Processing math: 59%

Tuesday, August 31, 2021

Elliptic integrals II

Catalan constant

Series involving K(p) and the Catalan constant


Today we show the proof of this result posted by @seriesbot_q on Twitter: β(2)=π4n=0(2nn)224n(2n+1) This series turned out being a slight variation of the series expansion of the complete elliptic integral of the first kind K(p) .

Proof

Recall the definition of the complete elliptic integral of the first kind: K(p)=π2011p2sin2θdθ|p|<1

First, we are going to obtain its series expansion.

In a previous post we proved the following result: 11x=n=0(2n1)!!(2n)!!xn0x<1 We know that 1sinθ1 |p|<1 p2sin2θ<1 Hence 11p2sin2θ=n=0(2n1)!!(2n)!!p2nsin2n(θ) Integrating from 0 to π2 K(p)=π2011p2sin2θdθ=π20n=0p2n(2n1)!!(2n)!!sin2n(θ)dθ=n=0(2n1)!!(2n)!!p2nπ20sin2n(θ)dθ=n=0(2n1)!!(2n)!!p2n10w2n(1w2)12dw(wsinθ)=12n=0(2n1)!!(2n)!!p2n10sn+121(1s)121ds(sw2)=12n=0(2n1)!!(2n)!!p2nB(n+12,12)=12n=0(2n1)!!(2n)!!p2nΓ(n+12)Γ(12)Γ(n+1) From the properties of the double factorial we have: (2n1)!!=2nΓ(n+12)π (2n)!!=2nn!=2nΓ(n+1) Hence K(p)=12n=0(2n1)!!(2n)!!Γ(n+12)Γ(12)Γ(n+1)p2n=π2n=0(2n1)!!2(2n)!!2p2n Recall the we can also write (2n1)!!(2n)!!=(2nn)22n Therefore, this is the expansion series of the complete elliptic integral of the first kind: K(p)=π2n=0(2n1)!!2(2n)!!2p2n=π2n=0(2nn)24n2p2n Now, integrating from 0 to 1 10K(p)dp=π2n=0(2nn)224n(2n+1) The right hand side is the series that we are looking for. Then, we just have to find the solution to the integral in the left hand side: 10K(p)dp=10π2011p2sin2θdθdp=π201011p2sin2θdpdθ=π20sinθ011w2sinθdwdθ(wpsinθ)=π201sinθsinθ011w2dwdθ=π201sinθarcsin(sinθ)dθ=π20θsinθdθ We solved this last integral in another post and the solution turned out being 2β(2). Hence 10K(p)dp=2β(2) Therefore β(2)=π4n=0(2nn)224n(2n+1)

Monday, August 30, 2021

Taylor's theorem II

Series from Twitter

Consequences of the series expansion of 11x


Today we show the proof of the following series found on Twitter: n=1(2nn)22nnxn=n=1(2n1)!!n(2n)!!xndx=2ln(21x+1) n=1(2nn)22nn2=n=1(2n1)!!(2n)!!n2=π262ln2(2) In reality these series are just a transformation of a very simple series as we will show

Proof

We previously showed the proof of the following series expansion with the help of Pochhammer polynomials: 11x=n=0(2n1)!!(2n)!!xn=n=0(2nn)22nxn=2F1(12,1;1;x) This series, as simple as it is, will allow us to calculate other integrals and series much easier and this is one of those cases.
We will start subtracting 1 from each side of the equation: 11x1=n=1(2n1)!!(2n)!!xn Now divide by x 1x1x1x=n=1(2n1)!!(2n)!!xn1 Integrating from 0 to x: x01x1x1xdx=n=1(2n1)!!(2n)!!x0xn1dx=n=1(2n1)!!n(2n)!!xndx The integral in the left hand side has the following antiderivative: 1x1x1xdx=2ln(1x+1)+C Hence n=1(2n1)!!n(2n)!!xndx=2ln(1x+1)+2ln(2)=2ln(21x+1) This is our first result: n=1(2n1)!!n(2n)!!xndx=2ln(21x+1) If we divide by x again we have: n=1(2n1)!!n(2n)!!xn1dx=2ln(1x+1)+2ln(2)x Integrating from 0 to 1: n=1(2n1)!!n2(2n)!!=102ln(1x+1)+2ln(2)xdx We can also write the left hand side with this form: n=1(2n1)!!n2(2n)!!=n=1(2nn)n222n This follow from the fact that: (2n1)!!=(2n)!2nn! (2n)!!=2nn! (2n1)!!(2n)!!=(2n)!22nn!2=122n(2nn) Therefore: n=1(2nn)n222n=102ln(1x+1)+2ln(2)xdx Integrating by parts: Let du=1xv=2ln(1x+1)+2ln(2) then u=ln(x)dv=1(1x+1)1x 102ln(1x+1)+2ln(2)xdx=ln(x)(2ln(1x+1)+2ln(2))|1010ln(x)(1x+1)1x=10ln(x)(1x+1)1x=210ln(1w2)w+1dw=210ln((1+w)(1w))w+1dw=210ln(1+w)w+1dwI1210ln(1w)w+1dwI2 The integral I1 is easily solvable given that ln(1+w)w+1dw=12ln2(1+w)+C Hence I1=12ln2(2) For I2: I2=10ln(1w)w+1dw=10n=0(w)nln(1w)dw=n=0(1)n10wnln(1w)dw=n=0(1)n10wn(ddt|t=0+(1w)t)dw=ddt|t=0+n=0(1)n10wn(1w)tdw=ddt|t=0+n=0(1)nB(n+1,t+1)=limt0+n=0(1)n(ψ(0)(t+1)ψ(0)(n+t+2))B(n+1,t+)=n=0(1)n(ψ(0)(1)ψ(0)(n+2))B(n+1,1)=n=0(1)n+1Hn+1n+1=j=1(1)jHjj(j=n+1) Now recall the generating function for the harmonic numbers: n=1Hnzn=ln(1z)1z Dividing by z n=1Hnzn1=ln(1z)z(1z)=ln(1z)zln(1z)1z Integrating from 0 to 1 n=1Hn(1)nn=10ln(1z)zdz10ln(1z)1z=10ln(1+z)zdz+10ln(1+z)1+z Hence n=1Hn(1)nn=10ln(1+z)zdz+I1 We just have to calculate 10ln(1+z)zdz 10ln(1+z)zdz=10k=1(1)k+1zk1kdz=k=1(1)k+110zk1kdz=k=1(1)k+1k2(Dirichlet eta function)=η(2)=π212 Therefore I2=n=1Hn(1)nn=π212+12ln2(2) Hence, from (1),(2),(3),(4),(5) we can conclude: 102ln(1x+1)+2ln(2)xdx=2I12I2=π262ln2(2) Therefore
n=1(2nn)22nn2=n=1(2n1)!!(2n)!!n2=π262ln2(2) n=1(2nn)22nnxn=n=1(2n1)!!n(2n)!!xndx=2ln(21x+1)

Wednesday, August 25, 2021

Integral of the day II

Integral related to the Sinc function

Integral related to the Sinc function


We show the proof of this mysterious integral:
0sinxxln(a2+cos2xb2+cos2x)=πln(a2+1+ab2+1+b)|a|,|b|>1
The integral resisted various techniques (residues, laplace transform) but following a hint suggested by @mvs_rpi we managed to obtain the proof: It turned out that this integral is a Lobachevsky integral.

