Clever Geek Handbook
📜 ⬆️ ⬇️

Quasianalytic function

Quasianalytic functions in mathematical analysis are a class of functions that, loosely speaking, can be completely reconstructed from their values ​​in a small area (for example, on the boundary of a region). This property greatly facilitates the solution of differential equations and the study of other problems of analysis. Since this property holds for analytic functions (see Complex analysis ), the class of quasianalytic functions contains the class of ordinary analytic functions and can be considered as its extension [1] .

Content

Definitions

Single Variable Functions

One of the many defining features of an analytic function : let a functionf(x) {\ displaystyle f (x)}   infinitely differentiable at all points of the segment[a,b] {\ displaystyle [a, b]}   and let there be a numberA {\ displaystyle A}   (depending on the function) such that for all pointsx∈[a,b] {\ displaystyle x \ in [a, b]}   the inequality holds:

|f(n)(x)|⩽Ann!{\ displaystyle | f ^ {(n)} (x) | \ leqslant A ^ {n} n!}  (one)

Then the functionf(x) {\ displaystyle f (x)}   analytic (the converse theorem is also true) [2] .

Jacques Hadamard in 1912 proposed to generalize the above inequality, replacing the sequencen! {\ displaystyle n!}   general sequenceM={Mk}k=0∞ {\ displaystyle M = \ {M_ {k} \} _ {k = 0} ^ {\ infty}}   positive real numbers . He defined on the interval [ a , b ] the class of functions C M ([ a , b ]) as follows:

Every functionf {\ displaystyle f}   from the class is infinitely differentiable ( f ∈ C ∞ ([ a , b ])), and at all points x ∈ [ a , b ] and for allk⩾0 {\ displaystyle k \ geqslant 0}   the condition is satisfied:

|dkfdxk(x)|⩽Akk!Mk{\ displaystyle \ left | {\ frac {d ^ {k} f} {dx ^ {k}}} (x) \ right | \ leqslant A ^ {k} k! M_ {k}}  (2)

where A is some constant (depending on the function).

If we take the sequence M k = 1, then, according to what was said at the beginning of the section, we obtain exactly the class of ordinary real analytic functions on the interval [ a , b ].

A class C M ([ a , b ]) is called quasianalytic if for any function f ∈ C M ([ a , b ]) the uniqueness condition holds: ifdkfdxk(x)=0 {\ displaystyle {\ frac {d ^ {k} f} {dx ^ {k}}} (x) = 0}   at some point x ∈ [ a , b ] for all k , then f is identically equal to zero.

Elements of a quasianalytic class are called quasianalytic functions . The above condition means that two functions that coincide at some point along with all their derivatives coincide everywhere. In other words, the values ​​of the function on an arbitrarily small section completely determine all its values.

Functions of Several Variables

For functionf:Rn→R {\ displaystyle f: \ mathbb {R} ^ {n} \ to \ mathbb {R}}   and for a set of indicesj=(jone,j2,...,jn)∈Nn {\ displaystyle j = (j_ {1}, j_ {2}, \ ldots, j_ {n}) \ in \ mathbb {N} ^ {n}}   we denote:

|j|=jone+j2+...+jn{\ displaystyle | j | = j_ {1} + j_ {2} + \ ldots + j_ {n}}  
Dj=∂j∂xonejone∂x2j2...∂xnjn{\ displaystyle D ^ {j} = {\ frac {\ partial ^ {j}} {\ partial x_ {1} ^ {j_ {1}} \ partial x_ {2} ^ {j_ {2}} \ ldots \ partial x_ {n} ^ {j_ {n}}}}}  
j!=jone!j2!...jn!{\ displaystyle j! = j_ {1}! j_ {2}! \ ldots j_ {n}!}  
xj=xonejonex2j2...xnjn.{\ displaystyle x ^ {j} = x_ {1} ^ {j_ {1}} x_ {2} ^ {j_ {2}} \ ldots x_ {n} ^ {j_ {n}}.}  

Thenf {\ displaystyle f}   called quasianalytic in the open areaU⊂Rn, {\ displaystyle U \ subset \ mathbb {R} ^ {n},}   if for each compactK⊂U {\ displaystyle K \ subset U}   there is a constantA {\ displaystyle A}   such that:

|Djf(x)|⩽A|j|+onej!M|j|{\ displaystyle \ left | D ^ {j} f (x) \ right | \ leqslant A ^ {| j | +1} j! M_ {| j |}}  

for all indices from the setj∈Nn {\ displaystyle j \ in \ mathbb {N} ^ {n}}   and at all pointsx∈K {\ displaystyle x \ in K}   .

The class of quasianalytic functions ofn {\ displaystyle n}   variables with respect to the sequenceM {\ displaystyle M}   on the setU {\ displaystyle U}   can be designatedCnM(U) {\ displaystyle C_ {n} ^ {M} (U)}   , although other notations are also found in the sources.

Quasianalytic classes for log-convex sequences

Assume in the above definitionMone=one {\ displaystyle M_ {1} = 1}   and sequenceMk {\ displaystyle M_ {k}}   non-decreasing. This sequence is called logarithmically convex if the condition is satisfied:

SequenceMk+oneMk {\ displaystyle {M_ {k + 1} \ over M_ {k}}}   increasing.

