中国科学院数学与系统科学研究院期刊网

15 July 1956, Volume 6 Issue 3
    

  • Select all
    |
    论文
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 347-362. https://doi.org/10.12386/A1956sxxb0027
    Abstract ( )   Knowledge map   Save
    The following mean-value theorems for Z_(n,k)(s) are proved:Theorem A. Let an be not an integer and 0 < a < kν - ν, then where c_1=k~2(2π)~1/2(2πν)~(-2(k-na-1))(2Γ(1+ν))~(2k-2)× × Γ(2(1 - ν)(k - na - 1) + 1) ζ(2(k - na)) and ε is any positive number.Theorem B. Let an be positive but not an integer and let 2 (n - 1) (kν - ν-a) + + 1 > max (2a, 4) then where c_3 = 2~(-2(kν-ν)+1/2[2(n-1)(kν-ν-α)-2α+2]~(-1)π~(-2(k-na))+3/2× × k~2 ν~(2na)Γ~2 (ν) ζ(2(k - na)).
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 363-373. https://doi.org/10.12386/A1956sxxb0028
    Abstract ( )   Knowledge map   Save
    Consider the following type of differential equations d~2x/dt~2+P(dx/dt)+Q(x)=0, where a is a real number, yi~2 > 0 (i=1, 2, ..., n), and Q (x) assumes one of the following forms: where b is a non-vanishing real number. Consider equivalent system dx/dt =y, dy/dt=-P(y)-Q(x). Constract the characteristic equation We proved that separates the phase plane into regions such that each region contains one singularity, the behavior of which determines the distribution of trajectories in this region, and the structure of the trajectories on the whole plane are determined simply by putting these regions together.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 374-388. https://doi.org/10.12386/A1956sxxb0029
    Abstract ( )   Knowledge map   Save
    This note is a sequel to a previous one in which the infinitesimal deformation of a hypersurface in regular Cartan space (so called by L. Berwald) has been generalized so as to depend upon the covariant tangential vector p_m as well as the coordinates x~i of a point on the hypersurface. Besides the notation newly introduced I shall use the same notation.Let us consider that a hypersurface S in a regular Cartan space is subjected to by an infinitesimal deformation of the type x~i=x~i+ξ~i(x,p)δt,(1) where the ξ~i is a function of both x~k and p_m, and is homogeneous of degree zero in p_m, so that ξ~i‖~o = 0.(2)As in the previous paper the equation for the variation of the mean extremal curvature of a general hypersurface in the Cartan space is found to be δα_ρ~ρ/δt=E-Gα_ρ~ρ.(3)Suppose that the original hypersurface S which we have to deform into a near one is minimal,namely, H≡1/n-1α_ρ~ρ=0.(4) In order that the deformed hypersurface S be minimal also, it is necessary and sufficient that the ξ~i shall satisfy the associate differential equation E=0.(5)Now decomposing, as usual, the vector ξ~i into a linear combination of x_α~i and l~i: ξ~i= λ~α x_α~i + V l~i(6) and substituting these components into the expression E, we are finally led to the associate equation of the normal form of Berwald where U_o is Kosehmieder invariant of the hypersurface S.It is quite remarkable that the vanishing of the coefficient of the term λ~α in (5) is due to the generalized Mainardi-Codazzi equations of the hypersurface S: (n-1) H_((k)) = α_k~ρ(ρ) + R_o~ρ_(kρ)+α_ρ~σP_o~ρ_(σk)-α_k~σP_o~ρ_(σρ)+α_ρ~σB_k~ρ_σ(8) and (4). Thus the expression for U_o U_o=(A~μα_ μ~ρA~μB_σ~σ_ρ -A~μA~να_μ~ρα_(ρν)-(n-1)(n-2)K+R_o~ρ_(oρ)-α_ρ~σP_o~ρ_(oσ), as originally given by L. Berwald, is reestablished in quite a different way.(9)
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 389-404. https://doi.org/10.12386/A1956sxxb0030
    Abstract ( )   Knowledge map   Save
    Let X be a random variable, whose distribution law is F(x). Arranging the results of n independent trials on X, we obtain the corresponding variational series: ξ~1≤ξ_2≤…≤ξ_n.(1) In this paper the limit of the common distribution of two middle terms and that of a side term of constant rank and a middle term are investigated. Two general theorems are obtained:Theorem 1. Suppose that as n→∞, k_1/n→λ_1,k_2/n→λ_2, 0<λ_1<λ_2<1. In order that for given sequences of constants a_n, (> 0), b_n and c_n (> 0), d_n, we have on the continuity points of Φ (x, y), it is necessary and sufficent that where and The increasing functions u(x) and v(y) are uniquely determined from Φ (x,y) by the following relations:Theorem 2. Let k_1 < k_2, where k_1 is fixed and k_2/n→λ_2 (n→∞) and 0 < λ_2<1. In order that for given sequences of constants A_n (> 0), B_n and C_n (> 0), D_n, we have, as n →∞ on the continuity points of Ψ (x,y), it is necessary and sufficient that: U_n(x) =n F(A_n x+B_n) →U(x) (n→∞), V_n(y)=F(C_n y +D_n)-λ_(k_2)/τ_(k_2)→V(y) (n→∞), where λ_(k_2) and τ_(k_2) as in Theorem 1. The increasing functions U(x) and V(x) are uniquely determined from Ψ (x,y) by the following relations:
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 405-417. https://doi.org/10.12386/A1956sxxb0031
    Abstract ( )   Knowledge map   Save
    Let be a homogenous space with as its fundamental group. A system of linearly independent almost periodic functions Φ(P)={(P),…, (p)}in (in the sense of J. v. Neumann) is called a system of generalized spherical functions, if to every transformationσ∈, there corresponds a transformation (σP)= aii(σ) (P) (i=1, 2,…, n), where A (σ) = (a_(ii) (σ)) gives a normal representation of group E. Cartan and H. Weyl have both studied these functions in case the group is compact. The present author studied them again, keeping the group arbitrary, but restricting the functions in question to the almost periodic functions, in view of applying v. Neumann's theory of almost periodic representations of groups.The main result may be described as follows: take a unitary representation U(σ) from each of the classes of equivalent irreducible normal representations of To each U(σ), there corresponds h orthonormal systems of generalized spherical functions Φ_1(P)={…, },…, Φ_h(P) = {…,} (h≤n), where n is the degree of U(σ). The totality of all such functions thus obtained from a complete orthonormal system in Approximation theorem is also proved.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 418-425. https://doi.org/10.12386/A1956sxxb0032
    Abstract ( )   Knowledge map   Save
    We shall say that a series sum from n=1 to ∞ (u_n) is summable (R, 1) to s if the series converges for all values of h(≠0) in some neighbourhood of the origin, and It is known that (R, 1) is not regulad. However, Hardy and Littlewood proved the following theorem: Suppose that a_n=O(1/n),b_n=O(1/n),(1) then a necessary and sufficient condition that the series should converge, at a point x, to s, is that it should be summable (R,I) to s.Thus under the condition (1), the ordinary convergence and the method (R, 1) are equivalent. Let be the Fourier series of the function f(x), and We define the Riemann-Gibbs set of {R_h (x)}at x_o to be the aggregate of values λ = =lim R_h(x_o+α(h)), where α(h) tends to zero with h and α(h)/h→β(-∞≤β≤+∞).In the present note we establish the following two theorems:Theorem 1. For the series is the closed interval the Riemann-Gibbs set of{R_h (x)}at x=0Theorem 2. If series (2) is the Fourier series of a function f(x) of bounded variation, and if ξ is a point of jump of f(x), then the Riemann-Gibbs set of {R_h(x)}at point ξis the closed interval of length | f(ξ + 0) - f(ξ-0) | centred round 1/2{f(ξ+0) + f(ξ-0)}.In establishing Theorem 2, we need the following lemma.Lemma. If series (2)is the Fourier series of a function f(x) of bounded variation, and x_o is a point of continuity of f(x); then {R_h(x)}is uniformly convergent at x_o.For a series (2) satisfying condition (1), we need not to distinguish the ordinary convergence from (R, 1) summation, so far as the summability problem is concerned. But, when we consider the problem of Gibbs phenomenon, the situation are quite different, as the interval of Riemann and Gibbs is shortened in the case of (R, 1) summability.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 426-432. https://doi.org/10.12386/A1956sxxb0033
    Abstract ( )   Knowledge map   Save
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 433-451. https://doi.org/10.12386/A1956sxxb0034
    Abstract ( )   Knowledge map   Save
