Loading [MathJax]/jax/output/SVG/jax.js
Research article Special Issues

Dynamics of a nonlinear differential advertising model with single parameter sales promotion strategy

  • Received: 07 October 2021 Revised: 19 November 2021 Accepted: 28 November 2021 Published: 14 March 2022
  • Advertising and sales promotion are two important specific marketing communications tools. In this paper, nonlinear differential equation and single parameter sales promotion strategy are introduced into an advertising model and investigated quantitatively. The existence and stability of period-nT (n = 1, 2, 4, 8) solutions are investigated. Interestingly, both period doubling bifurcation and inverse flip bifurcation occur at different parameter values in the same advertising model. The results show that the system enters into chaos from stable state through flip bifurcation and enters into stable state from chaos through inverse flip bifurcation. An effective control strategy, which suppresses flip bifurcation and promotes inverse flip bifurcation, is proposed to eliminate chaos. These results have some significant theoretical and practical value in related markets.

    Citation: Junhai Ma, Hui Jiang. Dynamics of a nonlinear differential advertising model with single parameter sales promotion strategy[J]. Electronic Research Archive, 2022, 30(4): 1142-1157. doi: 10.3934/era.2022061

    Related Papers:

    [1] Kwang-Wu Chen, Fu-Yao Yang . Infinite series involving harmonic numbers and reciprocal of binomial coefficients. AIMS Mathematics, 2024, 9(7): 16885-16900. doi: 10.3934/math.2024820
    [2] Zhi-Hong Sun . Supercongruences involving Apéry-like numbers and binomial coefficients. AIMS Mathematics, 2022, 7(2): 2729-2781. doi: 10.3934/math.2022153
    [3] Shelby Kilmer, Songfeng Zheng . A generalized alternating harmonic series. AIMS Mathematics, 2021, 6(12): 13480-13487. doi: 10.3934/math.2021781
    [4] Wenbin Zhang, Yong Zhang, Jiachen Wu . A Ramanujan-type supercongruence related to harmonic numbers. AIMS Mathematics, 2025, 10(5): 11444-11464. doi: 10.3934/math.2025521
    [5] Xuqiong Luo . A D-N alternating algorithm for exterior 3-D problem with ellipsoidal artificial boundary. AIMS Mathematics, 2022, 7(1): 455-466. doi: 10.3934/math.2022029
    [6] Jiaye Lin . Evaluations of some Euler-type series via powers of the arcsin function. AIMS Mathematics, 2025, 10(4): 8116-8130. doi: 10.3934/math.2025372
    [7] Ling Zhu . Asymptotic expansion of a finite sum involving harmonic numbers. AIMS Mathematics, 2021, 6(3): 2756-2763. doi: 10.3934/math.2021168
    [8] Andrea Ratto . Higher order energy functionals and the Chen-Maeta conjecture. AIMS Mathematics, 2020, 5(2): 1089-1104. doi: 10.3934/math.2020076
    [9] Aleksa Srdanov . Invariants in partition classes. AIMS Mathematics, 2020, 5(6): 6233-6243. doi: 10.3934/math.2020401
    [10] Kholood M. Alsager, Sheza M. El-Deeb, Ala Amourah, Jongsuk Ro . Some results for the family of holomorphic functions associated with the Babalola operator and combination binomial series. AIMS Mathematics, 2024, 9(10): 29370-29385. doi: 10.3934/math.20241423
  • Advertising and sales promotion are two important specific marketing communications tools. In this paper, nonlinear differential equation and single parameter sales promotion strategy are introduced into an advertising model and investigated quantitatively. The existence and stability of period-nT (n = 1, 2, 4, 8) solutions are investigated. Interestingly, both period doubling bifurcation and inverse flip bifurcation occur at different parameter values in the same advertising model. The results show that the system enters into chaos from stable state through flip bifurcation and enters into stable state from chaos through inverse flip bifurcation. An effective control strategy, which suppresses flip bifurcation and promotes inverse flip bifurcation, is proposed to eliminate chaos. These results have some significant theoretical and practical value in related markets.



    Let N be the set of natural numbers with N0=N{0}. For nN0, the following harmonic-like numbers are given by partial sums:

    ● The classical harmonic number and its quadratic counterpart

    Hn=n1k=01k+1andH2n=n1k=01(k+1)2.

    ● Skew harmonic number and its quadratic counterpart

    On=n1k=012k+1andO2n=n1k=01(2k+1)2.

    ● Alternating harmonic number and its quadratic counterpart

    ˉHn=n1k=0(1)kk+1andˉH2n=n1k=0(1)k(k+1)2.

    Among them, there are the following simple but useful relations:

    H2n=On+Hn2,H22n=O2n+H2n4;ˉH2n=OnHn2,ˉH22n=O2nH2n4.

    Harmonic numbers and their variants have been studied since the distant past and are involved in diverse fields (cf. [1,5,6,7,11,17,18]) such as analysis of algorithms in computer science, various topics of number theory, and combinatorial analysis. In this paper, we aim at presenting several algebraic identities about infinite series involving harmonic-like numbers, the binomial coefficient (3nn) in numerators, and a free variable "y".

    For nN0 and an indeterminate x, the shifted factorials (cf. [3, §1.1])) are usually defined by

    (x)0=1and(x)n=x(x+1)(x+n1)fornN.

    Denote by [xm]ϕ(x) the coefficient of xm in the formal power series ϕ(x). Then we can express harmonic-like numbers (see [11,18]) by extracting the initial coefficients of x from the quotients of shifted factorials as below:

    [x](1+x)nn!=Hn,[x2](1+x)nn!=H2nH2n2,[x]n!(1x)n=Hn,[x2]n!(1x)n=H2n+H2n2,[y](12+y)n(12)n=2On,[y2](12+y)n(12)n=2(O2nO2n),[y](12)n(12y)n=2On,[y2](12)n(12y)n=2(O2n+O2n).

