx. Now suppose also that Mh(Txn) → y. Since Mh is closed, it follows then that T x D ( M h ) and y = MhTx.
The operators Ak and Nk are defined analogously.
Proposition 6. Let 0 be the vector subspace of L2(Rd) spanned by the basis vectors { ϕ m } m W d. Then for k ∈ {1, 2, , d}, Ck|0 and Ak|0 have closures given by Ck and Ak, respectively (see subsection A.4 for the notion of closure).
Proof. We need to show that the graph of Ck, denoted Gr(Ck), is equal to the closure of the graph of Ck|0, i.e., to G r ( C k | 0 ) ¯ (see to subsection A.1 for the notion of graph). It is clear that G r ( C k | 0 ) G r ( C k ). Using this and the fact that Ck is a closed operator, we have
G r ( C k | 0 ) ¯ G r ( C k ) ¯ = G r ( C k ) .
Going in the other direction, take (f, Ckf) ∈ Gr(Ck). Now m W d a m ϕ m where am = 〈f, ϕm〉. Let fN be given by
f N = m W N d a m ϕ m where W N d = { m W d | 0 m 1 N , , 0 m d N } .
Observe that fN0. Moreover
lim N f N = f and lim N C k f N = lim N m W N d ( m k + 1 ) a m ϕ m = C k f
in L2(Rd). Thus ( f , C k f ) G r ( C k | L 0 ) ¯ and so we have G r ( C k ) G r ( C k | L 0 ) ¯.
The proof for Ak follows similarly. □
Linking this new definition for Ck with our earlier formulas Equation (63) we have:
Proposition 7. If f S ( R d ) then
C k f = f x k + x k 2 f , a n d A k f = f x k + x k 2 f .
Proof. Let g = f x k + x k 2 f. Since f S ( R d ), we have gL2(Rd). So we can write g as g = j W d a j ϕ j where aj=〈g, ϕj〉. Let us examine these aj’s more closely. Observe
a j = g , ϕ j = f x k + x k 2 f , ϕ j = f , ϕ j x k + x k 2 ϕ j = f , j k ϕ j
where j i = j i for all i ∈ {1, …, d} except when i = k, in which case j k = j k 1.
Bringing this information back to our expression for g we see that
g = j W d j k f , ϕ j ϕ j = m W d m k + 1 f , ϕ m ϕ m where m is as defined above = C k f by ( 65 ) .
The second equality is obtained by letting m = j″ and noting that ϕj = 0 when j k is −1. The proof follows similarly for Ak. □

