Anda di halaman 1dari 8


5. 6. 7.

A. F. Vakulenko, "Unitary regularization in the many-particle scattering problem," Dokl. Akad. Nauk SSSR, 249, No. 4, 825-828 (1979). L. D. Faddeev, "The inverse p~oblem in the quantum theory of scattering. II," in: Current Problems in Mathematics [in Russian], Vol. 3, Moscow (1974), pp. 93-180. L. D. Faddeev, "On a model of Freidrichs in the theory of perturbations of the continuous spectrum," Tr. Mat. Inst. Akad. Nauk SSSR, 73, 292-313 (1964). N. Dunford and J. T. Schwartz, Linear Operators, Part III: Spectral Operators, Wiley-Interscience, New York (1971). B. S. Pavlov and S. V. Petras, "On the singular spectrum of a weakly perturbed multiplication operator," Funkts. Anal. Prilozhen., ~, No. 2, 54-61 (1970).


One presents a new variant of the theory of periodic approximations systems and C*-algebras, namely the

of dynamical

construction for each automorphism of the One

Lebesgue space of a Markov tower (or adic model) of periodic automorphisms. gives several examples.



In [i] we have outlined the proof of the theorm on the simultaneous uniform approxima2 tion of the multiplication and shift operators in the space of L-functions on a dynamical system with an invariant measure; more exactly: let i] with an invariant Lebesgue measure and let >0 there exist operators in L 2 , I I ~g U (~ ~ and S be an automorphism of the segment [0, for each ,

be a bounded measurable function; V c~ , such that U ~ and gU s - U ~ V~ M~ II<~

IIM~ - V ~ here Us

and the *-algebra generated by

is finite-dimensional; and N'g

is a unitary shift operator of the dynamical system, If ~S~)~) = ~ + ~ ) ~ i ) , ~ = ~ O ~ ,

is a multiplier,

is the operator norm.

then the problem consists in the UMU "4~'~=~~I.

simultaneous approximation of the operators,

satisfying the relation

This problem has been solved in [2] and has been formulated earlier by the author in [3, 4] and by Rieffel [5]. The above formulated theorem has applications in operator theory, in the of dynamical systems and in ~ -algebras; it can be considered as a

theory of approximations

variant of the noncummutative

constructive theory of operators

(see [4]).

The proof of the theorem is divided into two stages. the theory of periodic approximations

The first one is a new variant of

(Theorem 3 in [i]), namely the construction of the (see below) and is considered in this paper. -algebras and will be considered elsewhere. The

Markov towers of periodic automorphisms second stage refers to the theory of The approximations ~

in the theory of dynamical systems, as in the theory of equations, or the moment method (projective In

can be partitioned into two classes: Factor-approximations approximation) and subapproximations

or the method of nets (inductive approximation).

the first case we investigate the behavior of the operators on finite-dimensional


of functions or on a collection of sets and in the second case we investigate the behavior Translated from Zapiski Nauchnykh Seminarov Leningradskogo Otdeleniya Matematicheskogo Instituta im. V. A. Steklova AN SSSR, Vol. 115, pp. 72-82, 1982.

0090-4104/85/2805- 0667509.50

O 1985 Plenum Publishing Corporation


of the system on finite collection of points.

The approximation considered below is of the

second type; we approximate an automorphism by periodic automorphisms whose trajectories are included in the trajectories of the initial one, i.e., we approximate random permutation. This variant is related to the author's paper the automorphism of a

[6] on the scale of an sequence of

automorphism and makes use of a modification of the construction of a consistent periodic automorphisms, introduced here consists periods with respect tion; in return, found there (the lemma of Sec. 3 in [6]).

The important modification

in the fact that we do not require the constancy of the minimal

to all points for each of the periodic automorphisms of the approxima-

the consistency of the approximation steps is stronger: Each permutation

differs from the approximated automorphism only at one point of its layer and, as a function on the quotient space relative to the trajectories of a periodic automorphism with values in the (variable) group of permutations, is measurable relative to some additional partition term of the approximation (see below). Such

to the trajectory partition of the subsequent a rigid requirement excludes

in many respects the additional arbitrariness

in the approxima-

tion which one usually has.