    There exist many infinite series identities involving central binomial coefficients (see [12,14,19,25]). However, the closed form evaluations for the known series with (3nn) are rare. In a recent paper, Sun [23] reported several remarkable infinite series identities, detected by numerical experimentation. Among them, the following three (see [23, Eqs (2.12), (2.13), and (2.15)]) seem rather challenging:

    n=1(227)n(3nn)H2n=1+32{3ln23ln(33)}, (1.1)
    n=1(227)n(3nn)H3n=1+34ln(4(7+43)3)3ln2, (1.2)
    n=1(3+554)n(3nn){H3nH2n}=1+52ln(9352). (1.3)

    When making efforts to confirm these three formulae, two substantial novelties emerged. First, not only the harmonic numbers H3n and H2n are separable in the third series, but also independent series involving Hn and On can be evaluated explicitly in closed form (cf. Corollaries 6 and 10). Another surprising novelty (see Theorems 1 and 5) lies in the fact that these series are particular cases of more general algebraic identities containing a free variable "y", corresponding to the convergent rates of these numerical series. To our knowledge, most of the infinite series identities presented in this paper have not appeared previously in the literature.

    The rest of the paper is organized as follows. As preliminaries, we shall evaluate, in closed form, three series containing binomial coefficient (3nn), harmonic number Hn and a free variable "y" by integrating Lambert's series. Then in Sections 3–5, three classes of infinite series with harmonic-like numbers will be examined by applying the "coefficient extraction" method (see [8,13,24]) to cubic transformations for the 3F2-series, that result in several algebraic identities about these series and confirm or refine, as special numerical examples, three conjectured identities made recently by Sun [23]. Finally, the paper will end with Section 6, where concluding remarks and further problems are provided.

    In classical analysis and enumerative combinatorics, Lambert's binomial series are well-known (see Riordan [22, §4.5] and [9,10,16]), which can be reproduced as follows: Let T and τ be the two variables related by the equation τ=T/(1+T)β. Then

    (1+T)α=k=0αα+kβ(α+kβk)τkand(1+T)α+11+TTβ=k=0(α+kβk)τk. (2.1)

    As preliminaries, we are going to evaluate, in closed form, three series containing ternary coefficient (3nn) and a free variable "y" by integrating across the above two equations, that will be crucial for demonstrating infinite series identities involving harmonic numbers in the subsequent sections.

    For "α=0 and β=3", performing the replacement τxy/(1+y)3 in Lambert's second series, we can state the resulting equality as

    ϕ(T):=1+T12T=n=0(3nn)(xy)n(1+y)3n:x=T(1+y)3y(1+T)3. (2.2)

    Since the rightmost function is monotone around T=0, it determines an implicit inverse function T=T(x) in the neighborhood of x=0 with particular values "T(0)=0 and T(1)=y".

    By expressing the harmonic number Hn in terms of the integral

    Hn=101xn1xdx,

    we can insert Hn in the series (2.2) as follows:

    n=0(3nn)Hnyn(1+y)3n=10ϕ(T(1))ϕ(T(x))1xdx=(1+y)3yy012T(1+T)41+y12y1+T12T1T(1+y)3y(1+T)3dT=(1+y)3y01(12T)(1+y)(1+T)(12y)T(1+y)3y(1+T)3dT,

    where the change of variable x=T(1+y)3y(1+T)3 has been made so that the integral becomes explicit.

    For the sake of brevity, the notation "ρ" will be fixed, throughout the paper, for the algebraic number

    ρ:=32(2+1)1/332(21)1/310.105892543. (2.3)

    The convergence region of the preceding series is determined by

    1<27y4(1+y)3<1,ρ<y<12.

    This can be illustrated graphically as below:

    Figure 1.  Range of convergence.

    By decomposing the integrand into partial fractions

    1(12T)(1+y)(1+T)(12y)T(1+y)3y(1+T)3=3(1+T)(1+y)2(12y)32(1+y)2(12y)4+y×{4+y+yT+y(3+y)(1+y)4+y2y+4+yyT+y(3+y)+(1+y)4+y2y},

    we can explicitly compute the integral

    y01(12T)(1+y)(1+T)(12y)T(1+y)3y(1+T)3dT=3ln(1+y)(1+y)2(12y)+3(y4+y)2(1+y)2(12y)4+yln((1+y)(22yy2+yy2+4y)2)3(y+4+y)2(1+y)2(12y)4+yln((1+y)(22yy2yy2+4y)2).

    By cancelling the three terms containing ln(1+y) and then making substitution, we find the closed formula for the series as in the theorem below:

    Theorem 1 (ρ<y<1/2).

    n=1(3nn)Hnyn(1+y)3n=(1+y)3y01(12T)(1+y)(1+T)(12y)T(1+y)3y(1+T)3dT=3(1+y)ln(12y)2(2y1)+3(1+y)y2(12y)4+yln(2y(2+yy2+4y)2y(2+y+y2+4y)).

    For "α=1 and β=3", performing the replacement τxy/(1+y)3, we can rewrite Lambert's first series as

    ψ(T):=1+T=n=0(3nn)(xy)n(1+2n)(1+y)3n:x=T(1+y)3y(1+T)3. (2.4)

    Analogously, by putting the harmonic number Hn inside the series (2.4), we have

    n=0(3nn)Hnyn(1+2n)(1+y)3n=10ψ(T(1))ψ(T(x))1xdxx=T(1+y)3y(1+T)3=(1+y)3yy0(yT)(12T)dT(1+T)4{1T(1+y)3y(1+T)3}=y0(1+y)3(Ty)(12T)dT(1+T){T(1+y)3y(1+T)3}=y0(1+y)3(12T)dT(1+T)(1T2yTy23Ty).

    By decomposing the integrand into partial fractions

    (1+y)3(12T)dT(1+T)(1T2yTy23Ty)=31+y1+T(1+y)(3y+4+y)2y(T+(3+y)y+(1+y)4+y2y)(1+y)(3y4+y)2y(T+(3+y)y(1+y)4+y2y),

    we can explicitly compute the integral

    y0(1+y)3(12T)dT(1+T)(1T2yTy23Ty)=3(1+y)ln(1+y)(1+y)(3y+4+y)2yln((1+y)(3y+4+y)(3+y)y+(1+y)4+y)(1+y)(3y4+y)2yln((1+y)(3y4+y)(3+y)y(1+y)4+y),

    which is simplified into the closed formula as in the theorem below:

    Theorem 2 (ρ<y<1/2).

    n=1(3nn)Hnyn(1+2n)(1+y)3n=y0(1+y)3(12T)dT(1+T)(1T2yTy23Ty)=(1+y)4+y2yln(2y(2+y+y2+4y)2y(2+yy2+4y))32(1+y)ln(12y).

