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

Tuesday, April 26, 2022

Generalized hypergeometric functions XI

Nice hypergeometric series

Nice hypergeometric series involving the silver ratio δS


Today we show the proof of this result involving hypergeometric series and the silver ratio δS=1+2 proposed by @mamekebi 3F2[1,1,1232,32,1]=10arcsinh(x)x1+x2dx=π28ln2(1+2)2 The proof relies on some properties of the rising factorial and some identities of the dilogarithm function.

Proof

Recall the recursion formula for the argument and the degree of the rising factorial: (x)n+1=(n+x)(x)n=x(x+1)n and the duplication formula (x)2n=22n(x2)n(x+12)n Hence 3F2[1,1,1232,32,1]=j=0(1)j(1)j(12)j(32)j(32)j(1)jj!=12j=0(1)j(1)j(32)j(j+12)(1)jj!from (1)=j=0(1)j(1)j(32)j(2j+1)(1)jj! We previously showed the proof of this series expansion: arcsinh(x)x1+x2=j=022j(2jj)(2j+1)(1)jx2j=j=022jj!2(2j)!(2j+1)(1)jx2j=j=022j(1)j(1)j(1)2j(2j+1)(1)jx2j((1)j=j!)=j=0(1)j(1)j(32)j(1)jx2jj! from (1) and (2) Integrating from 0 to 1 10arcsinh(x)x1+x2dx=j=0(1)j(1)j(32)j(2j+1)(1)jj!=3F2[1,1,1232,32,1] I=10arcsinh(x)x1+x2dx=10ln(x2+1+x)x1+x2dx=21+21ln(w)(w+1)(w1)dw(wx2+1+x)=1+21ln(w)1wdw1+21ln(w)w+1dw=20ln(1t)tdw1+21ln(w)w+1dw(t1w)IBP=20ln(1t)tdwln(1+2)ln(2+2)112ln(1w)wdw=02ln(1t)tdwln(1+2)ln(2+2)012ln(1w)wdw+01ln(1w)wdw=Li2(2)ln(1+2)ln(2+2)Li2(12)+Li2(1) From the identity Li2(z)Li2(1z)+12Li2(1z2)=π212lnzln(z+1) If we put z=2 Li2(2)=Li2(12)12Li2(1)π212ln2ln(2+1)(3) From the identity Li2(z)+Li2(1z)=π26ln2(z)2 If we put z=12

Therefore Li2(12)=Li2(12)π26ln2(21)2(4) Using the fact that Li2(1)=π212 and from (3) and (4) we have: I=π28+ln2ln(2+1)ln(1+2)ln(2+2)+ln2(21)2=π28ln2(1+2)2 3F2[1,1,1232,32,1]=10arcsinh(x)x1+x2dx=π28ln2(1+2)2

Sunday, April 17, 2022

Generalized hypergeometric functions X

Series involving powers of the binomial coeffient

Series involving the binomial coefficient


Today we show the proof of this series proposed by @qq3023625451 n=0(1)n(1+4n)210n(2nn)5=2Γ4(34) In the proof we use some properties of the generalized hypergeometric function pFq

Proof:

The Pocchhammer symbol (rising factorial) is defined as the product: (x)n=x(x+1)(x+n1)=n1j=0(x+j) while the generalized hypergeometric function is defined pFq[a1,a2,...,apb1,b2,...,bq;z]=n=0(a1)n(a2)n(ap)n(b1)n(b2)n(bq)nznn!

Note that (1)n=n!