The Lobachevsky integral formula states that:

If f(x) is a continous function satisfying the π-periodic assumption f(x+π)=f(x), and f(πx)=f(x), for 0x<

then
0sin2xx2f(x)dx=0sinxxf(x)dx=π20f(x)dx
Clearly
f(x)=ln(a2+cos2xb2+cos2x)
is a π periodic function.

Therefore we just have to calculate
π20ln(a2+cos2xb2+cos2x)dx
If we split the integral in two:
π20ln(a2+cos2xb2+cos2x)dx=π20ln(a2+cos2x)dxI1π20ln(b2+cos2x)dxI2
if we find I1 automatically we also have I2


Proof:

First, we will show that π20ln(a2+cos2x)dx=πln(a2+1+a)πln(2)|a|>1 I1=π20ln(a2+cos2x)dx=π20ln(a2(1+1a2cos2x))dx=π20ln(a2)dx+π20ln(1+1a2cos2x)dx=πln(a)+π20k=1[(1)k+1kcos2kxa2k]dx(|a|>1)=πln(a)+k=1(1)k+1ka2kπ20cos2kxdx=πln(a)k=11k(1a2)kπ20w2k(1w2)12dw(wcosx)=πln(a)12k=11k(1a2)kπ20sk+121(1s)121ds(sw2)=πln(a)12k=11k(1a2)kB(k+12,12)=πln(a)12k=11k(1a2)kΓ(k+12)Γ(12)Γ(k+1) Recall that Γ(k+12)=(2k1)!!π2k Γ(k+1)=(2k)!!2k Hence Γ(k+12)Γ(k+1)=(2k1)!!π(2k)!! Therefore I1=πln(a)π2k=11k(1a2)k(2k1)!!(2k)!! In another post we proved that 11x=k=0(2k1)!!(2k)!!xk Then, substracting 1 and dividing by 1x afterwards (the same trick we used in this post): 11x1=k=1(2k1)!!(2k)!!xk 1x1x1x=k=1(2k1)!!(2k)!!xk1 Integrating: x01x1x1xdx=k=1(2k1)!!(2k)!!x0xk1dx=k=11k(2k1)!!(2k)!!xk But note that 1x1x1xdx=2ln(1x+1)+C Therefore k=11k(2k1)!!(2k)!!xk=2ln(1x+1)+2ln(2) If we put x=1a2 k=11k(2k1)!!(2k)!!(1a2)k=2ln(1+1a2+1)+2ln(2) Therefore I1=πln(a)π2k=11k(1a2)k(2k1)!!(2k)!!=πln(a)+πln(1+1a2+1)πln(2)=πln(a)+πln(a2+1a2+aa)πln(2)=πln(a)+πln(1a(a2+1+a))πln(2)=πln(a)πln(a)+πln((a2+1+a))πln(2)=πln(a2+1+a)πln(2) Therefore I1=π20ln(a2+cos2x)dx=πln(a2+1+a)πln(2) I2=π20ln(b2+cos2x)dx=πln(b2+1+b)πln(2) Hence, we have the desired result π20ln(a2+cos2xb2+cos2x)dx=πln(a2+1+ab2+1+b) By the Lobachevsky integral formula: 0sinxxln(a2+cos2xb2+cos2x)=πln(a2+1+ab2+1+b)|a|,|b|>1

Sunday, August 22, 2021

Laplace transform III

Sinc and Sinch functions

Laplace transform of the Sinhc and Sinc functions


We show the proof of the Laplace transforms of these slight variations of the Sinc and the Sinhc function from two problems posted by @BossGercek here and here. Both can be obtained from each other if we allow the parameters b,c to have complex values. 0eaxsinhbxxdx=ln(a+bab) 0eaxsincxxdx=iln(a+icaic) Taking care where the inverse hyperbolic and trigonometric functions are defined, if we put b=c=1 we get the classic values for the Laplace transform of the Sinc and Sinhc functions: 0eaxsinhxxdx=arccoth(a) 0eaxsinxxdx=π2arctan(a)=arccot(a) If we put a=1 and maintain b,c as free parameters we have: 0exsinhbxxdx=arctanh(b) 0exsincxxdx=arctan(c) From the above we have the particular value: 0exsinxxdx=π4

Proof

Differtiating under the integral sign:
dda0eaxsinhbxxdx=0eaxsinhbxdx={Lsinhbx}(a)
Hence
{Lsinhbx}(a)=0eax(ebxebx2)dx=120ex(ab)dx120ex(a+b)dx=12(1ab1a+b)
therefore
0eaxsinhbxxdx=12(1ab1a+b)da+C=ln(a+bab)+C
Taking the limit as b0sinbx0C=0
0eaxsinhbxxdx=ln(a+bab)
For the Lapalce transform of the Sinc function:

Recall that
sinx=isinhix
Therefore if we allow b=ic :
0eaxsinhicxxdx=ln(a+icaic)i0eaxsinhicxxdx=iln(a+icaic)
Hence 0eaxsincxxdx=iln(a+icaic)

Now we will show a particular value:

Recall that in the complex plane arctan(z)=i2ln(1+iz1iz) Hence if a=1 0exsincxxdx=iln(1+ic1ic)=arctan(c) Moreover, if c=1 0exsinxxdx=iln(1+ic1ic)=arctan(1)=π4 Therefore 0exsinxxdx=π4

Saturday, August 21, 2021

Laplace transform II

Another beautiful Laplace transform

Laplace transform related to an infinite product


We show the solution of this integral posted on Twitter by @diegorattaggi. The solution turned out to be the logarithm of the golden ratio. We also find a general formulation for this integral wich turned out to be a Laplace transform with a beutiful solution. 0exe2xe3x+e4xx(1e5x)dx=ln(1+52) As a corollary we have the Laplace transform: 0esx[exe2xe3x+e4xx(1e5x)]dx=ln[Γ(1+s5)Γ(4+s5)Γ(2+s5)Γ(3+s5)] Moreover, the integral and the Laplace transform can be further generalized:

If a1+a2+...+an=b1+b2+...+bn=c with ai,biN{0}
Hence 0ea1x+ea2x+...+eanxeb1xeb2x...ebnxx(1ecx)dx=ln(Γ(a1c)Γ(a2c)Γ(anc)Γ(b1c)Γ(b2c)Γ(bnc)) 0esx[ea1x+ea2x+...+eanxeb1xeb2x...ebnxx(1ecx)]dx=ln(Γ(a1+sc)Γ(a2+sc)Γ(an+sc)Γ(b1+sc)Γ(b2+sc)Γ(bn+sc))

Proof
We show the proof of the particular case, the generalization can be shown in a similar manner.