Thus we arrive at the concept of a Markov tower or adic autoThe construction of these models or In

morphism model, whose existence is proved in this paper. approximations in concrete cases is a complex problem;

in Sec. 4 we give some examples.

Sec. 2 we give the definitions

and in Sec. 3 we give the proof of the fundamental

theorem. (in the

We recall that the existence of a periodic automorphism, sense of the theory of measure) lemma;

approximating uniformly

the given one, is the content of the classical Rokhlin--Halmos consistent in some sense, the

the question of the sequences of periodic approximations,

has been considered in different ways in a series of investigations survey in [8]. [7]) and the projective approximation

(see, for example,

(by multivalued mappings)

has been introduced see [9]) must

We also mention that in our view such an approximation


be an object of investigation also in hydrodynamics tion with the approximations, 2.

(in its metric variant)

in close connec-

so fruitfully investigated

in the works of O. A. Ladyzhesnkaya.

Markov Compactum and Adic Transformation In the sequel, for each automorphism of the Lebesgue space we construct a tower of But first we define an appropriate model space and its transform, (i..e, to establish an isomorphism) towers. ~ , ~=i~... M~ =(~J~) ; let %~-i ~ an arbitrary pair; in other

periodic automorphisms.

to which we wish to reduce

words, we define a class of canonical Let M I , M~ , .~ ~ ~ = i , ~4, '''

be a sequence of natural numbers -matrices,

be a sequence of Z +

where the matrix

has dimension

. ~ = i,...

We shall assume that none of these matrices has zero rows or columns. ~=0~{ ~=0 and the number of elements is I~I=~ ~[ ~I) whose vertices are partitioned into finite , ~0 ; the arcs of the j of the

We consider the graded graph "levels" ~ o = [ ~ ] , ~ , ....

graph join only vertices of adjacent levels and the number of arcs joining vertex ( ~i ~-i )-st level with vertex , + =O,i~.'. ~ of the r level is equal to ~K

, K=~,...,%~, j=l, .... each vertex is joined ~ 6~0. ) of the pre~_~ and

By virtue of the conditions on the matrices,

at least with one vertex of the next level and (except for the vertex vious level. By a path in the graph we mean a sequence of arcs


, where


have a common vertex from the the paths of the graph ~([

~ - t h level,




be the space of all (A neighborhood

M ~I~=~)

' equipped with the weak topology.

of a path is the collection of all paths coinciding with the given one up to some places; obviously, ~([~) is a compactum). By a Markov compactum with parameters ). If ~---- 9 and ~-_- M ~ , then [~,

Definition i. ~[~) , ~ =~{~il

M~ ~ i=i,...,J we


the Markov

is said to be stationary. In topological dynamics one considers frequently stationary Markov compacta and their two-sided shifts. another kind. We shall consider arbitrary Markov compacta and their transformations of ~ ) ~ and of the measures on them, in , will be considered elsewhere; -algebras constructed over the

Various properties of the compactum

dependence on the asymptotic behavior of the matrices they are related, graph ~ , in particular,

to the properties of A F

as over the Bratteli diagram [i0]. to the case when the elements of ~ consist of zeros and ones.

We restrict ourselved

In this case the graph does not have multiple arcs and a path is a succession of vertices; this assumption does not diminish at all the generality since by introducing new vertices one can get rid of multiplicities. In this case the Markov compactum acquires a usual description: ~ ~ , consisting of all sequences [~I for

it is a closed subset in the compactum which ~ ~*i = i


We give a direct description of the structure of a Markov compactum. Proposition I. with a distinguished lowing properties:

Every Markov compactum is a totally disconnected separable compactum X sequence of finite partitions ~ ~ ~lJ , possessing the fol-

l) V
2) for each K>~i and any set ~ of the partition element ~ the correspondence

defines a homeomorphism



denotes the product of the partitions,

i.e., the partition consisting of all pos6 is the partitioning into isolated points. ~ such that ~(~ ~ ~(9) , where ~/~ denotes the

sible intersections of the multiplied partitions; Condition i) means that for ~(~ is the element of ~ ~ ~ ~ there exists ~ .


