1
$\begingroup$

I am working on fractional derivatives which are defined by taking the Cauchy Integral formula and letting the order of the derivative be non-integer. Specifically, \begin{equation} f^{(\alpha)}(z)=\frac{\Gamma(\alpha+1)}{2\pi i}\oint f(t)(t-z)^{-(\alpha+1)}\text{d}t \end{equation} Where our contour in the $t$ plane begins at the origin, encloses $z$ once in the positive sense and returns to the origin. Now, for simplicity let's assume that $f(t)$ is analytic in a region containing both the origin and $z$. Then, our contour integral has a branch on the line segment from the origin to $z$. So, we can deform our contour to go from the origin, up the line segment, loop around $z$, down the line segment and back to the origin. The paper I am following expresses this contour as $$f^{(\alpha)}(z)=\frac{\Gamma(\alpha+1)}{2\pi i}\int_{0}^{z^{+}} f(t)(t-z)^{-(\alpha+1)}\text{d}t$$ They then immediately claim that $$f^{(\alpha)}(z)=\frac{\Gamma(\alpha+1)}{2\pi i}\left[1-\exp{(-2\pi i(\alpha+1))}\right]\int_{0}^{z} f(t)(t-z)^{-(\alpha+1)}\text{d}t$$ Where the integral is now your standard Riemann integral, that can be evaluated via the Fundamental Theorem of Calculus. My question is: why is this true? There is something about contour integrals that I am not aware of that they are using to make the jump from Eq. 2 to Eq. 3.

Also, if it helps, I believe I am working with a proto-Riemann-Liouville derivative, as Eq.3 becomes the Riemann-Liouville derivative when substituting out appropriate coefficients via the the Euler-Reflection formula for the Gamma function.

$\endgroup$
1
$\begingroup$

It comes directly from tracking the branch of $(t-z)^{-(\alpha+1)}$, that in the returning line from $z$ to $0$ the $(t-z)^{-(\alpha+1)}$ is multiplied by $\exp(-2\pi i(\alpha+1))\neq 1$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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