Consider the Laplace transform:
0esx[exe2xe3x+e4xx(1e5x)]dx Then
dds0esx[exe2xe3x+e4xx(1e5x)]dx=0esx[exe2xe3x+e4x(1e5x)]dx Hence
0esx[exe2xe3x+e4x(1e5x)]dx=0esxn=0e5nx(exe2xe3x+e4x)dx=n=00ex(s+5n)(exe2xe3x+e4x)dx=n=00[ex(s+5n+1)ex(s+5n+2)ex(s+5n+3)+ex(s+5n+4)]dx=n=0[0ex(s+5n+1)dx0ex(s+5n+2)dx0ex(s+5n+3)dx+0ex(s+5n+4)dx]=n=00[ex(s+5n1)ex(s+5n2)ex(s+5n3)+ex(s+5n4)]dx=n=0[1s+5n+11s+5n+21s+5n+3+1s+5n+4] Then
0esx[exe2xe3x+e4xx(1e5x)]dx=0esx[exe2xe3x+e4x(1e5x)]dxds+C=n=0[1s+5n+11s+5n+21s+5n+3+1s+5n+4]ds+C=n=0[1s+5n+1ds1s+5n+2ds1s+5n+3ds+1s+5n+4ds]+C=n=0[ln(s+5n+1)ln(s+5n+2)ln(s+5n+3)+ln(s+5n+4)]+C=n=0ln[(s+5n+1)(s+5n+4)(s+5n+2)(s+5n+3)]+C Therefore
If s0 0exe2xe3x+e4xx(1e5x)dx=n=0ln[(5n+1)(5n+4)(5n+2)(5n+3)]+C If s 0=limsn=0ln[(s+5n+1)(s+5n+4)(s+5n+2)(s+5n+3)]+C=n=0ln[lims(s+5n+1)(s+5n+4)(s+5n+2)(s+5n+3)]+C=n=0ln(1)+CC=0 Then
0exe2xe3x+e4xx(1e5x)dx=n=0ln[(5n+1)(5n+4)(5n+2)(5n+4)]=ln[n=0(5n+1)(5n+4)(5n+2)(5n+3)] Recall the following infinite product representation for a quotient of products of Gamma functions (proof in the Appendix):

If mj=1aj=mj=1bj then: k=0(a1+k)(a2+k)(am+k)(b1+k)(b2+k)(bm+k)=Γ(b1)Γ(b2)Γ(bm)Γ(a1)Γ(a2)Γ(am) Hence, we can write:
n=0(5n+1)(5n+4)(5n+2)(5n+3)=n=0(n+15)(n+45)(n+25)(n+35)=Γ(25)Γ(35)Γ(15)Γ(45)=21+5 Therefore
0exe2xe3x+e4xx(1e5x)dx=ln(1+52) As a corollary we have the Laplace transform:
0esx[exe2xe3x+e4xx(1e5x)]dx=ln[Γ(1+s5)Γ(4+s5)Γ(2+s5)Γ(3+s5)]

Appendix: Euler's infinite product representation for a quotient of Gamma functions

Recall the Euler's definition of the Gamma function Γ(z)=limn(n+1)zn!z(z+1)(z+2)(z+n) Supposse that mj=1aj=mj=1bj Hence Γ(b1)Γ(b2)Γ(bm)Γ(a1)Γ(a2)Γ(am)=limnmj=1[(n+1)bjajnk=0aj+kbj+k]=limn(n+1)mj=1ajmj=1bjmj=1nk=0aj+kbj+k=limnnk=0mj=1aj+kbj+k Therefore: k=0(a1+k)(a2+k)(am+k)(b1+k)(b2+k)(bm+k)=Γ(b1)Γ(b2)Γ(bm)Γ(a1)Γ(a2)Γ(am) if mj=1aj=mj=1bj

Laplace transform I

Laplace transform related to Harmonic numbers

What is the Laplace transform of exln(1ex)


We repost here the solution of this problem posted by @SrinivasR1729 on Twitter.
0eaxexln(1ex)dx=Ha+1a+1

Proof 0eaxexln(1ex)dx=10(1w)aln(w)dw(w1ex)=10(1w)a(ddt|t=0+wt)dw=ddt|t=0+10wt(1w)adw=ddt|t=0+B(t+1,a+1) Recall that the derivative of the complete beta function: ddaB(a,b)=B(a,b)(ψ(0)(a)ψ(0)(a+b)) Therefore: 0eaxexln(1ex)dx=ddt|t=0+B(t+1,a+1)=limt0+(ψ0(t+1)ψ0(a+t+2))B(t+1,a+1)=(ψ0(1)ψ0(a+2))B(1,a+1)=(γψ0(a+2))B(1,a+1) where we used the fact that ψ(0)(1)=γ, the Euler constant.

Recall that a hamonic number for real and complex values can be expressed with the following relation: Hx=ψ(0)(x+1)+γ Also note that B(1,a+1)=Γ(1)Γ(a+1)Γ(a+2)=Γ(a+1)Γ(a+1)(a+1)=1a+1 Hence 0eaxexln(1ex)dx=(γψ0(a+2))B(1,a+1)=Ha+1a+1 0eaxexln(1ex)dx=Ha+1a+1

Thursday, August 19, 2021

Residue theorem IV

Residue theorem for winding numbers

Beautiful integral involving a product of cosine functions



Today we show the proof of this beautiful result posted by @SrinivasR1729 on Twitter:

2π0cos(1a+x)cos(x)dx=πcos(1a) 4π0cos(1a+x)cos(1a2+x)cos(2x)=πcos(a+1a2) 8π0cos(1a+x)cos(1a2+x)cos(1a3+x)cos(3x)=πcos(a2+a+1a3) 16π0cos(1a+x)cos(1a2+x)cos(1a3+x)cos(1a4+x)cos(3x)=πcos(a3+a2+a+1a4) 2mπ0[cos(1a+x)cos(1a2+x)cos(1am+x)]cos(mx)=πcos(am1+am2+am3+...+1am)
Proof

Recall that under the assumptions of the residue theorem with winding numbers, if γ is a closed curve: γfdz=2πini=1{Res(f,zi)I(γ,zi)} where I(γ,zi) is the winding number of γ with respect to zi.