In condition 2) the s ~ b o l

factorization of ~ the following: ~/~ for If

with respect to the partitioning and the meaning of the condition is ~,~6~, then there exists ~ ~/~ by ~ for which for ~ the projection onto is the same as for


is the same as for

~ and onto ~

If one denotes the number of elements of and ~ by M~ , then the compactum ~ X

and the intersection matrix of ~-i ~ [~M~)) ;

from this proposition becomes ~


denoting by

the partitioning of

into the classes of paths passing through

a given vertex of the

~ -th level, we can see that the conditions of the proposition holds.

We shall use the definition from the proposition. 669

A subset of a Markov compactum , will be said to

that is measurable with respect to

~ ~

for some ~ ; a

be cylindrical; " ' [ . ~ ~

these subsets form an algebra, denoted by


transformation leaving

invariant and

.9 ~

fixed is K -cylindrical.

Each E -

cylindrical invertible transformation of a Markov compactum is a periodic transformation whose trajectories lie on the elements of the partition T ~ . By a transformation of class

we mean a homeomorphism

for which there exists a nowhere dense closed invariant set T is a ~-cylindrical (peri-

such that, in the complement of any of its open neighborhood, odic) transformation, where ~ depends on the neighborhood. (periodic)

Thus, the transformation of class

are limits of ~-cylindrical

transformations in the sense of uniform conver-

gence on any closed set lying in the complement of some nowhere dense set. An example of a transformation of class restrict ourselves in the sequel, ~ of a very general form, to which we shall

is obtained in the following manner. ~ by natural numbers from i to ~ ~i , ~=~... for all On the

We index the elements of the partitions Thus, we can introduce a lexicographic assuming space ~ (~,...,~)~(~,00.,~) , if

ordering on the set of elements ~ = ~


~=0,~,.o,,~ ; ~_~ > ~_~~

there emerges a partial ordering in which the sequences coinciding from some place the order type in each class of comparable elements is either , or ~ i~" , or

are comparable; ~+, or ~_

A Markov compactum together with the ordering of the partiZ , one by type

tions is said to be minimal if all except two classes are ordered by type ~+ and one by type ~ Z_ : El< K~ < ...

Minimality can be expressed in a simple manner as a condi-

tion on the matrices Proposition 2. "...'~,~i ordering on

There exists a sequence

such that the matrices


do not have zero elements ~



it does not depend on the

; by the change of the indexing only the one-sided orbits are changed. By an adic transformation of a Markov compactum with ordered partitions T which assigns to an element (If ~ ~ the element then T ~/ which is is not de-

Definition 2. ~ ,

we mean a transformation

directly larger in the sense of the ordering. fined.) ~+ If and ~ ~_ is minimal,

does not exist,

the we define an adic transformation by combining the semiorbits

Proposition 3. class

An adic transformation of a minimal Markov compactum belongs to the

The proof follows directly from the definition of the ordering, verse set in this case consisting of one point origin of the orbits Remark. (i.e., ~+ , respectively).

the exceptional and in~and the

(the end of the orbits of type

The trajectory partition of an adic transformation ~ ~ ~ , where the symbol ~ is the set

is the tail partition of intersection). i.e., of the substitu-


Thus, an adic shift is the limit of cylindrical tions of the first coordinates of the sequence. sive substitutions


It can be visualized as a tower of succes~I~ ~ .... It is the model to

of the elements of the partitions

which we reduce an arbitrary ergodic automorphism of the Lebesgue space (see the examples in Sec. 4).



