Skip to main content
Springer logoLink to Springer
. 2018 Oct 5;2018(1):272. doi: 10.1186/s13660-018-1870-0

Windschitl type approximation formulas for the gamma function

Zhen-Hang Yang 1,2, Jing-Feng Tian 1,
PMCID: PMC6182422  PMID: 30363740

Abstract

In this paper, we present four new Windschitl type approximation formulas for the gamma function. By some unique ideas and techniques, we prove that four functions combined with the gamma function and Windschitl type approximation formulas have good properties, such as monotonicity and convexity. These not only yield some new inequalities for the gamma and factorial functions, but also provide a new proof of known inequalities and strengthen known results.

Keywords: Gamma function, Windschitl type approximation formula, Monotonicity, Convexity, Inequality

Introduction

For x>0, the classical Euler’s gamma function Γ and psi (digamma) function ψ are defined by

Γ(x)=0tx1etdtandψ(x)=Γ(x)Γ(x), 1.1

respectively. The derivatives ψ, ψ, ψ, are known as polygamma functions. The gamma function has various important applications in many branches of science. For this reason, scholars strive to find various better approximations for the factorial or gamma function by using different ideas and techniques, for instance, Ramanujan [1, p. 339], Burnside [2], Gosper [3], Alzer [4], Shi et al. [5], Batir [6, 7], Mortici [812], Nemes [13, Corollary 4.1], [14], Qi et al. [15, 16], Feng and Wang [17], Chen [1821], Yang et al. [2225], Lu et al. [2628], Xu et al. [29]. Some properties of the remainders of certain approximations for the gamma function can be found in [4, 16, 23, 3035].

In this paper, we are interested in Windschitl’s approximation formula (see [36]) given by

Γ(x+1)W0(x)=2πx(xe)x(xsinh1x)x/2,as x. 1.2

As shown in [21, Eq. (3.18)], the rate of Windschitl’s approximation W0(x) converging to Γ(x+1) is like x5 as x, and like x7 if one replaces W0(x) with

W1(x)=2πx(xe)x(xsinh1x+1810x6)x/2 1.3

by an easy check. These show that W0(x) and W1(x) are more accurate approximations for the gamma function. In 2009, Alzer [37] proved that for all x>0,

2πx(xe)x(xsinh1x)x/2(1+αx5)<Γ(x+1)<2πx(xe)x(xsinh1x)x/2(1+βx5) 1.4

with the best possible constants α=0 and β=1/1620. Recently, Lu, Song and Ma [27] extended Windschitl’s formula to an asymptotic expansion:

Γ(n+1)2πn(ne)n[nsinh(1n+a7n7+a9n9+a11n11+)]n/2 1.5

as n with a7=1/810,a9=67/42525,a11=19/8505,, and proved that there exists an m such that, for every x>m, the double inequality

[xsinh(1x+1810x76742525x9)]x/2<Γ(x+1)2πx(x/e)x<[xsinh(1x+1810x7)]x/2 1.6

holds. An explicit formula for determining the coefficients of nk (nN) was given in [19, Theorem 1] by Chen. Another asymptotic expansion

Γ(x+1)2πx(xe)x(xsinh1x)x/2+j=0rjxj,as x, 1.7

was presented in the same paper [19, Theorem 2].

Let us consider the four new Windschitl type approximation formulas, as x, which are

Γ(x+1)2πx(xe)x(xsinh1x)x/2exp(11620x5):=W01(x), 1.8
Γ(x+1)2πx(xe)x(xsinh1x)x/2exp(11620x51118,900x7):=W02(x), 1.9
Γ(x+1)2πx(xe)x(xsinh1x)x/2(1+11620x5):=W01(x), 1.10
Γ(x+1)2πx(xe)x(xsinh1x)x/2(1+11620x51118,900x7)=W02(x). 1.11

The aim of this paper is, by investigating the monotonicity and convexity of the functions

xlnΓ(x+1)lnF(x),where F=W01,W02,W01,W02,

to establish some new sharp inequalities between the gamma function Γ(x+1) and Windschitl’s approximation formula W0(x). As a by-product, a concise proof of Alzer inequalities (1.4) is presented, and a strengthening for Lu et al.’s inequalities (1.6) is given.

The rest of this paper is organized as follows. In Sect. 2, three lemmas are given, which are crucial to the proofs of our results. In Sect. 3, five monotonicity and convexity results for the functions constructed from the gamma function and Windschilt’s formula are proved. Some new inequalities between the gamma or factorial functions with Windschilt’s formula are established in Sect. 4. In Sect. 5, numeric comparisons of several better approximation formulas are presented.

Lemmas

To prove our results, we need three lemmas as follows.

