Arithmetical Aspects of Power Series
[65/98%]The product of any consecutive integers is divisible by the product of the first positive integers.
The product of integers in arithmetic progression is divisible by if the common difference of the progression is prime to .
The product of the differences of any integers is divisible by the product of the differences of the first positive integers. (The product of the differences of variables , , consists of factors.)
Solution 96 of Part V.
For what integers is not divisible by ?
For what integers is not divisible by ?
What is the exponent of the highest power of a prime number that is a divisor of ?
In how many zeros does 1000! end when expressed in the decimal system?
Let and be positive integers. Then the integer is divisible by .
Cf. Problem 8.
If and are positive integers, then is an integer.
Let be an integer, a positive integer, . Show that only prime factors of divide the denominators of the coefficients of the power series with rational coefficients
Let be a prime factor of and let respectively be the highest powers of that divide , where denotes the denominator of the th coefficient in the power series expansion of ; . Calculate .
We term the function an algebraic function, if it satisfies an equation of the form where are polynomials, .
The proper perspective for evaluating the particular result contained in 138, 139 is given by the following general theorem due to Eisenstein: ``If the power series with rational coefficients represents an algebraic function of , then there exists an integer , such that has integral coefficients.'' [153.]
Find the smallest integer with the property that all the coefficients of the power series expansion are integers.
Prove Eisenstein's theorem in the simplest special case, namely for rational functions.
We say that the power series
with rational coefficients ( integers, , ) satisfies the Eisenstein condition, if there exists an integer such that is divisible by for . If two power series and with rational coefficients satisfy the Eisenstein condition, then the series , , , also satisfy the condition, and also (assuming ) and (assuming ).
If the two power series with rational coefficients
satisfy the Eisenstein condition, then the series
also satisfies the condition.
Show that the Eisenstein condition is equivalent to the following pair of simultaneous conditions:
The denominators contain only finitely many different prime numbers.
is bounded, .
Expand the two functions
in powers of . Do the power series obtained satisfy the two conditions (1), (2) in 144? Are the functions represented algebraic?
By a hypergeometric series we mean a series of the form
If and are rational numbers but not both integers, and if is a positive integer, then it is possible to choose a positive integer such that all the coefficients in the hypergeometric series become integers. (The function
for example is not an algebraic function [Problem 90 of Part I]: the Eisenstein condition is necessary but not sufficient for a function to be algebraic.)
Assume that , and are not all rational, but all the coefficients in the hypergeometric series are rational. Then and are the roots of an irreducible quadratic equation with rational coefficients.
In the case discussed in 147, does not represent an algebraic function of . Prove this from Eisenstein's theorem!
A rational function (quotient of two polynomials), whose expansion in ascending powers of has only rational numbers as coefficients, is the quotient of two polynomials with rational coefficients.
Let the power series with rational coefficients satisfy an equation of the form
where are polynomials, . Show that also satisfies an equation of the same form, where the polynomials occurring in it have coefficients that are rational integers.
Let the power series with rational coefficients satisfy an algebraic differential equation, i.e. an equation of the form
where denotes a rational entire function (multinomial) of its arguments. Show that must satisfy an equation of the form
where denotes a rational entire function with integral coefficients.
Let the functions , () be regular in a neighbourhood of the origin and related by the equation
Then the power series
if is suitably chosen, satisfies an equation of the form
where , , , and , , is a rational entire expression with integral coefficients in , possibly divided by a power of . (Function-theoretic preparation for 153, 154.)
If the power series with rational coefficients satisfy the Eisenstein condition, and if the power series with rational coefficients satisfies the equation
then also satisfies the Eisenstein condition.
Let and be integers, , , and denote the greatest prime factor that divides by . We say that the power series
satisfies the Tchebychev condition if is bounded, and the Hurwitz condition if is bounded, . For the Tchebychev condition as well as for the Hurwitz condition analogous theorems to 142, 143, and 153 hold.
Let be power series with rational coefficients, and set . If one of the following six equations holds
then is a uniquely determined rational number, given as a rational and entire function of the coefficients of in the expansions of , which therefore can have only such prime numbers in its denominator as occur also in the denominators of the coefficients mentioned above. [Cf. op. cit. 144.]
Let be power series with rational coefficients, and set . If one of the following six equations holds
then is a uniquely determined rational number, given as a rational and entire function of the coefficients of in the expansions of , which therefore can have only such prime numbers in its denominator as occur also in the denominators of the coefficients mentioned above. [Cf. op. cit. 144.]
We term the power series with integral coefficients primitive if the numbers have no common divisor apart from . Prove that the product of two primitive power series is a primitive power series.
If the power series with integral coefficients represents a rational function, then it can be expressed in the form , where and are polynomials with integral coefficients and . [155.]
Let
be a positive number expressed in the decimal system, , . The power series with integral coefficients represents a rational function if is rational, and does not represent a rational function if is irrational.
Let there be only finitely many distinct numbers in the infinite sequence of numbers . Then the power series represents a rational function if and only if the sequence of coefficients after a certain number of terms is periodic.
Let be an integer, , and a polynomial with integral coefficients. In the power series expansion of the coefficients are integers that form a periodic sequence modulo any prime number ; the length of the period is a power of .
Let be an integer and not divisible by the odd prime number . The coefficients in the power series expansion of are periodic mod , where the length of the period is a proper or an improper divisor of .
Let the integers and be as in 160. In the power series expansion of the coefficients are periodic mod . The length of the period is always a divisor of . It is or is not a divisor of according to or .
The sequence of Fibonacci numbers (cf. Problem 53 of Part VII) is periodic mod , where is any integer. Calculate the length of the period for all prime numbers less than 30.
In a power series expansion with integral coefficients that represents a rational function the coefficients, after a certain number of terms, form a periodic sequence mod ( arbitrary).
The algebraic function yields on expansion in powers of a power series with integral coefficients. These coefficients are for no odd prime number periodic mod . (163 cannot be extended to algebraic functions.)
The radius of convergence of a non-terminating power series with integral coefficients is .
For the background of 165--173 see G. Pólya: Math. Ann. 77, pp. 497--513 (1916), also Proc. Lond. Math. Soc., Ser. 2, 21, pp. 22--38 (1923), and F. Carlson: Math. Z. 9, pp. 1--13 (1921). For heuristic considerations see also MPR, 2, p. 46, ex. 14.
A non-terminating power series with integral coefficients that is convergent in the interior of the unit circle cannot represent a bounded function there.
If a power series with integral coefficients represents an algebraic function that is not a rational function, then its radius of convergence is .
In 167 the upper bound 1 is the best possible. In other words: If , then there exists a power series with integral coefficients that represents an algebraic function that is not a rational function and whose radius of convergence is .
Let the coefficients of the polynomial
be real. The analytic function defined by the power series
is
-
a rational function, if the numbers are all rational,
-
not a rational function, if the numbers are not all rational. (Cf. Problem 168 of Part II.) [Problem 85 of Part I.]
If the sequence of non-negative integers is bounded and contains infinitely many terms that are different from 0, then the series
does not represent a rational function.
If the sequence satisfies the same assumptions as in 170, then the series
does not represent a rational function.
Denote by the ``cross sum'' of the number , i.e. the sum of its digits. E.g. we have . The power series
has the unit circle as its natural boundary.
In an infinite table of logarithms let the logarithms of all the positive integers be written underneath each other in their natural order, in such a fashion that decimal places of the same order appear in the same vertical column. If we interpret the digits in any vertical column as the coefficients of a power series, then we obtain a power series that cannot be analytically continued.
In a different formulation we are required to prove the following theorem: Let be an integer and in the decimal representation of let the th digit after the decimal point be denoted by respectively. Then the analytic function defined by the power series
has as its natural boundary the circle .
If is a power series with H-integral coefficients, then
are also power series with H-integral coefficients.
If and are power series with H-integral coefficients, then
are also power series with H-integral coefficients, and under the further assumption that also
If is a power series with H-integral coefficients and , then
is also a power series with H-integral coefficients, [174].
If and are power series with H-integral coefficients and , then is also a power series with H-integral coefficients.
The function , well defined by the differential equation
and the initial conditions , , may be expanded in a power series with H-integral coefficients.
For every prime number , we have
For every composite number greater than 4 we have
The Bernoulli numbers are defined as the coefficients in the power series expansion
(Problem 154 of Part I). We have
In general we have
where is an integer and denote those prime numbers that exceed by one the divisors of . [Expand first and then in ascending powers of and apply 179--181.]
The coefficients in the series
where denotes the function defined in 178, are rational numbers and moreover the denominator of is not divisible by any square other than 1 and is divisible only by such prime numbers as are . [Expand first and then in ascending powers of .]
The differential equation
has a well-defined solution of the form
( is a particular elliptic function, termed a ``lemniscate function''; its period-parallelogram is a square.) The numbers are rational, and moreover the denominator of is not divisible by any square other than 1, and is divisible only by such prime numbers as are . [Show that , cf. 183. Expand first and then in ascending powers of ; as a function of satisfies a linear differential equation of second order.]
If a power series with H-integral coefficients satisfies a homogeneous linear differential equation with constant coefficients, then the sequence of coefficients is periodic mod after a certain number of terms, where is an arbitrary integer. [Examples of this are found already in 179, 180.]
The power series
shows that theorem 185 cannot without further assumptions be extended to homogeneous linear differential equations whose coefficients are polynomials.
If a transcendental entire function has a power series expansion with H-integral coefficients, then we have for the maximum modulus (Chapter 1 of Part IV) of
Let the power series in descending powers of
converge outside some circle. Let the numbers be rational. If the series assumes integral values for infinitely many integral values of , then is rational and
(Generalization of 93.) [Hurwitz--Courant, p. 30.]
The series
represents an integer for infinitely many integral values of .
Deduce 114 from 188.
Let the power series
that is not everywhere divergent represent integers for all sufficiently large integral values of . Then is an integral-valued polynomial.
If two polynomials are such that they assume integral values at the same points of the complex plane, then either their sum or their difference is a constant. (This constant is naturally an integer.)
If is the degree of and the degree of , then we have for sufficiently large
where all determinations are admissible for .