Loading [MathJax]/jax/output/HTML-CSS/jax.js

Deriving the power series of the generating function of the Bernoulli numbers.

In this thread, we are going to derive the series of xex1. This series is especially important because it is the generating function of the Bernoulli numbers. I have tried many times in vain to derive the series, but with the precious help of members of Mathstack Exchange, I could finally derive it.

The series is especially difficult because the n order derivative is very complex as we progress. For example:

f(x)=xex1f(x)=ex(x1)+1(ex1)2f(0)=1(01)+111=10

We can of course rely on the L'Hopital rule to calculate the limit of x approaches 0 in all n order derivatives but it is a hugely complicated calculation.

Another way must be found to derive the series for this function. The method is to rely on the equality: (xex1)(ex1x)=1

We can easily expand ex1x into its power series by using the series ex=+n=0xnn!

ex1x=1+x+x22!+x33!+...1x=1+x2!+x23!+x34!+...=+n=1xn1n!=+n=0xn(n+1)!

Now, denote the series of xex1=+n=0bnxn and the series of ex1x=+n=0anxn.

The coefficient of the +n=0anxn is an=1(n+1)!

We can rewrite the equation in (1) as followed:

(1+x+x22!+x33!+...)+n=0bnxn=1+0x+0x2+0x3+...

=(+n=0bnxn)(+n=0anxn)=1

We can compare the product of the coefficients bn and an on LHS to the coefficients of the "series" on the RHS in (4) and form this system of equations:

[x0]a0b0=1

[x1]a0b1+a1b0=0

[x2]a0b2+a1b1+a2b0=0

[x3]a0b3+a1b2+a2b1+a3b0=0

[x4]a0b4+a1b3+a2b2+a3b1+a4b0=0

[x5]a0b5+a1b4+a2b3+a3b2+a4b1+a5b0=0

[x6]a0b6+a1b5+a2b4+a3b3+a4b2+a5b1+a6b0=0

Refer to this thread to review what is a Cauchy triangle when multiplying 2 power series:

From the system of equations above, we can easily deduce the value for bn:

a0=1

b0=1/a0=11=1

b1=a1b0/a0=(121)=12

b2=(a2b0+a1b1)/a0=(13!1+12!(12))=112

b3=(a3b0+a2b1+a1b2)/a0=(14!1+13!(12)+12(112))=0

b4=(a4b0+a3b1+a2b2+a1b3)/a0=(15!1+14!(12)+13!112+(120))=1720

If we continue on, we will find that bn when n is odd are all 0, except that b1=12

We only need to compute for bn when n is even, and these are even values of bn

b0=1

b2=12

b4=1720

b6=130240

b8=11209600

b10=147900160

So the coefficients for the series of xex1 will be:

xex1=112x+112x21720x4+130240x611209600x8+147900160x10+...

To make the Bernoulli numbers appear, we divide the denominator by n!. For example, for the fourth term , we divide 720 by 4! and we obtain 130, which the fourth Bernoulli number. We can rewrite:

=1121!x+162!x21304!x4+1426!x61308!x8+56610!x10+...

The numbers colored in green are Bernoulli numbers.

Thus the general formula of the series is: 
xex1=+n=0Bnxnn!

Proving two formulae involving two trigonometric series of Euler using Euler identities

I have recently read the book "Тригонометрические ряды от Эйлера до Лебега" (Trigonometric series from Euler to Lebesgue) and on page 12, there are 2 intriguing formlae derived by Euler in 1748. The book doesn't specify from which publication these two series come. Further up on that page, there is a mention of the "введения в анализ бесконечных" (Introduction to the analysis of the infinite), which was published also in 1748. Thus I suspect that these two formulae may come from there.

The above two formulae are as followed:

1rcos(x)12rcos(x)+r2=1+rcos(x)+r2cos(2x)...=1++n=1rncos(nx)

rsin(x)12rcos(x)+r2=rsin(x)+r2sin(2x)...=1++n=1rnsin(nx)

In the above book, the author states that Euler didn't provide the radius of convergence for these series, he wrote: "никаких указаний на области применимости этих и подобных им разложений у Эйлера нет" (No indications of range of applications are provided by Euler for these and similar expressions) (page 13)

The radius of convergence should be |r|<1 

At r=1 and r=1, Euler obtained:

12=cos(x)+cos(2x)+cos(3x)+...

12=cos(x)cos(2x)+cos(3x)+...

Both of these trigonometric series diverge xR

I wish to provide a derivation of these two formulae. This is derived from an answer of Student A Level, a member of Mathstackexchange, the site is here

We proceed by working with the RHS of the equation (1) and (2) and try to turn it into the LHS.

We denote the series in (1) as C, and the series in (2) as B

C=1+rcosx+r2cos2x+r3cos3x+

S=rsinx+r2sin2x+r3sin3x+

We will use the Euler identity eix=cos(x)+isin(x) and einx=cos(nx)+isin(nx)

We multiply S with imaginary number i=1, and obtain the new series:

iS=risin(x)+r2isin(2x)+r3isin(3x)...

Summing these two series, we have

C+iS=1+r[cos(x)+isin(x)]+r2[(cos(2x)+isin(2x)]+r3[cos(3x)+isin(3x)]...

Using Euler identity, we have:

C+iS=1+reix+(reix)2+(reix)3+(reix)4+...

This is, in fact, a geometric series, and the general formula is:

11kt=1+kt+(kt)2+(kt)3+(kt)4+...

Let r=k and t=eix

We have 11reix

Let r=k and t=eix

We have 11reix=1+reix+(reix)2+(reix)3+(reix)4+...

On the LHS: we mutiply the fraction with 1reix, so we will have

1reix(1reix)(1reix)=1r(cos(x)irsin(x))1reixreix+r2

This is the tricky part of the proof, we wish to turn

1reixreix+r2=1r(eix+eix)+r2=12rcos(x)+r2

We can work backward and realize that cos(x)=eix+eix2, so we need to multiply reixreix with 2r(eix+eix2). So we will have:

1reixreix+r2=12r(eix+eix2)+r2=12rcos(x)+r2

Thus we have:

1r(cos(x)isin(x)12rcos(x)+r2=1rcos(x)+irsin(x)12rcos(x)+r2=1rcos(x)12rcos(x)+r2+irsin(x)12rcos(x)+r2

Equating the imaginary and the real part, we thus have:

A=1rcos(x)12rcos(x)+r2=+n=1rncos(nx)

B=rsin(x)12rcos(x)+r2=+n=1rnsin(nx)

Ramanujan, the "Euler" of 20th century

      The first time I heard of the name Ramanujan was in 2017. At that time, I was wondering on the internet, searching information about i...