Lemma 1

The inequalities

xx2+7184x4+1314x2+27560<ψ(x+12), 2.1
xx4+22766x2+42372640x6+15544x4+329176x2+3754928<ψ(x+12)<1xx4+6736x2+256945x4+3518x2+4071008 2.2

hold for x>0.

Proof

The inequality (2.1) was proved in [38, Remark 2.2].

Let

g1(x)=ψ(x+12)1xx4+6736x2+256945x4+3518x2+4071008,g2(x)=ψ(x+12)xx4+22766x2+42372640x6+15544x4+329176x2+3754928.

Then we have

g1(x+1)g1(x)=ψ(x+32)1x+1(x+1)4+6736(x+1)2+256945(x+1)4+3518(x+1)2+4071008ψ(x+12)+1xx4+6736x2+256945x4+3518x2+4071008=921,600×[x(2x+1)2(x+1)(1008x4+1960x2+407)]1×(1008x4+4032x3+8008x2+7952x+3375)1>0.

Hence, we conclude that

g1(x)<g1(x+1)<<limng1(x+n)=0,

which proves the first inequality of (2.2).

Analogously, we have

g2(x+1)g2(x)=ψ(x+32)(x+1)((x+1)4+22766(x+1)2+42372640)(x+1)6+15544(x+1)4+329176(x+1)2+3754928ψ(x+12)+x(x4+22766x2+42372640)x6+15544x4+329176x2+3754928=58,982,400×[(2x+1)2(4928x6+17,360x4+9212x2+375)]1×(4928x6+29,568x5+91,280x4+168,000x3+187,292x2+117,432x+31,875)1<0.

It then follows that

g2(x)>g2(x+1)>>limng2(x+n)=0,

which proves the second formula of (2.2). This completes the proof. □

Lemma 2

The inequalities

t2sinh2t>113t2+115t42189t6, 2.3
t2sinh2t>113t2+115t42189t6+1675t8210,395t10 2.4

hold for all t>0.

Proof

The inequalities in question are equivalent to

h1(t)=(2189t6115t4+13t21)cosh2t12t2+1>0

and

h2(t)=(210,395t101675t8+2189t6115t4+13t21)cosh2t12t2+1>0

for t>0, respectively.

Expanding into a power series yields

h1(t)=(2189t6115t4+13t21)n=022n+1(2n+2)!t2n+1=2189n=3(2t)2n32(2n4)!115n=2(2t)2n8(2n2)!+13n=1(2t)2n2(2n)!2n=0(2t)2n(2n+2)!+1:=11890n=3(n3)×p5(n)(2n+2)!(2t)2n,

where

p5(n)=40n5+60n4122n3543n2296n+1050.

We assert that p5(n)>0 for n3, since p5(n) can be written as

p5(n)=40(n3)5+660(n3)4+4198(n3)3+12,399(n3)2+15,832(n3)+6561,

which is evidently positive for n3. Hence h1(t)>0 for all t>0. While

h2(t)=h1(t)+(210,395t101675t8)n=022n+1(2n+2)!t2n=11890n=3(n3)×p5(n)(2n+2)!(2t)2n+210,395n=5(2t)2n29(2n8)!1675n=4(2t)2n27(2n6)!:=1415,800n=5(n5)×p9(n)(2n+2)!(2t)2n,

where

p9(n)=160n91200n8+2368n71768n6+2354n5+14,845n46403n370,782n257,384n+138,600.

It is easy to check that

p9(n)=160m9+6000m8+98,368m7+921,112m6+5392,514m5+20,270,695m4+48,258,997m3+68,827,423m2+51,883,321m+15,041,130>0,

for m=n50, which proves h2(t)>0 for t>0. The proof is complete. □

The following lemma offers a simple criterion to determine the sign of a class of special polynomials on given interval contained in (0,) without using Descartes’ Rule of Signs, which plays an important role in studying certain special functions, see, for example, [39, 40]. A series version can be found in [41, 42].

Lemma 3

([39, Lemma 7])

Let nN and mN{0} with n>m and let Pn(t) be an nth degree polynomial defined by

Pn(t)=i=m+1naitii=0maiti, 2.5

where an,am>0, ai0 for 0in1 with im. Then there is a unique number tm+1(0,) satisfying Pn(tm+1)=0 such that Pn(t)<0 for t(0,tm+1) and Pn(t)>0 for t(tm+1,).

Consequently, for a given t0>0, if Pn(t0)>0 then Pn(t)>0 for t(t0,) and if Pn(t0)<0 then Pn(t)<0 for t(0,t0).

Monotonicity and convexity

Theorem 1

T he function

f0(x)=lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)

is strictly decreasing and convex on (0,).