    Alternatively, rewriting Lambert's first series

    (1+T)α1α=n=11α+nβ(α+nβn)τn:τ=T/(1+T)β

    and letting "α0 and β=3", we deduce the identity

    σ(T):=3ln(1+T)=n=1(3nn)(xy)nn(1+y)3n:x=T(1+y)3y(1+T)3. (2.5)

    By inserting Hn in the above series, we obtain an integral expression

    n=1(3nn)Hn(xy)nn(1+y)3n=10σ(T(1))σ(T(x))1xdxxT(1+y)3y(1+T)3=3y0ln(1+T1+y)(12T)(1+y)3(1+T)(Ty)(1T2yTy23Ty)dT=3y0ln(1+T1+y){1Ty31+T+y2+3y+2TyT2y+Ty2+3Ty1}dT.

    In view of the partial fraction decomposition

    y2+3y+2TyT2y+Ty2+3Ty1=1T+y(3+y)+(1+y)4+y2y)+1T+y(3+y)(1+y)4+y2y),

    it suffices to evaluate, for a parameter λ subject to T+λ0 for T(0,y), the parametric integral

    J(λ,y)=y0ln(1+T1+y)dTT+λ=Li2(1+y1λ)Li2(11λ)+ln(λλ1)ln(1+y),

    with particular cases

    J(1,y)=12ln2(1+y)andJ(y,y)=π26+ln(1+y)ln(y1+y)Li2(11+y).

    Hereafter, the dilogarithm function (see Lewin [20, §1.1]) is defined for all complex values of z by

    Li2(z)=z0ln(1τ)τandLi2(z)=n=1znn2for|z|1.

    By making substitution and then simplifying the terms concerning logarithms and dilogarithms in the resulting expression, we find the explicit formula as in the following theorem:

    Theorem 3 (ρ<y<1/2).

    n=1(3nn)Hnynn(1+y)3n=y03dT1+Tln(1T(1+y)3y(1+T)3)=π22+3Θ(y)+3ln(1+y)ln(y(1+y)3),

    where, for simplicity, Θ(y) denotes the sum of five dilogarithmic values

    Θ(y)=Li2(yy2+4y2)+Li2(y+y2+4y2)Li2(yy2+4y2+2y)Li2(y+y2+4y2+2y)Li2(11+y).

    According to Bailey [3], the classical hypergeometric series reads as

    1+pFp[a0,a1,a2,,apb1,b2,,bp|z]=k=0zkk!(a0)k(a1)k(a2)k(ap)k(b1)k(b2)k(bp)k.

    We shall confine ourselves only to examining the series with |z|<1 and none of the parameters in the numerator and denominator being a non-positive integer, which imply that the corresponding series is well-defined and convergent.

    There are many transformation and summation formulae for hypergeometric series (see Bailey [3] and Brychkov [4, Chapter 8]) in the literature. In 1928, Bailey [2] discovered an important cubic transformation (see the next section). Half a century later, Gessel and Stanton [15, Eq (5.4)] found the following counterpart

    3F2[1+a3,2+a3,3+a31+a+c2,2+ac2|27y4(1+y)3]=(1+y)1+a12y×4F3[1+a3,a,c,1ca3,1+a+c2,2+ac2|y4]. (3.1)

    Under the parameter replacements "a6ax,c2cx", this transformation can be reformulated as

    n=0(y4(1+y)3)n(1+6ax)3nn!(1+3axcx)n(12+3ax+cx)n=(1+y)1+6ax12y{1+6cxk=1(y4)k(k+2ax)(1+6ax)k1(1+2cx)k1(12cx)kk!(1+3axcx)k(12+3ax+cx)k}, (3.2)

    where "y(ρ,12)", or equivalently, 27y4(1+y)3(1,1) so that the series are convergent.

    The above transformation is an analytic equality with respect to x in the neighborhood of x=0. Equating the constant terms on both sides yields the algebraic identity below:

    n=0(3nn)yn(1+y)3n=1+y12y.

    As illustrated in the introduction, harmonic-like numbers can be expressed in terms of x-coefficients from quotients of shifted factorials. By examining the coefficients of x across the above equation, we find that the corresponding equation involving harmonic numbers can be written as

    n=1yn(1+y)3n(3nn){6aH3n(3ac)Hn2(3a+c)On}=6a1+y12yln(1+y)+6c1+y12yk=1(k1)!(12)k(y4)k.

    The rightmost series admits the closed form expression, as in the lemma below:

    Lemma 4 (4<y<4).

    W(y)=k=1(k1)!(12)k(y4)k=4y4+yArcTanhy4+y=4y4+yln4+yy2.

    Proof. Recall the beta function (see [21, §16])

    B(a,b)=10xa1(1x)b1dx=Γ(a)Γ(b)Γ(a+b),where(a)>0and(b)>0.

    We can express W(y) as a definite integral

    W(y)=k=1(k1)!(12)k(y4)k=k=1B(k,12)(y4)k=k=1(y4)k10xk1dx1x=10dxx1xk=1(xy4)k=10ydx1x(4+xy)=102y4+yT2ydT,

    where the last line is justified by the change of variable 1xT. Then the formula in the lemma follows by evaluating the last integral explicitly.

    Therefore, we find the general formula below:

    n=1yn(1+y)3n(3nn){6aH3n(3ac)Hn2(3a+c)On}=6a1+y12yln(1+y)+6c1+y12yW(y).

    By specifying two parameters, a and c, we deduce four independent series

    a=0:n=1yn(1+y)3n(3nn)ˉH2n=31+y12yW(y),c=0:n=1yn(1+y)3n(3nn){H3nH2n}=1+y12yln(1+y),c=3a:n=1yn(1+y)3n(3nn){H3n2On}=1+y12yln(1+y)31+y12yW(y),c=3a:n=1yn(1+y)3n(3nn){H3nHn}=1+y12yln(1+y)+31+y12yW(y).

    Combining the above formulae with Theorem 1 and then keeping in mind the equalities

    H2n=ˉH2n+Hn,On=ˉH2n+Hn2,H3n=(H3nHn)+Hn,

    we establish further closed formulae for four series with a free variable "y".