Also recall that the circle γ(t)=eit0t2πn has an index I(γ,0)=n with respect to 0

Case n=1 (the circle winds one time around 0)

The fact that x varies from 0 to 2π suggest we can consider x as an argument of a point z on the unit circle C centered at the origin; hence we write z=eix0x2πdx=dziz
Then
I=2π0cos(1a+x)cos(x)=2π0(ei(1a+x)+ei(1a+x)2)(eix+eix2)dx=2π0(eiaeix+eiaeix2)(eix+eix2)dx=122C(eiaz+eiaz)(z+1z)izdz=122iC{eiaz+1z(eia+eia)+eiaz3}dz Hence, if g(z)=eiaz+1z(eia+eia)+eiaz3
Res(g,0)=eia+eia Therefore z=0 is the only singularity of g and: I=2π0cos(1a+x)cos(x)dx=2πi(122i)Res(g,0)I(C,0)=1=π2(eia+eia)=πcos(1a) 2π0cos(1a+x)cos(x)dx=πcos(1a)

Case n = 2 (the circle winds 2 times around 0)
I=4π0cos(1a+x)cos(1a2+x)cos(2x)dx=4π0(ei(1a+x)+ei(1a+x)2)(ei(1a2+x)+ei(1a2+x)2)(ei2x+ei2x2)dx=4π0(eiaeix+eiaeix2)(eia2eix+eia2eix2)(ei2x+ei2x2)dx=123iC(eiaz+eiaz)(eia2z+eia2z)(z2+1z2)zdz(z=eix)=123iC(eiaz2+eia)(eia2z2+eia2)(z4+1)z5g(z)dz It is easy to see that Res(g,0)=eiaeia2+eiaeia2=ei(1a+1a2)+ei(1a+1a2) Therefore, again z=0 is the only singularity of g and I=4π0cos(1a+x)cos(1a2+x)cos(2x)dx=2πi(123i)Res(g,0)I(C,0)=2=22πi(123i)(eiaeia2+eiaeia2)=π2[ei(1a+1a2)+ei(1a+1a2)]=πcos(1a+1a2)=πcos(a+1a2) 4π0cos(1a+x)cos(1a2+x)cos(2x)=πcos(a+1a2)

Case n = m (the circle winds 2m1 times around 0)

I=2mπ0(mj=1cos(1aj+x))cos(mx)=2mπ0(mj=1ei(1aj+x)+ei(1aj+x)2)(eimx+eimx2)dx=2m12π0(mj=1eiajeix+eiajeix2)(eimx+eimx2)dx=12m+1C[mj=1(eiajz+eiajz)](zm+1zm)dziz(z=eix)=1i2m+1C[mj=1(eiajz2+eiaj)](z2m+1)z2m+1dz Therefore if g(z) = \frac{\left[\prod_{j=1}^{m} \left(e^{\frac{i}{a^j}}z^2 +e^{-\frac{i}{a^j}}\right)\right]\left(z^{2m} + 1\right)}{z^{2m+1}} Then \operatorname{Res}(g,0) = \prod_{j=1}^{m} e^{\frac{i}{a^j}} + \prod_{j=1}^{m} e^{-\frac{i}{a^j}} = e^{i(\frac{1}{a}+\frac{1}{a^2}+...+\frac{1}{a^m})} + e^{-i(\frac{1}{a}+\frac{1}{a^2}+...+\frac{1}{a^m})} Hence \begin{align*} I= \int_{0}^{2^m \pi } \left(\prod_{j=1}^{m} \cos\left( \frac{1}{a^j}+x\right)\right)\cos(mx) =& 2 \pi i \left( \frac{1}{i2^{m+1}}\right) \operatorname{Res}(g,0) \underbrace{I(C,0)}_{=2^{m-1}}\\ =& \frac{\pi}{2}\left(e^{i(\frac{1}{a}+\frac{1}{a^2}+...+\frac{1}{a^m})} + e^{-i(\frac{1}{a}+\frac{1}{a^2}+...+\frac{1}{a^m})}\right)\\ =& \pi \cos\left(\frac{1}{a}+\frac{1}{a^2}+...+\frac{1}{a^m}\right)\\ =& \pi \cos\left(\frac{a^{m-1}+a^{m-2}+a^{m-3}+...+1}{a^m}\right) \end{align*} Hence \boxed{\int_{0}^{2^m \pi } \left[\cos\left( \frac{1}{a}+x\right)\cos\left( \frac{1}{a^2}+x\right)\cdots\cos\left( \frac{1}{a^m}+x\right)\right]\cos(mx) = \pi \cos\left(\frac{a^{m-1}+a^{m-2}+a^{m-3}+...+1}{a^m}\right)}

Saturday, August 14, 2021

Feynman's trick IX

Feynman's trick IX and Mellin transform

Mellin transform of \sin x and \cos x



We show the proof of the following result posted by @infseriesbot \int_{0}^{\infty} \frac{(\sin x - \cos x) \ln x}{\sqrt{x}} dx = \frac{\pi^{\frac{3}{2}}}{\sqrt{2}}

