Loading [MathJax]/jax/element/mml/optable/BasicLatin.js

Sunday, October 31, 2021

Euler's work VII

The complete Beta function

Series expansion of the complete beta function


We show the proof of the following series posted by @infseriesbot 1x+x1!(x+1)+x(x+1)2!(x+2)+...=πsin(πx)

Proof:

This series is just a special case of the series expansion of the complete beta function: B(x,y)=j=0(1y)jj!(x+j) To prove this, consider the definition of the beta function: B(x,y)=10tx1(1t)y1dt and recall the series expansion of (1t)y1 (1t)y1=j=0(1)j(y1j)tj Hence 10tx1(1x)y1dt=10tx1j=0(1)j(y1j)tjdt=j=0(1)j(y1j)0tx+j1dt=j=0(1)j(y1j)1(x+j) Recall the reflection formula for the upper indice (vm)=(1)m(v+m1m) and the fact that the Pochhammer polynomials can be expressed as the quotient of two gamma functions: (x)n=Γ(n+x)Γ(x) Then we have j=0(1)j(y1j)1(x+j)=j=0(jyj)1(x+j)( from (1))=j=0(jy)!j!(y)!1(x+j)=j=0Γ(jy+1)j!Γ(1y)1(x+j)=j=0(1y)jj!(x+j)( from (2)) Therefore B(x,y)=j=0(1y)jj!(x+j) Now, if we put y=1x we have: j=0(x)jj!(x+j)=B(x,1x)=Γ(x)Γ(1x)Γ(1) Recall the Euler's reflection formula Γ(x)Γ(1x)=πsin(πx)xN Hence j=0(x)jj!(x+j)=1x+x1!(x+1)+x(x+1)2!(x+2)+...=πsin(πx)

Generalized hypergeometric functions IV

Another nice misterious series

Another nice involving 3F2


We show the proof of this nice series posted by @PDemichel : n=0(4n)!256nn!4=πΓ2(58)Γ2(78)

It turned out that we proved it earlier using different formulas to calculate 3F2.

Proof: In the past we have used the duplication or triplication formula for Pochhamer polyonomials or rising factorials. It turned out that this formulas are particular cases for the general formula (Proof in the Appendix): (x)nm=mnmmj=1(x+j1m)n With this formula, we can expand (4n)! in the following way: (4n)!=Γ(4n+1)Γ(1)=(1)4n=44n(14)n(12)n(34)n(1)n Additionally, we know that (1)n=n!. Hence, n=0(4n)!256nn!4=n=044n(14)n(12)n(34)n(1)n256n(1)n(1)n(1)nn!=n=0(14)n(34)n(12)n(1)n(1)nn!=3F2(14,34,12;1,1;1) Now recall the Watson’s Sum formula for 3F2: 3F2(a,b,c12(a+b+1),2c;1)=Γ(12)Γ(c+12)Γ(12(a+b+1))Γ(c+12(1ab))Γ(12(a+1))Γ(12(b+1))Γ(c+12(1a))Γ(c+12(1b)), Therefore n=0(4n)!256nn!4=3F2(14,34,12;1,1;1)=Γ(12)Γ(1)Γ(1)Γ(12)Γ(12(14+44))Γ(12(34+44))Γ(12+12(34))Γ(12+12(14)) Hence n=0(4n)!256nn!4=πΓ2(58)Γ2(78)

Appendix

We show the proof of this result for Pochhammer polynomials:
(x)nm=mnmmj=1(x+j1m)n Consider: (x)nm=x(x+1)(x+2)(x+nm1) Factorize (x)nm in the following way: (x)nm=A1A2A3An where A1=x(x+1)(x+2)(x+m1)A2=(x+m)(x+m+1)(x+m+2)(x+2m1)A3=(x+2m)(x+2m+1)(x+2m+2)(x+3m1)An=(x+(n1)m)(x+(n1)m+1)(x+(n1)m+2)(x+nm1)
Now factor m out of each factor in each Aj
A1=mmxm(x+1m)(x+2m)(x+m1m)A2=mm(xm+1)(x+1m+1)(x+2m+1)(x+m2m)A3=mm(xm+2)(x+1m+2)(x+2m+2)(x+3m1m)An=mmB0(xm+n1)B1(x+1m+n1)B2(x+2m+n1)B3(x+mn1m)Bm Now it touned out that column B0=mnm and Bj=(x+j1m)n Therefore (x)nm=B0B1B2Bm=mnmmj=1(x+j1m)n

Generalized hypergeometric functions III

Another misterious series

Series involving 2F1


Today we show the proof of this nice series posted by @infseriesbot: n=0(3n)!54nn!3=πΓ(23)Γ(56)

Proof

First recall that the Pochhammer polyonomial or rising factorial can be expressed as a quiotinent of two gamma functions: (x)n=x(x+1)(x+2)(x+n1)=Γ(n+x)Γ(x) Hence (3n)!=Γ(3n+1)Γ(1)=(1)3n Now we will use the triplication formula for Pochhamer polynomials: (x)3n=33n(x3)n(x+13)n(x+23)n Is easy to show the proof of this: By definition (x)3n=x(x+1)(x+2)(x+3n1) If we factorize in this form: (x)3n=[x(x+3)(x+6)...(x+3n3)]nterms[(x+1)(x+4)(x+7)...(x+3n2)]nterms[(x+2)(x+5)(x+8)...(x+3n1)]nterms Consider A=x(x+3)(x+6)(x+3n3)=3n(x3)(x3+1)(x3+n1)=3n(x3)n B=(x+1)(x+4)(x+7)(x+3n2)=3n(x+13)(x+13+1)(x+13+n1)=3n(x+13)n C=(x+2)(x+5)(x+8)(x+3n1)=3n(x+23)(x+23+1)(x+23+n1)=3n(x+23)n Hence (x)3n=ABC=33n(x3)n(x+13)n(x+23)n Therefore (3n)!=(1)3n=33n(13)n(23)n(1)n Now, using the fact that (1)n=n! n=0(3n)!54nn!3=n=033n(13)n(23)n(1)n54n(1)n(1)nn!=n=0(13)n(23)n(1)n(12n)1n!=2F1(13,23;1;12) Now using the following identity for 2F1: 2F1(a,b;12a+12b+12;12)=πΓ(12a+12b+12)Γ(12a+12)Γ(12b+12). we have n=0(3n)!54nn!3=2F1(13,23;1;12)=πΓ(1)Γ(16+12)(26+12)=πΓ(23)Γ(56) n=0(3n)!54nn!3=πΓ(23)Γ(56)

Friday, October 15, 2021

Generalized hypergeometric functions II

Misterious series

Series involving 3F2


Today we show the proof of this nice series posted by @infseriesbot n=0(4n)!28nn!4=πΓ(58)2Γ(78)2 The proof will rely on some properties of the Pochhammer polynomials (rising factorials) and the generalized hypergeometric functions

Proof

Consider the Pochhammer polynomial: (x)n=x(x+1)(x+2)(x+n1)=n1j=0(x+j)n=1,2,3,... also called rising factorial.

Pochhammer polynomials can be expressed as the quotient of two gamma functions: (x)n=Γ(n+x)Γ(x) Therefore (4n)!=Γ(4n+1)=(1)4n Pochhammer polynomials also obey the following duplication formula: (x)2n=4n(x2)n(1+x2)n Therefore (4n)!=(1)4n=42n(12)2n(1)2n=42n[4n(14)n(34)n][4n(12)n(1)n]=44n(14)n(34)n(12)n(1)n From the definition of the Pochhammer polynomials is easy to show that (1)n=n!
Therefore
S=n=0(4n)!28nn!4=n=044n(14)n(34)n(12)n(1)n28n(1)n(1)n(1)n1n!=n=0(14)n(34)n(12)n(1)n(1)n1n!=3F2(14,34,121,1;1) where 3F2 is the Generalized Hypergeometric function

3F2 satisfy the following identity: 3F2(a,b,cd,e;1)=Γ(e)Γ(d+eabc)Γ(ea)Γ(d+ebc)3F2(a,db,dcd,d+ebc;1) whenever Re(d+eabc)>0 and Re(ea)>0

Therefore S=3F2(14,34,121,1;1)=πΓ(34)23F2(14,14,121,34;1) 3F2 also satisfy the Dixon’s Well-Poised Sum: 3F2(a,b,cab+1,ac+1;1)=Γ(12a+1)Γ(ab+1)Γ(ac+1)Γ(12abc+1)Γ(a+1)Γ(12ab+1)Γ(12ac+1)Γ(abc+1) Hence 3F2(14,14,121,34;1)=Γ(98)Γ(34)Γ(38)Γ(54)Γ(78)Γ(58)π=Γ(38)22πΓ(58)2 Therefore S=Γ(38)22Γ(34)2Γ(58)2=πΓ2(58)Γ2(78) Hence, we can conclude n=0(4n)!28nn!4=πΓ(58)2Γ(78)2

Thursday, October 14, 2021

Euler's work VI

Strange series involving the Catalan constant

Series related to the Catalan constant


We show te proof of the following result. n=0(2n)!!(n+1)(2n+5)(2n+1)!!(2n+1)(2n+3)2=1 The proof involves one of the series representation of the Catalan constant.

Proof

First note that (n+1)(2n+5)(2n+1)(2n+3)2=12(2n+1)+12(2n+3)2 Hence S=n=0(2n)!!(n+1)(2n+5)(2n+1)!!(2n+1)(2n+3)2=n=0(2n)!!(2n+1)!![12(2n+1)+12(2n+3)2]=12n=0(2n)!!(2n+1)!!(2n+1)S1+12n=0(2n)!!(2n+1)!!(2n+3)2S2 In a previous post we evaluated S1: S1=n=0(2n)!!(2n+1)!!(2n+1)=2β(2) where β(2) is the Calatan constant.

For S2:

S2=n=0(2n)!!(2n+1)!!(2n+3)2=n=0(2n+2)!!(2n+3)!!(2n+3)(2n+2)=j=1(2j)!!(2j+1)!!(2j+1)(2j)(n=j1) Note that n=1(2j)!!(2j+1)!!(2j+1)2β(2)1=j=1(2j)!!(2j+1)!!2jS3j=1(2j)!!(2j+1)!!(2j+1)(2j)S2 We just have to find S3:

In another post we found the following series expansion: arcsin(x)1x2=j=0(2j)!!(2j+1)!!x2j+1 Dividing by x2 arcsin(x)x21x2=j=0(2j)!!(2j+1)!!x2j1=1x+j=1(2j)!!(2j+1)!!x2j1 j=1(2j)!!(2j+1)!!x2j1=arcsin(x)x21x21x Integrating from 0 to 1: S3=j=1(2j)!!(2j+1)!!2j=10arcsin(x)x21x21xdx I=arcsin(x)x21x21xdx=arcsin(x)x21x2dxln(x)+C arcsin(x)x21x2dx=wsin2(w)dw(warcsin(x))IBP=wcot(w)+cot(w)dw=wcot(w)+ln(sinw)+C=1x2arcsin(x)x+ln(x)+C Therefore I=arcsin(x)x21x21xdx=1x2arcsin(x)x+C Hence S3=10arcsin(x)x21x21xdx=[1x2arcsin(x)x]10=1 Therefore S2=S3β(2)+1=22β(2) S=12S1+12S2=β(2)+1β(2)=1 We can conclude: n=0(2n)!!(n+1)(2n+5)(2n+1)!!(2n+1)(2n+3)2=1

Monday, October 11, 2021

Residue theorem V

Nice series involving hyperbolic functions

Series involving hyperbolic cotangent


We show the proof following result about series involving hyperbolic functions posted by @infseriesbot: πn=1cothπnn4m1=ζ(4m)+2m1n=1(1)k1ζ(2k)ζ(4m2k) The proof will rely on the residue theorem.

Proof

In order to calculate the residue, we will need the expansion in zeta numbers of coth(πx) and cot(x) coth(πx)=1πx2πn=1(1)nζ(2n)x2n1 cot(πx)=1πx2πn=1ζ(2n)x2n1

  1. To obtain these expansions we need the Fourier series expansion of cos(ax)

  2. Since cos(ax) is defined for all x, and represents a continuous, piece-wise smooth, odd function is everywhere equal to its Fourier series, which is absolutely and uniformly convergent.

    Therefore,
    cosax=n=cneinxan where cn=12πππcos(ax)einxdx Hence cn=12πππcos(ax)einxdx=12πππ[eiax+eiax2]einxdx=14πππeix(an)dx+14πππeix(an)dx=14π[ei(an)ei(an)i(an)]+14π[ei(a+n)ei(a+n)i(a+n)]=14π[eaeπineπiaeπini(an)]+14π[eπiaeπin)eπiaeiπni(a+n)]=eiπn(eiπaeiπa)4πi[1an+1a+n]=(1)nsin(aπ)2π[2aa2n2]=(1)nasin(aπ)π(a2n2) Therefore cos(ax)=n=(1)nasin(aπ)π(a2n2)einx=sin(πa)πa+j=1j=(1)jasin(aπ)π(a2j2)eijx+n=1(1)nasin(aπ)π(a2n2)einx=sin(πa)πa+n=1(1)nasin(aπ)π(a2n2)einx+n=1(1)nasin(aπ)π(a2n2)einx(n=j)=sin(πa)πa+n=1(1)nasin(aπ)π(a2n2)(einx+einx)=sin(πa)πa+2πn=1(1)nasin(aπ)cos(nx)π(a2n2)=2πsin(πa)[12a+n=1(1)nacos(nx)(a2n2)] Hence cos(ax)=2πsin(πa)[12a+n=1(1)nacos(nx)(a2n2)]
  3. With this result we can obtain the expansion in partial fraction of coth(x) and cot(x):

  4. If we put x=πcos(nπ)=(1)n in the Fourier series expansion of cot(ax) then cos(aπ)=2πsin(πa)[12a+n=1a(a2n2)] If we divide by sin(aπ) cot(aπ)=2π[12a+n=1a(a2n2)] Now, if we allow a to take complex values: let a=iv cot(ivπ)=2π[i2vn=1iv(v2+n2)] Multiplying by i coth(vπ)=2π[12v+n=1v(v2+n2)] We have found the expansion in partial fractions of cot and coth cot(xπ)=2π[12x+n=1x(x2n2)]x±1,±2,±3,.. coth(xπ)=2π[12x+n=1x(x2+n2)]x±i,±2i,±3i,..
  5. Another consequence is the expansion in zeta numbers of coth(x) and cot(x):

  6. Suppose that |x|<1, using (2) we have cot(xπ)=1πx2πn=1x(n2x2)=1πx2πn=1xn2(1(xn)2)=1πx2πn=1xn2j=0(xn)2j=1πx2πj=0n=11n2j+2x2j+1Rearrangement=1πx2πj=0ζ(2j+2)x2j+1=1πx2πk=1ζ(2k)x2k1 The rearrangement of the sums is possible due to the fact that the series j=0(xn)2j is absolutely convergent and the series ζ(v)=n=11nv is convergent. In a very similar way we can show the proof of coth(x) Therefore cot(πx)=1πx2πk=1ζ(2k)x2k1|x|<1 coth(πx)=1πx2πk=1(1)kζ(2k)x2k1|x|<1
  7. We will apply the residue theorem


  8. It is easy to show that if f(z)=πcoth(πz)z4m1 lim where C_{N} be a square with vertices \left(N+\frac{1}{2}\right)(1+i), \left(N+\frac{1}{2}\right)(-1+i), \left(N+\frac{1}{2}\right)(-1-i), \left(N+\frac{1}{2}\right)(1-i)

    Therefore, we can apply the following result : \sum_{\substack{n=-\infty \\ n\neq 0}}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} = -\sum \left\{\textrm{Residues of } f(z)\pi\cot(\pi z) \textrm{ at the sungularities of } f \right\} Given that the function f(z) has a singularity at z=0 and a countable number of singularities at \displaystyle z=in \quad n\in\mathbb{N}\setminus\left\{0\right\}

    We have: \sum_{\substack{n=-\infty \\ n\neq 0}}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} = -\sum_{n=1}^{\infty}\operatorname{Res}\left(\pi\cot(\pi z),in\right) -\sum_{n=1}^{\infty}\operatorname{Res}\left(f(z)\pi\cot(\pi z),-in\right) -\operatorname{Res}\left(f(z)\pi \cot(\pi z),0\right) With the L'Hopital rule can be proven that \lim_{z \to \pm in} (z\mp in)\cot(\pi z) \coth (\pi z) = \mp \frac{i\coth(n \pi)}{\pi} \quad \forall n>0 \therefore \operatorname{Res}\left(f(z)\pi\cot(\pi z),\mp in\right) = \lim_{z \to \pm in} \frac{(z\mp in)\pi^2 \cot(\pi z)\coth(\pi z)}{ z^{4m-1}} = \frac{\pi \coth(n \pi)}{ n^{4m-1}} Therefore \sum_{\substack{n=-\infty \\ n\neq 0}}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} = -\sum_{n=1}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} -\sum_{n=1}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} -\operatorname{Res}\left(f(z)\pi \cot(\pi z),0\right) 2\sum_{\substack{n=-\infty \\ n\neq 0}}^{\infty}\frac{\pi\coth(\pi n)}{n^{4m-1}} = -\operatorname{Res}\left(f(z)\pi \cot(\pi z),0\right)
    We just have to find the residue at z = 0.

    For that end, we will make use of the expansion in zeta numbers of \cot and \coth

    Note \begin{align*} \coth(\pi z)\cot(\pi z) =& \left[ \frac{1}{\pi z}- \frac{2}{\pi}\sum_{k=1}^{\infty}(-1)^k\zeta(2k) z^{2k-1}\right]\left[\frac{1}{\pi z}- \frac{2}{\pi}\sum_{k=1}^{\infty}\zeta(2k) z^{2k-1} \right] \quad (\textrm{from (4) and (5)})\\ &= \frac{1}{\pi^2 z^2} -\frac{2}{\pi^2} \sum_{k=1}^{\infty}\zeta(2k) z^{2k-2} - \frac{2}{\pi^2} \sum_{k=1}^{\infty}(-1)^k\zeta(2k) z^{2k-2} + \frac{4}{\pi^2} \underbrace{\sum_{k=1}^{\infty}\sum_{j=1}^{k} (-1)^{j}\zeta(2j)\zeta(2k-2j+2)z^{2k}}_{\textrm{Cauchy product}}\\ &= \frac{1}{\pi^2 z^2} -\frac{4}{\pi^2} \sum_{k=1}^{\infty}\zeta(4k) z^{4k-2} - \frac{4}{\pi^2} \sum_{k=1}^{\infty}\sum_{j=1}^{k} (-1)^{j}\zeta(2j)\zeta(2k-2j+2)z^{2k}\\ \end{align*} Hence \frac{\pi^2 \cot(\pi z)\cot(\pi z)}{z^{4m-1}} = z^{-4m-1-2}- 4\sum_{k=1}^{\infty}\zeta(4k) z^{4k-4m-1} + 4 \sum_{k=1}^{\infty}\sum_{j=1}^{k} (-1)^{j}\zeta(2j)\zeta(2k-2j+2)z^{2k-4m+1} The residue at z=0 implies that z^{4k-4m-1}= z^{-1} \Longrightarrow 4k-4m-1 = -1 \Longrightarrow k=m z^{2k-4m+1} = z^{-1} \Longrightarrow 2k-4m+1 = -1 \Longrightarrow k=2m-1 Therefore \operatorname{Res}\left(\frac{\pi^2\cot(\pi z)\coth(\pi z)}{z^{4m-1}}, 0\right) = -4\zeta(4m) + 4\sum_{j=1}^{2m-1} (-1)^{j}\zeta(2j)\zeta(4m-2j) Therefore 2\sum_{\substack{n=-\infty \\ n\neq 0}}^{\infty} \frac{\pi \coth(\pi n)}{n^{4m-1}} = 4\zeta(4m) + 4\sum_{j=1}^{2m-1} (-1)^{j-1}\zeta(2j)\zeta(4m-2j) Given that the function \displaystyle f(z) = \frac{\pi \coth(\pi z)}{z^{4m-1}} is even we have 4\sum_{n=1}^{\infty} \frac{\pi \coth(\pi n)}{n^{4m-1}} = 4\zeta(4m) + 4\sum_{j=1}^{2m-1} (-1)^{j-1}\zeta(2j)\zeta(4m-2j) Hence, we can conclude \boxed{\sum_{n=1}^{\infty} \frac{\pi \coth(\pi n)}{n^{4m-1}} = \zeta(4m) + \sum_{j=1}^{2m-1} (-1)^{j-1}\zeta(2j)\zeta(4m-2j) }

Saturday, October 2, 2021

Integral of the day VI

Nice integral of the day

Integral involving the Apéry's constant \zeta(3)


Today we show the solution to this integral posted by @sounansya_29 on Twitter \int_{0}^{\infty} \frac{x\ln(\sinh x)}{\sinh x \cosh x} dx = -\frac{7}{16}\zeta(3)
The solution involves the integral representation of the Dirichlet lambda function as well as some known results of Euler sums.

Proof

\begin{align*} I=\int_{0}^{\infty} \frac{x\ln(\sinh x)}{\sinh x \cosh x} dx =& 2\int_{0}^{\infty} \frac{x\ln(\sinh(x))}{\sinh(2x)} \quad \left( \sinh x \cosh x = \frac{1}{2}\sinh(2x)\right)\\ =& 2\int_{0}^{\infty} \frac{x\ln\left(\frac{1}{2}(e^{x}-e^{-x})\right)}{\sinh(2x)}\\ =& -2\ln(2)\int_{0}^{\infty} \frac{x}{\sinh(2x)}dx + 2\int_{0}^{\infty} \frac{x\ln(e^{x}(1-e^{-2x}))}{\sinh(2x)}dx\\ =& -2\ln(2)\underbrace{\int_{0}^{\infty} \frac{x}{\sinh(2x)}dx}_{I_{1}} +2\underbrace{\int_{0}^{\infty} \frac{x^2}{\sinh(2x)}dx}_{I_{2}}+ 2\underbrace{\int_{0}^{\infty} \frac{x\ln(1-e^{-2x})}{\sinh(2x)}dx}_{I_{3}}\\ \end{align*} Recall the beautiful integral representation of the Dirichlet lambda function: \lambda(v) = \frac{1}{2\Gamma(v)} = \int_{0}^{\infty} \frac{t^{v-1}}{\sinh(t)}dt Therefore, for I_{1}, I_{2} we have: I_{1} = \int_{0}^{\infty} \frac{x}{\sinh(2x)}dx = \frac{1}{4}\int_{0}^{\infty} \frac{w}{\sinh(w)} dw = \frac{1}{2}\Gamma(2) \lambda(2) = \frac{\pi^2}{16} \quad (w \mapsto 2x) I_{2} = \int_{0}^{\infty} \frac{x^2}{\sinh(2x)}dx = \frac{1}{8}\int_{0}^{\infty} \frac{w^2}{\sinh(w)} dw = \frac{1}{4}\Gamma(3) \lambda(3) = \frac{7\zeta(3)}{16} \quad (w \mapsto 2x)
For \displaystyle I_{3} we can split the integral in two:
\begin{align*} I_{3} = \int_{0}^{\infty} \frac{x\ln(1-e^{-2x})}{\sinh(2x)}dx =& 2\int_{0}^{\infty} \frac{x\ln(1-e^{-2x})}{e^{2x}-e^{-2x}}dx\\ =& -\frac{1}{2}\int_{0}^{1} \frac{\ln(w) \ln(1-w)}{(1-w)(1+w)}dw \quad (w \mapsto e^{-2x})\\ =& -\frac{1}{4}\underbrace{\int_{0}^{1} \frac{\ln(w) \ln(1-w)}{(1+w)}dw}_{J_{1}} - \frac{1}{4}\underbrace{\int_{0}^{1} \frac{\ln(w) \ln(1-w)}{(1-w)}dw}_{J_{2}} \end{align*} J_{1} = \int_{0}^{1} \frac{\ln(w) \ln(1-w)}{(1+w)}dw \stackrel{IBP}{=} \underbrace{\ln(w)\ln(w+1)\ln(1-w)\Big|_{0}^{1}}_{=0} - \underbrace{\int_{0}^{1} \frac{\ln(w+1)\ln(1-w)}{w} dw}_{K_{1}} + \underbrace{\int_{0}^{1}\frac{\ln(w)\ln(w+1)}{1-w}dw}_{K_{2}} \begin{align*} K_{1}=\int_{0}^{1}\frac{\ln(1-x)\ln(1+x)}{x} dx= & \int_{0}^{1}\sum_{n=1}^{\infty} \frac{(-1)^{n+1}x^n}{n}\frac{\ln(1-x)}{x} dx\\ =&\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n}\int_{0}^{1} x^{n-1}\ln(1-x) dx\\ =&\sum \frac{(-1)^{n+1}}{n}\int_{0}^{1}\sum_{n=1}^{\infty} x^{n-1} \left[\frac{d}{dt}\Big|_{t=0+} (1-x)^t\right] dx\\ =& \frac{d}{dt}\Big|_{t=0+} \sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n}\int_{0}^{1}\sum x^{n-1}(1-x)^t dx\\ =&\frac{d}{dt}\Big|_{t=0+} \sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n}B(n,t+1)\\ =& \sum_{n=1}^{\infty} \frac{(-1)^{n}}{n}\left(\psi^{(0)}(n+1)+\gamma\right)B(n,1)\\ =& \sum_{n=1}^{\infty} \frac{(-1)^{n}}{n^2}H_{n}\\ \end{align*} \begin{align*} K_{2} = \int_{0}^{1}\frac{\ln(w)\ln(w+1)}{1-w}dw = &\int_{0}^{1} \frac{\ln(1-s)\ln(2-s)}{s}ds \\ =& \int_{0}^{1} \frac{\ln(1-s)\ln\left(2\left(1-\frac{s}{2}\right)\right)}{s}ds\\ =&\ln(2) \int_{0}^{1} \frac{\ln(1-s)}{s}ds + \int_{0}^{1} \frac{\ln(1-s)\ln\left(1-\frac{s}{2}\right)}{s}ds \\ =& -\ln(2) \int_{1}^{0} \frac{\ln(1-s)}{s}ds - \int_{0}^{1} \frac{\ln(1-s)}{s} \sum_{n=1}^{\infty} \frac{s^{n}}{2^nn}ds \\ =& -\ln(2)\operatorname{Li}_{2}(1) - \sum_{n=1}^{\infty} \frac{1}{2^nn}\int_{0}^{1} s^{n-1}\ln(1-s)ds \\ =& -\frac{\ln(2)\pi^2}{6} - \sum_{n=1}^{\infty} \frac{1}{2^nn}\int_{0}^{1} s^{n-1} \left(\frac{d}{dt}\Big|_{t=0+} (1-s)^t\right) ds\\ =& -\frac{\ln(2)\pi^2}{6} - \frac{d}{dt}\Big|_{t=0+} \sum_{n=1}^{\infty} \frac{1}{2^nn}\int_{0}^{1} s^{n-1} (1-s)^t ds\\ =& -\frac{\ln(2)\pi^2}{6} - \frac{d}{dt}\Big|_{t=0+} \sum_{n=1}^{\infty} \frac{1}{2^nn}B(n,t+1)\\ =& -\frac{\ln(2)\pi^2}{6} + \sum_{n=1}^{\infty} \frac{H_{n}}{2^nn^2}\\ \end{align*} \begin{align*} J_{2} = \int_{0}^{1} \frac{\ln(1-w)\ln(w)}{1-w}dw =& \underbrace{-\frac{1}{2} \ln(w)\ln^2(1-w)\Big|_{0}^{1}}_{=0} +\frac{1}{2}\int_{0}^{1} \frac{\ln^2(1-w)}{w}dw \\ =& \frac{1}{2}\int_{0}^{1} \frac{\ln^2(1-w)}{w}dw \\ =& -\frac{1}{2} \int_{0}^{1} \sum_{n=1}^{\infty} \frac{w^n}{n} \frac{\ln(1-w)}{w} dw \\ =& - \frac{1}{2} \sum_{n=1}^{\infty} \frac{1}{n} \int_{0}^{1} w^{n-1} \left(\frac{d}{dt}\Big|_{t=0+} (1-w)^t \right) dw \\ =&\frac{d}{dt}\Big|_{t=0+} - \frac{1}{2} \sum_{n=1}^{\infty} \frac{1}{n} \int_{0}^{1} w^{n-1}(1-w)^t dw\\ =& \frac{d}{dt}\Big|_{t=0+} - \frac{1}{2} \sum_{n=1}^{\infty} \frac{1}{n}B(n,t+1)\\ =& \frac{1}{2} \sum_{n=1}^{\infty}\frac{H_{n}}{n^2} \end{align*} From the generating function (I will add the proof of the generating functions to the appendix later) \sum_{j=1}^{\infty} \frac{H_{j}(1-z)^j}{j^2} = \frac{1}{2}\ln^2(z)\ln(1-z)+\operatorname{Li}_{2}(z)\ln(z) - \operatorname{Li}_{3}(z) + \operatorname{Li}_{3}(1-z) If we put z = \frac{1}{2} we have: \sum_{n=1}^{\infty} \frac{H_{n}}{2^nn^2} = \zeta(3) - \frac{\pi^2\ln(2)}{12} If z \to 0 \sum_{n=1}^{\infty} \frac{H_{n}}{n^2} = \zeta(3) while from the generating function \sum_{j=1}^{\infty} \frac{H_{j}(z)^j}{j^2} = \frac{1}{3}\ln^3(1-z) +\ln(1-z)\operatorname{Li}_{2}\left(\frac{z}{z-1}\right) + \operatorname{Li}_{3}\left(\frac{z}{z-1}\right) +2 \operatorname{Li}_{3}(z) \quad z\in\left[-1,\frac{1}{2}\right] If we put \displaystyle z = -1 we have \sum_{n=1}^{\infty} \frac{(-1)^{n}}{n^2}H_{n} = -\frac{5}{8}\zeta(3) Therefore J_{1} = -K_{1} + K_{2} = -\sum_{n=1}^{\infty} \frac{(-1)^{n}}{n^2}H_{n} + -\frac{\ln(2)\pi^2}{6} + \sum_{n=1}^{\infty} \frac{H_{n}}{2^nn^2} = \frac{13}{8}\zeta(3) - \frac{\pi^2\ln(2)}{4} J_{2} = \frac{1}{2} \sum_{n=1}^{\infty}\frac{H_{n}}{n^2} = \zeta(3) I_{3} = -\frac{1}{4}J_{1} -\frac{1}{4}J_{2} = -\frac{21}{32}\zeta(3)+\frac{\pi^2\ln(2)}{16} I = -2\ln(2)I_{1} + 2I_{2} +2I_{3} = -2\ln(2)\frac{\pi^2}{16} + 2\frac{7\zeta(3)}{16} -\frac{42}{32}\zeta(3)+\frac{\pi^2\ln(2)}{8} = -\frac{7}{16}\zeta(3) Hence, we can conclude \boxed{ \int_{0}^{\infty} \frac{x\ln(\sinh x)}{\sinh x \cosh x} dx = -\frac{7}{16}\zeta(3)}

Series of the day

Series involving the digamma and the zeta functions The sum \displaystyle \sum\frac{1}{(n+1)^pn^q} ...