    Theorem 5 (ρ<y<1/2).

    (a)n=1(3nn)ˉH2nyn(1+y)3n=3+3y12y4y4+yln4+y+y2,(b)n=1(3nn)H2nyn(1+y)3n=3+3y24yy4+yln(2(1+y+y2+4y)2y(2+y+y2+4y))(3+3y)ln(12y)2(12y),(c)n=1(3nn)Onyn(1+y)3n=3+3y24yy4+yln(1+y+y2+4y12y)(3+3y)ln(12y)4(12y),(d)n=1(3nn)H3nyn(1+y)3n=1+y24yln((1+y)2(12y)3)+3+3y24yy4+yln(2(1+y+y2+4y)2y(2+y+y2+4y)).

    Numerous infinite series identities with different convergence rates can be derived by assigning particular values for y in Theorems 1 and 5. Five groups of representative examples are recorded as corollaries. In particular, the second formula (b) and the last formula (e) displayed in Corollary 6 confirm (1.1) and (1.2) conjectured by Sun [23, Eqs (2.12) and (2.13)].

    Corollary 6 (y3352 in Theorems 1 and 5).

    (a)n=1(227)n(3nn)ˉH2n=332ln(31),(b)n=1(227)n(3nn)H2n=1+32{3ln23ln(33)},(c)n=1(227)n(3nn)Hn=334ln(43)+3ln(2+33),(d)n=1(227)n(3nn)On=3+32{ln(1+32)34ln3},(e)n=1(227)n(3nn)H3n=1+34ln(4(7+43)3)3ln2.

    Corollary 7 (y52 in Theorems 1 and 5).

    (a)n=1(18)n(3nn)ˉH2n=3(15)5ln(512),(b)n=1(18)n(3nn)H2n=3(2+5)5ln(3+52)3ln55,(c)n=1(18)n(3nn)Hn=32ln(3+52)+35ln(11+5510),(d)n=1(18)n(3nn)On=3(1+5)25ln(2+5)3ln525,(e)n=1(18)n(3nn)H3n=ln(11+55)+35ln(4+255).

    Corollary 8 (y525 in Theorems 1 and 5).

    (a)n=1(564)n(3nn)ˉH2n=6(51)11ln(552),(b)n=1(564)n(3nn)H2n=6511ln(25+9522)311ln(233510),(c)n=1(564)n(3nn)Hn=3511ln(26711952)611ln(7+510),(d)n=1(564)n(3nn)On=3(1+5)11ln(4+511)+3511ln5,(e)n=1(564)n(3nn)H3n=6511ln(4(4+5)11)+111ln(200(2667735)116).

    Corollary 9 (y110 in Theorems 1 and 5).

    (a)n=1(100729)n(3nn)ˉH2n=38313ArcCos(161200),(b)n=1(100729)n(3nn)H2n=98ln(65)38313ArcCos(6780),(c)n=1(100729)n(3nn)Hn=98ln(65)+38313ArcCos(79878000),(d)n=1(100729)n(3nn)On=916ln(65)316313ArcCos(720),(e)n=1(100729)n(3nn)H3n=38ln(3215)38313ArcCos(6780).

    Corollary 10 (y7352 in Theorems 1 and 5).

    (a)n=1(3+554)n(3nn)ˉH2n=6565ln(1+15+653+5),(b)n=1(3+554)n(3nn)H2n=3(1+5)4ln(2+53)+3/25+5ln(351+30654),(c)n=1(3+554)n(3nn)Hn=3(1+5)4ln(2+53)+3/25+5ln(113+185825113185825),(d)n=1(3+554)n(3nn)On=3(1+5)8ln(2+53)+3/85+5ln(354+21565),(e)n=1(3+554)n(3nn)H3n=1+54ln(11+556)+3/25+5ln(351+30654).

    We remark that the second formula (b) and the last series (e) in this corollary refine identity (1.3) conjectured by Sun [23, Eq (2.15)]:

    n=1(3+554)n(3nn){H3nH2n}=1+52ln(9352).

    Now, we turn to examine Bailey's cubic transformation (Bailey [2, Eq (4.05)])

    3F2[a3,1+a3,2+a31+a+c2,2+ac2|27y4(1+y)3]=(1+y)a×3F2[a,c,1c1+a+c2,2+ac2|y4]. (4.1)

    Under the parameter replacements "a1+6ax,c2cx", it can be reformulated as

    n=0(y4(1+y)3)n(1+6ax)3nn!(1+3ax+cx)n(32+3axcx)n=(1+y)1+6ax×{1+2cxk=1(y4)k(1+6ax)k(1+2cx)k1(12cx)kk!(1+3ax+cx)k(32+3axcx)k}, (4.2)

    where, as before, both series converge for "y(ρ,12)", or equivalently, 27y4(1+y)3(1,1).

    When x=0, the above series becomes the following simpler one:

    n=0(3nn)yn(1+2n)(1+y)3n=1+y.

    Instead, the coefficients of x across (4.2) result in the general formula as below:

    n=1yn(1+y)3n(3nn){6aH3n(3a+c)Hn2(3ac)On1+2n+4(3ac)n(1+2n)2}=6a(1+y)ln(1+y)+2c(1+y)k=1(k1)!(32)k(y4)k.

    For brevity, we introduce the notations U(y) and V(y) for the two series:

    U(y)=k=1(k1)!(32)k(y4)kandV(y)=n=0yn(1+y)3n(3nn)n(1+2n)2.

    They are evaluated in closed form by the following lemma:

    Lemma 11.

    (a)U(y)=2+24+yyln4+yy2,4<y<4;(b)V(y)=(1+y)3yln(y+1+y)(1+y),ρ<y<1/2.

    Proof. The first series can be evaluated by computing the integral as follows:

    U(y)=k=1(k1)!(32)k(y4)k=k=1B(k,32)(y4)k=k=1(y4)k10xk11xdx=101xdxxk=1(xy4)k=10y1xdx4+xy=102yT24+yT2ydT1xT=2+24+yyln4+yy2.

    For the second series V(y), by shifting forward the summation index nn+1, we can rewrite it as

    V(y)=n=0yn(1+y)3n(3nn)n(1+2n)2=y(1+y)3n=0yn(1+y)3n(3n+3n)13+2n.