Proof

First we can start differentiating under the integral sign: \int_{0}^{\infty} \frac{(\sin x - \cos x) \ln x}{\sqrt{x}} dx = \int_{0}^{\infty} \frac{(\sin x - \cos x) \left(\frac{d}{dt}\Big|_{t=0+} x^t\right) }{\sqrt{x}} dx = \frac{d}{dt}\Big|_{t=0+} \int_{0}^{\infty} x^{t+\frac{1}{2}-1}(\sin x - \cos x) dx The right hand side is the derivative of the Mellin transform of the function \displaystyle (\sin x - \cos x) at \displaystyle t+\frac{1}{2}: We can split the Mellin into two, the Mellin of \sin x and \cos x: \left\{ \mathscr{M} \sin x \right\} \left(t+\frac{1}{2}\right) = \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\sin x dx \left\{ \mathscr{M} \cos x \right\} \left(t+\frac{1}{2}\right) = \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\cos x dx Now we can expand \sin x and \cos x using the Euler's formula and the de Moivre theorem in order to apply the Ramanujan's master theorem: \sin(w) = \frac{e^{ix}-e^{-ix}}{2i} = \sum_{n=0}^{\infty} \frac{(ix)^n}{2in!} -\sum_{n=0}^{\infty} \frac{(-ix)^n}{2in!}=\sum_{n=0}^{\infty} \frac{(-x)^n\left[(-i)^n-(i)^n\right]}{2in!} By de Moivre's theorem (-i)^{n} = -(e^{-\frac{i\pi}{2}})^{n} = \cos\left(-\frac{n\pi}{2}\right)+i\sin\left(-\frac{n\pi}{2}\right) = \cos\left(\frac{n\pi}{2}\right)-i\sin\left(\frac{n\pi}{2}\right) i^{n} = (e^{\frac{i\pi}{2}})^{n} = \cos\left(\frac{n\pi}{2}\right)+i\sin\left(\frac{n\pi}{2}\right) Then \frac{\left[(-i)^n-(i)^n\right]}{2i} = \frac{-2i\sin\left(\frac{n\pi}{2}\right)}{2i} = -\sin\left(\frac{n\pi}{2}\right) Therefore \sin x = \sum_{n=0}^{\infty}\frac{(-x)^n[-\sin\left(\frac{n\pi}{2}\right)]}{n!}dw In a similar way we can show: \cos x = \sum_{n=0}^{\infty}\frac{(-x)^n[\cos\left(\frac{n\pi}{2}\right)]}{n!}dw Applying the Ramanujan's master theorem and using the fact that \sin is odd and \cos is even \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\sin x dx = \Gamma\left(t+\frac{1}{2}\right)\sin\left(\frac{t\pi}{2}+\frac{\pi}{4}\right) \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\cos x dx = \Gamma\left(t+\frac{1}{2}\right)\cos\left(\frac{t\pi}{2}+\frac{\pi}{4}\right) Therefore: \begin{align*} \frac{d}{dt}\Big|_{t=0+} \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\sin x dx = & \frac{d}{dt}\Big|_{t=0+} \Gamma\left(t+\frac{1}{2}\right)\sin\left(\frac{t\pi}{2}+\frac{\pi}{4}\right) \\ =& \lim_{t \to 0+} \frac{1}{2} -\Gamma\left(t+\frac{1}{2}\right)\left[ \pi \sin \left(\frac{\pi t}{2} +\frac{\pi}{4}\right) +2\cos\left(\frac{\pi t}{2} + \frac{\pi}{4}\right) \psi^{(0)}\left(t+\frac{1}{2}\right)\right]\\ =& \frac{\pi^{\frac{3}{2}}}{2\sqrt{2}} + \sqrt{\frac{\pi}{2}} \psi^{(0)}\left(\frac{1}{2}\right) \end{align*} \begin{align*} \frac{d}{dt}\Big|_{t=0+} \int_{0}^{\infty} x^{t+\frac{1}{2}-1}\cos x dx = & \frac{d}{dt}\Big|_{t=0+} \Gamma\left(t+\frac{1}{2}\right)\cos\left(\frac{t\pi}{2}-\frac{\pi}{4}\right) \\ =& \lim_{t \to 0+} \frac{1}{2} \Gamma\left(t+\frac{1}{2}\right)\left[2\cos\left(\frac{\pi t}{2} + \frac{\pi}{4}\right)\psi^{(0)}\left(t+\frac{1}{2}\right)- \pi \sin \left(\frac{\pi t}{2} +\frac{\pi}{4} \right)\right]\\ =& \sqrt{\frac{\pi}{2}} \psi^{(0)}\left(\frac{1}{2}\right) - \frac{\pi^{\frac{3}{2}}}{2\sqrt{2}} \end{align*} Hence, we can conclude: \int_{0}^{\infty} \frac{(\sin x - \cos x) \ln x}{\sqrt{x}} dx = \frac{d}{dt}\Big|_{t=0+} \int_{0}^{\infty} x^{t+\frac{1}{2}-1}(\sin x - \cos x) dx = \frac{\pi^{\frac{3}{2}}}{\sqrt{2}} \boxed{\int_{0}^{\infty} \frac{(\sin x - \cos x) \ln x}{\sqrt{x}} dx = \frac{\pi^{\frac{3}{2}}}{\sqrt{2}}}

Thursday, August 12, 2021

Pochhammer polynomials

Series involving double factorial

Series expansion of \displaystyle \frac{1}{\sqrt{1-x}}


We prove the following result posted by @infseriesbot
\frac{1}{\sqrt{1-x}} = \sum_{n=0}^{\infty} \frac{(2n-1)!!}{(2n)!!} x^n

Proof

\frac{1}{\sqrt{1-x}} = (1-x)^{-\frac{1}{2}} Recall that Newton extended the binomial theorem to allow real and complex numbers as exponents: (x+y)^r = \sum_{k=0}^{\infty} \binom{r}{k}x^{r-k}y^k \quad |x|>|y| where \binom{r}{k} = \frac{(r)_{k}}{k!} and \displaystyle (r)_{k} is the Pochhammer symbol (falling factorial): (r)_{k} = r(r-1)\cdots(r-k+1) Hence: \begin{align*} (1-x)^{-\frac{1}{2}} =& \sum_{n=0}^{\infty}\binom{-\frac{1}{2}}{n} (-1)^n x^{n} \\ =& \sum_{n=0}^{\infty} \frac{\left(-\frac{1}{2}\right)_{n}}{n!} (-1)^n x^{n} \\ \end{align*} One of the properties of the rising and falling factorial states that r^{(n)} = (-1)^n(r)_{n} where \displaystyle r^{(n)} is the rising factorial: r^{(n)} = r(r+1)(r+2)\cdots (x+n-1) Hence \sum_{n=0}^{\infty} \frac{\left(-\frac{1}{2}\right)_{n}}{n!} (-1)^n x^{n} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}}{n!} x^{n} Note that we can express the factorial as a the pochhammer polynomial of (1)^{(n)}: (1)^{(n)} = 1\cdot 2\cdot 3\cdots(n-1)n = n! Hence, \frac{1}{\sqrt{1-x}} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}}{n!} x^{n} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}}{(1)^{(n)}} x^{n} On the other hand, double factorials can be expressed with the following form: (2n-1)!! = 2^n \frac{\Gamma\left(\frac{1}{2}+n\right)}{\sqrt{\pi}} = 2^n \frac{\Gamma\left(\frac{1}{2}+n\right)}{\Gamma\left(\frac{1}{2}\right)} (2n)!! = 2^n n! = 2^n \Gamma(n+1) = 2^n \frac{\Gamma(n+1)}{\Gamma(1)} while the pochhammer polynomial can be expressed as the quotient of two gamma functions: r^{(n)} = \frac{\Gamma(n+r)}{\Gamma(r)} Hence, we can write: (2n-1)!!= 2^n \left(\frac{1}{2}\right)^{(n)} (2n)!! = 2^n (1)^{(n)} Therefore \frac{(2n-1)!!}{(2n)!!} = \frac{2^n \left(\frac{1}{2}\right)^{(n)}}{2^n (1)^{(n)}} = \frac{\left(\frac{1}{2}\right)^{(n)}}{(1)^{(n)}} Hence, we can conclude \frac{1}{\sqrt{1-x}} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}}{(1)^{(n)}} x^{n} = \sum_{n=0}^{\infty}\frac{(2n-1)!!}{(2n)!!} x^{n} This can also be expressed with the notation of the generalized hypergeometric function as \frac{1}{\sqrt{1-x}} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}}{(1)^{(n)}} x^{n} = \sum_{n=0}^{\infty} \frac{\left(\frac{1}{2}\right)^{(n)}(1)^{(n)}}{(1)^{(n)}} \frac{x^{n}}{n!} = {}_{2}F_{1}\left(\frac{1}{2},1;1;x\right) However, in the hypergeometric function notation it is common to use (x)_{n} to denote the rising factorial instead of the falling factorial, so it is very important to always check if we are referring to rising of falling factorials. \boxed{\frac{1}{\sqrt{1-x}} = \sum_{n=0}^{\infty}\frac{(2n-1)!!}{(2n)!!} x^{n}}