Proof

Differentiation yields

f0(x)=ψ(x+1)12ln(xsinh1x)+12xcoth1xlnx12x12,f0′′(x)=ψ(x+1)+12x31sinh2(1/x)32x+12x2.

Replacing x by (x+1/2) in inequality (2.1) leads to

ψ(x+1)>56(2x+1)(21x2+21x+23)35x4+70x3+85x2+50x+12,

and using which to f0′′(x) gives

f0′′(x)>56(2x+1)(21x2+21x+23)35x4+70x3+85x2+50x+12+12x31sinh2(1/x)32x+12x2=f01(1x).

Simplifying yields

f01(t)=12t3sinh2t+16t(36t5+42t480t3220t2210t105)12t4+50t3+85t2+70t+35=t12f02(t)(12t4+50t3+85t2+70t+35)sinh2t,

where

f02(t)=(36t5+42t480t3220t2210t105)cosh2t+(72t6+264t5+468t4+500t3+430t2+210t+105).

Expanding into a power series gives

f02(t)=(36n=222n4(2n4)!t2n+180n=122n2(2n2)!t2n+1210n=022n(2n)!t2n+1)+(42n=222n4(2n4)!t2n220n=122n2(2n2)!t2n105n=022n(2n)!t2n)+(72t6+264t5+468t4+500t3+430t2+210t+105)=n=4(n3)(36n3+19n+70)22n(2n)!t2n+1+n=4(84n4252n3209n2+157n210)22n2(2n)!t2n>0,

where the inequality holds due to

84n4252n3209n2+157n210=84(n4)4+1092(n4)3+4831(n4)2+7893(n4)+2450>0

for n4.

It then follows that f01(t)>0 for t>0, so f0′′(x)>0 for x>0. This yields f0(x)<limxf0(x)=0, which proves the desired result. □

Theorem 2

The function

f1(x)=lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)ln(1+11620x5)

is strictly increasing and concave on (0,).

Proof

Differentiation yields

f1(x)=ψ(x+1)12ln(xsinh1x)+12xcoth1xlnx12x12+5x(1620x5+1),f1(x)=ψ(x+1)+12x31sinh2(1/x)32x+12x259720x5+1x2(1620x5+1)2.

Since limxf1(x)=0, it suffices to prove f1(x)<0 for x>0. Replacing x by (x+1/2) in the right-hand side inequality of (2.2) leads to

ψ(x+1)<1303780x4+7560x3+12,705x2+8925x+3019(2x+1)(63x4+126x3+217x2+154x+60), 3.1

which indicates that

f1(x)<1303780x4+7560x3+12,705x2+8925x+3019(2x+1)(63x4+126x3+217x2+154x+60)+12x31sinh2(1/x)32x+12x259720x5+1x2(1620x5+1)2:=f11(1x),

where

f11(t)=t3cosh2t132t+12t25t7t5+9720(t5+1620)2+130t(3019t4+8925t3+12,705t2+7560t+3780)(t+2)(60t4+154t3+217t2+126t+63).

Using the inequality

cosh2t1>n=1422n(2n)!t2n=2t2+23t4+445t6+2315t8

yields

f11(t)<t32t2+23t4+445t6+2315t832t+12t25t7t5+9720(t5+1620)2+130t(3019t4+8925t3+12,705t2+7560t+3780)(t+2)(60t4+154t3+217t2+126t+63)=130×t9×f12(t)(t5+1620)2(t+2)(t6+14t4+105t2+315)(60t4+154t3+217t2+126t+63)<0

for t>0, where the inequality holds due to

f12(t)=8100t14+39,690t13+193,586t12+645,960t11+2,028,124t10+90,019,275t9+406,666,800t8+1976,029,740t7+6395,589,900t6+20,173,546,260t5+51,035,406,750t4+110,592,337,500t3+184,843,490,400t2+254,068,164,000t+101,627,265,600>0

for t>0. This implies that f2′′(x)<0 for all x>0, and the proof is complete. □

Theorem 3

The function

f1(x)=lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)11620x5

is strictly increasing and concave on (0,).

Proof

We clearly see that

f1(x)=f1(x)+D(11620x5),

where D(y)=ln(1+y)y. By Theorem 2, f1 is strictly increasing and concave on (0,), so if we prove xD(y) is strictly increasing and concave on (0,), then so will be f1, and the proof will be complete. Now we easily check that for x>0,

dD(y)dx=1324x6(1620x5+1)>0,d2D(y)dx2=1542970x5+1x7(1620x5+1)2<0,

which completes the proof. □

Theorem 4

The function

f2(x)=lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)11620x5+1118,900x7

is strictly decreasing and convex on (0,).

