Professional Documents
Culture Documents
V I: 1 (Ii) Inu I (O L
V I: 1 (Ii) Inu I (O L
B.S.Pavlov
Department of Mathematical and Computational
Physics, Institute for Physics, Leningrad
State University, St.Peterhoff
198904 Leningrad, USSR
(1) V i :
1
(ii) inu i = {ol
If dim~i~oO , the last condition is equivalent to
~±i = [Ran(A0 ~ i I ) ] ~
+ U~+I ~I
Ws:T0s , wZ:--es
A
form a Riesz basis in the subspace ~ = ~i + ~-i "
Proof is obvious.
u = fS (-+W
~s +s + ~ w ~ ) , ~s± g C
= A(A-iI)-~ + + (A-iI)-~- .
^AW s
+=-Ws , ^A W _e = W +s
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.
A = A0@) AI
J- =r~ ÷ , ~i)
where P is a self-adjoint operator, p = p ~ , in ~i "
and
*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)
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 "
A
D = D O + ~p , (4)
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
(A-iI)-lh+A(A-iI)-~ + +P(A-iI)-I~ + = 0
h = [A(A-iI)-I+ P(A-iI)-l]~ + ,
+
= (pAP + p ) - l p ( A - i I ) u
W+ = A(A-iI) -1 g , W _ = (A-iI) _ Ig
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
2. Krein formula
(Ap-~I)u = f (5)
A+iI
proof : The result follows from the relation (in an obvious notation)
I I+ A @i + ~ O i
@4- A-iI(I-P)A~-~ + A__~ii_I+
Y~:'~A @i ' (7)
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
f I + as
Q(~) = ~L-~ d(PEsP) ' (8)
^
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
= f (9)
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
References
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)
(MC~)f)Ct) = 4 ~ ( t ) f ( t ) , (1.10)
1 4 _
and
and
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
and
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
and
fl(~)adom(K), gll (~) & ~ - i ' g21 ( ~ ) ~ i ' for every gle'Q-i and
~ £ A C ( ~ I ) . Similarly, we get
nm
n---) +oo
Sn(tl = Jp(tl (2.12/
18
) 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= (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
and
A ( ~ ) -- ALI'](~ ) (2.22)
A(1) = I, (2.23)
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)
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
A
dom(K) = dom(K) $ (V + I ) F , (4.1)
h = f + gl + Vg1' (4.2)
Q= ~ ~. (4.3)
and
A ( ~ ) Q = QA(~)Q (4.6)
If,f] = o. (4.?)
[f,gl = 0 (4.8)
22
1
Q = ~ \~ ~,)" (4.10)
~. 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)
~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
References:
[I~ T.Kato:Integration of the equation of evolution in Banach
space, J. Math. Soc. Japan 5(1953), 208 - 234.
K.A.Makarov
Physical Institute, Leningrad State University,
1 Maya 100, Petrodvoretz Leningrad 198904
J(/,(u),6(v)) = o
A~ = A 0'~~ (4)
I Dex( % ) O )
D(A ) = (7)
0 Din(A)
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 :
(12)
and the external boundary form can be written in the terms of the
asymptotic boundary values Uo, Ul:
(;o): (15)
with some Hermitian 2x2-matrix ~ .
~'00
,t-'=
rOlcll) ' Vjje~. (17)
to1
Hence we get the following expressions for scattering amplitude
f(k) and s-wave scattering matrix sO(k):
(i (22)
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
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
S(x,k) = I + ~ f ( x , k ) (30)
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.
(33)
+ 2 - i
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
~aDom(-A) O
and restrict the operator (-•)0 tc %~. Then the fcllcwing lemma
is valid.
u-_~ Im Gigo(U) ,
9;(u) •
S~(u) dy = lim ~ n u ds
~l
n ~ : {X~R6: dist(X,n) : #}, (38)
and
(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.
Acknowledgements
References
M. Demuth
Institute of Msthemetics, Mohrenstr. 39, 1086 Berlin-Mitte, G.D.R.
i. Introduction
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
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
(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
Tt Co ( ~ n ) C Lq(~n)
( Co - space o f continuous functions with compact support) and
l l T t f UL q llflILq , t ~ 0,
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
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 •
= 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
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
with V_ = max ( - V , O) .
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.
and
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
0enotatlon 15: Let (_~Lx, ~x' PF' ~(')) be the Feller process given
in Assumption A. Then we set
.-'- -- .
for all t ~ 0 .
:= ! - P •
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 .