If the sequenceMk {\ displaystyle M_ {k}}   logarithmically convex, then:

(Mk)one/k{\ displaystyle (M_ {k}) ^ {1 / k}}   also increasing.
MrMs⩽Mr+s{\ displaystyle M_ {r} M_ {s} \ leqslant M_ {r + s}}   for all(r,s)∈N2 {\ displaystyle (r, s) \ in \ mathbb {N} ^ {2}}   .

For logarithmically convexM {\ displaystyle M}   quasianalytic classCnM {\ displaystyle C_ {n} ^ {M}}   is a ring . In particular, it is closed with respect to multiplication and composition . The latter means:

If af=(fone,f2,...fp)∈(CnM)p {\ displaystyle f = (f_ {1}, f_ {2}, \ ldots f_ {p}) \ in (C_ {n} ^ {M}) ^ {p}}   andg∈CpM {\ displaystyle g \ in C_ {p} ^ {M}}   theng∘f∈CnM {\ displaystyle g \ circ f \ in C_ {n} ^ {M}}   .

Denjoy - Carleman theorem

The Denjoy-Carleman theorem was formulated and partially solved by Arno Denjoy ( Denjoy (1921 )) and fully proved in the work of Torsten Carleman ( Carleman (1926 )). This theorem provides a criterion for solving the question under which sequences M the functions C M ([ a , b ]) form a quasianalytic class.

According to the theorem, the following statements are equivalent:

  • C M ([ a , b ]) is a quasianalytic class.
  • ∑one/Lj=∞,{\ displaystyle \ sum 1 / L_ {j} = \ infty,}   WhereLj=infk≥j(k⋅Mkone/k) {\ displaystyle L_ {j} = \ inf _ {k \ geq j} (k \ cdot M_ {k} ^ {1 / k})}   .
  • ∑j(jMj∗)-one/j=∞,{\ displaystyle \ sum _ {j} (jM_ {j} ^ {*}) ^ {- 1 / j} = \ infty,}   where M j * is the largest logarithmically convex sequence bounded above by M j .
  • ∑jMj-one∗(j+one)Mj∗=∞.{\ displaystyle \ sum _ {j} {\ frac {M_ {j-1} ^ {*}} {(j + 1) M_ {j} ^ {*}}} = \ infty.}  

To prove that statements 3, 4 are equivalent to the 2nd, Carleman's inequality is used.

Example : Denjoy (1921 ) [3] indicated that ifMn {\ displaystyle M_ {n}}   given by one of the sequences

one,(ln⁡n)n,(ln⁡n)n(ln⁡ln⁡n)n,(ln⁡n)n(ln⁡ln⁡n)n(ln⁡ln⁡ln⁡n)n,...,{\ displaystyle 1, \, {(\ ln n)} ^ {n}, \, {(\ ln n)} ^ {n} \, {(\ ln \ ln n)} ^ {n}, \, {(\ ln n)} ^ {n} \, {(\ ln \ ln n)} ^ {n} \, {(\ ln \ ln \ ln n)} ^ {n}, \ dots,}  

then the corresponding quasianalytic class. The first sequence (of units) gives the usual analytical functions.

Additional properties

For a logarithmically convex sequenceM {\ displaystyle M}   The following properties of the corresponding class of functions hold.

  • CM{\ displaystyle C ^ {M}}   coincides with the class of analytic functions if and only ifsupj≥one(Mj)one/j<∞ {\ displaystyle \ sup _ {j \ geq 1} (M_ {j}) ^ {1 / j} <\ infty}   .
  • If aN {\ displaystyle N}   Is another logarithmically convex sequence in whichMj⩽CjNj {\ displaystyle M_ {j} \ leqslant C ^ {j} N_ {j}}   (hereC {\ displaystyle C}   Is some constant), thenCM⊂CN {\ displaystyle C ^ {M} \ subset C ^ {N}}   .
  • CM{\ displaystyle C ^ {M}}   stable with respect to differentiation if and only ifsupj≥one(Mj+one/Mj)one/j<∞ {\ displaystyle \ sup _ {j \ geq 1} (M_ {j + 1} / M_ {j}) ^ {1 / j} <\ infty}   .
  • For any unboundedly differentiable functionf {\ displaystyle f}   quasianalytic rings can be foundCM {\ displaystyle C ^ {M}}   andCN {\ displaystyle C ^ {N}}   and elementsg∈CM,h∈CN {\ displaystyle g \ in C ^ {M}, h \ in C ^ {N}}   such thatf=g+h {\ displaystyle f = g + h}   .

Weierstrass Division

Definition Functiong:Rn→R {\ displaystyle g: \ mathbb {R} ^ {n} \ to \ mathbb {R}}   called regular orderd {\ displaystyle d}   towardsxn {\ displaystyle x_ {n}}   , if ag(0,xn)=h(xn)xnd {\ displaystyle g (0, x_ {n}) = h (x_ {n}) x_ {n} ^ {d}}   andh(0)≠0 {\ displaystyle h (0) \ neq 0}   .

