License: CC BY 4.0
arXiv:2606.25468v1 [math.CA] 24 Jun 2026

An inequality for Jacobi polynomials: a complement to Finite Increment Theorem

Geno Nikolov Faculty of Mathematics and Informatics, Sofia University ”St. Kliment Ohridski”, 5 James Bourchier Blvd., 1164 Sofia, Bulgaria geno@fmi.uni-sofia.bg
Abstract.

Let P=Pn(α,β)P=P_{n}^{(\alpha,\beta)} be the nn-th degree Jacobi polynomial, which is orthogonal in [1,1][-1,1] with respect to the weight function (1x)α(1+x)β(1-x)^{\alpha}(1+x)^{\beta}, α,β>1\alpha,\beta>-1. For parameters (α,β)(\alpha,\beta) satisfying either αβ1/2\alpha\geq\beta\geq 1/2 or α1/2\alpha\geq 1/2, β=1/2\beta=-1/2, we prove the inequality

P(1)P(x)P(x)(1x),x[0,1],P(1)-P(x)\geq P^{\prime}(x)\,(1-x),\quad x\in[0,1],

which may be viewed as a complement to Finite Increment Theorem for Jacobi polynomials.


Keywords: Finite Increment Theorem, Chebyshev polynomials, Jacobi polynomials, nonnegative expansion.

2020 Mathematics Subject Classification: 41A17.

1. Introduction and statement of the result

In [3] we investigated two trigonometric inequalities proposed by M. S. Robertson [5] and R. Askey and G. Gasper [2], respectively (see (1.29) and (8.17) in [1]). These inequalities were restated in [3] in terms of the Chebyshev polynomial of the first kind

Tn(x)=cos(nθ),x=cosθ,T_{n}(x)=\cos(n\theta),\quad x=\cos\theta\,,

and the following refinement was proved:

Theorem A. Let nn\in\mathbb{N}, n4n\geq 4. Then

(1) Tn(x)+2x+2n2Tn(x)a(n)1xn2(n2Tn(x))0,x[1,1],T_{n}(x)+2-\frac{x+2}{n^{2}}\,T_{n}^{\prime}(x)-a(n)\frac{1-x}{n^{2}}\,\big(n^{2}-T_{n}^{\prime}(x)\big)\geq 0,\quad x\in[-1,1],

where

(2) a(n)={11+cosπn, if n is even,11+cos2πn, if n is odd.a(n)=\begin{cases}\frac{1}{1+\cos\frac{\pi}{n}}\,,&\textit{ if }\ n\ \textit{ is even},\\ \frac{1}{1+\cos\frac{2\pi}{n}}\,,&\textit{ if }\ n\ \textit{ is odd.}\end{cases}

The constant a(n)a(n) in (2) is the best possible in the sense that (1) fails for any larger constant. The equality in (1) is attained only at x=1x=1 and x=cosπnx=-\cos\frac{\pi}{n} if nn is even, and at x=±1x=\pm 1 and x=cos2πnx=-\cos\frac{2\pi}{n} if nn is odd.

The inequalities of Robertson and of Askey and Gasper are obtained with choosing in (1) a(n)=0a(n)=0 and a(n)=1/2a(n)=1/2, respectively. We note that Theorem A is true also for n=1, 2, 3n=1,\,2,\,3, in which cases the inequality in (1) becomes an identity.

For x[0,1]x\in[0,1] the inequality in (1) holds true with a larger constant a(n)a(n), namely with a=1a=1, yielding (cf. [3, Theorem 2])

Theorem B. For every nn\in\mathbb{N}, n3n\geq 3, the following inequality holds true:

(3) Tn(1)Tn(x)(1x)Tn′′(x),x[0,1].T_{n}^{\prime}(1)-T_{n}^{\prime}(x)\geq(1-x)\,T_{n}^{\prime\prime}(x),\quad x\in[0,1].