Proof

Differentiation yields

f2(x)=ψ(x+1)12ln(xsinh1x)+12xcoth1xf2(x)=lnx12x12+1324x6112700x8,f2′′(x)=ψ(x+1)+12x31sinh2(1/x)32x+12x2154x7+22675x9.

Since limxf2(x)=0, it suffices to prove f2′′(x)>0 for x>0. Replacing x by (x+1/2) in the left-hand side inequality of (2.2) leads to

ψ(x+1)>730(2x+1)(165x4+330x3+815x2+650x+417)77x6+231x5+560x4+735x3+623x2+294x+60,

and applying which to f2′′(x) gives

f2′′(x)>730(2x+1)(165x4+330x3+815x2+650x+417)77x6+231x5+560x4+735x3+623x2+294x+60+12x31sinh2(1/x)32x+12x2154x7+22675x9=f21(1x).

Making a change of variable t=1/x yields

f21(t)=730t(t+2)(417t4+650t3+815t2+330t+165)60t6+294t5+623t4+735t3+560t2+231t+7732t+12t2154t7+22675t9+t2t2sinh2t.

We distinguish two cases to prove f21(t)>0 for all t>0.

Case 1: t1. Application of inequality (2.3) gives

f21(t)>730t(t+2)(417t4+650t3+815t2+330t+165)60t6+294t5+623t4+735t3+560t2+231t+7732t+12t2154t7+22675t9+t2(113t2+115t42189t6)
=19450t9×p6(t)60t6+294t5+623t4+735t3+560t2+231t+77,

where

p6(t)=18,480t6+90,552t5+178,384t4+160,230t3+51,205t21617t539.

Clearly, p6(t)>0 for t1, so f21(t)>0 for t1.

Case 2: 0<t<1. Using inequality (2.4) yields

f21(t)>730t(t+2)(417t4+650t3+815t2+330t+165)60t6+294t5+623t4+735t3+560t2+231t+7732t+12t2154t7+22675t9+t2(113t2+115t42189t6+1675t8210,395t10)
=t11×q6(t)60t6+294t5+623t4+735t3+560t2+231t+77,

where

q6(t)=4693t614495t5+28811485t4+4816495t3+400,91920,790t2+157390t+1573270.

Since the coefficients of polynomial q6(t) satisfy the conditions of Lemma 3 and q6(1)=53,681/990>0, we find that q6(t)>0 for t(0,1), and then f21(t)>0 for t(0,1).

This ends the proof. □

Theorem 5

The function

f2(x)=lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)ln(1+11620x51118,900x7)

is strictly decreasing and convex on [4/3,).

Proof

We easily see that

f2(x)=f2(x)D(11620x51118,900x7),

where D(y)=ln(1+y)y. By Theorem 4, f2 is strictly decreasing and convex on (0,), so if we prove xD(y) is strictly increasing and concave on [4/3,), then so will be f2, and the proof will be complete. Now we easily check that for x4/3,

dD(y)dx=18100(25x233)(35x233)x8(35x2+56,700x733)>0,d2D(y)dx2=p11(x)x9(56,700x7+35x233)2<0,

where the last inequality holds due to

p11(x)=67,375x11180,180x9+114,345x7+122554x6223327x4+859190x2266275=385x7[(175(x2169)2+13889(x2169)+146581)]+11350x2(175x2319)2+242675[56(x2169)+59]>0,

which completes the proof. □

Inequalities

As is well known, analytic inequalities [4345] play a very important role in different branches of modern mathematics. Using the theorems presented in the previous section, we can obtain some new inequalities for the gamma function and factorial function related to Windschitl’s formula.

Corollary 1

Let W0(x) be defined by (1.2). Then the inequalities

max(1,exp(11620x51118,900x7))<Γ(x+1)W0(x)<1+11620x5<exp(11620x5) 4.1

hold for all x>0. If x33/35, then we have

1+11620x51118,900x7<exp(11620x51118,900x7)<Γ(x+1)W0(x)<1+11620x5<exp(11620x5). 4.2

Proof

The first and second inequalities in (4.1) follow directly from the monotonicity of f0, f2 and f1 on (0,) given in Theorems 1, 4 and 2, respectively, due to f0()= f2()=f1()=0. The third one holds due to a simple inequality 1+y<ey for y>0. The proof of inequalities (4.2) is similar, which completes the proof. □

Using the monotonicity of f0, f1 and f2 on (0,) and noting that

f0(1)=lne2πsinh1,f1(1)=ln(16201621e2πsinh1),f2(1)=lne28,349/28,3502πsinh1,

we immediately get the following corollary.