The rising factorial obey the duplication formula: (x)2n=22n(x2)n(1+x2)n and the recursion formula for the argument (n+x)(x)n=x(x+1)n Hence S=n=0(1)n(1+4n)210n(2nn)5=n=0(1)n(1+4n)(2n)!5210nn!10=n=0(1)n(1+4n)(1)52n210n(1)10n=n=0(1)n(1+4n)210n(12)5n(1)5n210n(1)10nfrom (1)=n=0(1)n4(n+14)(12)n(12)n(12)n(12)n(12)n(1)n(1)n(1)n(1)nn!=n=0(54)n(12)n(12)n(12)n(12)n(12)n(14)n(1)n(1)n(1)n(1)n(1)nn!from (2)=6F5[12,54,12,12,12,1214,1,1,1,1;1] One of the Whipple's identities states that 6F5[a,1+12a,b,c,d,e12a,1+ab,1+ac,1+ad,1+ae;1]=Γ(1ad)Γ(1+ae)Γ(1+a)Γ(1+ade)3F2[1+abc,d,e1+ab,1+ac;1] Hence S=6F5[12,54,12,12,12,1214,1,1,1,1;1]=1Γ(32)Γ(12)3F2[12,12,121,1;1] 3F2 obey 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, using Γ(32)=π2 and Γ(12)=π S=1Γ(32)Γ(12)3F2[12,12,121,1;1]=Γ(54)Γ(14)Γ(32)Γ(12)Γ(32)Γ(34)Γ(34)Γ(12)=4Γ(54)Γ(14)π2Γ(34)Γ(34)=Γ(14)Γ(14)π2Γ(34)Γ(34)(Γ(z+1)=zΓ(z))=2Γ(34)Γ(34)Γ(34)Γ(34)(Γ(1z)Γ(z)=πsin(πz))=2Γ4(34) n=0(1)n(1+4n)210n(2nn)5=2Γ4(34)

Saturday, April 16, 2022

Integral of the day XXXIV

Polylogarithm function

Integral involving the golden mean ϕ


Today we show the proof of the following integral posted by @integralsbot π0x252cosxdx=2πln2(ϕ)+π315 The proof relies on a "discrete" Laplace transform a technique that we have used in the past here and here

Proof:

Consider the following discrete Laplace transform (proof in Appendix 1): 1+2k=1ektcos(kx)=sinhtcoshtcosxt>0 If cosht=52 then t=ln(1+52)=ln(ϕ) sinh(1+52)=12 therefore 1252cosx=1+2k=1ϕkcos(kx) where ϕ=1+52

Hence I=π0x252cosxdx=12π0x252cosxdx=π0(1252cosx)x2dx=π0(1+2k=1ekln(ϕ)cos(kx))x2dx=π0[x2+2k=1ϕkcos(kx)x2]dx=π0[x2dx+2k=1ϕkπ0cos(kx)x2]dx=π0[x2dx+2k=1ϕkk3kπ0cos(w)w2]dxIBP=π33+2k=1ϕkk3[w2sin(w)|kπ02kπ0sin(w)wdw]IBP=π33+2k=1ϕkk3[(kπ)2sin(kπ)+2cos(w)w|kπ0+2πk0cos(w)dw]=π33+2k=1ϕkk3[(kπ)2sin(kπ)+2cos(kπ)kπ2sin(πk)]=π33+2k=1ϕk(k2π22)sin(kπ)k3sin(πk)=0+4k=1ϕkπcos(kπ)k2cos(πk)=(1)k=π33+4πk=1(ϕ1)kk2=π33+4πLi2(ϕ1) where we used the series expansion for the dilogarithm function Li2(z): Li2(z)=k=1zkk2 Landen and Euler proved (proof in the Appendix 2): Li2(ϕ1)=π215+12ln2(ϕ) Therefore π0x252cosxdx=2πln2(ϕ)+π315

Appendix 1

1+2k=1ektcoskx=sinhtcoshtcosxt>0 Proof: 1+2k=1ektcoskx=1+2(k=1ekteikx)=1+2(k=1ek(ixt))=1+2(1etix1)=1+2(eteixet)=1+2et(1cosxisinxet)=1+2et(cosxetisinx(cosxet)2+sin2x)=1+2et(cosxetisinxe2t2etcosx+1)=1+(cosxet+isinxcoshtcosx)=coshtcosxcoshtcosx+cosxet(coshtcosx)=sinhtcoshtcosx

Appendix 2

Li2(ϕ1)=π215+12ln2(ϕ)

Proof:

Consider the following derivative: ddxLi2(x1x)==[ln(1+x1x)](1x+11x)=ln(1x)(1x+11x) Integrating Li2(x1x)=Li2(x)12ln2(1x) Li2(x1x)+Li2(x)=12ln2(1x) If we put x=352 Li2(152)+Li2(352)=12ln2(1+52) Now we will show that 12Li2(x2)=Li2(x)+Li2(x) From the factorization (1xr)=(1x)(1wx)(1w2x)(1wr1x) where w=ei2πr Taking logarithms and dividing by x both sides ln(1xr)x=ln(1x)x+ln(1wx)x++ln(1wr1x)x Integrating 1rLi2(xr)=Li2(x)+Li2(xw)++Li2(wr1x)+C If x0 then C0 1rLi2(xr)=Li2(x)+Li2(xw)++Li2(wr1x) If we put r=2 12Li2(x2)=Li2(x)+Li2(x) From (1) and (3) we have Li2(x1x)+12Li2(x)Li2(x)=12ln2(1x) If we solve the equation x1x=x2x2x+1=0x=152 Hence from (4) we have 32Li2(352)Li2(512)=12ln2(1+52) If we put x=352 in the relation (Proof in the Appendix 2 here) Li2(x)+Li2(1x)=π26ln(x)ln(1x) we have Li2(352)+Li2(512)=π26ln(352)ln(512) Solving the system given by (5) and (7) we have Li2(352)=π21514ln2(352) Then from (2) and (8) we get Li2(152)=π215+12ln2(512) which is basically the same as Li2(ϕ1)=π215+12ln2(ϕ)

Friday, April 15, 2022

Integral of the day XXXIII

Polylogarithm function

An integral representation of the dilogarithm function


Today we show the proof of this integral posted by @Ali39342137 π0xarctan(sinx2+cosx)dx=π312π2ln2(2) In general we found π0xarctan(sinxa+cosx)dx=πLi2(1a)|a|1 In the proof we make use of Fourier series and some properties of the dilogarithm function Li2(z).

Proof:

Consider the following Fourier series (Proof in the Appendix 1): k=1pksin(kx)k=arctan(psinx1pcosx)0<x<2π,p21 Then, if p=q k=1(1)kqksin(kx)k=arctan(qsinx1+qcosx)=arctan(qsinx1+qcosx) Therefore arctan(sinx1q+cosx)=k=1(1)k+1qksin(kx)k1q1 Hence I=π0xarctan(sinxa+cosx)dx=π0xk=1(1)k+1(1a)ksin(kx)kdx=k=1(1)k+1(1a)kkπ0xsin(kx)dx=k=1(1)k+1(1a)kk3kπ0wsin(w)dx(wkx)IBP=k=1(1)k+1(1a)kk3[cos(x)x|kπ0+kπ0cos(w)dx]=k=1(1)k+1(1a)kk3[kπcos(πk)+sin(πk)]=πk=1(1)k(1a)kcos(kπ)k2cos(kπ)=(1)k+k=1(1)k+1(1a)ksin(πk)k3sin(kπ)=0=πk=1(1a)kk2=πLi2(1a) If a=2 π0xarctan(sinx2+cosx)dx=πLi2(12) From the relation (proof in the Appendix 2) Li2(z)+Li2(1z)=π26ln(z)ln(1z) if we put z=12 Li2(12)=π21212ln2(2) Therefore π0xarctan(sinx2+cosx)dx=π312π2ln2(2) π0xarctan(sinxa+cosx)dx=πLi2(1a)|a|1

Appendix 1 k=1pksin(kx)k=arctan(psinx1pcosx)0<x<2π,p21

Proof

Take the principal branch of the ln(z) function. Hence k=1pksin(kx)k=(k=1(peix)kk)=(ln(1peix))=arg(1peix)=arg(1pcos(x)pisin(x)) Since |p|<1 then 1pcos(x)>0 therefore k=1pksin(kx)k=arg(1pcos(x)pisin(x))=arctan(psinx1pcos(x))=arctan(psinx1pcos(x))(arctan is odd )

Appendix 2 Li2(z)+Li2(1z)=π26ln(z)ln(1z)