The equality in (3) occurs only for x=1x=1 and, if n2(mod4)n\equiv 2\,(\!\!\!\mod 4), for x=0\,x=0.

In the cases n=0, 1, 2n=0,\,1,\,2, inequality (3) becomes an identity. The Chebyshev polynomials of the second kind {Um}\{U_{m}\} are related to the Chebyshev polynomial of the first kind by

(4) Um(x)=sin(m+1)θsinθ=Tm+1(x)m+1,x=cosθ.U_{m}(x)=\frac{\sin(m+1)\theta}{\sin\theta}=\frac{T_{m+1}^{\prime}(x)}{m+1}\,,\quad x=\cos\theta.

Therefore, inequality (3) can be reformulated as

(5) Un(1)Un(x)(1x)Un(x),n0,x[0,1].U_{n}(1)-U_{n}(x)\geq(1-x)\,U_{n}^{\prime}(x),\quad n\in\mathbb{N}_{0},\ x\in[0,1].

According to the Finite Increment Theorem, Un(1)Un(x)=(1x)Un(ξ)\,U_{n}(1)-U_{n}(x)=(1-x)\,U_{n}^{\prime}(\xi) for some ξ(x,1)\xi\in(x,1), hence inequality (3) may be viewed as a complement to this theorem applied to UnU_{n}.

The Chebyshev polynomials of the first and the second kind are ultraspherical polynomials orthogonal in (1,1)(-1,1) with respect to the weight functions 1/1x21/\sqrt{1-x^{2}} and 1x2\sqrt{1-x^{2}}, respectively. On the other hand, the ultraspherical (called also Gegenbauer) polynomials belong to the family of Jacobi polynomials {Pn(α,β)(x)}n0\{P_{n}^{(\alpha,\beta)}(x)\}_{n\in\mathbb{N}_{0}}, which are orthogonal in [1,1][-1,1] with respect to the weight function wα,β(x)=(1x)α(1+x)βw_{\alpha,\beta}(x)=(1-x)^{\alpha}(1+x)^{\beta}, α,β>1\alpha,\,\beta>-1. Inequality (5) says that for (α,β)=(1/2,1/2)(\alpha,\beta)=(1/2,1/2) we have

(6) Pn(α,β)(1)Pn(α,β)(x)(1x)ddx{Pn(α,β)(x)},x[0,1].P_{n}^{(\alpha,\beta)}(1)-P_{n}^{(\alpha,\beta)}(x)\geq(1-x)\,\frac{d}{dx}\big\{P_{n}^{(\alpha,\beta)}(x)\big\}\,,\quad x\in[0,1]\,.

A rather natural question is: for which other parameters (α,β)(\alpha,\beta) the inequality (6) holds true ?

The aim of this note is to provide a partial answer of the above question. Our result is

Theorem 1.

The inequality (6) holds true in the following cases:

  • (i)  

    α12,β=12\alpha\geq\frac{1}{2},\ \beta=-\frac{1}{2}\,;

  • (ii)  

    αβ12\alpha\geq\beta\geq\frac{1}{2}\,.

Unless α=β=12\alpha=\beta=\frac{1}{2}, for n3n\geq 3 the equality in (6) is attained only at x=1x=1.

The special case α=12\alpha=\frac{1}{2}, β=12\beta=-\frac{1}{2} of Theorem 1(i) comes down to a trigonometric inequality, which to the best of our knowledge is new:

Corollary 2.

For every n,n2n\in\mathbb{N},\;n\geq 2, the following inequality holds true:

2n+1(n+2)sinnθsinθ+(n1)sin(n+1)θsinθ0,θ[0,π2].2n+1-(n+2)\,\frac{\sin n\theta}{\sin\theta}+(n-1)\,\frac{\sin(n+1)\theta}{\sin\theta}\geq 0\,,\quad\theta\in\big[0,\frac{\pi}{2}\big]\,.

The equality is attained only for θ=0\theta=0.