    The ordinary method of practical harmonic analysis is essentially an approximation of the following definite integrals by a fixed number (2n) of rectangles. As the function cos kx or sin kx oscillates between -1 and 1 2k times in the interval (0, 2π), it requires about 8k subintervals to get a reasonablly good approximation of the definite integral for a_k or b_k. It is obvious that as k increases, the number of subintervals has to be proportionally increased in order to insure the same accuracy. This explains why the ordinary method of practical harmonic analysis gives good approximations only for the first few coefficients.In this paper, the author approximates the curve of the function f(x) itself by a continuous curve consisting of 2n straight line-segments or n parabolic arcs. The Fourier coefficients of the approximating curve is then computed and used as approximations of the Fourier coefficients of the original function. Computation involved is almost as easy as the ordinary method. Approximation is found to be much closer than the ordinary method and more uniform for all k.The relation between approximations by the straight line-segment method A_k and B_k and those by the ordinary method α_k and β_k is found to be A_k/αk=B_k/β_k=sin~2 kπ/2n/(kπ/2n)~2.Two examples are given at the end of the paper.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 452-463. https://doi.org/10.12386/A1956sxxb0035
    Abstract ( )   Knowledge map   Save
    In the present note the author proves the following theorems:Theorem 1. Let R~n be a Riemannian manifold defined by a positive definite form gii(y)dy~idy~i, in which the functions gii(y) are assumed continuous in a coordinate neighborhood Let O denote the point (0, ..., 0). Then in order that the y's be normal coordinates at this point, it is necessary and sufficient that the identities gii(y)y~i = gii(O)y~i (i=1,…, n) hold true.For the particular case n=2, validity of this theorem was shown by Hartman and Wintner in 1952.Theorem 2. Let the metric of a Riemannian manifold be defined with a positive definite form ds~2=gii(y)dy~idy~i in which the functions gii(y) are continuous, and the y's are the normal coordinates at a point O(y=0). Then every hypersphere contained in the coordinates neighborhood and with 0 as center is intersected orthogonally by all the geodesic lines issued from the center.Theorem 3. Let the metric of a Riemannian manifold be defined with a positive definite form gii(x) dx~i dx~j in the neighborhood U of a point O. Suppose that the function gii(x) are continuous in U. Then the n functions y~1(x),…, y~n(x) of class C~1 will form a system of normal coordintes at O, if and only if the following conditions are satisfied:(i) y~1(x),…, y~n(x) are solutions of the differential equationsTheorem 4. Once we have obtained a complete solution of class C~3: of the differential equation subject to the conditions:(i)for the particular values x~i = x_o~i, c~i = c_o~i,(ii)everywhere in a certain delated neighborhood of the point O; Then a system of normal coordinates can be found withcut quadrature.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 464-471. https://doi.org/10.12386/A1956sxxb0036
    Abstract ( )   Knowledge map   Save
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 472-475. https://doi.org/10.12386/A1956sxxb0037
    Abstract ( )   Knowledge map   Save
    1. Let be the Fourier series of an L-integrable function f(θ). With the usual notation let Here taking the Borel's exponential definition, we have where K(x, t) = t~(-1)ν(x, t) sin(x sin t)(2) and ν(x, t) = e~(-x(1-cost)).(3) We shall be concerned with the behaviours at a point θ, which may suppose to be θ=0. Then (1) is the formula for finding the sum of the Fourier series of even function (t) at the point t=0 by Borel's method. As we know that the Borel's method does not sum all Fourier series at points of continuity. It is natural to inquire what happens when the function with an infinite limit. The result is as follows.Theorem. If (t) increases monotonically to +∞ as t→+0, then the Fourier series of (t) at point t=0 by Borel's method is summable to +∞ as x→+∞. 2. Proof of the Theorem. By (1), (2) and (3), we have Here we suppose that x is sufficiently large, and δ> 0 such that (t) is greater than any assignable number, however large when |t| <δ. Then we can show that So the theorem is proved
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 476-489. https://doi.org/10.12386/A1956sxxb0038
    Abstract ( )   Knowledge map   Save
