Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

What is the inverse Mellin transform of (s-1/2)^k on the vertical line Re(s)=a where 0 < a <1 and k is a natural number?

share|improve this question
    
I took the liberty of fixing your formatting –  Yemon Choi Jun 15 '13 at 21:13
    
thank you Yemon! –  Arian Berdellima Jun 15 '13 at 22:36
add comment

2 Answers

Maple command $with(inttrans):invmellin((s-1/2)^k,s,z,0..1)$ (see http://www.maplesoft.com/support/help/Maple/view.aspx?path=inttrans/invmellin for info) calculates it for concrete values of $k.$ For example, Maple produces $$ -1/32\,{\it Dirac} \left( z-1 \right) -{\frac {121}{16}}\,z{\it Dirac} \left( 1,z-1 \right) -{\frac {165}{4}}\,{\it Dirac} \left( 2,z-1 \right) {z}^{2}-$$ $${\frac {85}{2}}\,{\it Dirac} \left( 3,z-1 \right) {z} ^{3}-{\frac {25}{2}}\,{\it Dirac} \left( 4,z-1 \right) {z}^{4}-{\it Dirac} \left( 5,z-1 \right) {z}^{5} $$ in the case $k=5.$

share|improve this answer
1  
I might be inclined to suspect that the interactions of the powers of $z$ and the derivatives of Dirac delta interact in a way that simplifies the numbers in the outcome, maybe getting a constant multiple of the $k$-th derivative of Dirac delta at $1$. –  paul garrett Jun 16 '13 at 17:02
add comment
up vote 1 down vote accepted

In fact I was aiming at evaluating the expression:

$I_k=\frac{1}{2\pi i}\int_{a-i\infty}^{a+i\infty} (s-\frac{1}{2})^k x^{s-1} ds, 0 < a < 1, k\in N $ changing the variable $s-\frac{1}{2}=z$

one can rewrite the integral $I_k$ as:

$I_k=\frac{1}{2\pi i}\int_{a-\frac{1}{2}-i\infty}^{a-\frac{1}{2}+i\infty} z^k x^{z-\frac{1}{2}} dz$

Recall that the Dirac delta function is defined in the integral form as:

$\delta(u)=\frac{1}{2\pi}\int_{-\infty}^{+\infty} e^{ipu} dp$

One could rewrite it as:

$\delta(u)=\frac{e^{-cu}}{2\pi i}\int_{c-i\infty}^{c+i\infty} e^{zu}dz , Re(z)=c$

which would imply that:

${e^{cu}}\delta(u)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty} e^{zu}dz , Re(z)=c$

this in turn yields:

$\frac{d^k}{d u^k}{e^{cu}}\delta(u)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty} z^k e^{zu}dz , Re(z)=c, k\in N$

Now we could express the integral $I_k$ as a Fourier transform by making the change of variable $x=e^u$, hence:

$I_k=\frac{1}{2\pi i}\int_{a-\frac{1}{2}-i\infty}^{a-\frac{1}{2}+i\infty} z^k e^{u(z-\frac{1}{2})} dz$

which is the same as: $e^{u/2}I_k=\frac{1}{2\pi i}\int_{a-\frac{1}{2}-i\infty}^{a-\frac{1}{2}+i\infty} z^k e^{zu} dz$

set $c=a-\frac{1}{2}$ then:

$e^{u/2}I_k=\frac{d^k}{d u^k}{e^{cu}}\delta(u)$

Going to our old variable $x$ the above expression becomes:

$\sqrt x\ I_k=\frac{d^k}{d (ln(x))^k}{x^{c}}\delta(ln(x))$

Since $ln(1)=0$ then $f(x)=ln(x)$ has a root at $x=1$. This in turn would imply:

$\delta(ln(x))=\delta(f(x))=\frac{\delta(x-1)}{|f'(1)|}=\delta(x-1)$

Also $\frac{d}{d ln(x)}=x\frac{d}{dx}\rightarrow \frac{d^k}{d (ln(x))^k}=[x\frac{d}{dx}]^k$

Thus finally we get that:

$\sqrt x\ I_k=[x\frac{d}{dx}]^k(x^{c}\delta(x-1))\leftrightarrow I_k=\frac{1}{\sqrt x}[x\frac{d}{dx}]^k(x^{c}\delta(x-1))$

I hope this is a correct solution as it is an important step in a work that I have to complete. Any suggestions if this happens to be wrong would be more than appreciated.

share|improve this answer
    
Up to en.wikipedia.org/wiki/Mellin_transform , this integral is not the inverse Mellin trasform of $(s-1/2)^k$, assuming $k$ to be a natural number. –  Mark Jun 17 '13 at 4:52
    
It depends how you define at the first place the Mellin trasform. In my case I have define it as: $M(I_k , s)=\int_{0}^{\infty} I_k (v) v^{-s} dv , 0 < Re(s) < 1$ then its inverse has I have expressed should not be contradictory. –  Arian Berdellima Jun 17 '13 at 11:54
    
This is your personal definition. –  Mark Jun 17 '13 at 12:09
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.