Let beg {\ displaystyle g}   - regular order functiond {\ displaystyle d}   towardsxn {\ displaystyle x_ {n}}   . They say that the ringAn {\ displaystyle A_ {n}}   real or complex functions ofn {\ displaystyle n}   variables satisfies the Weierstrass division with respect tog {\ displaystyle g}   if for eachf∈An {\ displaystyle f \ in A_ {n}}   existq∈A {\ displaystyle q \ in A}   andhone,h2,...,hd-one∈An-one {\ displaystyle h_ {1}, h_ {2}, \ ldots, h_ {d-1} \ in A_ {n-1}}   such that:

f=gq+h{\ displaystyle f = gq + h}   whereh(x′,xn)=∑j=0d-onehj(x′)xnj {\ displaystyle h (x ', x_ {n}) = \ sum _ {j = 0} ^ {d-1} h_ {j} (x') x_ {n} ^ {j}}   .

Example : the ring of analytic functions and the ring of formal power series both satisfy the Weierstrass division property. If, however,M {\ displaystyle M}   logarithmically convex andCM {\ displaystyle C ^ {M}}   does not coincide with the class of analytic functions, thenCM {\ displaystyle C ^ {M}}   does not satisfy the Weierstrass division property with respect tog(xone,x2,...,xn)=xone+x22 {\ displaystyle g (x_ {1}, x_ {2}, \ ldots, x_ {n}) = x_ {1} + x_ {2} ^ {2}}   .

History

The key question of this topic is the ability of an analytic function to uniquely restore its “global appearance” from the values ​​of the function itself and its derivatives at an arbitrary regular point [4] . Emil Borel was the first to discover that this property holds not only for analytic functions.

In 1912, Jacques Hadamard formulated the question: what should be the sequenceMn, {\ displaystyle M_ {n},}   so that the above “ uniqueness condition ” is satisfied for any pair of functions from the corresponding class. Arno Denjoy in 1921 gave sufficient conditions for quasi-analyticity and a number of examples of quasi-analytic classes (see Denjoy (1921 )). A complete solution to the problem was given five years later by Torsten Carleman (see Carleman (1926 )), who established the necessary and sufficient conditions for quasianalyticity [1] .

Subsequently, S. N. Bernshtein and S. Mandelbright generalized the concept of quasianalyticity to classes of non-differentiable and even discontinuous functions. The simplest example is a set of solutions of a linear differential equation with continuous coefficients; the functions included in this solution, generally speaking, do not have an infinite number of derivatives [5] ..

Notes

  1. ↑ 1 2 Mathematical Encyclopedia, 1979 , p. 798.
  2. ↑ Mandelbright, 1937 , p. 10-12.
  3. ↑ Leontiev, 2001 .
  4. ↑ Mandelbright, 1937 , p. 9-11.
  5. ↑ Mountain, 1938 , p. 171.

Literature

  • Gorny A. Quasi-analytic functions // Uspekhi Matematicheskikh Nauk . - M. , 1938. - No. 5 . - S. 171–186 .
  • Leontyev A.F. Quasianalytic class of functions // Mathematical Encyclopedia (in 5 volumes). - M .: Soviet Encyclopedia , 1979. - T. 2. - S. 798-800.
  • Mandelbright S. Quasianalytic classes of functions. - M. — L .: ONTI NKTP, 1937.
  • Mandelbrot S. Adjacent rows, regularization of sequences. Applications, per. from the French., M.: Foreign Literature, 1955.
  • Carleman, T. (1926), Les fonctions quasi-analytiques , Gauthier-Villars   (fr.)
  • Denjoy, A. (1921), "Sur les fonctions quasi-analytiques de variable réelle", CR Acad. Sci. Paris T. 173: 1329–1331  
  • Hörmander, Lars (1990), The Analysis of Linear Partial Differential Operators I , Springer-Verlag, ISBN 3-540-00662-1  
  • Leont'ev, AF Quasi-analytic class // Hazewinkel, Michiel . Encyclopedia of Mathematics. - Springer Science + Business Media BV / Kluwer Academic Publishers, 2001. - ISBN 978-1-55608-010-4 .

Links

  • Cohen, Paul J. (1968), " A simple proof of the Denjoy-Carleman theorem ", The American Mathematical Monthly (Mathematical Association of America). - T. 75 (1): 26–31, ISSN 0002-9890 , DOI 10.2307 / 2315100  
Source - https://ru.wikipedia.org/w/index.php?title= Quasianalytic_function&oldid = 97703554


More articles:

  • Titov, Gennady Fedorovich
  • Snitko, Andrey Vladimirovich
  • Pikus, Evgeny Mikhailovich
  • Knyshevich, Pelageya Maksimovna
  • Lebedik, Peter Vasilievich
  • Shvedka, Oldrich
  • Indirect Discrimination
  • Kharlamov, Stanislav Olegovich
  • Heral, Joseph
  • Shamray Evgeny Aleksandrovich

All articles

Clever Geek | 2019