I consider for example the following function of two variables given by $$f(x,y)=\sum_{n=0}^{+\infty}\frac{\delta_{x}^{-m} \exp(\delta_{x})}{\delta_{y}^{-m}\exp(\delta_{y})}\left(\frac{x}{y}\right)^{n}$$ Where $\delta_{x}=\frac{d}{dx}$ and $\delta_{y}=\frac{d}{dy}$ and $m$ is a positive integer. My question is: when $x < y$ can we write f as $$f(x,y)=\frac{\delta_{x}^{-m} \exp(\delta_{x})}{\delta_{y}^{-m}\exp(\delta_{y})}\sum_{n=0}^{+\infty} \left(\frac{x}{y}\right)^{n}=\frac{\delta_{x}^{-m} \exp(\delta_{x})}{\delta_{y}^{-m}\exp(\delta_{y})}g(x,y)?$$ such that $g(x,y)=\frac{1}{1-\frac{x}{y}}=\frac{y}{y-x}$. If the formula above is correct can we prove it more explicitly. Precisely, constructing a corollary for which $\frac{\delta_{x}^{-m} \exp(\delta_{x})}{\delta_{y}^{-m}\exp(\delta_{y})}$ is a well-defined operator denoted by $A_{x,y}^{m}$. I need a response, if someone have an idea.

Best regards and thank you.

  • $\begingroup$ $\delta_{x}^{-1} $ is an integral operator. Those usually need a boundary condition? The exponentials are just shift operators. $\endgroup$ Sep 19, 2021 at 21:24
  • $\begingroup$ for example $\delta_{x}^{-2}T(x)$ is equal to $$\int_{0}^{+ \infty} \int_{0}^{+ \infty}T(x) dx?$$ for $x \in R^{+}$. $\endgroup$ Sep 22, 2021 at 16:10
  • $\begingroup$ Ok, so your boundary condition is that you start integrating at 0, that's fine. The upper limit of course must be variable - your result must again depend on $x$. So, $\delta^{-2}_{x} T(x) = \int_0^x dx' \int_0^{x'} dx'' T(x'')$. $\endgroup$ Sep 22, 2021 at 16:37
  • $\begingroup$ Thank you. It is simple to have $\delta_{x}^{m} x^{m}= m!$ But if we have $\delta_{x}^{-m} x^{-m}$ what can we say in this case? $\endgroup$ Sep 22, 2021 at 16:42
  • $\begingroup$ If you want to treat something like $x^{-m} $, you should probably use a different boundary condition. How would you integrate $x^{-m} $ starting at $x=0$? That's not well defined. $\endgroup$ Sep 22, 2021 at 16:46

1 Answer 1


To make sense of this, you might Fourier transform $$A_m=\frac{\delta_{x}^{-m} \exp(\delta_{x})}{\delta_{y}^{-m}\exp(\delta_{y})}=\delta_{x}^{-m} \delta_{y}^{m}e^{\delta_x-\delta_{y}},$$ which would give the operator $$\hat{A}_m=(k_y/k_x)^m e^{i(k_x-k_y)}.$$ Then $A_m$ acting on a function $f(x,y)$ could be obtained by inverse Fourier transformation, $$A_m f(x,y)=\int_{-\infty}^\infty \frac{dk_x}{2\pi}\int_{-\infty}^\infty \frac{dk_y}{2\pi}(k_y/k_x)^m \hat{f}(k_x,k_y)e^{i(x+1)k_x+i(y-1)k_y}.$$ The question whether $A_m$ is a well-defined operator would then boil down to whether this integration is well-defined. The function $\hat{f}(k_x,k_y)$ would need to vanish when $k_x\rightarrow 0$ to avoid the pole.

  • $\begingroup$ Firstly, $\frac{\delta_{x}^{-m} exp(\delta_{x})}{\delta_{y}^{-m} exp(\delta_{y})} \neq \delta_{x}^{-m} \delta_{y}^{m} exp(\delta_{x}-\delta_{y})$. Secondly, where those calculations came from? I don't understand your explications. $\endgroup$ Sep 20, 2021 at 16:51
  • $\begingroup$ Can you give me a reference on this topic? $\endgroup$ Sep 20, 2021 at 16:55
  • $\begingroup$ $\delta_y$ and $\delta_x$ commute, so you have two commuting operators $A=\delta_x^{-m}e^{\delta_x}$ and $B=\delta_y^{-m}e^{\delta_y}$; the ratio $\frac{A}{B}=AB^{-1}=\left(\delta_x^{-m}e^{\delta_x}\right)\left(e^{-\delta_y}\delta_y^{m}\right)=\delta_x^{-m}\delta_y^{m}e^{\delta_x-\delta_y}$, where in the last step I again used the commutation of $\delta_x$ and $\delta_y$. $\endgroup$ Sep 20, 2021 at 17:07
  • $\begingroup$ But the operator $e^{- \delta_{y}}= \sum_{k=0}^{+ \infty} \frac{(-1)^{k} \frac{d^{k}}{dy^{k}}}{k!}$ and $\frac{1}{e^{ \delta_{y}}}= \frac{1}{\sum_{k=0}^{+ \infty} \frac{\frac{d^{k}}{dy^{k}}}{k!}}$ they are not the same. Can we deduce a commutator here. Is there any propertie that is explaining this formula in a reference? $\endgroup$ Sep 21, 2021 at 10:53
  • $\begingroup$ $e^{-\delta_y}=\frac{1}{e^{\delta_y}}$, for the reason that for any operator $A$ such that the sum converges one has $\left(\sum_{n=0}^\infty (-A)^n/n!\right)\left(\sum_{m=0}^\infty A^m/m!\right)=1$. $\endgroup$ Sep 21, 2021 at 10:57

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.