VM := MP
(e - t ( l < ) G ) (x,y)
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
e- t ( K ) G 0 - 0 e- t K
e - t ( K ) G ~ - ~ e- t K
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
f dx f d y I (e-tCK)G 0 - 0 e ' t K ) ( x , y ) 1 2
Now let
e- t ( K ) G 0 - 0 e"tK , O< t = T ,
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 )
• ac(K) = %c((K)G) •
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
f j ( x i . . . . . x j . l ) ~ x j ~ g j ( xI . . . . . x.3_1), 3=2 . . . . . n ~ .
e-t(K)D O - 0 e- t K , 0 ~t ~ T ,
with ~ )n.
sup ; p(s,x,u) du
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)-/'~
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
References
V.D.Koshmanenko
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-
ven a sense as bilinear forms in H. But these forms are not olo-
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.
Hb H
vn ) v, vn ~ o, n ~ co . (2.1')
b[u] -- o }
5o = H. (2.3)
dense in L 2.
by the formula
w £ H \ H.
i
Thus
Let us decompose ~+I into two subspaces ~+I = Kb~ Rb' where
(i) b~l'Sk (b 1 % ),
(ii) Ran(Tb)~ H (k+1) \ H k ,
(iii) Kb is dense in Hk .
61
II % II
2
k ' o and
~(k+l) [ % - u ] -~ o, n ~ oo.
turbatious of A.
Proposltic n 3.1:
AR b = N, (3.2)
~-I = A -I + B -I (3.4)
Ab g = A~ f. (3.7)
64
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.
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
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
x = xo + x l i + x 2 j + x 3 k - - - - , ~ = Xo - x l i - x2j - xak.
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) = 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
S(x) := - 0 g(x) - 2l 14 , := s ( x )
~ : = { ~ E C ° ° ( ~ 4 ' f f - / ) ; '~l~+~lim~ o ( x ) = O , O ~ p E S }
• = [ 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
OA(x) = F(x)
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
the p-th moment, i.e. J la]Pu(dp) < c~ for p = 2, 3 , . . . , then A also has the p-th moment
p : A(x) ,-A(-~.) , x e ~ : l 4.
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
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
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
\ ]
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
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
with
xEAs
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 :
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
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.
Acknowledgements
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
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:
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)
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
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]):
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
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.
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
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
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.
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
I~ ~ n i b i , (nl,n2,n3) E ~ 3 .
i=l
91
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
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
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
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:
with
1
00 - - ~ ( b l + b2 + b3),
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.
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].
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 ~(-~,^).
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
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
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
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.
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
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
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
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
H0 + ~ 4 ~ ' , ~(t)> ,
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
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
+- -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
= ~. ,
o) • (9)
dt -I 0 I
105
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
2~ = (+M) 0 + iN Iq -lq
o ~ (-M Zq lq
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
- =
o) + i ~
(_, _,) , (11)
0 -tk Ie Ie
=
(me) (lili) + i~. , (12)
0 A Ii Ii
^
¢~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) .
~ = VU
v~ ~
't,,) ~k
, k = 1,...,4,
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 •
+ 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:
~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)
L ( A k) = ( A k - i) 2 _ M 2 ,
10g
zO:z,,
+
z~:z~, z~:~ ~ - ~ .
= fu ~ =
( ~ u ) ~u ~d = 0" (21)
fd Nd
= =
f(x) =
+ (22)
+ 16~T 3 R+ 3 dy 2
we obtain:
=~ Kn exp(i~nt)!3 dy I~n(X)~I~n(y)ITl~f(Y )
, exp(iA0t) .
: f~-~ f=
~pl ^
[~(@,~)~](p) = ~(p) + • foo(p2)<u~(Jpl) . (26)
2SN
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
^ ^ ilP~ ~ 2
~s±ul(p) = ~(p) + . A±(p2)<~(ipl) . (~)
2~
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
Yu~ A. Kuperin
Department of Physics, Leningrad State University
Leningrad 19890~, USSR
Abstract
1. Introduction
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
(ii) The compound state Cif decays into the outgoing chan~el
f E ~ , whose energy is equal to the energy of the incoming channel.
h ex u : - (~ + V(X)) u
+ +
where n. V + i s the normal derivative on the surfaces
dist (x,~-)--.~ ].
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.
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] :
( h- z ) ~ = O ) x E ~3\~ (5)
= ÷ 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 :
( ho -z ) u0 = 0 , x6 ~'\y (9)
<~,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).