The proofs are given in the next section. We provide a short proof of Theorem B for two reasons: firstly, for the readers convenience, and secondly, because Theorem B constitutes the main ingredient of our arguments.

2. Proofs

2.1. Proof of Theorem B

Denote by τ\tau the largest zero of Tn′′T_{n}^{\prime\prime}, then

cos2πn<τ<cosπn.\cos\frac{2\pi}{n}<\tau<\cos\frac{\pi}{n}\,.

Tn′′(x)T_{n}^{\prime\prime}(x) increases monotonically in [τ,1][\tau,1], therefore

Tn(1)Tn(x)=x1Tn′′(u)𝑑u(1x)Tn′′(x),x[τ,1],T_{n}^{\prime}(1)-T_{n}^{\prime}(x)=\int\limits_{x}^{1}T_{n}^{\prime\prime}(u)\,du\geq(1-x)T_{n}^{\prime\prime}(x),\quad x\in[\tau,1],

with the equality only when x=1x=1. Now the assertion of Theorem B obviously follows when n=3n=3, and we assume in what follows n4n\geq 4. Since the left-hand side of (3) is positive in [0,1)[0,1) while Tn′′(x)<0\,T_{n}^{\prime\prime}(x)<0 for x[cos2πn,τ)x\in\big[\cos\frac{2\pi}{n},\tau\big), it follows that (3) holds true with >">" sign when x[cos2πn,τ)x\in\big[\cos\frac{2\pi}{n},\tau\big). Thus, we need to prove the inequality in (3) for x[0,cos2πn)x\in[0,\cos\frac{2\pi}{n}) and n5n\geq 5. Let us set

(7) (x)=y(1)y(x)(1x)y′′(x),y=Tn(x).\mathcal{L}(x)=y^{\prime}(1)-y^{\prime}(x)-(1-x)y^{\prime\prime}(x)\,,\qquad y=T_{n}(x)\,.

By using the differential equation for yy (see, e.g., [4, Eqn. (1.92)])

(8) (1x2)y′′(x)xy(x)+n2y(x)=0(1-x^{2})y^{\prime\prime}(x)-x\,y^{\prime}(x)+n^{2}y(x)=0

and the representations

Tn(x)=cosnθ,Tn(x)=nsinnθsinθ,x=cosθ,T_{n}(x)=\cos n\theta,\ T_{n}^{\prime}(x)=n\,\frac{\sin n\theta}{\sin\theta},\quad x=\cos\theta,

we find

(0)={2n2, if n0(mod4),n2n, if n1(mod4),0, if n2(mod4),n2+n, if n3(mod4).\mathcal{L}(0)=\begin{cases}2n^{2},&\text{ if }n\equiv 0\,(\!\!\!\!\mod 4),\\ n^{2}-n,&\text{ if }n\equiv 1\,(\!\!\!\!\mod 4),\\ 0,&\text{ if }n\equiv 2\,(\!\!\!\!\mod 4),\\ n^{2}+n,&\text{ if }n\equiv 3\,(\!\!\!\!\mod 4).\end{cases}

Thus, (0)0\mathcal{L}(0)\geq 0 and the equality is attained only when n=4k+2,kn=4k+2,\ k\in\mathbb{N}. It remains to prove that (x)>0\mathcal{L}(x)>0 for x(0,cos2πn)x\in(0,\cos\frac{2\pi}{n}), and it suffices to verify this inequality at the points of local extremum of (x)\mathcal{L}(x) therein. Since (x)=(1x)y′′′(x)\mathcal{L}^{\prime}(x)=(1-x)y^{\prime\prime\prime}(x), the latter points are the zeros of y′′′(x)y^{\prime\prime\prime}(x) in (0,cos2πn)\,(0,\cos\frac{2\pi}{n}). Let t(0,cos2πn)t\in(0,\cos\frac{2\pi}{n})\, be a zero of y′′′(x)y^{\prime\prime\prime}(x). Note that

cos2πn=12sin2πn<12(2ππn)2=18n2,\cos\frac{2\pi}{n}=1-2\sin^{2}\frac{\pi}{n}<1-2\Big(\frac{2}{\pi}\frac{\pi}{n}\Big)^{2}=1-\frac{8}{n^{2}}\,,

therefore

0<t<18n2.0<t<1-\frac{8}{n^{2}}\,.