    Then, for "α=β=3", perform the replacement τxy/(1+y)3 in Lambert's second series

    n=0xnyn(1+y)3n(3n+3n)=(1+T)412T.

    Now multiplying this equation by x12 and then integrating across over x[0,1], we can proceed with

    n=0yn(1+y)3n(3n+3n)23+2n=10x12(1+T)412Tdxx=T(1+y)3y(1+T)3=(1+y)9yyy0T(1+T)3dT.

    By means of integration by parts, we can further compute

    y0T(1+T)3dT=2y1+y+y01T(1+T)dT=2ln(y+1+y)2y1+y.

    After substitution, the second formula in Lemma 11 follows accordingly.

    Therefore, we obtain the following reduced transformation:

    n=0yn(1+y)3n(3nn){6aH3n(3a+c)Hn2(3ac)On1+2n}=6a(1+y)ln(1+y)+2c(1+y)U(y)4(3ac)V(y). (4.3)

    By specifying two parameters, a and c in the above equality, we deduce four independent series:

    a=0:n=1yn(1+y)3n(3nn){ˉH2n1+2n}=(1+y)U(y)+2V(y),c=0:n=1yn(1+y)3n(3nn){H3nH2n1+2n}=(1+y)ln(1+y)2V(y),c=3a:n=1yn(1+y)3n(3nn){H3nHn1+2n}=(1+y)ln(1+y)+(1+y)U(y),c=3a:n=1yn(1+y)3n(3nn){H3n2On1+2n}=(1+y)ln(1+y)(1+y)U(y)4V(y).

    For the above four equations, first replacing U(y) and V(y) by their algebraic values in Lemma 11, and then putting the resulting equations in conjunction with that in Theorem 2, we establish, after some routine simplifications, four closed formulae as in the following theorem:

    Theorem 12 (ρ<y<1/2).

    (a)n=1yn(1+y)3n(3nn)ˉH2n1+2n=2(1+y)3yln(y+1+y)+2(1+y)4+yyln(4+yy2),(b)n=1yn(1+y)3n(3nn)H2n1+2n=2(1+y)3yln(y+1+y)3+3y2ln(12y)+(1+y)4+y2yln(2+2y2y2+4y2y(2+yy2+4y)),(c)n=1yn(1+y)3n(3nn)On1+2n=2(1+y)3yln(y+1+y)3+3y4ln(12y)+(1+y)4+y2yln(1+yy2+4y12y),(d)n=1yn(1+y)3n(3nn)H3n1+2n=2(1+y)+1+y2ln((1+y)2(12y)3)+(1+y)4+y2yln(2+2y2y2+4y2y(2+yy2+4y)).

    By assigning particular values for y in Theorems 2 and 12, we can deduce numerous infinite series identities with different convergence rates. In order to show the variety of implications, we record five representative classes of infinite series values.

    Corollary 13 (y52 in Theorems 2 and 12).

    (a)n=1(18)n(3nn)ˉH2n1+2n=22ln(3+101+2)(3+5)ln(1+52),(b)n=1(18)n(3nn)H2n1+2n=22ln(3+101+2)(645)ln(1+52)5ln5,(c)n=1(18)n(3nn)Hn1+2n=(553)ln(1+52)5ln5,(d)n=1(18)n(3nn)On1+2n=22ln(3+101+2)(92352)ln(1+52)52ln5,(e)n=1(18)n(3nn)H3n1+2n=(51)(2+ln2)(535)ln(1+52)5ln5.

    Corollary 14 (y3352 in Theorems 2 and 12).

    (a)n=1(227)n(3nn)ˉH2n1+2n=9+332ln(31)+96ln(3+223+2),(b)n=1(227)n(3nn)H2n1+2n=92ln(332)+332ln((3+1)3123)+96ln(3+223+2),(c)n=1(227)n(3nn)Hn1+2n=94ln(34)+332ln((3+1)4243),(d)n=1(227)n(3nn)On1+2n=34(33)ln(233)9ln343+96ln(3+223+2),(e)n=1(227)n(3nn)H3n1+2n=333+32ln(2332)+332ln(2+323).

    Corollary 15 (y324 in Theorems 2 and 12).

    (a)n=1(2+227)n(3nn)ˉH2n1+2n=33ln(6231)+9226ln(627+6(22)3),(b)n=1(2+227)n(3nn)H2n1+2n=33ln(312)(992)ln(1+23)+9226ln(627+6(22)3),(c)n=1(2+227)n(3nn)Hn1+2n=33ln(2+32+3)(992)ln(1+23),(d)n=1(2+227)n(3nn)On1+2n=(9292)ln(63)332ln(2+3)+9226ln(627+6(22)3),(e)n=1(2+227)n(3nn)H3n1+2n=626+33ln(312)(332)ln(3+223).

    Corollary 16 (y15 in Theorems 2 and 12).

    (a)n=1(25216)n(3nn)ˉH2n1+2n=1265ln(1+65)12215ln(1+2125),(b)n=1(25216)n(3nn)H2n1+2n=95ln(53)+1265ln(1+65)+6215ln(921215),(c)n=1(25216)n(3nn)Hn1+2n=95ln(53)+3215ln(2571321250),(d)n=1(25216)n(3nn)On1+2n=910ln(53)+1265ln(1+65)+3215ln(62115),(e)n=1(25216)n(3nn)H3n1+2n=125+35ln(203)+6215ln(921215).

    Corollary 17 (y110 in Theorems 2 and 12).

    (a)n=1(100729)n(3nn)ˉH2n1+2n=275arccot39395arccot39,(b)n=1(100729)n(3nn)H2n1+2n=275arccot393920arccos(6780)2720ln(65),(c)n=1(100729)n(3nn)Hn1+2n=93920arccos(79878000)2720ln(65),(d)n=1(100729)n(3nn)On1+2n=275arccot393920arccos(2740)2740ln(65),(e)n=1(100729)n(3nn)H3n1+2n=9593920arccos(6780)920ln(3215).

    Under the parameter replacements "a6ax,c2cx", Bailey's cubic transformation (4.1) can be reformulated alternatively as

    n=0(y4(1+y)3)n(6ax)3nn!(1+3axcx)n(12+3ax+cx)n=(1+y)6ax×{1+12acx2k=1(y4)k(1+6ax)k1(1+2cx)k1(12cx)kk!(1+3axcx)k(12+3ax+cx)k}.