,° ®
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)
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)
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
3
f~(y~)=<~=~2:%(-~)~o, ~ > (&). (2.
Hex
uo t
(25)
H~ =I - ~ y u + A u~ - S~w~+
+ 8,o, w&
- +
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.
3
Ua : >-" ~ab ( z ) fb " (28)
b:O
± 3 +
~ : ZI ~jb(z) 'fb , ~ : ~,2,3, (29)
b:O
( H- z ) ~(z) F = F
one can obtain a ~et of equations for the kernels of the operators
A~'R~,b- w~'~:b + w~
+ '~'b - (~y + z ) R~b =S~.b I ~ ' (32)
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.
[ n.v%] • =-~.
r=
(%(,) < "oh', ~=> * (39)
× ro(y~-y~, z- A ) ,
/
In view of eq. (39) the kernel G( X, X, z) of the quasiresolvent
G(z) obeys the boundary conditions
r~
3
(H ex + ~- W~(Z) - z) G(x,x,z)
I I
= 8 (x - x). (~)
3
G(z) = Ro(Z) - Ro(Z) Z v~(z) G(z). (45)
by the relation
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)
(I ÷ v~ ~o) ~= % = ~= ~o (~9)
u o : Ro( ) >-- V# ,
(H ex - z) Uo(X ) = O , X ~ F • (54)
£~
÷ : <Uo , , (56)
135
÷
~: = Q~ (~) ~ (97)
+
(-ny+A.-z) u = e~(Y~) w: - e:(Y~)w+~ + (581
( H - z )~ = 0 . (59)
Hence we have proven that the Faddeev equations (53) are spectral-
equivalent to the SchrGdimger equation (59).
~. Discussion.
Acknowledgments.
References
R.A.Minlos
Moscow State University, Moscow, Lenin's Hill
where ~ I
= (x I ,x2,x3) xj=xi=x g ~ij and Ai~ , Bi~ are functions
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.
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])
+ 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) -
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)
, mlm2m3 z ,j(p) -
I
~'~I s12 + ~ s2 + $3
I s~ - z
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.
~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
/ m m, V/2
sin T i = !((mj+mi~mk
~ K +mi~/A , 0 < ~i ~ i~)" . (22)
%.
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
÷ m3 p2
V - m ( m + m3 ) - z u(p) +
) %) ÷ '
=ZCp,z)) m = m ~ =½ .
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
I. Introduction
~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
(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,~ •
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)
~ - , k',s=4
~o ~- ~ < ~>~o,
~,S:4
Y : - a % * (I+K I % ) -~ .
= A- ~ j "
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
: g j- ° i .
BI_ ( ~ I ( z ) ) - 1 B
= o (15)
det & O (z)
B B2+
~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.
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 :
Zs '-s s "" @ ( $S ( 1 + Zs
2o)
z'_ \ z-1 I \ z
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
~/'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,
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.
+ 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.
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-
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
6. Faddeev equations
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]:
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
Acknowledgements
References
Alexandr I. Mogilner
Institute of Metal Physics, Academ~ of Sciences,
Ural Branch, Sverdlovsk, Kovalevskaya street, 18, USSR
Abstract
H = - ~I m,n
~ Jm,n (S~S~ + bS~S~) + D ~ 2( Sn~ ) (I .I)
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
x = 2 Q-If(n}X~n~ (1.4)
(n~ g
'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
Hf = FN f + bSBf + Vf (I .6)
F = D(I-2S) + SJ , J = ~r Jr '
162
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
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~
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.
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)
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.
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)
and the functions Vn(kt(n) ,k"t(n)) are the kernels of the N-magnon
kinematic interaction operator.
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 ~
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 .
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.
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].
X~n'}(K) = ~ ei(K,R )
%N rni (3.2)
~ rn I- rn 2 ,''', rnN_1- rn N }
Let us now suppose that O < b < < 1 . It is easy to show that the ope-
167
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
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.
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
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
Eo= ~ ( ~ - 2 s ) + D N(~-~.____.~ z
-
D + o (3.~1)
and the effective mass tensor components of this state are given b~
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
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
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.
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.
• 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
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.
References
B.S.Pavlov, P.B.Kuzasov
Department of Mathematical and
Computational Physics, Institute for
Physics, Leningrad State University,
I9890@ Leningrad,St.Peterhoff, USSR
where :
uo ~u; ~u o
[ uo] u; - uo i~T ~ %n ~
+ wi u~ wi
ui = ui + + - + ~i
*
< AioUi.vi ~
> - <ui,AioVi b : ~ -vl
- - ui
+ v-~[ --- i ui ~ V i (1.2)
(1.3)
We shall consider the symmetric operator ~ 0 acting in the Hilbert
space
.