Proof Li2(z)=0zln(1t)tdtIBP=ln(z)ln(1z)z0ln(t)1tdt=ln(z)ln(1z)+1z1ln(1w)wdw(w1t)=ln(z)ln(1z)+01ln(1w)wdw+1z0ln(1w)wdw=ln(z)ln(1z)+01ln(1w)wdw01zln(1w)wdw=ln(z)ln(1z)+Li2(1)Li2(1z) Therefore Li2(z)+Li2(1z)=ln(z)ln(1z)+ζ(2)=ln(z)ln(1z)+π26

Thursday, April 7, 2022

Application of Ramanujan's master theorem II

Good old Ramanujan's master theorem

A useful Mellin transform related to the Bernoulli and zeta numbers


Today we show the proof of this Mellin transform posted by @integralsbot 0xsn2(xex1nk=0Bkxkk!)dx=Γ(sn)ζ(sn)0<(s)<1 The proof relies on the Ramanujan's master theorem, some properties of the Bernoulli numbers and the binomial coefficient.

We also show the proof of this nice application: an integral representation for the Apery's constant (also posted by @integralsbot ) : 0(1x2+x212xex1)1x4dx=ζ(3)8π2

Proof:

We will apply the Ramanujan master theorem: If a complex-valued function f(x) has an expansion of the form f(x)=k=0φ(k)k!(x)k then the Mellin transform of f(x) is given by 0xs1f(x)dx=Γ(s)φ(s) where Γ(s) is the gamma function.

The Bernoulli numbers are defined through the generating function xex1=k=0Bkxkk! Therefore xex1nk=0Bkxkk!=k=0Bkxkk!nk=0Bkxkk!=k=n+1Bkxkk!=j=0Bj+n+1xj+n+1(j+n+1)! The zeta function are related to the Bernoulli numbers through the following expression Bn=ζ(1n)nn=2,3,4,.. Hence xex1nk=0Bkxkk!=j=0Bj+n+1xj+n+1(j+n+1)!=j=0ζ(jn)(j+n+1)xj+n+1(j+n+1)!=j=0ζ(jn)xj+n+1(j+n)!=xn+1j=0ζ(jn)j!(1)j(j+n)!(x)jj!=xn+1j=0ζ(jn)j!n!(1)jn!(j+n)!(x)jj!=xn+1j=0ζ(jn)(1)jn!(n+jn)(x)jj! The formula (vm)=(vm+1)mm! relate the binomial coefficient to the Pochhammer polynomial (rising factorial) while the Pochhammer polynomial obey the reflection formula (x)n=(1)n(xn+1)n Hence xex1nk=0Bkxkk!=xn+1j=0ζ(jn)(1)jn!(n+jn)(x)jj!=xn+1j=0ζ(jn)(1)j(j+1)n(x)jj!from (1)=xn+1j=0ζ(jn)(jn)n(x)jj!from (2) Back to the integral: 0xsn2(xex1nk=0Bkxkk!)dx=0xs1(j=0ζ(jn)(jn)n(x)jj!)dx Denote f(x)=j=0ζ(jn)(jn)n(x)jj! φ(j)=ζ(jn)(jn)n Applying the Ramanujan's master theorem I=0xsn2(xex1nk=0Bkxkk!)dx=Γ(s)φ(s)=Γ(s)ζ(sn)(sn)n Finally, recall that the rising facotiral can be expressed as the quotient of two gamma functions: (x)n=Γ(n+x)Γ(x) Hence I=Γ(s)ζ(sn)(sn)n=Γ(s)ζ(sn)Γ(sn)Γ(s)=Γ(sn)ζ(sn) Therefore, we can conclude 0xsn2(xex1nk=0Bkxkk!)dx=Γ(sn)ζ(sn)0<(s)<1

Application: Integral representation of the Apery's constant

0(1x2+x212xex1)1x4dx=ζ(3)8π2

Proof

Put n=2 in the Mellin transform, then 0xs4(1x2+x212xex1)dx=Γ(s2)ζ(s2) Under some technical conditions the Mellin transform converges uniformly and we can take the limit as s0 lim What is the limit on the right hand side? \begin{align*} \lim_{s \to 0} -\Gamma(s-2)\zeta(s-2) =& \lim_{s \to 0} -\Gamma(s-2)\zeta(s-2)\\ =& \lim_{t \to -2} -\Gamma(t)\zeta(t)\\ =& \lim_{t \to -2} -\frac{\zeta(t)}{\frac{1}{\Gamma(t)}}\\ =& \lim_{t \to -2} -\frac{\zeta(t)}{\frac{t(t+1)(t+2)}{\Gamma(t+3)}} \quad \left(\Gamma(t+1) = t\Gamma(t) \right)\\ =& -\frac{\lim_{t \to -2} \zeta'(t)}{\lim_{t \to -2} \frac{3t^2+6t+2-(t^2+3t+2)\psi(t+3)}{\Gamma(t+3)}} \quad \textrm{(L'Hôpital's rule )}\\ =& -\frac{\zeta'(-2)}{2}\\ \end{align*} The derivative of the zeta function at the negative even integers is given by \zeta^{\prime}(-2n) = (-1)^n \frac {(2n)!} {2 (2\pi)^{2n}} \zeta (2n+1) If we put n=1 \zeta^{\prime}(-2) = - \frac {\zeta (3)} { 4\pi^2} Hence \lim_{s \to 0} -\Gamma(s-2)\zeta(s-2) = \frac {\zeta (3)} { 8\pi^2} Therefore we can conclude \boxed{\int_0^{\infty} \left(1-\frac{x}{2}+\frac{x^2}{12} -\frac{x}{e^x-1}\right)\frac{1}{x^4} dx = \frac {\zeta (3)} { 8\pi^2} }

Wednesday, April 6, 2022

Integral of the day XXXII

Fun integral related to the polygamma

Integral involving the polylogamma function \psi^{(n)}(z)


Today we show the proof of this nice result posted by @Ali39342137 \int_{0}^{1} \ln^2(x)\ln(x^2+x+1) dx = \frac{2}{3}\psi^{(1)}\left(\frac{4}{3}\right)+\frac{8}{9}\zeta(3) +\frac{4\pi^3}{27\sqrt{3}} +2\ln\sqrt{27} -6 +\frac{\pi}{\sqrt{3}}-\frac{\pi^2}{3} The proof relies heavily on properties of the digamma, the trigamma and the tetragamma functions \psi(z),\psi^{(1)}(z), \psi^{(2)}(z) and the polygamma function \psi^{(n)}(z) in general.

Proof:

\begin{align*} I = \int_{0}^{1} \ln(x)\ln(1+x+x^2)dx =& \int_{0}^{1} \ln(x)\ln\left(\frac{1-x^3}{1-x}\right)dx\\ =& \underbrace{\int_{0}^{1} \ln(x)\ln\left(1-x^3\right)dx}_{J}- \underbrace{\int_{0}^{1} \ln(x)\ln\left(1-x\right)dx}_{K} \end{align*} \begin{align*} J = \int_{0}^{1} \ln^2(x)\ln\left(1-x^3\right)dx =& -\int_{0}^{1} \sum_{j=1}^{\infty} \frac{x^{3j}}{j}\ln^2(x) dx \\ =& -\sum_{j=1}^{\infty}\frac{1}{j} \int_{0}^{1} x^{3j}\left(\frac{d^2}{dt^2}\Big|_{t=0+} x^t \right) dx\\ =& \frac{d^2}{dt^2}\Big|_{t=0+} -\sum_{j=1}^{\infty} \frac{1}{j}\int_{0}^{1} x^{3j+t} dx\\ =& \frac{d^2}{dt^2}\Big|_{t=0+} -\sum_{j=1}^{\infty}\frac{1}{j(3j+t+1)} \\ =& -\sum_{j=1}^{\infty}\left[ \frac{d^2}{dt^2}\Big|_{t=0+} \frac{1}{j(3j+t+1)} \right]\\ =& -2\sum_{j=1}^{\infty}\frac{1}{j(3j+1)^3} \\ =& 6\sum_{j=1}^{\infty} \frac{1}{(3j+1)^2} + 6\sum_{j=1}^{\infty} \frac{1}{(3j+1)^3} -2\sum_{j=1}^{\infty} \frac{1}{j(3j+1)} \quad \textrm{(partial fractions)}\\ =& 6\sum_{n=0}^{\infty} \frac{1}{(3n+4)^2} + 6\sum_{n=0}^{\infty} \frac{1}{(3n+4)^3} -2\sum_{n=0}^{\infty} \frac{1}{(n+1)(3n+4)} \quad (n=j-1)\\ =& \frac{2}{3}\sum_{n=0}^{\infty} \frac{1}{\left(n+\frac{4}{3}\right)^2} + \frac{2}{9}\sum_{n=0}^{\infty} \frac{1}{\left(n+\frac{4}{3}\right)^3} -\frac{2}{3}\sum_{n=0}^{\infty} \frac{1}{(n+1)\left(n+\frac{4}{3}\right)} \end{align*} Recall that the polygamma function has the series representation \psi^{(m)}(z) = (-1)^{m+1}\, m! \sum_{k=0}^\infty \frac{1}{(z+k)^{m+1}} and that the difference between two digamma functions may be expressed as \psi(x)-\psi(y) = (x-y)\sum_{j=0}^{\infty} \frac{1}{(j+x)(j+y)} \tag{*} Therefore J = \int_{0}^{1} \ln^2(x)\ln\left(1-x^3\right)dx = \frac{2}{3}\psi^{(1)}\left(\frac{4}{3}\right)-\frac{1}{9}\psi^{(2)}\left(\frac{4}{3}\right)-2\psi\left(\frac{4}{3}\right)+2\psi(1) It is well known that \displaystyle \psi(1) = -\gamma

For \displaystyle \psi\left(\frac{4}{3}\right) we can use the following formula: \psi\left(\frac{1}{3}+J\right) +\frac{\pi}{2\sqrt{3}} = -\gamma -\ln\sqrt{27} + \sum_{j=1}^{J} \frac{3}{3j-2} \quad J=0,1,2,... If we put \displaystyle J=1 \psi\left(\frac{4}{3}\right) = -\gamma -\ln\sqrt{27} +3 -\frac{\pi}{2\sqrt{3}} For \displaystyle \psi^{(2)}\left(\frac{4}{3}\right) Recall the recurrence formula for the polygamma function: \psi^{(n)}(x+1) = \psi^{(n)}(x) - \frac{n!}{(-x)^{n+1}} \quad n=1,2,3,... If we put n=2 and \displaystyle x = \frac{1}{3} \psi^{(2)}\left(\frac{4}{3}\right) = \psi^{(2)}\left(\frac{1}{3}\right)+54 From the reflection formula (-1)^n\psi^{(n)} (1-x) - \psi^{(n)} (x) = \pi \frac{\mathrm{d}^n}{\mathrm{d} z^n} \cot{\pi x} if we put n=2 and \displaystyle x=\frac{2}{3} \psi^{(2)}\left(\frac{1}{3}\right) - \psi^{(2)}\left(\frac{2}{3}\right) = -\frac{8\pi^3}{3\sqrt{3}} while from the multiplication formula k^{m+1} \psi^{(m)}(kz) = \sum_{n=0}^{k-1} \psi^{(m)}\left(z+\frac{n}{k}\right)\qquad m \ge 1 If we put m=2, k=3 and \displaystyle z=\frac{1}{3} 27\psi^{(2)}\left(1\right) = \psi^{(2)}\left(\frac{1}{3}\right) + \psi^{(2)}\left(\frac{2}{3}\right) + \psi^{(2)}(1) Hence \psi^{(2)}\left(\frac{1}{3}\right) = 13\psi^{(2)}(1) - \frac{4\pi^3}{3\sqrt{3}} Finally, using the expression \psi^{(n)}(J) = (-1)^{n+1} n! \left[ \zeta(n+1)-\sum_{j=1}^{J-1} \frac{1}{j^{n+1}} \right] \psi^{(2)}(1) = -2\zeta(3) Therefore \psi^{(2)}\left(\frac{1}{3}\right) = -26\zeta(3) -\frac{4\pi^3}{3\sqrt{3}} \psi^{(2)}\left(\frac{4}{3}\right) = -26\zeta(3) -\frac{4\pi^3}{3\sqrt{3}}+54 However, \displaystyle \psi^{(1)}\left(\frac{4}{3}\right) do not have an expression in elementary functions. Therefore J = \int_{0}^{1} \ln^2(x)\ln\left(1-x^3\right)dx = \frac{2}{3}\psi^{(1)}\left(\frac{4}{3}\right)+\frac{26}{9}\zeta(3) +\frac{4\pi^3}{27\sqrt{3}} +2\ln\sqrt{27} -12 +\frac{\pi}{\sqrt{3}} \begin{align*} K = \int_{0}^{1} \ln^2(x)\ln(1-x)dx =& -\int_{0}^{1}\sum_{j=1}^{\infty}\frac{x^j}{j}\ln^2(x)dx\\ =& -\sum_{j=1}^{\infty}\frac{1}{j} \int_{0}^{1}x^j \left(\frac{d^2}{dt^2}\Big|_{t=0} x^t\right)dx\\ =& \frac{d^2}{dt^2}\Big|_{t=0} -\sum_{j=1}^{\infty}\frac{1}{j}\int_{0}^{1}x^{j+t}dx\\ =& \frac{d^2}{dt2}\Big|_{t=0} -\sum_{j=1}^{\infty}\frac{1}{j(j+t+1)}\\ =& -\sum_{j=1}^{\infty}\left[\frac{d^2}{dt^2}\Big|_{t=0} \frac{1}{j(j+t+1)}\right]\\ =& -2\sum_{j=1}^{\infty} \frac{1}{j(j+1)^2}\\ =& 2\sum_{j=1}^{\infty} \frac{1}{(j+1)^2} + 2\sum_{j=1}^{\infty} \frac{1}{(j+1)^3} -2\sum_{j=0}^{\infty} \frac{1}{j(j+1)} \quad \textrm{(partial fractions)}\\ =& \underbrace{2\sum_{m=2}^{\infty} \frac{1}{m^2} + 2\sum_{m=2}^{\infty} \frac{1}{m^3}}_{m=j+1} -2\sum_{j=0}^{\infty} \frac{1}{j(j+1)}\\ =& 2\zeta(2)+2\zeta(3)-4 -2\sum_{j=0}^{\infty} \frac{1}{j(j+1)}\\ =& \frac{\pi^2}{3}+2\zeta(3)-4 -2\sum_{n=0}^{\infty} \frac{1}{(n+1)(n+2)} \quad (n=j-1)\\ =& \frac{\pi^2}{3}+2\zeta(3)-4 -2\psi(2)+2\psi(1) \quad \textrm{ (from (*))} \\ =& \frac{\pi^2}{3}+2\zeta(3)-4 -2(1-\gamma)-2\gamma \\ =& \frac{\pi^2}{3}+2\zeta(3)-6 \\ \end{align*} Therefore I = \int_{0}^{1} \ln^2(x)\ln(x^2+x+1) dx = K-J = \frac{2}{3}\psi^{(1)}\left(\frac{4}{3}\right)+\frac{8}{9}\zeta(3) +\frac{4\pi^3}{27\sqrt{3}} +2\ln\sqrt{27} -6 +\frac{\pi}{\sqrt{3}}-\frac{\pi^2}{3} We can conclude \boxed{\int_{0}^{1} \ln^2(x)\ln(x^2+x+1) dx = \frac{2}{3}\psi^{(1)}\left(\frac{4}{3}\right)+\frac{8}{9}\zeta(3) +\frac{4\pi^3}{27\sqrt{3}} +2\ln\sqrt{27} -6 +\frac{\pi}{\sqrt{3}}-\frac{\pi^2}{3}}

Series of the day

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