    Both series in the above equation converge subject to the same condition "y(ρ,12)", or equivalently, 27y4(1+y)3(1,1) (as before).

    Equating the constant terms from both sides yields the algebraic identity below

    n=1(3nn)ynn(1+y)3n=3ln(1+y).

    Instead, the coefficients of x result in the following infinite series identity involving a free variable "y" and two linear parameters a and c:

    n=1yn(1+y)3n(3nn){6aH3n(3ac)Hn2(3a+c)Onn2an2}=9aln2(1+y)+6ck=1(k1)!k(12)k(y4)k.

    Let U(y) and V(y) stand for the two series

    U(y)=k=1(k1)!k(12)k(y4)kandV(y)=n=1ynn2(1+y)3n(3nn).

    They admit closed formulae as in the lemma below:

    Lemma 18.

    (a)U(y)=2ln2(2y+4+y),4<y<4;(b)V(y)=3{π26+lnyln(1+y)2ln2(1+y)Li2(11+y)},ρ<y<1/2.

    Proof. The first series U(y) can be reformulated as

    U(y)=k=1(k1)!k(12)k(y4)k=k=1B(k,12)(y4)k=k=1(y/4)kk10xk1dx1x=10dxx1xk=1(xy/4)kk=10ln(1+xy4)dxx1x=10dTTln(1+(2+y)T+T2(1+T)2)x4T(1+T)2=10dTT{2ln(1+T)ln(1+4T(y+4+y)2)ln(1+4T(y4+y)2)}=π26+Li2(4(y+4+y)2)+Li2(4(y4+y)2).

    Then the closed formula for U(y) in the lemma follows by applying the dilogarithmic equation

    Li2(x)+Li2(x1)=π26ln2x2,wherex>0.

    For the second series, we can rewrite it by reindexing nn+1 as

    V(y)=n=1ynn2(1+y)3n(3nn)=3y(1+y)3n=0yn(1+y)3n(3n+2n)1(1+n)2.

    Now specializing Lambert's second series by "α=2 and β=3" and τxy/(1+y)3

    n=0xnyn(1+y)3n(3n+2n)=(1+T)312T,wherex=T(1+y)3y(1+T)3,

    then integrating twice across this equation, we can proceed with

    n=0yn(1+y)3n(3n+2n)1(1+n)2=10dttt0(1+T)312Tdx=10(1+T)3lnx12Tdx=(1+y)3yy0dT1+Tln(T(1+y)3y(1+T)3)x=T(1+y)3y(1+T)3=(1+y)3y{y0dT1+Tln(T(1+y)3y)3y0ln(1+T)1+TdT}=(1+y)3y{π26+lnyln(1+y)2ln2(1+y)Li2(11+y)}.

    By substitution, the closed formula for V(y) in the lemma follows consequently.

    Thus, we have the following reduced transformation:

    n=1yn(1+y)3n(3nn){6aH3n(3ac)Hn2(3a+c)Onn}=9aln2(1+y)+6cU(y)+2aV(y). (5.1)

    By specifying two parameters, a and c, we deduce four independent series

    a=0:n=1yn(1+y)3n(3nn)ˉH2nn=3U(y),c=0:n=1yn(1+y)3n(3nn){H3nH2nn}=32ln2(1+y)+V(y)3,c=3a:n=1yn(1+y)3n(3nn){H3n2Onn}=32ln2(1+y)+3U(y)+V(y)3,c=3a:n=1yn(1+y)3n(3nn){H3nHnn}=32ln2(1+y)3U(y)+V(y)3.

    For these four equations, replacing U(y) and V(y) first by their algebraic expressions and then relating the resulting equations to that in Theorem 3, we find, after some simplifications, four infinite series identities as below:

    Theorem 19 (ρ<y<1/2). Letting Θ(y) be defined as in Theorem 3, the following algebraic identities hold:

    (a)n=1yn(1+y)3n(3nn)ˉH2nn=6ln2(2y+4+y),(b)n=1yn(1+y)3n(3nn)H2nn=π22+3Θ(y)+3lnyln(1+y)92ln2(1+y)+6ln2(2y+4+y),(c)n=1yn(1+y)3n(3nn)Onn=π24+32Θ(y)+32lnyln(1+y)94ln2(1+y)+6ln2(2y+4+y),(d)n=1yn(1+y)3n(3nn)H3nn=2π23+3Θ(y)+4lnyln(1+y)5ln2(1+y)+6ln2(2y+4+y)Li2(11+y).

    By assigning five particular values for y in Theorems 3 and 19, five classes of infinite series identities are displayed in the following corollaries:

    Corollary 20 (y52 in Theorems 3 and 19).

    (a) n=1(18)n(3nn)ˉH2nn=32ln2(1+52),(b) n=1(18)n(3nn)H2nn=π22+32ln22+3Θ(52)+6ln(1+5)ln(1+54),(c) n=1(18)n(3nn)Hnn=π22+3Θ(52)+92ln(1+5)ln(1+54),(d) n=1(18)n(3nn)Onn=π24+32ln22+32Θ(52)+154ln(1+5)ln(1+54),(e)n=1(18)n(3nn)H3nn=2π23+112ln22+3Θ(52)+ln(1+5)2ln((1+5)17238)Li2(1+54).

    Corollary 21 (y525 in Theorems 3 and 19).

    (a) n=1(564)n(3nn)ˉH2nn=32ln2(5+510),(b) n=1(564)n(3nn)H2nn=π22+3Θ(525)+32ln(5+58)ln(64+1285)+32ln2(5+510),(c) n=1(564)n(3nn)Hnn=π22+3Θ(525)+32ln(5+58)ln(64+1285),(d) n=1(564)n(3nn)Onn=π24+32Θ(525)+34ln(5+58)ln(64+1285)+32ln2(5+510),(e) n=1(564)n(3nn)H3nn=2π23+3Θ(525)Li2(5+58)+32ln2(5+510)+ln(5+58)ln(8(5+5)754).

    Corollary 22 (y13 in Theorems 3 and 19).