< ~ 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~
/
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
<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
ui(x,y) = f(x,y) ui
u[
u~ Di( kz
2 )
Di( ) = O5
+ + m
(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 )
IRi 2 + Isl a : 1
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.
~Uo~
i expi( +kyy)
183
~-~ 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
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.
- (~-t 2) + = o (2.1)
2 DI( h-c~ ta)
t ~,:t 2
f ~ - - - D , : 0 D,: co
/-
/
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
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 :
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.
One can check directly that R+ = I holds for k~ < A; in this case
the incoming wave is totally reflected by the potential barrier.
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.
• 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
+~(+ 2
I~12D2 2 el
2 2 ~ + (r ~ +~') I°(12
2 D1
188
~%~12D2
~+~'+ 2 ~
I~I2 2
2 nI 2 + (~+~') ~2D2
u0(x,y,z) ~ u ~0 + u0 + o(RI)
4~ R1
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:
ik
G O = -- I H (01)(kRlsine) exp(ik(z+Zo)C°Sg) S(kcose) sino de
8~
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
u0 = S exp-ikzz o , k z < O
2) u0
~+ : 0 ~- : B- e x p ~ z o
~i = - A x y + Ai + g(x,y)
(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.
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
O. Introduction
by t h e H a m i l t o n l a n
H = H0 + V, (i)
xd I Xd=O I Xd=O
(2)
(for d = i, cf. Ref, 1). Since the boundary forces are usually
tonian HQ i n s t e a d of (1).
macroscopic scale.
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
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
2. Ergodic s u r f a c e potentials
resp. ~ L the
HQ, restriction of H~~ o n t o ~ L X [0, L] r e s p . A L X ~+,
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
~nL(E) = N ( H~ Q~ , L , E ) / L d - I ~
= N(K~,E, L ,O)/L d-1 (7)
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
u( x , O )
~x d = Q(x/L) u(x,O), x e /\ L' (11)
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)
and
(i) (i)
Q2(x) = sup { ~ ( y / L ) : y E /\L ) for x E /~L
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)
done.
llm N(HQ,L,E)/Ld-i =
L-.-- oo
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
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
+ *
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
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)
Proposition 2
Every self-adjoint extension Ku is uniquely specified by the
following boundary conditions: f = (fl,f2) ~ D(K u) iff
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] .
PH u c H u P (i. 15)
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)~ 2 + ~ B ( k ) ~ z = i. (i.18)
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).
= co ~2
HO,z - A ; D(Ho. 2) = Co( \ {0~) (2.1b)
t- t- +
0 s = (f ,0) ; 02 = (O,fo) ; 0 = 0
Proposition 3
Every self-adjoint extension fulfilling (2.3) is specified
uniquely by the following boundary conditions.
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
[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
Lkx -Lkx
~i (X) = e + A(k) e
(2.5)
~=(r) = B(k) H <i~ (kr)
o
(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
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
I = 2e ; T(E) dE (2.10)
E '
F
dU -i
(2.11)
212
References
dU /i ///
</////
\
Ev =
= i;
1 , /9 =
1.5;
1,
2; 2.5; 3
i 2 eU
dU
dI
Er = 1 , ~ = .i, /9 = i,
.......... _
..... I I ! I
i0 20 eU
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)
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]
Copper contact
R=5.7.o. ,T =1.2K
dUldl
15
F•
c
dU 2
Jm
0 5 10 15 20 25
Voltage [mV]
I. Yu. Popov
Department of Mathematical
and Computational Physics, Physical Faculty,
Leningrad State University, 1 Maya lO0,
Petrodvoretz, Leningrad 198904
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)
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 )
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}.
• _-'; +~).
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).
~u 9v
J [u, v ] = ( NB-- V - ~ an ) ds,
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).
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 .
in in in
lim (Gex( x, Xo, k) - Re Gex( x, Xo, k 0 )) = D ex ( k ) .
x-~x 0
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
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 )
s(o~,9, k) = S e x ( ~ , k)
± k +ex( Xo,
9 ,l)+ex( Xo,~,k) k2
2 ZC(Din(~) + Dex(~))
223
i I~n(Xo) t
k= kn +
+ o ( in -a II, ol),
~u
u) I = O.