Corollary 2

For nN, the inequalities

1<n!2πn(ne)n(nsinh1n)n/2α0,β1(1+11620n5)n!2πn(ne)n(nsinh1n)n/2<1+11620n5,
exp(11620n51118,900n7)<n!2πn(ne)n(nsinh1n)n/2α2exp(11620n51118,900n7)

hold with the best constants α0=e/2πsinh11.000,34, β1=1620e/(16212πsinh1)0.999,72 and α2=e28,349/28,350/2πsinh11.000,30.

The proof of inequalities (1.4) presented by Alzer [37] seems to be somewhat complicated. With the aid of the first and second inequalities in (4.1), we can give a new and simpler proof.

Proof of inequalities (1.4)

The sufficiency for the inequalities (1.4) to hold for x>0 follows by the first and second inequalities in (4.1). The necessary condition for the left-hand side inequality of (1.4) to hold for x>0 follows from the following relation:

limx0+lnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)ln(1+αx5)ln(1/x)={12if α=0,92if α0.

While the necessary condition for the right-hand side of (1.4) to hold for x>0 follows from the limit relation

limxlnΓ(x+1)ln2π(x+12)lnx+xx2ln(xsinh1x)ln(1+βx5)x5=11620β0.

This completes the proof. □

The following corollary offers a strengthening for Lu et al.’s inequalities (1.6).

Corollary 3

The inequalities

[xsinh(1x+1810x76742,525x9)]x/2<(xsinh1x)x/2exp(11620x51118,900x7)<Γ(x+1)2πx(x/e)x<(xsinh1x)x/2exp(11620x5)<[xsinh(1x+1810x7)]x/2 4.3

hold for x>c, where c=0 for the second, third and fourth inequalities, while c=x00.43738 for the first one, here x0 is the unique solution of the equation

1x+1810x76742,525x9=0

on (0,).

Proof

Clearly, the second and third inequalities of (4.3) follow by the first and second inequalities in (4.1). It remains to prove the first and last inequalities of (4.3).

(i) The last one is equivalent to

x2ln[xsinh(1x+1810x7)]>x2ln(xsinh1x)+11620x5,

or equivalently,

h3(t)=lnsinh(t+1810t7)lnsinht1810t6>0

for t=1/x>0. Denote by l(t)=lnsinht and t2=(t+t7/810). Then by Taylor formula we have

h3(t)=l(t2)l(t)1810t6=(t2t)l(t)+l(t)2!(t2t)2+l(ξ)3!(t2t)31810t6,

where t<ξ<t+t7/810. Since l(t)=2(cosht)/sinh3t>0, we get

h3(t)>1810t7coshtsinhtt142×81021sinh2t1810t6:=t6×h31(t)2×8102sinh2t,

where

h31(t)=810tsinh2t810cosh2t+810t8.

Due to

h31(t)=540t4+144t6+1017t8+810n=5(n1)22n(2n)!t2n>0,

we conclude that h3(t)>0 for t>0.

(ii) To ensure that the first inequality holds, it is necessary to establish

xsinh(1x+1810x76742,525x9)>0

for x>0, for which it suffices so show that

1x+1810x76742,525x9=85,050x8+105x213485,050x9>0.

By Lemma 3, the numerator in the above fraction, as an 8th degree polynomial, has a unique zero x0 on (0,). Numeric computation gives x00.437,38.

Now the first inequality is equivalent to

x2ln[xsinh(1x+1810x76742,525x9)]<x2ln(xsinh1x)+11620x51118,900x7,

or equivalently,

h4(t)=ln[sinh(t+1810t76742,525t9)]ln(sinht)1810t6+119450t8<0

for t=1/x(0,1/x0), where 1/x02.28632 is clearly the unique zero of the polynomial

t1t1(t)=t+1810t76742,525t9

on (0,). In view of l(t)=1/sinh2t<0, we have

h4(t)=l(t1)l(t)1810t6+119450t8<(t1t)l(t)1810t6+119450t8=(1810t76742,525t9)coshtsinht1810t6+119450t8=185,050t6sinht(105sinht105tcosht+134t3cosht99t2sinht)=185,050t6sinhtn=34(268n99)(n1)(n2)(2n1)!t2n1<0,

which completes the proof. □

Remark 1

Clearly, the proof of Corollary 3 can also be regarded as a new proof of Lu et al.’s inequalities (1.6). Moreover, our proof gives the minimum value of m, i.e., min(m)=x00.437,38, such that the the double inequality (1.6) holds for all x>x0.