    (a)n=1(964)n(3nn)ˉH2nn=32ln2(7+136),(b) n=1(964)n(3nn)H2nn=π2232ln(43)ln(643)+3Θ(13)+32ln2(7+136),(c) n=1(964)n(3nn)Hnn=π2232ln(43)ln(643)+3Θ(13),(d) n=1(964)n(3nn)Onn=π2434ln(43)ln(643)+32Θ(13)+32ln2(7+136),(e) n=1(964)n(3nn)H3nn=2π23ln(43)ln(10243)+3Θ(13)Li2(34)+32ln2(7+136).

    Corollary 23 (y3352 in Theorems 3 and 19).

    (a) n=1(227)n(3nn)ˉH2nn=32ln2(31),(b) n=1(227)n(3nn)H2nn=π22+32ln(54)ln(23)+6ln(1+3)ln(1+32)+3Θ(3353),(c) n=1(227)n(3nn)Hnn=π22+32ln(1+33)ln(27(1+3)34)+3Θ(3353),(d) n=1(227)n(3nn)Onn=π24+34ln(1+32)ln((1+3)52)+34ln(54)ln(23)+32Θ(3353),(e) n=1(227)n(3nn)H3nn=2π23+32ln2(1+32)Li2(1+33)+3Θ(3353)+ln(1+33)ln(35(1+3)724).

    Corollary 24 (y324 in Theorems 3 and 19).

    (a) n=1(2+227)n(3nn)ˉH2nn=32ln2(1+32+6),(b) n=1(2+227)n(3nn)H2nn=π22+3Θ(324)+32ln2(1+32+6)+32ln(1+23)ln(27(1+2)2),(c) n=1(2+227)n(3nn)Hnn=π22+3Θ(324)+32ln(1+23)ln(27(1+2)2),(d) n=1(2+227)n(3nn)Onn=π24+32Θ(324)+32ln2(1+32+6)+34ln(1+23)ln(27(1+2)2),(e) n=1(2+227)n(3nn)H3nn=2π23+3Θ(324)+32ln2(1+32+6)Li2(1+23)+ln(1+23)ln(35(1+2)34).

    By integrating Lambert's series and manipulating the cubic transformations for 3F2-series through the "coefficient extraction method", we have shown several algebraic formulae for infinite series containing binomial coefficient (3nn), harmonic-like numbers, and in particular, a free variable "y". As showcases, three classes of infinite series were examined in detail from Sections 3–5. More variants of these series exist and can be treated analogously. For instance, the transformation formula (3.1) under the replacements "a1+6ax,c2cx" becomes

    3F2[1+2ax,23+2ax,43+2ax1+3ax+cx,32+3axcx|27y4(1+y)3]=(1+y)2+6ax12y×4F3[43+2ax,1+6ax,2cx,12cx13+2ax,1+3ax+cx,32+3axcx|y4].

    Then by carrying out the same procedure as in the preceding sections, we can establish the algebraic identities for five infinite series, as in the theorem below:

    Theorem 25 (ρ<y<1/2). Letting Λ(y) be the function defined by

    Λ(y)=3ln(12y)+2yy2+4yln(2y(2+y+y2+4y)2y(2+yy2+4y)),

    we have the closed algebraic formulae for five infinite series:

    (a) n=1yn(1+y)3n(3n+1n)ˉH2n+1=2(y2)(1+y)2(12y)y2+4yln4+yy2,(b) n=1yn(1+y)3n(3n+1n)H2n+1=(1+y)212y{2y4y2+4yln4+yy2Λ(y)2},(c) n=1yn(1+y)3n(3n+1n)Hn=(1+y)24y2Λ(y),(d) n=1yn(1+y)3n(3n+1n)On+1=(1+y)212y{2y4y2+4yln4+yy2Λ(y)4},(e) n=1yn(1+y)3n(3n+1n)H3n+1=(1+y)212y{2y4y2+4yln4+yy2Λ(y)2+ln(1+y)}.

    From the algebraic identities exhibited for the series with harmonic-like numbers of the first order, it is natural to ask: what would happen for the corresponding series involving harmonic-like numbers of the second order? Our computations manifested that the resulting expressions are very complex and convoluted, which can be exemplified by the following attempt made by the authors. Extracting the coefficient of x2 across (3.2), we can derive, after a number of reductions and simplifications, the following formulae:

    (a)n=1yn(1+y)3n(3nn){ˉH22n+H22n}=6y+62y1{U(y)+W(y)},(b)n=1yn(1+y)3n(3nn){(H3nH2n)2+H22nH23n}=1+y12yln2(1+y),(c)n=1yn(1+y)3n(3nn){(2OnH3n)2+4O2nH23n}=1+y12y{ln2(1+y)+6ln(1+y)W(y)10U(y)12W(y)},(d)n=1yn(1+y)3n(3nn){(HnH3n)2+H2nH23n}=1+y12y{ln2(1+y)6ln(1+y)W(y)2U(y)};

    where W(y)andU(y) are as in Lemmas 4 and 18, respectively, and W(y) is given by

    W(y)=k=1(k1)!(12)k(y4)k¯H2k=(1y4+y)ln2(4+y+y2)y4+yLi2(2yy+4+y).

    There are altogether nine harmonic-like numbers appearing in the above four series

    {H2n,O2n,H23n;H2n,O2n,H23n;HnOn,HnH3n,H3nOn}.

    To evaluate the nine corresponding independent "nuclear series", it is necessary to find five similar independent equations. It seems quite a tough problem that the authors failed to arrive at solutions, even though we did succeed in determining the following long and ugly expression involving six terms in dilogarithms (that prevents us from making further simplifications). Perhaps the only value of this formula lies in its existence.

    n=1(3nn)H2nyn(1+y)3n=y03(1+y)3ln(T(1+y)3y(1+T)3)dT(12y)(1+T)(T2y+Ty2+3Ty1)=π2(1+y)2(12y)+3(1+y)ln2(1+y)12y+3(1+y)lnyln(1+y)12y3(1+y)12yLi2(11+y)+9(1+y)12yy4+yln(1+y)ln(2+y+4y+y22)+3(1+y)2(12y)(1+y4+y){3Li2(2y(1+y)(y4+y))3Li2(2yy4+y)+Li2(2y3/2y(3+y)(1+y)4+y)}+3(1+y)2(12y)(1y4+y){3Li2(2y(1+y)(y+4+y))3Li2(2yy+4+y)+Li2(2y3/2y(3+y)+(1+y)4+y)}.