Wednesday, August 11, 2021

Generalized hypergeometric function

Integral found on Twitter

Solution to integral involving {}_{3}F_{2}


We show the proof of the following result posted by @SrinivasR1729 \int_{0}^{\infty} \frac{\ln(1+e^{-x})}{\sqrt{1-e^{-x}}}e^{-\frac{x}{2}}dx = 2\pi \ln\left(\frac{1}{2}+\frac{1}{\sqrt{2}}\right)

Proof

Recall that {}_{p}F_{q}(a_{1},a_{2},...,a_{p};b_{1},b_{2},...,b_{q};x) = \sum_{j=0}^{\infty} \frac{(a_{1})(a_{2})\cdots(a_{p})}{(b_{1})(b_{2})\cdots(b_{q})} \frac{x^{j}}{j!} where the symbol \displaystyle (x)_{j} refers to the Pochhammer polynomial (also called rising factorial): (x)_{j} = x(x+1)(x+2)\cdots(x+j-1) Hence, \begin{align*} \int_{0}^{\infty} \frac{\ln(1+e^{-x})}{\sqrt{1-e^{-x}}}e^{-\frac{x}{2}}dx =& \int_{0}^{1} \frac{\ln(1+w)}{\sqrt{1-w}\sqrt{w}} dw \quad (w \mapsto e^{-w})\\ =& \int_{0}^{1} \sum_{k=1}^{\infty} \frac{(-1)^{k+1} w^k}{k} \frac{1}{\sqrt{1-w}\sqrt{w}} dw\\ =& \sum_{k=1}^{\infty} \frac{(-1)^{k+1}}{k} \int_{0}^{1} w^{k-\frac{1}{2}}(1-w)^{-\frac{1}{2}} dw\\ =& \sum_{k=1}^{\infty} \frac{(-1)^{k+1}}{k} B\left(k+\frac{1}{2},\frac{1}{2}\right)\\ =& \sum_{j=0}^{\infty} \frac{(-1)^{j}}{j+1} B\left(j+\frac{3}{2},\frac{1}{2}\right) \quad (k=j+1)\\ =& \sum_{j=0}^{\infty} \frac{(-1)^{j}}{j+1} \frac{\Gamma\left(j+\frac{3}{2}\right)\Gamma\left(\frac{1}{2}\right)}{\Gamma(j+2)}\\ =&\sum_{j=0}^{\infty} \frac{(-1)^{j}}{j+1} \frac{\Gamma\left(\frac{3}{2}\right)\Gamma\left(\frac{1}{2}\right)\left(\frac{3}{2}\right)_{j}}{\Gamma(2) (2)_{j}} \quad \left(\Gamma(t+j) = \Gamma(t)(t)_{j}\right)\\ =& \frac{\pi}{2}\sum_{j=0}^{\infty} \frac{(-1)^{j}}{j+1} \frac{\left(\frac{3}{2}\right)_{j}}{(2)_{j}} \quad \left(\Gamma\left(\frac{1}{2}\right)=\sqrt{\pi}, \; \Gamma\left(\frac{3}{2}\right)=\frac{\sqrt{\pi}}{2}\right)\\ =&\frac{\pi}{2} \sum_{j=0}^{\infty} (-1)^{j} \frac{\left(\frac{3}{2}\right)_{j} (1)_{j} }{(2)_{j} (2)_{j}} \quad \left( \frac{(1)_{j}}{(2)_{j}}=\frac{1}{1+j}\right) \\ =&\frac{\pi}{2} \sum_{j=0}^{\infty} (-1)^{j} \frac{\left(\frac{3}{2}\right)_{j} (1)_{j} (1)_{j} }{(2)_{j} (2)_{j} (1)_{j}} \\ =&\frac{\pi}{2}\sum_{j=0}^{\infty} \frac{\left(\frac{3}{2}\right)_{j} (1)_{j} (1)_{j} }{(2)_{j} (2)_{j} } \frac{(-1)^{j}}{j!} \quad (j! = (1)_{j}) \\ =&\frac{\pi}{2}{}_{3}F_{2}\left(\frac{3}{2},1,1;2,2;-1\right) \\ =&2\pi \ln\left(\frac{1}{2}+\frac{1}{\sqrt{2}}\right) \end{align*} Therefore \boxed{\int_{0}^{\infty} \frac{\ln(1+e^{-x})}{\sqrt{1-e^{-x}}}e^{-\frac{x}{2}}dx = 2\pi \ln\left(\frac{1}{2}+\frac{1}{\sqrt{2}}\right)}

Hurwitz-zeta function

Hurwitz zeta function

Integral representation of the Hurwitz-zeta function


We show the proof of the following result posted by @integralsbot: \int_{0}^{1} \frac{x^n \ln^{a-1} \frac{1}{x} }{1-x} dx = \Gamma(a) \left( \zeta (a)- H_{n}^{(a)} \right) This result is the integral representation of the Hurwitz-zeta function. Recall that: \zeta(a,n+1) = \left( \zeta (a)- H_{n}^{(a)} \right) The proof for the case a \in \mathbb{R} is well known, so here we show a proof for a \in \mathbb{N} using the Feynman's trick just for the sake to give a different proof.

Proof

For \displaystyle a\in \mathbb{N}
\begin{align*} \int_{0}^{1} \frac{x^n \ln^{a-1} \frac{1}{x} }{1-x} dx = & \int_{0}^{1} \frac{x^n \left(\frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+} \frac{1}{x^t}\right) }{1-x} dx\\ =& \frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+}\int_{0}^{1} \frac{x^{n-t} }{1-x} dx\\ =& \frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+}\int_{0}^{1} \sum_{k=0}^{\infty}x^kx^{n-t} dx\\ =& \frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+} \sum_{k=0}^{\infty} \int_{0}^{1}x^{k+n-t} dx\\ =& \frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+} \sum_{k=0}^{\infty} \frac{1}{(k+n+1-t)} \\ =& \sum_{k=0}^{\infty} \left[\frac{d^{a-1}}{dt^{a-1}}\Big|_{t=0+} \frac{1}{(k+n+1-t)} \right] \\ =& \sum_{k=0}^{\infty} \left[\lim_{t \to 0+} (a-1)! \frac{1}{(k+n+1-t)^a} \right] \\ =& (a-1)! \sum_{k=0}^{\infty} \frac{1}{(k+n+1)^a} \\ =& (a-1)! \zeta(a,n+1) \\ =& \Gamma(a) \left( \zeta (a)- H_{n}^{(a)} \right) \\ \end{align*} \boxed{ \int_{0}^{1} \frac{x^n \ln^{a-1} \frac{1}{x} }{1-x} dx =\Gamma(a) \left( \zeta (a)- H_{n}^{(a)} \right) }