Fundamental Theorem THEOREM. Let CX,~) ; 0~o be a Lebesgue space and let ~ be its ergodic automorphism with $ , invariant and an auto(with ordered eleare iso-

invariant measure ~ with respect to morphism S/=~0) X

is the countable algebra of measurable sets in XocX , Xo~ ~ o

There exists a set of full measure ~'X, =xo ~i as well as a

minimal Markov compactum CX,, $/, ~o) and

ments of the partitions (~)

), such that ~:X, )~ ,

(~, T , ~)

morphic, i.e., one has a mapping ( ~ tion.

defined everywhere on

Xo and ~ ' ~ =~o

is the algebra of cylindrical sets) and ~$'~'I=T ,

where T is an adic transforma-

Our problem consists in the construction of a sequence of partitions required properties.


with the

But we start with the construction of periodic approximations, which

will correspond to cylindrical periodic transformations in the image. If ~ is an arbitrary measurable set, ~>0 , in X , ~B ~) then each ergodic autorelative to the subset =~CB)~) and ~ B ) =

morphism can be represented as an integral over the derivative (see [ii]). ~C~,$) We fix an automorphism and we form partitions K~)

in the following manner.

consists of at most a countable number of elements and the elements of ~eB ~) are ordered in the natural

Bo=B ,.Bt:S%\Bo,,. ,B~:SB~,\ x~ B~, ~sOs

manner; the element ~ since Let , 5 where ~ ~(8)(~)) ; ~ is ergodic, we have if

containing the point and %5 ~ ) < co and ~

, consists of all points of the form If %6[~)= I~C~)[~)I , ~8,5 =Su~55 ~s <~) 9 ;

are mutually complementary.

almost everywhere; we denote

S~B)~ = S ~

S~e~B) C~)




~ ~ ) ~ )


is a

periodic automorphism with period depending on the point. LEMMA. Let ~0 be an everywhere dense algebra of measurable sets, invariant with re5 There exists a sequence of sets %A~<~176 for all ' ~AI <

spect to the ergodic automorphism ~<9 ~ ) :0 Proof.


A4D ~



for which

We select an arbitrary

' A,



We denote



a ~ '



is the union of all but a finite number of elements of the partition ~ ~ ~Ai < ~ A~ and since A4 is invariant ~

having the measure to ~= and therefore



is measurable relative We conwe As a

lies in ~ and on

We note that A~c A~ X ,S


DA~,~ <co

sider , ~ = ~A~

, the derivative of A4 , ~4

A4 and

, ~A~ < ~ ~ ~ , A~

apply the same method to result, we construct ' =~ ,

~'z as for

in the first step. Since


, %~,$ < c o

and, therefore .

~A~,S <co "

we have

.~& < ~ + ~ = ~

Continuing this process, we and

obtain a system of sets intended to prove. We denote Obviously, ~.=





~A~,S < o o ,

which is what we

6{~h.) ~ ~ ~

the required periodic automorphisms have been constructed. =0 Now we construct a sequence of refinements K ( A ~ )


which brings in the structure of a Markov compactum. Let ~./~ 6 be an increasing sequence of finite partitions (here 6 is the partition of sets from ~o and generating 0~o ~ We identify Let

into points .~0~0 ), consisting X/~) and A~

, associating to an element its unique point from


% :
Obviously, Let ~ ~_~ = ~ ;~) V.p~ 9 for all ~ since

--,- A,,.



~ = ~

We define by induction the pattitions ~

~=~,$, . . . .



of ~K

inherit the ordering of Since ~CAK~

~(AK; SAKe) .

verify that the required properties hold. we have ~ ~) "~'~'~=I ~ and, therefore,

We V %C~C~;S~))...~.IC~CAK;~A~_~)) -t -I , ~(A~; ~) , ~ ~ ,

the last partltlon" " is complementary to S~ By 9 Now, ~

i.e., to the trajectory partition of the periodic automorphism since ~[~ ~'<~(~;~) V~ ; consequently,~_i~~fn p ~ = 6 .


the sequence

is Markov and in one element ~ D e ~

the automorphism

$~_~ has the same periods for remain fixed.