Differentiating (8) at x=tx=t and using y′′′(t)=0y^{\prime\prime\prime}(t)=0, we express y′′(t)y^{\prime\prime}(t) through y(t)y^{\prime}(t) as

y′′(t)=n213ty(t).y^{\prime\prime}(t)=\frac{n^{2}-1}{3t}\,y^{\prime}(t).

By plugging this expression in equation (8) with x=tx=t, we obtain

(9) y(t)=3n2tn21(n2+2)t2y(t),y^{\prime}(t)=-\frac{3n^{2}t}{n^{2}-1-(n^{2}+2)t^{2}}\,y(t)\,,

and consequently

(10) y′′(t)=n2(n21)n21(n2+2)t2y(t).y^{\prime\prime}(t)=-\frac{n^{2}(n^{2}-1)}{n^{2}-1-(n^{2}+2)t^{2}}\,y(t)\,.

Now we replace y(1)=n2y^{\prime}(1)=n^{2} and the expressions from (9) and (10) in (t)\mathcal{L}(t) to obtain

(11) (t)=n2[1+n21(n24)tn21(n2+2)t2y(t)].\mathcal{L}(t)=n^{2}\Big[1+\frac{n^{2}-1-(n^{2}-4)t}{n^{2}-1-(n^{2}+2)t^{2}}\,y(t)\Big]\,.

The quotient in front of y(t)y(t) in the last expression is positive. Indeed, this is obvious for the numerator, and for the denominator follows from

t2<t<18n2<13n2+2=n21n2+2.t^{2}<t<1-\frac{8}{n^{2}}<1-\frac{3}{n^{2}+2}=\frac{n^{2}-1}{n^{2}+2}\,.

Since y(t)1y(t)\geq-1, we conclude from (11) that

(t)n2[1n21(n24)tn21(n2+2)t2]=t[n24(n2+2)t)]n21(n2+2)t2.\mathcal{L}(t)\geq n^{2}\Big[1-\frac{n^{2}-1-(n^{2}-4)t}{n^{2}-1-(n^{2}+2)t^{2}}\Big]=\frac{t\big[n^{2}-4-(n^{2}+2)t)\big]}{n^{2}-1-(n^{2}+2)t^{2}}\,.

The numerator of the last quotient is positive, which is seen from

t<18n2<16n2+2=n24n2+2.t<1-\frac{8}{n^{2}}<1-\frac{6}{n^{2}+2}=\frac{n^{2}-4}{n^{2}+2}\,.

Hence, (t)>0\mathcal{L}(t)>0 and the proof of Theorem B is accomplished.

2.2. Proof of Theorem 1

We start with the special case (α,β)=(12,12)\,(\alpha,\beta)=\big(\frac{1}{2},-\frac{1}{2}\big) of Theorem 1(i). Apart from a constant multiplier, the Jacobi polynomial Pn(1/2,1/2)(x)P_{n}^{(1/2,-1/2)}(x) coincides with the Chebyshev polynomial of the fourth kind

(12) Wn(x)=sin(n+12)θsinθ2,x=cosθ.W_{n}(x)=\frac{\sin\big(n+\frac{1}{2}\big)\theta}{\sin\frac{\theta}{2}},\qquad x=\cos\theta\,.

This representation and (4) imply

(13) Un(x)+Un1(x)=sin(n+1)θsinθ+sinnθsinθ=sin(n+12)θsinθ2=Wn(x).U_{n}(x)+U_{n-1}(x)=\frac{\sin(n+1)\theta}{\sin\theta}+\frac{\sin n\theta}{\sin\theta}=\frac{\sin\big(n+\frac{1}{2}\big)\theta}{\sin\frac{\theta}{2}}=W_{n}(x)\,.