Golden ratio

Integral involving the Golden ratio

Integral that relates the secant function and the Golden ratio


Today we repost here the solution to this integral posted on Twitter by @integralsbot: \int_{0}^{\infty} \frac{x^{\frac{\pi}{5}-1}}{1+x^{2\pi}} dx = \phi In the proof we made use of the integral representation of the \sec x function, a trick that we have used extensively in other occasions (link 1, link 2)

Proof


I= \int_{0}^{\infty} \frac{x^{\frac{\pi}{5}-1}}{1+x^{2\pi}} dx = \frac{1}{\pi}\int_{0}^{\infty} \frac{y^{\frac{4}{5}}}{1+y^2} dw \quad (y \mapsto x^{\pi}) Recall the integral representation of \displaystyle \sec(w) (proof below): \sec(w) = \frac{2}{\pi} \int_{0}^{\infty} \frac{y^{\frac{2w}{\pi}}}{1+y^2} dy Therefore if \displaystyle w= \frac{2\pi}{5} I = \frac{1}{\pi}\int_{0}^{\infty} \frac{y^{\frac{4}{5}}}{1+y^2} dy = \frac{\sec\left(\frac{2\pi}{5}\right)}{2} = \frac{1+\sqrt{5}}{2} = \phi Hence \boxed{\int_{0}^{\infty} \frac{x^{\frac{\pi}{5}-1}}{1+x^{2\pi}} dx = \phi }

Appendix

We reproduce here the proof of the integral representation of the sec function again:
The proof can be obtained trough contour integration in a branch cut but here we will use a more straightforward approach with the help of the Ramanujan's master theorem

Let |a|<1

Recall \frac{1}{1+y^2} = \int_{0}^{1} t^{y^2} dt Therefore \begin{align*} \int_{0}^{\infty}\frac{y^a}{1+y^2}dy =& \int_{0}^{\infty}\int_{0}^{1} y^a t^{y^2} dtdy\\ =&\int_{0}^{1} \int_{0}^{\infty} y^a t^{y^2} dydt\\ =& \frac{1}{2}\int_{0}^{1} \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dwdt \quad (w \mapsto y^2) \end{align*} The integral \displaystyle \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dw is the Mellin transform of \displaystyle t^w at \displaystyle \frac{a+1}{2} Recall that t^w = e^{w\ln(t)} = \sum_{n=0}^{\infty} \frac{\ln^n(t)w^n}{n!} = \sum_{n=0}^{\infty} \frac{(-\ln(t))^n(-w^n)}{n!} Therefore, by Ramanujan's master theorem \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dw = \Gamma\left(\frac{a+1}{2}\right)\left[-\ln(t)\right]^{-\frac{a+1}{2}} Therefore \begin{align*} \frac{1}{2}\int_{0}^{1} \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dwdt =& \frac{1}{2}\int_{0}^{1} \Gamma\left(\frac{a+1}{2}\right)\left[-\ln(t)\right]^{-\frac{a+1}{2}} dt\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\int_{0}^{1}\left[-\ln(t)\right]^{-\frac{a+1}{2}} dt\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\int_{0}^{1}v^{-\frac{a+1}{2}}e^{-v} dv \quad (v\mapsto -ln(t))\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\Gamma\left(\frac{1-a}{2}\right) \end{align*} Finally, recall the Euler's reflection formula \Gamma(z)\Gamma(1-z)= \frac{\pi}{\sin \pi z} = \pi \csc \pi z Therefore \int_{0}^{\infty}\frac{y^a}{1+y^2}dy = \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\Gamma\left(\frac{1-a}{2}\right) = \frac{\pi}{2} \csc\left(\pi \frac{1+a}{2}\right)=\frac{\pi}{2} \sec\left(\pi \frac{a}{2}\right) If we let \displaystyle x = \pi \frac{a}{2} then \displaystyle a=\frac{2x}{\pi}. Then \boxed{\frac{2}{\pi}\int_{0}^{\infty}\frac{y^\frac{2x}{\pi}}{1+y^2}dy = \sec x}

Saturday, August 7, 2021

Catalan constant

Integral related to Catalan constan

Integral involving Catalan constant


Today we show the proof of the following result found on Twitter: \int_{0}^{\frac{\pi}{2}} \int_{0}^{\infty} \frac{1}{\sqrt[y]{x^{\pi}}+1} dx dy= 2\beta(2)

Proof

I = \int_{0}^{\frac{\pi}{2}} \int_{0}^{\infty} \frac{1}{\sqrt[y]{x^{\pi}}+1} dxdy = \int_{0}^{\frac{\pi}{2}} y\frac{2}{\pi}\int_{0}^{\infty} \frac{w^{\frac{2y}{\pi}-1}}{w^2+1} dwdy \quad (w \mapsto x^{\frac{\pi}{2y}}) Recall the integral representation fo \sec(x) (proof in the Appendix): \sec(x) = \frac{2}{\pi}\int_{0}^{\infty} \frac{t^{\frac{2x}{\pi}}}{t^2+1}dt Hence \frac{2}{\pi}\int_{0}^{\infty} \frac{w^{\frac{2y}{\pi}-1}}{w^2+1} dw = \sec\left(y-\frac{\pi}{2}\right) = \csc(y) = \frac{1}{\sin(y)} Therefore I = \int_{0}^{\frac{\pi}{2}} \frac{y}{\sin(y)}dy = 2\int_{0}^{\infty} \operatorname{arccot}(e^w) dw \quad \left(w \mapsto \ln\left(\cot\left(\frac{y}{2}\right)\right)\right) Recall the series expansion for \operatorname{arccot}(x) \sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)} x^{-2n-1} Hence \begin{align*} I =& 2\int_{0}^{\infty} \operatorname{arccot}(e^w) dw \\ =& 2\int_{0}^{\infty} \sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)} e^{-w(2n+1)}dw\\ =& 2\sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)}\int_{0}^{\infty} e^{-w(2n+1)}dw\\ =& 2\sum_{n=0}^{\infty} \frac{(-1)^n}{(2n+1)^2}\\ =& 2\beta(2) \end{align*} \boxed{\therefore \int_{0}^{\frac{\pi}{2}} \int_{0}^{\infty} \frac{1}{\sqrt[y]{x^{\pi}}+1} dxdy = 2\beta(2)}

Appendix: Integral representation of \sec x

How do we know that \displaystyle \sec x= \frac{2}{\pi}\int_{0}^{\infty}\frac{y^\frac{2x}{\pi}}{1+y^2}dy ?