all points, the elements of ~4 ~ Finally, all elements of all ~

being moved while the elements of ~=~ .~ lie in ~ ~0. with respect to the partitions

Now we define a Markov compactum flect ~ e : 0~ ~ into the space ~ X4 =~


and we re-

considered as the infinite product of the finite sets

with the weak topology. ~)~ ,..)~ ~ is a

From what has been said it is clear that the mapping monomorphism and the closure of the image of a set of ~c ~ ; the image ~ is the algebra of cylini~l~ '


complete measure is a Markov compactum drical sets; the Markov property of where the elements ~ ~

follows from the Markov property of ~

are ordered, and the transformation

goes into the adic shift as ~ ; ~)~ for this and

one can see from the construction. we note that from the erogodicity of E E ~s there exists ~ = ~(~,6) ~ and

Finally we verify the minimality of ~

there follows that for any elements ~5~)~ E) > 0

for which

; this means that the

matrix of the interactions Sec. 2). Remarks i.

~S consists of positive elements (see Proposition 2,

The theorem is proved.

Our method of construction of


is not economical since the refinement of ~ ~K ; as far as possible,

is not accompanied by the lengthening of the trajectories of ~ must be "almost" an independent complement to ~

The construction of economical

realizations is a complex problem (see the examples). 2. By the theorem of [12], there exist an invariant algebra consisting of strictly ergodic sets. ~o and a relatively

ergodic automorphism,

If in the theorem one starts

from this algebra, then the corresponding adic transformation will be strictly ergodic. 3. [~i The asymptotic properties of the matrices of the intersections of the partitions contain ample information about the metric invariants of the automorphism of the

scale type (see [6]). 4. The theorem can be easily generalized to nonergodic automorphisms; in this case the

Markov compactum will not be minimal. 5. The theorem is valid for any countable group G of automorphisms for which the

Rokhlin--Halmos lemma holds.



Examples i. Let ~ be an ergodic automorphism with rational spectrum, i.e., with a spectrum PI, P2,". ; Pil~ ;

which is the union of finite subgroups of roots of unity of order P~ = ~ 9 Z

In this case the adic realization is the following. and the transformation T is given by r~

The compactum

= ~Z/~

= ~+i , addition being con~ ~ is Pl ~Z ' the

sidered in the additive group ~ matrices M~ = { ~ } , ~ j~---i

Since the Markov compactum If ~ = ~ , then

is the additive group of

integer ~ -adic numbers which explains the term "adic automorphism." 2. then ~ Let ~ =C~+~) ~ , where ~e[0,1) .If =L~[O,i) is irrational, Let ; for

is ergodic and its adic realization can be given in the following manner. be the expansion of o& into a continuous fraction; ~ = [~] ~ of order ~ x ~+i : ~'~ =i d~


~=.~_1~,%~=[~il=n~+~ for j~ ~ ; ~i~=0

we consider the matrices ~-~+i ~ =~ ;

~=~,~, ....

In other words,

The Markov compactum

~C ~

is constructed on ~




The adic shift in

it is isomorphic to the shift with measure (on ducing ~ *

; more precisely: there exists an isomorphism of the space

there is a unique measure, invariant relative to the adic shift) reIndeed, in this case the isomorphism is the Borel isomorphism ). This model is the modification of the

to an adic shift.

of the topological spaces (circumference and ~

constructions from [2, 13]; however, there, the constructions are used for other purposes. At the suggestion of the author, M. and L. Gandel'sman have established the isomorphism which has been mentioned above. Another adic realization of the rotation of the circumference is the folllowing: ~:iall =$o=~ ,



_ [= ~ ~ _ ~ I ]+it

~=i,~,.. ; M~ ~i~,~ =0 .

is a matrix of order i.e.,

M~=(~" ) ,




. . . . . .

. ' . . . o

. . . .


J. . . . . . . . . .

Thus, in each matrix


there is exactly one zero (see Sec. i).