Now using (5), we obtain for x[0,1]x\in[0,1]

Wn(1)Wn(x)=Un(1)Un(x)+Un1(1)Un1(x)(1x)Un(x)+(1x)Un1(x)=(1x)Wn(x)0.\begin{split}W_{n}(1)-W_{n}(x)&=U_{n}(1)-U_{n}(x)+U_{n-1}(1)-U_{n-1}(x)\\ &\geq(1-x)U_{n}^{\prime}(x)+(1-x)U_{n-1}^{\prime}(x)\\ &=(1-x)W_{n}^{\prime}(x)\geq 0\,.\end{split}

For n3n\geq 3 the equality occurs only for x=1x=1. Indeed, according to Theorem B, the equality can only occur at x=0x=0 or x=1x=1, but it is not possible for both Un(1)Un(x)(1x)Un(x)U_{n}(1)-U_{n}(x)-(1-x)U_{n}^{\prime}(x) and Un1(1)Un1(x)(1x)Un1(x)U_{n-1}(1)-U_{n-1}(x)-(1-x)U_{n-1}^{\prime}(x) to vanish at x=0x=0.

We shall need the next lemma, which provides a positive expansion property of Jacobi polynomials.

Lemma 3.

The Jacobi polynomials obey the following properties:

  • (i)

    If α>γ>1\alpha>\gamma>-1, then

    Pn(α,β)=k=0nck,nPk(γ,β)with all ck,n>0.P_{n}^{(\alpha,\beta)}=\sum_{k=0}^{n}c_{k,n}\,P_{k}^{(\gamma,\beta)}\quad\text{with all }\ c_{k,n}>0\,.
  • (ii)

    If γ>α>1\gamma>\alpha>-1, then

    Pn(γ,γ)=k=0nck,nPk(α,α)with all ck,n0,P_{n}^{(\gamma,\gamma)}=\sum_{k=0}^{n}c_{k,n}\,P_{k}^{(\alpha,\alpha)}\quad\text{with all }\ c_{k,n}\geq 0\,,

    and ck,n>0\,c_{k,n}>0\, if kn(mod2)\,k\equiv n\ (\!\!\!\mod 2).

The explicit formulae for the coefficients {ck,n}\{c_{k,n}\} are given in [1, Eqns. (7.33) and (7.34)] (another reference for part (ii), which concerns the ultraspherical polynomials, is [6, pp. 95–96]).

The general case of Theorem 1(i) is a consequence from the case (α,β)=(1/2,1/2)(\alpha,\beta)=(1/2,-1/2) and Lemma 3(i) with γ=12\gamma=\frac{1}{2} and β=12\beta=-\frac{1}{2}. If α>12\alpha>\frac{1}{2} and x[0,1]x\in[0,1], then

Pn(α,1/2)(1)Pn(α,1/2)(x)(1x)ddx{Pn(α,1/2)(x)}=m=0ncm,n[Pm(1/2,1/2)(1)Pm(1/2,1/2)(x)(1x)ddx{Pm(1/2,1/2)(x)}]0.\begin{split}&P_{n}^{(\alpha,-1/2)}(1)-P_{n}^{(\alpha,-1/2)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{n}^{(\alpha,-1/2)}(x)\big\}\\ &=\sum_{m=0}^{n}c_{m,n}\,\Big[P_{m}^{(1/2,-1/2)}(1)-P_{m}^{(1/2,-1/2)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{m}^{(1/2,-1/2)}(x)\big\}\Big]\geq 0\,.\end{split}

For n3n\geq 3 the equality is attained only at x=1x=1 since all summands with indices m3m\geq 3 in the above sum are positive at x=0x=0, and so is the sum.