( ~n @O
in
Gex( x, XO, A ) = ( 22T~x- Xol) -I +-
+ (4F) -1 ( H + ( e ~ ) - i d ) in (ix - Xo O - 1
+o(i) .
Theorem:
9G D
lira O ' G ~ ( x, y, k) = ( x, y, k); x a @ C ~ , y~ @~
~-, oo B nx
225
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
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
_- ~ +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
~G C
gin + geX = (
(s - Xo, nF ) ~ (s, xo) ds) - l ( s ~ I -
Cex
-J ( s, xo) ( s, ds
cin ~nC ~ Xo) -
~2G~n , ex ~2~in'~D ex
gin, ex = ( ( x, x O) ( x, ko)) 1
n2 ~ n2 Xo' x = x0
z I% +
n
+N(~-~O)~ItN~p 12 (ANm+p
0 _ ~ )-i (~°m+p - ~o)-l +
N-1
0
l*Nm+p+j 12 (~m+~j - ~o )-l = O.
j=l
References
W. Karwowski
tential forces.
H = -A + V.
Very often this formula has only intuitive value and requi-
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
both the point interaction and that described by the bona fi-
de p o t e n t i a l s .
boundary conditions on S.
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
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
6+0.
cal potentials.
(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
+
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
Remarks
any n.
Examples
Put
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)
i) f (X_~0) = f (X_{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)
the angle ~ , tg n = I£ I /P 3.
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
tants ~,B,V.
E ÷ 0 E ÷ ~
ACKNOWLEDGEMENT
done.
REFERENCES
B.S .Pavlov
Department of Mathematical and
Computational Physics, Institute for
Physics, Leningrad State University,
19890# Leningrad, St.Peterhoff, USSR
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
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.
(°/
lation- invariant boundary conditions
~s t~z 1
In both cases the spectral structure of the resulting self-adjoint
operators
D~ = I D ext
o ~ Dintl
o + L~
exp-ist b°o(k)
\rlo rll
then the dispersion equation becomes more simple
As # A for s ~(Sl,S2,S3,...,sn)
int( k ) : ~ ÷ ~ ; w s ÷ ~ : w s
Ts
we get for the boundary values in the exterior space
+ext = ~fext( k )
~s s "
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
Hph = ~ T1 + ~ o
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
(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 ,
M = { x=m, s=(m,m,m,...,m)}
- 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
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
- ~ ~ 7. (A) n = - C A in
nEZ I o ~ o
( )( -u(s)
~s
:F In']
~s
+ (s)
) , F : I ,
d2
- ~ u~ - fr
on
- Iroll~Crll - d<~, ~ ~ - l ] Z
n 6 ZI
%c~-n~uc~ :
= % u(x) ;
( ]~oo- D ( k , t ) )
Here D ( h ,t) is one-dimensional lattice sum,
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
-
( As® It+Is@A t)
•
s6Z ,
Ls (8)
~I
-y2t(~)y~( v)]~l ÷ 7S t I ~st(~) [ +S ~ (v) -- ~S~(~) ~ ( V ) ] "
- 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
d2 d2
* d 2~) 7 ut (IO)
{(%)o-(~+d-~ i 5 2 = f~'
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
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.
References
Ca) ~ c~ ,
(b) ~, g" , ~'" are bounded,
- + + V(s,u)~ (1.2a)
for ~D , where
d o ~ d+ m 2~+
d .... "" l
E =4~E~ (1.5a) n
/' I 2
, !
independently of d . One can use the d X
=
j~1
.:
(-I)J+I rj e qj(1-y/d) ,
r = Cr (1.6a)
Jk
Cj k = I ( I - e -2qj) ~ (j2+k2_~) (1.6b)
4 = o.93 (1.5b)
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
ds 2 -
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
a b b)
S(k) = b a b (2.7)
b b a
with
-2ikB
b =
[I - i k ( A + 2 B ) ] [ I - i k ( A - B ) ]
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) =
(uj(lj)) (o)/
\uj(lj)/ = ~J uj,(o)j , j = 2,3
kuj
265
e2 T(E)
'i-T(E) (2.11)
References
I. Introduction
o ,V - oV> : vo - u° ds ÷
SO
+
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
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.
m (x) = [ ~ e + D<"
(~)_~i ] Z3 i
n6 yES n (12)
The equations (12) together with the Bloch conditions (ll) give
one can rewrite eq. (13) as an integral equatiom with the kernel K :
we get
~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
References