ZERO-RANGE INTERACTIONS WITH AN INTERNAL STRUCTURE

B.S.Pavlov
Department of Mathematical and Computational
Physics, Institute for Physics, Leningrad
State University, St.Peterhoff
198904 Leningrad, USSR

The purpose of this paper is to outline a general scheme for constru-

ction of zero-range interactions with an internal structure. This
construction can be applied to various situations of physical interest
- see [I-3]. Here we concentrate on the mathematical aspects of the
problem.

A zero-range interaction with an internal structure is usually con-

structed as a self-adjoint extension of a direct sum of operators ;
typically one of them is a differential operator and the other is an
abstract one. The standard yon Neumann theory works well for differen-
tial operators, though it is not very suitable for concrete calcula-
tions. A serious difficulty arises, however, if the abstract operator
is not densely defined. The purpose of our construction is twofold :"

(i) to reformulate the standard theory in terms of boundary condi-

tions making it better adapted to practical calculations,

(ii) what is more important, to generalize it in such a way that it

can handle also non-densely defined operators. Notice that it
is necessary once we want to treat finite-dimensional Hilbert
spaces.

The scheme of the construction described here has been published

first in Ref. 1 and discussed further in [2-5] ; see also the litera-
ture in [5].

I. The self-ad~oint extensions

Let A be a self-adjoint operator in a Hilbert space ~and let

U ~ ( A - i I ) ( A + i I ) -I be its Cayley transform. Let ~ be some minimal
generating subspace of A , i.e.,

(1) V i :
1
(ii) inu i = {ol
If dim~i~oO , the last condition is equivalent to

(iii) cos(~ i , U ~ i) ~sup llulllvll

which is stronger in general.

For further purpose, we denote ~ - i m U ~ i . It is useful also to define
analogous subspaces ~ , ~ for any complex ~ outside ~(A) ; they
are related by ~=U~ with U ~ = (A-dI)(A-T~I) -I.
Now we are going to restrict the operator A . We introduce the
following linear set in the domain D of A ,

DO = (A-iI)-1~A--i = (A+iI)-1~ti- '

and restrict the operator A to A O ~ A ~ D O • Our aim is to construct

all self-adjoint extensions Ap of A 0 . In distinction to the Stan-
dard theory, we do not require A 0 to be densely defined. Then the
adjoint A 0 may not exist, and consequently, the definition of defi-
ciency subspaces must be used in the form

~±i = [Ran(A0 ~ i I ) ] ~

The following assertion is valid :

Lemma I : (i) The deficiency subspaces ~±i of A0 coincide

with ~ ± i "
(ii) The operator A0 is densely defined iff D ±i = 0
The analogous assertion can be proven for the subspaces ~ corres-
ponding to any ~ ~ kg(A) .

Proof : (i) follows directly from the definition of DO and (ii) is

a consequence of the relation

<(A~il)u0,@±i > = (u0,(A±il)@±i> = 0

which is valid for every U O ~ DO ' ~±i ~ ~ ± i N D since A±iI maps DO

into ~±Ai .

The construction of the extensions Aft proceeds in a usual way. We

restrict the^Cayley transform U to U O~ U~_i ^and construct all
isometries Up : ~_i--7~+i . Combining U 0 and Uff , we get a unita-
ry operator whose Cayley pull-back is the sought extension Ap .
Next we need a technical lemma :

Lemma 2 : If the condition (iii) is fulfilled and ~@s~ is a Riesz

basis (see [6]) in ~i ' then the vectors ~W~, We} defined by

+ U~+I ~I
Ws:T0s , wZ:--es
A
form a Riesz basis in the subspace ~ = ~i + ~-i "
Proof is obvious.

Hence any vector u~ can be decomposed as

u = fS (-+W
~s +s + ~ w ~ ) , ~s± g C

Introducing the vectors ~ ± = ~ s ~ s±e s ~ # i , we c~n rowrito this

decomposition as

= A(A-iI)-~ + + (A-iI)-~- .

Using the basis { W + , W s } , we define further the linear operator

: ~-~ by the relations

^AW s
+=-Ws , ^A W _e = W +s

If A0 is densely defined, then It suggests that we can~= A 0~

use this operator for construction of the extensions even if A 0 does
not exist.

Remark : In most of applications, in particular, those mentioned

above [I-5] we use formally the symbol A 0 . In the case of non-densely
defined operators, it should be understood as described here.

Lemma__~ : Under the hypotheses of Lemma 2, the following formula is

valid

qcu) f~v)> @¢~) F~v)>

whore u,v are arbitrary vectors of ~ with the decompositions

~ f[~)w~+~;~.)w:~= ,c, i i ) ~ + ¢ u ) + ~, ii) IS ~v)

 v = Z ~S~+<v>w~++~cv~w~ = ~ c ~ - ~ - ~ + c v ~ + ~,-i~-'~-~v~
8

~roof : ~t follows from the d e f i n i t i o n of the basis ~,',,~ ~ that

<~+s,~+~> -<W~,V"~>-- ~es,Ot> ,

<W~,Wt~-<.~,~t>= 0
for all s,t . The result is now obtained by a straightforward calcu-
lation.

Lemma 5 provides us with the universal symplectic boundary form

J<u,~ -- ~ A u , v ) - 4 ~ , A v ) =~'{-~u~,~÷Cv~> - ~ + ~ u ~ , ~ - C v ~ > ,

where the rhs gives the expression of J(u,v) in terms of the vec-
tors ~ ± . The latter play here the role of usual boundary values
known from the theory of differential operators.

If A is an orthogonal sum of self-adjoint operators,

A = A0@) AI

on the Hilbert space ~= ~0 ~ ~I ' then its boundary form equals to

the sum of the boundary forms of its parts,

J(u,v) = J0(u0,v0) +J1(Ul,Vl) ,

where u = (Uo,U I) and v = (Vo,V I) .

Now we are going to formulate the m a i n result :

Theorem I : Under the assumptions (i) and (ii), all self-adjoint

extensions .~°f the operator A0 are determined by Lagrange hyper-
planes in ~ on which the form J vanishes. Each of the hyperplanes
can be specified by the boundary conditions

J- =r~ ÷ , ~i)
where P is a self-adjoint operator, p = p ~ , in ~i "

Remark : In order to describe all the self-adjoint extensions, one

can start with a fixed hyperplane. All other Lagrange hyperplanes can
be obtained from the initial one by J-unitary transformations.
Proof : What we need is to relate the boundary conditions (I) to the
isometries Ur : ~ _ i - - ~ i . We d e f i n e Up as t h e C a y l e y t r a n s f o r m o f
the operator Ap obtained as the restriction of ~ to the linear
m a n i f o l d d e t e r m i n e d by t h e b o u n d a r y c o n d i t i o n s ( 1 ) . U s i n g t h e r e l a -
tions

(~-iI) Z ( ]+s W+s + ~ s W s ) : Z ((~s-i~

+ ) W"+ - - (~s++i~s)Ns)-
~- =
S S

: A(A-iI)-1(~--i~ +) -(A-iI)-1(~++i] -) : ( p - i I ) ~ + --- v

and

(A+il) Z + + +fsWs ) = A(A-iI)

(IsWs -~(~-+i$-) -(A-i~) -1(y+_i.C) _-
S

= (A+iI) (A-iI) -I (P+ iI)l + - w

*k
we see that the Cayley transform Up of Ap map-s ~-i into ~'i
A
according to v = U p w , i.e.,

Up = ( P - i l ) ( P + il)-1(A+il)-1(A-il) (2)

Hence it differs from U by the factor (p-iI)(P+iI) -I only, and

the latter is a unitary operator in ~ i " However, if ~ runs through
the set of all self-adjoint operators on ~i ' the formula (2) yields
a parsmetrization of all isometries ~ i-~i . The only exception is
the original isometry U r = U . Using now the Cayley pull-back of Up ,
we get the sought self-adjoint extension Ap . The only exception is
again the original operator A to which the boundary condition
~+= 0 corresponds.
^
Corollar 2 1 : The domain Dp of the extension A2 consists of the
vectors

u = u 0 + A ( A - i I ) - 1 ~ + + (A-iI)-ly - = u O + [ A ( A - i I ) -I + (A-iI)-1]f +,
(3)
where u0~ D O and ~+ runs through ~i "

It means that the domain D~ of A2 can be represented in the form

A
D = D O + ~p , (4)

where ,~p is the Lagrange hyperplane in ~ specified by t h e b o u n d a r y

condition (I). To get this representation, we employ the relations
between A and U ,

Dp (~p+iz)-12= 1 ,, ,,
: ~(z-up)~t~
2 I ( z-up)~t_i =
- i + ~-[

1
- ~i(z-u)~_ti +~(z-up)kt
. ~ ~ ""
_~ -- Do+ ~tr

The last decomposition is a direct sum if ~i C D and the operator

P A P + P i s invertible, where P is the projection to ~i " In fact,
in this case DO and ~p are linearly independent for

(A-iI)-lh+A(A-iI)-~ + +P(A-iI)-I~ + = 0

if and only if P(A+p)~ + = 0 , i.e., ~ + = 0 and h =0 . Then both the

components in the decomposition u=u0+h , where U o ~ D O and h ~ ; ,
are uniquely determined,

h = [A(A-iI)-I+ P(A-iI)-l]~ + ,

+
= (pAP + p ) - l p ( A - i I ) u

In conclusion, let us illustrate the construction described above.

Example : Consider the simplest situation involving a non-densely

defined operator in a finite-dimensional Hilbert space ~ , d i m ~ = n .
Let A= s~ses~.,es> be a self-adjoint operator in ~ with a simple
spectrum, ~s ~ ~ t for s ~ t . Let ~i = ~gj be the subspace spanned
by a vector g = ~s gsee " Then ~i- = ~ g ' ~ ' where g'=

= ~s (~s+i)(~s-i)-Igses and the elements

W+ = A(A-iI) -1 g , W _ = (A-iI) _ Ig

form a basis in ~=~i+~_i . The Lagrange line specified by the

boundary condition ~ - = ff~+ is spanned by the vector

~+ Z (~s +P) (~s-i)-Igses '

S

and the decomposition (4) is a direct sum if ~Ag,g~+P ~ 0 . We

have

S
 Ax = Z ~s(~s-i)-1[hs-<h,g~gs+~+(~s+P)gs]e s '
S
+
Apx : Z (~s-i)-1[~s(hs-<h'g>gs)+~ (/'~(s-1)gs]es
S

so A and Ap differ by a rank-one operator only,

(Ap-A)x = -]+(x) (A+iI)g ,

where ~+(x):[<Ag,g~+P] -1~(A-iI)x,g~

2. Krein formula

In this section, we are going to derive Krein's formula for resol-

vents. Let A p = A 0 + ~ P be one of the self-adjoint extensions con-
structed in the preceding section. To obtain its resolvent, we have
to solve the equation

(Ap-~I)u = f (5)

for all fg~. First we prove

Lemma 4 : The deficiency subspaces ~ i and ~ i_ corresponding to i

and ~ , respectively, are related by

A+iI

proof : The result follows from the relation (in an obvious notation)

0 = ~(A - ~ I ) D o , ~ > = <(A - iI)Do,(A + i I ) - 1 ( A - & I ) ~ a >.

Hence (A+iI)-I(A-JG)~&~ (A-iI)D 0 and Lemma I applies.

The relation (6) defines an isomorphism of ~ i with ~ which

ascribes to every deficiency vector ~ i G ~i a deficiency vector
e4~ ~.
Lemma ~ : The deficiency vectors 84 and 0i are related by

I I+ A @i + ~ O i
@4- A-iI(I-P)A~-~ + A__~ii_I+
Y~:'~A @i ' (7)

where P is the orthogonal projection onto ~ i " The first~ term

on the rhs belongs to D O , while the remaining ones to ~ .
 10

Proof is based on the relations (6) and

A + iI = ( A - i I ) - I ~ I + ~ A + A ( A - & I ) ]

together with

(A-iI)-I(I-P)(A-&I)-I(I+&A) ~i E D O

The formula (7) shows that ~ can be expressed in the form (3) if
we choose

+
= Oi ,
-I
~ - = P(A-J~I) (I+&A) ~i : Q(.A)~i

The function Q(~) whose values are operators in ~ i is a very

important object in our theory ; it is called Schwartz integral of
the operator A associated with the deficiency subspace ~i :

f I + as
Q(~) = ~L-~ d(PEsP) ' (8)

where Es is the spectrsl measure of A . It is clear that Q(.) is

an operator-valued R-function, i.e., its imaginary part is positive
in the upper halfplane. Since ~i is supposed to be a generating
subspace we see that Q(~) contains all the spectral information
about A . Hence it plays a crucial role in construction of the resol-
vent of A

Theorem 2 : The resolvent (Ar-~I) -I is given by the Krein's formula

(Ap-aI) -I = (A-&I) -I + ( A + i I ) P [ P - Q ( a ) ] - I p (A-iI)(A-4I) -I

Proof : According to Corollary I, we look for a solution to (5) in

the form (3) with u O = (A-iI)-Iv & , where ~ + and ~- are related
by the condition (I) and v~ ~i " Using the identities

^
AA(A-i~) - 1 ~ + =-(A-i~)-1~ -
,

A(A-iI)-I~ - = A(A-iI)-I~ +

A
which follow from the definition of A , we rewrite the eq.(5) in the
form
 11

(A~-~I)u = (A-~I)(A-iI)-Iv&+ (A_di)(A_ii)-1~ - _(I+ A)(A-iI)-I~ + =

= f (9)

Applying the projection P to the both sides and multiplying by the

bounded operator (A-iI)(A-~I) -I , we get

PJ+ = J- = P(A-~I)-I(I+~A)~++p(A_iI)(A-~I)-If ,

and therefore

~+ = ~p - Q(d)~-1 p (A_ii)(A_~i)-If

Substituting this to (9) we find v~ so (3) yields the sought

vector u .

References

I B.S.Pavlov, Teor.Mat. Fiz.59 (1984),159-169.

2 Yu.A.Kuperin, K.A.Makarov, B.S.Pavlov, Teor.Mat.Piz.69 (1986),
100-114.
3 V.M.Adamian, B.S.Pavlov, Leningrad Branch of Steklov Institute
Zapisky 149 (1986),7-23.
4 Yu.A.Kuperin, K.A.Makarov, S.P.Merkuriev, A.K.Motovilov,
B.S.Pavlov : Quantum scattering theory with energy-dependent
potentials, part II, Vilnius 1986 ; pp.28-73.
5 B.S.Pavlov, Sov.Math.Uspekhi 42 (1987), No.6, 99-131.
6 I.C.Gohberg, M.G.Krein : Introduction to the Theory of Linear
Non-selfadjoint Operators, Nauka, Moscow 1965 (in Russian).
7 M.G.Krein, Acad. Sci.USSR Doklady 52 (1946),657-660.
 EVOLUTION EQUATIONS
AND
SELFADJOINT EXTENSIONS

Hagen Neidhardt +)
Laboratory of Theoretical Physics
Joint Institute for Nuclear Research
141980 Dubna, USSR

1. Introduction
Let ~ be a separable Hilbert space with scalar product <.,.) and
norm 11.11. We assume that we have a family [H(t)1 t ~ 1 of selfadjoint
operators o n ~ . For instance, this can be a family of SchrSdinger
operators ~-A + q(t)~ t ~IRI on L2(~ n) which naturally arises from a
time-dependent potential q(.,t). With ~H(t)}t E~I we associate an
evolution equation

i ~ = H(t)u(t), (1.1)

u(t)It= s = x, (1.2)

x~ , t,s ~ 1 . The problem is to find the solution operator of this

evolution equation, i.e. a strongly continuous family ~U(t,s)~tpsET~
of unitary operators on ~ such that in some sense u(t) = U(t,s)x is
a solution of (1.1) and (1.2). To attack this problem there are se-
veral possibilities.
The first investigations of this problem were carried out by
T.Kato [I~ who returned to it several times [2-6]. But there is a
lot of other authors who were interested in the same problem, among
them H.Tanabe, P.E.Sobolevskij, J.A.Goldstein, S.G.Krein, J.R.Dor-
roh, A.Yagi etc.. The methods used by these authors are similar to
those of the theory of ordinary differential equations.The main
task is to find suitable equations, which approximate the original
equation and which are easy to solve, and to show that the arising
approximation solution operators converge to that of the original
problem.

+)On leave of absence from Karl-Weierstra~-Institut ftLr ~thematik,

Akademie der Wissenschaften der DDR, 1086 Berlin, DDR
 13

An application of the abstract results to the time-dependent

Schr~dinger operator ~ - A + q(t)~te ~ can be found in ~7,8].
In the following we use an approach to this problem which is
quite different from the mentioned one and which has been proposed in
[9,10~. This approach goes back to ideas, which can be found by
O.A.Lady~enskaja [11,12], J.L.Lions [13,14,15~, G.DaPrato [16,17],
M.Iannelli [18], G.DaPrato and P.Grisvard ~19~ and L.Paquet [ 2 ~ . In
application to the present Hilbert space situation the method can
be described as follows.
Let z be a complex number, Im(z) @ O, such that for every x ~
the map

(H(.) - z) -1 x : ~x 1 '' ~~ (1.3)

is measurable. By L 2 ( ~ 1 , ~ ) we denote the Hilbert space of all

measurable ~ -valued functions on ~ I such that f(.)~ f(.)ll is Bochner
Integrable. The scalar product (.,.) of L 2 ( ~ 1 , ~ ) is given by

(f,g) = ~T~1<f(t),g(t~dt (1.4)

f , g e L 2 ( ~ l , ~ ). With the family ~H(t)}tE&l we associate a selfad-

joint operator H on L 2 ( ~ I , ~ ) defined by

dom(H) = ~ f £ L 2 C q ~ x l , ~ ) : f C t ) a domCH(t)) a . e . taTPx1,

(1.5)
H(t)f(t)E L 2 ( ~ 1 , ~ ) ,

(Hf)Ct) = HCt)f(t), (1.6)

f ~ dom(H). The selfadjointness of H is guaranteed by the measura-

bility condition (1.3).
Further we consider the shift group e-i~Ko , ~ e ~ I, on L2(~I,~)
given by

(e-i~Ko f)Ct) = fCt -~), (1.7)

t ~1,~ f L2(~I,~ ). It is easy to show that K O = -id , where

- i ~ i s defined via the Fourier transform as the usual multiplica-
tion operator induced by the independent variable t.
We assume that the set dom(K),
 14

@om(K) = dom(K O) ~ dom(H) (1.8)

is dense in L2(~I,~).-- Obviously, this condition implies some re-

strictions on [H(t)~te~l.U but they are not very hard. Then we in-
troduce the operator ~,

Kf = Kof + Hf, (1.9)

fa dom(K). The operator K is densely defined and symmetric but in

general not closed. Moreover, denoting by M ( ~ ) , ~ C ~ ( q ~ I ) , the
bounded multiplication operator induced-by ~(.), i.e.

(MC~)f)Ct) = 4 ~ ( t ) f ( t ) , (1.10)
1 4 _

tions ~ : with compact support, we find

M(q~)dom(K) & dom(K) (1.11)

and

~(4~)f - M(~)# = -i~(~)f, (1.12)

~e = ~ , for every f&domCK) and every ~aC1o(~l). Let C ( q ~ l , ~ )

the set of all continuous functions f~.;:
~ \~, ~I > % . Then,in ad-
dition we have

dora(K) & L2(~1,~) O C(~1,~). (1.13)

This inclusion follows from dom(K) & dom(K o) ~ L2(I~I,~ )

n c( ~ 1 , ~ ).
Denoting by K the closure of K i t i s not hard to show [10]
that K fulfils the conditions (1.11) - (1,13), too. This means we
have

M(~)dom(K) ~ dom(K) (1.14)

and

KM(~)f - M(~)Kf = -iM(~)f (1.15)

 15

for every f~dom(K) and every ~ a C 1 o ( ~ I) as well as

dom(K) ~ L2('lR, l,~) ~ C(1"~,1,"~)o (1.16)

The last fact needs some more efforts than the other two.
Ass1~ming now for a moment that K is selfadjoint and that the
conditions (1.14) and (1.15) are fulfilled, then in accordance
with J.S.Howland [21] there is a measurable family [ U ( t , s ) ~ t , s a ~
of unitary operators such that the representation

(e-ilK f)(t) = U(t,t-~)f(t-~), (1.17)

f6L2(T~I,~),-- is valid. Furthermore, if (1.16) is satisfied and

the set

(dom(K))t = ~ x ~ : ~fKdem(K), x = f ( t ) ~ , (1.18)

1, i s dense i n f o r every 1, then be

continuously chosen [221.
As has been pointed out in KlOB this continuous two-parameter fa-
mily ~ U ( t , s ) ~ t , s ~ 1 of unitary operators can be regarded as the
first candidate for a solution operator of the evolution equation
(1.1) and (1,2). In [I0~ additional conditions were made which
really allow to verify the usual properties of a solution operator
for ~U(t,s))t,s a~1 . Several applications of this approach to evo-
lution equations were given in [23~.
If now K is not selfadjoint, then naturally the problem arises
to extend K to a selfadjoint operator K preserving the conditions
(1.14) - (1.16) and (1.18). But simple examples show that that is not
always possible. Hence?we need conditions which guarantee the ex-
istence of such an extension K of K. To find such conditions will
be the aim of the present note.
As the first step on this way we concentrate our attention on
the conditions (1.14) and (1.15)and forget (1.16) and (1.18).
Therefore~the problem will be now to find conditions allowing the
existence of symmetric, maximal symmetric and selfadjoint exten-
sions of K obeying (1.14) and (1,15).
Besides,a further motivation of this problem comes from the
so-called Weyl commutation-relation [24]. Let Q be the selfadjoint
multiplication operator on L2( ~ I , ~ ) induced by the variable t,
i.e.
 16

(qf)(t) = tfCt), (1.19)

f a d o m ( Q ) = ~ f ~ L 2 ( 1 ~ I,~): tf(t)aL2(1~ I,~)~. If K is selfadjoint

and obeys (1.14) and (1.15), then using the considerations of
J.S.Howland [211 ~K,Q~ forms a Weyl pair, i.e.

e-i~K e-i%Q = ei ~ e-ilQ e-i~K , (1.2o)

~ , % g R 1. If K is not selfadjoint but obeys (1.14) and (1.15), then

K,Q~ forms in some sense an incomplete Weyl pair. Hence,to solve
the above proposed problem means to complete ~K,Q~ to a Weyl pair
~K,Q~, K~K. From this vlew-point the paper is related to that of
B.Fuglede [25~, N.S.Poulsen [26], P.E.T.J~rgensen [27S, P.E.T.J~r-
gensen and P.S.Muhly [28~, W.J.Phillips ~2~, K.SchmGdgen [30,31~,
G.Dorfmeister and J.Dorfmeister ~32~.
In the end we remark that unfortunately page restrictions make
it impossible for us to present the proofs of the results. The
reader, who is interested in proofs, must be refered to [3~ •

2. The associated representation

We introduce the set A C ( ~ I) of all loca}ly absolutely continuous
functions ~ : ~1~__~I such that ~ and ~ belong to L ~ ( ~ I ) .
Lemma 2.1. Let K be a symmetric operator on L 2 ( ~ I , ~ ) obeying
(1.14) and (1.15). Then,for e v e r y ~ A C ( ~ I) we have

M(~)dom(K ~) ~ dom(K ~) (2.1)

and

Z~M(~)f - M(~)K~f = -iM(+)f, (2.2)

f ~ dom(K ~).
The proof is obvious. Since every symmetric extension of K
is a restriction of K ~ to some domain ~ , dom(K) ~ ~ dom(K~),
the problem to obtain an extension of K preserving (1.14) and
(1.15) reduces to the existence of a domain ~ , dom(K) E ~
dom(K ~) such that

M(,~)~ c_ ~ (2.3)
 17

is valid for every ~ E C ( ~ 1 ) and K = K ~3 is symmetric.

By ~ - i and ~ i we denote the defect spaces

~-i = L 2 ( ~ 1 ' % ) E) ima(K + i) = ker(K~ - i) (2.4)

and

~i = L2(~1'~) ~)ima(K - i ) = ker(K ~ + i ) , (2.5)

where ima(.) is the range of an operator. The ordered pair

~dim(~_i),dim(~i) ~ is called the defect index of the operator K.
It is well-known that there exists a selfadjoint extension of K if
and only if dim(~_i) = dim(~i).
Now the domain dom(K~) admits the decomposition

dom(K~) = dom(K) $ ~-i $ q~i' (2.6)

where X $ Y means X ~ Y = {0~ [341. Because of (2.6) we have

M(q~)g 1 = f l ( ~ ) + g11(~) + g21(~), (2.7)

fl(~)adom(K), gll (~) & ~ - i ' g21 ( ~ ) ~ i ' for every gle'Q-i and
~ £ A C ( ~ I ) . Similarly, we get

M(~)g 2 = f2(~p) + g12(~) + g22(~), (2.8)

f2(~)-Cdom(K), g 2 1 ( ~ ) E ~ _ i , g 2 2 ( ~ ) & ~ i , for every g 2 ~ i and

~ AC(I-~I). Hence,we can introduce the linear operators Aij (~),

gij(~) = Aij(~)g j, (2.9)

i,J = 1,2, acting between the spaces ~ _ i and ~i"

Further we say a sequence ~ n } n _ I of elements of A C ( ~ I) con-
verges pointwise to an element ~ of A C ( ~ I) if

s#p.LlSn;.r, (. 11 + < (2.101

lim ,=,o~n(O) = ~ ( 0 1 , (2.11/
n~+

nm
n---) +oo
Sn(tl = Jp(tl (2.12/
 18

for a.e. t ~ l . It is not hard to see that (2.10) - (2.12) imply

lim ~n(t) = ~ ( t ) for every t £ ~ .
n--->+~o
Moreover, we introduce the Hilbert space % = ~ - i x ~ i with
the scalar product

) g2 % (f2'g2)'

(~]E~ ' /gll£~'g2 f l ' g l & q ~ - i ' f2'g2E~i and the norm " . ' ~ .
Equipping the Hilbert s p a c e % with the bilinear form L.,.3,

If,g] (Jr,g)%,
= (2.14)

f,g & % , where J is a bounded operator on~W~ given by

55 transforms into a Krein space. An introduction to the theory of

Krein spaces can be found in [35,36]. By means of the bilinear form
[.,.3 it is possible to define the J-adjoint operator of a linear,
densely defined operator on ~3~,36]. We denote the J-adjoint of an
operator A on 3~ by A £ ~ We have

Ar"*~]= JA~J. (2.16)

We introduce the map A C ( ~ 1) B ~ >~(%), where ~ ( ~ ) i s the

set of linear operators o n ~ defined by

A(~)f \121(#) t22(~ ) f2

f= (f
Theorem 2.1. The map A C ( ~ I ) ~ ' ~(~4j) defined by (2.17) saris-
fies the following properties:
(i) For e v e r y ~ 6 AC(qRI)~ A ( ~ ) is a bounded operator on ~ such
that

~A(~p)ll% ~ I~IIL~(~I ) + II+llL~(~l ). (2.18)

(ii) For every + I E A C ( ~ I) and ~ 2 & A C ( ~ I) we have

 19

A('~I) + A('~2) = A(4P1 + 4P 2) (2.19)

and

A('~P1)A(8~2) = A ( ~ 1 ~ 2 ) = A(~2)A('~I). (2.20)

(lli) If the sequenee {~n~n=l of elements of A C ( ~ 1) pointwise

converges to ~ A C ( ~ I ) , then

e-lim A ( ~ n ) = A(~). (2.21)

n--@ + ~
The convergence is understood as a strong convergence on ~ regar-
ded as a Hilbert space.
(iv) For every ~ A C ( ~ I) the relation

A ( ~ ) -- ALI'](~ ) (2.22)

holds, where ~ is the complex conjugated function to ~ .

We remark that

A(1) = I, (2.23)

I ~ AC(~I), holds. By equipping the set AC(~1) with the norm

II.fiAt(~I),

II~IIAC(,~I) = /I~IIL~(~I) + U~L~,, ( ~ 1 ) , (2.24)

~ A C ( ~ 1), AC(~I) forms a symmetric commutative Banach algebra

with identity. The involution is given by ~ - * ~ .
According to M.A.Najmark and R.S.Ismagilov ~37] the map ~I
A ( ~ ) is a symmetric representation of the symmetric commuta-
tive Banach algebra A C ( ~ I) on the Kreln space ~ . As we can
define such a symmetric representation of A C ( ~ 1) on ~ - i x ~ i
for every symmetric operator K satisfying (1.14) - (1.15)~we call
the map ~ --> A(~ ) the associated representation of K.
It is unclear how characteristic is the associated representa-
tion. This means.Let ~--* A ( ~ ) and ~ - - * A ' ( ~ ) be the associated
representations of K and K' such that they Lre (J,J')-equivalent. If
K and K' are completely non-selfadjoint, does it follow from this
assumption that K and K' are unitarily equivalent?
 20

Furthermore,the problem is unsolved if every symmetric repre-

sentation~ ~ A ( ~ ) of the symmetric commutative Banach algebra
A C ( ~ I) cam be regarded as the associated representation of some
symmetric operator K satisfying (1.14) and (1.15).

~, The associated ~roup

We set ~ w (t) = exp(-iWt), ~ ,t ~ . The functions ~ belong to
A C ( ~ I) for every ~ ~ I . Hence,we can define

U(~ ) -- A(~-). (3.1)

Theorem ~.I. The map ~ 1 8 ~ _ ~ U ( ~ ) ~(~) forms a one-parameter

strongly continuous group on ~ such that

~IU(~ )II~ ~ I + I~"| , (3.2)

u~(~ - ) = u ( - ~ ), (3.3)
~a~l.
Theorem 3~I is a consequence of Theorem 2.1. A linear boumded
operator U on a Krein space satisfying UUe]~ U -1 is called a
J-unitary operator. Consequently, U(~ ), ~ ~ I , is a one-para-
meter strongly continuous group of J-unitary operators o n ~ . We
call U(~ ),~ ~ I , the associated group of the symmetric operator
K. The one-parameter strongly continuous group U(~ ), ~ E ~ I, has
the representation

U(~ ) =
e -i~L , (3.4)

E~I, where L is called the associated generator of the symmetric

operator K. The operator L is J-selfadjoint, i.e. L E~3 = L.
Corollary 3.2. The spectrum spec(L) of the associated generator L
of the symmetric operator K is contained in ~ I , i.e. spec(L) ~ ~ I .
This property is a consequence of (3.2).

4. Symmetric extensions
Now we want to find conditions in terms of the associated represen-
tation or the associated group which guarantee the existence of sym-
metric extensions K of K obeying (1.14) and (1.15). Naturally,we
use the theory of J.von Neumannn on symmetric extensions of symme-
tric operators. In accordance with J.von Neumann,every symmetric
 21

extension K of a symmetric operator K is characterized by a par-

tially isometric operator V acting from F @ ~b_i onto G ~ ~ i "
The domain dom(K) is given by

A
dom(K) = dom(K) $ (V + I ) F , (4.1)

i.e. every element h adom(K) has the unique representation

h = f + gl + Vg1' (4.2)

f ~ dom(K), gl & F ~ 3b_i. We introduce the projection operator

Q= ~ ~. (4.3)

on the Hilbert s p a c e ~ , i.e. Q~ = Q = Q2.

Lemma 4.1. Let K be a symmetric extension of K domain of which is
given by (4.1). The symmetric extension K satisfies the conditions

M(<h)dom(~) ~ dom(K) (4.4)

and

~((W )f - ~(~ ) ~ = -ix( J~)f, (4.5)

f ~ dom(K), for e v e r y ~ e A C ( ~ I) if and only if the associated re-

presentation of K obeys

A ( ~ ) Q = QA(~)Q (4.6)

for every ~ AC(~I).

Condition (4.6) means that ~ = Q ~ is an invariant subspace
of A ( ~ ) , ~ ~ AC('~'~).
Now the notion of a neutral subspace of a Krein space will be
crucial in the following. An element f of the Krein s p a c e ~ is
called a neutral one if

If,f] = o. (4.?)

A subspace ~ o f ~ is called a neutral one if 36 consists of neu-

tral elements. It is not hard to show that we have

[f,gl = 0 (4.8)
 22

for f and g arbitrary elements of a neutral subspace.

A neutral subspace 3~ is said to be maximal neutral if ~ is
not contained in any other neutral subspaoe of ~ e x c e p t S . Every
maximal neutral subspace is either maximal non-negative or maximal
non-positive or both.
If the last case happens , it means, if the neutral sub-
space is maximal non-negatlve and non-positive, then the subspace
is called hypermaximal.
Now a simple calculation shows that J ~ = Q ~ is a neutral sub-
space. Hence we have established one direction of the following
Theorem 4.2. The symmetric operator K obeying (1.14) and (1.15)
A
has a symmetric extension K obeying (4.4) and (4.5) if and only if
the associated representation A ( ~ ) , ~ A C ( ~ 1 ) , o f K has a non-
zero neutral invariant subspace.
The converse is based on the following
Lemma 4.3. The orthogonal projection Q on ~ fulfils

Q~*]Q = QQ[*3= O (4.9)

if and only if there is a partial isemetry V: ~-i--* ~ i such that

1
Q = ~ \~ ~,)" (4.10)

The partial Isometry is uniquely determined.

Theorem 4.2 admits a further refinement.
Corllary 4.4. The symmetric operator K has a maximal symmetric ex-
tension K obeying (4.4) and (4.5) if and only if the associated re-
presentation A ( ~ ) , ~ ~ A C ( ~ I ) , of K has a maximal neutral inva-
rlant subspace. Furthermore, K has a selfadjoint extension obeying
(4.~) and (4.5) if and only if A ( ~ ) , ~ a A C ( ~ I ) , possesses a hyper-
maximal neutral invariant subspace.
We note that the results can be carried over the associated
group.
Corollary 4.5. The symmetric operator K has a symmetric extension
obeying (4.4) and (4.5) if and only if the associated group U(~ ),
~I, of K has a non-zero neutral invariant subspace. There is a
A
maximal symmetric K if U(~ ), ~ 1 has a maximal neutral inva-
riant subspace and there is a selfadjoint K if and only if U(~ )
has a hypermaximal neutral invarlant subspace.
 23

The advantage of Corollary 4.5 is that the investigations can

be reduced to a single operator, namely, the associated generator L.

~. Reduction
Let K be an extension of K such that the conditions (4.4) and (4.5)
are fulfilled. If K is not selfadjoint, then we can associate with
K a new associated representation A(q~), @ 6 A C ( ~ I ) . Naturally,
the question arises on the connection of the two associated repre-
sentations A ( ~ ) and A ( ~ ) of K and K, respectively. Let K ba cha-
racterized by the partial isometry V: V~-i---*~ i " With V we
connect the orthogonal projection P,

V~ V O
P = A~ (5.1)
~0 W ~ '
whose range coincides with F x G. A
Proposition 5.1. Suppose that the symmetric extension K K obeying
(4.4) and (4.5) is given by the partial isometry V: -~-i-~A ~ i " If
K is not selfadjoint, then the associated representation A ( ~ ) of
is defined on the Krein space ~ ,

A
= (z...~- P ) ~ , (5.2)

and can be obtained from the associated representation A ( ~ ) of K

by

A(~) -- ( I ~ - P ) A ( ~ ) ~ ( I ~ - P)~, (5.3)

~a AC(~I).

6o Concluding Remarks
The problem allows a further treatment assuming that the defect in-
dex of K is finite or one of the defect numbers dim(~b_i) and
d i m ( ~ i) is finite. In these cases methods of the Pontragin space
can be applied which allow far going conclusions. For instance
it is possible to show that the closed symmetric operator K is self-
adjoint if its defect index is finite (dim(~_i) ~ + ~, dim( ~ i ~
+oo) and the conditions (1.14) - (1.16), (1.18) are fulfilled.
Thus,if K obeys (1.14) - (1.16) and has a finite defect index, then
the condition (1.18) must be violated at some points. It can be
proved that these points coincide with the spectrum of the associated
 24

generator L. Furthermore, the problem can be deeper analyzed if the

dimension of the Hilbert space ~ is finite. But we must remark that
our efforts end in failure ass~m!ng d i m ( ~ ) = + ~ and an arbitrary
defect index. Moreover, we have found that the results fail in
this general case. For instance it is not true that the defect in-
dex of K arises only from a violation of (1.18). In the case dim(~) =
= +~o it can happen that all the conditions (1.14) - (1.16), (1.18)
are fulfilled but K is not selfadjoint. Unfortunately, the last case
is important in application to evolution equations.

References:
[I~ T.Kato:Integration of the equation of evolution in Banach
space, J. Math. Soc. Japan 5(1953), 208 - 234.

E2~ T.Kato: On linear differential equations in Banach spaces,

Comm. Pure Appl. Math. 9(1956) 479 - 486.

[3] T.Kato: Abstract evolution equation of parabolic type in Ba-

nach and Hilbert spaces, Nagoya Math. J. 19(1961),93 - 125.

[4~ T.Kato, H.Tanabe: On the abstract evolution equation, Osaka

Math. J. 14(1962), 107 - 133.

T.Kato: Linear evolution equations of "hyperbolic" type, J.

Fac. Science, University of Tokyo, Sect. IA, Mathematics,
17(1970), 241 - 258.

[6~ T.Kato: Linear evolution equations of "hyperbolic" type, II,

J. Math. Soc. Japan 25(1973), 648 - 666.

[7~ M.Reed, B.Simon:"Methods of Modern Mathematical Physics I:

Fourier Analysis, Selfadjointness", Academic Press, New-York-
San Francisco-London 1975.

[87 K.Yajima: Existence of solutions for Schr6dinger evolution

equations, Comm. Math. Physics 110(1987), 415 - 426.

C9~ H.Neidhardt: "Integration yon Evolutionsgleichungen mit Hilfe

yon Evolutionshalbgruppen", Dissertation, AdW der DDR,
Berlin 1979.
 25

~10~ H.Neidhardt: On abstract linear evolution equations, II,

Preprint, AdW der DDR, Institut fur Mathematik, P-~TH-07/81,
Berlin 1981.

[11~ O.A.Lady~enskaja: On the solution of operator equations of

different types, Doklady Akad. Nauk SSSR 102(1955), 207 - 210
(in Russian).

[12] O.A.Lady~enskaja: On the solution of non-stationary operator

equations, Mat. Sbornik 39(1956), 491 - 524(in Russian).

[13] J.L.Lions: Equations diff~rentielles ~ coefficients op~rateurs

non borngs, Bull. Soc. Math. France 86(1958), 321 - 330.

~I~ J.L.Lions: Sur certaines ~quations aux deriv~es partielles

coefficients o~erateurs non born~s, J. Anal. Math. 6(1958),
333 - 355.

E15] J.L.Lions: Equations diff~rentielles du premier ordre dans un

espace de Hilbert, C.R. Acad. Sci. Paris, s~rie A, 248(1959),
1099 - 1102.

6] G.DaPrato: Weak solutions for linear abstract differential

equations in Banach spaces, Advances in Mathematics 5(1970),
181 - 245.

~7] G.DaPrato: Sums of linear operators, in: "Linear Operators and

Approximation, II", Proceedings of the Conference held at the
Oberwolfach Mathematical Research Institute, Black Forest,
March 30 - April 6, 1974, pp. 461 - 472 (eds. P.L.Butzer,
B.Sz.-Nagy).

[18~ M.lannelli: On the Green function for abstract evolution

equations, Bolletino U.M.I. (4) 6(1972), 154 - 174.

~9S G.DaPrato, P.Grisvard: Sommes d'op~rateurs lin~aires et

~quatlons diff~rentielles op~rationelles, J. Math. Pures
Appliqu~es 54(1975), 305 - 387.

[2~ L.Paquet: Equations d'evolutlon pour op~rateurs locaux et

Equations aux deriv~es partlelles, C.R. Acad. Sci. Paris,
serie A, 284(1977).
 26

K2~ J.S.Howland: Stationary scattering theory for time-dependent

h~m~Itonians, Math. Ann. 207(1974), 315 - 335-

K22S H.Neidhardt: On abstract linear evolution equations, I, Math.

Nachr. 103(1981), 283 - 298.

[23~ H.Neidhardt: On abstract linear evolution equations, III,

Preprint, AdW der DDR, Institut fGr Mathematik, P-MATH-05/82,
Berlin 1982.

~24~ M.Reed, B.Simon:"Methods of Modern Mathematical Physics I:

Functional Analysis", Academic Press, New-York-San Francisco-
London 1974.

[25~ B.Fuglede: On the relation PQ - QP = -II, Math. Scand. 20(1967),

79 - 88.

~6~ N.S.Poulsen: On the canonical commutation relations, Math.

Scand. 32(1973), 112 - 122.

~7~ P.E.T.J~rgensen: Selfadjoint operator extensions satisfying

the Veyl commutation relations, Bull. (New Series) Amer. Math.
Soc. I(1979), I, 266 - 269.

[28~ P.E.T.J~rgensen, P.S.Muhly: Selfadjoint extensions satisfying

the Weyl operator commutation relations, J. d'Analyse Math.
37(1980), 46 - 99.

~ W.J.Phillips: On the relation PQ - QP = -il, Pac. J. Math.

95(1981), 435 - 441.

[3~ K.SchmUdgen: On the Heisenberg commutation relation I, J. Funct.

Analysis 50(1983), 8 - 49.

[3~ K.SchmUdgen: On the Heisenberg commutation relation II, Publ.

RIMS, Kyoto Univ. 19(1983), 601 - 671.

~3~ G.Dorfmeister, J.Dorfmeister: Classification of certain pairs

of operators (P,Q) satisfying ~P,Q~ = -lid, J. Fumct. Analysis
57(1984), 301 - 328.
 2y

[3 0 H.Neidhardt: Symmetric extensions preserving additional con-

ditions, Preprint, AdW der DDR, Institut f~r ~athematik,
P-MATH-16/82, Berlin 1982.

~34~ N.I.Achieser, l,M.Glasmann: "Theorie der linearen 0peratoren

im Hilbert-Raum", Akademie-Verlag, Berlin 1975.

[35] J.Bognar: "Indefinite inner product spaces", Ergebmisse der

Mathematik und ihrer Grenzgebiete, Bd. 78, Springer Verlag,
Berlin-Heidelberg-New York 1974.

[36J T.Ja.Azizov, I.S.Jodvidov: "Basics facts about linear operators

in spaces with indefinite metric", Izd. "Nauka", Moskva 1986
(in Russian).

137] M.A.Naimark, R.S.Ismagilov: Representation of groups and alge-

bras with indefinite metric, Itogi nauki, mat. anal. 1968,
AN SSSR, Institute Sci. Information, Moskva 1969.
 ENERGY-DEPENDENT INTERACTIONS AND THE EXTENSION THEORY

K.A.Makarov
Physical Institute, Leningrad State University,
1 Maya 100, Petrodvoretz Leningrad 198904

The aim cf this lecture is to present a short

summary cf some aspects a mathematical const-
ruction developed at Leningrad University for
twc-body exactly solvable models with point
interactions [1-3, 5-9].

I. M.G.Krein's formula fcr the ~enerallzed resolvent and boundary-

form representation

We start in this paper with a reformulation of von Neumann's

theory of self-adjoint extensions in terms of boundary forms which
based on the following theorem L2J:

Theorem 1 Let ~ i be a deficiency subspace ccrresponding to}%=i

of a densely defined symmetric cperatcr A O in a Hilbert
space H with equal deficiency indices. Let A be a self-adjcint ex-
tension of A O. Then the adjoint operator A~ can be described as

Dom(A~) = [ u: u = (A-iI)'1%~ + A(A-il) - I ~ + +

(I)
+ (A-iI)-I ~-' ~-~g~i' ~e ~i "

Hence to each ueDcm(A O) , vectors ~-~(u) are associated.

The boundary fcrm of the adJoint operator A~ can be written in

terms cf these "bcundary values" as

<A u. v>- <u AoV>--

(e)
- <~+(u), ~-(v)> - j(~ (u), ~ (v)).
 29

The self-adjoint extemsions of A 0 are in one-to-one correspondance

with the bundle of Lagrange planes of the boundary form, i.e., the
planes in ~ i + ~ - i where the boundary form (2) vanishes. All
such Lagrange planes can be reduced to the surface in ~ i + ~ - i '
defined as follows

L~= [u 6 Dom(Ao): ~-(u) = ~ E ( u ) I (3)

where [" is a self-adjolnt operator in ~ i" Thus

J(/,(u),6(v)) = o

for any u, v e L r . Finally the self-adjcint extension A r c f the

operator A 0 is determined as

A~ = A 0'~~ (4)

Let A 0 be a restricticn of a self-adjoint operator A in H. Then

for operators A and A r the following theorem (see [2]) hclds

Theorem 2 The resolvents of self-adjoint extensions A, Ap of the

symmetric operator A 0 are related by

(A z - AI) -I = (A- ~i)-I + A--~

r~|A+iI~'~-P ~ I +AA p)-Ip
(5)

where P denotes the projection onto ~ i in H.

The relation (5) is known as M.G.Krein's formula [4]. It makes

many interesting model problems in quantum mechanics solvable
[2,3,6-9].

Now we are going to apply the above results to the two-channel

situaticn: H = H e x ~ H in, A = A ex ~ A in, ~ i = ~~ i ex~.
~ glw7
* iin "
Let us denote

Dex'in(~) = P ~I+ ~A P IHex,in (6)

Then we can define the operator D ( A ) in H as

 30

I Dex( % ) O )
D(A ) = (7)
0 Din(A)

One can notice that D(}k ) is an operator-valued analytical func-

tion with positive imaginary part in the upper halfplane ( the
so-called R-function ). It is the main functional parameter of
the method discussed here.

Lemma I Theabcundary values" ~ ( u ) cf the solution to the ad-

joint equation (A~- A l)u = f are related by

~-(u) ~-il f + D( x ) ~ +(u)

: P A-Z-TT (8)

where ~-+(u) ~xcu) e ~nCU).

=

In the two-channel case, H = H e X ~ H in, the parametrizing opera-

tcr P has a block structure:

Vee C ei )
V : ' Fei = C +i e ' ~ee : ~ ee'
+ Cil = C +ii"
Vie Vii
The resclvent R C ( X ) = (At- AI) -1 has a block structure toc~
its components can be reconstructed from the "pure external" block
R~e(A) alone [8]. Lemma q allows to obtain for R V ( ~ ) the fol-
lowing formula
AeX+il ex Qex(X) A ex- il ex (9)
R%e(A) = (Aex- Alex) -I + AeX_Alex Aex_ Alex
where Iex is the identity operator in H ex and Qex(A) is the ope-
rator-valued R-function in ~ i :

Qex(A) = {(Pee-Dex(A)) - Pei (Vii-Din(A))-q~ie} -I. (qO)

The formulae fcr resolvents cf self-adjolnt extensions are va-

lid also for non-densely defined symmetric operator A 0 with equal
deficiency indices, in particular for finite-dimensional internal
cperatcrs, dim H in <c~ ( see the contribution of B.S.Pavlov to
this volume ).
 31

2. Zero-range pctential models and some generalizations

The simplest example of exactly solvable quantum model based cn

Krein's formula is represented by point interaction cf particles
with an internal structure [2,3,6-9].

Let A in be some self-adjolnt operator in Hilbert space H In and

let ~ be a generating element of Ain:

Hin= Vexp(iAint) (11)

The restriction of A in to the domain

Dcm(A~) = (~in _ iI)-I { e} ~

will be dencte~ as A 0in . The only deficiency vector of the operator

A im corresponding tc the point A = i coincides with ~ . If the do-
main Dcm(A in) is dense in H In, i.e., if ~ ~ Don(A), then the domain
Dom(A in~ ) of the adjcint operator A ~ ~ is given by the formula
(1),where ~Z i = {~}, ~ + = ~-+~ and ~-+~c. The boun~ farm
(2) now looks like

(12)

Let A ex be Laplace cperatcr in ~3 and A~ x = (-A) O is its rest-

fiction to the set of all smooth functions u, u aL2(R3), vanishing
in the nelghbourhcod of the origin. Then the domain of the adJoint
operator is [5]

Dom(A = u(x) : u O u I ÷ o(1)I (13)

and the external boundary form can be written in the terms of the
asymptotic boundary values Uo, Ul:

Jex(u, v) = - U o ~ + Ul~-o . (14)

The set of the self-adjoint extensions cf the operator (-~O)

AOin is parametrized by Lagrange planes of the combined boundary
form J = jex+ jin. The simplest realization of such a Lagrange pla-
 32

ne is given by the boundary condition

(;o): (15)
with some Hermitian 2x2-matrix ~ .

Let us calculate the external components of scattered waves

which are the eigenfuncticns cf the absolutely-continuous spect-
rum cf the self-adjcint operator correspcnding to the boundary
condition (15). These components lock as follows
ei k l x l k V~> 0 (16)
Ue = e-lk <x,~> + f(k) # ~ l'x'(' ' =

The boundary condition (15) and Lemm~ 1 lead tc the relation

~'00
,t-'=
rOlcll) ' Vjje~. (17)
to1
Hence we get the following expressions for scattering amplitude
f(k) and s-wave scattering matrix sO(k):

f(k) : ~( Uoo- 4--~-)-

ik Iro112(["11-Din(~))-1J
-1 ' (~8)
sO(k ) = ( Poo+i~/'~" ) - lCOl 12( P11_Din(~))-I • (19)
(roo-i~'~[) - IVo~ 2 (ClI-D~(~b) -1

The scattering matrix obtained above is ncn-trivial in the s-cha-

nnel only. Our approach fails fcr higher momenta 1 = 1,2,..., be-
cause the correspcndlng multipcls are not square integrable. There
exist two ways of constructing cf solvable models of an atom with
higher momenta. The first one is based on use of an indefinite
metric [10] • The other can be d c n e ~ i n L2(R 3) but one has to sac-
rify locality cf the interaction [5,8,9J.

Let ( - ~ ) 0 be the restriction cf ( - ~ ) to the linear set of all

smooth functions vanishing on the sphere Z R = ~ x: Ix~ = R I. The
domain of the adjoint cperatcr ( - A ) O consists cf functions u(x):

u(x) = Uo(X , llx- J -1 °xp(i 'x-sl u (s) ds (2o

 $3

where Uo~DOm(-~ )0 and ~ u ~ W23/2(ZR) • The boundary form is equal

to

J(u, V) = i~(- u ) ??nv ~u (-V) Ids =

z~ (21)

Let A in be a Hermitian operator in an internal Hilbert space

H in. We denote by AOin some symmetric restriction of A in with the
defiolency su space n) = corresponding to the point
=i, d i m ~ i ~ = n. Then the sought self-adjoint extension of the
operator (-~ )0 @ A ~ can be constructed as Priedrlchs extension
of the below bcunded operator Ar. The operator A F is defined by
the boundary conditions

(i (22)

Here ~;~&~lin, rik = ~ ki" We obtain the simplest picture of

p-scatterlng by taking
= ~11 = 0, YlmiS one of the 1=1 spherical functions on ~I and
~ Cn. This leads to the boundary condition

[~n u]l~R = - ( ~, Din()~) ~ ) <u I~R' Ylm)Y1m (23)

where ( , -~ means the scalar product in c n ~ ~ and <-

means the scalar product in L2(~R).

The boundary ccnditien (23) can be reformulated in terms of the

generalized energy-dependent potential V(~ ) [87:

to ~ L2( ~3 ).
Here ~ means ~-function with support at the sphere Z 2" The
R
form of the potential V(}~) allows us to obtain an explicit solu-
tion for the Lippmann-Schwinger equation for the resolvent
 34

G(A) = GO(,~) - GO(,X) v(~) G(,x) (25)

This solution locks like

G(X) = QO(,~) + <Go(>,)- , Y l m > G o ( A ) Ylm x

(26)
,
- <Go(A)Ylm, Ylm> 1 -1

We denote here GO(A) the free resclvent of the Laplace operator in

R3:

GO(A) = ( - 4 - ~)-I .

Now we can obtain the expression for the external component of the
scattered wave as the coefficient cf the spherical wave in the
asymptctics of the Green function G(x,x I, A) when [x11-~ ~ [11].
The free Green function locks like

a0(x,~', ~ 1 = I ~21/4 ! x-~'l 1/2

amd has ~he asymptotics
A eiV~l xq
Go(X'X"A) "" e-i<x'x'>V~ q-rlx-x"l "
Ix'i~ .o
This gives the asymptotical expression for the G(x,x I, Ik), defi-
ned by (261:
iV~lx, I .o(~,k-~) , k= I~Z = ~
G(x,x l, A ) "" e
^ k-.% m~
IX'l->- Zl. -~ Ix-x'l (87)
k=2=_ x •

where uO(x,k-~) is the external component of the scattered wave with

a momentum ? :
m ,%
u0(x,~l = e i < x , k > + ~eiV~R 4s,k>Ylm(~) d~ ,
z, (281
2
, G0(x,~',,k )~l,,(s) a~"
(;,Din(k2) ~- ) -1 - S i a~ d;' Go(R~,R~,k2)Ylm(~)~
z~ z~
The asymptotics cf uO(x,k-J when Ix I--> co,
+,% - ~
uO(x,k-') ~ ei(x'k> + f(x,k) • ik ~ I / ( ~ Z l x l )
 35

leads to the following scattering amplitude f(x,k):

f(x,k) = ~ e i k R < s ' k > Y l m ( S ) d~ j e i k ~ ( s ' X > Y 1 m ( ~ ) d~ ~

zr~ Z~ ,, - t (29)
'^
^ l J
9S) k2)~
z~z~
Finally) for t h e S - m a t r l x we g e t :

S(x,k) = I + ~ f ( x , k ) (30)

where f(x,k) is given by (29).