Theorem 1(ii) follows from Theorem B and Lemma 3 in a similar manner. For γ12\gamma\geq\frac{1}{2} and x[0,1]x\in[0,1] Lemma 3(ii) and Theorem B imply

Pn(γ,γ)(1)Pn(γ,γ)(x)(1x)ddx{Pn(γ,γ)(x)}=m=0ncm,n[Pm(1/2,1/2)(1)Pm(1/2,1/2)(x)(1x)ddx{Pm(1/2,1/2)(x)}]0.\begin{split}P_{n}^{(\gamma,\gamma)}(1)&-P_{n}^{(\gamma,\gamma)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{n}^{(\gamma,\gamma)}(x)\big\}\\ &=\sum_{m=0}^{n}c_{m,n}\,\Big[P_{m}^{(1/2,1/2)}(1)-P_{m}^{(1/2,1/2)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{m}^{(1/2,1/2)}(x)\big\}\Big]\\ &\geq 0\,.\end{split}

If n3n\geq 3, then the equality is attained only at x=1x=1. Indeed, on account of Theorem B and Lemma 3(ii), depending on whether nn is odd or even, the summand corresponding to m=3m=3 or m=4m=4 in the last sum is positive at x=0x=0. Therefore, the sum is positive at x=0x=0, too.

Invoking again Lemma 3(i), we conclude from the case just considered that if α>β12\alpha>\beta\geq\frac{1}{2} and x[0,1]x\in[0,1], then

Pn(α,β)(1)Pn(α,β)(x)(1x)ddx{Pn(α,β)(x)}=m=0ncm,n[Pm(β,β)(1)Pm(β,β)(x)(1x)ddx{Pm(β,β)(x)}]0\begin{split}P_{n}^{(\alpha,\beta)}(1)-&P_{n}^{(\alpha,\beta)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{n}^{(\alpha,\beta)}(x)\big\}\\ &=\sum_{m=0}^{n}c_{m,n}\,\Big[P_{m}^{(\beta,\beta)}(1)-P_{m}^{(\beta,\beta)}(x)-(1-x)\,\frac{d}{dx}\big\{P_{m}^{(\beta,\beta)}(x)\big\}\Big]\geq 0\end{split}

and the equality is attained only for x=1x=1. Theorem 1 is proved.

2.3. Proof of Corollary 2

We show that Corollary 2 is an alternative representation of the inequality

(14) Wn(1)Wn(x)(1x)Wn(x)0,x[0,1].W_{n}(1)-W_{n}(x)-(1-x)W_{n}^{\prime}(x)\geq 0,\quad x\in[0,1]\,.

From (12) we obtain

Wn(x)=1sinθddθ{sin(n+12)θsinθ2}=1sinθsin2θ2[12sin(n+12)θcosθ2(n+12)sinθ2cos(n+12)θ]=12sinθsin2θ2[12[sin(n+1)θ+sin(nθ)](n+12)[sin(n+1)θsin(nθ)]]=12sin2θ2[(n+1)sin(nθ)sinθnsin(n+1)θsinθ].\begin{split}W_{n}^{\prime}(x)&=\frac{-1}{\sin\theta}\,\frac{d}{d\theta}\Big\{\frac{\sin\big(n+\frac{1}{2}\big)\theta}{\sin\frac{\theta}{2}}\Big\}\\ &=\frac{1}{\sin\theta\sin^{2}\frac{\theta}{2}}\Big[\frac{1}{2}\sin\Big(n+\frac{1}{2}\Big)\theta\cos\frac{\theta}{2}-\Big(n+\frac{1}{2}\Big)\sin\frac{\theta}{2}\cos\Big(n+\frac{1}{2}\Big)\theta\Big]\\ &=\frac{1}{2\sin\theta\sin^{2}\frac{\theta}{2}}\Big[\frac{1}{2}\big[\sin(n+1)\theta+\sin(n\theta)\big]-\Big(n+\frac{1}{2}\Big)\big[\sin(n+1)\theta-\sin(n\theta)\big]\Big]\\ &=\frac{1}{2\sin^{2}\frac{\theta}{2}}\,\Big[(n+1)\,\frac{\sin(n\theta)}{\sin\theta}-n\,\frac{\sin(n+1)\theta}{\sin\theta}\Big]\,.\end{split}

Consequently,

(1x)Wn(x)=(n+1)sin(nθ)sinθnsin(n+1)θsinθ.(1-x)\,W_{n}^{\prime}(x)=(n+1)\,\frac{\sin(n\theta)}{\sin\theta}-n\,\frac{\sin(n+1)\theta}{\sin\theta}\,.

By replacing W(x)W(x) from (12) and plugging the above expression in (14), we arrive at the equivalent trigonometric inequality

(15) 2n+1(n+2)sinnθsinθ+(n1)sin(n+1)θsinθ0,θ[0,π2],2n+1-(n+2)\,\frac{\sin n\theta}{\sin\theta}+(n-1)\,\frac{\sin(n+1)\theta}{\sin\theta}\geq 0\,,\quad\theta\in\big[0,\frac{\pi}{2}\big]\,,

with the equality sign occurring only for θ=π/2\theta=\pi/2. The proof of Corollary 2 is complete. Let us point out to the equivalent to Corollary 2 statement given in terms of the Chebyshev polynomials of the second kind:

Corollary 4.

For every n,n2n\in\mathbb{N},\;n\geq 2, the following inequality holds true:

(n+2)[Un1(1)Un1(x)](n1)[Un(1)Un(x)]0,x[0,1].(n+2)\big[U_{n-1}(1)-U_{n-1}(x)\big]-(n-1)\big[U_{n}(1)-U_{n}(x)\big]\geq 0,\quad x\in[0,1]\,.

The equality is attained only for x=1x=1.

We conclude this note with the following comment.

Remark 1.

A challenging task is to find domain 𝒟2\mathcal{D}\subset\mathbb{R}^{2} formed by the pairs (α,β)(\alpha,\beta) such that for all n,n3\,n\in\mathbb{N},\;n\geq 3 the inequality

Pn(α,β)(1)Pn(α,β)(x)(1x)ddx{Pn(α,β)(x)},x[0,1],P_{n}^{(\alpha,\beta)}(1)-P_{n}^{(\alpha,\beta)}(x)\geq(1-x)\,\frac{d}{dx}\big\{P_{n}^{(\alpha,\beta)}(x)\big\}\,,\quad x\in[0,1]\,,

holds true. Some experiments carried out with Wolfram Mathematica indicate that the region provided by Theorem 1 is a proper subset of 𝒟\mathcal{D}, e.g., if, say, β>2\beta>2, then there are pairs (α,β)𝒟(\alpha,\beta)\in\mathcal{D} with α<β\alpha<\beta.

3. Acknowledgment

This research is partially supported by the Bulgarian National Research Fund under Contract KP-06-N62/4

References

  • [1] R. Askey: Orthogonal Polynomials and Special Functions, SIAM, Philadelphia, (1975).
  • [2] R. Askey, G. Gasper: Positive Jacobi polynomials sums II, Amer. J. math 98 (1976), 709–737.
  • [3] G. Nikolov: Some inequalities for Chebyshev polynomials, In Constructive Theory of Functions. Sozopol 2019, (Edited by B. Draganov, K. Ivanov, G. Nikolov and R. Uluchev), pp. 181–193, Prof. Marin Drinov Academic Publishing House, Sofia, (2020).
  • [4] T. Rivlin: Chebyshev Polynomials. From Approximation Theory to Algebra & Number Theory, Second edition, Dover Publications, Inc., Mineola, New York (2020).
  • [5] M. S. Robertson: The coefficients of univalent functions, Bul. Amer. Math. Soc. 51 (1945), 733–738.
  • [6] G. Szegő: Orthogonal Polynomials, 4th edition, AMS Colloquium Publications, Providence, RI (1975).