    Denote by S the class of functions f(z) a_ν z~ν regular and univalent in the unit circle |z|<1, by S the subclass of S each function w=f(z) of which maps |z|<1 on a domain D_f starshaped with respect to w=0, and finally by S the subclass of S,whose function f(z) are characterized by the condition The class K is the subclass of S such that the image D_f obtained by w=f(z) of K is convexWe write σ_n(z)=z+a_2z~2+…+a_nz~n, the sections of the expansion f(z)Rachmanof proves that, if f(z) ∈ K, then all the sections σ_n (z) except σ_4 (z), are Univalent in the circle |z| <1/2. In the present paper, we settle Rachmanof's problem on σ_4(z). Indeed, what is more, we can prove the followingTheorem A. If f(z) a_νz~ν∈S,then 2σ_n(1/2 z)∈S,n=1,2,3…, and 2σ_n(1/2 z)∈S, n=1, 2, , , , 6, …. Moreover, 3σ_n(1/3 z)∈S, provided that n≠3. Indeed, 4σ_n (1/4 z)∈K; furthermore, if f(z) ∈ S, then 4σ_n(1/4 z)∈S, n 1, 2, 3, …The factors 1/2,1/3,1/4,in the respective cases are not allowed to be increased.The proof is based chiefly on the principle of subordination. In fact, we can establishTheorem B. If f(z)∈S, then and there exists an increasing function a(θ) with a(θ) = a(θ+0) (0≤θ<2π), a(2π) - a(0) = 2π, such that and that This last formula of representation implies f(z)∈S.Hence we can deduce the following precise estimates): z = re~(iθ), 0 ≤ r < 1, All the signs of equality in these relations hold good when and only when f(z)≡f(z)=z/1-ηz,|η| =1.Furthermore, by a famous theorem of C. Caratheodory, we can prove the followingTheorem C. Suppose f(z) a_ν z~ν ∈S. For any integer m ≥2, consider (a_2, a_3,…, a_m) as a point P_m in the 2_(m-2) dimensional real Euclidian space. Then P_m falls into the smallest convex body V_m, containing the closed curve Γ_m, which is defined parametrically by (e~(-iθ), e~(-i2θ),…, e~(-i(m-1)θ)) ,m ≥ 2, 0 ≤ θ ≤ 2π. The case P_m∈Γ_m realizes if and only if f(z)≡f(z)=z(1-zη)~(-1), |η| = 1.Finally, we give some relations ,between a_2 and a_3.Theorem D. if f(z)=z+a_2 z~2+a_3 z~3+…∈S, then the equality holds only for the function f(z)=z[1-ze~(-i arc cos(±1/4))]~(-1); and |a_3-1/2 a_2~2|≤1/2, the equality holds only for the function f(z)=z(1-ηz)~(-λ) (1+ηz)~(-(1-λ)),|η| =1, 0≤λ≤1.
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 490-499. https://doi.org/10.12386/A1956sxxb0039
    Abstract ( )   Knowledge map   Save
    Let the function f(z)=c_n z~n be regular and schlicht in the unit circle |z|<1. The main theorem in this note runs that if n greater than an absolute number N, then |c_n|
  • Acta Mathematica Sinica, Chinese Series. 1956, 6(3): 500-513. https://doi.org/10.12386/A1956sxxb0040
    Abstract ( )   Knowledge map   Save
    Professor L. K. Hua has pointed out to me that we may improve result on Goldbach problem by the combination of the methods of Selbergand Brun-The purpose of this paper is to carry out this idea and to prove the following two results:Theorem 1. Every sufficiently large even integer is a sum of a product of at most 3 primes and a product of at most 4 primes.Theorem 2. There are infinitely many integers n, such that n is a product of at most 3 primes and n+2 is a product of at most 4 primes.It can be expected that by the present method with some complicated numerical calculations we can prove that every sufficiently large even integer is a sum of two products each of which has at most 3 primes.In succeeding papers, the author will give the proofs of the following results.1°. If F(x) denotes a irreducible integral valued polynomial of degree k without any fixed prime divisor, then there are infinitely many integers x, such that F(x) is a product of at most [2.1 k] primes.2°. Let π(N; F(x)) be the number of primes represented by F(x) as x=1, 2,…,N. Then we have π(N; F(x))≤2e~γ μF N/log N + o(N/log N ), where μF is a constant depending on F(x) only and γ denotes the Euler's constant.Assuming the truth of grand Riemann hypothesis, that is, assuming the real parts of all zeros of all Diricblet's L-functions L(s, X) are ≤1/2, we have the following.3°. Every sufficiently large even integer is a sum of a prime and a product of at most 4 primes.4°. Let Z_2 (N) be the number of twin primes (p, p+2)not exceeding N. Then where ε is any given positive number and the constant implied by the symbol "O" depends on εonly.