Numeric comparisons

By the asymptotic expansion listed in [46, Eq. (6.1.40)]

lnΓ(x+1)ln2π+(x+12)lnxx+n=1B2n2n(2n1)x2n1, 5.1

we easily verify that our four approximation formulas W01(x), W01(x), W02(x) and W02(x), defined by (1.8), (1.10), (1.9) and (1.11), respectively, have the following limit relations:

limxlnΓ(x+1)lnW01(x)x7=limxlnΓ(x+1)lnW01(x)x7=198340,200,limxlnΓ(x+1)lnW02(x)x9=limxlnΓ(x+1)lnW02(x)x9=143170,100.

Also, for another approximation formula W1(x) defined by (1.3), we have

limxlnΓ(x+1)lnW1(x)x7=163340,200.

Denote the two approximation formulas generated by the double inequality (1.6) by

WL1(x)=2πx(xe)x[xsinh(1x+1810x7)]x/2, 5.2
WL2(x)=2πx(xe)x[xsinh(1x+1810x76742,525x9)]x/2. 5.3

We have

limxlnΓ(x+1)lnWL1(x)x7=6785,050,limxlnΓ(x+1)lnWL2(x)x9=1917,010.

These, in combination with Corollaries 2, 1 and 3, show that the approximation formula W02(x) given by (1.9) is the best among those listed above, which can be seen from comparison Table 1.

Table 1.

Comparison among W02 (1.9), WL2 (5.3), W1 (1.3), W01 (1.10) and WL1 (5.2)

n |W02(n)n!n!| |WL2(n)n!n!| |W1(x)n!n!| |W01(x)n!n!| |WL1(x)n!n!|
1 3.065 × 10−4 5.655 × 10−4 1.832 × 10−4 2.754 × 10−4 4.686 × 10−4
2 1.098 × 10−6 1.629 × 10−6 2.668 × 10−6 3.449 × 10−6 5.030 × 10−6
5 3.956 × 10−10 5.367 × 10−10 5.743 × 10−9 7.054 × 10−9 9.681 × 10−9
10 8.221 × 10−13 1.098 × 10−12 4.710 × 10−11 5.738 × 10−11 7.794 × 10−11
20 1.630 × 10−15 2.172 × 10−15 3.727 × 10−13 4.531 × 10−13 6.138 × 10−13
50 4.300 × 10−19 5.715 × 10−19 6.129 × 10−16 7.445 × 10−16 1.008 × 10−15
100 8.404 × 10−22 1.117 × 10−21 4.791 × 10−18 5.819 × 10−18 7.877 × 10−18

Conclusion

In this paper, we provide four Windschitl type approximation formulas for the gamma function, and prove that those functions, involving the gamma function and Windschitl type functions, have good properties, including monotonicity and convexity. From these facts we obtain some new sharp Windschitl type bounds for the gamma and factorial functions. These sharp inequalities, together with numerical comparisons, illustrate that W02(x) defined by (1.9) is the best approximation formula among those mentioned in Sect. 5.

Moreover, we give a simple proof of Alzer’s inequalities (1.4), and improve and strengthen Lu et al.’s inequalities (1.6).

It is worth mentioning that our proofs of Theorems 15 are subtle and interesting, since the approximations deal with the gamma and hyperbolic sine functions, and it is difficult to establish their monotonicity and convexity by usual methods. Evidently, Lemmas 2 and 3 play important roles.

Acknowledgements

The authors would like to express their sincere thanks to the anonymous referees for their great efforts to improve this paper.

Authors’ contributions

All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.

Funding

This work was supported by the Fundamental Research Funds for the Central Universities (No. 2015ZD29) and the Higher School Science Research Funds of Hebei Province of China (No. Z2015137).

Competing interests

The authors declare that they have no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Zhen-Hang Yang, Email: yzhkm@163.com.

Jing-Feng Tian, Email: tianjf@ncepu.edu.cn.