~ (FC[~;],



the required compactum. 3. A shift on tori and so much more the automorphisms with continuous spectrum involve Here the Perron algorithm and the generalized expansions into a continuous It is interesting to pose inverse problems, i.e., to investigate the The Pascal graph,

more complications. fraction are useful.

ergodic properties of an addic shift in given Markov compacta and graphs.

the Young graph generate examples of shifts about which one does not know anything. LITERATURE CITED i. 2. A. M. Vershik, "Uniform algebraic approximation of the shift and multiplication operators," Dokl. Akad. Nauk SSSR, 259, No. 3, 526-529 (1981). M. Pimsner and D. Voiculescu, "Imbedding the irrational rotation C*-algebra into an AFalgebra," J. Operator Theory, 4, 201-210 (1980).

*Missing symbol in Russian original -- Publisher.


I0. ii. 12. 13.

A. M. Vershik, "Countable groups which are close to finite groups," in: F. P. Greenleaf, Invariant Means on Topological Groups [Russian translation], Mir, Moscow (1973). A. M. Vershik, "Is the uniform algebraic approximation of the multiplication and convolution operators possible?." J. Soy. Math., 26, No. 5 (1984). M. A. Rieffel, "Irrational rotation of C*-algebras," in: Internat. Congress Math., Helsinki (1978). A. M. Vershik, "Four definitions of the scale of an automorphism," Funkts. Anal. Prilozhen., ~, No. 3, 1-17 (1973). A. B. Katok, Ya. G. Sinai, and A. M. Stepin, "The theory of dynamical systems and general transformation groups with invariant measure," J. Soy. Math., 7, No. 6 (1977). A. M. Vershik, "Multivalued mappings with invariant measure (polymorphisms) and Markov operators," J. Sov. Math., 23, No. 3 (1983). A. M. Vershik and O. A. Ladyzhenskaya, "The evolution of measures determined by the Navier--Stokes equations and on the solvability of the Cauchy problem for the Hopf statistical equation," J. Soy. Math., i0, No. 2 (1978). E. G. Effros, The Dimension Group. Preprint. I. P. Kornfel'd, Ya. G. Sinai, and S. V. Fomin, Ergodic Theory [in Russian], Nauka, Moscow (1980). G. Hansel and J. P. Raoult, "Ergodicity, uniformity and unique ergodicity," Indiana Univ. Math. J., 23, No. 3, 221-237 (1973). E. G. Effrosand C. L. Shen, "Approximately finite C*-algebras and continued fractions," Preprint.




UDC 517.9

For second-order quasilinear degenerate elliptic equations, having the structure of ~ , ~ ) -elliptic equations in a bounded domain ~c~ , ~ , one estab-

lishes theorems of existence and uniqueness

for the generalized

solutions of the

first boundary-value problem, bounded together with their first order and also of first and second order. (A,O)

~ -derivatives of

The case of linear second-order

-elliptic equations are separately considered.

In the early sixties, O. A. Ladyzhenskaya and N. N. UraL'tseva have constructed theory of solvability of boundary-value elliptic and parabolic equations [i, 2].


problems for quasilinear second-order uniformly These results have formed the necessary foundaproblems for quasilinear

tion for the subsequent development of the theory of boundary-value elliptic and parabolic equations.

The overwhelming majority of the subsequent investigations

in this area is based to a certain extent on the mentioned results of Ladyzhenskaya and Ural'tseva. In this respect, the present paper is not an exception; we investigate the

question of the existence and uniqueness of the regular solutions of the first boundaryvalue problem for a class of quasilinear degenerate second-order elliptic equations. The

results obtained here are new even for the case of linear equations with a nonnegative characteristic form. Other results regarding the existence and the uniqueness of regular solu-

Translated from Zapiski Nauchnykh Seminarov Leningradskogo Otdeleniya Matematicheskogo Institutaim. V. A. Steklova AN SSSR, Vol. 115, pp. 83-96, 1982.


0090-4104/85/2805- 0674509.50

9 1985 Plenum Publishing Corporation