    Therefore, to evaluate the series containing binomial coefficient (3nn) and harmonic-like numbers of the second and/or higher orders, one needs to find out different approaches. The interested readers are enthusiastically encouraged to make further explorations.

    C. Li: Writing, computation and review; W. Chu: Methodology and supervision. Both of authors have read and approved the final version of the manuscript for publication.

    The authors declare they have not used Artificial Intelligence tools in the creation of this article.

    The authors express their gratitude to anonymous referees for their careful reading, critical comments and valuable suggestions that make the manuscript improved substantially during the revision.

    Prof. Wenchang Chu is the Guest Editor of special issue "Combinatorial Analysis and Mathematical Constants" for AIMS Mathematics. Prof. Wenchang Chu was not involved in the editorial review and the decision to publish this article. The authors declare that there are no conflicts of interest regarding the publication of this paper.



    [1] G. Luo, Y. Liu, Q. Zeng, Su. Diao, F. Xiong, A dynamic evolution model of human opinion as affected by advertising, Physica A, 414 (2014), 254–262. https://doi.org/10.1016/j.physa.2014.07.055 doi: 10.1016/j.physa.2014.07.055
    [2] J. B. Ford, S. Bezbaruah, P. Mukherji, et al., A decade (2008-2019) of advertising research productivity: A bibliometric review, J. Business Res., 136 (2021), 137–163. https://doi.org/10.1016/j.jbusres.2021.07.030 doi: 10.1016/j.jbusres.2021.07.030
    [3] Z. Guo, J. Ma, Dynamics and implications on a cooperative advertising model in the supply chain, Commun Nonlinear Sci. Numer. Simul., 64 (2018), 198–212. https://doi.org/10.1016/j.cnsns.2018.04.017 doi: 10.1016/j.cnsns.2018.04.017
    [4] X. Chen, Y. Zhu, L. Sheng, Optimal control for uncertain stochastic dynamic systems with jump and application to an advertising model, Appl. Math. Comput., 407 (2021), 126337. https://doi.org/10.1016/j.amc.2021.126337 doi: 10.1016/j.amc.2021.126337
    [5] R. Y. Chenavaz, G. Feichtinger, R. F. Hartl, P. M. Kort, Modeling the impact of product quality on dynamic pricing and advertising policies, Eur. J. Oper. Res., 284 (2020), 990–1001. https://doi.org/10.1016/j.ejor.2020.01.035 doi: 10.1016/j.ejor.2020.01.035
    [6] L. Leong, T. Hew, K. B. Ooi, Y. K. Dwivedi, Predicting trust in online advertising with an SEM-artificial neural network approach, Expert Syst. Appl., 162 (2020), 113849. https://doi.org/10.1016/j.eswa.2020.113849 doi: 10.1016/j.eswa.2020.113849
    [7] L. Weng, Z. Huang, J. Bao, A model of tourism advertising effects, Tour. Manag., 85 (2021), 104278. https://doi.org/10.1016/j.tourman.2020.104278 doi: 10.1016/j.tourman.2020.104278
    [8] Q. Wang, J. Wen, P. Zhang, Oscillation analysis of advertising capital model: Analytical and numerical studies, Appl. Math. Comput., 354 (2019), 365–376. https://doi.org/10.1016/j.amc.2019.02.029 doi: 10.1016/j.amc.2019.02.029
    [9] M. Nerlove, K. J. Arrow, Optimal advertising policy under dynamic conditions, Economica, 39 (1962), 129–142. https://doi.org/10.2307/2551549 doi: 10.2307/2551549
    [10] M. L. Vidale, H. B. Wolfe, An operations research study of sales response to advertising, Oper. Res., 5 (1957), 370–381. https://doi.org/10.1287/opre.5.3.370 doi: 10.1287/opre.5.3.370
    [11] X. Qiu, G. Zhu, Y. Ding, K. Li, Successive lag synchronization on complex dynamical networks via delay-dependent impulsive control, Physica A, 531 (2019), 121753.
    [12] N. Bandyopadhyay, B. Sivakumaran, S. Patro, R. S. Kumar, Immediate or delayed! Whether various types of consumer sales promotions drive impulse buying?: An empirical investigation, J. Retail. Consum. Serv., 61 (2021), 102532. https://doi.org/10.1016/j.physa.2019.121753 doi: 10.1016/j.physa.2019.121753
    [13] R. McColl, R. Macgilchrist, S. Rafiq, Estimating cannibalizing effects of sales promotions: The impact of price cuts and store type, J. Retail. Consum. Serv., 53 (2020), 101982. https://doi.org/10.1016/j.jretconser.2019.101982 doi: 10.1016/j.jretconser.2019.101982
    [14] E. Crespo-Almendros, S. Del Barrio-García, Online airline ticket purchasing: Influence of online sales promotion type and Internet experience, J. Air Transp. Manag., 53 (2016), 23–34. https://doi.org/10.1016/j.jairtraman.2016.01.004 doi: 10.1016/j.jairtraman.2016.01.004
    [15] H. Jiang, Z. Feng, G. Jiang, Dynamics of an advertising competition model with sales promotion, Commun. Nonlinear Sci. Numer. Simul., 42 (2017), 37–51. https://doi.org/10.1016/j.cnsns.2016.05.007 doi: 10.1016/j.cnsns.2016.05.007
    [16] F. M. Bass, A new product growth model for consumer durables, Manage. Sci., 15 (1969), 215–227. https://doi.org/10.1287/mnsc.15.5.215 doi: 10.1287/mnsc.15.5.215
    [17] J. Guckenheimer, P. Holmes, Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields, New York: Springer-Verlag, 1983. https://doi.org/10.1007/978-1-4612-1140-2
  • This article has been cited by:

    1. Chunli Li, Wenchang Chu, Generating Functions for Binomial Series Involving Harmonic-like Numbers, 2024, 12, 2227-7390, 2685, 10.3390/math12172685
    2. Artūras Dubickas, On the Range of Arithmetic Means of the Fractional Parts of Harmonic Numbers, 2024, 12, 2227-7390, 3731, 10.3390/math12233731
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2455) PDF downloads(93) Cited by(4)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog