Section 6

Supplementary Problems

[7/14%]
Show
LaTeX
Problem 4.206

We assume that the sequence of complex numbers z1,z2,…,zk,…z_{1}, z_{2}, \ldots , z_{k}, \ldots has the following three properties:

  1. The series

    ∑k=1∞zk,∑k=1∞zk2,…,∑k=1∞zkn \sum _{k=1}^{\infty } z_{k}, \qquad \sum _{k=1}^{\infty } z_{k}^{2}, \qquad \ldots , \qquad \sum _{k=1}^{\infty } z_{k}^{n}

    are convergent.

  2. ∣z1∣≥∣z2∣≥⋯≥∣zk∣≥⋯\left|z_{1}\right| \geq \left|z_{2}\right| \geq \cdots \geq \left|z_{k}\right| \geq \cdots.

  3. There exist two real numbers α\alpha and β\beta such that

    0<β−α<2πnn+1,α≤arg⁡zk≤β,for k=1,2,3,…. 0 < \beta -\alpha < \frac{2\pi n}{n+1}, \qquad \alpha \leq \arg z_{k} \leq \beta , \qquad \text{for } k=1,2,3,\ldots .

Then ∣z1∣n+∣z2∣n+⋯+∣zk∣n+⋯\left|z_{1}\right|^{n}+\left|z_{2}\right|^{n}+\cdots +\left|z_{k}\right|^{n}+\cdots is convergent.

The conclusion holds under a wider assumption: The series listed in (1) need not be convergent; it is sufficient to assume that their partial sums are bounded. Yet the conclusion becomes invalid if we drop assumption (2) or if we require less than the first part of (3), namely that

0<β−α≤2πnn+1. 0 < \beta -\alpha \leq \frac{2\pi n}{n+1}.
?
Remark.
?

Cf. Problem 36 of Part III. G. Polya, problem: Jber. deutsch. Math. Verein. Section 2, 35, p. 48 (1926). Solution by L. Tchakalov, op. cit. 37, pp. 30-33 (1928).

Problem 4.207

The functions f1(z),f2(z),…,fn(z)f_{1}(z), f_{2}(z), \ldots , f_{n}(z); g1(z),g2(z),…,gn(z)g_{1}(z), g_{2}(z), \ldots , g_{n}(z) are regular analytic in a region of the zz-plane where

∣f1(z)∣2+∣f2(z)∣2+⋯+∣fn(z)∣2=∣g1(z)∣2+∣g2(z)∣2+⋯+∣gn(z)∣2. \left|f_{1}(z)\right|^{2}+\left|f_{2}(z)\right|^{2}+\cdots +\left|f_{n}(z)\right|^{2} = \left|g_{1}(z)\right|^{2}+\left|g_{2}(z)\right|^{2}+\cdots +\left|g_{n}(z)\right|^{2}.

If f1(z),f2(z),…,fn(z)f_{1}(z), f_{2}(z), \ldots , f_{n}(z) are linearly independent, then so also are g1(z),g2(z),…,gn(z)g_{1}(z), g_{2}(z), \ldots , g_{n}(z), and

gν(z)=cν1f1(z)+cν2f2(z)+⋯+cνnfn(z) g_{\nu }(z) = c_{\nu 1}f_{1}(z)+c_{\nu 2}f_{2}(z)+\cdots +c_{\nu n}f_{n}(z)

for ν=1,2,…,n\nu =1,2,\ldots ,n; the constants cνλc_{\nu \lambda } form a unitary matrix.

?
Remark.
?

R. Nevalinna and G. Polya, problem: Jber. deutsch. Math. Verein. Section 2, 40, p. 80 (1931). Solution by H. Schmidt, op. cit. 43, pp. 6-7 (1933).

Problem 4.208

We consider the approximation of the entire function g(z)g(z) by polynomials. We define the integer

k=k(r,α) k = k(r,\alpha )

(termed the degree of approximation of the function g(z)g(z) on the circle ∣z∣=r\left|z\right|=r to accuracy α\alpha) as the degree of the polynomial P(z)P(z) of the lowest possible degree for which

∫02π∣g(reiφ)−P(reiφ)∣2 dφ<α−2∫02π∣g(reiφ)∣2 dφ; \int _{0}^{2\pi } \left|g(re^{i\varphi })-P(re^{i\varphi })\right|^{2} \, d\varphi < \alpha ^{-2} \int _{0}^{2\pi } \left|g(re^{i\varphi })\right|^{2} \, d\varphi ;

(r>0r>0, α>1\alpha >1). Then

lim⁡α→∞klog⁡klog⁡α=lim⁡r→∞log⁡klog⁡r=ρ, \lim _{\alpha \to \infty } \frac{k\log k}{\log \alpha } = \lim _{r\to \infty } \frac{\log k}{\log r} = \rho ,

where we keep rr fixed when computing the first limit and α\alpha fixed when computing the second limit; ρ\rho is the order of g(z)g(z).

?
Remark.
?

Cf. Problem 123 of Part III. See problem 51 for the definition of order. G. Polya, problem: Jber. deutsch. Math. Verein., Section 2, 34, pp. 97-98 (1925). Solution by G. Szego, op. cit. 35, pp. 86-89 (1926).

Problem 4.209

Let g(z)g(z) be an entire function whose growth does not exceed the minimum type of order 11; that is, g(z)e−ε∣z∣g(z)e^{-\varepsilon \left|z\right|} remains bounded in the whole zz-plane provided that ε\varepsilon is a fixed positive number. Assume that the sequence

…,g(−n),…,g(−2),g(−1),g(0),g(1),g(2),…,g(n),… \ldots , \quad g(-n), \quad \ldots , \quad g(-2), \quad g(-1), \quad g(0), \quad g(1), \quad g(2), \quad \ldots , \quad g(n), \quad \ldots

is bounded. Then g(z)g(z) reduces to a constant.

?
Remark.
?

Cf. Problem 220(4), Problem 254 of Part III. G. Polya, problem: Jber. deutsch. Math. Verein., Section 2, 40, pp. 80-81 (1931). Solutions op. cit. 43 (1933) by L. Tchakalov, pp. 10-11 and G. Szego, pp. 11-13; comments by G. Polya, pp. 67-69. See also R. E. A. C. Paley and Norbert Wiener: Fourier transforms in the complex domain, Amer. Math. Soc. Colloquium Publ. 19, New York 1934, pp. 81-83.

Problem 4.210

We consider square matrices with nn rows and nn columns, and let capital letters denote such matrices. Let

g(z)=a0+a1z+a2z2+⋯+anzn+⋯ g(z) = a_{0}+a_{1}z+a_{2}z^{2}+\cdots +a_{n}z^{n}+\cdots

be an entire function and ZZ a matrix. Then the series

g(Z)=a0+a1Z+a2Z2+⋯+anZn+⋯ g(Z) = a_{0}+a_{1}Z+a_{2}Z^{2}+\cdots +a_{n}Z^{n}+\cdots

converges to a well-defined matrix.

Given the matrix AA, when does there exist a matrix ZZ such that

(∗)g(Z)=A  ? (*) \qquad g(Z) = A \; ?
?
Remark.
?

See Picard’s theorem, for example in Ahlfors pp. 297-298, Hille Vol. II pp. 219-221 or Titchmarsh pp. 277-284. Cf. problem 194. G. Polya, problem: Jber. deutsch. Math. Verein., Section 2, 40, p. 80 (1931). Solution by G. Szego, op. cit. 43, pp. 7-9 (1933).

Problem 4.211

Let f(z)f(z) be a single-valued analytic function which is regular everywhere including the point z=∞z=\infty, except at the point z=1z=1. In a neighborhood of the point z=0z=0 it has the expansion

f(z)=a0+a1z+a2z2+⋯+anzn+⋯ , f(z) = a_{0}+a_{1}z+a_{2}z^{2}+\cdots +a_{n}z^{n}+\cdots ,

and in a neighborhood of the point z=∞z=\infty the expansion

f(z)=b0+b1z+b2z2+⋯+bnzn+⋯ . f(z) = b_{0}+\frac{b_{1}}{z}+\frac{b_{2}}{z^{2}}+\cdots +\frac{b_{n}}{z^{n}}+\cdots .

If there exists a constant kk such that both ann−ka_{n}n^{-k} and bnn−kb_{n}n^{-k} are bounded (n=1,2,3,…n=1,2,3,\ldots), then f(z)f(z) is a rational function.

?
Remark.
?

G. Polya, problem: Jber. deutsch. Math. Verein. Section 2, 40, p. 81 (1931). Solutions op. cit. 43 by N. Obreschkov, pp. 13-15, G. Szego, pp. 15-16, and N. Tchebotarev, pp. 16-17.

Problem 4.212

Let the numbers a,b,m1,m2,m3,…a, b, m_{1}, m_{2}, m_{3}, \ldots be real, and satisfy

0<m1<m2<m3<⋯ , 0 < m_{1} < m_{2} < m_{3} < \cdots , lim⁡n→∞nmn>b−a2π>0. \lim _{n\to \infty } \frac{n}{m_{n}} > \frac{b-a}{2\pi } > 0.

Furthermore, let f(x)f(x) be continuous in the interval [a,b][a,b]. Then it follows from

∫abf(x)cos⁡mnx dx=∫abf(x)sin⁡mnx dx=0 \int _{a}^{b} f(x) \cos m_{n}x \, dx = \int _{a}^{b} f(x) \sin m_{n}x \, dx = 0

for n=1,2,3,…n=1,2,3,\ldots, that f(x)f(x) vanishes identically. (A fraction of the whole system of Fourier functions is closed over the corresponding fraction of the whole interval.)

?
Remark.
?

Cf. problem 198. G. Polya, problem: Jber. deutsch. Math. Verein., Section 2, 40, p. 81 (1931). Solution by O. Szasz, op. cit. 43, pp. 20-23 (1933). For a stronger result, see Paley and Wiener, op. cit. problem 209, pp. 83-85. See also Norman Levinson: Gap and Density Theorems. Amer. Math. Soc. Colloquium Publications 26, New York 1940, pp. 3-5.