Thus we have demonstrated that one can use no indefinite metric

even in the case 1 ~I, and obtained S-matrlx (30) has rather com-
plicate analytical structure. All the results were practically
obtained on the algebraic level of computations.

3. Boundary conditions on thin manifolds

The nex~ natural question is: how one can generalize the scheme
of the boundary conditions model to the case of higher codlmensicn
("thin") manifolds ? The answer can be obtained using the embedding
theorems which allow to state boundary conditions on manifolds of
lesser dimensions.

Let us consider the Laplacian - in R n and let L m be a hyper-

surface in R n, dim L m = m. Restrict - ~ to the symmetric operator
(-A) 0 specified by its domain
C~ n
Dom(-A )0 = Co(R \ Lm)

Notice that ( - ~ ) 0 is e.s.a, if c o d i m L m in R n ~4. For c o d i m L m < 4 ,

the deficiency subspace corresponding to ~=i of ( - ~ ) 0 consists of
the elements

~p(X) = ~ (~n)P Q(~-s,i)~p(S) ds (32)

with such distribution ~ p ( S ) for wich 0 ~ (x) E L2(Rn) • Here

P
 $6

G(x,s,i) means the Green function of (-~), ~ n is the normal de-

rivative~ onl Lm" In order to find the class Of~p with 0 # p ~ L2(R n)
or ~;p ~ W2(Rn),we use the momentum representation:
J P
jOp(S) ~ .~p(k. ) .

Simple calculations show that

(33)

Here k IIdenotes ~he momentum tangential to Lm and k~i k~. The

following assertion is valid

Lemma 2 Suppose that

+ 2 - i

Then 0 p # L2(R6) if ccdim Lm< g-2p and 9 p satisfies the fcl-

Iowlng conditions :

codim Lm 0 belongs to #1 belongs to

the cl@ss the cl@ss
w 3/2 2
2 ~21 not exist

3 ~21/2 not exist

Here ~ are Sobolev classes of distributions. For p ~2 the corres-

ponding Scbclev class does not exis$.

If one is interested in the situation ~ ~ wlo(R6), it is neces-

sary to use the p=O distribution only and ~o£W21/2(R ).
 37

4. Three- body quantum problem with point interactions

It is interesting to apply the above technique to the three-body

problem with point interacticns. After the separation of the cen-
tre of masses the conflgurational space of the three-dimensional
three-body system im L2(R6). There are three 5-dimensicnal planes
in R 6 corresponding to the pair point interactions. We denote L
the union of them. Hence~, H = L2(R6), dim L = codim L = 3,
= -~, Dom(~) = W~(R6).

Let us restrict the Hamiltonian -~ to the symmetric cperatcr

~ O = (-~)O with the dcmain
6
Dom(- A)O = CO(R \T,) .
Acccrding to Lemma 2 the deficiency subspace of ~ 0 corresponding
to A =i coincides with the linear set of functions

L
The Friedrichs extension cf ~ 0 leads to the original operator ~.
That is why we can use procedure described above. Namely, let
us consider the linear set

~= u : u + -~u-X-~ + (%- i)-I~ - ;

~+= (~.,.i)-ljo ~ , (36)

~aDom(-A) O

and restrict the operator (-•)0 tc %~. Then the fcllcwing lemma
is valid.

Lemma ~ Let u, v ~ . Then

< ~ o u, v > - <u, ~ o V > = < ~-(u), ~+(v)>T2(R6 ) -

<~+(u), ~-(V)>L2(R6 ) = <Ira G igO(u), ? ~ ( v ) >L2(L) -

-< f~(u), Im Gi#o(V)>L2(L) . (37)

 38

So we have the natural variables (ccnjugated in the sense cf Sc-

bclev classes) cf the symplectic form:

u-_~ Im Gigo(U) ,
9;(u) •

Hew ene can calculate ~ ~, u- fcr a given u ~ ? The answer is gi-

ven by

Lemma 4 The integral ~ ~ ( u ) dy fcr any .fA is given by

S~(u) dy = lim ~ n u ds
~l
n ~ : {X~R6: dist(X,n) : #}, (38)
and

u- = Im Gi~ O = lim (u(X) - S Re G i ~ o÷( U ) dy)

On the base cf Lemma $ we can write the bcundary ccnditicn

(39)
(;I
for the thin manifold L in R 6. Such ccnstructicns have been used
in~5] for ccnstructing the belcw bounded Hamiltcnian in three-body
system with pcint interacticn2 and internal structure. As it was
shown in [5] cne can not construct such a self-adjcint cperatcr by
the extension theory methcds withcut using an additional Hilbert
space.

Thus in this sltuaticn rather easy methcds based on yon Neumann

theory allow tc sclve rather hard and important prcblem related to
the many-body systems with singular interactions.
 39

Acknowledgements

I want to express my gratitude to B.S.Pavlov, Yu.A.Kuperin and

Yu.B.Melnlkov for the assistance in this work.

References

1. V.M.Adamjan, B.S.Pavlcv, Zap.Nauch.Sem.LOMI ~ (1986), 7.

2. B.S.Pavlov, Uspekchi Matem.Nauk @2 (1987), 99.
3. B.S.Pavlov, Tecr.Mat.Fiz. 59 (19~), 345.
4. M.G.Kreln, Dckl.Akad.Nauk SSSI~, 52 (1946), 657.
5. B.B.Pavlev, Matem.Sbcrnik, 1988 (~o appear).
6. Yu.A.Kuperin, K.A.Makarov, B.S.Pavlov, Tecr.Mat.Fiz. 69 (1986),
100.
7. Yu.A.Kuperin, K.A.Makarcv, Yu.B.Melnikcv, Tecr.Mat.Fiz. 7~
(1988), 103.
8. Yu.A.Kuperin, K.A.Makarcv, S.P.Merkurlev etal., ITP-Budapest
Report N441, Budapest, 1986.
9. Yu.A.Kuperin, K.A.Makarcv, Yu.B.Melnlkcv: in "Theory of Quan-
tum Systems with Strong Interactioms", Kalinia, 1987, p.63.
10. Yu.G.Shcndlu, Teor.Mat.Fiz. 64 (1985), 432.
11. S.P.Merkurlev, L.D.Faddeev: ~ a n t u m Scattering Theory in Few-
Body Systems, Moscow, Nauka, 1985 (in Russian).
 ON PERTURHATION~ FOR SELF, ADJOINT GENERATORS OF FELLER PROCESSES

M. Demuth
Institute of Msthemetics, Mohrenstr. 39, 1086 Berlin-Mitte, G.D.R.

i. Introduction

The aim of this article is to explain some spectral theoretical

consequences for generators of FeIier processes if they are per-
turbed by reguiar and singular potentiais. The content is based
on both the book by van Casteren [i] on generators of strongly
continuous semigroups and on the report by Demuth [2] on scat-
tering theory for generators of Markov processes. In both mono-
graphs Feiier semigroups are introduced and investigated. In [ i ]
the Feynman-l<ac formaiism for reguiar potentials is studied exten-
siveiy, In [2] strongiy continuous semigroups with singuiar poten-
tials over unbounded regions are included. In the present article
some further connections between these aspects in the Feller
semigroup theory are sketched.
In section[2] the Feller semigroups are introduced in L ~ ( ~ n ) .
This definition is extended to L 2 ( ~ n) such that the Feiier genera-
tors become seifadjoint. Reguiar and singuiar perturbations are
considered in sections 3 and 4, respectively.The l<ato-ciass
potentiais are described. The singularity regions are ciassified.
In section 5 sufficient conditions are given, such that certain
differences of Feller semigroups are Hilbert-Schmidt or trace
class operators. The conditions are satisfied even in the case of
star-like singuiarity regions.
For the sake of shortness the theory is expiained roughly,
proofs are omitted, if they are known from the literature, and
sketched for the new resuits. Moreover the content of the article is
compressed. The interesting reader is referred to a more detailed
version of this articie which wiii be pubiished in the Hathema-
tische Nachrichten.
 41

The r e s u l t s were p r e s e n t e d at the workshop on " A p p l i c a t i o n s of

the S e l f a d j o i n t - E x t e n s i o n s T h e o r y i n Quantum P h y s i c s " , Dubna, 1987.
The a u t h o r thanks D r . P. Exner f o r the k i n d i n v i t a t i o n and h o s p i -
tality d u r i n g t h i s workshop.

2. U n p e r t u r b e d R e n e r a t o r s of F e l l e r processes

The main c o n n e c t i o n between the s t o c h a s t i c and s p e c t r a l a n a l y s i s

i s g i v e n by the t h e o r y o f s t r o n g l y c o n t i n u o u s s e m i g r o u p s , T h e r e f o r e ,
at f i r s t , i t w i l l be i n t r o d u c e d the semigroup of the F e l l e r p r o c e s s ,
i . e. the F e l l e r semigroup, c o n s i d e r e d i n t h i s a r t i c l e .
Denotation 1: Probability space ( see [ 3 ] )
Let ( ' ( i x ' ~ x ' PF) be a p r o b a b i l i t y space, where ,r~ x i s the set
of all continuous functions ~(.) , mapping [ 0 , ~ ) i n t o ~n, s t a r t i n g
at x . PF i s a p r o b a b i l i t y measure on ~ x .
Assumption A: Let ( ' ~ x ' 3 x " PF ' ~ a ( ' ) ) be e F e l l e r p r o c e s s homo-
geneous i n time w i t h the t r a n s i t i o n function

P(t,x,E) :: PF { w : ~(t),E , ~(0): x} ,

x ~ ~n , t G [ 0 , ~ ) , E a Bore1 set i n ~ n . Assume a t r a n s i t i o n

d e n s i t y f u n c t i o n p, mapping ( 0 , ~ ) ~ R n ~ ~n )~, such t h a t

P(tix,E) :: ~ p(t,x,y) dy
E
with

0 ~ p(t,x,y ) 4

for t ) O, x , y ~ ~ n

Definition 2: F e l l e r semigroup i n L ~
Let p be a d e n s i t y of a p r o c e s s s a t i s f y i n g Assumption A. Then the
F e l l e r semigroup i s d e f i n e d by

(Ttf)(x) :: j p(t,x0y) f(y) dy, t)O,

and

(mof)(x) :: f(x)
~or f ~ L~( ~ n ~) (essentially bounded f u n c t i o n s ) .

For s p e c t r a l t h e o r e t i c a l c o n s i d e r a t i o n s it is n e c e s s a r y to d e f i n e
the F e l l e r semigroups i n L 2 ( ~ n ) o r more g e n e r a l i n L q ( ~ n ) , l~q~m.
 42

Definition 3: Feller semigroup in L q

Let LTt, t ~ O~ be a Feller semigroup in C ® ( ~ n ) (continuous
functions vanishing at infinity). The semigroup is said to act in
Lq ( ~ n ) if it has the following properties:

a) There is a T > O , M~ I such that

Tt Co ( ~ n ) C Lq(~n)
( Co - space o f continuous functions with compact support) and

~Ttfll Lq ~ M llf llLq l

0 ~ t E T .
b) lira ~ T t f - f II Lq = O for f • Co( ~ n ) .

Remark 4: T h i s d e f i n i t i o n is taken from van C a s t e r e n ~1 7 p . 16.

Because C o ( ~ n ) i s dense i n L q ( ~ n ) t h e s e m i g r o u p can be e x t e n d e d
to L q. Here we d e n o t e t h i s e x t e n s i o n by t h e same symbol T t .
The o b j e c t i v e in this section is to present conditions such t h a t
the generators of the Feller semigroups considered are selfad3oint.
A first step in this direction i s to s t u d y s y m m e t r i c s e m i g r o u p s .
Then we r e s t r i c t us to s e l f a d j o i n t semigroups in L2(~n) with
selfad~oint generators.
Proposition 5: Let T t be a Feller semigroup which is symmetric,~i.e.

~(Ttf)(x) g(x) dx : ~ f(x) (Ttg)(x) dx

for f, g ~ Co(~n). Then

l l T t f UL q llflILq , t ~ 0,

for f G C o ( ~ n) and t h e s e m i g r o u p acts in Lq(~n), l~q~. For

q=2 i t extends to a s e l f a d j o i n t semigroup.

Proof see [ 1 ] p. 19.

Definition 6: Feller g e n e r a t o r
Let Ttbe a contractive Feller semigroup in L2(~n). Then i t s
generator K is given by
K f :-- lim t -I (1 - Tt) f

with
dora K := { f , f~ L2(~n), lira t - 1 (1 - T t ) f exists ~ .

By means o f K we r e w r i t e
Tt f =: e - t K f , t _t 0 •
 43

Proposition 7: The generator of a Feller sere.group in L2(~ n) is

selfad3oint if the semigroup is symmetric.
Proof see ~lJ p . 138 •

One can f i n d sufficient conditions for the transition density

function such t h a t the corresponding Feller semigroup acts in L 2 ( ~ n) .

Assumption B: L e t (JPLx , ~ x ' PF' ~2(") ) be a F e l l e r process

homogeneous i n t i m e p o s s e s s i n g a t r a n s i t i o n density function p°
In addition we assume f o r this denlity function

p(t,x,y) ~= a ( 2 r t ) " n / 2 e- b Ix-yl2/2t

for 0 ~ t ~ T, x ~ A n
-- 0
y e ~n w i t h some p o s i t i v e constants a, b •

Proposition 8 : The F e l l e r semigroup of a process satisfying

Assumption B is a contractive Co-semigrou p acting in L2(~n). If
p(t,x,y) = p(t,y,x) f o r 0 4 t -~ T this e x t e n d s to a s e l f a d j o i n t
sere,group w i t h a selfadjoint generator•

Proof: The c o n t r a c t i v e Co-property was shown i n [ 2 ~ p . 36ff. The

selfsdjointnese of t h e s e m i g r o u p and t h e g e n e r a t o r follows on
eccount of the P~opoeitions 5 and 7 o Q•e.d.

C o r o l l a r y 9: L e t A s s u m p t i o n B be s a t i s f i e d • Then i t holds for all

f e L2(~n), x~ ~n

(e - t K f)(x) = f f(y) p(t0x,y) dy

= f f(~(t)) eF(d~ ) .
J~x
Proof: The f i r s t equation follows by t h e D e f i n i t i o n 6 . The second
equation follows by t h e d e f i n i t i o n of t h e measure P F ( . ) . P r o p o -
sition 8 implies that these equations are true for all f g L2(~n).

Examples,:
1 . The main example i s the Wiener process. Its density function is
symmetric. Its generator is the Laplace operator in L2(~n) o
2o A n o t h e r e x a m p l e s a r e t i m e - homogeneous d i f f u s i o n processes with
a strongly elliptic, bounded, H61der-continuous diffusion matrix
and a bounded, H~lder-continuous drift vector.
Further examples are given in [2] .
 44

3. Regular perturbations of Feller generators

For introducing t h e Feynman-Kac f o r m a l i s m there are two p o s s i b i l -
O

ities. One can s t a r t with the unperturbed semigroup e- t l < , uses t h e

Trotter-product f o r m u l a , and c o n s t r u c t s the perturbed semigroup
e- t ( K + v ) where V is t h e m u l t i p l i c a t i o n operator with a potential
function V ( . ) : ~ n ~ ~ ° (see e . g . Theorem 3 , 8 . in [2] ).
The o t h e r possibility is to define a function

(Ptf)Cx) = J e x p [ - f V C ~ ( s ) ) ds] f ( ~ ( t ) ) PxCd~) (I)

.O.w o

x e Rn, and study then the properties of Pt f in dependence of V (see

e,g. [ I ] p. 2 O f f . ) . Here wedeserlbe the second way and define the
most i m p o r t a n t class of potentials ( see Theorem 11) :

Definition 10: Kato-class potentials

Fix a Feller-semigroup Tt, t -~ 0 ° A m o r e l m e a s u r a b l e function V
mapping ~n ~ is said to be i n K a t o ' s class with respect to Tt
if it satisfies the following conditions:

1) S e t t i n g V+=max(V,O), it holds
t
J(TsV+) (x) ds ( ~ (2)
0
for x 6 ~ n t :" O.
2) It holds

lim sup . J (TsV_)(x) ds = O (3)

with V_ = max ( - V , O) .

Theorem 11: L e t V be a K a t o - c l a s s p o t e n t i a l with respect to the

F e l l e r s e m i g r o u p e- t K c o n s i d e r e d i n L 2 ( ~ n ) .
Then t h e r e is a closed linear operator H w h i c h e x t e n d s K+V and
which generates a strongly continuous positivity preserving semi-
g r o u p e- t H . If e-tK is symmetric, then H is selfadjoint and g e n e -
rates a selfadjoint semigroup, represented in equation (1).

Proof see [1] p. 22ff, 122, 171ff.

In the following we i d e n t i f y H with I<+V and e - t H w i t h e- t ( K + V ) .

A g a i n we m e n t i o n the following examples. I f e" t K i s the semigroup
of the Wiener process then H is a selfadjoint Schr6dinger operator.
Note g e n e r a l , if e- t K is the semigroup of a canonical diffusion
 45

p r o c e s s such t h a t K i s an e x t e n s i o n o f a second o r d e r d i f f e r e n t i a l
operator with variable coefficients, then H e x t e n d s the p e r t u r b e d
o p e r a t o r K+V, ( f o r explicit conditions for the c o e f f i c i e n t s see
[2] p. 76 f f ) .
C r u c i a l f o r Theorem 11 are the K a t o - c l a s s p o t e n t i a l s . For
Feller processes s a t i s f y i n g Assumption B the K a t o - c l a s s can be
characterized in more detail.

Corollary 12: Assume a Feller process satisfying Assumption B. Take

0 ( t ~ T. Then a potential V mapping ~n~ ~ , n ~ 3 , is a Kato-class
potential with respect to the corresponding Feller semigroup if

V+(x) I x - y ( "n*2 dx ( ~ (4)

and

lim sup f V_(x) I x - y l -n+2 dx = 0 . (5)

P r o o f : That f o l l o w s i m m e d i a t e l y from the e s t i m a t e i n Assumption B,

An e q u i v a l e n t condition for potentials satisfying (5) is the

following.

Proposition 13: A potential satisfies the c o n d i t i o n i n (5) i f and

only if

Zim sup j ds s-n*z / V (x+y) dy = 0 . (6)

Proof; Integration by p a r t s y i e l d s

S V_(y) I x - y l "n+2 dy
Ix-yl~
-n+2
= r J V_(x+u) du
1~1 ~
(7)
+ (n-2) f ds s- n + l f v_(x+u) du.
0

From (5) follows

sup f V (X+U) du : o ( r n-2)

x6R" Iwli~ -

as r~0. The same f o l l o w s from the c o n d i t i o n in (6). Thus i n both

cases the f i r s t term i n (7) v a n i s h e s as r ~ 0 . Q.e.d.
 46

Remark 14: The c o n d i t i o n

, ,
in (6) is satisfied for V • Lq ( ~ n ) #

n/2• q ~ . That implies for instance V_ 6 L 2 ( ~ 3 ; + L~(~3) such

that Coulomb p o t e n t i a l s are included.A necessary condition for (5)
is

fV_(x) dx = o(r n'2}

I)ll 1t f
as r - ~ O .

Introducing the conditional measure of the processes considered

the Feller semigroups turned out to be integral operators.

0enotatlon 15: Let (_~Lx, ~x' PF' ~(')) be the Feller process given
in Assumption A. Then we set

.-'- -- .

and we d e n o t e w i t h py,t(.) the conditional measure c o r r e s p o n d i n g to

this p r o c e s s ,

Theorem 16: Let e- t K be a F e l l e r semigroup of a process satisfying

Assumption A (see the Definitions 3 and 6 } . Its density function is
assumed to be s y m m e t r i c , L e t V be a B o r e l measurable function from
n to ~ which belongs to K a t o ' s class.
Then t h e s e m i g r o u p e" t ( K + V ) consists of integral operators the
kerne]~of which are given by

(e-t(K÷v)) (x,y) :: /y,( exp L- fV(~(s) ds] py't(d.) . (8)

o
The k e r n e l s are continuous on (O,ee) ~r R n x R n and s y m m e t r i c .

Proof: The p r o o f is given in [1] p. 35ff using essentially

Khas'minskii's Lemma. A n o t h e r proof is given in [2] p. 35ff for
potentials w h i c h can be a p p r o x i m a t e d by bounded o n e s .

Corollary 17: I f V is a bounded measurable function from An-~

i t holds

(e-t(K+V))(x,y) = etess ~u~Jv_(x)l p(t,x,y)

where p is the density function of the Feller process considered.

Proof: This follows immediately from the representation in (8),

 47

Besides this trivial estimation of the kernel a more general one

i$ given recently by van Casteren in [4] .

Proposition 18: Let K be the Laplace operator and assume a Kato-

class potential V, i.e. V shall satisfy the conditions in (4) and
(5). Then there are positive constants ~ and ~ such that

(e-t(K+V))(x,y) ~ ~ (2~t) -n/2 e ~t e-lX-yl2/2t

for all t ~ 0 .

4 . Singular perturbations of Feller 9 e n e r a t o r s

In this section we n e g l e c t the regular perturbations for t h e sake

of simplicity, i.e. we assume V ( x ) m 0 0 b u t we c o n s i d e r singular
perturbations of Feller generators which arise by i n c r e a s i n g poten-
tial barriers up t o infinity. This region over which the potential
barriers are defined is called singularity region, Describing
that we have t o introduce some f u r t h e r denotations.

Definition 19: Singularity region

Let G be a closed region in ~n with a piecewise ~i boundary. Let
~ G ( . ) be the corresponding indicator function and P be a projection
operator in L 2 ( ~ n ) given by

(Pf)(x) := ~G(X) f( x ) , f • L2(~n)

Horeover, we denote with ~ the complementary projection operator

with respect to P, i.e.

:= ! - P •

With 0~ we d e n o t e the canonical embedding o p e r a t o r from L 2 ( ~ G )

to L Z ( ~ n ) , i.e.

D := 1L2 ( ~ n \ G) -> L 2 ( ~ n ) .

L e t 0 be t h e a d j o i n t of it such t h a t

D~ D = 1 L 2 ( ~ n ) -P =

D O~ = 1L2(~n\ G) •
 48

As m e n t i o n e d a b o v e , s i n g u l a r p e r t u r b a t i o n s of Feller generators
a r e i n t r o d u c e d by i n c r e a s i n g p o t e n t i a l s over the s i n g u l a r i t y region
G .

Theorem 20: Assume a Feller process which satisfies Assumption B.

Let e -tK be the corresponding semigroup with the Feller generator
K considered in L 2 ( ~ n ) . Moreover, assume a positive parameter N
and define

VM := MP

which is given by t h e m u l t i p l i c a t i o n operator

vM(.) = M ~G(') "

Set I<M := K+V M . Then dom KM = dom K and I<M g e n e r a t e s a contractive

C o - s e m i g r o u p , 6"tKN , i n L 2 ( ~ n ) .
Then t h i s semigroup has a s t r o n g limit as N ~ , which is denoted
with Ut , i.e.

Zim 11 e -tKM f " U t f II = 0 , t ~ O, (9)

for all f~ L2(~n). The l i m i t Utcan be r e p r e s e n t e d pointwise by

(Utf)(x) =f~):~(s)~ G,Vs, s~ { O , t ] } f(~)(t)) PF(d~) (10)

for a.a. x$ ~n\ G and a l l f6 L2(~n) .

Moreover U t t L 2 ( ~ n \ G) i s a c o n t r a c t i v e CO s e m i g r o u p . D e n o t i n g
i t s g e n e r a t o r w i t h (K)G t h i s s e m i g r o u p c o n s i s t s o f i n t e g r a l operators,
t h e k e r n e l ~ o f w h i c h a r e g i v e n by

(e - t ( l < ) G ) (x,y)

= f Z~):~(s)#G, Vs, s~[O,t] ~ py't(dF ) " (11)

Proof: The p r o o f is given in [2] p. 4 3 f f and 5 1 f f .

I n the l a s t s e c t i o n t h e s e l f a d j o i n t n e s s o f t h e g e n e r a t o r s I<
and K+V was s t u d i e d . I n dependence o f K a l s o the s i n g u l a r l y per-
turbed generator (K)G can be d e s c r i b e d .
 49

Theorem 2 1 : L e t K be a s e l f a d j o i n t , nonnegative operator. Let

be g i v e n as i n D e f i n i t i o n 19. Suppose t h a t L 2 ( ~ n % G ) ~ dom K i s
dense i n L 2 ( ~ n \ G ) , and assume Kf = ~l<f f o r a l l f e L 2 ( ~ n \ G)
dom K. Set a g a i n KM = K + MP • M~O D and choose zL~res KH
(resolvent sets) .
Then t h e s t r o n g limit of (z - K M ) - I e x i s t s , i.e.

s-lim (z - KM) - 1 :: RG(Z ) ,

where RG(Z ) i s a pseudo-resolvent. Moreover, RG(Z)~ L 2 ( ~ n \ G) is

t h e r e s o l v e n t o f (K)G, and (K)G is the Friedrichs-extension of
I< #Cdom I < n L 2 ( ~ n \ G)~ .

Proof: The p r o o f is g i v e n by B a u m g ~ r t e l , Demuth i n [5~ .

Remark 22: Regular perturbations c o u l d be i n c l u d e d into the conside-

rations of this and t h e n e x t s e c t i o n by u s i n g the estimations of
t h e form g i v e n i n C o r o l l a r y 17 and P r o p o s i t i o n 18.

5. Spectral properties of Feller generators

Here t h e i n f l u e n c e of the singular perturbations for the spec-

trum o f F e l l e r generators is discussed by i n v e s t i g a t i n g the differ-
ences { s e e D e f i n i t i o n 19)

e- t ( K ) G 0 - 0 e- t K

of Feller s e m i g r o u p s mapping L 2 ( ~ n) -~ L 2 ( ~ n \ G) . The d e n o t a t i o n s

are taken from the sections above.

Theorem 2 3 : Assume t h a t e- t l < i s a Feller semigroup of a Feller

process which satisfies A s s u m p t i o n B. W i t h p we d e n o t e a g a i n the
density of the process.
Then t h e s e m i g r o u p d i f f e r e n c e

e - t ( K ) G ~ - ~ e- t K

is a Hilbert- SChmidt o p e r a t o r for any t ~ (O,TJ if

sup r p(s,x,u) du • L I ( IR n XG, d x ) .

./

se (#,~) G
(Compare t h i s w i t h the Kato-class conditions for the regular poten-
tial parts in Definition 10~ .
 50

Proof: e- t ( K ) G 0 - D e- t K is an i n t e g r a l operator. Because o f the

representations in Corollary 9 and Theorem 20 the f o l l o w i n g esti-
m a t i o n s are v a l i d :

f dx f d y I (e-tCK)G 0 - 0 e ' t K ) ( x , y ) 1 2

"-(~X fdy l ~/.,~(¢@:.Q~ ' S o , Soe ( O , t ) , a ) ( S o ) , G ~ PFY't(d(a)l 2

const .fdx f Z t . : E s o, ~(So) , G ~ PF(d~)

Now let

em--Lxo--°' xl ..... Xmi' ~m--tJ , z;~,~,

be a partition of [O,t] . The trajectories are continuous. The

measut, e of the trajectories which only touch the boundary of G is
zero (see Ginibre ~6 ] ) . Therefore

X[~:3So' Soe ( o , t ) ~(So) e G

But the l a s t expression is smaller than

sup f p{s,x,u) /~'G(U) du . Q.e.d.

se(~e) ~

Theorem 24: Let the a s s u m p t i o n s of the Theorem 23 be s a t i s f i e d .

Then the semigroup d i f f e r e n c e

e- t ( K ) G 0 - 0 e"tK , O< t = T ,

is a trace class operator , if

(1+ I x l ) g sup f p(s,x,u) du E L l ( ~ n \G, dx) (12)

s e (o,~)
with ~" > n.

Proof: Let Q be the multiplication operator, given by

 51

(Qf)(x) := (1 + Ixl) ~/2 f(x) ,

x 6 ~ n, ~ >n, f 6 L2( ~ n) . Then we r e w r i t e the semigroup difference

e
-2t(K)G 3 - D
e-2tK

= e- t ( K ) c 3 Q-1 g 0 ~ ( e- t ( K ) G 3 - D e- t K )

+ ( e-t(K)G 3 - 3 e - t l < )Q Q - l e ' t K .

Because o f A s s u m p t i o n B t h e o p e r a t o r s e- t ( K ) c D Q-1 and Q-1 e - t K

are Hilbert-Schmidt operators. Corresponding to t h e p r o o f o f Theorem
23 t h e o p e r a t o r s g 3 ~ ( e - t ( K ) G J - 3 e- t K ) and ( e - t ( l < ) G o - O e - t K ) Q
are also Hilbert-Schmidt operators, if the condition in (12) is
fulfilled. Q.e.d.

Consequences 2,5: I f one has s e l f a d j o i n t Feller generators K and a l s o

selfadJoint (K)G, then the Hilbert-Schmidt property i n Theorem 23
implies a stable essential spectrum, i.e.

%ss (K) = ~ess((K)e) "

That follows by a g e n e r a l two-space criterion g i v e n by B r C n i n g ,

Demuth, G e s z t e s y i n [ 7 J .
If the trace claas condition i n Theorem 24 i s satisfied one can
use t h e i n v a r i a n c e principle of the mathematical scattering theory
to o b t a i n for the absolutely continuous spectra

• ac(K) = %c((K)G) •

The a b s t r a c t condition i n Theorem 24 a d m i t s unbounded s i n g u -

larity regions. Here we a p p l y i t to s t a r - l i k e regions, A star is a
u n i o n o f p e a k s . Thus i t suffices to c o n s i d e r a single peak o f such
a star.

Definition 26: Peak o f a s t a r

For t h e sake o f s i m p l i c i t y we assume t h a t t h e peak D i s un-

bounded i n the direction of the xl-axis. Let x • ~n be decomposed
into (xl, Xn_l), x1 6 ~, ~n-1 ~ ~ n - 1 . Xn_l s h a l l be o r t h o g o n e l
 52

to the xl-axis. W i t h P x l we d e n o t e the plane through (xl, E)

orthogonal to the xl-axis. Let D(xl) be t h e o r t h o g o n a l projection

of Px? D onto PO " I D ( x l ) ! is the (n-1)-dimensional volume of

D(xI) •
Then D i s assumed t o satisfy the following conditions:
- D is c o n n e c t e d and c l o s e d .
- F o ~ a l l ~1 ~ ×] ~ ° and f o ~ a l Z x l ! x~O i t holds D(xl)~D(x~).
-. D~O) i s bounded i n ~n-1 .
- The b o u n d a r y o f D i s piecewise in ~1 The b o u n d a r y can be e s t i -
mated by f j , gj, j=2 ..... n , mapping ~ j - 1 into ~ such t h a t

D(xl) := {~n-1 = (x2 ..... Xn) :

f j ( x i . . . . . x j . l ) ~ x j ~ g j ( xI . . . . . x.3_1), 3=2 . . . . . n ~ .

Because D ( x l ) i s contained in D(O) fj and gj are bounded functions.

Therefore we can define

d := 2 sup ~up ( ] xj_ )l lgj(~j_1) J )

Theorem 27: Assume a F e l l e r process which satisf~eaAssumption B. Let

D be a peak of a s t a r according to the D e f i n i t i o n 26.
Then the se.migroup d i f f e r e n c e

e-t(K)D O - 0 e- t K , 0 ~t ~ T ,

is a trace class operator, if t h e peak D s a t i s f i e s

t f ix11 lD(x 1)1 dxl I '

with ~ )n.

Note t h a t unbounded D a r e a d m i t t e d . But t h e amount o f the

contraction of t h e peak must have a s u f f i c i e n t l y h i g h power i n
Ixll'~as (xll-~ ~.

Proof: The p r o o f ideas are sketched. One has to p r o v e that the

condition i n Theorem 24 i s satisfied. The main d i f f i c u l t i e s arises
from such t r a j e c t o r i e s starting near the boundary of D. T h e r e f o r e
 53

one i n t r o d u c e s a l a r g e r r e g i o n ~ , ~ D , the boundary of which has

e d e c r e a s i n g d i s t a n c e to ~D as I X l i - > ~ . Set

(xz) .= -/'`-- xj == gj + ( l + l X l l ) ' / ~ ;

j = 2,...,n

w i t h / x 7 ~ > n. The volume of ~ is f i n i t e . Hence i t suffices to

estimate

sup ; p(s,x,u) du

for x @ ~ n \ ~ . Because of Assumption B t h l s l s s m a l l e r than

c ..1" I x - u l - n e- c l x - u l 2 du

-~ ~n-1
( c i s some c o n s t a n t ) . For Xn_16 \ ~(xl) and Un_ 1 ~ O(Ul)
the d e f i n i t i o n of D" p r o v i d e s

I -1 - Un-ll = (1 , Ixll)-/'~

This estimation i s used to prove that the integral

f dx ( I + I x I ) E ; du I x - u l -n e- c l x - u l 2 du

is f i n i t e if

S IXll ~ ID(X 1)l dx 1 < Oo j ~ ~ 0 Q.e.d.

 54

References

1. J. van C a s t e r e n : Generators of strongly continuous semigroups,

Pitman; Boston, London, Melbourne, 1985.

2. M. Demuth, R e p o r t Inst. Hath. Berlin, 1985, R-Hath-04/85.

3. A. Friedman: Stochastic differential equations and a p p l i c a t i o n s ,

Vol.1. ; Academic P r e s s ; New Y o r k , San F r a n c i s c o , London,1975.

4. O. van C a s t e r e n , Integral kernels for SchrGdinger type

equations, Preprint Univ. Antwerpen, 1986, nr.86-32 .

5. Ha B a u m g & r t e l , H. Demuth, Rep. H a t h . Phys. 1979, v.15, p.173 .

G° O. G i n i b r e : Some a p p l i c a t i o n s of functional integration in

statistical m e c h a n i c s and quantum f i e l d theory. -
In: Statistical m e c h a n i c s and quantum f i e l d theory, Les Houches
1970, ed. by C. D e W i t t , R. S t o r a ; Gordon and B r e a c h ; New Y o r k ,
London, Paris; 1971, p. 327 .

7. E. B r i n i n g , M. Demuth, F. G e s z t e s y , Lett. Math. Phys. 1987,

v. 13, p. 69 •
 SINGULAR PERTURBATIONS DEFINED BY FORMS

V.D.Koshmanenko

Institute of Mathematics, Ukrainian Acad. of Sol.

252601, Kiev, USSR

We give a precise sense to the notion of singular perturbation.

It is 8 bilinear form h in a Hilbert space H with a regular

(closable) component br = o. Further we propose a classification

of singular bilinear forms with respect to a fixed selfadJoint ope-

rator A ~ o in H. Finally we present a construction of the sin-

gularly perturbed operator A b. Our definition of Ab is based on

the interpretation of b as a boundary condition for a fixed

selfadJoint extension of the symmetric operator A o = A ~Ker b.

I. Introduction
A singular perturbation of a selfad~oint operator A in Hil-
bert space H is a formal mathematical expression given on a li-

near set G~D(A) and which is equal to zero on a subset Go C G

supposed to be dense in H. As a rule such expressions can be gi-

ven a sense as bilinear forms in H. But these forms are not olo-

sable, and therefore they have not an operator representation in

H. In physics such expressions are used for heuristic description

of point, relativistic-point and other types interactions. All the-

se interactions are concentrated within a "small" space-time region

(with Lebesgue measure zero) which is the reason for the lack of

correct operator sense for these so-called singular interactions.

In spite of zero measure of the interaction regions many of them

represent nontrivisl perturbations.

At the present time, there is an increasing number of publics-

 56

tions dealing mainl2 with point interactions (see, for example, pa-
pers by S.Albeverio, R.Hoegh-Krohn and others [I-4]). The key pro-
blem is to give a reasonable interpretation to the Schrodinger ope-
rator with a point interaction. In general, there are several appro-
aches to this problem. But a complete and consistent theory of sin-
gular perturbations or point interaction alone has not been formu-
lated Up to now.

In the present paper, we are going to develop the following

point of view. A singular perturbation can be interpreted in terms
of a boundary condition corresponding to a fixed selfadJoint ex-
tension of some Hermitian operator which is common for both the
free and the singularly perturbed Hamiltonians. For a wide class
of singular perturbations, we formulate this interpretation in a
mathematically correct way.

The idea to make use of the selfadJoint-extenslon theory to

study the formal Schr~dluger operator A~ = -~ + ~ ( x ) belongs to
Berezin and Paddeev [5] • They proposed to interpret A~ as the
family of all selfadJolnt extensions of the Hermltian operator
A ° = - ~ ~Go, where G~D (-~) consists of the functions u(x)
such that u(o) = o. The question about cholse of a particular
selfadJolnt extenslou in general is open. We are going to show that
in the general case there is a one-to-one correspondence between
some class of slugular perturbations and the set of all self-
adJoint extensions of a symmetric operator which is analogous to
A o. To formulate the correspondence rule, we treat the singular
perturbations in a non-additive way. The information about a per-
turbatlou is contained in boundary conditions of some new (diffe-
rent from the free one) selfadJoint operator.
 5y

2. Definltion of a sinKu!ar perturbation

Let G be a linear dense subset in a complex Hilbert space H.

A map b: G x G - * C such that b(u,v) is linear in u gG and

conjugate linear in vg G is called a bilinear form on G with

the domain Q(b) = G. The inner product (',') in H restricted

to G gives an example of a billnear form on G. It will be deno-

ted in the following as a. A form b is possitive if bins

b(u,u) ~ o for all ugG.

Definition 2.1: A positive bilinear form b on G is singu-

lar in H (notation b I s) if for earh u E G, there exists a

sequence {Un)CG, n = 1,2, ... such that

Un H.. u, b [Un] , o, n ~ co (2.1)

We write bgF s if bla.

Let Hb denote the Hilbert space which is obtained by comple-

tition of G equipped with the (quasi)inner product (u,V)b=b(u,v).

In particular Ha= H. The fact that b is singular in H means

that Hb and Ha are orthogonal in following sense.

P,ropo,sltion 2.1 [6] :

b I a <---> Ha+ b = Ha ~ Hb (2.2)

Note that (2.2) implies bla~ ~a±b. It means that for

all v g G C - H b, there exists a sequence {v n ~c G such that

Hb H
vn ) v, vn ~ o, n ~ co . (2.1')

It is easy to see that u and v in (2.1) and (2.1') can be ar-

bitrary vectors from H and H b.

 58

The condition (2.1) is fulfilled automatically if the set

b[u] -- o }

is dense in H. In what follows we study the singular forms for

which GO is dense in H, i.e.

5o = H. (2.3)

Examp!e: Let H = L 2 (Rd), d = 1,2, ... , A = - ~ and

G = S(R d) be the Schwarts space. The biliuear form

b ° (u,v) = u(o) v(o), u, v 6 S ( R d)

is singular in L2 since G O = { u 6 S ( R d) I u(o) = o~ is

dense in L 2.

Next we introduce a class of singular bilinear forms of rank

one. Let H~H ~H. be a rigged Hilbert space; H -~ G. We as-

cribe to each vector w 6 H_ the biliuear form bw ou G defined

by the formula

bw(u,v) = <u,w> <.,v> ,

where ~','~ is the duality between H+ and H .

Proposition 2.2: The form bw is singular in H iff

w £ H \ H.
i

Proof. Let G O = Ker b w. Since b~ is a form of rank one, the

equality (2.3) is fulfilled only if w E H_ \ H.

Consider a negative Sobolev space H_ = W2 k (Rd), k = 1,2 . . . .

Then to each distributlou w 6 W~ k \ L 2 the above defined singular

form bw in L2 corresponds. In particular if k >d/2, then

6 W2 k and for w = ~ we have b w = b~.

 89

Let A be a positive selfadjoiut operator H, and consider

a set G such that

Gr-D(A), AGcG, A~G = A. (2.4)

Each biliuear form b ou G such that the = Ker b set G is

o
dense in H will be called the singular perturbation of A. Our
aim is to construct the corresponding perturbed operator which we
demote as A b. Before explaining our constructiem of Ab, we
give a classification of the singular perturbations introduced
above.

For this purpose we introduce the A-scale of Hilbert spaces:

• .. H _ k ~ H_I ~ H o ~ H O H I ~ . . . Hk~... G, (2.5)

where H÷k, k = o,1, ... is the completion of G with respect

m

to the inner products

(U,V)+k a+k(U,v) ((A+1}~ku,v), u, v £ G . (2.6)

Definition 2.2: The bilinear form b on G is k-singular

(k = o,1, ...) with respect to A (notation b ~a k) if b is
singular in Hk but it is closable in ~+1" The set of all k-
singular forms we denote as FSk.

Theorem 2.1 [7] : For each positive bilinear form b in H

with Q(b) = G~-D(A) the following decomposition is valid

b = br + bso + bsl + "'" ' bskE Fsk ' k=o,1, ...

where br~ br is the largest part of b which is closable in

H. All components b r, bsk, k = o,1, ... , are uniquely deter-
mined by b.
 6O

Note that the decomposition of a bilinear form b on the re-

gular br and the singular b s - b - br parts has first appeared
in KS] (see also [6] ).

Let b be an arbitr~y positive form on G. When b belongs

to Fsk ? Assume that b is bounded or closable in Hk+ I for
some fixed k : b~b (k+1). Then a selfadJoint operator Vb in
Hk+ I is associated with b (k+1) due to the representation theo-
rem

~Ck+1)Cu,v) =(vb u, v)k+I, u,v cDCvb) ~ Qc~Ck+1)) .

We can also associateL with ~(k+1) the operator

T b = I - 1k+l Vb: Hk+l ~ H-ck+1)

where Ik+ 1 : H_(M+I) --* Hk+ I is the unitary operator which

is obtained as the closure of the map

H_(M+ 1) ~ u --~ (A+1)-(k+1) u E Hk+ I, u E G.

Thus

b(u,v) = <TbU,V~ , u,v ~ ( V b ) = D ( T b ) C St+ 1.

Let us decompose ~+I into two subspaces ~+I = Kb~ Rb' where

Kb ~ Ker Vb = Go(k+1) and Rb=EbI. The restriction of ~(k+1)

to Rb we d e n o t e a s b }.

Theorem 2.2: Let b ~>~ mb> o in Rh . Then the follo-

w i n g three conditions are equivalent:

(i) b~l'Sk (b 1 % ),
(ii) Ran(Tb)~ H (k+1) \ H k ,

(iii) Kb is dense in Hk .
 61

Proof. Let b 6 Fsk, i.e., b I ak- Assume that (ii) is not

valid, i.e., that for some u 6 D (T b) the vector w = TbU
belongs %o H_k and . ~ o. Then ~(k+1) [u] > o and since
b I ak, there exists a sequence ~}c- G such that a k [Un]--=

II % II
2
k ' o and
~(k+l) [ % - u ] -~ o, n ~ oo.

This fact means that ~(k+l) [Un] - * -b (k+l) [U ] ~= 0

since I~ (k+l) ( ~ , u) l = I<%, Tbu>l = 1 < % , . >1%

II%llk I1" II_k , O, but t h i s i s impossible i f b~o. Hence
(i) ~ (ii). Let no. Ran(T b) ~ H_k = { o} . Assume that
(iii) is not valid, i.e., that for a non-zero vector v 6 ~ ,
it holds (Kb, v) k = o. It means that Ik+ I Ik-1 v 6 R b = Ran(Vb),
where the last equality follows from b ~ $ m b > o. Hence
Ik I v 6 Ran(Tb), but this is impossible if Ik I v E H_E. Thus
(ii) ~ ( i i i ) . The implication (iii)--~ (i) is equivalent to
that (2.3)~(2.1).

3- Construction o f t h e sin~ularl~perturbed operator

$
Let A = A >o and b ~ Fs with Q(b) = G which
satisfies (2.4). Assume that G O m Ker b is dense in H. The
problem is %o ascribe to a pair A, b a unique selfadJoin% opera-
tor Ab which is interpreted the singularly perturbed operator.
We start with some preliminaries.

Denote Ao = A ~ GO • The operator Ao is symmetric, What

are its deficiency indices n~Ao)?

Theorem 3.1 [7]: Suppose that b belongs %o FSo or Fsl,

and b ~ ~ m b > o. Then

n + (A o) = dim H b = dim Ran(V b) = dim Ran(Tb). (3.1)

 02

If b ~ Fsk, k ~ 2, the deficiency indices are zero, Go

is dense in H2 and Ao is essentially selfadjoint. In addition,
A is different from the Friedrichs extension A~ of Ao if

k = o, while A = Ace for k = I.

In particular, if b = b W, where w 6 H_2 \ H, then the

numbers nl(A o) = I because dim Hbw = I. On the other hand, if

w 6 H_k \ H_2, k >2, (i.e., bw 6 F ), then u~(A o) = o

Sk-1
and A o = A.
Thus all forms b 6 FSk with k ~ 2 are the trivial per-

turbatious of A.

Let b 6 FSo or FSl and b~ ~ mb > o in H 2.

It is convenient to put A~m > o. Then the A - scale (2.5)
is determined equivalently by the inner product (u,v)+_k = (A~ku,v)
*
instead of (2.6). Denote the range of Ao by M and put N=Ker A o
so that H = MeN. The space H2 decomposes naturally into the

orthogoual sum H 2 = Kb~Rb, where K b = Go (2) and R b = Ran(V~.

The following equality is crucial for construction of Ab •

Proposltic n 3.1:

AR b = N, (3.2)

where A = 1 -I : H 2 - ~ H is the unitary operator.

Proof. We have (ARb, M) = (ARb, AG O ) = (R b, Kb)+2 = O

which means that ARb~ N. Ou the other hand, (A-IN,Kb)+2=

(AA-IN, AK b) = (N, M) = o, i.e., A -I N e--

-Rb •

Now we can associate with b the operator Bb lu N de-

fined by
Bb -- A V : A -I~-- I -I Vb~I • (3-3)
 63

Since b S ~ m b ~o, the operator Bb is positive and iuvertible.

For construction of Ab, we need the following well-known abstract
result ( K9, 10] ; see also 111, 12] ).

Theorem 3.2: Let Ao ~ m > o be a closed Hermitian ope-

*
rator such that Ker Ao ~ N ~ o. Then each positive selfadjoint
extension A of Ao is given by the formula

~-I = A -I + B -I (3.4)

where B=B >I o acts in N. The correspondence between

and B is biJective.

This theorem establishes a parametrizatiou of the set of all

positive selfadJoint extensions of Ao iu terms of auxiliary
operators B = B*~ o iu N. In really, these operators play a
role of the boundary conditions for extension A. Due to (3.2)
and (3.3) this is just the parametrizatiou by described above
singular biliuear forms. If we put Abe---~, then (3.4) implies

Ab-1 = A -I~ + B b -I (3.5)

The singularly perturbed operator Ab is specified by A and

b if b belongs to Fso or ~Sl.

Next we want to describe the domain of Ab . The vector

g belongs to D(A b) if it has the following representation

g = f + Bb-IPNAoof, f E D(Aoo ), (3.6)

where PN is an orthogoual projection with the range N .

In addition, we have

Ab g = A~ f. (3.7)
 64

Note that (3.6) and (3.7) is true if b ~> mb > o in H2

and the domains fulfil D ( A ~ ) = D(A~ I) = H. A general descrip-

tion of D(~) has been obtained in [12 ] .

Each operator A appearing in (3.4) is associated with the

form

E ~ b A = bA~ + bB , Q(b) = Q(bA~ ) + Q(b B)

where ~ means the direct sum~ in other words h 6 Q(b) if

h = u +v for u 6q(b A ), v6Q(bB) C- N and b [h ] =

bAoo[ U ] + bB [ v ] (here bA~ is the form corresponding to

A oo and bB corresponds to B). Vice versa the form bB
corresponding to B = Bb, where b belongs to Fso or
Fsl , may be expressed directly by means of b and A:

bB I v ] " ( B b v,v) = (AV~b A-1 v,v) = (AVb~ A-lv, AA-lv)

I' -I v, A-Iv)+2 =
= (VbA ~(2) [A-Iv ], A-lv 6 Q(~(2) ).

The operator Ab is therefore associated with the direct form sum:

bAb[h] = bAoo[ u ] + ~(2) [ A - l v ] , h = u ; v. (3.8)

This decomposition is correct since Q(bAoo)O N = { o }

and N r-Ran(A ).

We have noticed already that A is identical with the Frie-

drichs extension Aoo only if b 6 FSl. lu the case
b 6 Fso, we can define the perturbed operator using the usual
form sum in view of the following fact
 65

Theorem 3.3 [7S: The bilinear form bA1~b A + b is clo-

sable in H iff b C PSo.

The operator AI associated with closure of bA + b is

different from Ab defined by (3.5) or (3.8). In other words, the
form - sum and selfadjoint - extensions methods yield different re-
sults in the case b E PSo. On the other hand, the form-sum
method is not applicable for b E PSl, since bA + b
is not closable in this case. In addition, we have the following
abstract negative result.

Theprem 3.4 [13 ~ : Let a sequence of forms

~bn >r- FSo,
n = 1,2 ... converges to b £ F in the following sense:
sI
S.roS s
Vbn ~ Vb , Kb n K b, n ~ oo (3.9)

Then the sequence of operators An which are associated with

b A + bn converges to the unperturbed operator, An--*- A in
the strong resolvent sense.

The replacement of An by Abn means a regularization,

which leads to a non-trivial effect:

~heorem 3.5 [13 ] : Under the assumptions of theorem 3.4,

the replacement of An by Abn gives Abn ~ Ab in strong

resolvent sense, where Abu and Ab ere defined by A, b u
and b according to (3.5) and (3.8).

It means that the dependence of Ab on b is continuous

in the strong resolvent sense for both Ab in H and Vb
in H 2.
 66

References
I S. Albeverio, J.E.Penstad and R.H~egh~Krohn, Singular perturba-
tions and nonstandard analysis, Trans. Am. Math. Soc. 2~2,
275-295 (1979).
2 S.Albeverio, F.Gesztesy, R.H~egh-Krohu, W.Kirsch, Oupoint
interactions in one dimension, J. Oper. Th. 12, 101-126 (1984).
3 A.Grossmann, R.H~egh-Krohn, M.Mebkhout, The one-particle theory
of periodic point interactions, Commun. Math. Phys. 77, 87-110
(1980)
4 P. Seba, Some remarks on the -interaction in one dimension,
Rep. Math. Phys. 24, 111-120 (1986).
5 F.A.Berezin, L.D.Paddeev, A remark on Schradiuger equation
with a singular potential, Sov. Math. Dokl. ~, 372-375 (1961).
6 V.D.Koshmaueuko, An operator representation for nonclcsable
quadratic forms and the scattering problem, Soviet Math. Dokl.
20, 294-297 (1979)
7 V.D.Koshmaueuko, A classification of singular perturbation of
selfadjoint operators, prep~int 82, 34, Institute of Math.,
Kiev, 1982 (in Russian).
8 B.Simou, A cannonical decomposition for quadratic forms with
applications to monotone convergence theorems, J. Funct. Anal.
28, 377-385 (1978).
9 M.G.Krein, The theory of self-adjoint extensions of semibonded
Hermitian transformations and its applications. I, Rec. Math.
(Math. Sb.), 20 (62), 431-495 (1947) (in Russian).
10 M.Sh.Birmau, On the self-adJoint extensions of positive defini-
te operators, Math. Sb. 38, 431-450 (1956) (in Russian).
11 W.G.Paris, Self-AdJoint Operators, Lecture Notes in Math.
433, 1975.
12 A.Alouso, B.Simon, The Birman-Kreiu-Vishik theory of self-ad-
Joint extensions of semibouuded operators, J. Oper. Th. ~,
251-270 (1980)
13 V.D.Koshmauenko, On the uniqueness on a singularly perturbed
operator, to appear iu Acad. Sci. USSR Dokl. (1988).
 Covariant markovian r a n d o m fields in
four s p a c e - t i m e d i m e n s i o n s with
nonlinear e l e c t r o m a g n e t i c interaction

by

S e r g i o Albeverio*,**,***, R a p h a e l Hcegh-Krohn~,t,
Koichiro Iwata*,~

Abstract
Vv~econstruct covariant r~nttom--
- a_ vector fie~uslaover 4-dimensionM ova,~.~.xm.*"
~ as ~.,1,,,;n,,~ of
. . . . . . . . . .

a system of first order coupled stochastic partial differential equations, best interpreted as
equations for quaternionic valued random fields. The fields are covariant under the proper
Euclidean transformations. We give necessary and sufficient conditions in terms of a given
source of the infinitely divisible type, for the fields to be covariant also under reflections.
In the case of a Gaussian white noise source the fields are Euclidean free electromagnetic
potential fields and have the global Markov property. The fields with Poisson white noise
source can be used as approximation of the Gaussian fields, with better support properties.

* Fakult~it fiir M a t h e n m t i k , Ruhr-Universit~it, D 4630 B o c h u m 1 ( F R G ) ;

S F B 237 B o c h u m - E s s e n - D f i s s e l d o r f
** B i B o s R e s e a r c h C e n t r e
*** C E R F I M R e s e a r c h Centre~ L o c a r n o ( C H )
M a t e m a t i s k I n s t i t u t t , U n i v e r s i t e t e t i Oslo, Oslo (N)
t~ D A A D - F e l l o w s h i p
D e c e a s e d 24. J a n u a r y 1988

To appear in Proc. Dubna Conference 1987, Edts. P. Exner, P. Seba, Lect. Notes Phys.,
Springer (1988)
 70

1. I n t r o d u c t i o n

Local relativistic quantum field theory was invented more than sixty years ago to
provide a synthesis of quantum mechanics and the (special) theory of relativity. In recent
years models of local interacting relativistic quantum fields of scalar, vector or gauge type
have been constructed in space-times of dimension less than 4, see e.g. [AHK1], [AFHKL],
[DST], [GJ], [JLM].
In the case of 4-dimensional space times only partial results are known, see e.g. [DST],
[AFHKL].
In the present paper we exhamine the possibility of constructing a four space-time dimen-
sional theory describing quantum fields of the electromagnetic type, with a formal action
which is not necessarily of the canonical type "kinetic energy minus potential term", but
rather kinetic energy minus a term involving a nonlinear function of suitable linear com-
binations of derivatives of the field.
There is some relation of such models with those of non linear electromagnetic field the-
ories, like BormInfeld theory [BI]. Such nonlinear electromagnetic field theories had been
introduced as approximations to Maxwell fields and our models can also be looked upon in
the same spirit (and we prove indeed a result in this sense). Let us also remark that very
recently the interest of Born-Infeld's action has been reactivated by the discovery that it
describes heuristically the full effective self-interaction of vector fields in the Abelian limit
in open bosonic strings (and superstrings), see e.g. [FT], [CLNY], [CF].
Our models exploit in an essential way the 4-dimensionality of the physical space-time,
which permits to identify it, as a vector space, with the space g-/of quaternions 1).
The fields are given as solutions of a system of coupled stochastic first order partial dif-
ferential equations, having a natural formulation in terms of quaternions. The possibility
of writing such equation relies on the isomorphism SO(4) -~ (SU(2) × SU(2))/2~2. The
Euclidean vector generalized random fields {At(x), x E ~ 4 ~ g_/, r = 0, 1,2,3), identi-
fied with quaternion fields A(x), satisfy stochastic partial differential equations of the form
OA(x) = F(x), with F(x) a quaternionic-valued infinitely divisible field (see e.g. [K1], [Ku],
[Su]) with suitable transformation properties under the proper Euclidean group SO(4)A~ 4,
0 being the basic 1-order quaternionic differential operator with unit coefflcients2).
We discuss the transformation properties of A under reflections as well as Markovian prop-
erties of the fields. In the case of F being Ganssian white noise A is the free electromagnetic
Euclidean potential field. We exhibit a way to approximate the latter field by fields Ap
defined by taking F to be a Poisson type white noise.
We also point out that the fields A can be obtained as continuum limits of corresponding
lattice fields, which makes appear their action as being heuristically given by
Jf (IdivAI, IE- BO dx, withA(x) = (Ao(x), A(x)), x = (xo,~)E ~ × ~3,
j~ _ 0Oz0 _~_ grad~A0 /~ - rot~ A, for suitable real valued functions f on ~/.
Let us also remark that the present work is connected with previous work (see e.g.
[AHKH1-3], [AHK 6], [AHKHK], [Ka] and references therein) in which Markov and quan-
tum fields associated to 1-codimensional hypersurfaces, instead of points, in ~ a were
constructed. For d = 2 such "cosurface fields" can be identified, on closed contours, with
 7]

quantum gauge fields; for d = 4 they include free electromagnetic fields and more gener-
Mly 3-forms with values in the Lie algebra of compact semisimple Lie groups, providing
(by duality) a natural extension of electromagnetic fields to "coloured fields" (this relies
on the realization of ~ 4 and the Lie algebra u(2) of U(2) as the space of quaternions
[AHK3]). The constructed cosurface can also be connected to vector fields, using again the
4-dimensionality of space-time, and these fields satisfy the stochastic partial differential
equation discussed ([AHK2]).
We finally remark that the present paper extends the work of [AHK4] and makes precise
the point first overlooked in [AHK4a] (but shortly remarked in [AHK4b]) that A is not
time reflection invaxiant in the non Gaussian case.
 72

2. A covariant quaternionic partial differential equation

We shall consider a covariant partial differential equation over ~ 4 . This type of equations
can only be considered over ~ 1 , ~ 2 and ~ 4 and their existence is tied to that of the asso-
ciative fields of real, complex resp. quaternionic numbers. In this paper we shall consider
the physical situation with underlying space J~rt4, and the equation is best formulated using
quaternions, as in [AHK2-4]. Let g ' / b e the field of quaternionic numbers and {1, i , j , k}
be its canonical basis.
As a real vector space Ht is isomorphic to ~ 4 by

H-I g x o l + x l i + x 2 j + xak, )(Xo,Xl,x2,xa) • ~ 4 .

We regard ~ as being inbedded in g - / b y identifying t • ~ with t l • g-/, then g'/forms a

real associative algebra with the identity 1 under the multiplication rules : i s = j2 = k 2 =
- 1 and ij = - j i = k.
There is a distinct automorphism of g-/called the conjugation :

x = xo + x l i + x 2 j + x 3 k - - - - , ~ = Xo - x l i - x2j - xak.

As in the case of C we write

1
Rex : = ~(x +~) =x0

I r n x : = ~1 ( x - ~) = x l i + x 2 j + xak

Later we also use the notation ~ for I m x. We see that the square root of the nonneg-
ative quantity x~ = i x is equal to [x[, the ~ 4 - n o r m of x, under the above mentioned
isomorphism g - / ~ Kt4, and moreover

1
x . y : = 7 (Ix + yl - Ix - yl 2) = R e = Re
9c

Sp(1) := {a • g-/; [a[ = 1) is a subgroup of the multiplicative group g-/× := g'/\{0} and
it is isomorphic to SU(2). By g I 9 x , , axb -a • g-I for a, b • Sp(1) we have a surjective
homomorphism Sp(1) x Sp(1) , SO(4), whose kernel is {(1, 1 ) , ( - 1 , - 1 ) ) - 2~2, and
hence [Sp(1) x Sp(1)]/2~2 ~- SO(4).
We consider the following two distinct Sp(1) x Sp(1) actions on ~4-valued functions on
~ 4 : identifying ~ 4 with g-/, the first one is given by

A(x) , a A (a-l(x -y)b) b -a x, y • ~ 4 , (a, b) • Sp(1) × Sp(1) (i)

and A obeying this rules is called a covariant 4-vector field.

The second one is given by

A(x) , bA (a-l(x -y)b) b-! x, y e Kt 4 , (a, b) • Sp(1) x Sp(1) (ii)

 73

and A obeying this rule is called a covariant scalar 3-vector field.

We define a bilinear form by

:= . A(x) = ne . A(x) ex

and extend this as the distributional pairing in the natural way. Note that (., .) is invariant
under Sp(1) × @(1) actions (i) and (ii).
Let
0:= 0 i 0 0 _ k 0_.0_ 0 i 0 . 0 0
Ox---~- - ~ x ~ - J - ~ x 2 Oxa and 0 : = ~ 0 x 0 + 0-~Xl+)O~-x2 + k s '
then 0(9 = (90 = A, the Laplacian in fit4. Consider two variables x, x ~ E ~ 4 related by
x' = a - l x b for some (a, b) E Sp(1) x Sp(1) and define 0' and ~7 in the same way as 0 and
(9. Then it is easily seen that -70 = a-lOb and 0 ~ = b-lOa. Therefore, if A is a covaxiant
4-vector field, then F = OA is a covariant scalar 3-vector field. This is well understood,
3

if we introduce a 1-form a := E Ai dxi, the orientation adapted to {1,i,j, k} and the

i----0
associated Hodge duals. In fact,identifying anti-self dual 2-forms with 3-vector fields, we
have ( * d ' a , da - * d~r) = (Fo , F).
We note that the equation OA = F is not covariant under reflections, since F corresponds
to an anti-self dual 2-form.
1
We denote by g Green" s function to - A , i.e., g(x) 27r21x[2 and set - -

S(x) := - 0 g(x) - 2l 14 , := s ( x )

then we see that

a S ( x ) = -OO g(x) = - A g(x) = 5(x) , OS(x) = 5(x) ,
where 5 is the Dirac distribution. In order to give a precise meaning to the inverse of 0
(resp (9) we introduce the following space

~ : = { ~ E C ° ° ( ~ 4 ' f f - / ) ; '~l~+~lim~ o ( x ) = O , O ~ p E S }

(S -- S ( ~ 4 , if/) is the Schwartz test space of rapidly decreasing test functions).

It is easily seen that 2" ~ ~0 ) c5~o E S is bijective and the inverse map is given by
S 9 ~ ' 5' * ~ E 2., where

• = [ 3(x - v) dv $

Jt~ 4
Using this isomorphism we introduce a locally convex topology on 2.. Note that the injec-
tion t : S ¢--*2. is not dense and hence t* : 2.~ ) S ~ is not injective, since {vh~ ; ~o e S} is
not dense in S.
We have the
 74

Theorem 1 : Let A be a covariant 4-vector field and F be a covariant scalar 3-vector

field. T h e n the elliptic 1-order partial differential equation

OA(x) = F(x)

is covariant. If F belongs to 2"~, OA = F has a unique solution in S ~ given by A = S * F.

We interpret A as a classical electromagnetic Euclidean potential (in the Feynman gauge)
and Ei := O o A i - OiAo , i = 1,2,3, resp. Bi = OjAk -- OkAj ( i , j , k ) cyclic permutation
of (1, 2, 3) as electro resp. magnetic fields. F corresponds to E~ - Bi.

3. R a n d o m fields as solutions of a q u a t e r n i o n i c partial differential

e q u a t i o n w i t h r a n d o m source

We shall now consider the equation OA = F in sect. 2, in the case where F is a generalized
r a n d o m field over ~ 4 with values in H-/. We assume that {F(x)} and { bF(a -1 (x - y)b)b -1 }
have the same finite dimensional distributions for all ((a, b), y) E Sp(1) x Sp(1) x ~ 4 and
we call such F an invariant scalar 3-vector (generalized) r a n d o m field. From the result in
sect. 2 we see that the H-/-valued (generalized) random field A related to F by the equation
OA = F is invariant, in the sense of law, under proper Euclidean transformations. We shall
call such A an invariant 4-vector (generalized) random field (or also, for short, as in the
title, a covariant r a n d o m field). We have:
Theorem 2 : If F is an invariant scMar 3-vector generalized r a n d o m field realized as a
2"~-valued r a n d o m variable, then OA = F has a unique solution A = S * F realized as an
S~-valued r a n d o m variable. A is an invariant 4-vector r a n d o m field. •

In what follows we further assume that F is independent at every point, i.e., if we restrict
its characteristic functional to 8, then taking translation invariance into account we have

with ¢ a continuous negative definite function on ~/. Because of its Sp(1) adjoint invari-
ance, ¢(bAb -1) = ¢(A), ¢ has the following L4vy-Khinchine representation:

¢(~) = - 4 - : - Y ~ 0 ~ + o0~2
2 0+ 21~,12
+ i (1-t-v/Z-1A" o~X(o,,)(1~1)-j:s~,.<,),,(d<:<), ),e

with Sp(1) adjoint invariant Ldvy measure ~ (~(b d a b - ' ) = v(da)) and • E ~ , a o , a > O.
We call ¢ resp. (~, a0, a, v) the Ldvv characteristics of F. If it is possible to extend
the domain of C F ( ' ) to Z, then F is realizable as a 27'-valued r a n d o m variable. To
 75

/ k
this end we assume that ¢(~) = 0 {]$l~ +e) as ,k --~ 0 for some ¢ > O. Indeed under
this assumption the characteristic function CF(') defined on 2" is uniquely determined by
exp ¢(~(x))dx) , ~ E S, and consequently the I ' - v a l u e d r a n d o m variable F is
uniquely characterized by the $'-valued r a n d o m variable e* o F. For we see from Sobolev§
inequality that

]]S*~(x)]]L~ < {// r 2 ]I~(y)l

x _ y ] 3 dy p dx)~ ~ < C [l lln,,
4
P> 3 ' q .
1
-
1
p
+
1
-4"

Theorem 3 : Let F be a translation invariant 2~/-valued generalized r a n d o m field over

2~4 independent at every point with L~vy characteristic ¢. T h e n F has the properties as
in T h e o r e m 2 and A solving OA = F has a distribution given by :

In particular A is an invariant 4-vector generalized r a n d o m field. If the Ldvy measure ~ has

P

the p-th moment, i.e. J la]Pu(dp) < c~ for p = 2, 3 , . . . , then A also has the p-th moment

Z [(~1, A ) . . . (~p, A)] , e $,

as a continuous linear functional on $®P .
We call such F with the properties in Theorem 3 an invariant scalar 3-vector generalized
r a n d o m field of the infinitely divisible type.
Concerning reflection invariance of A, the situation is utterly changed according to whether
F is Gaussian distributed or not. By a reflection we mean the following 2~2-action on
covariant 4-vector fields :

p : A(x) ,-A(-~.) , x e ~ : l 4.

If A is p-invariant as well, then A is invariant under full Euclidean transformations. Before

going into general cases, we first note that A is invariant under the reflection p when ¢(,k)
depends only on ~0 = Re ~, i.e., a = 0 and v is supported by ~ \ { 0 } = {a E H-/× ; I m a =
0}. Indeed since the operators g *. and div • commute with p, we have

- R e $ * p~(x) = Re 0g * p~(z) = g * Re 0 (p~(x))

= g • div p~(x) = - R e S * ~(-~)

for x E 2F/4 , ~ E S, and therefore

 76

We shall now discuss the case of a pure Poisson source, i.e., a0 = a = 0 and v # 0. We
assume that v has a compact support in g_/x and fl = 0 for simplicity. Let {(hi, xi)}i°°=~ be
the Poisson point process on ~/-x x ~ ' with Ldvy measure v(da)® dx, then F is realizable
as an g-/-valued r a n d o m measure over ~t~4 by using {(Oti,Xi)}i=l:°~
oo

F(x) = Zai6{~,}(x)
i=1
and therefore A solving OA = F one has the following representation :
oo X -- Xi
A(~) = S , F(~) = ~ : ~ - _ 7,1" "'
i----1

We now see what happens if we perform the reflection p. Because of the invariance of the
~, oo oo
L : v y measure u(da)®dx the law of { ( ~ / , - i)}/=1 is equal to that of {(hi, xi)}i=l, so that
we have
oo

pA(x) = - A ( - ~ ) = - Z
i=1
~i (-,~
r21 -
-

~ -
x~)
xi["
Z
= 00 '~ : 1 ~
i=a
- (-~i)l' =
i=1
ai : F ; - ;,1 ~ '

where d stands for the law equivalence. Suppose that pAd=A, then OpAdOA = F. However
this does not hold unless I m al = 0 Vi a.s., since it follows from the definition of S that

OpA( x ) d=O
i=1
oo
X -- Xi ~ oo

i----1

and hence there are subsets of S ' which have zero measure for O(pA) and measure 1 for
F. Hence A in the pure Poisson case is not p-invariant, unless I m ai = 0 Vi a.s..
Let us look as a contrast to the case F Gaussian, i.e. v = 0, and see how one recovers the
reflection invariance. We have

m, Is* ¢(x)l 2 dx = (Og • ~, Og • ~) = - ( g , ~, OOg • ~)

= (9.~,~)

= [ 9(~ - v) ~(~) dxdy

J
and thus we have
t
/ Is p¢(x)l2 d x = / IS*¢(x)f2dx ¢~S-
4 J.~4

Combining this with the fact R~ ~ * p~(x) = Re ~ * ~(-~) we get

This implies CA(pC) = CA(f), ¢ E S, as far as v vanishes. This fact is a striking contrast
to the case of a pure Poisson source. We summarize these results in the following
 77

Theorem 4 : Let F be an invariant scalar 3-vector generalized r a n d o m field of the

infinitely divisible type. Then A solving OA = F is p-invariant iff the LSvy measure v
associated with F is supported by ~t\{0}.
From the point of view of the Euclidean field theory, A for fl = 0 , a0 = a = 1 , v = 0
corresponds to the free electromagnetic potential field in the F e y n m a n gauge and A for
fl = 0 , a0 = 0 , a = 1 , v = 0 corresponds to that in the Coulomb gauge.

4. S o m e f u r t h e r p r o p e r t i e s of t h e c o n s t r u c t e d r a n d o m fields

It is well known that the passage from Euclidean fields to relativistic fields is possible in
general situations where the Osterwalder-Schrader (reflection) positivity (see e.g. [G3])
holds under time reversal p. In the following we shall first see how we can construct
from A an Osterwalder-Schrader ( - O.S.) positive field by taking into account the gauge
invariance of the underlying equation OA = F (if A is a solution, then A + v~X with X
harmonic, i.e. A X = 0, also solves OA = F ) in the Gaussian case, v = 0 in the notation
of section 3. Let 8dr be the subspace of 8 consisting of all ~ with R e O~ = div ~ = 0.
We note that Sdf is a Euclidean invariant test function space and {(~,A) ; ~ E Sdf) is a
family of gauge invariant random variables, which we shall call the Euclidean transversal
field with the gauge potential A. Since R e S * ~ = R e g , O ~ = 0 for ~ E 8df, the cova~-iance
functional of the transversal field is equal to

whatever the parameter a0 is. Suppose that ~ E Sdl has its support in J~t+ x ~t 3. By using
the partial Fourier transformation, we have
oo 0 dk= .

Next we apply the integration by parts formula to the dr-integral, then, since div ~ = 0, it
follows that

/0 O0 e-lr, lt~o'~,.)(~)dt = _vrzi /0 00 e-lkl Ikl-lk. ((t, .)(g)dt

and similary

S~0 elr,i,(p()'~s, .)(-~)ds = -vCi L 0° e-lkl'lkl-~k

~ i~ --~• ~'(s,
A .)(-~)ds.
O0
Hence we obtain the positivity

(¢'P¢) = i.. .-i,i,¢(,,.)(k)a, _ L e-I'l'iki-'k'¢("')(k)d' 2--i

 78

This implies the O.S.-positivity of the transversal field : Vzi EaT, i = 1 , . . . , n

z i 2 j E [eVrZ-f(~"A) e -vrZY(~j'pA)] ~ 0 , ~j E ,.qdf , supp [~j] C ~ + × .g~3 .

i,j=l

As usual as we obtain the physical Hilbert space 7"l spanned by

"{evrZY(~'A) ; ~ E Sd$, supp [~] C ~ + X ~ a } with the inner product naturally introduced
by the O.S.-positive condition and the symmetric contraction semigroup acting on 7~,
which is determined by
E [eX/-Z-f(5,r-,A) e-v~-f(LpA)] ,

where {r,} is the shift along the x0-axis : v t A ( x ) = A(xo - t, ~).

The negative H of the generator of the aboVe semigroup is the physical energy operator.
Using this operator we can construct relativistic potential fields as operator-valued distri-
butions (with test functions in Saf). These fields can be identified with the electromagnetic
free potential fields.
Remark 1 : It is possible to show that above Euclidean transversal electromagnetic
potential fields {(~, A>} have the Markov property with respect to arbitrary open subsets
of ~4, in the sense that, extending (~, A) to all ~ E S' with (~, ~) < co and denoting by
~ h -- a ((~, A>; ~ E 8', (~, ~) < co, div ~ = 0, supp [~] C A) V A/" the a-algebra generated by
the fields in the Borel region A, and the zero measure sets .hf (with respect to the measure
associated with A), then for any open D C ~ 4 , ~ D is conditionally independent of ~D*
given ~-,OD, w h e r e / ) is the closure of D, D c - ~:~4 _ D, OD is the boundary of D.
This is proven using the Fock space or Wiener-chaos decomposition of 7"/. For some related
discussions see e.g. [LS] and references therein.
It is also known that the Markov property holds even for { (~, A) }, with ~ not restricted to
be in Sdf ("non transversal fields "), provided one takes the Feynman gauge/~ = 0, a0 =
a = 1, u = 0. In fact this holds also for any Gaussian field defined by

\ ]
with

(with ~ the Fourier transform of ( and c the ratio ~o/a in the notation of section 3). ¢ = 1
corresponds to Coulomb gauge, where one only has the Markov property when restricting
( to be in Sdf.
That even in the non purely Ganssian case u ¢ 0 one should still have Markovian properties
is suggested by the fact that 0 is a first order partial differential operator. However this is
not yet fully mathematically settled. One difficulty is due to the bad spectral properties
of 0 -1 (OA = F being a "zero mass" equation).
In related positive mass equations it is possible to prove the 0-Markov property in the sense
of Kusuoka [K], see [I]. Let us also remark that Surgailis has discussed related problems
in the case where ~ 4 is replaced by _~2, see [Su2].
 79

Remark 2 : As remarked in [AHK3], it is possible to associate to the quaternionic valued

field A in the general case, a component wise 3-form w = (w#, # = 0, 1, 2, 3).
In fact let ao =- A , al -- - i A , a2 = - j A , a3 = - k A . Then a =_ ~ a~,dx~ is also a
#
3
quaternionic valued 1-form. We have ~ 0 r a~ = OA = F.
/~----0
Let w be the Hodge dual of a, then dw = F, in the sense that dwt, = F#, # = O, 1, 2, 3,
where w~ is looked upon as a 3-form over/R 4 and F , is looked upon as a 4-form over htt4.
dw = F can be written as w(OB) = F ( B ) , for any measurable B C ~ 4 , where by definition
w(OB) = fB dw (in analogy with the corresponding formulae which hold when w and B
are smooth), w is then a Markov Euclidean invariant cosurface in the sense of [AHKH1], a
stochastic integral in the sense of [AHKH2]. In [AHK3] the relation dw = F is extended to
the case where the w~ are 3-forms with values in a Lie algebra g containing that of U(2).
w is then a g-valued Markov cosurface.
Remark 3 : All considerations of this section, with invariance properties suitable reinter-
preted, can also be made for the case where the region ~4 on which the fields are defined is
replaced by an open domain B with boundary OB. Let in fact ( F , PB) be the generalized
random field defined by
Ep~(e'/-~(~'F))=exp(-f~ ~p(~(x))dx), with ¢ as in Theor. 2. Let Ou be defined by

closure in L2(dx) from 0 on C°°0(B ; ~tt4). Let SB be the fundamental solution to OB. SB
has the same local behavior as S. The analogue of Theor. 2 holds then with S replaced by
SB, yielding a solution of the equation 0B A = F. A is rotation invariant if B is rotation
invariant.
Let # e be the probability measure giving the distribution of the field A. (A,#B) is a
locally Markov field in the sense of [AHK7], [Ne]. (A, #B) converges weakly as B T ~/4 to
(A, #), with (A, #) given by Theor. 2.
Remark 4 : It is possible to discuss a "lattice approximation" of the field A constructed
in Sect. 3.
Let/i>0, 2g~= { ~ n , n e 2 g ~ . } , A~-=Afq2g~ for any bounded subset Aof~tt4.
Let PA6 (') be the probability measure on/Tarh~ given by

dPA,(F) = (Z~) -Ih'l exp(-W6(F)) 1-[ dF(x),

xEA~

with

xEAs

with f8 a positive function o n / R 4 s.t. f6(7)=f~(l~01, I~1) V ~ 2R4,

Z6 = f~4 e-~4/~('Y)d7 < cx) and
 80

exists for all A C ff~4 and is a L6vy-Khinchine function having the same properties as the
function ¢ entering Theor. 2.
Remark 5 : An example is given by the convolution semigroup {Pt ) , t > 0 of probability
densities associated with ¢. Namely we choose f6(7) = - 6 - 4 1 o g {P,4(~47)616}, then

Let us define
d#~(A) =_-K~e -W~(°~A) 1-I dA(x)
xEA6
with A(x) : A~ ~ g-I, K6 a constant making d#~ into a probability measure and 06 a
discrete version of 0. It is possible to show that (A,#~) converges weakly as 6 $ 0 to the
continuum limit (A, #) described in Sect. 3.
Finally we remark that the field (A, #) constructed from a L~vy characteristic ¢ = Cp of
Poisson type can approximate the free electromagnetic Euclidean field arbitrary well.
In fact let us choose the L~vy characteristic ~r of A to be in s.t., e.g. for r0, r > 0 :

~,,. (l~ol, I,~l)--3 [,%o(~O)+ ,~_,-o(~o)] p(I,~l)/(8~-,,4),

with p the restriction of Lebesgue measure to I~l = r.
Then

- ¢ ( , k ) = 3 cos,k0r0 eV/n-~lY'l~c°~°r2(sinO)dOdqa- 47rr2 / (47rr4)ro,'~° 2

I=r

Calling/Zr the probability measure given by ¢, we have that (#r, A) converges in this case
for r --* 0 weakly to the free Euclidean electromagnetic potential field. This can be used
to study interactions with matter, see [AIW].
Exploiting the support properties of A one can study local perturbations of the field (#p, A).
Let v be a ~tbvalued Borel measurable function on ~ a s.t., for I,kl ~ oo,
4
v(A) = v ( l : ' l ) = O(l~,l'~), ,~ < 5 '

v bounded on compacts.
Let #B be as in Remark 3, with ¢ = Cp. Then

Bv(IA(z)I) dx e LI(#B),

for any B C ~ a bounded measurable. Thus if in addition v is bounded from below, then

d#BV(A) =_ Z ; 1 e- fB -(IA(*)l)d* d#B(A) ,

with ZB the normalizing constant, is a well defined probability measure. (A, #B v) is locally
Markov.
For v suitable, e.g. v > 0 one gets weak limits points as B ,7 ~ 4 .
In this way we can create new locally Markov random fields, covariant under the proper
Euclidean group.
 81

Footnotes

1) This is similar to the association of ~ 2 with complex numbers. Our use of quaternions
is different from one done in a large literature involving quaternionic (and octonionic)
Hilbert spaces for the study of elementary particle models (see e.g. [A] and references
therein). In fact our use is more similar to the one done in relation with classical elec-
tromagnetic fields, starting with Maxwell. Our approach has been partly announced
in [AHK2-4]. On the basis of this announcement Osipov [O] has given an extension,
renouncing of course associativity, to 8-space-time dimensions by using octonions.

2) Euclidean (generalized) random fields as solutions of stochastic differential equations

have been discussed before in [AHK2,3,4]. For lower space-time dimension or Gaus-
sian fields (free fields) see [AHK1,5], [Ca], [GuL], [Gua], [Ha], [JLM], [RS], [Su2] and
references therein.

Acknowledgements

Raphael Heegh-Krohn reported on a previous version of this work in Dubna. On Jan-

uary 24, '88 Raphael suddenly died. In great sorrows we deeply mourn his departure and
acknowledge our great indebtness to him. We also greatfully acknowledge great stimula-
tion received from Prof. Z. Haba by his pointing out at an early stage the natural use
of quaternionic calculus in electromagnetism. We are very grateful to Prof. S. Kusuoka
for his patience and constructive criticism of a previous version of the paper. We also
thank Professors B. Gawedzki, R. Gielerak and R. Streater for helpful criticism on previ-
ous versions of this work. The kind invitation of the first and second author to the Dubna
Conference is gratefully acknowledged, as well as the DAAD support to the third author.
 82

References

[A] Adler, S.: Quaternionic quantum field theory, Comm. Math. Phys. 104, 611-656
(1986)
[AFHKL] Albeverio, S.; Fenstad, J. E.; Hcegh-Krohn, R.; LindstrCm, T.: Nonstandard Methods
in Stochastic Analysis and Mathematical Physics, Academic Press, Orlando (1986)
[AHK1] Albeverio, S.; Heegh-Krohn, R.: Diffusion fields, quantum fields and fields with
values in groups, in Stochastic Analysis and AppNcations, 1-98, Edt. M. Pinsky~
M. Dekker, New York (1984)
[AHK2] Albeverio, S.; Heegh-Krohn, R.: Euclidean Markov fields and relativistic quantum
fields from stochastic partial differential equations in four dimensions, Phys. Letts.
B177, 175-179 (1986)
[AHK3] Albeverio, S.; Hcegh-Krohn, R.: Quaternionic non abelian relativistic quantum fields
in four space-time dimensions, Phys. Letter B 189, 329-336, (1987)
[AHK4] Albeverio, S.; HCegh-Krohn, R.: a) Construction of interacting local relativistic quan-
tum fields in four space-time dimensions, Phys. Letts. B, 200, 108-114 (1988) ;
b) Err. Phys. Lefts. B, Febr. (1988)
[AHK5] Albeverio, S.; Hcegh-Krohn, R.: Quasi-invariant measures, symmetric diffusion pro-
cesses and quantum fields, Proc. Int. Coll. Math. Math. Quantum Field Theory,
CNRS 248,11-59 (1976)
[AHK6] Albeverio, S.; Hcegh-Krohn, R.: Some recent interactions between mathematics and
physics in connection with generalized random fields, Proc. 1st World Congress of
Bernoulli Society. Tashkent 1986, VNU-Press (1988)
[AHK7] Albeverio, S.; H~egh-Krohn, R.: Uniqueness and the global Markov property for
Euclidean fields. The case of trigonometric interactions, Comm. Math. Phys. 68,
95-128 (1979)
[AHKH1] Albeverio, S.; Hcegh-Krohn, R.; Holden, H.: Markov cosurfaces and gauge fields,
Acta Phys. Austr., Suppl. XXVI, 211-231 (1984)
[AHKH2] Albeverio, S.; Heegh-Krohn, R.; Holden, H.: Stochastic multiplicative measure, gen-
eralized Markov semigroups and group valued stochastic processes and fields, J. Funct.
Anal., 78, 154-184 (1987)
[AHKH3] Albeverio, S.; HCegh-Krohn, R.; Holden, H.: Random fields with values in Lie groups
and Higgs fields, in Stochastic Processes in Classical and Quantum systems, Proc.
Ascona, 1985, Edts. S. Albeverio, G. Casati, D. Merlini, Lect. Notes Phys. Springer,
Berlin (1986)
[AHKHK] Albeverio, S.; Hcegh-Krohn, R.; Holden, H.; Kolsrud, T. : Representations and
construction of multiplicative noise, Stockholm and BiBos-Preprint (1987), to appear
in J. Funct. Anal. (1988)
[AIW] Albeverio, S.; Iwata, K.; Welz, B. : in preparation
[BI] Born, M.; Infeld, L.: Proc. Roy. Soc. A 144, 425 (1934)
[Ca] Carlen, E. : The stochastic mechanics of the scalar fields, pp. 40-60 in A. Truman,
I.M. Davies (Edts.), Stochastic Mechanics and Stochastic Processes, Lect. Notes
Maths. 1325, Springer, Berlin (1988)
ICE] Cecotti, S.; Ferrara, S.: Supersymmetric Born-Infeld Lagrangians, Cern-Preprint
Dec. 1986
 83

[CLNY] Callan, C. G.; Lovelace, C; Nappi, C. R.; Yost, S.A.: Princeton Preprint PUPT-1027
(1986)
[DKK] Derrik, G. H.; Kay-Kong, W.: Particle motion and interaction in nonlinear field
theories, J. Math. Phys. 9, 232-240 (1968)
[DST] Davies, I.M.; Simon, B.; Truman, A.: Edts. Proceedings IAMP Conf., Swansea
(1988)
[FT] Fradkin, E. S.; Tseythin, A. A.: Phys. Lett. 163 B, 123 (1985)
[CJ] Glimm, J.; Jaffe, A.: Quantum Physics, 2n Ed., Springer, Berlin (1987)
[GuLl Guerra, F.; Loffredo, M. I.: Stochastic equations for the Maxwell Field, Left. Nuovo
Cim. 27, 41-45 (1980)
[CUR] Guerra, F.; Ruggiero, P.: A new interpretation of the Euclidean-Markov field in the
framework of physical Minkowski space-time, Phys. Rev. Letts. 31, 1022-1025
(1973)
[Cr] Gross, L. : A Poincar~ Lemma for connection forms, J. Funct. Anal. 63, 1 (1985)
[Ha] Haba, Z.: Stochastic equations for some Euclidean fields, in Stochastic Processes
in classical and Quantum System, Proc. Ascona, Edts. S. Albeverio, G. Casati, D.
Merlini, Lect. Notes Phys. 262, 315-328, Springer, Berlin (1986)
[i] Iwata, K., work in preparation
[JLM] Jona-Lasinio, G.; Mitter, P.K.: On the stochastic quantization of field theory, Comm.
Math. Phys. 101,409-436 (1985)
[Ka] Kaufmann, A.G.: Stetigkeit von gruppenwertigen stocha~tischen Kofi£chen, Diplom-
arbeit, Bochum (1987)
[K1] Klauder, J . R . : Measures and support in functional integration, in Progress in
Quantum Field Theory, Ed. H. Ezawa, S. Kamefuchi, Elsevier Science, 31-56 (1986)
[Ku] Kusuoka, S.: Markov fields and local operators, J. Fac. Sci. Univ. Tokyo 1A, 26,
199-212 (1979)
[LS] LSffelholz, J. : Faraday§ law and quantum theory, Karl-Marx Universit£t, Leipzig,
Preprint (1987)
[Ne] Nelson, E. : Probability theory and Euclidean field theory. In "Constructive Quantum
Field Theory", Edts: G. Veto, A. Wightman, Springer, New York (1973)
[O] Ossipov, E.P. : Euclidean Markov fields from stochastic differential equations in
eight-dimensional spaces, Novosibirsk Preprint (1987)
[R6] RSckner, M. : Traces of harmonic functions and a new path space for the free quantum
field, J. Funct. Anal. 7__99,211-249 (1988)
[Sul] Surgailis, D.: On infinitely divisible self-similar random fields, Z. Wahrsch. th. verw.
Geb. 58, 453-477 (1981)
[Su2] Surgailis, D.: On covariant stochastic differential equations and Markov property of
their solutions, Universitd di Roma, Preprint, Fisica (1979)
 Point interaction Hamiltonians
for c r y s t a l s w i t h r a n d o m d e f e c t s

by
S. A l b e v e r i o 1,2,3,~, R. Figari ~,3,4, F. G e s z t e s y 6,
R. Hcegh-Krohnt, H . H o l d e n 5, W . K i r s c h 1,3

Abstract:
We give a short report on work done in recent years on solvable models for quantum
mechanical crystals (crystals with point interactions, thus three dimensional extensions of Kronig
Penney's model). We discuss the mathematical definition of the Hamiltonian and its spectral
properties in the case of perfect crystals, as well as in the case of crystals with deterministic or
randomly distributed point defects. We also discuss the connection of such point interactions
Hamiltonians with the study of scattering by a large number of small randomly placed scatterers.

1. Introduction

In this paper we shall report on some recent mathematical work on models describing crystals
with defects which are randomly distributed. In the formulation of these quantum mechani-
cal models so called point interactions arise; these are interactions localized at points of the
perturbed crystal, and are felt by the particle (electron) moving in these crystals (in the usual
one-electron approximation of the motions of electrons in a crystal). The crystals we discuss
here are mainly three dimensional (but similar results are obtained for two and one dimensional
crystals). Despite extensive and very interesting work developed in recent years on point inter-
actions in three dimensions, described e.g. in monographs [AGHKH], [DO], it seems that quite a
few physicists, mathematical physicists and mathematicians still believe that point interactions
only are possible in one dimension--an immediate association being with the Kronig-Penney
model (since it has entered standard text books in solid state physics). This entire workshop
has been a proof Of how active is the research concerning three dimensional models with point
interactions. We hope that the present contribution might also help eliminating eventually the
above mentioned prejudice. In fact all is done in one dimension with point interactions can be
done also in three dimensions, provided of course the point interactions are correctly defined.
We shall report here mainly on work of the authors and their coworkers, in particular, as far
as random perturbations are concerned, F. Martinelli. We refer to other contributions in this

1. Fakult£t fiir Mathematik, Ruhr-Universit£t Bochum, D-4630 Bochum 1 (FRG)

2. BiBoS Research Centre
3. SFB 237 Bochum-Essen-Dfisseldorf
4. Dipartimento di Scienze Fisiche, Universit£ di Napoli (Italy)
5. Matematisk~Institutt, Universitetet i Trondheim, N-7034 Trondheim-NTH (Norway)
6. Department of Mathematics, University of Missouri, Columbia, Mo 65211 (USA)
7. CERFIM Research Centre, Locarno (CH)
t Matematisk Institutt, Universitetet i Oslo (Norway)
Deceased 24 January 1988
 88

volume for complementary topics, see also e.g. [Pa] (and references therein). Our basic reference
for this paper is the monography [AGHKH], to which we also refer for more complete references.
For the reader at his first contact with point interactions let us start by answeringbriefly the
question:

1.1 What are point interactions ?

A point interaction at the origin 0 in the d-dimensional Euclidean space ~ d should be a pertur-
bation ("potential") localized at 0 of the free Hamiltonian, thus the Hamiltonian (Schrbdinger
operator) describing this interaction has the form (in suitable units)

H = " - & + AS(x)" (1.1)

as an operator in L2(IRa, dx) (square integrable functions over ~ d ) . Does H exist as a well
defined self-adjoint operator, is it non trivial in the sense of being different from the free part
- A ? More generally point interactions at a subset Y of R d should be

H = "- A + E Ay~(x - y)" (1.2)

yEY

in L2(.~ d, dx ).
Models of this type, with different choices of Y, occur in nuclear physics, solid state physics,
electromagnetism, see [AGHKH] and references therein. It is well known that there is a "no go
theorem" for Y discrete (without accumulation points) if d > 4, - A being already essentially
self-adjoint on C ~ ( ~ a - {0}) (C ~ functions of compact support in the complement of the
origin) if d > 4. I).
It is also well known that, as an application of Krein's theory, as first discussed by Berezin
and Faddeev, when d < 3 for Y consisting of only one point there exists a 1 parameter family
indexed by ct E ~ , of different realizations - A a # - A of H. The parameter c~ determines for
d = 2,3 a renormalized coupling constant. For d = 1, a is simply A. The way - A a arises is
perhaps best seen by an heuristic argument (justifiable by nonstandard analysis [AFHKL]):

(-A+AV-k2) -' = a k - G k A +VG} Vak, (1.3)

with Gk - (--A -- k 2)-1, k 2 # 0, as computed rigorously for V say bounded and continuous.
Setting then formally V(x) = $(x) in this formula one sees that, for d = 2, 3, one has to choose
!.~___ -G0(0) - ct to compensate the singularity of IGk(0)l ( = +eo for d = 2,3). By this choice
of ~ we get [I + VGk] = ,k - a. As suggested by this (1.1) can then be defined using the final
result of this formal computation, namely as the selfadjoint operator - - A a in L2(~ d, dx), d < 3
with resolvent kernel given by

( - A . - k') -~ (x,y) = a , ( ~ - ~1 - C~(~) ~ - ~ C~(~). 0.4)

1) It is a different story if instead of L2(.nrld, dx) one considers some other spaces, as in some
uses of point interactions in electromagnetic theory, see references to work by Grossman and
Wu in [AGHKH]. For recent results on point interactions situated on non discrete subsets Y
see in addition to [AGHKH] and contributions to these proceedings, [Sra], [ABrR], [AMaZ1-3],
[AFHKL], [AFHKKL], [go], [HI, [Pan], [Te] and references therein.
 89

Remark: The mentioned justification of the above formal computation by nonstandard analysis
yields that, for d = 3, and 6 infinitesimal, - & + ~=(e)6,(x), with 8,(x) a nonstandard realization
of the ~-function (in standard terms, 8~(z) is a delta sequence for g ~ 0), is near standard and
defines --A~, ¢ --A if ~,(¢) = [_4%g + , r ' ~ a ] ~ e 3, a E ~ . A similar result holds also for d = 2.
See [AFHKL] and references therein. (1.4) or the observation in the remark give a realization
of - A o as a ~ - interaction or point interaction of strength a at the origin.
Remarks 1) It is useful to remark that setting a = +co formally in (1.4) the r.h.s, reduces to
Gk, so that it is natural to define - A + ~ ___-- A .
2) Besides the mentioned two ways to define - A ~ , the one by the resolvent and the nonstandard
analytic one, there exist other ways to define -A~, e.g.
a)"by boundary conditions": formulated for d = 3, - A ~ can be characterized as the extension
of --A on G ~ ( ~ 3 - {0}) functions with domain such that if f = D ( - A ~ , ) and f(x) = ](r),

b) "by resolvent limits": define for ~ > 0, H~ = - & + $(¢)¢-2V(x/~), with V in Rollnik's
class, (1 +1" I) V Z L1(~3), and $ E C ' ( ~ ) , $(0) = 1, $'(0) # 0 such that - 1 is an eigenvalue
of vsignVGov, v = IVl~ with eigenfunctions 7~¢, j = 1 , . . . , N in L2(~3,dx) such that the
"resonance functions" Gov~j are not in L2(~i~3, dx) for some j . Then H~ converges in norm
resolvent sense as e ~ 0 to - A ~ , with a = - ~'(o) This is an approximation of - A m by
local potentials, there are also approximations by non local potentials, see again [AGHKH].
c) Another useful construction of --A¢~, which provides probabilistic tools for the study
of point interactions, has been first discussed in [AHKS]. Let, for d = 3, ~o(x) = a ~ - - ~ F '
a E ~ . Let Ho be the self-adjoint positive operator in L 2 ~ , 3, I~J 2 dz), uniquely associated
with the Dirichlet form E ( / , f ) = ~ f IV/12J~l~dz in L~(~, I~offd~)
(in the sense that Hgf, H g f = E ( f , f ) , Vf e D(HJ) = D(E) ). Then we can define
- A a by -A~, --" ~,~/~,~X1 _ (4~ra)~. /7/~, on C ~ ' ( ~ 3 - {0}) is given by - A - ~ . V, with
~a -= V In 7>a. /~a generates a diffusion Markov symmetric semigroup in L~(~3; ]7~a]2dx), with
invariant measure 17~ol2dx.
Having solved the problem of the construction of self-adjoint realizations of the one source point
interaction, by one of the above methods, it is not difficult to extend the solution to other cases,
of the type (1.2.) with Y consisting e.g. of N points in ~ 3 or a discrete subset of ~ 3 (see
below). One can also discuss the case where Y is'some other suitable geometrical measure zero
subsets of ~t~a like e.g. S ~ (see e.g. [AGS]) or the path of Brownian motion on ~ d , d _< 5, run
in [0,t] (see [AFHKL], [AFHKKL]). In this article we shall discuss some situations where the
particle and the centers are in ~ 3 , for other cases in particular finite and infinitely many centers
in ~ d , d = 1, 2, see [AGHKH] and also e.g. [DO], as well as contributions to this conference, in
particular by P. Exner and P. ~eba.

1.2 N-centers point interactions

Before going over to the ease of infinitely many centers, let us consider the N-centers ease, given
heuristically by (1.2.) with Y consisting of N points ("sources") in ~ 3 . We give strengths
{au,y q Y} corresponding to the sources y E Y. The point interaction Hamiltonian for the
sources y, with strengths my, denoted by -A~,,y, is given in terms of its resolvent by

(1.5)
y,y' E Y
 90

with
= - a, - v') + ¢) ,v,y' e Y (1.6)
y,ys

where G/~(z) = G~(z) for z # 0 and Gk(z) = 0 for z = 0.

Remark. Actually there exist N 2 self-adjoint extensions of - A restricted to C ~ funtions of

compact support outside the sources y E Y, only N of which are covered by above definition.
However only the ones given above do correspond to separated boundary conditions at the
sources, the others are "non local" involving non separated boundary conditions, see [DaGr],
[Bra]. Having the above resolvent it is possible to discuss in details spectral properties and
scattering for N centers, see [AGHKH].

1.3. The case of infinitely many centers

We shall consider the case of Hamiltonians given formally by (1.2) with Y an infinite subset of
m s, discrete in the sense that

Y = {y~,i E z~V} with inf¢¢j, l y ¢ - Yj'I > 0.

We give sources a i , j E z~¢ and denote for any 12 C Y, 12 finite, by & the restriction of a to 12.
( )_1 as 12 1"Y.
One then defines the Harniltonians for point interactions with sources Y and strengths a by the
limit in the sense of strong resolvent convergence of --Ae,,f. -- k 2
That this limit exists is easily seen by using monotonicity arguments, see [AGHKH]. Since by
(1.2) we have good control on the approximants it is possible to get information on the limit,
especially in the case where (a, Y) have suitable symmetry properties, see [AGHKH]. Such a
case is the one of crystals, which we shall handle in the next chapter.

2. Periodic uoint interactions and crystals.

We shall consider in the one electron model of a solid the case of a multiatomic crystal or a
perfect alloy, with point sources located at the points of a subset Y of ~ 3 of the form

a Bravais lattice and Yo a finite number of points o f / R 3. Let f" = ~ 3 / A be the basic periodic
or primitive cell or Wigner-Seitz cell, i.e.

Let bj be dual basis vectors in the sense that

aibj = 2rc~51j, i,j = 1,2,3.

Let
I~ ~ n i b i , (nl,n2,n3) E ~ 3 .
i=l
 91

r is called the dual lattice.

i=l

is the so called Wlgner-Seitz cell of the dual lattice I" or Brillouln zone. f" is the dual group
of A and is the basic periodic cell or primitive cell of the dual lattice. One has the direct
decomposition L 2 (/~,I2(F)) = f~12(r)dO. We c • look upon e + 7,7 e r as coordinates
corresponding to/3 E / R 3. Our periodic Hamiltonian/~r, describing crystals (in Fourier space),
is unitarily equivalent to a direct integral:

,,,/ae~(O)dO
for s o m e / t ( e ) acting in 12(r).
The study of the spectrum of/~r is then reduced to the study of the spectrum of H(8).
I f / t is the momentum space realization of a Hamiltonian, then /~(e) is called the reduced
Hamiltonian. E.g. if H is - A , then -,~(e) is the operator of multiplication by ]7 + 8[ 2 in/2(F)
and the spectrum of - £ ( 8 ) is the discrete set IF+6[ 2, so that a ( - A ) = U a(-£(8))d~
consists
eE£
of bands, the spectrum in each band being purely absolutely continuous. We shall now discuss
the Hamiltonians corresponding to point interactions located at the points Y0 + A. Formally it
is given by - A + V(x), with

v(~) = - ~ ~ ~ 6 ( ~ - ~ - ~), (2.1)

~j E ~'0 AEA
jffil,...,N

with (unrenormalized) strengths #1 e ~ , j = 1, ..., N (with N number of points in Y0).

The following Theorem, proved in [AGHKH], shows how to construct the point interaction
Hamiltonian corresponding to the interaction given by the above formal expression (2.1):

T h e o r e m 2.1 Let
N
1
+ K

5 j=l
with K > 0 a cut-off, (,) the scalar product on/2(F), and
¢~(e,.) --- x . (~ + e)~ -~+~)yj ,
X/¢ being the characteristic functions of the closed ball of radius K centered at the origin. We
have that

D (/tK(0)) = D (-/X(0)) = { g E/2(F)[' 7er

~ 17 + 01'1g(7)12 < ~ }"

K -1
Choose pJ¢ = (aj + ~-~2) , with aj E ~ . Then HK(e) converges in norm resolvent sense as
K ---* co to a self-adjoint operator -/~a,^,y0(6), the reduced Hamiltonian for point interactions
on A + Y0. This operator is given by its resolvent through the formula
N
(-,L,^,y.(o) - k~)-' = a~(e) + If'l-' ~ [-ro,^,.o( h,e)l.,-' (F_~,.,, (0, .), .) F~,,, (0, .),
j,j'=l
 92

[ro,^,r0(k, 0)]i, i, - ~ z ' - gk(yj - yi', 0),

with
ei(7+0)~ ]
7Er,[7+el_<gl7 + {~1 - -

gk(x, 0) =
(2r)-ae-lO~lirao -47rK + E (]7 + 0[2 - k2) -1 b~l for x E A.
7EPJT+0[<K
e-i(7+o)y
F~,y(0,7) - 17 + 012 - k 2 ' G k ( O ) - - (17 + 0[ 2 - k2) - 1 .
The periodic point interaction Hamiltonian with sources on A + Yo and strengths ayj = c%+A
(independent of A E A) is given by --A~,A+yo ~ f ~ (--Aa,A,Yo(0)) dO, hence determined by the
above resolvent. •
In the simplest case Y0 = {0), writing --/~o,A for --]X~,A+{0}, we have the following spectral
results:
T h e o r e m 2.2 a (--/~o,A(0)) is purely discrete consisting of isolated eigenvalues of finite
oo
multiplicity. We have ~ - IF + 0[2 = U I,(O), with I0(6) = (-oo,02), In(O) being bounded

disjoint open intervals for n >__1, each containing exactly one eigenvalue E~,^(8) for -/~o,^(~).
These eigenvalues are increasing in a. In addition a point EA(O) E IF + 012 is an eigenvalue
of --Ao,A(8) of multiplicity m >_ 1 if and only if there exist 3'.... ,7-, E F such that EA(0) =
17o + o1" . . . . . I'~,.,, + ol ~.
For the proof see [AGHKH]. Having this result one then gets information on the spectrum of
the point interaction Hamiltonian for a crystal A:

T h e o r e m 2.3 The spectrum a ( - A o , ^ ) , of a crystal with point interactions on a lattice A,

with strengths a equal at each point of the lattice, is purely absolutely continuous and has the
form of the union of two intervals

with
1
00 - - ~ ( b l + b2 + b3),

E~'A =- min {E'~:a(o),l]b_12 } = mineeh [E~'a(O)] ,

with
[b_ I < bi, j = 1,2,3, b_ E {bl,b2,b3}.
We have E7 '^ > 0 for all c~ e ~ . Moreover Eo'^(Oo) < 0 if a < r~0,A ~ g0(0, 00) (in this case we
have thus effectively a gap!). The spectrum of a (-Ao,A) is monotone increasing in ~. One has

and
E~'^(O°) ---* l lO°~- as avt ---* +°°
There exists an az,^ E ~ such that
v.>_olA
(i.e. for large enough strengths the gap doses).
 93

Remark. a) There exist some extensions of this result to the case where the basic cell consists
of more than one point i.e. IY01 > 1. E.g. in such a case a (-A~,,y0+^) n ( - c o , 0) consists of at
most ]YoI disjoint closed bands, see [AGHKH] and references therein.
b) One can approximate --At,,r0+A in the norm resolvent sense by scaled short range Hamilto-
nians, which can be exploited for obtaining information on crystals with interactions not of the
point interaction type.

3. R a n d o m point interactions, defect crystals.

We consider the models of Sect. 2 with the sources on Y0 + A replaced by random sources
located at the random subset Y(w) of ~ 3 , with w a point in a probability space f~, Y(w) being
for each w a countable subset {yi(w),j e ~W} of ~ 3 such that infij, e]~iyi(w ) -yi,(w)l > O.
Let a(w) -- {ct~j(w),j e iN} be a Y-indexed family of random strengths (real valued variables).
For each w, by the methods indicated in Sect. 1.3, we can define a point interaction Hamiltonian

A particularly interesting case is the following. Let Xx, A E A be i.i.d. {0,1}-valued random
variables associated with the points of a Bravais lattice A0. Set p = P(X~, = 1). Choose Y(w)
to be the set of occupied sites in A i.e. Y(w) = {A E A,X~(w) = 1}. Assume {ax, A E Y}
are i.i.d, random variables with supp P~,0 compact. Then H,~ has the interpretation of a point
interaction Hamiltonian describing a crystal with randomly distributed defects. If A(w) = A
then it is natural to talk of a random alloy, with types of alloys described by the state space of
a. Using the fact that both c~ and Y are i.i.d, we have the following theorem:

Theorem 3.1 (//~,,w E f~) is an ergodic family of self adjoint operators in L2(~ri~3) (relative
to the natural shift operator in path space). The spectrum a(H,~) and its different parts like
cr,ss(H,,,), and the closure %(H,~) of the point spectrum of H,,, are non random subsets of/R,
almost surely. Moreover, the discrete spectrum ad(H,~) is void, almost surely. Finally, for any
r E ~ there exists a subset £tr of f~ of probability 1 such that r is not an eigenvalue of finite
multiplicity of H~, for w E fl~.
Remark. This result belongs to a type of results established in various contexts by Pastur,
Kirsch-Martinelli, Englisch-Kfirsten, see e.g. [AGHKH], [Ki], [KiMa] and references therein.
It is useful to refer to the family q'(w) = {ax(w),Xx(w),A E A} as to a "stochastic poten-
tiM". Let H(O(w)) = H,,, = --Aa(w),y(t¢) be the corresponding Hamiltonian. We call any
- { ( ~ , T/A) E suppPt,0 x {0,1}, A E A} an admissible potential. The set of all admissible po-
tentiais is denoted by .,4. Let us set A(¢) - {A E A, r/A = 1} and H ( ¢ ) - -A~,(O),^(¢). We call
¢ E .A ueriodic with periods Li, i = 1,2,3 if there exist linearly independent Li E A - {0} such
that ~A+L~ = ~ and ~/a+L~ = T/a for all A E A and all i = 1,2,3. This means that the charges
as well as the occupied sites are Li-invariant. We call 7~ the family of all periodic admissible
potentials. We then have

Theorem 3.2 The spectrum a ( H ( ¢ ) ) of the Hamiltonian H ( ¢ ) for any admissible potential
is contained in ~,, with ~ = a(H(q,(w))) for almost every w E f L Moreover ~ = U a ( H ( ¢ ) ) =
~qA

U
 94

Proof." This also follows from the references mentioned in the previous theorem, see e.g.
[AGHKH].

Remark: One can show in addition that:

a) Given p < 1 the set ~ depends only on suppP¢, 0

b) E has a band structure, see [AGHKH].

It is possible to give a rather detailed study of the variation of the negative part of the spectrum
of H(¢(w)) as one removes some of the point interactions. In fact we have

T h e o r e m 3.3
Let p -= inf[suppP~,,],
v -- sup[suppPa0].

Assume that either E~'^(0) < E0~'^(S0) or suppPa 0 = [#, v]. Then the assumption p = 0 implies

(3.1)
= ~(-~.,^) u ~(-~,^).

If v < a0,n we have

N (-~,0) = [E~'^(0),E~'A(e0)]

and i f p > a l , ^ we have E = a ( - A , , a ) .

Remark. a) If 0 < p < 1 the same results hold, but (3.1) should be replaced by

= [E0~'A(0), E;'^(00)] U [0, oo) = a(--A.,A) U a(--A~,A) U [0, oo).

b)For p > al,A we have the result that a(H,~) = a(--A~,^) almost surely, i.e. the spectrum
does not change if starting from the random Harniltonian we create or switch off points with
arbitrary strength.
c) For results on random Hamiltonians formally given by

H(w) = -A + ~ ~(x -- b(s,w))ds

with b : [0, t] ~ ~ a a Brownian motion in ~ d , d < 5 see [AFHKL] [AFHKKL] and references
therein. Such Hamiltonians arise in the description of the scattering of a quantum mechanical
particle by a polymer. Moreover, via the Feynman-Kac formula, they also enter the construction
of polymer measures of the heuristic form

exp [-~ fot ~ot ,~(b(sl)-b(s2))dslds2] dP(b)dP(b),

with (b, P), (b, P ) two independent Brownian motions in ~'Rd, d < 4. The densities of such mea-
sures also occur in Symanzik's representations of self-interacting scalar quantum fields described
by interaction densities v(s) which are functions of s 2, like e.g. in the (~04)a-model, see [AFHKL].
 95

4. Point interactions and scattering by a large number

of small randomly placed scatterers

In the definition of the N center point interaction (1.5.) all the physical constants characterizing
the strength of the interaction, the energy of the quantum particle and the mutual distances of
the scatterers are contained in the matrix Fa,y(k).
In particular they appear in r as the inverses of three lengths: the a y , y E Y = {Y1,...,YN}
represent the inverses of an "effective linear size " of the scatterers, (see [AGHKH]), k is propor-
tional to the inverse of the wave length associated with the quantum particle and the Gk(y - y')
are proportional to the inverses of the distances between the scatterers. In this section we shall
study some limits N ---* 0% different from those studied in Sect. 2,3, but also of physical
relevance cfr e.g. [Lo]. We report here on work in [FHT] - [FOWl.
In the following the orders of magnitude of the above lengths will be expressed as powers of
the number N of the scatterers and the volume of the system will be considered fixed and
finite (of course only the dimensionless ratios of the lengths are to be considered as meaningful
parameters). We shall consider separately three situations, which we shall denote by cases 1),
2), 3).
Case 1:
inf v,v'eY]Y - Y ' ] = 0 ( N - i ) , ]c - 1 ---- 0 ( 1 ) , O~y 1 ~-~ 0 ( g - 1 )

. In this case there is a large number N of scatterers in a finite volume, each one of strength being
of order N -1. The wavelength of the quantum particle is large with respect to the scatterers
effective length and to the interparticle distance.
Physically, in this case one expects a finite effective potential depending only on the local
strength per unit volume of the scatterers.
Case 2:

In this case the wavelength is of the same order as the interparticle distance. Each scatterer has
infinitesimal strength but the "surface per unit volume" is kept constant.
The limit problem is expected to contain information about the local statistics of the interparticle
distances. It is the relevant limit, for example, in modelling scattering of neutrons by liquids
(scattering due only to the nuclei of the atoms in the liquid). In fact this kind of scattering
experiments are often performed to investigate the range of solid-like order in the interparticle
distances in fluids.
(~ase 3: All the lengths are of order N - i . This case is in fact a genuine infinite volume limit
case. There are no negligible terms in the F matrix.
This case is the relevant one for modelling amorphous or perfect crystalline solids, as discussed
in Sect. 2,3. In what follows we give results for cases 1) and 2), for dimension d = 3 ( the ease
d = 2 can be worked out as the d = 3 case, whereas the case d = 1 is much simpler).
(~ase 1 Let Y(/~)= ~y!N),...,y(,N)~beasequence
k " --) of N-tuples of points in K/3 such that

N
~N) .__.
'= NToo

with p(=) _> 0, p(.) C L l ( ~ 3 ) N L 2 ( ~ 3 ) , f~s p(x)dx = 1, the convergence being weak convergence
of probability measures on ~ 3 .
 96

Some technical assumptions on the distribution of the y~N) will be needed; e.g.

inf i,/ y~N) y~N) > c N - a f o r s o m e a E ( 1 , l )

1 x--', 1
<_ C, VN
"---i ~
N N) N) ~

If, for example, the y~N) are chosen to be N independent, identically distributed (i.i.d.) random
points of ~ 3 with common distribution density p, the above stated assumptions are satisfied by
any configuration of a set of measure increasing to 1 when N goes to infinity.
Let or(N) = {oty, y E y(N)}. We are looking for the existence of a limit operator for the sequence
--ANa¢~),V(N ) and for an explicit characterization of the limit. If all the a~N) are bounded away
from 0 and inf 1~i,o~N y~g) _ y~N) >> 1, as a first order in a perturbation expansion, we get,
as N -----~oo:

and
(-t,N~(~),v(~) - k~) -~ (~,y) ~ a~(~,~)+
N
-kN~Gk(x,y~N))Gk(y~N),y) (w~N')-' (4.1)
jml

If the a~h') are chosen to be the values a (y~N)) in y~N) of a function which is continuous
(outside a set of p(x)dx measure 0 ) and 0 < a < [a[ < b < co, the right hand side of (4.1)
converges to
Gk(x,y) + kl. , )~(z) kl. ,y)dz =
= ( - ~ - k~)(x,y) + [ ( - ~ - k~)-' ~ ( - a - ks) -'] (x,y) (4.2)
(4.2) is the first term of a perturbation expansion of ( - A + ~ - k2); up to the first order
--ANo,(~),y(n ) behaves for large N like ( - A - ~).
In fact the result is true up to any order in the expansion. One has in particular the following

T h e o r e m 4.1.: Under the assumptions made before

s - lim (--AN~,(N),y(N) + ~)--1 =

NTe~
( - A - /9 + ,~
a
)' = Ap

for ,k > 0 sufficiently large.

For a detailed proof and further comments see [FHT]. If the y~N) are i.i.d, random points with
common distribution density p the above stated theorem can be looked upon as an operational
law of large number. The corresponding central limit theorem can also be proved:

T h e o r e m 4.2.: For any f, g E L 2 ( ~ 3) the random variable

N½(f,[(-AN~(~,,,y(~,)+)~)-I-AXp]g)=~(N)(Y (N))
converges in distribution when N goes to +co to the gaussian random variable ~x with mean 0
and variance.
E(~X) x x a - ~ ApfApg)L
=(ApfApg, x x ~ - ( A ~J , ~ -1 A~9)L,
x 2
(here L~ = L2(1Fl3, pdx ) ).
For the proof and further comments see [FHT] and [FOT].
 97

Remark: If V is any function in L I ( ~ 3) n L2(~t 3) and we take p, cx as

p(~) = IV(:~)l//,,, IY(~)ldx

O/(X) = (sign V)(~)//~Iv(~)ldx
so that ~ = V, then Theorems 4.1, 4.2 express the fact that any one particle Hamiltonian with
a potential V of class L 1 N L 2 can be arbitrarily well approximated by a Hamiltonian with zero
range potential on an increasing number of points.
(~ase 2: In experiments of neutron scattering, neutrons with a wavelength of a few/~ngstrCms
are used to investigate samples of condensed matter. The average interparticle distances are of
the same order of the wavelength, while the range of the interaction of the neutrons with the
nuclei of the atoms is of the order
10 -13 cm.
N
A first order expansion for the differential cross section da/d~ for a formal potential 27rb~'~(x -
i=l
y~N)) (in the usual units such that h = m = 1) gives

(dd__~) (x)=b ~ N ,~(',,,,~-'Jc,,~)

i,1

where z is the transferred momentum and y~N) are the positions of the scatterers.
Notice that the formal expansion cannot be continued beyond the first order term since infinities
due to the singularity of the Gk at coinciding points would appear in each higher order term.
For the N-centers point interaction the above mentioned result is exact for N large in the
scaling described above as pertaining to case 2 (which in a box of volume proportional to N

can be redefined as ly, - yJl = 0(1), k-g = 0(1),N½a~-' = 0(1)). In fact let ~ NJ be
,y(N)
the differential cross section corresponding to N-center zero range interactions, all of the same
strength N-½ a - I , placed at the points {y~N), ..., y(NN)}. Under some technical assumptions on
the distribution of the y~N), it is possible to prove that

lim (~--~) - - - 2 . .1. i e x - , ,k(y!~)-y~N)'~

x, ' , ----0 (4.3)
N/"oo N½a'y(N) Na2 i,j

For this see [DFZ].

Notice that if the y~N) are distributed according to an homogeneous point process in ~t 3 of
density p and if the static pair correlation function given by

dr--
4rr 2
is decaying fast enough to p for large r, the common limit of the two quantities appearing in
(4.3) is ~ + a-~-~(x) (~ denotes the Fourier transform of g).
It should be stressed that for the N-centers point interaction the terms of the perturbation ex-
pansion, disappearing in the limit N ~ oo, are explicitly known.
 98

Acknowledgements

This report is based on a lecture given by the first named author when he was visiting Dubna with
Raphael Hcegh-Krohn in the fall '87. On January 24th, 1988, Raphael suddenly passed away.
He had been a standing source of inspiration for all of us and we deeply mourn his departure. We
thank J. Brasche, G.F. Dell'Antonio, P. Exner, P. ~eba, W. Karwowski and L. Streit for many
interesting and simulating discussions. The kind invitation of the first and fourth author to the
Dubna Conference is gratefully acknowledged.

References

[ABrR] S. Albeverio, J. Brasche, M. RSckner, Dirichlet forms and generalized Schr6dinger operators,
in preparation, to appear in "Lectures on Schr6dinger Operators", from the Nordic Summer
School in Mathematics 1988, Ed. A. Jensen, H. Holden, Lect. Notes Phys., Springer 1989
FHKKL] S. Albeverio, J.E. Fenstad, R. Hcegh-Krohn, W. Karwowski, T. LindstrCm,
Schr6dinger operators with potentials supported by null sets, in preparation, to appear in
Proc. Syrup. in Memory of R. Hcegh-Krohn
[AFHKL] S. Albeverio, J.E. Fenstad, R. HCegh-Krohn, T. LindstrCm, Nonstandard Methods in
Stochastic Analysis and Mathematical Physics, Academic Press, Orlando (1986)
[AGHKHl S. Albeverio, F. Gesztesy, R. HCegh-Krohn, H. Holden, Solvable Models in Quantum Me-
chanics, Springer, New York (1988)
[AGS] J.P. Antoine, F. Gesztesy, J. Shabani, Exactly solvable models for sphere interactions in
quantum mechanics, J. Phys. A 20, 3687-3712 (1987)
[AHKSI S. Albeverio, R. H~egh-Krohn, L. Streit, Energy forms, Hamiltonians and distorted Brow-
nian paths, J. Math. Phys. 18, 907-917 (1977)
[AMaZI] S. Albeverio, Ma Zhiming, Nowhere Radon smooth measures, perturbations of Dirichlet
forms and singular quadratic forms, in preparation, to appear Proc. Bad Honnef Confer-
ence, Ed. Christopeit et al. (1988)
[AMaZ2] S. Albeverio, Ma Zhiming, Additive functionals, smooth nowhere Radon and Kato class
measures associated with Dirichlet forms, in preparation
[AMaZ3] S. Albeverio, Ma Zhiming, On the perturbations of Dirichlet forms, in preparation
[Bra] J. Brasche, Perturbations of self-adjoint operators supported by null sets, Ph.D. Thesis,
Bielefeld (1988), and papers in preparation
[CaS] C. Carvalho, L. Streit, in preparation
[DaGr] L. Dabrowski, H. Grosse, On nonlocal point interactions in one, two and three dimensions,
J. Math. Phys. 26, 2777-2780 (1985)
[DFZ] D. Dfirr, R. Figari, N. Zanghi, in preparation
[DOI Y.N. Demkov, V.N. Ostrovskii, The Use of Zero-Range Potentials in Atomic Physics (in
Russian), Nanka, Moscow 1975; transl. Plenum Prss
[FttT] R. Figari, H. Holden, S. Teta, A law of large numbers and a central limit theorem for the
Schr6dinger operator with zero-range potentials, J. Stat. Phys. 51 (1988) 205-214
[FOT] R. Figari, E. Orlandi, S. Teta, The Laplacian in regions with many small obstacles, -
fluctuations around the limit operator, J. Stat. Phys. 41,465-487 (1985)
[HI Herczyfiski, J., On SchrSdinger operators with distributional potentials, Warsaw Preprint
[gi] W. Kitsch, contribution to "Lectures on SchrSdinger Operators", from the Nordic Summer
School in Mathematics 1988, Ed. A. Jensen, H. Holden, Lect. Notes Phys., Springer 1989
[KiMal W. Kirsch, F. Martinelli, Some results on the spectra of random SchrSdinger operators
and their applications to random point interactions, pp. 223-244 in "Stochastic Methods
in quantum theory and stochastic mechanics", Ed. S. Albeverio, Ph. Comte, M. Sirugue-
Collin, Lect. Notes Phys. 173, Springer Berlin (1982).
 99

[Ko] Koshmanenko, V.D., Singular perturbations defined by forms, BiBoS-Preprint, July 1988.
[Lo] Lovesey, Theory of Neutron Scattering from Condensed Matter. Vol. 1. Nuclear Scattering,
Vol. 2. Polarization Effects and Magnetic Scattering, Clarendon Press, Oxford 1984.
[Pa] B.S. Pavlov, The theory of extensions and explicitely soluble models, Rus.. Math. Surv.
42, 127- 168 (1987)
[Pan] Pantid, D., Stochastic calculus on the distorted Brownian motion, Belgrade Preprint (1986)
[Tel Teta, A., Quadratic forms for singular perturbations of the Laplacian, SISSA Preprint, in
preparation
 SCATTERING ON A RANDOM POINT POTENTIAL

B.S.Pavlov, A.E.Ryzhkov
Physical Institute, Leningrad State
University, Ulyancvskaya St. 1,
Petrcdvcretz, 198904 Leningrad, USSR

1. Introduction

The fact that the Bchr~dinger equaticn with a zero-range poten-

tial can lead to explicitly solvable mcdels was disccvered in 1934
when Fermi fcrmulated his famous deuteron model. 0nly thirty years
later, however, an exact mathematical meaning was given to the
Fermi heuristic pseudcpctential i n t h e w c r k cf Berezin and Faddeev
~I~. They showed that the ccrrespcnding Hamiltcnian is nothing else
then a self-adjcint extensicn cf a suitably defined symmetric opera-
tcr (see [2] fcr a recent review).
It has been shcwn by one cf the present authors [3] that the
structure cf the standard point interaction models can be enriched
substantially when the self-adjcint extensions are constructed in a
larger Hilbert space. This idea yields various mcdels cf zero-range
interacticn with an additicnal internal structure.
In our previous paper [4] this method has been used to con-
struct an explicitly sclvable model cf scattering cn a point object
whose internal structure depends cn a stcchastic prccess. In the
present paper, investigation cf the mcdel is ccntinued. We carry out
the spectral analysis of the averaged Hamiltonian which turns cut
tc be a dissipative operator. We show that its spectrum contains a
complex absclutely ccntinucus branch. We show, moreover, that the
mcdel is explicitly solvable. All spectral properties cf the ave-
raged Hamiltcnian can be derived from an algebraic "dispersicn"
equation. We construct also the averaged scattering matrix and in-
vestigate its prcperties in the limit when the velccity c£ the sto-
chastic process tends tc zerc. We compare the obtained expressions
to the "deterministic" S-matrices which ccrrespcnd tc the ~ixed sto-
chastic states cf the system.
 101

The paper is organized as follows. In Section 2 we construct

the Hamiltonian describing a Schr~dinger particle interacting with
a point-like object whose internal structure depends on a two-state
Markov process. We investigate also the evcluticn operator corres-
ponding to this stochastic Hamiltonian. Averaging this evolution
operator with respect to trajectories of the Markov process with
fixed initial and final states, we get a strongly continuous semi-
group. This semigrcup acts in the quantum-stochastic space
which is the tensor product cf the space L2(R 3) ~ C 2 of quantum
states and the space ~2 of states of the stochastic process under
consideration. Its generator ~ turns out to be a dissipative ope-
rator which is nearly normal in the sense that the commutator cf
resolvents corresponding to ~ and its adjcint operator ~ is a
flnite-rank operator only. This fact makes it pcssible to express
spectral properties of ~ m in explicit form.
In Secticn 3 we fcrmulate an eigenfuncticn-expansion thecrem
for the operator ~z . The thecrem can be proved using the expli-
A
cit formulae for the resclvent cf ~g~ which have been cbtained in
[4] • The ccncluding Section is devcted to construction and inves-
tigaticn of the averaged scattering cperatcr S which is cf the
form

s-lim exp(i~t) J exp(2i~at) J" exp(iAt) : ~ , (I)

t--~ +@@
where ~ is the unperturbed Hamiltcnian and J is the correspon-
ding identificaticn operator. We derive there the limit correspon-
ding tc "freezing"of the stochastic variable; the obtained exppres-
sion for limiting scattering amplitude is compared tc the "deter-
ministic" scattering amplitudes f+, f- ccrrespcnding to fixed
stochastic states

lim fm = 1/4 (f+ + f-).

~-~0
The factor I/4 appears because the contributicn corresponding to
the relaxation (complex) branch cf the absclulely continuous
spectrum cf ~ vanishes. This is a consequence cf the S-matrix
definition (1) and cf the fact that the ccntributicn cf the real
spectral branch tc the scattering amplitudes f+ and f- is always
non-zero. An alternative definition of the S-matrix (based on the
ccrrespcnding wave operators) gives (after freezing the stochastic
factor) a classical limit corresponding to the situation without
stcchasticity.
 102

Our model can describe, for instance, a situation when a

neutrcn beam is bombarding a massive nucleus localized in a magnetic
substance with a small dispersion. It was shown in ~5~ that for
high encugh temperature the magnetic momentum ~(t) of each domain
cf the magnetic substance conserves its modulus
i~ = oc~st
and suffers a Brownian mcticn cn a sphere of a constant radius. If
the substance is magnetized along one axis and the temperature is
not too high, this mcticn can be ccnsidered as a Markcv prccess
with twc states. We suppose that the nucleus has only two internal
states and that the correspcnaing Hamiltcnian can be expressed as
a sum of two terms:

H0 + ~ 4 ~ ' , ~(t)> ,

where ~ is the so-called gy~cmagnetic factor cf the nucleus,

is ~ e vectcr whcse ccmpcnents are Pauli matrices

Ro = d i a g { X O , '

where "~0'
~fl are the two pcssible "levels" cf nucleus.
The motion cf the magnetic momentumS(t) causes shifts of the
nuclear "levels" ~0' ~I' which become time dependent. In this
situaticn one can investigate the problem of the averaged (with
respect tc the n~gnetic momentum trajectcries) elastic scatterir~
on the nucleus. To solve this problem, we shall neglect frcm the
very beginning the direct interaction of the neutron beam with the
magnetic field. It is cbvious that this technical assumption should
be remcved in a more detailed analysis where the Laplacian must be
replaced by a Pauli cperatcr which includes the direct interaction
of the neutron magnetic momentum with the substance magnetic momen-
tum
Let us remark that the basic features of the proposed model
remain valid also in other stochastic scattering problems.
 103

2. Ccnstructicn cf the model

We start with the following quantum Hamiltcnian of the (neu-

trcn + nucleus)- system

where the Laplacian ( - ~ ) acts in the space L2(R 3) and the internal
nuclear Hamiltonian H 0 + ~ < ~ , ~(t)> acts in C 2 (recall that
H0 = diag{10' ~ I ~ ) " We describe the substance magnetic mcmentum
M(t) by a Markcv process with two states ~0M ) and (O 1 correspon-
..p

ding tc the magnetic mcmentum M directed up er down alcng the

z-axis, respectively. Suppcsing fcr simplicity that T = I, we get
the following expressicn of the internal Hamiltcnian

+- -9,
Fcr a fixed time t, the interaction between the external and
internal channels is ccnstructed using the same method as in [31.
We restrict the Laplacian (-A) : t (-AO) to a linear set of W 2-
smcoth functicns which equal zero in a neighbcurhccd cf the pcint
x = O. Constructing then the corresponding adjcint operator (-AS)
we find that its domain is given by

D(-~) = ~u(x) - 4~xl

u- +
u+ + u0;
Uo(X)£h(-~O) u + £C}. (2)

The bcundary form cf the operator (-/kS) is given by

<(-A6)u,v> - <u,(- a S ) v > = u+v - - u-v +, (3)

where v +, v- correspond to the deccmpositicn of the vector v£

D (- A~),

v(x) = ~ Iv- + v + + Vo(X) with Vo(X) E D(-/k0).

We restrict also the internal operator A(M) w-~A?(~) in such a

way that the generating vector ~ = ccnst.(~) turns cut to be
its deficiency element: A~(~)~ = iV. Intrcducing a "real basis"

w- = (A - iI)-IA~ , w + = (A - iI)-1"? , (4)

 104

we can write an arbitrary element frcm the domain cf A~ in the

form [3]

= + ~-w- + ÷, ~ = (a-il)-1~ ~ D ( ~ ) . (5)

ry fcrm cf the cperatcr A~ reads

<~,~> - <~,A~> = ~ +~-- ~-~+. (6)

The operator (- ~ 0 ) ~ A~ has deficiency indices (2,2). It has

a four-parameter family cf self-adjcint extensions which is para-
meterized by Lagrange planes cf the global boundary form (3) + (6).
In cur model, we chccse the extermicn ~ specified by the following
bc undary conditions

u- = a~,~+ , ~ = ~ u +, Im~ = O. (7)

The operator ~g obtained in such a way describes the Hamiltcnian cf

the (neutron + nucleus)-system. Since ~ = ~(t), this Hamiltonian is
time-dependent. The ccrrespcnding evcluticn operator U(t) restric-
ted to a fixed trajectory of the process M = ~(t) is a scluticn to
the equaticn

II ~~U = "~ (M'(t))U, U It=O -- lq = I e ~ I i ' (8)

where I e and I i are the identity operators in the external and

internal spaces, respectively.
Together with the stochastic evclution described by the equa-
ticn (8), we shall consider the "deterministic" evolutions corres-
ponding to the Hamiltcnians ~ (+M), ~ (-M) in which the magnetic
momentum is fixed in the up-state cr in the down-state. On the time
intervals where ~(t) is constant, the evolution equation (8) is
solved by the time-ordered exponentials [~] ccrrespcnding to the
cperators ~ (+M) and ~ (-M), respectively.
In order te calculate the averaged evcluticn cperatcr we must
introduce a measure on the trajectcries cf the stochastic prccess
~(t). The most simple way hcw tc do it is to use the equation for
the transition prcbabilities p = (p+(t), p-(t)), the solving matrix
cf which represents a scluticn cf the fcllcwing equaticn:

= ~. ,
o) • (9)
dt -I 0 I
 105

Knowing the probabilities p+(O), p-(O) cf the up- and down-

orientations of ~(0), we can calculate the corresponding probabili-
ties at an arbitrary instant

p-(t}/ \p-(t)/
The equation (9) is called equaticn cf stochastic evolution. It
allows to calculate the probability of the beam cf trajectcries cf
M(t) which are in the states ~ = ± M at the instants t = sA ,
s = O~1~...tn :
n

P~n%n_1 ...%0 =~{exp[~sC-~ _~)~l}~s~s_q "

s=1
In this way the measure on the space of trajectcries can be intro-
duced. After that, one can calculate the averaged evcluticn cperatcr
cn the set cf trajectories with fixed initial and final states using
the Trotter formula. It was carried cut in [~ cn a "physical level"
of rigcur. A complete prcof fcr the case of smccth interacticns can
be fcund in [6].
Let us ccnaider the quantum-stochastic space ~ , introduced
above, i.e., tensor product of L2(~ 3) @)C 2 = ~(.q with the space
of the stochastic states R 2, ~ = ~ q ~ q . Consider the operator
in this space which is represented by the following block matrix:

Furthermore, let us ccnsider the operator which acts cn the stochas-

tic variables as the matrix (-~ _ ~ ) . It is represented by the
blcck matrix:

The following linear combination

2~ = (+M) 0 + iN Iq -lq
o ~ (-M Zq lq

is the averaged Hamiltcnian as the following result shows

 106

Theorem [4]: The quantum evolution cperatcr averaged over the set
cf all trajectories cf the magnetic mcmentum starting in the sto-
chastic state ~ at t = 0 and ending=in the stochastic state
at t = T coincides with the element Oa~(T) of the operator matrix
which satisfies the following differential equation:

1 ~U = ~U, ~ t=O =
T ~t 0 Iq

Together with ~ we shall ccnsider the corresponding unper-

turbed cperatcr ~^ 0 the quantum part of which is simply the orthc-
gcnal sum cf cperatcrs, acting in the external and internal space,
respectively. The external and internal parts of this cperatcr are:

- =
o) + i ~
(_, _,) , (11)
0 -tk Ie Ie

=
(me) (lili) + i~. , (12)
0 A Ii Ii

where A u = A(+M), A d = A(-M).

^0
The unperturbed operator ~ = (-~) ~ ~ is normal and its
spectral characteristics can be calculated explicitly. For example,
the spectrum of this operator is the sum of the spectrum cf (-~)
whose spectrum is purely continuous and consists of two branches:
= k 2 and ~ = k 2 + 2i , Im k = O, and cf the spectrum of the
operator ~ which consists cf fcur simple eigenvalues:

A.~,IQ) = 2,.0,1 + i~ + _ ~2 . (13)

Operator ~^ differs from ~^ 0 by a finite-rank operator only,

and this fact allows
^
us tc calculate all important spectral charac-
teristics of ~ (see [z~]).
The discrete spectrum cf the operator ZSz ccnsists also of
four simple eigenvalues which equal for small ccupling parameter
to
 107

^
¢~2i F
0/1
(~.+
~-
~(~))
~'1(~) = X~'I(~) T ~2~T (U 2 -' ~2)II2 + O(K}), (1~)

where the functions FO,I(j~) are for ~ < < ~0,1 ' i.e., for small
velocity of the stochastic process equal to

FO,I(~) = ~_~- -M 2 + X 0 a£ - % ) -

- 1) - "~'0,1] + 0(~¢2) .

In Ref. [4] we have calculated the eigenfuncticns corresponding

to the discrete spectrum cf the operator ~ :

~ = VU

v~ ~
't,,) ~k
, k = 1,...,4,

where v u, v a ~ T.a(~3), ~lu, '~ C ~a a~a X.1,2=X~(#-,,),

A~(Z~). ~he
A

X3,4. = external part or the e i g e ~ u ~ c t i o ~ ~re equal

to:

~k(X) = I V u ~ (x) = I , ~ Coexp ( i~_~k,,x, ) (! ~ +

\va/~. ~ 8~I~I

÷ C~exp(i'~Ak-2i~"~x~) -I ' (15)

Co, ~ = const. #-(CoB ~ , i ~ B o) when ~ = ~ 1 , 2 '

CO, ~ = co,~t. ~ ( q B ~ * i~ B~) when ~ = X3,~ '

where we use the following notation:

A~ = 4 0 ± M, A I = 4 1 T M, B 0,1
+ = A ±0,1 - i,
(16)
o,1 = ~ ,1 + i~ -X, o,q = (i~ - X)A ,1 - 1 •

The absolute continuous spectrum (in the sense of [7]) of the

A

cperator ~ coincides with the absolute ccntinuous spectrum of

 108

unperturbed operator ~^0 , because ~ z differs frcm it by a fi-

A

nite-rank operator only. Mcreover, the operator ~ has nc eigen-

values Of infinite multiplicity.
The sc-called "scattered waves" play the role of eigenfunctions
of ~ . The eigenfuncticns in which the initial plane wave is
symmetric with respect to the stochastic variables ccrrespcnd ~o the
~+ branch cf the spectrum (we call it the stable branch). They are
cf the following form:

+ c) + •
The complex branch ~+ + 2i~ of the spectrum of ~ will be
called the relaxaticn branch. It corresponds to the eigenfuncticns
in which the initial plane wave is antisymmetric with respect tc
the stochastic variables:

~as(X,~,~[) = e x p ( - i ~ . < x , ~ > ) ( _ 11) + (18)

exp,i~..,xl) (~) +f c.~.~exp(i-~~.lxl) (_11)

f01 (A) #~ Ixl 11 "J~" zl.~' ix I

Let us remark that using the methods cf the extensicn theory in

L2(~ 3), we restrict cur~elves automatically tc the case when the
scattering takes place only in the s-channel because the singular
scluticns in the higher channels are nct square integrable. This is
why the quantities fi,k in (17), (18) dc nct depend oh the angular
variables. The explicit expressions I cr £i,k were obtained in [~].
For example,

~2
foo (-~-) -- --4-
~ I~k-
1
(-1)k+m.M-~
,L(Ak:) -- x (19)
k, m=OL

~k + (-1)k+m'M- i 2 i ~ ( ~ k - i) I x m
T'I(A ,~k)

where we use the following notation:

L ( A k) = ( A k - i) 2 _ M 2 ,
 10g

LI(A Ak) (A k + i~ k) 2 M 2 + a~ 2 (20)

zO:z,,
+
z~:z~, z~:~ ~ - ~ .

3. Eigenfunction-expans ion theorem

To formulate the eigenfuncticn-expaneicn theorem we have to

nctice first that the evolution cf special elements cf quantum-
stochastic space whose "internal" ccmpcnents are initially equal to
zero is traced cnly:

= fu ~ =
( ~ u ) ~u ~d = 0" (21)
fd Nd

The ~cllcwing asserticn is valid

A
Thecrem. Let a vector f ~ ~ have the form (21) and denote its
external part as

= =

Then the fcllcwing representaticn for the vector-function f(x)

holds almcst everywhere (in the Lebesgue measure sense):

f(x) =

+ (22)
+ 16~T 3 R+ 3 dy 2

The vector-valued function ~ n ( X ) in the rhs cf (22) is the exter-

nal part cf the eigenfunction ccrrespcnding to the n-th eigenvalue
of ~ (see (15)); the vectcr-valued ~n(X) is the eigenfunction
 110

corresponding to the n-th eigenvalue cf ~ (this functicn has a

structure analogcus to that of ~n(X) and Kn are some constants
which can be calculated explicitly. The functions ~s(x,~, k) and
~as(X,O,k) are the eigenfuncticns corresponding tc different^
branches of the continuous spectrum of the operator ~ z (see the
expressions (17) and (18), where ~ = k2 and ~ = k2 + 2i res-
pectively), and the functions ~s(y,~ ,-k), ~as(Y,~,-k) defined
by the expressions analogcus to (17), (18) are the eigenfunctions
A
corresponding to the ccntinuous spectrum of ~ •

The prcof of this theorem can be carried cut by the standard

method cf integrating the bilinear form of the resolvent of the
operatcr ~ around its spectrum. The present theorem will be
called the eigenfunction-expansicn theorem.
Using this result in combination with the Fourier method we
can construct the external part cf the averaged evcluticn cperatcr.
Introducing the nctaticn

[Pext f] (x)= [~]ext(X),

we obtain:

[Pext(U(t)f)l (x) = [Pext(exp(i ~ m t)fl(x) =

=~ Kn exp(i~nt)!3 dy I~n(X)~I~n(y)ITl~f(Y )

~f(y) + ~ +exp(i(~2+2i~)t)k2d~ 3~ ~as(X'0'k)

~

The expression (23) can be used tc compute the averaged scat-

tering operator.
 111

4. $cat terin~ operator

In the final part cf this paper, we shall describe ~he result

cf calculaticn of the averaged scattering operator for the mcdel
under ccnsideraticn. The averaging cf the quantum evcluticn leads
A
tc an evcluticn-cperator semigrcup with the generator ~ :
^
U(t) = exp(i ~ t ) , t>0 .

which acts in the quantum-stochastic space ~t~ . From the formula

(23) we see that the ccntributicn cf the relaxation branch tends tc
ze~c for t - - * ~ . Hence in scattering process cnly the ccntribu-
ticn cf the stable branch of spectrum of the operator ~ must be
considered. We chccse the restricticn of the unperturbed operator
( - ~ ) to the stable invariant subspace, corresponding to the real
branch ~+ cf the absclute continuous spectrum cf ( - ~ ) as a
ccmpariscn cperatcr. This leads to the following chcice cf the iden-
tification operator: J = P0' where P0 is the prcjectcr tc the sub-
space consisting of the functicns which are symmetric with respect
t o t he stochastic variables. This subspace is invariant under the un-
perturbed external cperatcr. In this way we eliminate the relaxation
branch cf the unperturbed-operator spectrum and the scattering mat-
rix lccks as follows
A A
~(og, - A 0 ) = s-lira exp(iA0t) J exp(2i ~ t ) J*
t-~+~
(24)
A

, exp(iA0t) .

This expreesicn i2 close to the S-matrix derived in [8]. The cnly

difference is that in the present case the evcluticn cn the comple-
mentary subspace of antisymmetric states is not unitary.
The restricticn of the operator ( - ~ ) to the symmetric sub-
space ccincides simply with the Laplaclan on two-component vectors

Of the type ~ = (~)with f E W~(R3). Hence we can write (-~)

instead of ( - ~ 0 ) in the formula (24) inserting at the same time
the projection operator into J, or equivalently, using the operator
~ which averages the stochastic states,
 112

Also we have tc insert in J the operation cf the dimension doub-

ling (~) cf the vector f:

: f~-~ f=

Taking into ccnsideraticn all these facts, we obtain the averaged

scattering operator L2(R 3) ~-~ L2(R 3) in the following form:
B

~(~,9~) = s-lim exp(i~t) ~. J U(2t) J" (~) exp(i&t) . (25)

t.-p+~o

Using the expression (23) for the averaged evcluticn operator,

we can obtain the following expression for the averaged scattering
matrix in the momentum representation:

~pl ^
[~(@,~)~](p) = ~(p) + • foo(p2)<u~(Jpl) . (26)
2SN

Here ^u = ~ u is the Fourier transform of the function from the

Schwartz class ~(~3), which is dense in L2(R3); ~ ( ~ , ~ ) denotes
the scattering matrix in the momentum representation and • u > is
the average of the function ~(p) = ~(~pl° ~ ) over angular variables,

~(Ipl) = I !2d u(Ipl.~) • (2?)

The kernel of the averaged S-matrix can be written as follows

i
(p,p,) = ~(p - p,) + ~ foo(p2) g(p2 - p, 2), (28)

cr equivalently,
^ i ~(Ip~ -Ip'i)
S(p,p') = ~(p - p') + ~ fo0(P 2) (29)
8~ ~ IPl
In conclusion, let us consider a very slow Marker stochastic
process when the parameter ~ tends to zero. Using the explicit
expression (19) for the value fo0(p2), we cbtain:
 113

~2
lira foo(P 2) = .{ a+(p 2) + a-(p2)} , (3o)
~,"~ 0 zl.

where

ik+(p2) : ~ (A k +_ (-1)k.M)p2 + I
~k + (-1)k" M - p2 " (31)
k=O

Thus, the acticn cf the averaged S-matrix on the function ~(p) is

in this case given by the following expressicn:

g(~, O) = u(p) + ' - ~+(p2) + ~-(p2) <~>(IPl). (52)

2~
We can see that the averaged scattering amplitude in this limit
ceincides with foo(P 2) and equals in the s-channel tc
c~. 2
(33)
~0 4

On the ether hand, we can calculate the "deterministic" S-mat-

rices S ± ccrresponding tc the fixed stochastic states, i.e., the
S-matrices correspcnding to the operators ~ ( + M ) , ~ ( - M ) , respec-
tively. They are given by

^ ^ ilP~ ~ 2
~s±ul(p) = ~(p) + . A±(p2)<~(ipl) . (~)
2~

The ccrrespcnding scattering amplitudes are nct trivial in the

s-channel enly and are equal to:

f± = ~ 2 A±(p2) . (35)

It means that fo is equal tc I/4 (f+ + f-) and not to I/2 (f+ +
f-). It is ccnnected with the fact that calculating the values f+
and f- we autcmatically take in$c account beth branches cf the
spectrum cf the unperturbed operator which ccincide for ~ = 0. In
the limit, hcwever, when t tends tc inliniSy for ~ 0 the con-
tributien frcm the ccmplex branch vanishes.
 114

References

1 F.A.Berezin, L.D.Faddeev, DAN SSSR, 1~7(1961), 1011-101# (in

Rue s Jan)
2 S.Albeverio, F.Gestesy, R.H~egh-Krchn, H.Hclden : Sclvable
~cdels in Quantum Mechanics, Springer, Berlin 1988
3 B.S.Pavlcv, Tecr. i Matem. Fizika (TNLF), 59(1986), 3#5-353 (in
Rue s ian)
# B.S.Pavlcv, A.E~Ryzhkcv, Prcblemy Matem. Fiziki, 1987, iss. 12,
54--82 (in Russian)
5 G.N.Belczerskij, K.A.~akarov, B.S.Pavlcv, Leningrad State Univ.
Vestnik, 1982, iss. 1, N #, 12-18 (in Russian)
6 S.E.Cheremshantsev : Some Generalizations cf the Fey~man-Kac
Fcrmula, Leningrad 1982, VINITI N 2195-82 Dep. (in Russian)
7 L.A.Sachnovich, DAN SSSR, !6_7_(1966), 760-763 (in Ruffian)
8 B.S.Pavlcv, Prcblemy M a t e m ~ i z i k i , 1982, iss. 10, 183-208 (in
R us s Jan)
 FADDEEV EQUATIONS FOR THREE COMPOSI~E PARTICLES

Yu~ A. Kuperin
Department of Physics, Leningrad State University
Leningrad 19890~, USSR

Abstract

A general formulation of the quautum scattering

theory for a system of few particles, which have an
internal structure, is given. Due to freezing out the
interaal degrees of freedom in the external channels
a cez~ain class of energy-dependent potentials is
generated. By means of potential theory modified
Faddeev equations are derived both in external and
imteraal chan~els. We prove the fredholmity of these
equations, what provides a sound basis for solving
the addressed scattering problem.

1. Introduction

The no,relativistic quautum mechanics for two and three composite

particles has been constructed in different forms and generalized
schematically for systems with an~ number of particles ~1 - 12S .
Nevertheless, up to now it is rather a heuristic scheme than a mathe-
matically well-defined physical theory, since it is not clear how the
basic objects of this formalism, i. e., the energy-dependent poten-
tials, cau be treated from the operator point of view. In fact, al-
ready in the two-body problem there is no self-adJoint ( s.a. )
Hamiltenian, which could be generated by a time-dependent unitary
group of operators. In the three-body case the original Faddeev equa-
 118

tions are also not directly applicable due to the absence of the
corresponding s.a. Hamiltonian. There are many other practical and
theoretical questions which should be answered before one can seri-
ously applay the formalism to real systems. The critical point of
all these questions is the operator interpretation of the theory:
how to include the energy-dependent potentials into the few-body
Hamiltonlans in a mathematically consistent way?
In order to overcome these difficulties we introduce a new class
of multichannel few-body scattering models. The channels appearing
in these models can be divided into two parts: the so-called exter-
nal channels describing the standard two - and three- particle scat-
tering, and some additional (internal) channels describing the in-
ternal structure of the particles. The Hamiltonians describing the
dynamics in the external channels are the usual two - and three -
particle SchrSdinger operators, while the internal-channel Hamilto-
nians are given by some abstract operators having only purely dis-
crete spectrum. The coupling of the e ~ e r n a l and internal channels
is in our model realized in the framework of the s.a. extensions
theory [13 - 15] leading in such a way automatically to a s.a. Hamil-
ionian of the global system.
In order to introduce the energy-dependent potentials we proceed
further as follows: we exclude the internal channels from the glo-
bal Hilber~ space obtaining in such a way the modified Faddeev equa-
tions for the external-channel components of the global Green's
function ~16, 17] . We prove that these equations are of the Fredholm
type. The energy-dependent potentials appear naturally when projec-
ting the global Hamiltonian on t h e subspace of the external channels.
The Fredholm property of the corresponding modified Faddeev equations
justifies in some sense the applications to the three-partlcle scat-
t ering problem.

2. Two-body problem

In this section, we construct a s.a. Hamiltonian for two particles

possessing a nontrivial internal structure, restricting ourselves to
the two-body processes only. We construct also in the external chan-
nel the corresponding energy-dependent potential and investigate its
properties.
 119

The physical content of the model is fixed by the following

assumptions:

(i) In the reaction Xi + Yi ~ Cif , Xf + Yf , where the

indices i , f denote the initial and final channels, respectively,
and belong to an index set ~ , the internal structure of the
colliding objects is manifested only at relative distances of the
order of the characteristic radius of the particles and not in the
asymptotic states.

We separate also the two-body configuration space ~3 ( with the

centre of mass removed ) into two domains V- , such that ~3 =
= V - U V + • Let V- be the part of the space ~3 where the rela- 1
rive distance between the particles is bounded: r i < b V ~ ( Vol V-)~
~ . Physically the compact domain V- can be interpreteted as
the reaction domain, and the complement V + = R3\ V- as the region,
where the particles are "asymptotical~_y free". Let us denote as
the common boundary of the domains V •

We assume that there exist an interaction in the channel i ~

such that a phase transition occurs at the surface ~ which gives
rise to the formation of a compound state Oil with a finite llfe-
time. At the same time, at r < ~ two types of d~namics are
possible: 1) in Cif there is an admixture of the "cluster dyna-
mics" of the asymptotic region V + ; 2) with a certain probability
there is formed in Cif a new state u In possessing additional deg-
rees of freedom not contained in the asymptotic region V + .

(ii) The compound state Cif decays into the outgoing chan~el
f E ~ , whose energy is equal to the energy of the incoming channel.

From the mathematical point of view, the proposed model is based,

as we have already mentioned, on the theory of ex~ensions of symmet-
ric operators. In the two-body case, it is assumed that the global
dynamics, which takes into account the interaction of the internal
and external degrees of freedom, is specified by a s.a. operator h
of a special structure. Namely, it acts on the space ~ = ~ e x ~ ~ i n
ex
where ~ is the Hilbert state space describing the motion of
of the particles with the internal degrees of freedom neglected, and
~in is the Hilbert space of the states corresponding to the in-
dependent dynamics in the internal degrees of freedom. The method of
 120

constructing the operator h in the space ~ is the following.

Suppose that s.a. operators h ex and A act on the spaces g~ex
and ~ i n respectively. Then the orthogonal sum h e x ~ A describes
the independent dynamics in the external and internal degrees of fre-
edom. An interaction between these dynamics can be "switched in" as
follows. We restrict the operators h ex on ~ e x and A on ~ in
to symmetric ones h~ x and A 0 and construct all s.a. extensions
of the operator h~ x ~ A 0 on the space 9~ . Then each s.a. ex-
tension h can be interpretated as a total Hamiltonian defining
the coupled dynamics. The nature of the interaction between the ex-
ternal and internal degrees of freedom is regulated both by the me-
thod of restricting the operators h ex and A and by the choice
of the extension scheme.

We emphasize that in the proposed model the pairs hex , ~ex

and A , can have quite different structures; in particular,
we ma~ take for hex , g~ex a differential operator and an infi-
nite- dimensional function space,respectively, but for ~ i n , A
a finite - dimensional space and any s.a. matrix on it. This "fre-
edom" has a lot of advantages, but it leads to a difficulty associ-
ated with the fact that the domain of the restricted operator A o
is not dense. However, this difficulty can be overcome, and moreover,
as it is shown below, one can match the internal and external Hamil-
tonians by means of boundary conditions. The advantage is that one
can describe explicitly the S matrix in this w~y and study in de-
tail its analytical prope~ies. It can be shown that the S matrix
inherits the spectral properties and characteristics of the operator
A only ( in the case of a finite - dimensional matrix A this
me~n~ the set of its eigenvalues ). Finally, the fact that the ope-
rators h ex and A can be of a different kind makes it possible to
model situations in which the dynamics in ~ e x and ~ i n can differ
substantially. An example of such a situation is the hadron-hadron
scattering in the bag model [4 - 6] . In hadron-hadron collisions the
external dynamics at low and intermediate energies is nonrela~Ivistic
and can be described by a SchrSdinger equation, whereas within the
compound quark bag the dynamics is assumed ~o be essentially rela-
tivistic and given by an or~hogonal sum of single - particle Dirac
operators.
 121

~, Hamiltonians of the External and Internal Channels.

We follow here the general construction described in L13-17~ •

Let us assume that the dynamics of the external channel is given by
~he s,a. Hamiltonlan h ex defined as follows

h ex u : - (~ + V(X)) u

in the Hilbert space g~ex = L2 (~3) . The potential v(x) repre-

sents the so-called peripheral interaction ( e.g., a meson - exchan-
ge potential ) of strongly interacting particles and will be assumed
to decrease rapidly as Ixl --~ oo and to be sufficiently smooth.

In our model we shall restrict the s.a. Hamiltonian h ex to the

symmetric operator h o with the domain D( h o ) = C o ( R3\~ ), where
oo
C O (R3\ ~ ) is the class of infinitely differentiable functions,
which vanish together with all derivatives in the neighbourhood of
the surface ~ . The adjoint operator h o has a nontrivial boun-
dary form jex( .,. ) , namely

Je c . , f > : < hu,f> - <u, > : (1)

= lira ( I" dS ( ~ . V u ~ - un-V~)-

S aa ( n . ~ ' u f - un.g f )),

+ +
where n. V + i s the normal derivative on the surfaces
dist (x,~-)--.~ ].

Now we asuume that t h e dynamics o £ the internal degrees of freedom

wlthou~ a couplin 6 ~o the external ch~n~el ~ e x is given by an ar-
bitraz~j s.a. operator A acting in some Hilber~ space ~ i n . The
important question of the model is the following: how to pa~ametrlze
the boundary form jin(. ,.) for an arbitrary symmetric operator A o ?
 122

The general answer was obtained in Ref. [13~ • For ~he symmetric
operator A o with a non-dause domain D(Ao) this question was also
solved by B. S. Parlor (see his contribution in these proceediugs).
Following the general construction, the symmetric restriction A o
of the s.a. Hamiltonian A should be performed in our scheme via
its Ca~ley transform U = ( A - il) (A + il) -1 . Let us consider the
special isometric restriction U 0 = U ~ ( U* e )~, where ~ is a
generating elemaut of the operator A . The symmetric restriction
A o can be obtained as the inverse Cayley transform of the iso-
merry U o . Hence the operator A o has deficiency indices (1,1) and
the domain D(A~) of its adjoimt can be described in terms of
yon Neumamm theory : D(Ao ) = D(i o) + ~ ( ~ , U * @ ). Here i o is
the closure of the A o and ~ := ~ ( 0 , U ~ @ ) is the span of
deficiency elements @ and U * ~ . Let us emphasize that though
yon Neumsmm formulas cannot be directly used in the case of a men-
densely defined operator A o the description they give for the
domain of A o remains valid.

i~ is convenient to introduce some new basis in ~ : w + =

=(I/2)( U*@ + ~ ), w- =~1/2i)( U * @ - @ ). In accordance with the
first yon N e u m ~ n ' s theorem an arbitrary element u E D(A~) can be
decomposed as

u= ~+ ~+ w+ + ~ - w - , u ~ D ( i o) ,
+
where ~- (u)+ are the so-called boundary values of the element u .
In term~ of ~ , the boundary form of the operator A o reads
. - +
A o w = ~ w , and it can be written as the symplectlc form in the
boundary-value space [13, 15] :

Jin( ul,f 1) : <A; u1, 1> - <u 1, A; = (2)

= ~'(=1). ~+(~1)- ~+(ul) . ~-(~1) , u I, ~'1 e ~ i n .

Afber we have constructed the boundary forms jex,~n(.,.) , the

nex~ step is to construct a s.a. extension h of the operator
h o ~ A o , acting in the or~hogonal sum ~ex~ ~In • In accor-
dance with our general method one should find such boundary conditi-
ons for which the total boundary form vanishes, i.e., for which
 123

jex( e ) .) + jin( e ) e ) =O • It car be show~ that such conditions cau

be written , for instauce, as

[ n . v uo] ~ s-CuI) ~ , (3)

8*(uI) = <u°'~> == S dS uo-~ . (#)

Here u ~ , ~ = O,q , are the external and internal channel elements,

respectively. We use the notation [n.V u~y := n . ~ u- - n.~ u + to
denote the Jump of a function n - V u across ~ . Here the super-
scripts -, + denote the limits taken from V- aud V+,respecti -
rely. The function ~ E L2(~ ) is the parameter of the model.

The investigation of the wave functions ~ = ( Uo, uq) is

based on the two-chonnel SchrSdimger equation

( h- z ) ~ = O ) x E ~3\~ (5)

with the boundary conditions (3), (4) . In order to obtain the

energy-depemdent potential we consider further omly the external
channel ~ e x • For~this purpose we have to exclude the internal
boun~o~y values ~ from (3) and (@) . This can be done using the
the following relation [q3] :

s" = A ( z ) ~+, (6)

where (z) is the Schwartz integral of the spectral measure

d ~ E A ~ , ~ > of the s.a. operator A ,

(z) := 4 ( 1 + z A) (A - zl) -1 ~ ,~ = (7)

= ÷ c t -. d o>.

Taking into account the relation (6), we obtain from (3), (4) the
following energy-dependent boundary conditions in the external
space ~ex :

[n.VUo] Y = - A(=) < % , ~ > ~ . (s)

 124

In accordance with (5) the extermal component u o obeys the e q u a -

tlon

( ho -z ) u0 = 0 , x6 ~'\y (9)

In order to obtain the differential equation in the external con-

figuration space ~3 it ks convenlent to use a quaslpotential appro-
ach ( see, e.g., [19, 203 ). Let us consider the quaslpotautial
V(z) acting on the function u as

W(z) u : = - ~ ~(z) < u, ~ > ~ . (lO)

Here g~ ~t is the distribution, usually called the simple layer [21],

that acts on the set of sufficiently smooth functiom~ f in the
following way

<~,f>.= ds~f .
The boundaz-j-value problem ~n the ex~emaal space ~ ex can be writ-
ten in ~erms of quasipotential ~(z) as

( h ex+v(z) -s ) uo = 0 , (11)

where the variable x now runs over the whole configuration spa-
ce ~3 . One can state that (9) is equivalent to the boundary -
value problem (8), (9).

We conclude this section with the following ren~Lrks:

1. The operator h , which is the total Hamiltonlau in the

two-bo~y system, is a self-adJolnt operator and hence the scattering
for the pair ( h, h ex ~ A ) can be investigate~ in a mathematically
correct way. it should also be noted that in ou~ model we are able
to simulate an arbitrary complicated internal StrUcture of parblcles
due to the general nature of the internal self-adJolnt operator A .
The physical nature of the Hamiltonlan A can be interpreted in
different wa~s. In the non-relativistlc framework it might be, for
example, a few-bod~y Schr6di~er operator with co,fining potenti-
als [18, 19] .
 125

2. As it follows from (7) and (10) the energy dependence o~ the

potentials cannot be arbitrary. It is given by the Schwartz inte-
gral ~ (z) , which is real on the real axis and is an analytic
function in the upper half-pla~e Im z > 0 with the positive ima-
ginary par~ Im A (z) > 0 . It can be shown that such interactions
ensure the anal~icity and ~m!tarity of the correspond~g scattering
matrix [1@, 16] .

3. In the model, described above, the quasipotential W~(z) is

separable and of rank one. The generalization to an arbitrary rank
of W(z) is trivial. For this purpose one should increase the
dimension of the deficiency subspaces ~ = { e ) , ~ * = { U*@ } and
change in a self-consistent way the functionals ~ , ~. , ~ > by
arbitrary finite-rank operators B , B .

4. The Three-bod~ Problem.

We consider in this section a system of three panicles having

a nontrivial internal structure. A total s.a. Hamiltonian H gover-
ning the dynamics of external end internal degrees of freedom is an
most impo~ant object in the three-body analysis.

In order to describe the kinematics of the system we use the

usual Jacobi coordinates x~ , y~ , ~ = 1,2,3, which we combine into
the six-component vector X = {x~ , y~} belonging to the external
co~iguration space ~6 ( with the centre of mass removed ).

Let ~ = ~ X ~ be cylinders in ~6 and C = U ~ • The star-

ring point of the method is the description of the two-bod~ Hamil-
ionian

in the three-body configuration space ~6. Here hg is the s.a. two-

bod~ Hamiltonign defined in the previous section, ly and Ig are
the identity operators in the spaces L2( ) and ~ = ~ e x + ~ ,
 126

respectively.~ and - ~ y ~ is the Laplacian defined on its natural

domain ~ ( ~ ) . The operator H~ is essentially self-ad~oint on the
domain

,° ®

The closure H~ of the operator H~ is the s.a. operator, which will

be danoted by the same symbol H~ .

The domain D(H~) can be also described in terms of boundary con-

ditions. Namely, let ~ -- ( Uo, u~ ) E D(H~). Then the external
component u o is a W 2 - smooth function outside of F~ . The in-
ternal component u ~ e ~ u := ~ i n ~ L 2 ( ~ ) can be decomposed
into the sum

u~ .. ~u + &(y~)
÷ ~ .,. 6~.(y~)w;. , ~u~, D ( _~_o ) , (lZl.)
+_
where w are the deficiency elements of the symmetric operator
A~o , which is the restriction of the s.a. operator A ~ , and

in
H~:0 := A~o(~ "l'y + I~(~) ( - A y ) . (15)

Let us note that the formula (14) is a simple consequence of the

special structure of the space ~ n e

The functions ~ ~ D(H~) satisfy the boundary conditions

It should be noted that the boundary conditions (16), (17) have a

two-body character (see the previous section ). The only difference
is that the $~(y~) are now functions of the variable Y~ E ~ •

We are now ready to construct the total three-bod~ Hamiltonian

H . Let us consider the space
3
 127

a~d s ~ e t r i c operator H0 in ~ ,

~o ~I == { (-~x

Hin
~o u
+ Z c~ v ( x ) )

, ~ = 1,2,3,
uo ,
(19)

with the domain

O0

D(HO ) .= Co (~ek r ) • 7_ ® D(~o) (20)

Any s.a. ex~emsion H of the operator H 0 is a total three-body

Hamiltonian describing the whole d~namics in both extermal and in-
ternal channels, lu accordance with the yon Neumanm theory all such
extensions cam be obtained by the extension of the operator H 0 on
its deficiency subspaces. So we shall extend the domain D(Ho) to
a linear set D(H) in the following way

% = ~'o + Z ~o(-'1).,% , u"o ~ % ( ~ % r ) (2-,)

D ~( ) = +

and impose on the D (9) the boumdary oomzlitloms (16), (17). Here
R0(z) = (H ex - z) "1 is the resolvent of the s.a. operator H ex =
= - A X + ~ v~(x~) and j o are densities of the simple-layer
potentials givem on the cylinders r~ , ~ = 1,2,3. The correspon-
N

di~g extension of the operator H 0 on the domain D (H) with the

boundary conditions (16), (17) will be denoted H . luterms of the
demslties jo these conditions cam be written as

3
f~(y~)=<~=~2:%(-~)~o, ~ > (&). (2.

Let us now state some important facts about the operator H .

Theorem I. The domain D (H) is dense in the space ~ and

the operator H given by H 0 ~ D (H) is symmetric and bounded from
below.
 128

The proof of this statemaut will be given elsewhere.

The last step is now to construct a self-adjoint extension H

of the symmetric operator ~ obeying the following conditions:

1. The translation-invariant boundary conditions (16), (17)

hold on D(H).

2. The Hamiltonian H is bounded from below.

For this purpose we shall c h o o s e the Friedrichs extension [23]

H of the operator H . On the domain D(H) which can be des-
cribed as usual [23] , the action of H is given by

Hex
uo t

(25)
H~ =I - ~ y u + A u~ - S~w~+
+ 8,o, w&
- +

= ( Uo, U~) , c~= 1,2,3,

with the boundary conditions (16), (17).

3. Resolvemt Equations.

This section deals with the Fredholm-type equations for the re-
solvent R(z) of the s.a. Hamiltonian H . As in the case of the
energy-independent interactions [22, 2@] , these equations represent
a stax~ing point for the three-bod~ scattering problem.

First we shall derive the differential equations for the resol-

vent components Rab(Z) corresponding to the decomposition of
into the sum (18),

~(z) _-~Rab(Z) } , ,,, b = 0 , 1 , 2 , 3 . (26)

 129

Here the indices a, b sta~d for the external ( a,b = O ) amd

internal ( a,b = 1,213 ) subspaces ~ 0 = L2(~6) a~d % ~ n , ~ = 1,2,3,
respectively.

Because R(z) is the resolvent of the s.a. operator H it satis-

fies the usual rslations

Rab(Z ) = ~ a ( Z ) ; a,b = 0,I,2,3. (27)

We shall introduce the following notations: Let F be an arbi-

trary element of ~ and ~ = R(z) F , i.e., F = ( fo,fl, f2' f3 )
and

3
Ua : >-" ~ab ( z ) fb " (28)
b:O

Then due to (25) and (28) one gets

± 3 +
~ : ZI ~jb(z) 'fb , ~ : ~,2,3, (29)
b:O

~here ~'~ + are the operators which mapping ~ n into L 2 ~ ) for

b = 0 , - -and~
~- ~ i n into L 2 ( ~~) for b ~ O. vThe
v
relation (29)
can be considered as the definitiem of these operators.

Let R~b(Z) denote the operators

R~b fb = ( ~ b - '< " ~ b - ";' ~;'b ) ~b - oo)

Then using the idautity

( H- z ) ~(z) F = F

one can obtain a ~et of equations for the kernels of the operators

( H ex - z ) Rob(z ) = 80b I 0 , (31)

 130

A~'R~,b- w~'~:b + w~
+ '~'b - (~y + z ) R~b =S~.b I ~ ' (32)

with the following boundary conditions

n.V ~ob]r~ = - (~, ~b, (33)

~+b' = ~" ROb'' ~ ~> ~' r= (34)

The differentlal equations (31), (32) for the exCe~al ROb and
internal R~b components of the resolvent R(z) serve as a star-
ting point for construction of the Faddeev equatioms.

We shall rewrite the conditions (33) and (34) in terms of the

interaal Hamiltonians A~ . For this purpose we use the relation

~b.= Q~(z) ~ b ' + ~b<(A - il) (H~

i~ - z) -I. , ~ > (35)

which can be obtained by arguments analogous to the two-body case

[13, 153 • Here

in = A~@ I~@ ( - A (36)

and Q~(z) is the generalization of the Schwartz integral in the

three-body co~Lfiguratiom space

% (z) = ~( ~ + ( ~y~÷, ~ A~ (HT- "~-~ e~, e~ > (3~)

I~ accoz~la~ce with (15) this operator is an integral operator
with the kez~el

Q ~ ( y ~ - y ~ / , z) = 2 ~ i1 ~ dt .~~ (t) r~ (y~ - y~ ' z -t).

L~ (38)

Here to(Z) = ( - ~ ~ - z) -I is the resolvent of the Laplacian,

~ (t) is the t w o - b o g Schwartz imtegral, and th# co,tour L~
emcircles the spectrum of A~ .

The operators ~b cam now be excluded from (33) and ( ~ ) by

 131

virtue of the relation (35) :

[ n.v%] • =-~.
r=
(%(,) < "oh', ~=> * (39)

÷S~b < CA -iI) (Ein-z) -1 • , Oo~ ).

If the internal-channel Hamiltonians A~ have the discrete spectra

z
~p(A~) = { ~ s } only, thau the kernels Q ~ ( y ~ - y~ , z) should be
written in the form

% (y, - yj, --) = E (~ ÷ (~)2) < ~ e . , e.> (40)

s

× ro(y~-y~, z- A ) ,

where ~s are the spectral projectors of the operators Ao~ .

Notice that internal Hamiltonians are used for describing the

internal channels, e.g., with quark confinement.

6. The External-Chan~el Faddeev Equations.

The study of the total resolvent R(z) ca~ be reduced to construe

tlng the external-chan~el component Roo(Z) only. In order to see
this, the eqs. (31) - (37) should be used. Let the component Roo(Z)
be given. From (3@) we can get ~ o ( Z ) which we use for substitution
into (35) to ge~ ~jO(Z) . Then from (32) and from ~he definition
(3o) of the ~b(z), lm z ~ O, i.e.,

R~b = R~b + w+ ~+b + w~ ~ b '

one can obtain R~o(Z ) . It gives the components R~o(Z), ~ = S,2,3.

Then the components RO~ (m) , ~ = 1,2,3, cam be found from (27).
Thus we shall now treat further om the component Roo(Z ) only; for
the simplicity, it will be denoted by G(z) . It should be noted that
G(z) is the so-called Krein's resolvent [25, 15B and it has the
corresponding propez~ies.
 132

/
In view of eq. (39) the kernel G( X, X, z) of the quasiresolvent
G(z) obeys the boundary conditions

r~

As in the two-body case these conditions can be written i~ terms

of quasipotentials

W~ (z)~ := ~ V~ (z) ~ , (42)

where V ~ (z) ks the integral operator in L2(~) with the kernel

v (x,x,z) = - ~ ( x ) Q~(y~ ° y~' , z) ~(x~) (~3)

In accordance with (31), (@I) and (@2) we obtain the following

equation

3
(H ex + ~- W~(Z) - z) G(x,x,z)
I I
= 8 (x - x). (~)

To derive an integral equation for Krein's quasiresolvent one can

use the usual procedure. Namely, applying the operator Ro(z ) =
= (H ex - z) -I to (g4) we obtain the resolvent identity for G(z) :

3
G(z) = Ro(Z) - Ro(Z) Z v~(z) G(z). (45)

From this equation which is of the Lipp_m~n-Schwinger type the ope-

rator G(z) can be expressed explicitly in terms of generalized
operators

M~(z) =v~(z) s(z) (;%6)

by the relation

G(z) = Ro(Z) - ~o(Z) }- ~(z) (47)

Hence we have reduced the problem of investigating the quasiresolvent

G(z) to study of the operators M (z) [ 20] .
 133

The next problem is to derive the Faddeev equations from eq. (@7).
Applying the operators V~(z) to (@7) one can write this equation
in the form

(z + v= ~o) ~= = v= ~o - v~ ~o Z m (48)

~ollowing Faddeev's method we have to inver~ the operator I + V ~ R O.

This inversion can be performed explicitly in terms of the two-body
operator G~(z) = ( H ~ - z) -I which is the resolvent of the s.a.
operator H ~ . The following formula can be easily verified

(I ÷ v~ ~o) ~= % = ~= ~o (~9)

This relation yields in a straightforward way the equations

m(z) = w'~ Q=,(z) - v = Q~(z) ~_ m (z) (50)

which have the structure of the Faddeev equations.

Nevertheless to emsure that these equations are actually Faddeev

equations, one must prove the following statement:

Theorem 2. Let ~ b e the densities of the simple-layer poten-

tials, ~ ( z ) := ~ F ~ ( z ) and ~ = ( ~I' ~2' ~3 ) . Then:

I. Equations (50) rewritten in terms of densities ~w,

(z) = 6o(Z) + ~(z) # (z) , (51)

are of Fredholm type and Bn , n > Nma x , with a sufficiently large

Nma x , is a compact operator in an appropriate Banach space.

2. Equations (50) or (51) are spectral-equivalent to the origi-

nal SchrSdlnger equation with the s.a. Hamiltonlan H .

The proof of the first statement proceeds in a standard way [ 2 2 ] .

 134

The second statement of the theorem is much more subtle in con-

trast to the case of energy-independent interactions. In particular,
we must show that the homogeneous equations

(z) = B(z) ~(z) (52)

have a nontrivial solution, iff z 6 ~p(H), where ~ p(H) is the

discrete spectrum of the s.a. operator H .

Let ~ be the solution of the homogeneous equations (52) re-

written in the form

~(z) = -v~ (z) G= (z) ~ ~#(z) (53)

Consider the function

u o : Ro( ) >-- V# ,

which is evidently the simple-layer potential given on the hyper-

surface F = U ~ and hence it satifies the equation

(H ex - z) Uo(X ) = O , X ~ F • (54)

In order to find the appropriate boundary conditions one must apply

the operator I + V ~ R O to eq. (53). Taking into account (49) and
the properties [21 ] of the simple-layer potential : [ n.V u O ] =
= - ~ , we find the boundary conditions

[n.Vuo] C v (z) uo . (55)

Iterating the eq. (53) one finds that u 0 E ~2(~6\ r ) at Imz

0 as well at z = E + iO, E 6 R, and furthermore, that
. 2,R3~
WUo , ~ > ~ w2~ yJ •

Now we shall express the internal functions u~ in terms of the

external component u O . To this end one must take into account the
representation u ~ in the form (22) and the relations (23), (24) as
well as (35), which express the relation between 8 6 and u 0 :

£~
÷ : <Uo , , (56)
 135

÷
~: = Q~ (~) ~ (97)

The functions ~a may be then found as solutions of the equations

+
(-ny+A.-z) u = e~(Y~) w: - e:(Y~)w+~ + (581

By virtue of equation (56) the functions 86 • W ( ) , and

hence 8~ ~ W2~(~3 ) . This means that ~I = ( u 0 , ua) belongs
to D(H) and, in accordance with (5~) - (58) , ~ is an eigau-
vector of the s.a. operator H ,

( H - z )~ = 0 . (59)

This equation implies that ~ = 0 if z ~ p ( H ) and hence u O = 0

in this case. In other words we have proven that eq. (51) has a
unique solution, if z ~ ~p(H) .

On the contrary let ~ be an eigenvector of the Hamiltonian H .

Then one must repeat the derivation of (50) for densities ~ =

= - V~ u 0 , which obey the equations (53).

Hence we have proven that the Faddeev equations (53) are spectral-
equivalent to the SchrGdimger equation (59).

Consequently, the Fredholm alternative can be applied to (Sq) and

the properties of densities ~ can be investigated. Knowing these
properties we can stud~v the behaviour of the resolvent R(z) by
(@7) • The wave functions are determined too. Their asymptotic form
cam be investigated and their completeness in the total space
established using the methods of Ref. [ 222 .
 136

~. Discussion.

In this paper we have presented a new approach towards a mathe-

matically correct study of the scattering theory for few-body sys-
tems with energy-dependent potentials. The main result is that
treating of such systems in usual configuration space ks inconsis-
tent from an operator point of view. We have demonstrated that the
energy dependence of the potentials is generated by the internal
structure of the interacting particles. This very dependence,
however, turns out not be arbitrary, since it is given by some
class of operator-valued R-f,,nctions, including in particular
Schwartz integrals as described above.

The main effect incorporated in our scheme is the possibility to

separate the contributions from two-body and three-bo~yforces.
From the geometrical point of view the three-bod~ forces are con-
nected with boundary conditions, which m a y b e stated on the m a ~ o l d
F O = ~ ~ " The deficiency subspaces of the operator H 0 ,
corresponding to the manifold r 0 , are parametrized by simple-
layer densities belonging to the Sobolev class W2-3/2 . In order
to conserve the pair character of the boundary conditions (16), (17)
we do not include such deficiency elements into the domain D (H) .
But it is clear that three-body forces can be included into our
consideration without a drastic change of the formulation.

Acknowledgments.

I would like to thank m~ collegues K. A. Makarov, S. P. Merkuriev,

A. K. Motovilov and B. S. Parlor, ~ho collaborated with me on the
problems discussed in this paper. I am also indebted to L. D. Faddeev
for encouragement and fruitful discussions and to P. Exner and
v
P. Seba for some essential remarks, which improved the content of
the m~nuscript.

References

I. Feshbach ~. Ann. Phys. ~, 357(1958); ibid. I_9, 287(1962).

2. Schmid E.W. Prec. of the Intern. Sympos. on-~ew Particle
Problems in Nuclear Physics, Dubna, 1979, p.17@.
3. Wildermuth K., Tang Y.C. A Unified Theory of Nucleus,
N.Y., Academic Press , 1977.
 137

4. Hill D.A., Wheeler J.A. Phys. Rev. 89, 1102(1953).

5. Alt E.O., Grassberger P., Sandhas W . ~ u c l . Phys. B2,
167(1967).
6. Jaffe R.L., Low F.E. Phys. Rev. D19, 2105(1979).
7. Simomov Yu.A. Phys. Left. BS07, I(1981).
8. Narodetskii I.M. : Few-Bod~ Problems in Physics ( Ed. by
Faddeev L.D. and Kopaleishvili T.I. ), Singapore, World
Scientific, 1985.
9- Vauzani V., Cattapan G. Phys. Rev. C19, 1168(1979).
10. McKellar B.H.J., McKay C.M. Austr. ~ . Phys. 36, 607(1983).
11. Orlowski M. Helv. AcSa Phys. 56, 1053(1983).
12. McTavish J.P. J. Phys. GS, 10AV(1982).
13. Pavlov B.a. Toot. Mat. ~ z . 59, 3@5(198@).
1@. Adamjan V.M., Pavlov B.a. Zap. Nauch. Semln. LOMI,
149, 7(1986) •
15. l~aglov B.S. Uspekhi Mat. Nauk @2, 99(1987).
16. Kuperin Yu.A., Makarov K.A., Pav~ov B.a. Teor. Mat. Fiz.
63, 78(1985) ~ ibid. 69, 100(1986).
17. Kuperin Yu.A., Makarov K.A., Merkuriev S.P., PavAov B.S.,
Motovilov A.K. Preprlat Budapest Univ. ~P-Budapest-Repoz~
No ~J+1, Budapest, 1986.
18. Kuperin Yu.A., Kvitsinsky A.A., Merkuriev S.P., Yarevsky E.
Proc. VIII I~ter. Semln. on High Energy Problems, Dubna,1987
19. Kvitsinsky A.A., Kuperin Yu.A., Merkuriev S.P., Motovilov A. i
Yakovlev S.L. Elem. Par~. and Atom. Nucl. 17, 267(1986) (in
Russian).
20. Merkuriev S.P., Motovilov A.K. Left. Math. Phys. 7,
@97(1983) ; Preprint lust. des Sciences Nucl@aires de
Grenoble ISN 83.41, 1983.
21. Vladimi~ov V.S. Distributions in Mathematical Physics,
Moscow, Nauka, 1979 (in Russian).
22. Merkuriev S.P., Faddeev L.D. Quautum Scattering Theory
for Few-Bod~ Systems, Moscow, Nauka, 1985 (in Russian).
23. Birmam M.S., Solomjak M.Z. The Spectral Theory of Selfadjo-
int Operators in the Hilber~ Space, Leningrad, Leningrad
Unlv. Press, 1980 (in Russian).
2@. Faddeev L.D. Trudy Matem. In-ta AN SSSR 6~ ,1 (1963).
25. Krein M.G. Doklady AN SSSR 52, 657(19@6) 7-
 ON THE POINT INTERACTION OF THREE PARTICLES

R.A.Minlos
Moscow State University, Moscow, Lenin's Hill

The problem of point interaction of three indistinguishab-

le particles (Bosons) has been studied by several authors K.A. Ter-
Martirosjan, G.V.Skornyakov [I] , G.S.Danilov [2~, R.A.Minlos and
L.D.Faddeev [3] , [ 4 ] •

The Hamiltonian of the three-particle system was treated

in [3] ,[4~ by means of the theory of self-adjoint extensions of
symmetric operators see, for example,J5] . Such an approach to the
study of point interactions of two particles has been proposed
first by F.A.Berezin and L.D.Faddeev [6]. In this note we consider
the general case of point interaction of three distinguishable
particles.

The Hamiltonian of such a system should be according to [ ~

defined as some self-adjoint extension of a symmetric operator

H = -(I/2m1~A Xl -(I/2m2)Ax2 -(I/2m3~x3 (I)

acting in the space ~ = L~(R3) 3~ of the wave functions~(x I ,x2,x3)

xie R 3 with the domain

where q j C [R3) 3 are the h~perplanes

~ij --[(xl,x2,'9~03) 3 : xi --xj, I_. i~ j.~3~ (3>

The domain of the operator H ~ consists of the functions
~/(x I ,x2 ,x 3) which

(a) are smooth outside an~ ~ij

(b) in a neighbourhood of Pij behave asymptoticall~ ss

~CXl ,x2,x3)Z Aij~) t + sijC ~)

~xi_xj , xl,x j ~x (4)
 139

where ~ I
= (x I ,x2,x3) xj=xi=x g ~ij and Ai~ , Bi~ are functions

on Uij in general distributions are allowed; an exact description

in terms of Fourier transforms will be given below . It is natural
to consider the following symmetric extensions of the operator H.
Let H ~ E _
ij ~ H~ be the contraction of the operator H * to the
domain

O H~- = ~ DH~ : Bi~ = £i~Aij, I~ i~ ~ 3 ~ (5)

where E = ~li~ ~ are real constants. The operators H7 are adjoints

to some symmetric extensions H£ of the operator H. These symmetric
extensions of H - we call them Ter-Martirosian-Skornyakov extensions
are natural from man~ points of view. In particular, they are direct
analogues of the Hamiltonians of two-particle point interaction [6].

As we shall see below, however, the Ter-Martirosian-Skor-

n~akov extensions are not self-adjoint: the operator H£ have non-
zero deficiency spaces and apparently all their self-adjoint exten-
sions (translation-invariant) are unbounded from below. This pheno-
menon has been discovered first by the author of [2 3 and [3] for
indistinguishable particles. In the case, however, when only two
particles are identical fermions and the third particle is different
from them, the operator n Z(a) ,CI ,2) = HE I ~,(1,2)
a turns out to
a
be self-sd~oint and below bounded ~here L2,(1,2 ) is the space of wave
functions antisymmetric in the variables Xl,X 2 ) . This result has
been established by the present author together with M.Shirmatov
[8] . In the present note, we give a detailed formulation of the
results together with a concise derivation.

Passing to the Fourier transform of the wave functions

,~(._Xl,X2,X3 ) > ~f(pl,p2,p3 ) , pi E R3

and introducing the total momentum
P = Pl + P2 + P3
we rewrite the operator H to the form

H : (il2M) P2 + h (6)
It acts on the tensor product ~(R3~ ® h ( P ,dv) where Pc(R3) 3
 140
is the manifold
=[(kl,k2,k3)~.(R3)3 : k I + k2 + k 3 = O~
equipped with the natural measure dv.

The quantity M in (6) is defined as M = m1+m2+m3 . The self-

adjoint operator (I/2M)P2 acts in ~ ( R B) while the symmetric
operator h acts in ~ ( r I ,dv) as
(hf)(k I ,k2,k 3 ~ =(41/2m1) kl2 +(I/2m2) ~ +41/2m3) k~) f(k I ,k2 ,k3)
(7)
Its domain is of the form
Dh = f~ L ~ ,dr) : hf~L2(fl ,dv), f~k 1,k2,k3]u(ki)dv = 0,i=1,2,
C8)
where the function u(k) fulfils the condition

2 + i>i/2
3 •(k ufk) 2 d3k ~ ~O (9)
The domain Dh~ of the ad~oint operator h~ consists of functions of
the following form (see [3])

f(k) + Z Uj,k(ki ) + (10)

2mi 2-m3 k~ + I

+ Z v,k(ki)
(2~ I k~ + ~ k~ + 12-~3k~" + I) z

where f ~ D h and uj, k , vj, k belong to the class (9) and h* acts
according to the formula

~h~g)(k1'k2'k3) = I ~m I k2 + 2 ~ k2 + 2-m3
I k~g(k 1,~,k 3) -

- Z Uj,k(ki ) " Q11)

The element g~ Dh~ determined in the representation (10) by a
rapidly decreasing function uj, k has the following asymptotic
behaviour
 141

C
g(k 1,k2,k3> dv i'j = 4~i,~ui,~s>N + Bi,~Cs ) + 0(I) , N--~

where ^ ^
i,j =
is a 3-dimensional submanifold of ~ (equipped with the measure dv s
end Bi, ~ belongs to the class (9~. The aforementioned extension
H£ of the operator H is obtained by an extension h E of h the
ad~oint hE of which is defined on the domain
Bi, j = 4 ~ , ~ C i , j ui,ji I ~ i ~ jE3 (13)

We shall now compute the resolvent (h~ - mE) -I of the

operator h£ . Setting f = (h~ - zE)-Ig we find from (11) that

f(s) = g(s) + Z Ui,~(Sk) ~ S = (Sl,S2,S3)6~ (14)

I
--
2m I
Sl2 + 1
~_~2 s2 + 2"~3 s~
I - z

According to ~12)-(14) the functions ui,~, I~ i~ ~ 3 solve the

following system of equations

, mlm2m3 z ,j(p) -

I
~'~I s12 + ~ s2 + $3
I s~ - z

= ~,j~p,z) ~/ I g(s~ ~ dvi,

jp
.t~i~c ~ S~ + I - Z
Let T be the symme operator in L23 = I ~ ( R 3 ) ~ LI~R3) ~ L2~R3)

I[M/~ j3 p2 ÷ ~') ui,jCp)

(T u)i,jCp) = 2~2(2~i,j > 3/2(vm~sm
_[ ui,k(so~ ÷ Uj,k(Si} dvi,J
P
3 ~3 s~
+I

defined on finite functions .

 142

Using the same arguments as in [3], we find that self-

adjoint extensions of h~ are obtained from extensions of the opera-
tor T and that the deficiency indices of h E and T are equal.

Theorem I
The operator T on L2(3)(and consequently, the operator h£ on L ~ ,dv))
has non-zero, finite and mutually equal deficiency indices.

Proof.: The operator T commutes with the operators Fg, g ~ O 3 of the

rotation group representations in L2(3)
Cvgu)i,j(p) : g o3, 0v)
and is reduced by every space Lp(3~I) whose elements are triples
of the functions

u~,~p) = ~Z ul'mz'j,kt'Pl)'\Yl(~)~)m (.18)

Here I P I , @ , ~ are the spherical coordinates of p, y l ( ~ , ~ ) are

the spherical funcions of the weight i [see [7])and u i,
l'mj are
functions of f ~ l
Let T I be the part of T acting in the space L2(3~I) . We
shall consider first the case i = 0 i.e. the spherically symmetric
ui,j(p). By the calculation similar to those of K3~ and [4"] , it is
possible to check that the domain of (TI)~ , i=0 consists of triples
l ui,j~ of the form
ui,j~p) = Z c ' qI,U
' IP~S~ + ~ui, j(p ) (19)
2
p + I
where ui, ~ are functions from the domain of the closure TI= 0 and
cq,j solves the following homogeneous system of equations

~mIMmym~)-I/2 Sq c ~ Ic j -W6t~,j[W66i,
~mj/L~k, j ) iSq/2 s jSq)C ,j -
mi f~k.i )iSq/2

where Sq are real zeros of the determinant of this system

 143

D s = mlm2m~M s3ch - Mmlmym3

E(m1+m3~2(m2+m3) 2 sh2(~3s) + (m1+m2~2(m3+m2~ 2 s h ~ 2 s ~ +

2 ~ m i+m~) 2 ~ sh~is)~ .~21)

+ (my+el) 2(m3+ml )2 s h ~ i S ~ _ ~Mmlm2m3)3/2 ,

The angles ~ , %' W3 in (20) are defined by the relations

/ m m, V/2
sin T i = !((mj+mi~mk
~ K +mi~/A , 0 < ~i ~ i~)" . (22)
%.

It is easy to check that the functions s-3D(s) is even, negative-

valued at s = 0 and positive for Isl>>1. Thus there are at least
two real zeros of the function D(s) and the total number of its
zeros is finite. It means that the operator Tl= 0 has non-zero,
finite and mutually equal deficiency indices. The operators T1,
1 = 1,2,... can he treated in a similar way and moreover it is
possible to prove that they are essentially self-adjoint for large
enough 1. Hence the theorem is proved.

Remark:
The self-adjoint extensions of h£ may be unbounded from below
similarly as it was proved for identical particles bosons in [3]
and K4] . In order to prove such an assertion one has to know how
the coefficients C~,j in the representation (19) for the solution of
a system of equations similar to 0 5) behave assymptotically as
z---~(such a system arises when we are looking for the eigenfun-
ctions of the operator h£ with negative eigenvalues z)°The unboun-
dedness from below of the extensions of h E is an unpleasant property,
since only below-bounded Hamiltonians have a physical meaning. In
the particular case, however, when two particles are identical
fermions and the third one is of a different kind, the operator
h~ is self-adjoint and bounded from below. More ~xactly, let
LCa)
2,(I ,2)eL2( ~ ,dr) be the space of functions on D which are anti-
symmetric in the variables k 1,k 2 , i . e . ,~(k 1,k 2,k3~= - ~ k 2,k I,k3~ •
If m1=m 2 and g12 = 0, ~ 1 3 = E23 = ~ then this space is invariant
with respect to the operator h£ . Let ~ , ( 1 , 2 ) be the restriction
of h£ to ~2,(1,2) ' then the following theorem is valid:
 144

Theorem 2:
The operator h £,(1,2)
(a) on L~a~
2,~,2) is self-ad~oint, bounded from
below and its discrete spectrum is negative and finite.

/' h~,(I
The operator ~ (..a) , 2 h/~) on L2,
( s'~(I ,2~ is defined by the
formula

In order to write down the resolvent ~h.(a)r. ~ - zE~ -I of h(a}.. ^~

~ , ~1 ,L) ~,\19~ )
one has t o f i n d a s o l u t i o n t o t h e f o l l o w i n g i n t e g r a l e q u a t i o n f o r
the function u which is similar to ~15)

÷ m3 p2
V - m ( m + m3 ) - z u(p) +

) %) ÷ '

=ZCp,z)) m = m ~ =½ .

The corresponding homogeneous equation has not solutions of the form

( 1 9 ~ , which implies that ---~/h~a}1,2b is self-adjoint and bounded
from below. A detailed study of the quation (24) is needed to prove
the finiteness of the discrete spectrum.

The last result suggests the following interestig problem:

what kind of assumptions about statictics can ensure existence of
a self-adjoint and below-bounded operator of the form 4 3 for a
general system of n particles with a point interaction? In the case
n = 4 for example, there are the following possibilities:

(a) three identical fermions interacting with a fourth particle of

a different kind
(b) two different couples of identical fermions interacting with
each other
~c) a couple of identical fermions interacting with a couple of
bosons.

In conclusion, let us remark that the above described

scheme worked out for scalar particles can be extended to the case
of particles with spin.
 145

References
I. K.A.Ter-~rtirosian, G.V.Skornyakov: Sov.Phs, s.JETP 31 1956
775
2. G.S.Danilov: Soy.Phys. JETP 40 (1961) 498
3. R.A.Minlos, L.D.Faddeev: DokI.Acad.Sci.USSR 141 ~1961) 335
4. R.A.Minlos, L.D.Faddeev: Soy.Phys. JETP 41 [1961) 1850
5. M.Reed, B.Simon: Methods of Modern Mathematical Physics, vol.II
Academic Press, New York, 1975
6. F.A.Berezin, L.D.Faddeev: Dokl.Acad.Sci. USSR 137 ~961) 1011
7. I.M.Gelgand, R.A.Minlos, Z.Ya.Shapiro: Representations of the
rotation group and Lorentz group. Fizmatgiz, Moscow 1959
~in Russian)
8. R.A.Minlos: in " Quantum Field Theory and Quantum Statistics.
Essays in Honor of the 60-th Birthday of E.S.Fradkin, vol.I,
Adam Hilger, Bristol, 1987, p. 393
 A RESONATING- GROUP MODEL WITH EXTENDED CHANNEL SPACES

Yu.A.Kuperin, K.A.Makarov, Yu.B.Melnikov

Physical Institute, Leningrad State University,
I Maya 100, Petrodvoretz Leningrad 198904

A new version of the resonating-group model with exten-

ded relative-motion space which takes into account the
effect of additional two-body resonant channels is for-
mulated using the theory of self-adjoint extensions. It
is shown that after projecting on the original relative-
motion space we get an effective non-local energy-depen-
dent interaction which describes Pauli repulsion for
small intercluster distances. We propose also a genera-
lization of this schema for three-cluster systems.

I. Introduction

I~ this lecture, we propose a new method for exclusion of forbid-

den states in cluster scattering [I-61 based on the abstract spect-
ral deformation scheme of non-relativistic Hamiltonians [7-12 S. In
distinction to the standard resonating-group model (RGM), the exclu-
sion of Pauli,forbidden states is achieved not by truncation of the
relative-motion space, but using an extension of the standard RGM
sp ce RGM = ~ ~n Rare in are the spaces
referring to the internal-cluster degrees of freedom, ~ 0 is the
relative-motion space and A is the antisymmetrization operator. The
extended space A ~ i n @ ~ ~n@ [~0 ~ ~ I ] I includes the dy-
namics of compound resonances, which can arise when the intercluster
distance is small enough to allow overlapping of clusters. This dy-
namics is described by an additional Hilbert space .~I. Such exten-
sion of the space ~ R G M makes it possible to construct a total
self-adjoint Cs.a.) operator which describes the common dynamics of
all degrees of freedom of the system starting from the Hamiltonians
acting in the channels 5~ h b = 0 I The construction is based on
 147

the theor~ of s.a. extensions, and enables to obtain an effective

description of the class of effective energy-dependent interactions
veff(~), arising in 4 0 after the exclusion of the resonant channel
~I. As we show, the special choice of spectral characteristics of
the operator acting in 4 1 leads to the exclusion of one or several
forbidden states from the spectrum of the RGM Hamiltonian. Thus the
energy-dependent potential veff~z) imitates Pauli principle in ~0.
In distinction to the "operators" of the standard scheme EI-6S, the
new "Hamiltonian" with the additional effective potential veff(z)
obtained ~, our method acts in the complete (non-truncated) relative-
motion space.

The paper is organized as follows. Sec.2 is devoted to construc-

tion of the new Hamiltonian which imitates Pauli principle in two-
cluster systems. An illustration of the exclusion of one Pauli-for-
bidden state in such a system is presented in Sec.3. We demonstrate
also the connection between the presence of a linear term in the po-
tential veff(z) as a function of energy and the asymptotical "swit-
ching off" of Pauli principle in the limit of high energies z ~ in
cluster scattering. We propose also a generalization of this scheme
for three-cluster systems (See.4) and study the spectrum and the re-
solvent of the corresponding Hamiltonians (Sec.5). Faddeev equations
for such systems are discussed in Sec.6. Finally, Sec.7 contains so-
me conclusions; in particular, the possibility of exclusion of three-
body forbidden states is discussed.

2. The Hamiltonian for a two-cluster system

Let us consider binar processes of cluster scattering and bound

states in such systems. D~namics of the interior-cluster degrees of
in
freedom is defined by s.a. operators H a , a = I, 2, while the dyna-
mics of relative motion of clusters by a s.a. operator H 0 acting in
4 0 • We suppose that the operators H ain acting in Hilbert spaces
in have a purely discrete spectrum. A possible occurance of reso-
nant compound states shall be imitated with the help of an additio-
nal Hilbert space ~I in which a s.a. operator H I acts. If there is
no connection to the external channel ~ 0 the operator H I generates
the dynamics of resonant degrees of freedom. Let us consider the s.a.
 148

operator H 0 ~ H I in the space ,,.~0 ~ ~{I, which gives the indepen-

dent dynamics in spaces ~ 0 j ~ I In order to "swith in"' an inte-
faction between the channels ~ 0 and ~ I in accordance with gene-
ral scheme [7-12], one must restrict the s.a. operators H 0, H I to
symmetric ones H00' H~ and then construct all s.a. extensions of the
operator H~ ~ H~ in the space ~ 0 ~ ~j~1 • Each s.a. extension can
be interpreted as a total Hamiltonian of the whole system including
the resonant degrees of freedom.

Realization of this scheme includes the following steps:

I. Let s.a. operators H b, b = 0, I are defined on domains D (Hb)c
C ~b and let linear envelope ~ b = ~ @ k = nI be generating
subspaces of the discrete parts of these operators ( e ~ a-e the
corresponding generating elements). We restrict H b to the symmetric
operators H~:

Choosing the basis [W±b,k} k : nl -

W~,k = ~(Ub Okh + e bk.) and

with

in ~near envelope ~ { ~ b ' ~ b "I( where ~ b - U %*~ b ) ° " e ca.

describe the ~omain D ( ~ ) o f the ad~oint operator H0" b~ the formula
n

~b ~ D (H0b )
}(2)
The domain D (H0b)which appears in (I) is not dense in ~ b , and the-
refore the operator H b~ does not exist in general. One can overcome
u b~
thls difficult~ [ 7,8] by defining the action of H 0 on the elements
+
w~, k as

HOb Wb,k
+ = ~ Wb, k" (3)

Hence

Hob = HbID (g0b) + ( - i I [ ~ b + iII~)r~lgZb' ~I (4)

 149

2. The. self-adjoint extensions H o f the o p e r a t o r ~ • H~ in the

space ~ O ~ ~ I are specified ~ fixing the boundary conditions
(71 which ~nsure vanishing of the b@undar3 form ~ H t ~ , ?f) -
-,<~, H~> •

(5)
~o(~0) ~* a2 -~o(U0)
Here H0 = H + H~ w i t h ~ = ~ u b,~b=0'
] ~ : I Vb}b:~, and f u r t h e r m o r e
Bj. = B~, j = 1, 2, and B are n x n - m a t r i c e s , which play the r o l e o f
parameters of the model. Vectors ~ have the components ~b,~ •

Using the connection between the vectors ~,

n
b
b,s

and the relation (4), one can write the operator H defined b~ the

H =
Z~
z)
boundar~ conditions (5) in the matrix form

H VI
(7)

Here H b = PbHPb , where Pb are the projections on ~ b in. ~ = q40 ~)

~ 1 . V l n accordance with (5)-(7), the blocks of the Hamiltonian H
are :
n
~o ~o÷ ~o ~o ~o <. ~o> ~
V - ks ~ 9
n k,$=4

~ - , k',s=4
~o ~- ~ < ~>~o,
~,S:4

where j0,, j1, ~ are the following matrices

 150

Y : - a % * (I+K I % ) -~ .

The matrices Kb, b=0,1, are defined as follows

( K h ) i j = < H b Sb, ~ i b).

Let a matrix ~b. connect the vectors ~±

b :
(9,)
Using this relation one can exclude the resonant channel ~ I and ob-
tain from C5)-(9) closed formulas for PO H :
n
Po.: .o + veff(z) : .o <., 0o)
k,s=~
where the matrix D O has the following form

(DO)ij = a O ij + < . 0 8i0 ' e 0j > . 01 )

On the eigensubspaces ( H - z ) ~ = 0 the matrix elements ~ b ij(z)

depend on total energy in the c.m. frame and are equal to the
Schwartz integrals of the spectral measure d < ~ ~, ~ jb > of the
s.a. operator H b (see [7]):

~(z) = <(I ÷ zHb)(H b - z) -I 0~, o jb> :

= A- ~ j "

The external ~ 0 and resonant ~ I channel are coupled together by

the formula

A° o B*~' (% a' I)'B ~ (,3)

which iS a simple consequence of (5), (9).

We have found that our scheme of adding the resonant channel ~ I

leads automatically to existence of an additional energy-dependent
potential veff(z) = - Z ( D 0 1 ) k s < ' , ~0s > ~ 0 in the channel 4 0 .
This separable non-local interaction imitates antisymmetrization ef-
fects and, as we show below, leads to exclusion of the Pauli-forbid-
den states from the channel ~ 0 . The parameters (D01)k s (ie. mat-
fix elements A ij
I (z)) are determined by the positions of compound re-
sonances and can be fitted from the scattering data.
 151

The resolvent R(z) of the total Hamiltonian H has the matrix form
~Rab(Z)l, where Rab~ ~ b - - ~ 4 ~ a , and can be reconstructed fram its
"external" block G ( z ) -- RO0 (z) a l o n e [ 1 0 , 1 1 3 . S i n c e t h e e f f e c t i v e po-
tential veff(z) is separable, this block can be easily calculated:
n

ocz) : Oo(Z + ( ol)ks @-1(Z)) sq < % (z) • ,

k,S--4 ¢1: 4
where Go(Z) : (H 0 - z) -I and the matrix N(z) is defined by its mat-
rix elements

: g j- ° i .

In accordance with (9), the discrete spectrum of the operator H

defined by the boundary conditions (5) can be obtained as the set of
the solutions to tile dispersion equation

BI_ ( ~ I ( z ) ) - 1 B
= o (15)
det & O (z)
B B2+

where ~b~z), b=0,1, are to be calculated from (12).

3. Two-cluster s2sSem: an illustration

Analyzing the equation (15), we find that choosing parameters of

the model in a proper way, one can exclude effectivel2 a finite num-
ber of forbidden states. We are going to demonstrate here this phe-
nomenon in the simplest situation with one forbidden state, using
n = I, B = ~ e C and B I = B 2 = 0. In this case, the equation (15)
turns into

A0(z) : - ~-I • I (z) (I 6)

where ~ : I~I 2 is a real parameter playin~ the role of a coupling

constant of the channels ~ O and ~ I .

The above described scheme of coupling the operators H 0 and H I

employs Schwartz integrals of the form (12)as main parameters. They
 152

do not contain linear energy-dependent terms. However, the Riesz-

Herglotz theorem allows to extend the class of Schwartz integrals
to functions of the following type :

~b(z) = 11 X-+ Am
z d2b(A) + C~z + C~ , C bI ~ 0, C ~ E R (17)
m
Henceforth we shall use this more general representation for ~I (z)
assuming C I0 = C20 = 0 if b = 0 . The role of the linear energy-de-
pendent terms was discussed in Refs. 9, 13.

Let thus H I be a one-dimensional operator so that

A1(z) = ( 1 + z z l ) ( z , _ z ) - 1 < ~ 1 ~ 1 , 81> + CIzl + C21

and let -~0 be a forbidden state in the channel"°'~f0 with the energy
z0' (H0 - z^)
.u ~u = 0. Let us choose
r I further a resonance level z I in
channel ~ ! which equals to z0: z = z0. Then (16) turns into

ZS - Z
$
+ I + zQz < ~ I 01, 01>I
z0 - z
where 6~d(H 0) : {Zsl is the discrete spectrum of H 0. We pick a ge-
nerating element ~0 which differs slightly" from ~0 in the following
sense :

< e O ~0' 00 > = ; s ; I - ~ 0 ~<I and f s ~< I for s/0. (19)

Assuming then that the coupling constant o( is a small parameter

(the weak-coupling limit), we find that the relative shift between
the spectra ~Zsl and I Zsl of the operators H and H 0 p respectively
equals in the first-order of the perturbation theory to

Zs '-s s "" @ ( $S ( 1 + Zs
2o)

The external components U0s of the corresponding eigenfunctions

Us = ubs b% a,e

U0s = G0(.zls~ (p0 = ~ zq~q~'s- ~q (21)

 153

where ~ q are the eigenfunctlons of the operator H 0, (H 0 - Zq)~q=0.

z'_ \ z-1 I \ z

Figure I: Qualitative solution to the equation (18). 0

The solid curve represents the function ~ (z),
the dashed curve does the same for the func-
tion _ ~-I ~1(z ) which has the asymptotics
_~-Iclz;while ~ i m ~ lira ~ 0(z) = - Z ms's"

An analysis of the dispersion equation (18), illustrated on Fig.l,

leads to the following conclusions :
I. The energy of the forbidden state z 0 E ~ d ( H 0) dissapears from
the discrete spectrum of H ; all the remaining eigenvalues z ~ (0H) -
are shifted slight~ in accordance with (20). Moreover a new eigen-
value z+~>0 appears in the continuous spectrum of H such that value
zJ--~c~ as the coupling constant ~ - ~ 0 .
4

2. If At(z) contains a linear energy-dependent term, C I ~ 0, the

spectrum of the total Hamiltonian H is amended furthermore by an ad-
ditional level z ~ < 0 which is called sometimes a phantom [13]. On
the other hand, the equations (10), 01)and (17) exhibit the asymp-
totic "switching off" the antisymmetrizstion effects in the high-en-
ergy limit, i.e. for C I # 0 and z - ~ t h e effective potential V eff (z)
vanishes. These facts concerning the role of the linear energy-dep-
 154

endent term in the function A I ~z) can be summarised in the alterna-

tive: either the spectrum of the total Hamiltonian H contains phan-
toms at large negative energies or the S-matrix (.phase shifts) exhi-
bits an anomalous behaviour at high energies related to the fact
that veff(z) 7 ~ 0 as z - ~ .

4. The Hamiltonian for a three-cluster system

One of the most essential questions in the RGM scheme with additi-
onal resonance channels is how to include a two-body energy-depend-
ent effective potential into the three-body Hilbert space. The star-
ting point is to study a three-cluster system with a resonance in a
fixed pair ~ o n l y $). One should know the form of the pair potential
generated by such an additional resonance channel in this pair. The
Hamiltonian ~ for the three-cluster system with the resonance in-
teraction in the pair ~ only can be obtained [I~] from the pair Ha-
miltonian

H Z~)
H~ = v• ~ (22)
Z~ Hv

constructed in Sec.2 (see eq. (7)) b~ the formula

Here T~ is a s.a. operator, acting in a Hilbert space ~ w h i c h des-

cribes an effective non-resonant interaction of the third cluster
with the subsystem~ ~), IT~ and I~ are the identity operators in
the spaces ~ and ~ = ~ e~l, respeetlvely, where the spaces
~ (b;0,1; ~ fixed) coincide with the spaces ~ b of Sec.2. The ope-
rator ~ is e.s.a, in the space = ~ ~which can be repre-
sented as a direct sum ~= ~ ) ~, where ~ = ~ ® ~.

~) It should not be confused with the coupling constant ~= ~21 2

of the previous section.
x~)In the most simple situation T~ is the kinetic-energy opera-
tor.
 155

The elements ~..~" ~ . have the components {ual , a=O,~(, u0 , c ~ ,

~/'w" =
(H0
u ~ 6 ~ 1 and the operator ~

V~

'~
can be written in the matrix form :

IT~ + I , 0 ~ T~

"~
Z~

+ Ie(,1@ T~
) (24)
z~ v ® IT,

where I ,h are the identity operators in ~ . This relation can be

rewritten further as follows
A
,~.., = ,v..(, 0 + V~ ,

~,0 = I H0 @ ITs0 + I~'0~ T~ o ) (26)

0

Z
A I V~ ~ IT~

Hv ~9 IT~ + I~, I ~ T~ /
where V~ coincides with the operator V of Sec.2 (see the first one
A
of the relations (8)). The operator V~ is interpreted as the pair
potential, generated by the resonant channel ~ i in the three-clus-
ter system.

Our aim is now to construct the total Hamiltonian of the three-

cluster system with additional resonant channels in each pair sub-
system. To this purpose we make the following simplifying assumption
Suppose that there exist unitary identification operators IS'~ :
9[~ ~ :0 ~ t ~ ~). It allows us to drop in the following the index__
from the notation of the quantities associated with the space 4 0
L/
When the operators ~'~ exist the three-cluster Hamiltonian ~ can be
defined in the natural way :

The operator ~ is e.s.a, in the space

i) In particular, if H 0 = - ~x~+ ~3-~, y~) and T~ = - B ~ (where

x~,y~ are the Jacobi coordinates in~&), then the operators 7Ea~
exist. J
 156

Inthe ~ representation , the Hamiltonian ~ is given by the ope-

rator-valuedU m a t r i x ~ a b l a , b = 0 3 with the elements ~ 0 0 = H 0 @ I T +

+ I0,0 ~ T, ~ = Hv
~ ~T~ ÷ I~, I ~T~, ~0~ = Z~, ~ 0 = Z~, ~ , ~ =
=I, 2, 3; remaining elements equal zero. The last mentioned fact is
a consequence of the maximal simplicity of the chosen s.a. extension,

~I 711 o o B1o ~T
~; = o ~o ~2o ~ (29)
~ 0 o B33 B30 ~
~ Bmo B2o B3o Boo - ;
which corresponds to such a structure of the operator-valued matrix
~ . This choice supposes existence of a coupling between each reso-
nance channel and the exterior channel, while the resonance channels
are uncoupled mutually. This excludes existence of three-body forces
generated by an interior structure of the clusters.

5. Spectrum of the three-bod2 Hamiltonian

The spectrum of the Hamiltonian ~ can be formally obtained from

the dispersion equation

Detreg~(z) : o (30)
in terms of the s.a. extension parameters Bab (29) " Here~ Det~reg de-
notes the regularized operator determinant,B= [Bab(Z)~a,b=J0 is 4x4-

operator-valued matrix with the components ~@(= B~- Q-1 (z), ~ 0 =

: Bg0' ~O~= B~0' ~ =I, 2, 3; 300 = BOO + Q0(z) while all the re-
maining ~ a b = 0. Furthermore, Qa(Z), a : 0~ I, 2, 3, are n~n-opera-
tor-valued matrices with the elements

C~(",))~ = < ( ~ ~ ~ + " ~ ( ~ ~ - u))C" ~ ~ ÷

+ l a ~ ~Ta - Z i T ~ ) - 1 ~ , ~ > (31)

where H a are s.a. operators in ~ 1 when a= @( or in ~ 0 when a=0 and

$ a
i are their generating elements.
 157

A standard way to understand meaning of the dispersion equation

(30) is based on the study of the structure and analytical properti-
es of the resolvent R(z) = ( ~ - z) -I of the total three-body Hamil-
tonian ~ . The components Rab(z)of the 4x4-operator-valued matrix
R(z) satisfy the equations [10,11]
3

(~00 - z)R0b. (z') + Z R rZa,

b: : ~0bl0 . (32)

Z~*R Ob(z) + (~a.~- z) R~b = ~<bIa, o<= 1 ' 2, 3 " (33)

As it was shown in Refs.10,11, all the components Rab{Z ) can be re-

constructed from the component R00 (z) alone, and therefore the study
of ~he total resolvent R (z) can be reduced to the study of the "ex-
ternal" component R00 (z) only. The equation (33) for b=0 yields

~-~0 ('~ } = - ( "~-,.C z) -1 Z~Roo(Z)

* • (34)

Using then the equations (34) and (32), we obtain closed equations
for s (z)~%0 [z):
(H°@ ZT +Z0,0 ® ~ + Z W~Cz)- z)G(-) = 0
where the energy-dependent potentials are

(36)
Elere I 0 o and In are the identity operators in 3 0 a n d % 0 =
= N 0 @'~respectively. The operators Wa(z)can be expressed in terms
of two-body quantities, namely the resolvents (Hv - z) -I and
[Ta - z~ -I . Due to the special structure of the operator ~ , the
equations (36) turn into

W~(z) = 2 ~ ~ V~I~)(T~ -~+z) -1 d ~ (37)

r~
where the contour ~ encircles the spectrum of ~ . The relation (37)
hasbeen obtained using the identity

C~...~ z~t ~ ~(Hv~'-~I(T~-~+ z)-I d ~ (38)

r~
 158

6. Faddeev equations

The equation (36) can he rewritten in the following integral form

A A
G(z) = R(z) - R(a) I W~(z)G(z) , (39)

A
where R <z)=(~/00 - ~)-1. This aquation has a structure of Lipp, ann-
Schwinger equation and is not of the Fredholm t~pe. Hence it should
be^rearranged. Let us define the Faddeev components Ga(z)= ~ AI R _
- RW~(z)G(m), ~ = I, 2, 3. According to the relation (39), G = ~ G ~
holds which allows to obtain Faddeev equations for the operator~
G~(z) using the standard procedure [14]:

o~(z~ = # ~ , ~ (z) - %,(z~w~) ~ o~(z) (40)

where G~ (z) = ( ~ A - z)-1 is the resolvent of the operator (25).

One must check that the equation (40) is a Fredholm one. The proof
does not differ from the standard one [ 14 ] and can be found in
Refs.10,11.

7. Conclusions

The system of the equations (40)can serve as a mathematically cor-

rect base for studying the bound states and scattering in the three-
cluster system with exclusion of two-body Pauli-forbidden states in
the two-body subsystems. This exclusion is connected with the possi-
ble occurence of two-body compound resonances and differs from the
schemes proposed earlier which restricted the relatlve-motion space
[~-51.

The relations (32), (33) provide us with a solid justification for

the equation (40) from the operator point of view which can be hard-
ly obtained in the models with restricted relative-motion space.

One should remark also, that (40) is valid in 8 limited region of

energies only, i.e. below the break-up threshold of individual clus-
ter, because it does not take into account the reaction channels.
 159

T h e equations (40) can not be used also for exclusion of three-

body forbidden states, which can exist in the system. However, the-
se states can be easily excluded by adding an additional orthogonal
channel~to the space ~ ,which imitates the possible occurence of
three-bodyV resonances. TheV operators ~ , ~ , acting in the spaces
~ and % , c a n be coupled also by the extension-theory methods in 8
way similar to the one described above.

Acknowledgements

We are grateful to B.S.Pavlov and S.P.Merkuriev for useful dis-

cussions,

References

I. K.Wildermuth, Y.C.Tang: A Unified Theory of the Nucleas, Vieweg,

Braunschweig 1977.
2. S.Saito,lS.Okai , R.Tamagaki, M.Yasuno, Prog. Theor. Phys. 50
(1973), 561.
3. S.Saito, Prog. Theor. Phys. Suppl. 62 (1977), 11.
4. E.W.Schmid, Z. Phys. A2~7 (1980), I0-~.
5. V.I.Kukulin, V.G.Neudatchin, V.V.Pomersntsev, Yad. Fiz. 24 (197~
298.
6. B.Buck, H.Friedrieh, C.Wheatly, Nucl. Phys. A 27~ (1977), 246.
7. B.S.Pavlov, Teor. Mat. Fiz. 59 (1984), 357.
8. Yu.A.Kuperin, K.A.Makarov, B.S.Pavlov, Teor. Mat. Fiz. 63 (1985~
78.
9. Yu.A.Kuperin, K.A.Makarov, B.S.Pavlov, Teor. Mat. Fiz. 69 (1986) l
100.
10.Yu.A.Kuperin, K.A.Makarov, S.P.Merkuriev, A.K.Motovilov: in
"Theory of Quantum Systems with Strong Interactions", Kalinin
1987~ D.4.
11.Yu.A.Kuperin , K.A.Makarov, S.P.Merkuriev et al., ITP-Budapest,
Report N 441, Budapest 1986; Proc. of the 3rd School on Few-Bo-
dy and Quark-Hsdron Interactions, Vilnius 1986 , Part 2; p.28.
12.B.S.Pavlov, Usp. Mat. Nauk 42 (1987),6(258J, 99.
13.I.L.Grstch, Yu.S.Kalashnikova, I.~.Narodetskii et. al., Yad. Fiz.
42 ~ 985) , •
14.~.P.,~erkuriev , L.D.Faddeev: Quantum Scattering Theory in Few-
Body Systems, Nauka, ~oscow 1985 (in Russian).
 THE PROBLEM OF A FEW QUASI-PARTICLES IN SOLID-STATE PHYSICS

Alexandr I. Mogilner
Institute of Metal Physics, Academ~ of Sciences,
Ural Branch, Sverdlovsk, Kovalevskaya street, 18, USSR

Abstract

The problem of energy spectrum of a few quasi-particles in a cry-

stal is investigated. The form of the N-magnon Hamiltonian in a Hei-
senberg ferromagnet is obtained and general spectral properties of
the Hamiltonian as a cluster operator are demonstrated. The quasi-
particle spectrum in the strong coupling limit, the Efimov effect,
the current and noncurrent bound states are also discussed.

I. The Hamiltonian of e N-mafnon System in a Heisenberf Ferromagnet

Some problems in solid-state physics require a investigation of

a few quasi-particles in a crystal. One of the most important prob-
lems of this type is the spectral analysis of the Hamiltonian desc-
ribing a few quasi-particles on a lattice [I]. An important example
is, for instance, the N-magnon system in a Heisenberg model.

Let us consider a d-dimensional cubic lattice Z d each site of

which is occupied by a spin. One of the most simple Hamiltonians of
this system has the form:

H = - ~I m,n
~ Jm,n (S~S~ + bS~S~) + D ~ 2( Sn~ ) (I .I)

where Jm,n are the exchange integrals, Jm,n=Jm_n=Jr , J r ~ O ,

Jr <C1exp(-C2~rl)' CI'C2 ~0' S nz is the operator of spin projection
on the z axis at the n-th site, Sn (S~) is the operator lowering
(rising) the z component of the spin by one ~see ~23), b and D are
exchange and single-ion anisotropy coefficients, O ~ b ~ 1 , D~O.

Since the operator So= ~ SZn of the total spin projection on the
z axis commutes with the Hamiltonian H, it is possible to classify
the stationery states using its eigevalues. The highest eigenvelue
of S o corresponds to the ground state while the excited states cor-
 161

respond to the eigenvalues of So, which are by 1,2,...,N less than

the highest one. The wavefunctions corresponding to these states are
given by
N
= A -I ~] - xO (I .2
X<n} g i=I Sni

where Ag is the normalisation coefficient, g is the multiindex asc-

ribed to the set of sites (n}
: <nl...nN} , g = {gl...gs~ and gm is
the number of sites with m spin flips,
s
s~<N, ~ mg m : N
m=1
s m- I
Ag= m~__1 (Am)gmZ2 , Am= l~=o (2S-I)(I+I) CI .3)

It is clear that if gm ~ 0 for m ~ 2 S , then X(n}=O , i.e., there can not

be more than 2S spin flips at one site. Moreover the spin operators
corresponding to different sites commute and this implies that the
functions X<n ) are s~mmetric with respect to variable permutations.

The functions (1.2) are not eigenfunctions of the Hamiltonian (1.1~

The wsvefunctions corresponding to the N-msgnon stationary states
can be constructed as a linear combination of the functions X<n}:

x = 2 Q-If(n}X~n~ (1.4)
(n~ g

The constants Qg are choosen in such a way that

'x'2 = ({}If(n~2
where we are summing over all different functions f~n}"
s
Qg = (N! ~ ~m!)'gm)I/2 (1.5)
m=1

Substituting the functions (1.4) into the eigenfunction equation

for the Hamiltonian 41.1) and identifying the coefficients of X~n }
we get after a cumbersome calculations the expression for the Haml-
itonian acting in the space of N-magnon functions f(n}612sym(zdi~J:--
""

Hf = FN f + bSBf + Vf (I .6)

F = D(I-2S) + SJ , J = ~r Jr '
 162

V is the operator of the dynamical two-magnon interaction.

I
(Vf){n} = (i>-.#j(D~nl.,n~ - ~ Jni,n~)) f{n} {1.7)
B is the operator of kinetic e n e r ~ and kinematic magnon interaction~
N
(Bf){n) : i=~
= ~y - B ni'y f{(nkni)Y } (1.81

and {~nkn~ y} denotes the set ~n I ...hi_ I y hi+ l...nN}. The transfer
amplitude of the i-th spin flip from the site n i to the site ~ is
given by

where p is the number of spin flips on the site n i in the configura-

tion {n} and v is the number of spin flips on the site y in the
configuration {(n\n~ y}.

Let us note that the expressions CI.6)-0.9) are correct also for
N>2S. B~'Y=O for p,v>2S, i.e. the functions f~n}with gm~O for m > 2 S
do not contribute to the r.h.s, of (1.6). The equations for f~n}
with {n ~ such that gmiO for m >2S must be considered as sbsent~

Let us turn from the coordinate to momentum representation:

aN S e p(i N
f{n} = (2~)- 2-- (Td) N T ( k1"''kN ) j=IZ (kj,rnj) dk 1...dk N
.Io)
where T(k,...kM) : T ( { k ~ ) is the N-magnon wavefunction, ~U({k})6
~s~'m~(Td)N), k~' is the quasi-momentum of the j-th magnon, k 4 6 T d
(T d is a d-dimensional torus) and r n is the coordinate of ~he site
j. The quasi-momentum kj is measured @n units of a -I, where a is the
step of the lattice.

The expressions for the operators B and V have in the momentum

representation the following form:

× r~ ~ (kl-k~)U/({l~}) dkl...dki~ (1 .11)

l~i, j
W2(ki,kj,ki,"k'.)=j 2D - ~ (j(1 (ki_k+._ki_k:
J d)) J(2"(ki-k3+ki-kj
I "
)))
(I .,2)
 163

N
k 4
i=I = (Td) N VnCkt (n) ' t n) x

•
"~Kt{n)-Kt(n)) j~t(nf (kj-k~)~((k'})dk I. .dk N " " "
(1.13)

J(k) = 2 Jr exp(i(k,r)) e(ki) = J(ki)

r

is the kinetic energy of the i-th magnon and t(n) and kt(n) are the
n-element subsets of the sets <I ...N} and {k I ...kN} respectively.

Kt(n) = ~ l ( n ) kJ , Kt(n) = j~t(n)

~ k~ ,

C] 2S /CJ(kil+ J(kjI+J(ki)+J(kj))
• • (I .14)

I n-1
Vn(kt(n)'kt(n)) : 7 ~ ~ C(p-l) (n-p) ~
p=O t(p,i),i~tCn)

[J(j t(p ,i) ( i-kJ ÷ki ' (I. 1 5)

Coefficients Cp, v are given by recursive relations:

Cp,v =~eS+I-P~s(2S+I-v)' - Z C . , C I =I (I .16)

p'<p,v~ v p,v ' I

and the functions Vn(kt(n) ,k"t(n)) are the kernels of the N-magnon
kinematic interaction operator.

2. The General Spectral Properties of the Nrquasi-particle

Hamiltonia D on a Lattice as a Cluster Operato~

The formulas (1.6), (1.11,)-(1.16) define the N-magnon Hamiltonian

in the momentum representation. This operator turns out to be a par-
ticular case of the additive N-particle cluster operator ~3]. There
are also other examples of systems of a few quasi-particles on a lat-
tice (magnons, electrons, excitons) the Hamiltonians of which can be
written in the form (I .6), (I .I I)-(1.16) . (The expressions for the W
and V kernels are different for different problems). Owing to this
common feature, the following multichannel picture is characteristic
for spectra of all Hamiltonians of that type [3J.
N
For an3, N we can fix the total N-magnon quasi-momentum K=i~ I k i
and consider restriction of the operator H to the subspsce of wave-
functions corresponding to a fixed K. We denote a restricted operator
 164

as H(K) (see[4] for a more detailed analysis of the decomposition of

the operator R into the "layers" HCK)).

There always exist a main N-particle branche of the continuous

spectrum of the operator H(K), which corresponds to the free motion
of N noninteracting quasi-particles with quasi-momenta kl,...,k N
and energies
N N
e(ki) , ~ ki = K
i=I i=I
Moreover there exist a finite or countable number of p-particle bran-
ches of the continuous spectrum, p=2,..., N-I , which correspond to
the free motion of p clusters with quasi-momenta KI,...,K p and
energies

P
~- E~i (Ki)
i=I
Here K i is the quasi-momentum of the centre of mass of the i-th clus-
ter corresponding to the s-th bound state of n quasi-particles, n=
=2,...,CN~I-p). In the case n=1 we have one free quasi-particle. The
cluster energy E~i(K~ is a point of discrete spectrum of the Hamil-
A ~

tonian H(Ki) of the m-quasi-particle system. And finally the operator

H(K) can have the discrete spectrum corresponding to N-quasi-particle
hound states.

The algebraic structure of the Hamiltonian (1.6),(1.11)-(1.16) is

analogous to the structure of Shrodinger operator (SOl acting in the
space of N particles in a d-dimensional Euclid~an space R d. There is
an important difference however: in the case of SO we have always

e(k) = ~m k2 (2.1)
where m is the particle mass and k ~ R d. With this analogy we call
the few-quasi-particle Hamiltonian as a discrete SchrSdlnger opera -

tot (DSO .

There are man3~ mathematical difficulties in investigation of the

SO which are connected with the noncompactness of the space,unboun-
dedness of SO. On the other hand, the structure of spectral branches
is much more simple in the SO case (see[S]) because e(k) is then
given by the formula (2.1) while for DSO it can be of an arbitrary
form.
 165

The structure of the discrete spectrum Sdisc of DSO is also more

complicated than in the SO case C ~ d i s c denotes all isolated eigen-
values of finite multiplicity. For bound states embedded in the
continuous spectrum see ~4S,~6]). There is a theorem describing the
essential spectrum of the cluster operator which is the analogy of
the HVZ theorem for SO C7].

I. In the DSO case the N-particle bound state with energy E(K)
need not necessarily exist for all values of centre-of-mass quasi-
momentum.

2. The hound state energy E(K) may be higher than the energy cor-
responding to the continuous spectrum.

3- It is possible that e ( O ) < e ( k ) , k~O and E ( O ) ~ E ( K ) , KiO

(see §5 below).

4. In the SO case t w o w e a k l y i n t e r a c t i n g particles cannot create

a bound state for d=3. This is, however, possible, in the lattice
case if e ( k ) ~ k 4 for small k /2].

5. If d=3 in the SO case if and if any two particles out of the three
a bound state then the bound state of three particles also exists.
There are examples, however, that this does not hold on a lattice.
For instance two magnons with infinitesimal nearest neighbours inter-
action form a bound state while three magnons do not.

There are many defailed studies of the two particle case, but
only a few papers investigating the N ~ 3 case (see ~I] and referen-
ces therein) because of the difficulties connected with analysis of
the few-particle Hamiltonians. In the DSO case there are two limit
situations which can be investigated: the weakly and strongly inter-
acting quasi-particles. The weak interaction case was discussed in
our paper ~8].

~. Magnon Bound States in the Strong Coupling Limit

In the strong coupling limit (SCL) which has no analogy for SO

the quasi-particles on a lattice are motionless. An example of this
situation are the spin flips in Ising model with the nearest-neigh-
 166

bour interaction described by the Hamiltonian /1.1) with

S = ~, D=b=O, Jr=J for Irl=a, Jr=O for ~r~ > a

Here bSB=O and Ho= const + V where V is the operator of multiplicati-

on by a function in the coordinate representation. It is clear that
the wavefunctions X~n } appearing in (1.2) and describing the spin
flips at sites nl...n N in the regime when at most one flip occurs at
each site are the stationary states of the Hamiltonian corresponding
to the energies

: NJz - n n}J (3.1)

where n~n } is the number of the nearest neighbours pairs in the con-
figuration fn} and z is the number of the nearest neighbours to the
site (z=2d).

In view of the translational symmetry of the problem it is con-

venient to introduce another system of basis vectors:

X~n'}(K) = ~ ei(K,R )
%N rni (3.2)

and the configuration <n'}is defined by the set of (N-I) vectors

~ rn I- rn 2 ,''', rnN_1- rn N }

HoX{n~(K ) = E(n ~ x ~ n ~ K ) E(n,}= ~ JNz - n(n)J

We fix the conserved centre-of-mass quasi-momentum K. We will use

the term connected for all those configurations where for each spin
flip there exist a corresponding spin flip in the nearest neighbour
site. All others configurations will be called disconnected.

All states corresponding to the disconnected configurations are

infinitely degenerate and the part of the states corresponding to the
connected configurations has the same energy. Energies corresponding
to these states belong to the essential spectrum~ess. The remaining
part of states which correspond to connected configurations are fi-
nite degenerate and their energies lie in &disc"

Let us now suppose that O < b < < 1 . It is easy to show that the ope-
 167

rators H(b) = H O + ~ B form a Kato's analytic family [4] and that

IIBII < CJzN , C "I ('3.3)

We know from the perturbation theory [43 that under the perturbation
with b<<1 the points of ~ e s s transform into the segments O f ~ e s s
which are possibly limited by a finite or countable number of ~ d i s c
points. The quantity 6 majoring the dimension of these ~ess-Segments
and limiting points is equal to

b.
2 II all = b C J z N

Thus the criterion of the applicability of the perturbation theory is

b IIBII << rain I E(n}l I b<< 2 (3.4)

2 (n}1 ,{n}2 - E{n}2 z-'N

The perturbation removes partially the degeneration of the points

of 6disc. It can be shown that the energies of the N-magnon bound
states in SCL are contained in segments [E{n~-C b zJ ; E(n}+C b ZJ].
Let us note that for large N most energetically advantageous forms
are the "cubic drops" with n given by:

n : ~- z N d-1
z N - ~- d - C , CNI
d-1
E(n}= C ~ JN d (3.5)

It can be also shown that for N>>I, IK{<<I the following relati-
on holds :

d
E(n,I(K)=const + i=I
E_ --
( (b.d/Cd i>) 9 1 (3.o)
2eL ' mi~
where K i are the coordinates of the vector K on the main axes. One
can see from the formula (3.6) that the magnon-drop effective mass
is essentiall3, nonadditive. This is one more important difference
between DSO and SO.

Let us consider now a different model:

b=l, S>>I, N<<S, SJ<<IDI (3.7)

In this model spin flips are bosons interacting by pair contact po-
t%ntials on a lattice, Two quasi-particles attract each other with a
 168

strength D only if they are at the same site and do not interact
when they are localized at different sites. Under the condition (3.7)
the Hamiltonian H can be written as

H = V'+ T'+ o(S,D) (3.s)

I(v~)in). = ( 2i~j D ,~ i,3 ) fin~"
"
, N (3.9)
L(Tf)£n } = S i=IZ Zr Jr f((n\n i) (ni+r) }

The wavefunctions X{n } from (1.2) are the stationary states of the
operator V'with the energies

s m(m-1)
E~n } = ND(I-2S) + D Z gm ~ (3.10)
m=1
where gm were defined above. Under the conditions (3.7) the kinetic
energ~ operator T'can be considered as a small perturbation of the
potential enerE~7 operator V[ All the energy levels (3.10) are infini-
tel~ degenerate and transform under the perturbation into segments of
bess except for the nondegenerate lowest level with the energ~

= N~N-I)
Eo ND(I-2S) + D 2

This nondegenerate level corresponds to the bound state of N spin

flips at one site.

The N-msgnon bound state energy, becomes under the perturbation

Eo= ~ ( ~ - 2 s ) + D N(~-~.____.~ z
-
D + o (3.~1)

and the effective mass tensor components of this state are given b~

mi'~z D (N-I) (SJ) -N (3.12)

4. The Efimov Effect for Three ~uasi-Particles in a

Three-Dimensional Crystal

The existence of the phenomenon discovered by Efimov E9-113 is

rigorously proved for a system of three particles interacting by at-
tractive short-range two-body potentials. If there are no two-particle
 169

hound states and two of three pairs create a virtual state, then there
exists an infinite number of three-particle bound states with energi-
es approaching zero. The asymptotic expression for these energies is
given by

En~Wl-e-W2n , n=1,2,... (4.1)

where w I and w 2 are positive constants depending on masses of the par-

ticles and on the two-particle scattering length. It was shown in
[9-11] that the Efimov effect is caused fully by the two-particle
scattering amplitude singularity at E-~O and ki-~O. Under these con-
ditions, the singular part of the homogeneous Faddeev equation ~ =
=A(E)~ separates. (Here ~ i s the wavefunction of the three-particle
system and A(E) is an integral operator in the momentum representati-
on. The eigenfunctions of this equation are the statlonar 2 states of
the system and E are the corresponding energies. For more detail see
[12].) This singular part is non-Fredholm for E-@O. Its kernel As(q,~)
coincides with the kernel of A at ~q~,[q'~< A and As(q,q" ) =0 for
~,~q~>A and A small. The characteristic numbers of the operator A S
have an accumulation point st zero and the additional Fredholm term
(A-As) does not change the accumulation point in agreement with the
Weyl's theorem [10].

Let us note that for smalllkil < A t h e ,.~q~asi-psrticle kinetic ener-

gy is given in the general case by e(k)= ~m + O(k4)C we consider only
the highest -symmetry case). Therefore t'or Ikil< A t h e three-quasi-
particle Hamiltonian (I .6), (I .I I)-(I .I 3) coincides with the Hamilto-
nien of a three-particle system in an Euclidean spase written in the
momentum representation with the exception of the three-particle in-
teraction. The short range three-particle interaction adds, however,
only a Fredholm term to the Faddeev equation. Therefore the Faddeev
operator corresponding to DSO can be decomposed into the singular
part AsH which coincides with A s and the Fredholm term' A H= A Hs + ~ "
Thus for the three-quasi-particle system in s three-dimensional cry-
stal the Efimov effect takes place at K=O.

1. In analog, to SO [ t 3 ] in the case of dimensions d=1,2 or for more

than three quasi-particles ( N ~ 4 ) , the Efimov effect is absent.

2. In the case when two of the three particles are impurities, i.e.,
e1(k)=e2(k)=const (as an example consider a ferromagnetic crystal
with two ferromagnetic or antiferromagnetic impurities) the Efimov
 170

effect exists, i.e., there is infinite number of bound states of a

magnon with two impurities. The physical nature of these states is
clear: a virtual state of the magnon on one impurit 2 creates a long-
range interaction w i t h the second impurity. This second impurity can
be localized at any site generating in such a way an infinite number
of bound states.

3. In the case when one of the quasi-particles is an impurity (two

magnons in the ferromagnetic crystal with one antiferromagnetic impu-
rity) the Efimov effect occur if and only if both the magnon with
impurity and two magnons create virtual states. In fact there exists
onl 2 a finite number of bound states in the limit of noninteracting
magnons.

4. The effect is not present if a bound state rather than a virtual

state appears at the continuous spectrum threshold and the two-parti-
cal scattering length remains finite.

5. The effect is apparently absent if all three quasi-particles are

fermions with spin ~ (three electrons in the Hubbard model).

6. Let us mention one more system, where the Efimov effect takes
place: three magnons in an anisotropic ferromagnet with such an ani-
sotropic constant that two magnons at K=O create a virtual state.

The Efimov effect must lead to some observable peculiarities in

thermodynamics and in kinetic phenomena . The existence of the Efi-
mov effect for three bosons on a lattice interacting by two-particle
contact potentials was announced in CI].

~. Current and Noncurrent Ground States of a Few

Quasi-Particles on a Lattice without External Field

It is well known that the e n e r ~ of a system of a few particles

the dynamics of which is defined by SO is given by

E = E o + 2-M

where M as the total mass and E o is the internal energy of the sys-
tem, w h i c h does not depend on K. The ground state corresponds to K=O,
 171

i.e., the ground state is noncurrent. This is, however, not in the
case of quasi-particles on a lattice, where the internal energy de-
pends on the centre of mass quasi-momentum K. This dependence is ar-
bitrary in general and the question is whether the ground state of
the system on a lattice without external field is current or noncur-
rent, i.e., whether it has a quasi-momentum K#0 or K=0 ? This questi-
on is interesting not only from academical or methodological reasons,
but also from the viewpoint of possible applications. It is, moreover,
related also to the discussion on macroscopic current states in
conductors.

In what follows, we give an example of a one-dimensional system

with a current ground state. We formulate conditions under which the
ground state of a Hamiltonian describing a system with an arbitrary
number of quasi-particles is noncurrent.

Let us note, that if we have only one quasi-particle, then its

energy is given by e(k) = ~ Jrexp(i(k,r)) and the minimum m~n e(k) is
achieved st K=0 if and only if J r ~ 0 (Jr is the hopping matrix).

~he system is described in the coordinate representation by a

wavefunction ~(n~" The Hamiltonian can be represented in a form:
N
(H~)(n}= i=I~ ZJ(~n}ni)r
<n) (ni+r'}~((n\ni)(ni+r) } + i~#jVi,j~{n~ (5"I)

j{(n+c\n~+c) {ni+r+c) } .T {(n\n~) (hi+r)}

V£(ni+r) (nj+r)}= Vni,n j (n+c} ~ = ~<n~
The centre-of-mass quasi-momentum is conserved and the wavefunction
must have the form

• i(K,R) Z
{n') c {n'} x (n'}
2
<n'} C(n'} C(n'} =
, (5.2)
From the minimax arguments we know, thet the ground state energ 2
equals to

Emi n = m&n <%ulB}~> (5.3)

<~I~>
and the ground state wavefunction is described by the function
on which the minimum is achieved. Substituting (5.1) and ~5.2) into
~5.3) we get
T

We find that the N-quasi-particle Hamiltonian (5.1) has under the

conditions (5.5)a noncurrent ground state, i.e., the centre-of-mass
quasi-momentum of the ground state equals to zero.

Let us now give an example of a one-dimensional two-quasi-particle

system with a current ground state. Let [k[ and let the corres-
ponding dispersion law be given by

Further let the two-particle interaction be given in the momentum

representation by a rank-one operator with a kernel V(q)~(q'), where

Then the bound state energy E can be obtained from the equation
(U is the coupling constant) :
 173

K= ~ ~" 2U I
5'4 = I--/~-E , E = -,~
Hence E(2)<E(O}. We note that in this example the inequalities
J r ~ O , V r ~ O hold.

The ground state energy of a N-quasi-particle system with a kine-

tic energy fulfiling min e(k) = e(O~ and with repulsive interaction
between quasi-particles is given by Ne(O).

References

I. D.Mattis, Rev.Mod.Phys. 58 (1986), 361-379

2. A.M.Kosevich, B.A.Ivanov, A.S.Kovslev,"Nonlinear magnetisation
waves. Dynamical and topological solitons", Naukova Dumka, Kiev 1984
(in Russian)
3. R.A.Minlos, A.I.Mogilner, in '~athematicsl problems of statisti-
cal mechanics and dynamics", D.Reidel, New York 1986
4. M.Reed, B.Simon,"Methods of modern mathematical physics. 4: Ana-
lysis of operators "~, Academic Press New York 1978
5. M.Reed, B.Simon,"Methods of modern mathematical physics. 3: Scat-
tering Theory", Academic Press, New York 1979
6. A.I.Mogilner, VINITI Publications I019-B87,1987 in Russian
7. H.Zolondek, Teor.Mat.Fis., 53 (19821, 216 - 226 ~ n Russian~
8. A.I.Mogilner, Physics of Metals and Metallography, in print
(in Russian)
9. G.V.Efimov, Phys. Lett., B33 (1970), 563-564
10. D.R.Jafaev, Mat.Sbornik, 94 1974 , 567-593
11. R.D.Amado, J.V.Noble, Phys.Rev D5 1982 , 1992-2002
12. S.P.Merkurjev, L.D.Faddeev,"Quantum scattering theory for a few
particle systems", Moscow, Nauka,(1985~ (in Russian)
13. S.A.Vugalter, G.M.Zislin, Doklady Akademii Nauk SSSR, v.267
(1982) , p.784 - 786 (in Russian)
 SURFACES WITH AN INTERNAL STRUCTURE

B.S.Pavlov, P.B.Kuzasov
Department of Mathematical and
Computational Physics, Institute for
Physics, Leningrad State University,
I9890@ Leningrad,St.Peterhoff, USSR

Zero-range potentials with internal structure allow to const-

ruct quantum-mechanlcal models of various physical phenomena. In
this lecture, we discuss a particular class of operators of this
type, which can be used for description of surface phenomena in
solid-state phisics.

The concept of zero-range potentials with internal structure

proposed in Ill opens way to construction of new exactly solvable
quantum mechanical models. In the present paper this method is used
to discuss the surface effects. The boundary-contact problem without
internal structure was investigated rigorously earlier by B.P.Be-
linsk~ [ 2]. In distinction to the work of Karwowsky [3] who conside-
red the simplest model of a translationally invarlant surface, our
method allows to treat the translationally invariant surfaces as
well as periodic structures. Our approach is similar to the method
of Ref. [ @], where a similar mathematical construction was applied
to the description of energy-dependent two-body problems in few-body
physics.

The scattering on a surface, described in Sec.I is quite simi-

lar to the two-particle scattering. The spectrum of the correspon-
ding operator is purely continuous. It consists of two branches :
the branch of the scattered waves and the waveguide branch, which
represents an analogue to bound states. The eigenfunctions corres-
ponding to the first branch describe various processes of transmis-
sion and reflection of the plane waves. We consider here only the
simplest case of a planar surface, but all the results can be gene-
ralised for an arbitrary shaped surface. Such a generalisation,
however, does not yield an exactly solvable model.
 178

The presence of a potential barrier ( considered is Sec. 5 )

splits the scattering branch of the continuous spectrum into two
parts.The surface waves described in Sec.2 c a n propagate with a velo-
city which differs from the velocity of the "external" waves. At the
same time the wavegaide branch of the spectrum changes. The space
anisotropy of the sumface modifies the waveguide functions in a ty-
pical way. Sec.4 is devoted to description of a model of a zero-range
defect localised sear the surface with am internal structure. In the
last Seotios we discuss a surface with a periodical structure.

i. Formulation of the problem

Let ~ denote a plane in R 3,

r = ~x=(x,y,z) C R ' , z=O}

Let AO= - • be the self-adjoint Laplace operator in L2(R 3) with

the standart domain D(Ao)=W2(R3). We restrict this operator to
c2 (~3\F),
Aoo= ~eP c~(~3\F)
where CO(~ ) denotes the set of all infinitely differentiable func-
tions with a compact support contaised in ~-~. This symmetric opera-
tor has infinite deficiency indices. The boundary form of the ad-
joint operator AO0 restricted to the linear set of smooth functions
can be written in terms of the limits of functions and their deri-
vatives taken from the different sides of the plane ~ :
+
Uo(X,y) = Uo(X,y,±o)
+ +
-aUo (x,y) = "~ Uo (x,y,+-o)

Then the boundary form can be expressed as follows

÷ <~U_mo
( oo o,Vo>-<oo , ooVo}= II [-~ U_O]
f L~";J < Vo';
....
~o > [Vo] -

-[oo]. <Oo [ ~o,Vo-] -~ (T.I)

 179

where :
uo ~u; ~u o
[ uo] u; - uo i~T ~ %n ~

< Uo> :(Co ÷ Uo)i~

In order to add an internal structure we choose self-adjoint opera-

tors Ai, i = 1,2 acting in Hilbert spaces H i . In accordance with
the standart procedure [i], we restrict these operators to symmetric
operators with deficiency indices (1,1). An arbitrary element from
the domain of the adjoint operator Ale can be represented in the
following form :

+ wi u~ wi
ui = ui + + - + ~i

where u~i6~D(Aio) and w +~ are basis elements in the deficiency sub-

m
spaces [ 1] . The boundary form of the adjoint operator can be des-
crihed in terms of the "boundary" values u+, u[ :

*
< AioUi.vi ~
> - <ui,AioVi b : ~ -vl
- - ui
+ v-~[ --- i ui ~ V i (1.2)

Now we can construct symmetric operators with infinite defici-

ency indices. We start with self-adjoint operators defined in the
Hilbert spaces L2(R 2, H~) as follows :
D "~2ui -82u i
(Y~iu)(x'Y) " - ( ~ x = + 35~ )(x,y) + (Aiui)(x,Y)

We restrict these operators to the linear set of smooth functions

C~°(R2,D(Aio)). The boundary form of the adjoint operator restricted
to the set of smooth functions can be written as •

(1.3)
We shall consider the symmetric operator ~ 0 acting in the Hilbert
space

L2(R3) @ L2(R2,Hl) ® L2(R2,H2)

 180

whose elements are denoted by U=(Uo,Ul,U 2) as follow

~o: ~oo* RlO*~o

This operator has infinite deficiency. The boundary furm of its ad-
joint on the set of smooth functions ~epresents a sum of boundary
forms of the operators 3~00' J~ I0' ~ 20 :

.
< ~ o•~~ , V ~ - <~,~oV~ : z~ [ u~, 7~
v~ J
~t=O
This form vanishes on the domain D O ~ which is determined by the
boundary conditions :

-~ UO~

o~+: ~ < u° > ~; : ~ <~oOo

Thus the restriction of the adjoint operator T 0 to the domain D O
is a below semibounded symmetric operator. We can therefore const-
ruct Friedrichs extention ~ of the symmetric operator ~ 0 "
One can show that ~ d o e s not coincide with the original operator
so this Friedrichs extension is not trivial. The spectrum of this
operator is purely continuous. There exist two branches of the con-
tinuous spectrum : the scattering branch and the waveguide branch.
The external component of the scattered waves can be decomposed into
plain waves of three types :

reflected waves R exp i(~k,x)

/
transmitted waves S exp i(k,x)
(see Fig.l)
where :

~(kx,~,k z) : (kx,b,-k ~)
Fig. i
The external component satisfies the following equation :

- ~ u0 = k 2 u0

The boundary values for the external component are :

 181

<Uo> = ~ ( 1 + s + R ) exp,O,x~+Sy)
[Uo~ = ( S - l - R ) ~xpi(kxX+ 5/)
-'0 u 0 ~ ik z
: -- ( 1 + S - R ) exp i(kxx + ~ y )
<"On 2

1"%-u0] = ik z ( S - 1 + R ) exp i(kxx + kyy) (1.5)

Dn

The internal components are assumed to be of the following form :

ui(x,y) = f(x,y) ui

where f~W2 loc(R2) , uig H i . Let f(x,y) be given by

f(x,y) = exp i(kxX + By)

We obtain the following equations for elements of the internal spa-
ces H i :
2
AioUi : ( K2 _ ~ - ~ ) Ui = k z ui

It follows [i] that

u[
u~ Di( kz
2 )

where D i are Schwarz integrals of the internal operators A i on the

deficiency elements C~i :

Di( ) = O5

+ + m

Substituting Uo, u i into the boundary conditions (1.4), we can

calculate the transmission and reflection amplitudes :

(Io~12 2
2D I - ~ ) ik z
S =
lod 2 2
 182

(ikz)2 lal 2 2
2D I I~-~ 2
R = (1.6)
2
(ikz i~IE) (ikz
2D I ~-~2 )

They fulfil the identity :

IRi 2 + Isl a : 1

The internal components Ul, u2 of U can be easily calculated in

terms of the deficiency elements ~i"

Let us investigate now behaviour of the scattering matrics in

the case of a weak coupling between the internal and external chan-
nels,a--'O and ~--~0. The transmission amplitude S vanishes if

I~I 2 2

2D 1 ~- -2 = 0

The zero lines of this equation in the (k,t) - plane coincide (ap-
proximately) with the zero lines of the function DI/D2-- . The
reflection amplitude R vanishes if

I~I 2 1
D1 D2 =
(ikz)2

The eigenfunction of this type form the ~0,oo) branch of the conti-
nuous spectrum.

The external component of waveguide functions can be found in

the following way : we put

B + exp(-~z) exp i(kxX + kyy), z>0 , ~ > 0

Uo(X,y,z)
B- e x p ( ~ z ) exp i(kxX + kyy), z < O ~ > Q (1.8)

So the boundary values are

~Uo~
i expi( +kyy)
 183

<%U 0 > = "~e ( B- - B+ ) exp i(kxX + kyy)

7~n 2

[ Uo] _- ( B+ - B - ) e~pi(kxX+ 5y)

1
Uo ...... ( B+ ÷ B- ) exp i(kxx + kyy)
2
The conditions (1.4) now yield

~-~ u0 I~ 2

[Uo]
The constants B +, B- satisfy the following homogeneous linear system
of equations
l~l2 I~ 2
(~ + ) B+ + (~+ ) B-:O
2 DI 2 D1

~3i2D2 ~~2D 2
( l + ~ - - ) B+-( i+ ~ - - ) B-=O (l.lO)
2 2
which has a nontrivial solution only in the following cases
L~ 2
+ = 0
2 D1
~2D 2
I + ~( '' = 0 (1.n)
2

The solution of the first equation in (i.Ii)gives symmetric wave-

guide functions while the solution of the second equation gives the
antisymmetric ones. In the weak-coupling case the solutions of the
equations (1.11) are near to the zero lines of D 1 and D21, respec-
tively. The waveguide eigenfunctions correspond to the following
branches of the continuous spectzum :
[_~2,~) ,

where ~ i are arbitrary real solutions of one of the equations

(l.n).
 184

A screen with am internal structure allows us to construct a

resonance transmission of plane waves across the surface. The ex-
ternal components Soo of the scattering matrix depends on the nor-
mal component of the momentum k z. In this way we can obtain wave-
guide eigenfunctions in this surface model.

2. Velocity of waves in the internal space

In this section we are going to describe a model of the sur-

face with a special internal channel. The velocity of internal
waves is, in general, different from the velocity of the external
ones. Let 9~ i be the unperturbed internal operator

i = - ci2 ~ xy + Ai

The constants ciER + are velocities of the internal waves. One can
construct a self-adJoint operator following the procedure outlined
in the previous section.

The spectrum of such an operator is also purely continuous and

all eigenfunctions can be calculated in the same way as in the Sec.l.
The waveguide functions depends on the quasi-momentum t=(tx,ty) and
the same is true for the energy ~%. One can easily obtain the dis-
persion equations :

- (~-t 2) + = o (2.1)
2 DI( h-c~ ta)

-l+ (A- t2 ) i~12D2(h- c~ t a) : o (2.2)

Let us investigate the solution of these equations in the case of

weak coupling between the external and internal channels. There are
two different cases to be distinguished
i) c i > I
2) c i< 1
In the first case all energy bands are finite while in the second
case they are infinite. We shall prove this result graphically using
Fig.2 . Zero lines of the Schwarz integral D1(-c~t z) are para-
bolas ~ - c~
~
&
t 2 = const denoted by the dotted lines. In the
 185

case ~> 1 , the parabolas of this family intersect the parabola

=t or lie inside it (Fig.2a). If Cl< i then the parabolas lie
in the exterior of the parabola ~ = t 2 or intersect it (Fig.2b).

Solutions to the equations (2.1), (2.2) are situated near the

zero lines of the Schwarz integrals in the region ~ < t 2. They are
denoted by full lines on the figures. The projection of the solu-
tions on the energy axis yield the spectral bands of the correspon-
ding operator.

t ~,:t 2

f ~ - - - D , : 0 D,: co
/-
/

"----~ L:__- L-- Z-- L

Fig.2a .oo / . I " __

t / ~ 2
/.-;(3
// / o,

Fi~ ;. 2b ~ .~
The simplest generalisation of this model is connected with the
space anisotropy of the surface. Suppose that the unperturbed inter-
nal operator ~ i is of the form :
~2 ~2
2 c.2

In the case cix:Ciy , the diagrams on Fig.2 become three-dimensional.

Solutions to the corresponding equations lie near the zero surfaces
of the functions Dl( ~ - Clx2 t x2 - Cly2 ty2 ) and

D~I( ~ - c2x2 t x2 - o ~ ty2 )

 186

The picture is symmetric with respect to the energy axis. The wave-
guide functions can propagate nearly in a particular direction
in one of the following cases :

cix >> Ciy or Cix << Ciy

The corresponding energy bands are infinite if in some direction

the velocity of the ioternal wave is less than the velocity of the
external one

3. Surfaces with a potential barrier

Suppose now that the external unperturbed operator is of the

form ~ 0 = - ~ + q(z), where

q(z) : ~ O, z > O
I -A, z < 0
For some A ) O we can construct the internal structure in th~ same
way as in Sec.1. The model is now nonsymmetric with respect to the
plane F . Consequently the scattering matrices are different in the
positive and negative z-axis directions, (S+,R +) and (S-,R-) res-
pectively.

The external component of the scattered waves have the same

form, ms in the previous cases but the momenta of the incoming and
transmitted waves are different. If we have an incoming wave of a
momentum k' = ( k'x,~y,~'k'z )' kz' • 0 the reflected wave has the m o m e n -
turn T k ' and the transmitted one k = ( ~ ,, ~ , , k'v2 - A ). If
k~ - A <0, then the transmitted wave is damped. The standard pro-
cedure gives us the following results :
i~2 2
2 ik~ ( ~ )
2 D1
S÷= 2 ''
(ikz-i'~,i) (ik~-~)+ ( i k ~ - ' ~ ! , ) ( i k z- 2
2 ]~l ~ID 2 )
I,xl 2 2 I 412 2
( ik z- ) ( ik~+ ik~+ (i k z -
R+ = 2 D1 ,,, t )+( ..... D1 )

(ik z- I~(12,, (ik~- 2 l~I 2 2

2 D1 ) ~~l
2 " T) + (ik~ . . . . ) (ik z-
2 D1 ~u~,_2 )
 187

One can check directly that R+ = I holds for k~ < A; in this case
the incoming wave is totally reflected by the potential barrier.

The eigenfunctions of this type form the waveguide branch

~-A,oo) of the continuous spectrum.

The scattered waves corresponding to the normal component of

the incoming wave k z < 0 can be described in the same way :
2 l~l 2
2 ik z ( ~ D 2 2 D1 )

I~I 2 2 I ~ 12 2
- ) (ikz+ + (ikz+ , ) (ik~+

2
I~ 2 l~l 2 2
(ik~+ ..) (ik z- ,~--/,27n) + (ik z- ) (ik~+
2 D1 ',-'-a , 2 D1 ~ [ s ,u2 )
R- =
i~I 2 2 I ~I 2 2
(ik~+ .... (ikz+ (ikz+ ) (ik~+

These eigenfunctions form the branch ~0, co) of the continuous spec-
trum.
J~l 2 2 +
If ~ = 0 then the transmission amplitudes S- are
2 D1 I~ I~D2
equal to zero. while the moduli of the corresponding reflection am-
plitudes are equal to one.

The waveguide functions can be constructed in the standard form

• I
B exp(-~z) exp i(kxx +kyy) , z > O ~,~ > 0
Uo(X'Y'Z) = I B- exp(+~('z) exp i(kxX + ~ y ) , z<O ~ =~-2~-A

One can easily obtain the following dispersion equations :

+~(+ 2
I~12D2 2 el
2 2 ~ + (r ~ +~') I°(12
2 D1
 188

~%~12D2
~+~'+ 2 ~
I~I2 2

2 nI 2 + (~+~') ~2D2

Thus the potential barrier splits the scattering branch of the

spectrum into two parts I-A, °°) and I0, °°) and changing at the
same time the waveguide branch.

4. A zero-range defect near the surface

Let us restrict the self-adjoint operator of the section

1 to the linear set of smooth functions vanishing i n a neighbouzhood
of a point (0,0,-Zo), Zo> 0. The obtained symmetric operator z
O

has the deficiency indices (1,1). The external component of an ele-

ment from the domain of the adjoint operator (~fz~) * has the follo-
o
wing asymptotic behaviour in a neighbourhood of the point z o :
+

u0(x,y,z) ~ u ~0 + u0 + o(RI)
4~ R1

where Rl= (x,y,z+z o) . The boundary form of the adjoint operator

can be written in the standard way

o o
The deficiency element G(k) corresponding to the spectral parameter
k = J - ~ , Im k > O , has the external component of the following type:

i exp ikR I + Go , z<O

4qTR 1
GO= G0 , z > O
exp ikR 1
where G O is the result of scattering of the spherical wave
~R 1
on the surface with the internal structure. The unknown function G O
can be calculated with ths help of the following expansion (seeC5])
~-~ 2~
exp ikR 1 i8~ 1
- de d~ exp i(kxX+~y+kz(Z+Zo)) sin ~, z.> z
#~RI o
o o
 189

where kx = k sin e cos ~ k z = k cos

ky = k sin e cos b°
Plane waves coming to the surface stimulate reflected and transmit-
ted waves. The formulas (1.6) are valid in the case of damped waves
too, because all the functions involved are analytic in the upper
semiplane. For G O we get the following expressions :
the reflected wave
ik
GO = I de I d~ exp i(kxX+kyy-kz(Z-Zo) ) R(k cos ,) sin e

the scattered wave

ik 1 I d~ exp i(kxx+~y+kz(Z+Zo)) S(k cos e) sin e

GO = 8~ 2 de

where S and R are the transmission and reflection amplitudes intro-

duced in Sec.1. Symmetry of the model allows us to transform these
expressions to w/2 Loo

= H (kR' sin ~) exp(-ik(Z-Zo)COSe)R(kcose)sin~ d~

GO 8~T
and -'~/2+ L~o R' = I(x,y, z-z o) I

ik
G O = -- I H (01)(kRlsine) exp(ik(z+Zo)C°Sg) S(kcose) sino de
8~

The internal components can be easily calculated in the same way.

The boundary values GO,
+ G O for the deficiency element are the fol-
lowing : ~7/2 ~ °

G +o = 1 , - ik
G O = 4~ ~ ( 1 + 1 exp(ik cose 2z o) R(k cose) sine de )

0
The asymptotic behaviour at infinity can be investigated by the
method of Brechovskich [5] • If we fix a direction specified by an
angle e, then
exp i ~ l ~ ] ~_
Go ~ R(k cose) ,
~- , e E
4~Y R 1 2 2
exp l 3
GO 4 S(k cose) , e C- ~2' - f ~
4~N 1 2
This deficiency element is the kernel for the external component of
the resolvent of the self-adjoint operator described in Sec.1 .
 190

In order to construct the zero-range potential with an inter-

nal structure, let us choose a self-adjoint operator A 3 acting in
the Hilbert space H 3. We shall restrict it to a symmetric operator
A o with deficiency indices (l~l). The boundary form can be exp-
ressed in a standard way :

("3o ~3, v3) - ~°3,~3o'3) -- °; ~v3 - °3 v; =_ ~3,~37 ~° (~.~)

Next we define the symmetric operator N 0 = ~ z 5 ~ A30 acting in the
o
Hilbert space L2(R3)@L2(R2,H1)@L2(R2,H2)~3 .Its deficiency indices
are (2.2), boundary form is the sum of the boundary forms :
~, ~'~
_ ~.~, ~,; ~? _- I~,
~ ] e~ in ~,.~?
We can define the sought self-adjoi~t operator by restricting the
adjoint operator ~ ; to the linear set specified by the boundary con-
ditions :
+ +
u 0 = ~ u3
-u 7 -_ ~ uo
The continuous spectrum of N~ consists of eigenfunctions of two
types, the scattered waves and waveguide functions. The external
components of all eigenfunctions of the continuous spectrum can be
expressed in the following form :

U~o(~) :Uo~(,~) +9%(,~)

where UO
~ ( ~ ) are the eigenfunctions of the self-adjolnt operator
~ . The boundary values ~ of such function are the following :

l) u o~ + = 0 u~" = exp-ikzz o + R exp ikzZ o , k z ~ 0

u0 = S exp-ikzz o , k z < O

2) u0
~+ : 0 ~- : B- e x p ~ z o

Hence we can calculate the amplitude ~ from the boundary conditi-

ons and from the standard correlation between the boundary values
of the internal e e ~ n e m t
u;
J

where D 3 is the Schwarz integral of the operator A3 .

 191

The amplitude ~ j s_ expressed as follows

-u 0
÷
D3( GO +

The external components of the scattered waves has the following

asymptoties at infinity
exp ikR 1
u~(~) ~- uo (>~) + f(~,e)
~- R1

where f(k,~) = ~ R ( k cose), e C I ~ , 31 ~2

f(k,e) = S(k cose), e ~ I - 5T2, 2]

and the function f(k,~) is the scattering amplitude.

The eigenvalues of the discrete spectrum correspond to soluti-

ons of the equation :

In the case of weak c o u p l i n g , ~ - ~ 0 , the eigenvalues are situated

near the negative zeroes of D3(k ), while all positive solutions
of this equation correspond to resonances.

5. ~ periodic structure on the surface

There are different ways how to construct such a periodic

structure within our approach. One can choose the internal operators
A i dependent on the point (x,y) of the surface in some periodic way.
One can add also a periodic potential g(x,y) into the definition of
the operator ~ i :

~i = - A x y + Ai + g(x,y)

We are going to investigate here the simplest exactly solvable model.

We can replace the constants ~ and /3 in the boundary conditions
(1.4) by arbitrary periodic smooth functions a(x,y) and b(x,y) .
Let a, b be the following exponential functions

a(x,y) = c~ exp i(PlxX + PlyY )

b(x,y) = ~ exp i(P2xX + p2~y)
 192

Let us investigate the spectrum of the waveguide functions. The

external components constructed in a standard way :
I B exp-~z exp i( txX + tyy ) , z>O
Uo(X,y, z) =
B- exp ~ z exp i( txX + tyy ) , z<O
From the boundary conditions we get that the constants B +, B- sol-
ve the standard homogeneous system :
2 2
(~+ , i~l ) B+ + ( ~ + I~I ) B-= 0
2 Dl( ~ - (t - pl) 2) 2 Dl( -'[t-Pl)2i

(I+~
I~12D2 ( ~ -(t-P2)2 ) %i2D2(,~_(t_P2)2))
) B+ - ( i + ~ - B- = 0
2 2
The system has a nontrivial solution only if
icx~ 2
~+ ....... 0
2 DI( h -(t-Pl )~)

i~i2D2 ('~-(t-P2)2)
i+ ~ = 0
2
Let us investigate the solutions of the first equation which lead
to symmetric wavefunctions. The real solutions can be localised only
in the region h ( t 2 (see Fig.3) because ~e = ~ - ~ - ~ . In the
case of weak coupling between the external and internal channels, the
solutions lie near the zero surfaces of Dl(Z-(t-p) 2) (marked by the

~(PI" ~=0
/

Fig 3
dashed lines ), which are parabolas of the following type :
 193

~- ( t - p )2 = const
These parabolas are symmetric with respect to the axis t=p. Thus the
branches of the solutioms are izfinite (marked by the full line),
and they are not symmetric with respect to the energy axis.

If a waveguide function propagates in some direction, there is

no waveguide function of the same energy propagating in the opposite
direction. The same is true for the antisymmet~ic functioDs.

Beferences
1. B.S.Pavlov, Teor. Mat. Fiz. 59 (1984), 345-353
2. B.P.Beliosky, Proc. Acad. Scl. USSR 278 (198@), 1090-i094
3. W.Karwowsky, in this volume
4. Yu.A.Kuperin,... : Properties of few-body quark-hadron systems,
part II, Vilnius 1986, pp.28-73 (in Russian)
5. L.M.Brekhovskikh : Waves in sandwiched media, Namka, Moscow 1973
(io Rassiao)
SPECTRAL PROPERTIES OF THE LAPLACIAN WITH ATTRACTIVE

BOUNDARY CONDITIONS

Manfred Schr6der, Sektlon Mathematik, Karl-Marx-Unlverslt&t,

Ka.rl-Marx-Platz 4-I0, Leipzig 7010, German Democratic Republic

O. Introduction

Let a quantum p a r t i c l e be e n c l o s e d in a container G with

the boundary ~G. The d y n a m i c s of the particle can be d e s c r i b e d

by t h e H a m i l t o n l a n

H = H0 + V, (i)

where H is the Hamiltonian of a particle in a container with

o
indifferent boundary: I n an a p p r o p r i a t e system of units, one

has H ° = - ~ on L2(G) with Neumsnn b o u n d a r y conditions (b.c.),

and V i s the potential corresponding to the boundary forces.

It can be shown / 1 3 / that if G is the half-space ~ d÷ =

~d-1X ~+, and V has t h e p r o p e r t y that the function Q(xl,,.. ,

Xd_ 1) = S V ( x 1, . . . . Xd)dX d b e l o n g s uniformly locally to

O
Lp( ~ d - 1 ) ~ w h e r e p ~ d - 1 and p ~ 2, then t h e s e q u e n c e H° +

nV(xI,...,Xd_l,nxd) converges for n~o inthe norm-resolvent

sense to the operator HQ = - ~ on L 2 ( ~ ) with b.c.

xd I Xd=O I Xd=O
(2)

(for d = i, cf. Ref, 1). Since the boundary forces are usually

of short range, it seems t o be s e n s e f u l to consider the Hamil-

tonian HQ i n s t e a d of (1).

If the function Q is non-negative, then the spectrum o f HQ

coincides with that o f Ho, even i f we add t o both Hamiltonians

 195

some e x t e r n a l potential w. Such b . c . we c a l l repulsive. Thereas

if there are sufficiently l~rge regions, w h e r e ~ becomes n e g a -

tive, then there occur additional states - surface states -

with wave f u n c t i o n s cencentrated near the boundary. In this case

we c a l l the b.c. attractive.

In the following we w i l l consider three typical situations=

In Section I, we w i l l investigate the effect of local impurities

of a neutral surface. Section 2 is dedicated to ergodic surface

potentials which can be c a u s e d e . g . by t h e m o l e c u l a r forces at

the surface of a crystalline o r amorphous solid. In the last sec-

tion, we c o n s i d e r surface potentials varying remarkably only in a

macroscopic scale.

t. Localized surface potentials

If the dynamics of a quantum p a r t i c l e in a container G is de-

scribed by t h e H e m i l t o n i a n H = H~ + W, w h e r e Q c o r r e s p o n d s to the

surface forces and W i s an e x t e r n a l potential° then the spectrum

o f H can be f o u n d by i n v e s t i g a t i n g the solvability of the equation

Hu = Eu f o r any r e a l E. G e n e r a l l y , this requires methods d e s c r i b e d

d
in /2/. Zn a q u i t e simple way i t can be done i f w = 0 and G = 3+

(cf. /3/7. One o b t a i n s the following

Proposition 1: L e t Q be r e l a t i v e l y bounded w i t h respect to (_~)1/2

with a relative bound l e s s than 1. Then 0 > E E 6(HQ)(~pp(HQ),

~ess(HQ)) if and o n l y if 0 e 5(KQ,E)(~pp(KQ,E), ~ees(Kq,E)),

where ~ ( ~pp, ~ees ) denote the spectrum (pure point spectrum,

essential spectrum), and KQ, E = ( - ~ - E ) 1 / 2 + Q.

If the condition on Q i s satisfied then ~(HQ) is bounded from

below /6/. A sufficient condition on Q t o be f i t for Proposition 1

is that Q E L uniformly locally, with p > d-l, p ~ 2 /3/. We b e -

P
gin with t h e case Q ( L p . T h i s i m p l i e s that 6ess(KQ,E) = ~(Ko,E),
 196

hence, by P r o p o s i t i o n 1, Gess(HQ) = [ 0 , oo). Now l e t us d e n o t e by

N(A,t) t h e number ~f e i g e n v ~ l u e s (counting multiplicity) of the
self-adJoint operator A, w h i c h do n o t exceed t h e real number t .
From P r o p o s i t i o n 1 it follows that, for E < O, N(HQ,E)=N(KQ,E,O ) .

F o r d = 2, one can e s t i m a t e N(KQ,E,O ) by u s i n g the Birman-

Schwinger principle (cf. /4,11/).

p ropogition 2: L e t Q E L 2 ( ~ ) and ~ . ~ L 2 ( ~ ) ~ Lp(~), 1 < p ~2, and

let P + ~q = 1. Then t h e r e exists a constant Cp such t h e t for E < 0

N(H~,E) { Cp I1~_1I~( - E ) - 1 / q . (3)

(Here and f u r t h e r o n we d e n o t e by F t the positive (negative) part

o f F: F = (IFI - F)/2.)
In higher dimensions, one can use an a n a l o g y of t h e C w i k e l -
Lieb-Rosenbljum bound /10,11/ for the same p u r p o s e : For Q E
Lp(~d-1), with d > 2 and p > d - l , we s e t

~Q(q) = rues { x E ~d-1 ~(x) ~ q} , (4)

where mes d e n o t e s t h e Lebesgue m e a s u r e . Now we s t a t e (cf. /6/)

Proposition 3: L e t ~ E L p ( ]R d - l ) , with d ~ 2 and p • d - l . Then t h e r e

a r e two c o n s t a n t s Cp, d and C;,dW such that, for E < 0t

o
N(HQ,E) ( ~ rain { C p , d ( _ q ) d - 1 . C;.d (_E)-1/2 (_q)d} d~Q(q).
-oo (5)
F o r Q_ E Ld_ I f3 L d, P r o p o s i t i o n 3 i m p l i e s N(HQ,E)
Cp,d [I Q- I Idd--1I end N(HQ,E) ~ c ; , d II _li dd C-E) - 1 / 2 .

2. Ergodic s u r f a c e potentials

Now we s u p p o s e t h a t Q is some random e r g o d i c function in the

sense that any measurable set of realizations which is invsriant

under translations Q~(x)~-pQ~(x+y), where y runs over ~d-I or

some lattice ~ da- i = ( e l ~ )X'''X(ad-I ~ ) with a i > O, has p r o b a -

bility 0 or I .
 197

For given L > O, l e t ~ L = ~ - L / 2 , L / ~ d - I , and denote by HQ, L

resp. ~ L the
HQ, restriction of H~~ o n t o ~ L X [0, L] r e s p . A L X ~+,

with appropriate (e.g. Neumann) b . c . at a A L X tO, L] and ~ L X { L}

resp. at ~ / ~ L X ~+, and l e t analogously KQ,E,L
~ denote the re-

striction of KQ,E
~ onto /~L with t h e same b . c . The e s t i m a t e s of
Section 1 suggest that it should be s e n s e f u l to consider the qua n -

tity

n~(E) = N(HQ,L,E) / L d - 1 (6)

in the limit L~ oo . It turns out that the quantity

~nL(E) = N ( H~ Q~ , L , E ) / L d - I ~
= N(K~,E, L ,O)/L d-1 (7)

(cf. /5/) has the same l i m i t as nL(E). I f t h i s l i m i t e x i s t s we

cell it the density of surface states. But when does i t exist?

One p o s s i b l e answer is the

following /5/
~d-1)
Proposition 4: L e t Q E Lp( uniformly locally, with p • 2,
p >d-l, and l e t ~ be e r g o d i c with respect to translations in
~d-1 resp. ~ da - 1 "
If E x p ( { Q ~ ( O ) I p) < co r e s p .

a1 ad_ 1
Exp( S dx I " ' " S dXd-1I ~ ( X ) I p ) < Oo
o o
then there exists a non-negative, monotonously increasing function

na(E) on ~_, such that for all negative rational E almost surely

~c~
lim n (E) = lim nL(E) = nS(E).
L-~O o l L-- co

This result is the analogy to the non-randomness of the densi-

ty of s t a t e s of S c h r ~ d i n g e r operators with ergodic potentials

(cf. e.g. /7,8/ and r e f e r e n c e a therein).

The d e n s i t y of surface states can be e s t i m a t e d analogously as

in the previous section: Let

 198

~(q,) = lira rues ~ x ( AL: Q~:(x) _~ q} ILd-i; (8)

L-~Oo
the ergodicity of Q~ g u a r a n t e e s the existence and non-randomness
of this limit. Then t h e r e holds /6/

Proposition 5: L e t ~ as ' i n P r o p o s i t i o n 4 and C p , d , Cp, d as i n

Proposition 3. Then, for E • 0
o
nS(E) ~< ~ min{Cp,d (_q)d-l,Cp, d (_E)-I/2 (_q)d} d~Q(q). (9)
_oo
If the function Qw i s p e r i o d i c , then the a d d i t i o n a l spectrum

c~used by t h e b . c . is absolutely continuous /4/, and g e n e r a l l y the

co
ergodicity o f Q~0 i m p l i e s that the discrete s p e c t r u m o f HQ i s al-

most s u r e l y empty / 9 / . Furthermore, i n many t y p i c a l examples o f

ergodic Qw ( e . g . potentials of the s u b s t i t u t i o n a l alloy type) the

surface states form a dense pure p o i n t spectrum.

Now we assume H ~ = HQ + W~, where t h e e x t e r n a l potential W~ i s

the restriction of some e r g o d i c (with respect to ~d d =
o r 77a

~ ad - 1 x (a d Z ) , ad > O ) function U~ d e f i n e d on ~d onto ~ +d. If Uw

obeys a s t r e n g t h e n e d ergodicity property, namely t h a t any measu-

rable set of realizations which is invarient under translations

UU(x)~UV(Xl+Y1, .... Xd_l+yd_l,Xd), with y running over ]Rd-1 reap.

77ad-1, has p r o b a b i l i t y 0 o r 1, t h e n t h e methods o u t l i n e d in /2/

lead us t o p s e u d o d l f f e r e n t i a l equations which are e r g o d i c with re-

d-1
spect to ~d-i reap. 77e . Arguments of Refs. 7 and 8 s u g g e s t that

in this case - at l e a s t if Q~ and U w a r e s u f f i c i e n t l y well-behaved

- one has t o e x p e c t non-random surface s p e c t r u m and d e n s i t y of

surface states below t h e infimum and i n t h e gaps o f the s p e c t r u m

of -A + U ~: F o r E l • E2 l y i n g in the same gap one w i l l have

nS(E2) - nS(E1) = l i r a (N(HQ, L +W ,E2) -

L-*OO

N(HQoL + w u ' E i ) ) / L d ' l " (10)

If t h e domain G i s o f the kind ~d-m X ( ] R + ) m, w i t h m¢ d,

 199

then there can o c c u r new s t a t e s at the edges o f the domain. If the

potential and t h e b.c. satisfy sufficiently strong ergodicity con-

ditions then one cain d e f i n e a series of densities of edge s t a t e s

which will be a g a i n non-random (cf. /12/). This reflects the fact

that the e l e c t r o n i c properties of real ergodic systems - such as

(macroscopicslly) homogeneous c r y s t a l s , polycrystals , glasses,

alloys etc. - are sample-independent.

3. Macroscopic surface potentials

In this section we c o n s i d e r the ca~e o f surface potentiais

which are varying very slowly, so t h a t t h e y can be assumed as b e -

ing constant in any m i c r o s c o p i c region. Such p o t e n t i a l s can o c c u r

e.g. in metals as an e f f e c t of an e x t e r n a l electrostatic field,

or in containers with walls consisting of several materials. We

start with the simplest possibility: Let the region G be t h e co-

lumn A L X ~ + , with A L as i n the previous section, and c o n s i d e r

the operator H~, L = - ~ inside A LX ~+, with b.c.

u( x , O )
~x d = Q(x/L) u(x,O), x e /\ L' (11)

at the basis and Neum~nn b . c . at ~LX ~+. We n o t e that in this

case t h e r e holds an a n a l o g of Proposition % (cf. /5/). The d e n s i t y

of surface states can be e v a l u a t e d explicitely:

Proposition 6: Let the function Q be Riemann i n t e g r a b l e over A 1.

Then t h e r e holds the asymptotic formula (for E < O)

l i m N(HQ, L , E ) / L d - 1 =
L -.4-oo
m d-I (2~)-d+iAs((~2(x)- + E)+ )(d-1)/2 dd-lx' (12)

where ~ d-1 denotes the volume of the unit sphere in ~d-1.

 200

Proof= We divide the cube ~ L into cubes --'/\~i) of length L/[LI/2j,~"

where the square brackets denote the integer part. On the columns
N
i) X ~+ we define operators H~,L, i and HO
Q,L,i as - ~ inside the
columns, with b.c. (ll) at the basis and Neumann resp. Dirlchlet

b.c. at ~ / ~ L ( i ) x + (but o n l y Neumann b . c . at (~ALN~i))x]R+).

From Dirichlet-Neumann brackettlng arguments (cf. /IO/) it follows
that, for E < O,

N(H~,L,i'E) ~ N(H~,L,E) ~ ~ N(H~,L,i'E)" (z3)

Further, we i n t r o d u c e the piecewise constant functions

Ql(x) - inf {~(y/L): Y E A (i>} for x E / ~ L( i )

and
(i) (i)
Q2(x) = sup { ~ ( y / L ) : y E /\L ) for x E /~L

and t h e o p e r a t o r s H+~ , L , i = H~Nl , L , i and HQ,L,

- i = HQ
O2 , L , i . Since the

eigenvalues o f HN
~,L,i and HD
~,L,i depend on Q m o n o t o n o u s l y / 4 / , we

obta~in

. ÷
(14)
-< i[ ':)"
It i s e a s y t o see t h a t ~symptotically
+
N(HQ,L,i'E) ~ "[d-1 (2~)-d+i (L(QI2,2(x) + E)) (d-i)12, (15)

(with x E A L( i ) ) if Q1, 2 ( X ) ~ - ( - E ) I / , 2 and 0 e l s e w i s e . Obviously,

t h e sums on t h e l e f t and t h e right hand s i d e of (14), divided by

Ld-l, are lower resp. u p p e r D a r b o u x sums f o r the integral in (12).

S i n c e Q has been assumed t o be Riemann i n t e g r a b l e , the proof is

done.

If we now c o n s i d e r the region A L X r O, a L ] , with some a > O,

and a t this region the operator -A with b.c. (11) at the basts

and Neumann b . c . at the other surfaces, then the asymptotic formula

 201

(15), and hence ( 1 2 ) , remains true, independently of a. T h e r e f o r e

we can conclude that, if G l is some simply connected region with

smooth boundary in ~d, GL = LGI, HQ,L = _ ~ inside GL, with b.c.

~n(X) = Q(x/L)u(x), where ~ denotes the inner normal derivative
and O i s a Riemann i n t e g r a b l e function on ~ G1, then there holds
the following asymptotic formula

llm N(HQ,L,E)/Ld-i =
L-.-- oo

~d-I (2 ~)-d+l ~ ((~2(x) + E)+) (d-l)12 dS, (16)

with dS denoting the surface element of ~ G I. This can be proven

by considering only a thin layer ~ near a G I instead of the whole

region GI, and dividing the blown up region L~ l into columns of

width ~L l/2 perpendicular to a G L, where formula (15) holds.

References
l. H. Englisch and P. Seba, The stability of the Dirichlet and
Neumann boundary, Rapt. Hath. Phys. 2,3(1986) 73.
2. L. H6rmander, Pseudo-differentlal operators and non-elllptlc
boundary problems, Ann. Math. 83(1966) 129.
3. M. SchrOder, On the spectrum of Schr~dinger operators at the
half-space with a certain class of boundary conditions,
to appear in Z. Anal. Anw.
4. H. Englisch, M. Schr6der and P. ~eba, The free Laplacian with
attractive boundary conditions, to appear in Ann. Inst. H.
Poincar6.
5. M. Schr6der, Spektraleigenschaften yon SchrOdinger-Operatoren
mit zuf&lligen ergodischen Randbedingungen, in: Proc. 2 nd
Conf. Stochastic Analysis (IH Zwlckau 1996).
6. M. SchrSder, Estimates on the spectrum of SchrOdlnger operators
with attractive boundary conditions, submitted to Math. Nachr.
7. L.A. Pastur and A.L. Figotin, Ergodic properties of the dist-
ribution of the eigenvalues of certain classes of random self-
adJoint operators, Sel. Math. Soy. 3(1984) 69.
8. W. Kitsch and F. Martinelli, On the density of states of SchrO-
dinger operators with a random potential,D.Phys.A15(1982) 2139.
 202

g. H. Englisch and M. Schr~der, SchrSdinger operators with random

boundary conditions, to appear in: Proc. Int. Sem. Localiza-
tion in Oisordered Systems (Teubner, Leipzig 1987).
tO. M. Reed and B. Simon, Methods of Modern Mathematical Physics IV
(Academic Press, New York - San Francisco - London Ig78).
ii. B. Simon, Functional Integration and ~uantum Physics (Academic
Press, New York - San Francisco - London IgTg).
12. H. Engllsch and M. SchrSder, Bose condensation in disordered
systems. II. Surface and bound states, submitted to Physlca A.
13. H. Engllsch, Private communication.
 QUANTUM JUNCTIONS AND THE SELF-ADJOINT EXTENSIONS THEORY

O v m~
P.Exner , P. Seba
Joint Institute for N u c l e a r Research; Laboratory of Theoretical
Physics, 141 980 Dubna, USSR

The a i m of the p r e s e n t e d paper is to d e s c r i b e a simple model

of low-energy scattering on a q u a n t u m point contact.

Let us f i r s t outline the p h y s i c a l background of the model.

For a metallic contact,the common wishdom suggests a linear
relation between the applied voltage and the current (the Ohm's
law). This remains true, however, only if the size of the contact
is l a r g e enough. Once its d i a m e t e r becomes comparable with the
mean free path of the electrons in the metal scattering effects
appear which add a nonlinear contribution to the current.

Effects of this type were measured first at K h a r k h o v in a

pioneering experiment by Igor Yanson [i] , w h i c h gave rise to a n e w
research branch called now "Point Contact Spectroscopy". The
measured nonlinearity represents usually a few p r o m i l e to a few
percent of the total current and is v i s i b l e in the differential
resistance dU/dI. The second derivative d2U/dI 2 e x h i b i t s typically
a more complicated shape with peaks corresponding to the
electron-phonon interaction in the m e t a l involved.

There are two b a s i c types of the p o i n t contact experiments.

In the first of them, dubbed a pressure-type contact, a sharply
tipped wire is adjusted by a screw against a flat metallic

On leave of absence from the Nuclear Physics Institute,

Czechoslovac Academy of Sciences, 25068 ~e~ near Prague,
Czechoslovakia.
w*
On l e a v e of a b s e n c e from Nuclear Centre,Charles University, V
%" . ~ S
Holesovlckach 2, Prague 8, C z e c h o s l o v a k i a
 204

surface. The second type consists of two thin metallic films

separated by an insulating (oxide) layer which is perforated at
one point. In b o t h two types the contact region is t y p i c a l l y a few
micrometers in d i a m e t e r .

The "microscopic" theory of these contacts is a v e r y complicated

matter a n d we are not going to d i s c u s s it here. We shall restrict
our attention to the long-wave-limit situation only, when the de
Broglie wavelength of the electrons is m u c h l o n g e r than the contact
diameter. In t h i s case it is r e a s o n a b l e to e x p e c t that the contact
scattering would not depend on the detailed shape of the contact
region, but rather on the global geometry of the experiment. Our
aim here is to show that the "free" quantum mechanics together
with the geometry of the contact are alone responsible for the
measured shape of the current-voltage characteristics in the
long-wave limit.

I. A model for the thin-film contact

He consider here the simplest possible quantum-mechanical
model [2] in w h i c h a free electron moves on a manifold consisting
of two planes connected at one point (Fig.l). The state Hilbert
space ~ of our problem is a sum
of the spaces corresponding to
the upper and lower plane
respectively: ~ = L2(~2)~L2(~2) .
To construct the quantum / 2 ~
Hamiltonian
operator
we start with the /z/~
z -j~
•O
//
H = Ho, I Ho, 2 (i i) Fig.1

where Ho, j = - A w i t h D(Ho,j)= Co(~ \ 0 ) , j = 1,2, assuming the

connection point to be placed at the origin on each of the planes
; the particle is s u p p o s e d to be free outside the connection
point. The deficiency indices of the operators H are known to
o,j
be (i,i) [3], hence the deficiency indices of H are (2,2) and it
0
has a four parameter family of self-adjoint extensions. To
construct these extensions we proceed in a standard way. The
 205

+ *
deficiency subspaces 9~- = Ker (H ° + iD) are spanned by the vectors

±
~ , k = 1,2, where
+
~k = (fo "0) ' e~ = (O'fo) (1.2)

with
i~/4
f (x) = H(I~(~ x ) , ~ = e (1.3)
o o

and ~ are complex conjugated to ~. Introducing the polar

coordinates in e a c h of the two planes and decomposing the Hilbert
space

CO

L z ([Rz) = m=e_coL (~
.+ , rdr) ®
{Y}m LLn (1.4)

where
y (~o) = (2r~) t / 2 e imp°
n'l
(i. 5)
we c an decompose the operators H as
o,j
CO

H = e h ®D (1.6)
o,j m---CO m,j

where
2
dz
- 1 d m ,
- ; D(%.? (1.7)
hm"J dr z r dr r
All the operators hm, j a r e e.s.a, for m ~ 0 (see [3], Sec. X.1).
We get therefore

Proposition l:
All the self-adjoint extensions of the operator H are of the form
o
H = K eh , (i. 8)
U U
where

(1.9)

and Ku is a s e l f - a d j o i n t extension of the operator

KO = (ho, i ® 0)~(0 ® ho,2 ) (i.i0)

with the domain

D(Ko) = ~ = (f~,f2): fj(x)=fj(,x,), fj E Co~(~+\ {0 5 )}o

(Here U d e n o t e s a 2x2 unitary matrix U which is usually used in

the yon Neumann theory to p a r a m e t r i z e self-adjoint extensions.)
 206

Hence it is o n l y necessary to f i n d the operators K u. The most

simple way to do it is to use b o u n d a r y conditions. The deficiency
functions are, however, singular at the origin. Therefore the
corresponding boundary conditions must be written in terms of
regularized boundary values [4]

Lo(f) = lim f (r) ; L (f) =r~olim [f(r) - Lo(f) in r ]

~o fn ?

Proposition 2
Every self-adjoint extension Ku is uniquely specified by the
following boundary conditions: f = (fl,f2) ~ D(K u) iff

f~ ~ [ f E L 2 ( ~ + , r d r ) ; f , f ' E A C ( ~ +) and f '' + ?1 f ' E L 2 (~+,rdr)

for i = 1,2 and

(i) L o(fl) = aL o(fa ) + bL,(fa)

L I(fI) = cL o(fa ) + dL~(f2 ) (i.ii)

where the coefficients are given by

-i
a = u 212(uiI-i)+ ~(det(U) - U2a)]

b = __2i ui:[l-tr(U) + det(U)]

(i. 12)
c = ,i
7 ui:[2"+~ tr (U) + (~)z det (U)]

d = ui:[x(l-u22) + ~ (uil- det (U)) ]

and u are m a t r i x elements of a n o n d i a g o n a l unitary 2x2 m a t r i x ,

1.1
or
(ii) L o(f,) = AL i(f,)
(1.13)
L o(f ) = BE (f)

1 2i
with A , B ~ ~. Here 2~ = Li(f o) = ~ + ~-- (Z - in 2) and Z = 0.577..
is the Euler's constant.
Proof: See [2] .

Remark: The extensions defined by the b o u n d a r y conditions (ii)are

physically not interesting because they lead to H a m i l t o n i a n s which
 207

have are a direct s u m and h e n c e describe a system in which the

two p l a n e s are completely separated. In w h a t follows we restrict
ourselves therefore only to e x t e n s i o n s given by (i) .

He suppose that the two planes are physically equivalent,

i.e., we restrict our attention to the Hamiltonians Hu which
commute with the m o d i f i e d parity operator P exchanging the p l a n e s

P : (fi,f2) --~ (f2,fl) ; (fi,f2) ~ ~ (1.14)

It can be shown [5] t h a t all the H a m i l t o n i a n s which fulfil

PH u c H u P (i. 15)

form a two p a r a m e t e r subfamily corresponding to s y m m e t r i c matrices

U. T h i s subfamily can be described by the b o u n d a r y conditions (i)
from the Proposition 2 with the coefficients a,b,c,d given by

a = -d = cos (/9) + cos(F.) - sin(~)

sin ([)
(i. 16)
b = 2 I/2 cos (/9) + cos(~)
sin (/9)

c = 2 I/2 sin(~) - cos(/9)

sin (/9)

where /9,~ ~ (0,2hi are two real parameters.

Let us n o w investigate scattering of the p a r t i c l e on the

connection point. Our aim is to f i n d the transmission probability
from the u p p e r to the lower plane. Using the time-independent
approach we start with the function f = (fl,f2). w h e r e

fl (r) = H{a>o (kr) + A(k) H(*>

(kr)o
(i. 17)
fz (r) = B(k) H(*>o(kr)

and d e m a n d it to b e l o n g locally to D(Hu). A simple calculation

yields the coefficients A(k) ,B(k) which are expressed as
 208

2
k ~ k 2]
c -2a (7 + in y) - b [ ~-- +(;~ + in y )
A (k) =
k n k 2
C - 2a ( ~ + ~ + in ~ ) - b ( ~ + ;~ + in ~ )

in
B (k) =
k ~ k 2
c - 2a ( ~+ 7 + in ~ ) - b (~-~+7 + in~)

A(k) and B(k) are the reflection and transmission coefficients

respectively. It c a n be easily seen that they fulfill

[A(k)~ 2 + ~ B ( k ) ~ z = i. (i.18)

Thus we have obtained a nontrivial particle transmission from the

upper to the lower plane.

Before comparing the obtained transmission coefficient with

the experimental results we describe ~n short one more type of
a quantum point contact model, which corresponds to the pressure
type experiments.

2. A m o d e l for the pressure-type contacts

As already mentioned, in the pressure-type experiments a thin

wire is a d j u s t e d against a flat metallic surface. Fixing the basic
geometry and supposing that the linear dimension of the contact is
zero we obtain the simplest model in w h i c h the electron moves in a

Jl-/
manifold consisting of a half-line connected to a plane (Fig.2) .
The corresponding state Hilbert space is n o w given as
= L2 (~ -) ~ L2 (~2).

Such a quantum system has been

discussed in [6] and we summarize
here the results.
Since the electron is again
supposed to be free outside the
connection point the starting
operator H is given by (i.i) Fig.2
o
with
 209

= - d ; D (Ho.i) = c ~0( m - \ {o~) (2.1a)

Ho.i dx 2

= co ~2
HO,z - A ; D(Ho. 2) = Co( \ {0~) (2.1b)

The operator H has deficiency indices (i,i) and the corresponding

0
deficiency subspaces are spanned by v e c t o r s

t- t- +
0 s = (f ,0) ; 02 = (O,fo) ; 0 = 0

where fi(x) = 8 I/4 e x p ( Z x ) , and fo is given by (1.3).

Using the v o n N e u m a n n t h e o r y we get a four-parameter family of
possible self-adjoint extensions of the operator H o, which is
parametrized as u s u a l l y by 2x2 u n i t a r y matrices U. Comparing to
the p r e v i o u s case, one cannot eliminate some of the possible
Hamiltonians using the symmetry of the manifold. Physically it
is, however, reasonable to use only the Hamiltonians leading to a
time-invariant particle dynamics. Mathematically speaking this
means to use only such extensions Hu of H° for which the
equivalence

f e D(Hu) (=~ f ~ D(H u) (2.3)

holds. It can be easily shown that this condition leads to a

three-parameter family of a d m i s s i b l e Hamiltonians corresponding
to the m a t r i c e s U whose elements fulfil u = u
i 2 2~"
These extensions are d e s c r i b e d by

Proposition 3
Every self-adjoint extension fulfilling (2.3) is specified
uniquely by the following boundary conditions.

If f = (fl,f2) belongs to D ( H u) then

Lo(f z) = Afi ( O ) + Bf i ( O )
(2.4)
LI ( f ) = Cf i ( O ) + Df (O)
i
where the p a r a m e t e r s A,B,C,B are g i v e n by
 210

81/4 sin(~+ --4 ) - cos(/9) sin(a + -4)

A = -r~-
sin (/9)

81-'4 sin(~) cos ((9) - sin (~)

B = m
sin (/9)

[2
1 + 2(z-ln2)] (cos~ +sinacos/9)+[21 _ 2 ( z _ i n 2 )] (sin~ +cos~cos/9)
C =
8 I/4 s i n/9

1
8 I/4 (cos~+ cos~cos/9) + ~(z-in2)(sin~ - sin~cos/9)
D = - u
2
sin/9

with a,~ e [0,2-] and [9 E (0,2~) .

Let us now investigate the t r a n s i t i o n from ~ - to ~a in the

framework of £ime-independent scattering theory Using the
boundary conditions (2.4) and r e q u i r i n g the function f = (~i,~2)
with

Lkx -Lkx
~i (X) = e + A(k) e
(2.5)
~=(r) = B(k) H <i~ (kr)
o

to b e l o n g locally to D(Hu) , we find

2_~i[[in ~+ k 7][A + ikB 1 -C - ikD] + A + ikB

A (k) =
n2-~i[[in k5 + ;~][-A +ikB 1 +C - ikD] - A + ikB

(2.6)

2 i k (BC-AD)
B (k) =
~2--i[ [In 5k + 7 ] [ - A +ikB] +C - ikD] - A + ikB

Here A(k) and B(k) are again the reflection and transition
coefficients, respectively.
 211

Let us now turn to the e x p e r i m e n t and c o m p a r e the calculated

quantum transition coefficients with the m e a s u r e d nonlinearity of
the c u r r e n t - v o l t a g e characteristic.

3. C o m p a r i s o n with the experiment.

Before comparing our results w i t h the e x p e r i m e n t a l data we have to

relate the quantum transition coefficients to the measured
resistance of the point contact. This can be done using a formula
well known from the electrical transport theory [7]. If the
metals involved have the same Fermi energies then the current
through the contact is given by

i =_ 2_~
e I T(E) [fT(E) - fT(E - eU) ] dE (2.7)
o
where e is the e l e c t r o n charge, U is the a p p l i e d voltage and

E - R -i
f?(E) = [ i + exp[ F 11 (2.8)
kT

is the m e t a l electron-gas density at the t e m p e r a t u r e T and Fermi

energy E F. T(E) is the q u a n t u m transition probability which is
given s i m p l y by
T(E) = ]B(kZ)[ 2 (2.9)

where B(k) are the transition coefficients calculated in the

Sections 1 and 2, respectively.

The formula (2.7) becomes particularly simple in the zero-

temperature limit, when fT(E) becomes a step function
E +eU
F

I = 2e ; T(E) dE (2.10)
E '
F

Differenciating it by U we get f i n a l l y the sought formula for the

differential resistance

dU -i
(2.11)
 212

To be just we have to m e n t i o n at this point that the formula

(2.11) has been challenged, however, the alternative proposed in
Ref.8 differs only by a constant which is n o t important for our
purpose here.

In the models described in S e c t i o n s 1 and 2 there are two or

three adjustable parameters, respectively, with the help of which
we can fit the basic nonlinearity of the current-voltage
characteristic measured in the experiment. In order to give an
example we plot on Figures 3 and 4 the resistance corresponding
via the formula (2.11) to the quantum models constructed in
Sections 1 and 2 . The same plots are transformed on Figures 5
and 6 to the logarithmic scale. One has to c o m p a r e these curves
with the voltage dependence of the resistance measured in the
point-contact experiment [9] w h i c h is p l o t t e d on the Figures 7 and
8. One can see that the shape of the resistance obtained in our
models fits well the experimental curves. Moreover, the resistance
curves obtained in the two-plane and pressure-type models are very
similar and are in fact determined only by the geometry of the
manifolds involved.

On the other hand the models under consideration cannot

describe a more complicated structure observed in the second
derivatives of the current-voltage characteristic which is
connected with the particular electron-phonon interaction in the
metal used. One can, however, add an internal Hilbert space
Ln
to to the state Hilbert space ~ and model in such a way for
instance the polaron states localized in the contact region.
Constructing the self-adjoint extensions in the larger Hilbert
space one obtaines a more complicated structure of the transition
coeficients and hence also a more complicated structure of the
resistance plot. Adding the internal space it is possible to
mathematically combine the global manifold geometry, which is
responsible for the dU/dI plot with the detailed electron-phonon
interaction being responsible for the complicated structure of the
 213

dZu/dI 2 plot. The m o d e l obtained in s u c h a way is, however, very

similar to the zero-range interaction models described in this
volume and we omit the details here.

References

1 , I.K.Yanson: Zh. E k s p . T e o r . F i z . 66 (1974) 1 0 3 5

v
2. P.Exner, P . S e b a : L e t t . M a t h . Phys. 12 (1986) 193
3. M.Reed, B.Simon: Methods of M o d e r n Mathematical Physics II
Academic Press, New York 1975
4. W.Bulla, F.Gesztesy: J . M a t h . Phys. 26 (1985) 2520
5. P.Exner, P.Seba: A simple model of thin-film contacts in two
and t h r e e dimensions. Czech. J.Phys. B38 (1988) in p r e s s
6, P.Exner, P.Seba: J.Math. Phys.28 (1987) 386
7 K.C.Kao, H.Huang: Electrical Transport in Solids. Pergamon
Press, N e w York, 1981
8° R.Landauer: Phys.Lett.A. 85 (1981) 91
9 A.G.M. Jansen, A.P. van Gelder, P. W y d e r : J.Phys.C 13 (1980)
6073
 214

dU /i ///
</////
\
Ev =

= i;
1 , /9 =

1.5;
1,

2; 2.5; 3

i 2 eU

Fig.3: The resistance plots obtained in the thin-film model

for various self-adjoint extensions.

dU
dI

Er = 1 , ~ = .i, /9 = i,

= 0.i; 0.2; 0.3; 0.4; 0.5

.......... _

..... I I ! I

i0 20 eU

Fig.4: The resistance plots obtained in t h e pressure-type model

 215

dU
aT

Er = 1 , /3 = i,

= i; 1.5; 2; 2.5; 3

iiiiii
Fig.5: The resistance plots obtained in the thin-film model
for various self-adjoint extensions: the logarithmic scale

%
dU %
dI

/ E
F
= 1

= 0.i~
, ~ =

0.2;
.i,

0.3~
{9 = i

0.4~ 0.5

gl
/
m

.,-.1
.Q I~ II04

I . | ~ i

i0 20 in (eU)

Fig.6: The resistance plots obtained in the pressure-type model:

the logarithmic scale
 216

a)
Au'- 0.03°Io Mg contact
R ~ 2.1~, T= 1.5K
1-
.m
0
in
.~
"o

E
If)
c~

{ voltage (mV) 10

b)

I--'1

o
=

i.--I

0 2 4 6
Voltage U [mV]

Fig.7: A t y p i c a l r e s i s t a n c e plot obtained in a thin-film point

contact
a) the l o g a r i t h m i c s c a l e
b) the l i n e a r s c a l e
 217

Copper contact
R=5.7.o. ,T =1.2K

dUldl

15

F•
c
dU 2

Jm

0 5 10 15 20 25
Voltage [mV]

Fig.8: A typical resistance plot obtained in a p r e s s u r e type point

contant experiment
 THE EXTENSION THEORY AND DIFFRACTION PROB~Jk~S

I. Yu. Popov
Department of Mathematical
and Computational Physics, Physical Faculty,
Leningrad State University, 1 Maya lO0,
Petrodvoretz, Leningrad 198904

A model describing the scattering problem on a surface with a

small opening or a narrow slit is discussed in the framework of the
extension theory. This approach simplifies the problem considerably
and makes it explicitly solvable in many cases. A comparison to the
"real" diffraction problem is also presented.

Most of interesting diffraction problems cannot be solved in

an analytic form! let us recall boundary problems for Helmholtz
equation in domains with small openings, slit antennas, a circular
resonator with a running wave etc. From the physical point of view,
however, it is evident that the wave dynamics in such complicated
domains is a superposition, in some sense, of the "local" dynamics
relative to components of the domain, and that the spectral proper-
ties of Laplace operators in such domains depend on the spectral
properties of the "local" operators and on the quality of the chan-
nel coupling. The aim of the present lecture is to give an exact
meaning to these heuristic statements in the framework of the ex-
tensions theory.

1. The Neumann problem

Let ~ in be a compact domain in R 3 with a smooth boundary,

ex
= R 3 \ ~ i n . Consider the Laplace operators _ ~ i n , _ • e x in
 219

C2 in C2 ex with the Neumann boundery condition. Let us restrict

these operators to the linear sets of smooth functions vanishing
neer einPOlnt ~x~'~ ~ . The restricted operators will be denoted
as-d O , -A

Lemma fl]: The operator - A 0in has deficiency indices ( l, l)

The deficiency element which corresponds to the complex point ~ 0
coincides with the Green's function of the internal Neumann problem
Gin(x, Xo, kO) which has e pole at x O.

This lemma can be obteined from the esymptotic behaviour of the

Green's function near the pole on c~'~ :

Gin( x, Xo, k ) - 1 H 1
+ ~ In +
27Ulx - Xol ~" Ix - %1

I~ n (xo)l 2
+ c i n ( k o, x o) + o ( 1 ).
+ (x-), o) >--
n (Xn-A) (k n -k o)

Here H is the Gaussian curvature of the boundary at the point Xo,

ere the eigenfunctions of _ ~ i n corresponding to the eigenve-
@n
lues • n,and ~ 0 is a fixed value of the spectrel peremeter.

An analogous formule holds also for the external domein,

i H I
Gex( x, x O , k ) = - ~ In +
2~Ix - Xol 4~ Ix - Xol

~ ex( Xo,]))12
+ (X-A o) d? + CeX(Ao, Xo) + o ( I )
(~I 2 -),) ( ~ I 2 - ~ 0 )

where ~ e X ( x o , ~ ) ere the scettered weves in q-~ex. The assertion of the

lemme is velid for _alex too. It is important thet the singular
perts of these 8symptotics do no t depend on the spectrel parameter.

To construct en extension of - ~ 0 ' it is necessary to describe

the domain of its edjoint operator. For this purpose it is conve-
nient to introduce a "reel" basis W + W- in the direct sum of
 220

deficiency subspaces in the following way: If h is the normalized

deficiency element of - ~ O (for the sake of simplicity, we drop
temporarily the superscript "in" or "ex"), we set

W+ = ( - A - i) -I ( - A ) h, w'= (-n- i) -~ h.

In fact, we replace the basis {G( x, Xo, i), G( x, Xo, -i)} by the
new basis {W- = Im G( x, Xo, i)(Im G( x, Xo, i)) -1, W + =
= Re G( x, Xo, i)}. The domain of the adjoint operator can be
represented in the form:

D(-A 0) = {u: u = ( - a - i) -l v + a ÷
u w+ + a~ W-, v J- h}.

The boundary-form value for u, v ~ D ( - ~ O ) equals to

• _-'; +~).
J [u, v] = ( - A 0 u, v) - ( u, - A 0 v) = 2 (a u a v - a u a v

One can prove the last equality in the abstract form taking into
account the relations - ~ 0 W+ = W- and - D O W- = -W +. Alternatively,
one can integrate by parts in our concrete case using the asymptotic
formulas

i H i
W+ - + in + O ( i ),
2~Ix-xo~ ~zr Ix-xo~

W-=l+ o(1).

For any g ~ O, we get

~u 9v
J [u, v ] = ( NB-- V - ~ an ) ds,

where 2 ~ = ~x: Ix - Xol = g, x m Q } . The limit £ ~ 0 yields the

above formula.
in ex
Consider now the operator - ( A O + • 0 )" Its boundary form
coincides with the sum of boundary forms of internal and external
operators.
 221

The domain of a selfadjoint extension is a linea2 set of func-

tions, for which the boundary form vanishes. There exist an ex-
tensive family of boundary conditions leading to selfadjoint ope-
rators, but we restrict further on to one of them. It is determined
by the conditions

a -'in = a -'ex, -a +'in = a +'ex ( 1 )

in ex
Restricting the adjoint operator - ( A 0 + ~ 0 ) tO the domain
containing functions which fulfil the boundar~ condition ( 1 ) we
get a selfadjoint e x t e n s i o n - ~ of - ( A 0in + A~x).

We can construct the S-matrix corresponding to this operator.

First of all we note that the Green's function has a form:

in in in
Gex( x, Xo, k ) = ( k - A O) ~ Gex( x, s , k ) Gex( s, Xo, kO) d s

in in
+ i Im Gex( x, Xo, k O) + Re Gex( x, Xo, kO) ,

Im ~ 0 ~ 0 .

It is now evident that there exist a limit

in in in
lim (Gex( x, Xo, k) - Re Gex( x, Xo, k 0 )) = D ex ( k ) .
x-~x 0

Here D in, e x ( ~ ) can be represented as a Schwartz integral, and

we get

in
1 ~ n in 2
Din(x) : X( ) l~n ( Xo)~
n ~in - ~ ( ~nin _ ~o ) 2 + 3 o2

XO =6"0 + i ~ O ,
 222

~0

Dex(A) = (221-) -3 [ q'-~ds ((s -A,) -I -

0

s j I~ex( Xo,~qT)l a ~9.

Using the usual anzatz for the external part of scattered

waves,

+( x, ~,~,) =+ e x ( x , O , } . ) + ~:( Xo, O , k ) ~ x ( x, Xo,~)

and using the boundary conditions ( 1 ), we obtain

x0 ~
Gin( x, Xo, k ) , x a Q i n
oin(X) + pax( ~ )

q ' ( x, 9 , ~ . ) = +ex(x,O,A)

+ex( X 0 m 9 t A )
Gex( x, Xo,A), x e Q ex.
Din(~) + Dex( X )

Hence the S-matrix is given by

s(o~,9, k) = S e x ( ~ , k)

± k +ex( Xo,
9 ,l)+ex( Xo,~,k) k2
2 ZC(Din(~) + Dex(~))
 223

Here Sex(co, ~, k) is the S-matrix for the unperturbed external

problem. This simple formula allows us to find the discrete spect-
rum and resonances of - ~ . If the internal and external domains
are weakly connected, Lk O ~ - ~ , the resonanoes can be approximately
calculated (this situation corresponds to the small-opening case-
see below). The resonances appear from positive eigenvalues X
n
of the unperturbed internal operator and fulfil

i I~n(Xo) t
k= kn +

2 I1%eX(xo ,- )ll ,2(xl) (1, Cl n(Xo) lnl; o )

+ o ( in -a II, ol),

where k n = ~ n " It is important that we obtain also an estimate

for the imaginary part of the resonanoe, which characterize the
lifetime of this state. Within the standard approach, it is usually
a very difficult problem.

It is interesting to compare our model with the real diffrac-

tion problem on a surface with a small opening. Provided the exten-
sion is fixed, there exists only one free parameter: the point 9%
0
for which we have constructed the deficiency elements. It is rea-
sonable to require that the flux through the opening in our model
equals to the main part of the real flux as the size of the hole
approaches zero. In the two-dimensional case the comparison has
been made for many particular diffraction problems (for example,
two halfplanes connected through the interval, a half-elliptic
resonator connected to a halfplane etc.). It appears that the above
conjecture is true [2] if we choose ~0 ~ with k 0 ~ 2i d -1 e -~
where d is the size of the hole and ~ is the Euler's constant.
Later this conjecture has been proven by Gottlieb [3].

2. The Dirlchlet problem

It is interesting to construct such a model also for the Di-

richlet boundary condition. Unfortunately, the approach described
above leads to a trivial model. The reason is that the restriction
 224

of the Laplace operator to the set of functions vanishing in a

neighbov_rhood of x 0 leads to an essentially selfadjoint operator.
Nevertheless, we can build a non-trivial model, adding to the usual
Hilbert space L 2 solutions of the Helmholtz equation. These solu-
tions (for instance, a dipole or a higher multipole solution) can
become the deficiency elements of the restricted operator. This
approach leads, however, to an indefinite-metric space [4].

There exist another way connected with the mixed boundary-

condition problem

~u
u) I = O.
( ~n @O

The asymptotics of the corresponding Green's function near the

pole x 0 at the boundary is similar to the previous one [5]:

in
Gex( x, XO, A ) = ( 22T~x- Xol) -I +-

+ (4F) -1 ( H + ( e ~ ) - i d ) in (ix - Xo O - 1

+o(i) .

This property allows us to proceed with the scheme described

above. Then we require O" to go to infinity obtaining in this way
some results about the perturbed Dirichlet problem. It is crucial
for the limit procedure that the Green's function has the following
properties :

Theorem:

lira G ~ ( x, y, k) = G D ( x, y, k); x, y ~ ~Q,

9G D
lira O ' G ~ ( x, y, k) = ( x, y, k); x a @ C ~ , y~ @~
~-, oo B nx
 225

lim ( G ~ ( x, y, k) - G ~ ( x, y, kO)) = O, x ~OQ,

y--~x

lim 62 ( G ~ ( x, y, k) - G ~ ( x, y, ko)) =
y-~x

2G D
~2Q D
= lira ( ( x, y, k) ( x, y, ko)) = D(A), x ~ 9 Q .
2
y-~ x ~nx2 Bn x

Here G D ( x, y, k) is the Green's function of the Dirichlet problem.

All quantities appearing in the theorem exist if the boundary is
smooth.

The most natural formula is obtained if we choose the extension

specified by the boundary condition

a -,in = - a-, ex, a+,in = a+,ex

which differs from ( 1 ). If ~ - ~ , we get a formal expression

for the scattered waves:

s~ex

- ( x, x O, k); x a Q In,
Din( ) + Dex(k) ~ n

+ ( x, 9, k) = + e x ( x, 9 , k) -

B,~ex
( Xo, 9 , k)
8D
~n ( x, Xo, k); x ~ ~ e x ,
Din(~) + Dex(~)
 226

Here the function D i n ( A ) , Dex(k) depend on the extension parame-

ter AO"

We can compare the result with the solution of a real diff-

raction problem. Let F d = {x e @ ~ , Ix - Xol < d] be the opening.
The scattered wave in this case can be represented as [3~

+d( z,~, k) ~'n ( x, z, k) dz, x ~ ~in,

q r

+a( x, ~), k) ; + e x ( x, 9 , k) + (2)

~G~ x
.~ ~d( z,~, k) (x, z, k) dz, x~oex.
rd ~n

Using ( 2 ), we calculate first the function ~d ( x, ~, k) on

the sphere ~ d = Ix: ~x - Xo~ = d3. After that we employ the Green's
function G C for a ball to compute the value of ~ d ( x, ~ , k) on ~d"
As a result, we get the following integral equation for ~d:

"Cd(X,9, k) - ~ "~'d(Z,~), k) dz ~ (s, z, k) (s, x) d s -

q cin Bn

- S ~d(Z, ~, k) dz ~ (s, z, k) (s, x) ds =

I'd Cex B n

_- ~ +eX(s,V, k) - - ( s , x) ds,
Cex BnC

where C in' ex =~-d ~fLin' ex. For a sufficiently small d, the inte-
gral operator appears to be contractive, and we can solve this
equation by iterations. The parameter k 0 is chosen in such a way
that the flux through the hole in the real and model problem
 227

coincide The condition have the form of an algebraic equation:

~G C
gin + geX = (
(s - Xo, nF ) ~ (s, xo) ds) - l ( s ~ I -
Cex

-J ( s, xo) ( s, ds
cin ~nC ~ Xo) -

-J "8'n (S, Xo)-7 ( s , x o) ds),

Cex r ~c

where S d is the opening area, G~' in, ex ( x, y) is the Green's

function o~ the Laplace equation and

~2G~n , ex ~2~in'~D ex
gin, ex = ( ( x, x O) ( x, ko)) 1
n2 ~ n2 Xo' x = x0

It is essential that the spectral and geometrical characteristics

are separated in this equation.

Our model can be generalized to describe the case of a domain

with a narrow slit [61. In this situation it is necessary to impose
the boundary conditions on a line. The deficiency indices of the
Laplace operator restricted to the set of functions vanishing near
the slit are infinite,and the deficiency elements are the potentials
of simple layers with the density from the Sobclev's space H_l. In
order to satisfy the boundary conditions, which define the sought
selfadjoint operator, we have to solve an integral equation. In the
cases of simple domains this equation can be solved explicitly.
This is true, for example, when the variables in the internal and
the external problems separate, and the "connection" llne coincides
with one of the coordinate axis (a plane with a straight slit, a
cylinder or sphere with straight or circular slits [7] etc).
 228

3. A model of coupled resonators

Our construction is particularly useful if one has to inves-

tigate a system of coupled resonators [2, 8, 9]. Let us consider the
problem of a circular resonator with
running wave which can be reduced to
the spectral problem for Laplace ope-
rator with Neumann boundary condition
on a symmetric domain consisting of a
several identical circles coupled to a
ring (Figure 1). Let k n1 be the eigen-
values of the Laplacian inside the
circles, ~n being the corresponding
eigenfunctions. Let further A n0 be the
eigenvalues of th~ Laplacimn inside
Figure I. The resonator the ring and t n being the correspon-
ding eigenfunctions. Then the disper-
sion equation for ~ has the following form:

z I% +
n

+N(~-~O)~ItN~p 12 (ANm+p
0 _ ~ )-i (~°m+p - ~o)-l +

+ [(~ - 1) l@~m+pl 2 ( ~ mo+ p - A o )-I -

m

N-1
0
l*Nm+p+j 12 (~m+~j - ~o )-l = O.
j=l

Here N is the number of circles, p = O, I, ... N-I. The roots of

this equation are approximately equal to the eigenvalues of corres-
ponding "real" operator for such complicated domain.
 229

References

I. B. S. Pavlov, M. D. Faddeev, Zapisky of Leningrad Branch of

Steklov Inst. (LOMI) 126 (1983), 159.
2. M. M. Zimnev, I. Yu. Popov, Soviet J. Vych. Mat. and Mat. Fiz.
27 (1987), ~66.
3. V. Yu. Gottlieb, DAN SSSR 287 (1986), ll09.
4. I. Yu. Popov, DAN SSSR 25~ (1987), 330.
5. M. Yu. D~ozdov, I. Yu. Popov, Leningrad Univ. Vestnik, Set. ~,
No 3 (1987), 93.
6. B. S. Pavlov, I. Yu. Popov, Leningrad Univ. Vestnik No 19
(1983), 36.
7. I. Yu. Popov, Leningrad Univ. Vestnik, No 16 (198~), 79.
8. B. S. Pavlov, I. Yu. Popov, Leningrad Univ. Vestnik, No 4
(1985), 99.
9. B. S. Pavlov, I. Eu. Popov, Leningrad Univ. Vestzik, Set. l,
No 4 (1986), 105.
 HAMILTONIANS WITH ADDITIONAL KINETIC ENERGY
TERMS ON HYPERSURFACES

W. Karwowski

Institute of T h e o r e t i c a l Physics, University of W r o c ~ a w

Wroc~aw, Poland

In m a n y cases informations about a physical system can

be d e r i v e d from study of a p a r t i c l e in a f i e l d of e x t e r n a l po-

tential forces.

Such systems are d e s c r i b e d by Hamiltonians of the form

H = -A + V.

Very often this formula has only intuitive value and requi-

res additional input to a c q u i r e precise meaning of a s e l f -

adjoint operator in H i l b e r t space.

Sometimes the f o r c e s in q u e s t i o n are v e r y small away of

a set S of m e a s u r e zero. Then it s e e m s reasonable to c o n s i -

der the idealization. Namely a potential being infinite on S

and zero away of S.

The c o r r e s p o n d i n g Hamiltonian is e x p e c t e d to b e a l i m i t
 231

as e + 0 of H e = -A + V£, where V e is the m u l t i p l i c a t i o n by

a bounded measurable function. As e + 0 the supports of V

shrink to S a n d the maximum of V goes to infinity.

It seems natural to t a k e for V a sequence converging

E
to the Dirac ~ distribution and in fact one uses the symbo-

lic notation of the kind -A + f6(x-y)V(y)dy for the l i m i t of

s
H g as e ÷ 0. It t u r n s out however that unless S has co-dimen-

sion I one needs a slower growth of V .

c
In fact it h a s been shown [I] that with appropriate

choice of the sequence V , the H converges in the strong

resolvent sense to a p o s i t i v e selfadjoint operator. The prob-

lems of this kind include the so c a l l e d point interactions.

They have been extensively studied together with the scatte-

ring theory in the forthcoming book on Soluble Models in

Quantum Mechanics by S. A l b e v e r i o . R. Hoegh-Krohn, H. Holden,

F. Gesztezy. The authors consider also Hamiltonians with

both the point interaction and that described by the bona fi-

de p o t e n t i a l s .

It is c h a r a c t e r i s t i c that the "potential supported by

the set of L e b e s g u e measure zero" corresponds to a c e r t a i n

boundary conditions on S.

It is t h i s fact that bears similarity between those "po-

tentials" and the problem we are going to d i s c u s s .

Namely we shall also be interested in the Hamiltonians

with boundary conditions on the sets of Lebesgue measure ze-

ro, but these boundary conditions have different character

than those obtained from the potentials.

 232

As we shall see it is p o s s i b l e to g i v e a m e a n i n g of

a positive selfadjoint operator in L 2 ( ~ n ,dx) to the f o r m a l

expression

- A - ~6(Xn)An_ I , (I)

32 32
where A is u s u a l L a p l a c e o p e r a t o r a n d An_ I - 3x~ + "''+ 3 2
Xn- 1

We g a v e this f o r m u l a to i l u s t r a t e the o r i g i n e of the

difference between the "potentials" and our o p e r a t o r s . It

also explains the t i t l e of our talk.

Our s t u d y will be b a s e d on the q u a d r a t i c forms t e c h n i c s

but in the c a s e like t h a t of (I) G o r z e l a ~ c z y k (private c o m -

munication) has shown t h a t there is a (strong r e s o l v e n t li-

mit) of the o p e r a t o r s -A-u(e)X[_E,e](Xn)An_1 as E ÷ 0, w h e r e

X{-6,~] is the c h a r a c t e r i s t i c f u n c t i o n of the i n t e r v a l

[-6,e] and u(6) the real f u n c t i o n such that u(e) + ~ as

6+0.

The m o t i v a t i o n for our study is twofold. F i r s t it s h o u l d

have applications in a c o u s t i c . Second: the s u r f a c e p h y s i c s

is v e r y c o m p l i c a t e d . We w o u l d like to test o u r m o d e l s against

the s u r f a c e p h y s i c s fenomena. Perhaps there are e f e c t s that

will fit in the a d d i t i o n a l kinetic energy scheme r a t h e r then,

or t o g e t h e r with, the p o t e n t i a l s . It may p r o v e true or not,

b u t t h e r e m i g h t be a n o t h e r a r g u m e n t that m a k e s the w h o l e con-

 233

cept more acceptable for the physicists. Namely Dabrowski

(priv.com.) claims that the H a m i l t o n i a n s with the a d d i t i o n a l

kinetic energy on m a n i f o l d s can be d e s c r i b e d by the nonlo-

cal potentials.

After this introduction we shall state the main theorem

which shows how to define our operators. Then we c o n s i d e r se-

veral examples. They are very simple but n e v e r t h e l e s s they

exhibit effects of the kinetic energy terms.

Let n > 2 and ~ a nonempty open subset of ]Rn . Select

n
n constants C I , . . . , C n such that for s with 0,s,n-1

~s = { (x1'''''Xn)6~n ; Xs+1 = Cs+1'''''Xn = Cn} ~ ~"

If f 6 L 2 ( ~ n ;dx) we put f s ( x l , . . . , x s) E f(xl,...,Xs,

Cs+1,...,Cn). Also (Xl,...,Xs)6~ s will be the shorthand for

(Xl,...,Xs,Cs+1,...,Cn)6~ s.

I
Theorem. The following forms defined on C0(~ n) are p o s i t i v e

and closable in L2(~n,dX) :

+
ek(f'g ) = S ? n f ' V n g d x I ... dx n
n

fn-1 n-1
+ ~n-1~ Vn_ I "Vn_ig dXl • . .dXn_ I +
n-1
 234

fn- Ig n- Idx
+ Bn- I~n_i i.. •dXn_ I + . . . + e k f Vkfk'vgkdxl...dXk
Uk

+ Bk f fkgkdxl-..dx k + Bk_ I f fk-lgk-ldx I " " "dXk-1 '

Qk ~k-1

where k=1,2,...,n; ~i,...,~n_i > 0 , 80,B1,...,Sn_ I > 0

and ~ fOgO = fOgO = f(cl,...,Cn)g(cl,...,Cn).

There is one to one correspondence between positive closed

quadratic forms and positive self-adjoint operators. The

formal expressions corresponding to the operators given by

the above forms read:

Hk = - A - 6(x n -Cn)(en_iAn_1 - 8n_ I ) -

- 6 (Xn-C n) . . .6 (Xk+1-Ck+ I ) (~kAk-Sk) +6 (Xn-C n) . . .5 (Xk-C k) Bk_ I .

Remarks

I) This theorem is a simple corollary to the result of [2].

The proof is based on a version of the imbedding theorem for

the Sobolev spaces which implies the inequality:

f Vs fS'VsfSdx I. .dXs>C ~ fs-II2dxl dXs_1=cIIfS-III2%_ I

~s ~s-1
 235

2) If w e put ~i=0 for an i>k then the form is n o t closable.

3) If k=1 and ~0>0 then we get the point interaction for

any n.

4) Since -4 < H k w e can define Hkv = H k + V where V is a n y po-

tential small with respect to - A in the Kato sense.

Examples

Put

jR3 = ~3 ' ~2 = jR2

Let us consider following simple models

I) H = - A + V(x)

2) H = - A + 6(x3) B

3) H = -A- 6(x3)(~A 2 - B)

4) H = - A - d(x3)(~A 2 - B) + V(x)

where ~>0 ,~>0, V(x) = S 0; x3<0 , V<0.

V; x3 > 0

The resulting boundary conditions are

i) f (X_~0) = f (X_{0)

ii) ~(X~0) -~(X_[0) = (-&&2 + 6 ) f ( x , 0 ) .

 236

We shall examine propagation of the plane waves. Thus we put

e ipx +Ae i(px - P3X3 ) , x 3 <0

f(x)
=f
[ Be iqx , x 3 > 0,

2
where ~ = (PI'P2)' q = p ' q3 = P3 - V

We get

(P3 - q3 ) - i ( e p 2 + 13) 2P 3
n -- t S --
2 2
(P3 + q3 ) + i ( ~ p + ~) (P3 + q3 ) + i ( e p + B)

It is convenient to e x p r e s s A and B by the energy E=p 2 and

the angle ~ , tg n = I£ I /P 3.

We shall also consider IAI and the phase ~.

B = I + A

We get

2 2
P3 = E C O S

2 2
q3 = E c o s ~ - V

IPl 2 E sin 2

and hence
 237

I
[ (/Ecosq 2]
IAI = - /Ecos2q -V) 2 + (eEsin2q + B)2]
(/Ecosq + /E--~oos~-V) 2 + (~Esin2q + 6)

2/Ecosn (~Esin2q + 6)
tg~ =
(~Esin2q + 6)2 _ V

We o b t a i n our cases I),2),3),4) by s p e c i f i c a t i o n of the cons-

tants ~,B,V.

Let us d i s c u s s the r e f l e c t i o n amplitude for very small and

very large energies.

E ÷ 0 E ÷ ~

I) H=-£+V(x) [AI÷I , {=-~ IAI÷0 , {=-z

2) H = - A + 6 (x3)~ IAI+I , ~÷0 IAI÷0 , ~=~i2

n=0 IAI+0 ~+~12

3) H=-A 6>0,1A[+1, ¢÷0 n#0 IAI+I ~+0
-6 (x 3) (~A2-8)
q=0, A - 0 q=O, A -= 0
B=0
n~0, IAI÷0 ~÷~12 nMO, AI+I

4) H=-A+V(x) IAI÷I , ¢÷0 q=0 AI÷0 , %+~/2

-6 (x 3) (eA2-B) q~O AI+I , ~÷0

 238

These examples show that the additional kinetic energy terms

indeed change the character of the reflection and it should

be possible to discover the corresponding efect (if it

exists) in the experimental data. On the other hand the dis-

cussed models may be too simple to describe a real situation.

ACKNOWLEDGEMENT

The author acknowledges gratefully the h o s p i t a l i t y at the

Research Center BiBoS (Bielefeld) where this work has been

done.

REFERENCES

[I] Grossmann A . , H o e g h - K r o h n R., Mebkhout, Comm.Math. Phys.

77 (1980)87-I 10.
[2] Karwowski W., Marion J., J.Funct.Analys. 62(1985) 266-275.
 THIN LATTICES AS WAVEGUIDES

B.S .Pavlov
Department of Mathematical and
Computational Physics, Institute for
Physics, Leningrad State University,
19890# Leningrad, St.Peterhoff, USSR

This lecture is concerned with infinite lattices embedded into

a configurational space of a higher dimension. We shall call them
thin lattices. Such a structure is non-compact and the scattering on
it exibits some pecular features. In particular, the continuous spe-
ctrum of related operators consists of two branches :

(S) Scattered wave branch Z s: the corresponding eigenfunc-

tions have a form of free waves, reflected by the lattice. This
branch coincides with the spectrum of the "unpertumbed operator"
which acts on the space without any lattice.

(W) Waveguide branch Z w : the corresponding eigenfunctions are

localised in a neighbourhood of the lattice. In a periodic case they
have a form of Bloch waves.

system of scattered waves might not be complete in the corresponding

Hilbert space,i.e.,that the related wave operators might not be com-
plete. Usually one tries to exclude the waveguide branch, formula-
ting suitable conditions for completeness of the wave operators. On
the contrary, we shall discuss here several situations in which this
branch plays an important role.

i. 0he-electron model of a linear molecule

One of the first examples of a thin lattice was studied in 1966-

1968 by R.A.Subramanian (see Ilia, a postgraduate student of profes-
sor Yu.N.Demkov. He suggested a model of a long molecula based on
the self-adjoint extension ( - A ~ ) of Laplace operator with the
boundary conditions on the lattice
 242

(xs)l = I x=es, s 6 Z I, {e{ = i, x6R3~

1
Us 0
+ u s + o(i),
~ ( x - x s)

X-'~Xs~
o
Us = ~ US
i} •

I
The spectral properties of the operator (-n~) have been investi-
gated by S.Albeverio, R.H~egh-Krohn and others (see [2]). In Ref.3,
Yu.E.Karpeshlna suggested the following new representation of the
related lattice sums

ik+ Z exp ik}x sl

D(A ,t) = 4 ~ s~z(l,2,3) 4 T iXsl exp i(t,x s)

This analysis became a base for more refined crystal models with
two-dimensional lattices in R 3, and was used later on for construc-
ting point-interaction models with an internal structure (see [4]);
thin Z 1 and Z 2 lattices of zero-range potentials have been inves-
tigated in our paper [5]. We begin here describing the results ob-
tained there for Z 1 lattices .In a sense, they are characteristic
for all the waveguide situations discussed in the present lecture.

Let ~int = ~ ~ Es be an ortogonal sum of unitary equivalent

finitedimensionsl Hilbert spaces. Let A s be self-adjoint operators
in E s which are mutually unitary equivalent, and Ain t = ~ ~ As.
s6Z I
The starting operator is defined as s direct sum

( - 6 ) • Aim t in L2(R3 ) ~ ~int

of the kinetic energy operator ( - A ) and the "inner" operator

A int. The restriction of ( - A ) ~ ( - ~ ) o on the linear set
DeXto of a l l W2~ smooth functions in R3\ (Xs)l, which have asymptoti~
1
u(x) = us + u sO + o(I) ,
x-,-x s @ ~ l X-Xsl

creates the non-zero boundary form

 243

Jo(u,v) = < (- A)oU,V ~ - <u,( - ,~)o

~ v > = ,7' (sVs-uslv
u o~--~ s)
s6Z I

The restriction of the inner operator Aint--*Aint, o to the linear

set D Oint described in [4,5,6] also leads to nontrivial boundary
form Jl( u int, v int ) :

Jl( Hint, vint ) : Z (~(.)w~(v) - W~+(.)Ws(V) )

sEZ

The boundary form Jo + Jl of the operator (-z~) o + Aint, o vanishes

on the Lagrange plane L F Ci L2(R3 ) ~ 6 int given by the trans-

(°/
lation- invariant boundary conditions

~'(.) fez I s-t

O)
or

~s t~z 1
In both cases the spectral structure of the resulting self-adjoint
operators

= (- A)o • aint, o Dr,

D~ = I D ext
o ~ Dintl
o + L~

is similar. In what follows, we shall discuss only the first one.

The spectrum of ~ F is purely continuous and consist of two

branches: the scattered-waves branch ~ s and waveguide branch ~w"
The eigenfunctions corresponding to ~ s have a form of scattered
waves :
eik~x-se~
I exp-ik(x, O) + Toext (k,) Z exp-ik(se,~)
~(x,~ ), k) = sEZ I @ST~x-sel
int
exp-ik(se,~) ~ o (k,~)

for k = ~ and ~ E [0,~). The eigenf~nctions oorresponding to

have a form of Bloch-waves localised near the lattice :
 244

i exp-ist exp ik Ix-eel

~(x,t,k) = sEZ 1 4~Ix-sel

exp-ist b°o(k)

The energy ~ and the quasimomentum t of the waveguide branch are

connected by the dispersion equation

det D(k,t) s~Z s

where the R-function d(i~) = <(I+~A) ( A - h l ) - l ~ , ~

t
depends on the
spectral structure of the inner operator A=A s and the deficiency
vector e of its restriction in ~s" The lattice-sum D(h,t), cor-
responding to the one-dimensional lattice ~xs~ was calculated
explicitely by R.A.Subramanian :
1
D(A ,t) _ i n ( 2 ( c o s ~ - cos t) )-i
4ff
The branch of logarithm is fixed by the condition of analytical
continuability of D( h ,t) into the complex spectral plane of h and
real-valuednessof D(A,t) on the negative semiaxis. In the case
without "atomic orbitals overlaping" ,i.e., if l~s=O for s~O and

\rlo rll
then the dispersion equation becomes more simple

(~oo - D(~,t) ) ( ~ll - d()~) ) - I~oi 12 = 0 (2)

If the interaction between the internal and external channels is

weak, Iro114< l, this equation can be easily solved graphicaly.
In this case the zero-lines of eq.(2) in the (t,k)-plane, k= ~-~ ,
are approximately given by the zero lines of ]~oo- D =0 and
~ll- d = O. Our choice of the logarithm branch in D(~,t) yields
the equality Im D(A ,t) = n/4, n ~ Z l, on the domains cut from the
trip It- t by the lines +- t = 2 z, I. real
part of D( i~ ,t) has logarithmic singularities on these lines. The
waveguide branch ~ w found by R.A.Subramanian is given by projec-
tion of the zero-line "Foo- D = 0 onto the Z-axis. The posftive
part of this zero-lime is denoted as c(1 on Fig.1. The waveguide
branch ~ w ' which correspomds to the operator A is given by pro-
 245

jection of the lines ~2(see on the left of Fig.l). The quasista-

tionary bands corresponding to higher momenta n=l,2,.., have been
treated in Ref.7. In our model these bands arise from the inner
structure of the atoms,i.e., the internal structure of the zero-
range potentials.

This nontrivial structure of

the "atoms" considered in our mo-
del illustrates the relation bet-
ween the single-atom resonances
which are localised near the zero
lines ]~Ii- d ( h ) = 0 (see [ 6 ] )
and t h e g a p - b a n d s t r u c t u r e of w Zw ~ ....
One can a l s o e v a l u a t e t h e e f f e c -
t i v e mass of the corresponding
quasiparticle (d2~,/ dt2) -1. This
0 qr
mass appears to be very large near
the edges of the resonant bands
Fgl
and/or gaps. The fact that the waveguide band is situated in the
lower part of the spectrum,-~ < t ~ ~ , corresponds to the re-
quirement that the component of the three-dimensional momentum
k cos(e,~) = ~-~ (e, ~) tangential to the latOice is ma~orized by
the lattice quasimomentum t. Thus the "conservation law" for momen-
tum prohibites emission of an electron from the lattice into the
R 3 - space.

It is easy to include a finite number of impurities into the

lattice assuming

As = A for s ~ (sl,e2,s3,...,s n) and

As # A for s ~(Sl,S2,S3,...,sn)

Using the following ansatz for waveguide eigenfunctions

eXt(x,k) = 7 , exp ik}x-se~ ext
s~Z 4~lx-sel ~ s

int( k ) : ~ ÷ ~ ; w s ÷ ~ : w s
Ts
we get for the boundary values in the exterior space

}s--ext-----ik~e~ + Z e~ikls-r, Tre~

4~ --s r~s a~Is-r I
 246

+ext = ~fext( k )
~s s "

~int ,/~- '~As @> +int

The boundary conditions (1) give the equation

~s r~s ~,s-r~ r too vol ~t

= +int|
+int rIo Vll ~ s --
~s
which can be solved easily in terms of Fourier transforms of ~sext
+ int
and ~ s

~(q) = Z exp iqr ~ rext ' ~(q) = ~ exp iqr ~ ; i n t , _ ~ q < ~

r~Z feZ
and the lattice sum D(A ,q)

D( h ,q) ~f (q) ~oo ol '~(q)

(3)
d( h ) ~(q) + (K ~)(q) K'lo ]~Ii T(q)

Here K is the finite-rank integral operator in L2(-~T, ~) with the

kernel

K(q-p) = Z (ds(A) - d(A) ) exp is(q-p)

s £(Sl,S2,..,sn)
Excluding ~ from the last equation,

~(q) = - ( F o o - O(~ ,q) )-lFo I ~(q)

we get the Friedrichs-model-type equation for T

r o( too - D( ~ ,q) )-l Vol) ~ _ K '~ : o

(4)

The simplest equation of this type with a smooth kernel K was inves-
tigated by L.D.Faddeev [8~ • By a similar argument, one can show
that the waveguide branches of spectrum for the homogeneous and in-
homogeneous lattice coincide. The only difference is that some ad-
ditional (finitely many) eigenvalues and resonances can appear in
the inhomogeneous-lattioe case. The operators in the lhs of eq.(@)
 247

are comparable for k=0 and K~0 independently of ~ . The correspon-

ding scattering matrix which describes scattering processes in the
waveguide can be calculated explicitly.

The case of a halflattice, de=0 for s < 0, is also explicitly

solvable. The role of the integral operator K is then played by a
Henkel operator. One can combine the above discussed cases and dis-
cuss half-lattlces with impu~ities. Other posibilities such as
As--~A + for s ~ o o are interesting rather mathematically than from
the physical point of view. A similar consideration can be perfor-
med for a two-dimensional lattice in R 3. The homogeneous lattice
zs~2 , s ~ Z 2 is con=i~ore~ i~ ~5].

2. 0ne-elqctron model of a long elastic molecule

We take the simplest one-particle Hamiltonian (see also ~9])

in the form of a Jacobi matrix Hph in ~2(Z l) = H ph
1

Hph = ~ T1 + ~ o

where (TlU)s= Us_l, u g ~2(Z l) . The N-phonon Hamiltonian in the

harmonic approximation can be considered as the form sum

Hph= Hphl ® 1 2 ® . . . @ I N .+ l l ® H.~ h ® 1 3 . @. .®i N + i i @ 1 2 ~ . ® H phN

It is a selfadjoint operator in Hilbert space ~ p h = ~ i ~2 @

ph~ ph "'"
ph"
The simplest one-particle Hamiltonian of a long molecule can
be taken in the form of a one-dimensional Schrodinger operator
with a periodic point potential (the Kronig-Penney model) :
in L2(R) = ~ e l we define
d2

It is easy to combine Hph and Hel to construct the initial Ha-

milt onian

Hel®Iph + lel® Hph = H

which is a selfadjoint operator in the tensor product

 248

The spectrum of H is the algebraic sum of the spectra % 1 of ]~el

and ~ p h of ~ph " If Pel is the quasimomentum of Hel and X(p)
is the corresponding eigenfunction :

Hel X(p el): o~ el(Pel ) X(Pel)

and X(pp h) is the phonon eigenfunction with quasimomentum Pph

~ph X(Pph) : ~;h(Pph) Xph ,

then X(p) = X(Pe I) • X(pp h) is the eigenfunction of H with the quasi-
momentum P = ( P e l ' Pph ) corresPonding to the eigenvalue

c~ (p) = o4 el(Pel ) + O~ph(Pph)

The resolvent of H can be easily constructed in terms of its eigen-

function expansion. Its kernel

I
x(p,~) x(p,~)

G dp

i s a f u n c t i o n on t h e e l e c t r o n - p h o n o n c o n f i g ~ a t i o n space which is
t h e u n i o n of a c o u n t a b l e f a m i l y of c o p i e s of t h e r e a l a x i s A s

R~z~ = YzN A s

It fulfils the equation

(H - /~ I ) X 2'
G(s,r,2~) : (...,0, ~s~ ( S ( X - 2 . ) , O , . . )
If y is integer, y=m, then the additional singularity of G can be
described in the boundary-condition form :
/

I Gr ] (n) - h Gs(n) -- - ~ nm (S mr ,

here G s = G ~ A s . We interpret s as the coordinate of an

excited oscillator. Then the interaction between phonons and elec-
trons must be localized near the main diagonal

M = { x=m, s=(m,m,m,...,m)}

of the configuration space. The simplest way to switch in the in-

teraction is given by the selfadjcint-extension theory. Let us re-
strict H---H o to the operator defined on the linear set D O of all
 249

functions u, v ~ D(H) vanishing on the main diagonal. An integration

by parts leads to the boundary form

HoU,V ) - < U , H o V > = ~ {(~u~](m) - h urn(m) ) Vm(m) -

m~Z 1

- urn(m) ( [Vm](
, m ) - h Vm(m) ) }
Using the boundary conditions on the diagonal M

i
~u m ] ( m ) - h Um(m) = V Um(m) , Y~O, ImF=O (4)
we get a homogeneous one-dimensional Hamiltonian HW. Of course, one
can use slso boundary conditions depending on energy,r -~ ~ ( ~ ) ,
or to employ the next-to-main diagonals. Such assumptions also lead
to solvable models. However, we restrict ourselves here to the
simplest sort of boundary conditions (4). The spectrum of H con-
sists of two branches ~-7s and Z w . The scattering branch ~ s
coincides with ~ ( H ) . The corresponding scattered waves differ
from the electron-phonon Bloch functions by the waves reflected
from the diagonal, which plays here the role of a thin lattice

Tp=Xp÷ ZYmGm , %-- G(~ 'm'

~ ,~)
Using the translation invariance, we get ~ m = h°° e x p i ( P e l + P p h ) m
and
1
~o = ,
- X(p 0o) ( r+h + ~m ~ Z exp 3m(Pel+Pph) G( 0 m ~ +30) )

The scattering amplitude can be derived from this formula using the
asymptotic behaviour at infinity, ( x , s ) - ~ o o , ~ ~ R 2, I ~I =I

G(s, o, ~ ) (5)

~)
Here p =[pel,Pph ) i s t h e p o i n t on t h e i s o e -
nergetic surface E~, where the tangent line
is orthogonal to ~ . We assume h e r e , t h a t
there exist only one p o i n t of s u c h a s o r t
(the formula (5) has b e e n d e r i v e d by my s t u -
dent V.Evstratov). The a s y m p t o t i c $ ( 5 ) g i v e s
at infinity

(]~+h)-l+ 7 . exp im(Pel+Pph) GcO m Z +30)

~O'm'
G(s,o)
m~Z
 250

These functions describe the processes of electron-phonon scattering.

The corresponding amplitude contains Laue singularities and has, due
to the denominator, a resonant form. The second branch ~ v W corres-
ponds to the eigenfunctions which are localised near the diagonal
M and have the form of Bloch waves :

~fq = ~ exp i(q,m) Gm, m = ( m,m,...,m )

m6Z 1

The quasimomentum q can be calculated from the following dispersion

e quat ion

( h + V )-i + 7, exp i(q,m) G cO m h +iO) = 0 (6)

',O,m I
m6Z 1
This equation is similar to the equation (2) of Section I. The cor-
responding quasiparticles can be interpreted as pc!stone.

Using energy-dependent boundary conditions with an R-function

h=h(2%), we get the picture of waveguide bands, which is quite simi-
lar to the corresponding picture of section i. It is not difficult
to discuss the model of a crystal built of atoms of different sorts.
There are no difficulties to consider crystals in R3and crystals
with finite number of impurities. More difficult problems arise for
semiinfinite lattices and for lattices which contains stochastic
elements. If those elements are given by Markovian processes, then
the electron-phonon properties of the corresponding lattices can be
studied after averaging which yields nonhermitian Hamiltonians.

Of course, one can impose boundary conditions not only on the

main diagonal, but also on the diagonal planes

il..in: {<x,s):x:m, si:m, l, i ]

Then we get a many-body problem with many channels. It can be stu-
died with the help of appropriate technical tools (see [9]). In
this model, one can describe the scattering of polarons of different
sorts. These models are not solvable, but equations, which appear
in corresponding analysis are quite similar to Faddeev's equations
(see the zero-range version of them in [10, ll] ).
 251

3. Two quantum particles in a one-dimensional

periodic lattice

This is, in fact, a three-body quantum problem (see also [ 1 2 ] )

but it can be reduced to a one-dimentional integral equation which
gives again the waveguide-type eigenfunctions.

We construct a one-dimenticnalhard lattice of zero-range poten-

tials with an internal structuze. The one-dimentional one-particle
Hamiltoniao is a selfadjoint extention of the restricted "original"
Hamilt onian

- ~ ~ 7. (A) n = - C A in
nEZ I o ~ o

in I L2 (R) ~ ~ ' i (E)n ~ = L2(R) ~ ~ in =

nEZ
All the internal operators are unitary equivalent, (A)n~-A, and the
deficiency elements e s of the corresponding restrictions coincide
mutually, es=er=e . The external operator is restricted by the con-
dition of vanishing at x=s. Then (see [4])
1 ±
D(Ao)s = {Us: Us= A-il U + ~ +s W +s + ~ s Ws } '

d2 ~* ~w~, z> nc] .

We choose the extension ~r of ~ determined by the following
boundary conditions

( )( -u(s)
~s
:F In']
~s
+ (s)
) , F : I ,

which gives a SchrSdinger equation with energy-dependent potential

for the external components of eigenfunctions

d2
- ~ u~ - fr
on
- Iroll~Crll - d<~, ~ ~ - l ] Z
n 6 ZI
%c~-n~uc~ :
= % u(x) ;

here d(A) : @(A-Xl) -l ( l + A ~ ) e, e > . ~he speot~um of A~ can

 252

be derived from the dispersion equation

( ]~oo- D ( k , t ) )
Here D ( h ,t) is one-dimensional lattice sum,

o(~,t) _ ( 2 ~ ) - l s i n ~ ( cos~- cos t )-l.

In order to construct the two-body Hamiltonian we can start with

the operator A l ~ 12 + I 1 ~ A 2 in the space ~ 1 ~ ~ 2 " The de-
composition

contains the subspace ~ 1 ~ ~ 2 which can be neglected if we discuss

only the interaction of "free" particles with the lattice. Never-
theless, it must be taken into account if we want to consider the
direct interaction between the internal dynamics of the "atoms"
As, A t . We begin with the "original" two-body Hamiltonian

d2 d2 d2

d2
@ ( ® A~ nt + ( - ) ® 12
Ixl ~ ) @ ~ ((A)s®I t + Is@(A) t
s,tEZ 1
(7)
Here A int = ~
~ (A) s and I is the identity in ~ The inte-
s6Z I
faction between "free" particles can be described by a potential
in the first term, which is then replaced by a Schrodinger operator9

d2 d2 d2 d2
-

Here q can be Coulomb or Yukawa potential or even some separable

interaction in L2(R2) = L2(O,O°) ® L2(0,2~-) . The lattice effects
can be introduced by the above described restriction-extension pro-
cedure. The simplest two-body interaction can be also taken in
the form h(Xl-X~) , which is equivalent to. the boundary condition

= hu on the diagonal M of the configuration space R2

~n
Unfortunately, the last condition is not valid for fermions, which
 253

have antisymmetric eigenfunctions. Since they are continuous they

vanish on the diagonal M and the last boundary condition becomes
trivial. To rectify this situation we use f~ther a smooth potential
and rasctrict the external operator to the linear set of all smooth
functions, vanishing on the lines L1s : {~l_-s ,l:I,2, s ~ . ~he
restriction of the inner operators

int ~(_ d2 ix#~int . d2 -

Ais ® Ix2+Iis dx--~ ) ' 2t +k d ~ x l ) ~ I 2 t '

( As® It+Is@A t)

gives a nontrivial boundary form, written in terms of the boundary

values ~ + + +
~s(~2), ~ t (~), ~ st as
dXl+
~"
S ~t I u~t(u)uL(~)
~ -
sc-Z Ls

•
s6Z ,
Ls (8)

- ~ is(~)~s(V) ]d~l + ~--t~_Z!~ L T~t (u) r~ (~) -

~I
-y2t(~)y~( v)]~l ÷ 7S t I ~st(~) [ +S ~ (v) -- ~S~(~) ~ ( V ) ] "

Here the second term corresponds to the deficiency elements which

have singularities at the intersections of the lines L I~L~.
s it
can be shown that every intersection contributes by one deficiency
indices. The last term corresponds to the pumely internal term in
(7). The boundary form (8) vanishes on the linear set D r of all
elements u, u~D~* , which are subjected to the boundary conditions
with hermitian 2K2 matrices r, ~ s t

- u(s,x~

Vis(xl)
/ =V
~-~ (s,xI)
+
oS(~l) / (S ,x I) m L s
1

(uL s,tE Z I
(9)

~t ) = "V~t
g;t
 254

As far as the overlapping of the "orbitals" is nontrivial for the

neighbouming points only, we put ~ s t = ~s_t=O for Is-t 1 > I

It can be shmwn that the operator A o I D r is semibounded

from below. Its Friedzichs extensions plays the role of the two-body
Hamiltonian in the periodic lattice. To construct the resolvent of
this Hamiltinian, one has to solve the system of the equations

d2 d2

* d 2~) 7 ut (IO)
{(%)o-(~+d-~ i 5 2 = f~'

(As+At)oUst - A Ust = fst ' s,t6 Z I

with the boundary conditions (9).

In the case when the direct overlapping of orbitals is neglec-

ted the last terms in (7),(8),(9), s,tm Z I , must be dropped. Then
the the system (I0) can be reduced (see Ref.lO for details) to the
SchrSdinger equation in the external channel with the energy-depen-
dent additional potential containing pseudo-differential operators
on the lines:
d2 d2
- x )u÷

d2
1=1,2 dx 1
S
In the case when the internal operator A=Z ~sEs is ef a finite rank
d2
the pseudo-dlfferential operator d ( ~ + d~x ) has the following form:

dx= is

Finally, we derive the dispersion equation for the waveguide

branch ~ w of the spectrum. We assume that q ~ O and insert a new
function in the exterior channel, v= q.u . Using the~ the exterior
quasiresolvent G corresponding to q = 0 , we get the equation for v:
 255

v(x = I q(x v(y dy ,

•R 3
Rere q(x) = q(Ix I -x2~).

Taking into account the Bloch property of the waveguide eigen-

function with respect to the shift along the lattice x - ~ x + e ,
e=(1,1),we can reduce the last integral equation to a simpler one
with the operator K which is compact in the Yukava case;

v(x) = I (x) v(y)dy , 0<- Xl+X2 z ,

OWYi+Y2<- 2

whose kernel is the-lattice sum:

Kh(x,y,P)=
Z G~\(x,y+me) e ipm .
m~Z
Thus the dispersion equation for waveguide branch ~-w can be writ-
ten in terms of a Fredholm determinant containing the potential and
the quasiresolvent of the initial two-particle Hamiltonian without
interaction ( q=O ). This quasiresolvent can be calculated explicitly.

It is very interesting to investigate the localization of the

waveguide bands. This can be done numericaly on a computer.

The waveguide bands can exist even in the case of a repulsive

potential, q70 , because the effective mass of a particle is ne-
gative. One can also construct more realistic models using the elec-
tron-phonon blocks described in previous sections as initial opera-
tors. This will be done in another publication.

Im any case, we get different kinds of waveguide bands, which

correcpond to forming and dissociating of the combined quasipartic-
les similar to bipolarons. ~ore difficult analytical problems arise
in the case of many particles, when the bound states of three, foum
etc. electrons or polarons could appear.

One can also insert the interaction between different phonon

modes in our modell. To do it one takes as usually a harmonic term
in potentiai~. The restrictions-extension proceaure gives anoZher
possibility to switch in the inUerac~ion oe~wee~ different phonon
modes, based on a construction similar to that described in previ-
ous section. It will be done in another publication.
 256

The approach to the description of interactions suggested here

looks at glance very different from the standard one. In fact, it
is very close to it provided there are invariant subspaces of the
standard Hamiltonian which contain a finite number of particles on-
ly. Moreover, using the internal structure we are able to enrich the
spactral properties of the models without any analytical difficul-
ties.

It could provide physicists with many attractive possibili-

ties of heuristic considerations based on easily manageable formulae.

References

I. Yu.N.Demkov, V.N.Ostrovsky: The method of zero-range potentials

i~ atomic physics. Leningrad State University Publ. in Russian
2. S.Albeverio, F.Gesztesy, R.Hoegh-Krohn, H.Holden: Solvable
models in quantum mechanics. Springer Verlag, Berlin 1988
3. Yu.E.Karpeshina: Teor.Mat.Fiz. 57 ~1983) 414 - 423
4. B.S.Pavlov: Teor.Mat.Fiz. 72 (1985) , No 3
5. P.B.Kurasov, B.S.Pavlov: Teor.Mat.Fiz. to appear
6. B.S.Pavlov: Teor.Mat.Fiz 59 (1984) 345 - 354
7. G.V.Golubkov, F.I.Dalidchick, G.K.Ivanov: S o v . P ~ s . JETP 78
(1980) 1423 - 1434
8. L.D.Faddeev: Proc.of Steklov Inst. 73 (1964) 292 - 313
9. S.P.Merkuriev, L.D.Feddeev: Quantum Scattering Theory For
Few-Body Systems. Nauka, Moscow 1986 in Russian
10. B.S.Pavlov: Leningrad Univ.Vestnik, Ser.Mat. 1987
11. Yu.A.Kuperin, K.A.Makarov, S.P.Merkuriev, B.S.Pavlov, A.K.Moto-
volov: Feddeev equations with enegry.dependent potentals.
Preprint ISSN 013-426Y, ITP Budapest, November 1986
12. B.S.Pavlov: Sov.Math.Sbornik to appear
13. L.D.Faddeev, F.A.Berezin: DAN USSR 137 (1961) 1011 - 1014
 QUANTUM WAVEGUIDES

P.Exner, P.~eba, P.~ov~ek

Laboratory of Theoretical Physics,
JINR, 141980 Dubna, USSR

This lecture is concerned with the motion of a quantum particle on

stripes, tubes,layers and similar manifolds in ~ n . A study of this
problem might have started at every moment of the more-than-sixty-
years-long history of quantum mechanics. It did not, however, apart
of some trivial textbook examples, because an appropriate physical
motivation was lacking.

This situation has changed with the advent of modern microelectronics.

During recent years, various technologies have been developed which
allow to produce ultrathin layers and "wires" or extremely pure
metallic or semiconductor materials and build more complicated
structures such as sandwiches, graphs with many branching points etc.
of themKIS. Our knowledge of quantum-mechanical effects in such
systems is at present very poor.

Let us notice that similar problems for a constrained wave equation

have b e e n extensively studied in the classical waveguide theory. The
electron motion in the above mentioned structures is governed by the
Schr6dinger equation (with appropriate boundary conditions), so it is
natural to Call these systems quantum waveguides (or briefly,
Q-guides). The analogy is not only illustrative but useful too ;
recall that in the stationary approach, the free Schr6dinger and
wave equations are the same up to physical meaning of the spectral
parameter.

i. Bound states i n c u r v e d Q-guides

We shall consider a free spinless particle living on a curved planar

strip ~ of a width d with
Dirichlet boundary conditions
(Fig. l). It can model the electron d
motion (with spin neglected) on
a planar quantum wire or a thin
layer on a cylinder-shaped
substrate. Since the particle is
Fig. I
assumed to be free, the results
 258

are applicable provided the region of interest is smaller than the

m e a n free path of electrons ; it can be achieved. A more complete
description should include, however, a stochastic potential correspon-
ding to random impurities.

We choose one of the boundaries of ~ as the reference curve P and

introduce the natural curvilinear coordinates s,u on ~ . The strip
is then fully characterized by its width d and the (signed) curva-
ture ~(s) of ~ .

The state Hilbert space of the problem is L2(~,dxdy) and the

Hamiltonian is

H~=- ~Lm2AD , (1.1)

where AD denotes the Dirichlet Laplacian relative to ~. We shall

restrict our attention to the strips with an infinitely smooth boun-
dary which are curved essentially within a bounded region only, i.e.,
we adopt the following assumptions

Ca) ~ c~ ,
(b) ~, g" , ~'" are bounded,

(c) ~(s)~2_ for some 2_~(-d-I,0] ,

(d) ~2 , ~,% L(R, (I+ Is~) ds)
(e) ~ , ~" ~ L2(~,(1+s2)ds) .

We shall consider the non-trivial case (~ ~ 0 ) only.

The following theorems are proven in Eel. 2 . First we pass to a

straight-strip problem using the coordinates s and u :

1.1 Theorem : Assume (a). Then HQ is unitarily equivalent to a

self-adjoint operator H on L2(R X [0,d]) such that

D(H)~D, {?:~is c~,~(s,o)=~(s,d)=o for sG~,H~G~2~

and

- + + V(s,u)~ (1.2a)

for ~D , where

v(s,u) = - u2", 5 u2~ "2

4(i+uf12 + 2(i+ug15 - ~ (1+ug)~ (1.2b)
 259

A simple minimax estimate applied to the operator (1.2) now gives

1.2 ~heorem : Let A~ fulfil (a)-(d) . Then

(i) ~ess(H~) = ~E~,oO) , where E~= ~212 /2rod 2 ,

(li) there is d0 ~ 0 such that for all d < d O , the operator H~

has at least one bound state in [O,E~) .

The same argument yields a lower bound on d O , but a relatively poor

one. A slightly modified Birman-Schwinger technique can be used to
prove

1.3 Theorem : Let ~ fulfil (a)-(e), then a bound state E~[0,E~)

exists iff

In particular, if ~ is simply bent, i.e., ~ ( s ) ) 0 for all s~ ,

then

d o ~ d+ m 2~+

where ~+ m~x ~(s)

. and z 2d s .2
Simple examples show that the estimate (1.4) yields the critical
width of the same order of magnitude as the minimum curvature radius,
unless ~ exhibits steep changes.
The above results say nothing, however, about the gap between the
bound state E and the bottom of v - - ~ s(H~%)
s given by the first
transversal-mode energy. With this
aim, we present here the results of
a solvable model. It concerns the Y

rectangular Q-guide (Fig.2). Since

is self-similar in this case, we
I //
have /
f

d .... "" l
E =4~E~ (1.5a) n
/' I 2
, !
independently of d . One can use the d X

mode-matching procedure known from

Fig. 2
the waveguide theory ~3], expanding
 260

=
j~1
.:
(-I)J+I rj e qj(1-y/d) ,

where qj = ~ j 2 - E 2 and ~j = ~ sin(~jx/d) , and similarly for the

regions II and III. The regularity requirement then yields the
system of equations

r = Cr (1.6a)

for r = ~rj~ , where

Jk
Cj k = I ( I - e -2qj) ~ (j2+k2_~) (1.6b)

This system cannot be solved directly, because C is not compact on

12(j -I) . It can be reformulated, however, into another system which
is solvable and yieldsE4 ]

4 = o.93 (1.5b)

Let us comment briefly on physical applications of these results. The

most interesting among them is a possible existence of edge-confined
currents in thin films grown on a sharply edged surface. The spectrum
of the three-dimensional problem is,of course, continuous starting
at E , however, the electrons with energy less than E~ cannot
leave the edge. The transitions between E and the (two-dimensional)
continaous spectrum are also observable in principle, being manifes-
ted by infrared photons.

2. The lon~-wave approximation in Q-~uides

It is extremely difficult to get exact solutions describing particles

in Q-guides. At the same time, it might not be necessary in some cases.
Recall how useful are the point-interaction method in low-energy
processes when the particle "sees" a point instead of an interaction
potential which can be of a complicated structure but localized to
a region small comparing to the Broglie wavelength[5J.

One can try to use the same idea to describe the Q-guides replacing
them by appropriate graphs (Fig.3). The question is whether it will
work. Let us start with the curvature effects considered above.
We denote by Ed and E ,d the bound-state and first transversal-
mode energies for a given width d , respectively. Let further E be
 261

Fig. 3

the ground state energy of the operator

ds 2 -

on L2(~) with the natural domain. A straightforward estimation

procedure than gives

2.1 Theorem : Assume (a)-(e), then lim (E - E d) = E .

d~O+ ~,d
This is the rigorous version of heuristic conclusions obtained by
several authors ~ - 9 J .

The behavicur at branching points represents a more difficult

problem. The simplest case is represented by the so-caled Y-junction
(Fig.4). For the symmetric Y-junction
(~ = 2(~-~) ) and Dirichlet boundary
conditions, the electromagnetic
reflection and transmission coefficients
1811 ~ and ]s121 have been calculated
in Ref. 10 . A remarkable fact is that
their zero-frequency limits are indepen-
dent of ob and equal to
Fig.4
e-',olimIsl~l = 13 ' '~"olim Is121 = ~
2
(2.2)

Let us mention further that for Y-junctions with a smooth boundary

and Dirichlet b.c., existence of bound states can be proven in a
similar way as in the previous section (there is no bound state in
the Neumann case as it is well known from the classical waveguide
theory).

Now ene has to construct a suitable graph model for the Y-junction.
It has been done in Ref. 11 ; we review here the results. The
 262

appropriate graph consists of three halflines connected at one point ;

the corresponding state Hilbert space is ~ = L2(~ +) ~ L2(~ +) ~ L2(~+).
The construction of Hamiltonian starts from the operator

H 0 = H0, I @ H0, 2 @ HO, 3 , (2.3)

where Ho,j=-(~2/2m) d2/dx 2 defined on D(Ho, j) =C;(~ +) . The opera-

tor (2.3) has deficiency indices (3,3) and hence a 9-parameter family
of self-adjoint extensions. We restrict our attention to those among
them that commute with the operators Pjk permuting the halflines.
This leaves us with the 2-parameter family whose elements are speci-
fied uniquely by the boundary conditions

f;(0) = Af1(O) +Bf2(0) +Bfs(0) ,

f;(O) = Bf1(O) +Af2(O) +Bf3(0) , (2.4)

f3(0) = Bf1(0) +Bf2(0) +Afs(O)

with A,B~ ~ and two "exceptional, one-parameter classes specified

by the conditions

f~(o)+ f~(o) +f~(o) = o ,

f2(o)- f3(o) = c(f~CO)-f~(o)) , (2.5)
f1(o) -f3(o)= c(f~(o)- f~(o))
and

f;(o) = f (o) = q(o) , f'(o) ,

(2.6)
f1(0) + f2(0) + f3(0) = Df'(0) ,

where again C , D ~ . The S-matrices for scattering on the junction

corresponding to each of these Hamiltonians are easily calculated
to be

a b b)
S(k) = b a b (2.7)
b b a

with

a = . -1.,+ ikB- k2(A2+AB-2B 2)

[1- ik(A+2B)]D- lk(A-B)]
 263

-2ikB
b =
[I - i k ( A + 2 B ) ] [ I - i k ( A - B ) ]

for the boundary conditions (2.4), and

S(k) - 3(1+ikC) 3ikC-1 2 (2.8)

" 2 2 3ikC- I

S(k) - 3-ikD -2 1-ikD -2 (2.9)

-2 -2 1-ikD

for the remaining cases.

What is the physical meaning of the parameters appearing in the con-

ditions (2.4)-(2.6) ? A natural conjecture is that they are related
to the angles ~,~ specifying a given Y-junction. In particular, we
conjecture that to a symmetric Y-junction one of the "exceptional"
Hamiltoniana corresponds. This guess is supported by the following
argument : the S-matrices (2.8) and (2.9) satisfy

lim S(k) = ± ~ 2/3

[-I/3 2/3 2/3)
-I/3 2/3
k~O+ \ 2/3 2/3 -I/3

independently of the parameters C and D , which is just the feature

expressed by the relation (2.2). Of course, the quantum-mechanical
reflection and transmission coefficients are given by squared moduli
of the S-matrix elements, being therefore equal to I/9 and 4/9, res-
pectively. Purthermore, we see that a part of the Hamiltonians (one
half of them, roughly speaking) has a bound state corresponding to
a Role in the S-matrix. It suggests that just this part describes the
Y-junctions with Dirichlet boundary conditions.

Hence the long-wave approximation described above represents a useful

way in which one can replace the original problem by a mathematically
much more simple one. Among its potential applications, those related
to Aharonov-Bohm-type effects in microstructures
~12-14] are particularly interesting. We limit
ourselves here to one of them, namely to a
description of a semiconductor loop with two
external leads in a homogeneous electric
field (Fig. 5). In order to calculate the trans-
mission coefficient through the loop, one has
Fig. 5
 264

to "sew" using the boundary conditions (2.4) the wavefunctions

u1(x) = e -ikx + a e ikx u4(x) = b e -ikx

describing the electron behaviour on the leads, and the functions

u2,u 3 which solve the equations

_ ~22mu;'(~+ vj(x)5(x~ = Euj(x~

on the arms of the loop. Here E = ~2k2/2m and the potentials Vj
are given by the electric field (they should contain also a curvature-
dependent term - cf.(2.1) - but it is negligible in the physically
interesting situations).

The problem leads to a system of linear equations which yields the

following expressions for the transmission coefficient ~5,16 S

B2
T(E) = Ibl 2 -_ -l+~.g
-_-L Ic2*d 2 j2 , (~.1o~

where

(~12)= KJ~21C2(-k)-C1(-k)J']31S-1
(:12)
with

2ik
z I = (A I - B 1 ) z 2 , z2 =
I +ik(Bi-A1)

C1(k) =

I AI+ik(BI-AI ) (BI-A I ) ~(AI+B I )( 1-ikA I )+2ikB2J~

BI~I+ik(BI-AI)3 \ 1-ikA I -A 1-ik (B~-A2)

2
and C2(k) expresses similarly through the parameters A2,B 2 speci-
fying the second junction. Furthermore, ~ are the transfer matri-
ces

(uj(lj)) (o)/
\uj(lj)/ = ~J uj,(o)j , j = 2,3
kuj
 265

Transport properties of the loop are then given by Landauer formula

~17J : the conductance equals

e2 T(E)
'i-T(E) (2.11)

recall that ~/e 2 12906~) . Using the WEB-expressions for ~j ,

one can calculate G from (2.10) and (2.11) numerically for various
shapes of the loop and
parameters specifying the G
junctions. As an example,
we present here (Fig. 6) the 400 ~.
results for a loop of a 12
o
200 A GaAs wire of the 000
sketched shape wlth A I =
= B I = - A 2 = - B 2 = I . We see
that the conductance plot
exhibits sharp minima at 4
reasonably low field
intensities. Changing the 0
loop shape and parameters, -800 -400 0 400 800
we obtain similar pictures. field intensity
This interference effect is ~V/cmJ
particularly promising from Pig. 6
the viewpoint of possible
device applications.

In conclusion, let us mention that the long-wave approximation can

be used effectively in other fields too, e.g., in the electromagnetic
waveguide theory, for killing the humming noise in air-Conditioning
systems etc.

References

I H.Sakaki, in Proc. Symp. Foundations of Quantum Mechanics, Phys.

Soc.of Japan, Tokyo 1983 ; pp.94-110.
2 P.Exner, P.~eba : Bound states in curved quantum waveguides,
preprint, Bochum 1987.
3 R.Mittra, S.W.Lee : Analytical Techniques in the Theory of
Guided Waves, Macmillan, New York 1971.
4 P.Exner, P.~eba, P . ~ o v ~ S e k : A correct truncation procedure
for a rectangular planar waveguide, in preparation.
5 S.Albeverio, F.Gesztesy, R.H~egh-Krohn, H.Holden : Solvable
Models in Quantum Mechanics, Springer, Berlin 1988.
6 J.Marcus, J.Chem. Phys.45 (1966),4493-4499.
 266

7 H.Jensen, H.Koppe, Ann.Phys.63 (1971),127-140.

8 D.Picca, Lett.N.Cim. 34 (1982),449-452.
9 R.C.T.da Costs, Phys.Rev.A23 (1981),1982-1987 ; A25 (1982),
2893-2900.
10 R.Mehran, IEEE Trans.26 (1978), 400-405.
11 P.Exner, P.~eba : Free quantum motion on a branching graph I,II ,
preprints JINR E2-87-213,214, Dubna 1987.
12 R.A.Webb et al., Phys.Rev.Lett.54 (1985),2696-2700.
13 C.P.Umbach et al., Appl.Phys. Lett.~_Q (1987),1289-1291.
14 C.P.Umbach et al., Phys.Rev. Lett.56 (1986),386-389.
15 P.Exner, P.~eba : On feasibility of quantum interference transi-
stors, preprint JINR E2-87-346, Dubna 1987.
16 P.Exner, P.~eba, P . ~ o v ~ 6 e k : Quantum interference on graphs
controlled by an external electric fiels, preprint JINR E2-87-
707, Dubna 1987.
17 R.Landauer, Phys. Lett.A8~ (1981),91-93.
 AN EXACTLY SOLVABLE MODEL OF A CRYSTAL WITH
NON-POINT ATOMS

Yu.A. Kuperin, K.A.Makarov, B.S.Pavlov

Department of Mathematical and
Computational Physics, Institute for
Physics, Leningrad State University,
198904 Leningrad, St.Peterhoff, USSR

I. Introduction

The Schradinger equation with a periodic potential in R 3 pro-

vides the conventional one-electron solid-state model of a hard lat-
tice. This model is exactly solvable in the zero,range point inte-
ractions approximation (the so-called K~onlg-Penney model). The cor-
responding spectral picture seems to be rather poor. The reason is
that the traditional Fermi deuteron potential which describes an
isolated atom [ i - ~ has very simple spectral properties. The next
step is to provide the individual atoms with an internal structume
5 S using the method of the Hilbert-space extensions and to study
properties of the corresponding crystal lattice. In such models one
can obtain an information about the influence of the sharp atomic
resonances on the band structure of the crystal and get answers to
some other general questions (~6S,17J).

The shortcommings of the models described above are that they

can describe the effects of s-wave scattering only, because the
higher angular momenta can not be included into the zero-range appro-
ximation in a mathematically well - defined way ( see ~ 8~). This
difficulty can be overcome in the models of atoms with a finite size
core such as bag-like models C93 and their modifications ~lO,11~
in which the presence of the atomic internal structure is imitated
by an abstract Hamilt6nian.

In this paper we consider a three-dimensional cubic lattice in

R 3 . The individual atoms which form the lattice are supposed in our
model to be semitransparent closed surfaces equiped with an internal
structume. The "centres" of these surfaces are located on the latti-
ce sites n ~ Z 3. The construction of the Blcch waves in this model
 268

can be reduced to solution of a Schroedinger equation with special

energy-dependent boundary conditions on the interaction surfaces.
The boundary conditions are defined by the internal channel Hamil-
tonian of the scatterer and are responsible for both the spectral
properties of an isolated atom and the band structure of the crystal.
The aim of the present paper is to derive the dispersion relations
connecting the energy and the quasimomentum of Bloch waves in this
model.

2. The Hamiltonian of an individual atom

The Hamiltonian of the isolated atom with an internal structure

acts in the direct sum of the space L2(R3) describing the external
degrees of freedom and some abstract Hilbert space H of internal
degrees of freedom. The unperturbed dynamics in the external chan-
nel is determined by a Laplacian, while the internal - channel dy-
namics is governed by a n abstract bounded self-adjoint operator A.
We consider here the simplest case of coupling between the external
and internal channels setting suitable boundary conditions on a
closed surface S O in the exterior configuration space.

The Hamiltonian for the individual atom can be constructed from

the unperturbed operator - ~ ~ A in the following manner. We rest-
rict the external part A - ~ Z~o to the linear set of all smooth func-
tions vanishing together with their normal derivatives near the inte-
raction surface S O . Then we restrict the internal component of the
operator A to A o in such a way that a fixed vector e 6 H becomes a
deficiency element of A ° : A o e = i e [5]. Then the domain D(A o)
of the adJoint operator A~ can be described in terms of the so-
called "real" basis w+= A(A-iI)-le , w- =(A-iI)-le as follows

u I = E 1 + K+w + + ~-w- Ul~ D(A o) (1)

where
1u = (A-iI)-lw " , ~ ]. @ (2)
The total boundary form of the operator ~ o = (- Am)* ~ Am~* restric-
ted to domain U= ~(uo(x)' u l) : uo(x) is a smoth function in the
vicinity of S O , UlE H } can be written as

o ,V - oV> : vo - u° ds ÷
SO

+~-(u) 6+(v)- ~+(u) c-(v) (3)

 269

+
where n uls n ul< and are boondary values
of the element uI [5]. The Hamiltonian h of the individual atom is
non-trivial self-adjoint extension of the operator h o . Such an
extension can be described in terms of boundary conditions under
which the symplectic form (3) vanishes, for example ,

= B (4)
~- ~+
Here the self-adjoint operator B has the following matrix represen-
tation

where •e j ~ E R , ~ E L 2 ( S o) are parametres of the model and the func-

tion b° generates the functional

<u o,~ = lUo~dS

~S O
The spectral analysis of the Hamiltonian h has been performed in Ref.
12 . In particular, the exteroal components u o of the eigenfunctions
U : h U = h U , are known to be the solutions of the following emergy-
dependent boundary value problem
-~u ° = 'hu ° , (5)

n v Uo~ = ~ u o + (D(,~) - ~ i)-i (Uo, ~b

The boundary conditions in (5) follows from (4) combined with the
the equality ~ - D ( k ) 6+, where D(,h ) is the Schwarz integral [5~
of the self-adjoint operator A :
D(A) = < ( I + ,hA) (A-,hi) -I O , ~ )

The internal components u I of the wave functions U can be recon-

structed from the external ones. Thus it is sufficient to study the
boundary value problem (5) only.

The external components of the eigenfunctions belonging to the

point spectrum of the Hamiltinian h are the solutions of the problem
(5) • They can be sought in the form of simple-layer potentials

Uo(X) = I G(x-y,~-C) ~(y) dy --G~ (6)

Y6S o
1 exp i ~ I x - y {
Here G(x-y,~) =
~ Ix-y~
 270

is the Green,s function of the Laplacian and d i s choosen in such

a way that Im~ > 0.

To fulfill (5) the simple-layer density ~ must be a solution

of the integral equation

= Ye GS' + ( D ( , X ) - ~ ' i ) - I < G g , ~'xx~ (7)

The point spectmum of h consists of negative ~\ for which the eq~7)
has a non-trivial smooth solution. The eigenfunctions of the abso-
lutely continuous spectrum of the Hamiltonian h can be found by the
same technique. For this p~rpose it is s~fficient to add an incoming
plane wave exp i(k,x) with the wave vector k, k2= ~ ,to the simple-
layer potential (6). Then the corresponding density ~> can be found
as a solution to the inhomogeneous F~edholm equation associated
with (7).

3.The lattice Hamiltonia n

The unperturbed lattice Hamiltonian -& ~nEZ 2_* 3 An acts in the

direct sum of spaces L2(R3) and H n , where An=A act in the identical
copies H n of the Hilbert space H. Let us restrict the Laplacian to
the linear set of functions vanishing in a neighbourhood of the sur-
faces Sn, which are supposed to be copies of S o placed on sites of
the lattice and small enough to avoid intersecting. Then we restrict
the operators A n to symmetric operators using the procedure descri-
bed above. The total Hamiltonian is defined as a self-adJoint exten-
sion of the corresponding restricted operator defined on the linear
set of funbtions satisfying the boundary conditions (4) on every S n-
In this paper, we are going to study the external components oi~
Bloch waves and to derive the dispersion relations.

The external components of Bloch waves can be found as a s o l u -

tion of the equation - A u = k 2 u satisfying on the su~rfaces S n the
following boundary conditions

= + (D(~) -~'i)-l<u,~2(Sn ) , (a)

where we have used the Bloch's anzstz

u(x +n) = exp i(t,n) u(x) , n 6 Z3 (9)
 271

and t is the quasimomentum of the system. Such solutions can be

represented as s sum of simple-layer potentials corresponding to the
surfaces S n centred at the lattice sites n ~ Z 3 :
u(x) = Z I G(x-y,Cq)~n(y) dy (lO)
n6z 3
yES n
with the densities ~ n satisfying on S n the Bloch's boundary con-
ditions
~n(x+y) = ~o(X) exp i(t,n) , x~S o (ll)
To fulfil (8) the densities ~ n must be related mutually by

m (x) = [ ~ e + D<"
(~)_~i ] Z3 i
n6 yES n (12)

The equations (12) together with the Bloch conditions (ll) give

< ,~o ] ~ I G(x-Y+n'~)~°(t'y)eitn

~o (t,x) = ~Ye + O(~) -~i ~z~ (i~)
xg S O y ~S °

Introducing the lattice sum

1 exp i~Ix-y-m~
K(x-y,t,~) / .... exp i(t ,n) (15)
n6Z 5 $~ Ix-y-n 1

one can rewrite eq. (13) as an integral equatiom with the kernel K :

~o(t,x) = ~ Fe + D(~ ) - ~i ] ~ K(x-y,t,~) y o(t,y) dY(15)

F6S o

Inverting the operator which appears in the relation (15),

<.,~> ~ ~-l -l ~ .,~>
[~el + D ~ ) - ~ i j -~'e I - ~912 + ~'e(D('R) _ ~ i j (16)

we get

~ - ~ e K ~o I~L~ + ~e (D(×) _ ~i ) (l?)

If the imverse operator (I -~'eK) -1 exists ( this is true,e.g.,in
the weak-coupling case ~e~<l ) , we denote R = (I-~eK) -1 - I .
The eq. (17) in terms of this operator reads
 272

~o(t,x) .....
IT12 + ~e(D( hj _ ~ i ) ( ~ + ~ R(x-y,t,~)W(y) dy 0 ~
yES
o
and projecting the last equation on ~ L 2 ( S o ) , we get a homogeneous
algebraic equation. The conditions of its solvability yields the
following dispersion relation

~e(D(A) - ~ i ) = ~ R(t,[~) ~, ~ L 2 ( S o ) (19)

If ~ e = 0 this relation must be modified. To this end, let us return

to the eq.(18), which is solvable for ~ e = 0 if

D(h) - ~i = ~K(t,~-~) ~, ~L2(So ) (20)

which is the sought dispersion relation in the ~e= 0 case.

In conclusion, let us note that the role of the projection

< ,~>~ in the boundary conditions (8) can be played by an arbi-
trary self-adjoint operator in L2(So) or even by an operator-valued
R-function ( 13 ). This enriches the scattering matrix of the indi-
vidual atom, and therefore also spectral properties of the whole
crystal would be more complicated. In the model we have proposed here
all the partial waves can contribute to the scattering processes of
electrons on the crystal. This makes it possible to model properties
of real crystals outside the perturbation-theory framework.
Finally, the model can be easily generalized to crystal lat-
tices without the cubic symmetry. Moreover, it can describe crystals
containing different kinds of atoms. The only modification required
in these oasis is the refinement of the lattice sum.

References

i Yu.N.Demkov, V.N.0strovskii: The Zero-Range Potentials Approach

to Nuclear Physics, Leningrad State Univ. 1975(in Russian).
2 A.Grossmann,R.H~egh-Krohn, M.Mebkhout, J.Math.Phys. 21(1980),2376
3 Yu.E.Karpeshina, Tent. Mat. Fiz. 57(1983), 30~
E.Fermi, Nuovo Cimento, 11(1934) ,I~7
5 B.S.Pavlov, Tent. Mat. F ~ ~(198~),345 ,,
6 B.S.Pavlov, N.V.Smirnov, in Probl. Mat. Fiz. , Leningrad State
Univ. 1987, v.12, P.155
B.S.Pavlov, S.E.Cheremshantsev, l~eprint LOMI P-15-85, Leningrad
1985
 273

8 F.I.Dalidtchik, V.Z.Slonim, JETP Letters 31(1980), 122

9 A.W.Thomas, Adv. Nucl. Phys., 13(1984), i •
10 Yu.A.Kuperin, K.A.Makarov, B.S.~Pavlov, Teor. ~at.Fiz.6~(1986),lO0
ll Yu.A.Kuperin, K.A.Makarov, S.P.Merkuriev, B.S.Pavlov,-~.K.Moto-
vilov, ITP-Budapest-Report N $41, Budapest 1986
12 Yu.A.Kuperin, K.A.Makarov, S.P.~erkuriev, A.K.Motovilov, B.S.Pav-
lov, lToc. of the School on Few-Body Quark-Hadronic Systems, Vil-
nus 1986, p.28
13 F.V.Atklzson : Discrete and Continuous Bouodary Problems, Acade-
mic Press, New Lork 1964