The proof can be obtained trough contour integration in a branch cut but here we will use a more straigthforward approach with the help of the Ramanujan's master theorem

Let |a|<1
Recall \frac{1}{1+y^2} = \int_{0}^{1} t^{y^2} dt Therefore \begin{align*} \int_{0}^{\infty}\frac{y^a}{1+y^2}dy =& \int_{0}^{\infty}\int_{0}^{1} y^a t^{y^2} dtdy\\ =&\int_{0}^{1} \int_{0}^{\infty} y^a t^{y^2} dydt\\ =& \frac{1}{2}\int_{0}^{1} \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dwdt \quad (w \mapsto y^2) \end{align*} The integral \displaystyle \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dw is the Mellin transform of \displaystyle t^w at \displaystyle \frac{a+1}{2}
Recall that t^w = e^{w\ln(t)} = \sum_{n=0}^{\infty} \frac{\ln^n(t)w^n}{n!} = \sum_{n=0}^{\infty} \frac{(-\ln(t))^n(-w^n)}{n!} Therefore, by Ramanujan's master theorem \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dw = \Gamma\left(\frac{a+1}{2}\right)\left[-\ln(t)\right]^{-\frac{a+1}{2}} Therefore \begin{align*} \frac{1}{2}\int_{0}^{1} \int_{0}^{\infty} w^{\frac{a}{2}-\frac{1}{2}} t^{w} dwdt =& \frac{1}{2}\int_{0}^{1} \Gamma\left(\frac{a+1}{2}\right)\left[-\ln(t)\right]^{-\frac{a+1}{2}} dt\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\int_{0}^{1}\left[-\ln(t)\right]^{-\frac{a+1}{2}} dt\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\int_{0}^{1}v^{-\frac{a+1}{2}}e^{-v} dv \quad (v\mapsto -ln(t))\\ =& \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\Gamma\left(\frac{1-a}{2}\right) \end{align*} Finally, recall the Euler's reflection formula \Gamma(z)\Gamma(1-z)= \frac{\pi}{\sin \pi z} = \pi \csc \pi z Therefore \int_{0}^{\infty}\frac{y^a}{1+y^2}dy = \frac{1}{2}\Gamma\left(\frac{a+1}{2}\right)\Gamma\left(\frac{1-a}{2}\right) = \frac{\pi}{2} \csc\left(\pi \frac{1+a}{2}\right)=\frac{\pi}{2} \sec\left(\pi \frac{a}{2}\right) If we let \displaystyle x = \pi \frac{a}{2} then \displaystyle a=\frac{2x}{\pi}. Then \boxed{\frac{2}{\pi}\int_{0}^{\infty}\frac{y^\frac{2x}{\pi}}{1+y^2}dy = \sec x}

Monday, August 2, 2021

Taylor's theorem

Arcsin expansion

Series related to the power series of \arcsin(x)


We show the proof of the following result. The proof relies on the series expansion of arcsin. \sum_{n=1}^{\infty} \frac{(2n-2)!}{2^{2n}(n!)^2 } = 1-\ln(2) Proof

Lets start with the expansion series of \arcsin(x): \sum_{n=0}^{\infty} \frac{(2n)!}{2^{2n}(n!)^2 (2n+1)}x^{2n+1} = \arcsin(x) \quad \quad |x|\leq 1 Taking the derivative \sum_{n=0}^{\infty} \frac{(2n)!}{2^{2n}(n!)^2 }x^{2n} = \frac{d}{dx}\arcsin(x) = \frac{1}{\sqrt{1-x^2}} Multiplying by \frac{1}{x^2} and subtracting -\frac{1}{x^2} \sum_{n=1}^{\infty} \frac{(2n)!}{2^{2n}(n!)^2 }x^{2n-2} = \frac{1}{x^{2}\sqrt{1-x^2}} -\frac{1}{x^2} Integrating \int_{0}^{x} \sum_{n=1}^{\infty} \frac{(2n)!}{2^{2n}(n!)^2 }x^{2n-2} dx = \sum_{n=1}^{\infty} \frac{(2n-2)!(2n)}{2^{2n}(n!)^2 }x^{2n-1}= \int_{0}^{x} \left(\frac{1}{x^{2}\sqrt{1-x^2}} -\frac{1}{x^2} \right)dx Hence \int_{0}^{x} \left(\frac{1}{x^{2}\sqrt{1-x^2}} -\frac{1}{x^2} \right)dx = \left[ -\frac{\sqrt{1-x^2}}{x} +\frac{1}{x} \right]_{0}^{x} = -\frac{\sqrt{1-x^2}}{x} +\frac{1}{x} Therefore \sum_{n=1}^{\infty} \frac{(2n-2)!(2n)}{2^{2n}(n!)^2 }x^{2n-1}= -\frac{\sqrt{1-x^2}}{x} +\frac{1}{x} Integrating one more time from 0 to 1: \int_{0}^{1}\sum_{n=1}^{\infty} \frac{(2n-2)!(2n)}{2^{2n}(n!)^2 }x^{2n-1} dx=\sum_{n=1}^{\infty} \frac{(2n-2)!}{2^{2n}(n!)^2 }= \int_{0}^{1} \left(-\frac{\sqrt{1-x^2}}{x} +\frac{1}{x} \right)dx If x=\sin(w) and dx =\cos(w)dw \int_{0}^{1} \left(-\frac{\sqrt{1-x^2}}{x} +\frac{1}{x}\right)dx = \int_{0}^{\frac{\pi}{2}} \left[ \cot(w) - \cot(w)\cos(w)\right]dw Using the following identities \cos(w)\cot(w) = \csc(w)-\sin(w), \quad \cot(w)-\csc(x) = -\tan\left(\frac{w}{2}\right) \quad \textrm{and} \quad \sin(w)=2\sin\left(\frac{w}{2}\right)\cos\left(\frac{w}{2}\right) \begin{align*} \sum_{n=1}^{\infty} \frac{(2n-2)!}{2^{2n}(n!)^2 } =& \int_{0}^{\frac{\pi}{2}} \left[ \cot(w) - \cot(w)\cos(w)\right]dw \\ = & \int_{0}^{\frac{\pi}{2}} \left[ \cot(w) - \csc(w) +\sin(w) \right]dw \\ = & \int_{0}^{\frac{\pi}{2}} \left[ \cot(w) - \csc(w) +\sin(w) \right]dw \\ = & \int_{0}^{\frac{\pi}{2}} \left[ -\tan\left(\frac{w}{2}\right) +2\sin\left(\frac{w}{2}\right)\cos\left(\frac{w}{2}\right) \right]dw \\ = & \left[ 2\ln(\cos\left(\frac{x}{2}\right)) -2\cos^{2}\left(\frac{x}{2}\right) \right]_{0}^{\frac{\pi}{2}} = 1-\ln(2) \end{align*} Therefore \boxed{\sum_{n=1}^{\infty} \frac{(2n-2)!}{2^{2n}(n!)^2 } = 1-\ln(2) }

Series of the day

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