7. Properties of the Functions in S p ( R )

Our aim here is to obtain a complete characterization of the functions in S p ( R ). We will prove that S p ( R ) consists of all square-integrable functions f for which all derivatives f(k) exist for k ∈ {1, 2, , p} and
sup x R | x a f ( b ) ( x ) | <
for all a, b ∈ {0, 1, , p − 1} with a + bp − 1.
A significant tool we will use is the Fourier transform:
f ^ ( p ) = f ( p ) = ( 2 π ) 1 / 2 R e i p x f ( x ) d x .
This is meaningful whenever f is in L1(R), but we will work mainly with f in S ( R ). We will use the following standard facts:
  • maps S ( R ) onto itself and satisfies the Plancherel identity:
    R | f ( x ) | 2 d x = R | f ^ ( p ) | 2 d p
  • for any f S ( R ),
    f ( x ) = ( 2 π ) 1 / 2 R e i p x f ^ ( p ) d p
  • if f S ( R ) then
p f ^ ( p ) = i ( f ) ( p ) .
Consequently, we have
f sup ( 2 π ) 1 / 2 R | f ^ ( p ) | d p = ( 2 π ) 1 / 2 R ( 1 + p 2 ) 1 / 2 | f ^ ( p ) | ( 1 + p 2 ) 1 / 2 d p = ( 2 π ) 1 / 2 [ R ( 1 + p 2 ) | f ^ ( p ) | 2 d p ] 1 / 2 π 1 / 2 by Cauchy-Schwartz 2 1 / 2 [ f ^ L 2 + p f ^ L 2 ( R , d p ) ] 2 1 / 2 [ f L 2 + f L 2 ] by Plancherel and Equation ( 69 ) .
For the purposes of this section it is necessary to be precise about domains. So we take now A and C to be closed operators in L2(R), with common domain
D ( C ) = D ( A ) = { f L 2 ( R ) | n 0 n | f , ϕ n | 2 < }
and
C f = n 0 f , ϕ n n + 1 ϕ n + 1 , A f = n 0 f , ϕ n n + 1 ϕ n 1 .
Moreover, define operators C1 and A1 on the common domain
D 1 = all differentiable f L 2 with f L 2 and x f ( x ) L 2 ( d x )
and
C 1 f ( x ) = [ d d x + 1 2 x ] f ( x ) , A 1 f ( x ) = [ d d x + 1 2 x ] f ( x ) for all f D 1 .
We will prove below that C and C1 (and A and A1) are, in fact, equal.
For a function
f = n 0 a n ϕ n L 2 ( R ) ,
we will use the notation fN for the partial sum:
f N n = 0 N a n ϕ n .
Observe the following about the derivatives f N:
Lemma 8. If f D ( C ), then { f N } is Cauchy in L2(R).
Proof. Note that
f N = ( A C 2 ) f N . So f N f M L 2 1 2 A f N A f M L 2 + C f N C f M L 2 .
Now for M < N we have
A f N A f M L 2 2 = n = M + 1 N a n n ϕ n 1 L 2 2 = n = M + 1 N | a n | 2 n .
Likewise,
C f N C f M L 2 2 = n = M + 1 N a n n + 1 ϕ n + 1 L 2 2 = n = M + 1 N | a n | 2 ( n + 1 ) .
Since f D ( C ), we know n = M + 1 N | a n | 2 ( n + 1 ) tends to 0 as M goes to infinity. Thus { f N } is Cauchy in L2(R). □
Lemma 9. If f D ( C ) then f is, up to equality almost everywhere, bounded, continuous and {fN} converges uniformly to f, i.e., ║ffNsup→0 as N→∞.
Proof. It is enough to show that ║ffNsup→0 as M,N→∞. Note that
f M f N sup 2 1 / 2 ( f N f M L 2 + f N f M L 2 )
by Equation (70). Since fL2(R) we have f N f M L 2 0 as M, N and by Lemma 8 we have that f N f M L 2 0 as M, N Therefore {fN} converges uniformly to f. □
Next we establish an integration-by-parts formula:
Lemma 10. If f,gL2(R) are differentiable with derivatives also in L2(R) then
R f ( x ) g ( x ) d x = R f ( x ) g ( x ) d x .
Proof. The derivative of fg, being fg + fg′, is in L1. So the fundamental theorem of calculus applies to give:
a b f ( x ) g ( x ) d x + a b f ( x ) g ( x ) d x = f ( b ) g ( b ) ( a ) g ( a )
for all real numbers a < b.
Now fgL1, and so
lim N N f ( x ) g ( x ) d x = lim N N f ( x ) g ( x ) d x = 0.
Consequently, there exist aN < −N < N < bN with
f ( a N ) g ( a N ) 0 and f ( b N ) g ( b N ) 0 as N .
Plugging into Equation (74) we obtain the desired result. □
Next we have the first step to showing that C1 equals C:
Lemma 11. If f is in the domain of C1 then f is in the domain of C and
C f = C 1 f a n d A f = A 1 f .
Proof. Let f be in the domain of C1. Then we may assume that f is differentiable and both f and the derivative f′ are in L2(R). We have then
C 1 f , ϕ n = R ϕ n ( x ) [ d d x + 1 2 x ] f ( x ) d x = R f ( x ) [ d d x + 1 2 x ] ϕ n ( x ) d x by Equation ( 73 ) = n f , ϕ n 1 by Equation ( 53 ) .
Then
| | C 1 f | | 2 = n 0 | C 1 f , ϕ n | 2 = n 1 n | f , ϕ n 1 | 2 .
Because this sum is finite, it follows that f is in the domain D ( C ) of C. Moreover,
C f = n 0 n + 1 f , ϕ n + 1 = m 0 C 1 f , ϕ m ϕ m by Equation ( 76 ) = C 1 f .
The argument showing Af = A1f is similar. □
We can now prove:
Theorem 12. The operators C and C1 are equal, and the operators A and A1 are equal. Thus, a function fL2(R) is in the domain of C (which is the same as the domain of A) if and only if f is, up to equality almost everywhere, a differentiable function with derivative falso in L2(R) and with R | x f ( x ) | 2 d x < .
Proof. In view of Lemma 11, it will suffice to prove that D ( C ) D ( C 1 ). Let f D ( C ). Then
n 0 n | f , ϕ n | 2 < .
This implies that the sequences {C1fN}N≥0 and {A1fN}N≥0 are Cauchy, where fN is the partial sum
f N = n = 0 N f , ϕ n ϕ n .
Now
1 2 ( A 1 C 1 ) f N = f N , and 1 2 ( A 1 + C 1 ) f N ( x ) = x f N ( x ) .
So the sequences of functions { f N } N 0 and {hN}N≥0, where
h N ( x ) = x f N ( x ) ,
are also L2–Cauchy. Now, as shown in Lemma 9, we can take f to be the uniformly convergent pointwise limit of the sequence of continuous functions fN.
By Lemma 8, the sequence of derivatives f N is Cauchy in L2(R). Let g = lim N f N in L2(R). Observe that
f N ( y ) = f N ( x ) + 0 1 f N ( ( x + t ( y x ) ) ( y x ) d t .
Now 0 1 | f N ( x + t ( y x ) ) ( y x ) g ( x + t ( y x ) ) ( y x ) | d t | y x | f N g L 2 by the Cauchy-Schwartz inequality. Since f N g L 2 0 as N, we have
0 1 f N ( x + t ( y x ) ) ( y x ) d t 0 1 g ( x + t ( y x ) ) ( y x ) d t .
Because { f N } N 0 converges to f uniformly by Lemma 9, taking the limit as N → ∞ in Equation (78) we obtain
f ( y ) = f ( x ) + 0 1 g ( ( x + t ( y x ) ) ( y x ) d t .
Therefore f′ = gL2(R). Lastly, we have, by Fatou’s Lemma:
R | x f ( x ) | 2 d x lim inf N R | x f N ( x ) | 2 d x < ,
because the sequence {gN}N≥0 is convergent. Thus we have established that f D ( C 1 ). □
Finally we can characterize the space Sp(R):
Theorem 13. Suppose fSp(R), where p ≥ 1. Then f is (up to equality almost every where) a 2p times differentiable function and
sup x R | x a f ( b ) ( x ) | <
for every a, b ∈ {0, 1, 2, } with a + b < 2p. Moreover, Sp(R) consists of all 2p times differentiable functions for which the functions xxaf(b)(x) are in L2(R) for every a, b ∈ {0, 1, 2, } with a + b ≤ 2p.
Proof. Consider fS1(R). Then
f = n 0 a n ϕ n with n 0 n 2 | a n | 2 < .
In particular, f D ( C ). Moreover,
C f = n 0 a n n + 1 ϕ n + 1 A f = n 0 a n n ϕ n 1 .
From these expressions and Equation (79) it is clear that Cf and Af both belong to D ( C ). Thus,
B 1 B 2 f L 2 ( R ) for all B 2 { C , A , I } .
Similarly, we can check that if fSp(R), where p ≥ 2, then
B 1 B 2 f S p 1 ( R ) for all B 1 , B 2 { C , A , I } .
Thus, inductively, we see that
B 1 B 2 p f L 2 ( R ) for all B 1 , , B 2 p { C , A , I } .
(This really means that f is in the domain of each product operator B1 ···B2p.) Now the operators d d x and multiplication by x are simple linear combinations of A and C. So for any a, b ∈ {0, 1, 2, } with a + b ≤ 2p we can write the operator x a ( d d x ) b as a linear combination of operators B1…B2p with B1, , B2p ∈ {C, A, I}.
Conversely, suppose f is 2p times differentiable and the functions xxaf(b)(x) are in L2(R) forevery a, b ∈ {0, 1, 2, … } with a + b ≤ 2p. Then f is in the domain of C2p and so
n 0 | f , ϕ n | 2 n 2 p < .
Thus fSp(R).
The preceding facts show that if fSp(R) then for every B1, , B2p ∈ {C, A, I}, the element B1 ····B2p−1f is in the domain of C, and so, in particular, is bounded. Thus,
sup x R | x a f ( b ) ( x ) | <
for all a, b ∈ {0, 1, 2, } with a + b ≤ 2p − 1. □
We do not carry out a similar study for Sp(Rd), but from the discussions in the following sections, it will be clear that:
  • Sp(Rd) is a Hilbert space with inner-product given by
    f , g p = f , T d 2 p g
  • as a Hilbert space, Sp (Rd) is the d–fold tensor product of Sp(R) with itself.

8. Inner-Products on S(R) from N

For fL2(R), define
f t = { n 0 ( n + 1 ) 2 t | f , ϕ n | 2 } 1 / 2
for every t > 0. More generally, define
f , g t = n 0 ( n + 1 ) 2 t f , ϕ n L 2 ϕ n , g L 2 ,
for all f, g in the subspace of L2(R) consisting of functions F for which ||F ||t < ∞.
Theorem 14. Let fS(R). Then for every t > 0 we have ||f||t < ∞. Moreover, for every integer m ≥ 0, we also have
N m f = n 0 n m f , ϕ n ϕ n ,
where on the left Nm is the differential operator d 2 d x 2 + x 2 4 1 2 applied n times, and on the right the series is taken in the sense of L2(R, dx). Furthermore,
f m / 2 2 = f , ( N + 1 ) m f .
This result will be strengthened and a converse proved later.
Proof. Let m ≥ 0 be an integer. Since fS(R), it is readily seen that N f is also in S(R), and thus, inductively, so is Nmf. Then we have
f , N m f = n 0 f , ϕ n ϕ n , N m f = n 0 f , ϕ n N m ϕ n , f by Equation ( 62 ) = n 0 f , ϕ n n m ϕ n , f = n 0 n m | f , ϕ n | 2 .
Thus we have proven the relation
f , N m f = n 0 n m | f , ϕ n | 2 .
An exactly similar argument shows
f , ( N + 1 ) m f = n 0 ( n + 1 ) m | f , ϕ n | 2 = f m / 2 2 .
So if t > 0, choosing any integer mt we have
f t / 2 2 f m / 2 2 = f , ( N + 1 ) m f < .
Observe that the series
n 0 n m f , ϕ n ϕ n
is convergent in L2(R, dx) since
n 0 n 2 m | f , ϕ n | 2 = N 2 m f , f < .
So for any gL2(R, dx) we have, by an argument similar to the calculations done above:
N m f , g = n 0 n m f , ϕ n ϕ n , g = n 0 n m f , ϕ n ϕ n , g = n 0 n m f , ϕ n ϕ n , g .
This proves the statement about Nmf. □
We have similar observations concerning Cmf and Amf. First observe that since C and A are operators involving d d x and x, they map S(R) into itself. Also,
A f , g = f , C g ,
for all f, gS(R), as already noted. Using this, for fS(R), we have
ϕ n + m , C m f = A m ϕ n + m , f = ( n + m ) ( n + m 1 ) ( n + 1 ) ϕ n , f .
Therefore,
C m f = n 0 [ ( n + m ) ! n ! ] 1 / 2 f , ϕ n ϕ n + m .
Similarly,
A m f = n 0 [ n ! ( n m ) ! ] 1 / 2 f , ϕ n ϕ n m .
More generally, if B1, , Bk are such that each Bi is either A or C then
B 1 B k f = n 0 θ n , k f , ϕ n ϕ n + r ,
where the integer r is the excess number of C’s over the A’s in the sequence B1, , Bk, and θn,k is a real number determined by n and k. We do have the upper bound
θ n , k 2 ( n + k ) k [ ( n + 1 ) k ] k = ( n + 1 ) k k k .
Note also that
B 1 B k f 2 = ( B 1 B k ) * B 1 B k f , f = n 0 θ n , k 2 | f , ϕ n | 2 .
Let’s look at the case of Rd. The functions ϕn generate an orthonormal basis by tensor products. In more detail, if aWd is a multi-index, define ϕaL2(Rd) by
ϕ a ( x ) = ϕ a 1 ( x 1 ) ϕ a d ( x d ) .
Now, for each t > 0, and fL2(Rd), define
f t = def { a W d [ ( a 1 + 1 ) ( a d + 1 ) ] 2 t | f , ϕ a | 2 } 1 / 2 ,
and then define
f , g t = a W d [ ( a 1 + 1 ) ( a d + 1 ) ] 2 t f , ϕ a ϕ a , g ,
for all f, g in the subspace of L2(Rd) consisting of functions F for which ║Ft < ∞.
Let Td be the operator on S ( R d ) given by
T d = ( N d + 1 ) ( N 1 + 1 ) .
Then, for every non-negative integer m, we have
f m / 2 2 = f , T d m f .
The other results of this section also extend in a natural way to Rd.

9. L2–Type Norms on S ( R )

For integers a, b ≥ 0, and f S ( R ), define
f a , b , 2 = x a D b f ( x ) L 2 ( R , d x ) .
Recall the operators
A = d d x + 1 2 x , C = d d x + 1 2 x , N = C A
and the norms
f m = f , ( N + 1 ) m f .
The purpose of this section is to prove the following:
Theorem 15. The system of semi-norms given byfa,b,2 and the system given by the normsfm generate the same topology on S ( R ).
Proof. Let a, b be non-negative integers. Then
f a , b , 2 = ( A + C ) a 2 b ( A C ) b f L 2 a linear combination of terms of the form B 1 B k f L 2 ,
where each Bi is either A or C, and k = a + b. Writing cn = ⟨f, ϕn⟩, we have where
B 1 B k f L 2 2 = n 0 c n θ n , k ϕ n + r = n 0 | c n | 2 θ n , k 2 ,
where
r = # { j : B j = C } # { j : B j = A } ,
and, as noted earlier in Equation (90),
θ n , k 2 ( n + 1 ) k k k .
So
B 1 B k f L 2 2 n 0 | c n | 2 ( n + 1 ) k k k = k k f k / 2 2 k k f k 2 .
Thus ║fa,b,2 is bounded above by a multiple of the norm ║fa+b.
It follows, that the topology generated by the semi-norms || · ||a,b,2 is contained in the topology generated by the norms || · ||k.
Now we show the converse inclusion. From
f k 2 = f , ( N + 1 ) 2 k f L 2 f L 2 2 + ( N + 1 ) 2 k f L 2 2
and the expression of N as a differential operator we see that f k 2 is bounded above by a linear combination of f a , b , 2 2 for appropriate a and b. It follows then that the topology generated by the norms ║·║k is contained in the topology generated by the semi-norms ║·║a,b,2. □
Now consider Rd. Let a, bWd be multi-indices, where W = {0, 1, 2, …}. Then for f ∈ S(Rd) define
f a , b , 2 = { R d | x a D b f ( x ) | 2 d x } 1 2
These specify semi-norms and they generate the same topology as the one generated by the norms || · ||m, with mW. The argument is a straightforward modification of the one used above.

10. Equivalence of the Three Topologies

We will demonstrate that the topology generated by the family of norms || · ||k, or, equivalently, by the semi-norms ║ · ║a,b,2, is the same as the Schwartz topology on. S ( R ).
Recall from Equation (70) that we have, for f S ( R )
f sup 2 1 / 2 ( f L 2 + f L 2 ) .
Putting in xaDbf(x) in place of f(x) we then have
f a , b a linear combination of f a , b , 2 , and f a , b + 1 , 2 .
Next we bound the semi-norms ║fa,b,2 by the semi-norms ║fa,b. To this end, observe first
f L 2 2 = R ( 1 + x 2 ) 1 ( 1 + x 2 ) | f ( x ) | 2 d x π ( 1 + x 2 ) | f ( x ) | 2 sup π ( f sup 2 + x f ( x ) sup 2 ) π ( f sup + x f ( x ) sup ) 2 .
So for any integers a, b ≥ 0, we have
f a , b , 2 = x a D b f L 2 π 1 / 2 ( f a , b + f a + 1 , b ) .
Thus, the topology generated by the semi-norms ║ · ║a,b,2 coincides with the Schwartz topology.
Now lets look at the situation for Rd. The same result holds in this case and the arguments are similar. The appropriate Sobolev inequalities require using (1 + |p|2)d instead of 1 + p2. For f S ( R d ), we have the Fourier transform given by
( f ) ( p ) = f ^ ( p ) = ( 2 π ) d / 2 R d e i p , x f ( x ) d x .
Again, this preserves the L2 norm, and transforms derivatives into multiplications:
p i f ^ ( p ) = i ( f x i ) ( p ) .
Repeated application of this shows that
| p | 2 f ^ ( p ) = ( Δ f ) ( p ) ,
where
Δ = j = 1 n 2 x j 2
is the Laplacian. Iterating this gives, for each r ∈ {0, 1, 2, …} and f S ( R d ),
| p | 2 r f ^ ( p ) = ( 1 ) r ( Δ r f ) ( p ) ,
which in turn implies, by the Plancherel formula Equation (67), the identity:
R d | | p | 2 r | f ^ ( p ) | | 2 d p = R d | Δ r f ( x ) | 2 d x .
Then we have, for any m > d/4,
f sup ( 2 π ) d / 2 R d | f ^ ( p ) | d p = ( 2 π ) d / 2 R d ( 1 + | p | 2 ) m | f ^ ( p ) | ( 1 + | p | 2 ) m d p = K [ R d ( 1 + | p | 2 ) 2 m | f ^ ( p ) | 2 d p ] 1 / 2 by Cauchy-Schwartz ,
where K = ( 2 π ) d / 2 [ R d d p ( 1 + | p | 2 ) 2 m ] 1 / 2 < The function (1 + s)n/(1 + sn), for s ≥ 0, attains a maximum value of 2n−1, and so we have the inequality (1 + s)2m 22m−1(1 + s2m), which leads to
( 1 + | p | 2 ) 2 m 2 2 m 1 ( 1 + | p | 4 m ) .
Then, from Equation (102), we have
f sup 2 K 2 2 2 m 1 ( f L 2 2 + Δ m f L 2 2 ) .
This last quantity is clearly bounded above by a linear combination of ║f0,b,2 for certain multi-indices b. Thus ║fsup is bounded above by a linear combination of ║f0,b,2 for certain multi-indices b. It follows that ║xa Dbfsup is bounded above by a linear combination of ║fa,b,2 for certain multi-indices a′, b′.
For the inequality going the other way, the reasoning used above for Equation (97) generalizes readily, again with (1 + x2) replaced by (1 + |x|2)d. Thus, on S ( R d ) the topology generated by the family of semi-norms ║ · ║a,b,2 coincides with the Schwartz topology.
Now we return to Equation (102) for some further observations. First note that
Δ = 1 4 j = 1 d ( C j A j ) 2
and so Δm consists of a sum of multiples of (3d)m terms each a product of 2m elements drawn from the set {A1, C1,…, Ad, Cd}. Consequently, by Equation (95)
Δ m f L 2 2 c d , m 2 f m 2 ,
for some positive constant cd,m. Combining this with Equation (102), we see that for m > d/4, there is a constant kd,m such that
f sup k d , m f m
holds for all f S ( R d ).
Now consider f S p ( R d ), with p > d/4. Let
f N = j W d , | j | N f , ϕ j ϕ j .
Then fN → f in L2 and so a subsequence { f N k } k 1 converges pointwise almost everywhere to f. It follows then that the essential supremum ║f is bounded above as follows:
f lim sup N | f N | sup .
Note that fN → f also in the ║ · ║p–norm. It follows then from Equation (104) that
f k d , p f p
holds for all f S p ( R d ) with p > d/4. Replacing f by the difference f − fN in Equation (105), we see that f is the L–limit of a sequence of continuous functions which, being Cauchy in the sup-norm, has a continuous limit; thus f is a.e. equal to a continuous function, and may thus be redefined to be continuous.

11. Identification of S ( R ) with a Sequence Space

Suppose a0, a1, … form a sequence of complex numbers such that
n 0 ( n + 1 ) m | a n | 2 < , for every integer m 0.
We will show that the sequence of functions given by
s n = j = 0 n a j ϕ j
converges in the topology of S ( R ) to a function f S ( R ) for which an = ⟨f, ϕn⟩ for every n ≥ 0.
All the hard work has already been done. From Equation (106) we see that (sn)n≥0 is Cauchy in each norm ║·║m. So it is Cauchy in the Schwartz topology of S ( R ), and hence convergent to some f S ( R ). In particular, sn → f in L2. Taking inner-products with ϕj we see that aj = ⟨f, ϕj⟩.
Thus we have
Theorem 16. Let W = {0, 1, 2, …}, and define
F : L 2 ( R ) C W
by requiring that
F ( f ) n = f , ϕ n L 2
for all nW. Then the image of S ( R ) under F is the set of all a ∈ CW for which a m 2 = def n 0 ( n + 1 ) 2 m | a n | 2 < for every integer m ≥ 0. Moreover, if F ( S ( R ) ) is equipped with the topology generated by the norms ║ · ║m then F is a homeomorphism.

A. Spectral Theory in Brief

In this section we present a self contained summary of the concepts and results of spectral theory that are relevant for the purposes of this article.
Let H be a complex Hilbert space. A linear operator on H is a linear map
A : D A H ,
where DA is a subspace of H. Usually, we work with densely defined operators, i.e., operators A for which DA is dense.

A.1. Graph and Closed Operators

The graph of the operator A is the set of all ordered pairs (x, Ax) with x running over the domain of A:
G r ( A ) = { ( x , A x ) : x D A } .
Thus Gr(A) is A viewed as a set of ordered pairs, and is thus A itself taken as a mapping in the set-theoretic sense. The operator A is said to be closed if its graph is a closed subset of HH; put another way, this means that if (xn)n≥1 is any sequence in H which converges to a limit x and if limn→∞ Axn = y also exists then x is in the domain of A and y = Ax.

A.2. The Adjoint A

If A is a densely defined operator on H then there is an adjoint operator A* defined as follows. Let DA be the set of all yH for which the map
f y : D A C : x A x , y
is bounded linear. Clearly, DA is a subspace of H. The bounded linear functional fy extends to a bounded linear functional fy on H. So there exists a vector zH such that fy(x) = ⟨z, x⟩ for all xH. Since DA is dense in H, the element z is uniquely determined by x and A. Denote z by A*y. Thus, A*y is the unique vector in H for which
x , A * y = A x , y
holds for all xDA. Using the definition of A* for a densely-defined operator A it is readily seen that A* is a closed operator.

A.3. Self-Adjoint Operators

The operator A is self-adjoint if it is densely defined and A = A*. Thus, if A is self-adjoint then DA = DA* and
x , A y = A x , y
for all x, yDA. Note that a self-adjoint operator A, being equal to its adjoint A*, is automatically a closed operator.

A.4. Closure, and Essentially Self-Adjoint Operators

Consider a densely-defined linear operator S on H. Assume that the closure of the graph of S is the graph of some operator S ¯. Then S ¯ is called the closure of S. We say that S is essentially self-adjoint if its closure is a self-adjoint operator. In particular, S must then be a symmetric operator, i.e., it satisfies
S x , y = x , S y
for all x, yH. A symmetric operator may not, in general, be essentially self-adjoint.

A.5. The Multiplication Operator

Let us turn to a canonical example. Let ( X , , µ ) be a sigma-finite measure space. Consider the Hilbert space L2(µ). Let f : X → C be a measurable function. Define the operator Mf on L2(µ) by setting
M f g = f g ,
with the domain of Mf given by
D ( M f ) = { g L 2 ( μ ) : f g L 2 ( μ ) } .
Let us check that D(Mf) is dense in L2(µ). By sigma-finiteness of µ, there is an increasing sequence of measurable sets Xn such that ∪n≥1Xn=X and µ (Xn) < ∞. For any hL2(µ) let h n = 1 X n { | f | n } h. Than
| f h n | d μ n | h | 1 X n d μ n μ ( X n ) 1 / 2 h L 2 <
and so hnD(Mf). On the other hand,
h n h L 2 2 0
by dominated convergence. So D(Mf) is dense in H.
It may be shown that
M f * = M f ¯ .
Thus Mf is self-adjoint if f is real-valued.
A very special case of the preceding example is obtained by taking X to be a finite set, say X = {1, 2, …, d}, and µ as counting measure on the set of all subsets of X. In this case, L2(µ) = Cd, and the operator Mf, viewed as a linear map
M f : C d C d
is given by the diagonal matrix
( f 1 0 0 0 0 f 2 0 0 0 0 f 3 0 0 0 0 f d ) .
Now take the case where µ is counting measure on the sigma-algebra of all subsets of a countable set X. Let f be any real-valued function on X. Let D f 0 be the subspace of L2(µ) consisting of all functions g for which {g ≠ 0} is a finite set, and let D f 0 be the restriction of Mf to D f 0. Then it is readily checked that D f 0 is essentially self-adjoint. Consequently, the restriction of Mf to any subspace of Df larger than D f 0 is also essentially self-adjoint.

A.6. The Spectral Theorem

The spectral theorem for a self-adjoint operator A on a separable complex Hilbert space H says that there is a sigma-finite measure space ( X , , µ ), a unitary isomorphism
U : H L 2 ( μ )
and a measurable real-valued function f on X such that
A = U 1 M f U .
Expressing A in this way is called a diagonalization of A (the terminology being motivated by Equation (114)).

A.7. The Functional Calculus

If g is any measurable function on R we can then form the operator
g ( A ) = def U 1 M g ° f U .
If g is a polynomial then g(A) works out to be what it should be, a polynomial in A. Another example, is the function g(x) = eikx, where k is any constant; this gives the operator eikA.

A.8. The Spectrum

The essential range of f is the smallest closed subset of R whose complement U satisfies µ (f−1(U))=0. It consists of all λ ∈ R for which the operator Mf− λI=Mf−λ has a bounded inverse (wich is M ( f λ ) 1). This essential range forms the spectrum σ(A) of the operator A. Thus σ(A) is the set of all real numbers λ for which the operator A− λI has a bounded linear operator as inverse.

A.9. The Spectral Measure

Associate to each Borel set ER the operator
P E = M 1 f 1 ( E )
on L2(X, µ). This is readily checked to be an orthogonal projection operator. Hence, so is the operator
P A ( E ) = U 1 P E U .
Moreover, it can be checked that the association EPA(E) is a projection-valued measure, i.e., PA(∅) = 0, PA(R) = I, PA(E ∩ F ) = PA(E)PA(F), and for any disjoint Borel sets E1, E2, … and any vector xH we have
P A ( n 1 E n ) x = n 1 P A ( E n ) x .
This is called the spectral measure for the operator A, and is uniquely determined by the operator A.

A.10. The Number Operator

Let us examine an example. Let W = {0, 1, 2, …}, and let µ be counting measure on W. On W we have the function
N : W R : n n .
Correspondingly we have the multiplication operator MN on the Hilbert space L2(W, µ). Now consider the Hilbert space L2(R). We have the unitary isomorphism
U : L 2 ( R ) L 2 ( W , μ ) : f ( f , ϕ n ) n 0 .
Consider the operator N on L2(R) given by
N = U 1 M N U .
Then
N f = n W n f , ϕ n ϕ n
and the domain of N is
D N = { f L 2 ( R ) : n W n 2 | f , ϕ n | 2 < } .
Comparing with Equation (82) we see that
( N f ) ( x ) = ( d 2 d x 2 + x 2 4 1 2 ) f ( x )
for every f S ( R ).
Thus the self-adjoint operator N extends the differential operator d 2 d x 2 + x 2 4 1 2, and, notationally, we will often not make a distinction. In view of the observation made at the end of subsection A.5, the differential operator d 2 d x 2 + x 2 4 1 2 on the domain S ( R ) is essentially self-adjoint, with closure equal to the operator N.
The operator U above helps realize the operator N as the multiplication operator MN′, and is thus an explicit realization of the fact guaranteed by the spectral theorem.