References

  • 1.Ramanujan S. The Lost Notebook and Other Unpublished Papers. Berlin: Springer; 1988. [Google Scholar]
  • 2.Burnside W. A rapidly convergent series for logN! Messenger Math. 1917;46:157–159. [Google Scholar]
  • 3.Gosper R.W. Decision procedure for indefinite hypergeometric summation. Proc. Natl. Acad. Sci. USA. 1978;75:40–42. doi: 10.1073/pnas.75.1.40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Alzer H. On some inequalities for the gamma and psi functions. Math. Comput. 1997;66(217):373–389. doi: 10.1090/S0025-5718-97-00807-7. [DOI] [Google Scholar]
  • 5.Shi X., Liu F., Hu M. A new asymptotic series for the Gamma function. J. Comput. Appl. Math. 2006;195:134–154. doi: 10.1016/j.cam.2005.03.081. [DOI] [Google Scholar]
  • 6.Batir N. Inequalities for the gamma function. Arch. Math. 2008;91:554–563. doi: 10.1007/s00013-008-2856-9. [DOI] [Google Scholar]
  • 7.Batir N. Bounds for the Gamma function. Results Math. 2017;72:865–874. doi: 10.1007/s00025-017-0698-0. [DOI] [Google Scholar]
  • 8.Mortici C. A new Stirling series as continued fraction. Numer. Algorithms. 2011;56(1):17–26. doi: 10.1007/s11075-010-9370-4. [DOI] [Google Scholar]
  • 9.Mortici C. Improved asymptotic formulas for the gamma function. Comput. Math. Appl. 2011;61:3364–3369. doi: 10.1016/j.camwa.2011.04.036. [DOI] [Google Scholar]
  • 10.Mortici C. Further improvements of some double inequalities for bounding the gamma function. Math. Comput. Model. 2013;57:1360–1363. doi: 10.1016/j.mcm.2012.11.020. [DOI] [Google Scholar]
  • 11.Mortici C. A continued fraction approximation of the gamma function. J. Math. Anal. Appl. 2013;402:405–410. doi: 10.1016/j.jmaa.2012.11.023. [DOI] [Google Scholar]
  • 12.Mortici C. A new fast asymptotic series for the gamma function. Ramanujan J. 2015;38:549–559. doi: 10.1007/s11139-014-9589-0. [DOI] [Google Scholar]
  • 13.Nemes G. New asymptotic expansion for the Gamma function. Arch. Math. (Basel) 2010;95:161–169. doi: 10.1007/s00013-010-0146-9. [DOI] [Google Scholar]
  • 14.Nemes G. More accurate approximations for the gamma function. Thai J. Math. 2011;9(1):21–28. [Google Scholar]
  • 15.Guo B.N., Zhang Y.J., Qi F. Refinements and sharpenings of some double inequalities for bounding the gamma function. J. Inequal. Pure Appl. Math. 2008;9(1):17. [Google Scholar]
  • 16.Qi F. Integral representations and complete monotonicity related to the remainder of Burnside’s formula for the gamma function. J. Comput. Appl. Math. 2014;268:155–167. doi: 10.1016/j.cam.2014.03.004. [DOI] [Google Scholar]
  • 17.Feng L., Wang W. Two families of approximations for the gamma function. Numer. Algorithms. 2013;64:403–416. doi: 10.1007/s11075-012-9671-x. [DOI] [Google Scholar]
  • 18.Chen C.-P. Unified treatment of several asymptotic formulas for the gamma function. Numer. Algorithms. 2013;64:311–319. doi: 10.1007/s11075-012-9667-6. [DOI] [Google Scholar]
  • 19.Chen C.-P. Asymptotic expansions of the gamma function related to Windschitl’s formula. Appl. Math. Comput. 2014;245:174–180. doi: 10.1016/j.amc.2014.07.080. [DOI] [Google Scholar]
  • 20.Chen C.-P., Paris R.B. Inequalities, asymptotic expansions and completely monotonic functions related to the gamma function. Appl. Math. Comput. 2015;250:514–529. doi: 10.1016/j.amc.2014.11.010. [DOI] [Google Scholar]
  • 21.Chen C.-P. A more accurate approximation for the gamma function. J. Number Theory. 2016;164:417–428. doi: 10.1016/j.jnt.2015.11.007. [DOI] [Google Scholar]
  • 22.Yang Z.-H., Chu Y.-M. Asymptotic formulas for gamma function with applications. Appl. Math. Comput. 2015;270:665–680. doi: 10.1016/j.amc.2015.08.087. [DOI] [Google Scholar]
  • 23.Yang Z.-H. Approximations for certain hyperbolic functions by partial sums of their Taylor series and completely monotonic functions related to gamma function. J. Math. Anal. Appl. 2016;441:549–564. doi: 10.1016/j.jmaa.2016.04.029. [DOI] [Google Scholar]
  • 24.Yang Z.-H., Tian J.-F. Monotonicity and inequalities for the gamma function. J. Inequal. Appl. 2017;2017:317. doi: 10.1186/s13660-017-1591-9. [DOI] [Google Scholar]