B. Explanation of Physics Terminology

In quantum theory, one associates to each physical system a complex Hilbert space. [ 4 ] Each state of the system is represented by a bounded self-adjoint operator ρ ≥ 0 for which tr(ρ) = 1. An observable is represented by a self-adjoint operator A on . The relationship of the mathematical formalism with physics is obtained by declaring that
tr ( P A ( E ) ρ )
is the probability that in state ρ the observable A has value in the Borel set ER. Here, PA is the spectral measure for the self-adjoint operator A.
The states form a convex set, any convex linear combination of any two states being also clearly a state. There are certain states which cannot be expressed as a convex linear combination of distinct states. These are called pure states. A pure state is always given by the orthogonal projection onto a ray (1–dimensional subspace of ). If ϕ is any unit vector on such a ray then the orthogonal projection onto the ray is given by: Pϕ:ψ ↦ ⟨ψ, ϕϕ and then the probability of the observable A having value in a Borel set E in the state Pϕ then works out to be
P A ( E ) ϕ , ϕ .
Suppose, for instance, the spectrum of A consists of eigenvalues λ1, λ2, …, with Aun = λnun for an orthonormal basis {un}n≥1 of . Then the probability that the observable represented by A has value in E in state Pϕ is
{ n : λ n E } | u n , ϕ | 2 .
Thus the spectrum σ(A) here consists of all the possible values of A that could be realized.
To every system there is a special observable H called the Hamiltonian. The physical significance of this observable is that it describes the energy of the system. There is a second significance to this observable: if ρ is the state of the system at a given time then time t later the system evolves to the state
ρ t = e i t H ρ e i t H ,
where ℏ is Planck’s constant.
A basic system considered in quantum mechanics is the harmonic oscillator. One may think of this crudely as a ball attached to a spring, but the model is used widely, for instance also for the quantum theory of fields. The Hilbert space for the harmonic oscillator is L2(R). The Hamiltonian operator, up to scaling and addition of the constant 1 2, is
H = d 2 d x 2 + x 2 4 1 2 .
The energy levels are then the spectrum of this operator. In this case the spectrum consists of all the eigenvalues 0, 1, 2, … The creation operator bumps an eigenstate of energy n up to a state of energy n + 1; an annihilation operator lowers the energy by 1 unit.
In many applications, the eigenstates represent quanta, i.e., excitations of the system. Thus raising the energy by one unit corresponds to the creation of an excited state, while lowering the energy by one unit corresponds to annihilating an excited state.

Acknowledgments

A first version of this paper was written when Becnel was supported by National Security Agency Young Investigators Grant (H98230-10-1-0182) and a Stephen F. Austin State University Faculty Research Grant; Sengupta was supported by National Science Foundation Grant (DMS-0201683) and is currently supported by National Security Agency Grant (H98230-13-1-0210).
The authors are also very grateful to the three referees for their remarks and comments.

Author Contributions

This work is a collaboration between the authors Becnel and Sengupta.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Schwartz, L. Théorie des Distributions; Herman: Paris, France, 1950; Volume 1. [Google Scholar]
  2. Schwartz, L. Théorie des Distributions; Herman: Paris, France, 1951; Volume 2. [Google Scholar]
  3. Simon, B. Distributions and Their Hermite Expansions. J. Math. Phys. 1971, 12, 140. [Google Scholar] [CrossRef]
  4. Von Neumann, J. Mathematical Foundations of Quantum Mechanics; Princeton University Press: Princeton, New Jersey, USA, 1957. [Google Scholar]
  5. Glimm, J.; Jaffe, A. Quantum Physics; Springer Verlag: New York, NY, USA, 1987. [Google Scholar]
  6. Gelfand, I.M.; Vilenkin, N. Generalized Functions; Academic Press: New York, NY, USA, 1964; Volume IV. [Google Scholar]
  7. Becnel, J.; Sengupta, A. White Noise Analysis: Background and a Recent Application. In Infinite Dimensional Stochastic Analysis: In Honor of Hui-Hsiung Kuo; World Scientific Publishing Company: Singapore, 2008. [Google Scholar]
  8. Hida, T.; Kuo, H.-H.; Potthoff, J.; Streit, L. White Noise: An Infinite Dimensional Calculus; Kluwer Academic Publishers: Norwell, MA, USA, 1993. [Google Scholar]
  9. Kuo, H.-H. White Noise Distribution Theory; CRC Press: Boca Raton, Florida, USA, 1996. [Google Scholar]
  10. Becnel, J. Equality of Topologies and Borel Fields for Countably-Hilbert Spaces. Proc. Am. Math. Soc. 2006, 134, 313–321. [Google Scholar]
  11. Hermite, C. Sur un Nouveau Développement en Série des Fonctions. Comptes rendus de l’Academie des Sciences 1864, 14, 93–266. [Google Scholar]

Share and Cite

MDPI and ACS Style

Becnel, J.; Sengupta, A. The Schwartz Space: Tools for Quantum Mechanics and Infinite Dimensional Analysis. Mathematics 2015, 3, 527-562. https://0-doi-org.brum.beds.ac.uk/10.3390/math3020527

AMA Style

Becnel J, Sengupta A. The Schwartz Space: Tools for Quantum Mechanics and Infinite Dimensional Analysis. Mathematics. 2015; 3(2):527-562. https://0-doi-org.brum.beds.ac.uk/10.3390/math3020527

Chicago/Turabian Style

Becnel, Jeremy, and Ambar Sengupta. 2015. "The Schwartz Space: Tools for Quantum Mechanics and Infinite Dimensional Analysis" Mathematics 3, no. 2: 527-562. https://0-doi-org.brum.beds.ac.uk/10.3390/math3020527

Article Metrics

Back to TopTop