  • 25.Yang Z.-H., Tian J.-F. An accurate approximation formula for gamma function. J. Inequal. Appl. 2018;2018:56. doi: 10.1186/s13660-018-1646-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lu D. A new sharp approximation for the Gamma function related to Burnside’s formula. Ramanujan J. 2014;35(1):121–129. doi: 10.1007/s11139-013-9534-7. [DOI] [Google Scholar]
  • 27.Lu D., Song L., Ma C. A generated approximation of the gamma function related to Windschitl’s formula. J. Number Theory. 2014;140:215–225. doi: 10.1016/j.jnt.2014.01.023. [DOI] [Google Scholar]
  • 28.Lu D., Song L., Ma C. Some new asymptotic approximations of the gamma function based on Nemes’ formula, Ramanujan’s formula and Burnside’s formula. Appl. Math. Comput. 2015;253:1–7. doi: 10.1016/j.amc.2014.12.077. [DOI] [Google Scholar]
  • 29.Xu A., Hu Y., Tang P. Asymptotic expansions for the gamma function. J. Number Theory. 2016;169:134–143. doi: 10.1016/j.jnt.2016.05.020. [DOI] [Google Scholar]
  • 30.Qi F., Niu D.-W., Guo B.-N. Monotonic properties of differences for remainders of psi function. Int. J. Pure Appl. Math. Sci. 2007;4(1):59–66. [Google Scholar]
  • 31.Guo S., Qi F. A class of completely monotonic functions related to the remainder of Binet’s formula with applications. Tamsui Oxf. J. Inf. Math. Sci. 2009;25(1):9–14. [Google Scholar]
  • 32.Shi X.-Q., Liu F.-S., Qu H.-M. The Burnside approximation of gamma function. Anal. Appl. 2010;8(3):315–322. doi: 10.1142/S0219530510001643. [DOI] [Google Scholar]
  • 33.Qi F., Guo B.-N. Some properties of extended remainder of Binet’s first formula for logarithm of gamma function. Math. Slovaca. 2010;60(4):461–470. doi: 10.2478/s12175-010-0025-7. [DOI] [Google Scholar]
  • 34.Qi F., Guo B.-N. Integral representations and complete monotonicity of remainders of the Binet and Stirling formulas for the gamma function. Rev. R. Acad. Cienc. Exactas Fís. Nat., Ser. A Mat. 2017;111(2):425–434. doi: 10.1007/s13398-016-0302-6. [DOI] [Google Scholar]
  • 35.Yang Z.-H., Tian J.-F. Two asymptotic expansions for gamma function related to Windschitl’s formula. Open Math. 2018;16:1048–1060. doi: 10.1515/math-2018-0088. [DOI] [Google Scholar]
  • 36. Smith, W.D.: The gamma function revisited. http://schule.bayernport.com/gamma/gamma05.pdf (2006)
  • 37.Alzer H. Sharp upper and lower bounds for the gamma function. Proc. R. Soc. Edinb. 2009;139A:709–718. doi: 10.1017/S0308210508000644. [DOI] [Google Scholar]
  • 38.Yang Z.-H., Chu Y.-M., Zhang X.-H. Sharp bounds for psi function. Appl. Math. Comput. 2015;268:1055–1063. doi: 10.1016/j.amc.2015.07.012. [DOI] [Google Scholar]
  • 39.Yang Z.-H., Chu Y.-M., Tao X.-J. A double inequality for the trigamma function and its applications. Abstr. Appl. Anal. 2014;2014:702718. doi: 10.1155/2014/702718. [DOI] [Google Scholar]
  • 40.Yang Z.-H., Tian J. Monotonicity and sharp inequalities related to gamma function. J. Math. Inequal. 2018;12(1):1–22. doi: 10.7153/jmi-2018-12-01. [DOI] [Google Scholar]
  • 41. Yang, Z.-H., Tian, J.: Convexity and monotonicity for the elliptic integrals of the first kind and applications. arXiv:1705.05703 [math.CA]
  • 42.Yang Z.-H., Qian W.-M., Chu Y.-M., Zhang W. On approximating the arithmetic-geometric mean and complete elliptic integral of the first kind. J. Math. Anal. Appl. 2018;462:1714–1726. doi: 10.1016/j.jmaa.2018.03.005. [DOI] [Google Scholar]
  • 43.Tian J., Wang W., Cheung W.-S. Periodic boundary value problems for first-order impulsive difference equations with time delay. Adv. Differ. Equ. 2018;2018:79. doi: 10.1186/s13662-018-1539-5. [DOI] [Google Scholar]
  • 44.Tian J.-F. Triple Diamond-Alpha integral and Hölder-type inequalities. J. Inequal. Appl. 2018;2018:111. doi: 10.1186/s13660-018-1704-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Tian J.-F., Ha M.-H. Extensions of Hölder’s inequality via pseudo-integral. Math. Probl. Eng. 2018;2018:4080619. doi: 10.1155/2018/4080619. [DOI] [Google Scholar]
  • 46.Abramowttz M., Stegun I.A. Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables. New York: Dover; 1972. [Google Scholar]

Articles from Journal of Inequalities and Applications are provided here courtesy of Springer

RESOURCES