I'm reading a paper and don't understand some of the calculations. We are given an integral equation with asymptotic boundary conditions

$\rho_+(u)=\frac{1}{2\pi} \int\limits_{|v|>\mu}^{}\mathrm{d}v\,\frac{2\hbar \rho_+(v)}{(u-v)^2+\hbar^2}$

$\rho_+(u)=\ln(|u|)-\frac{1}{2} \ln\left(\Xi\right)+O(u-1),\,u\rightarrow\infty$

$\rho_+$ is a density of zeros an $\ln(\Xi)$ can be seen as a density of particles. But the exact meaning is not important for further calculations. The first integral ranges over $\mathbb{R}$ exceot the interval $[-\mu,\mu]$, where we have $\rho_+=0$.

Now a new function $\rho_-$ eis defined as $\rho_-(u)=\begin{cases} 0 &\mbox{if } |u|\geq\mu \\ -\frac{1}{\pi \hbar}[\rho_+(u)-\frac{1}{\pi}\int\limits_{|v|>\mu}^{}\mathrm{d}v\,\frac{\hbar \rho_+(v)}{(u-v)^2+\hbar^2} & \mbox{if } |u|<\mu \end{cases}$

with which we can rewrite the first integral equation as

$\pi\hbar\rho_- +(1-K_+)\rho_+=0$

with $K_+\rho(u)=\int\limits_{-\infty}^{\infty} k_+(u-v)\rho(v) \mathrm{d} v$ and $k_+(u)=\frac{1}{\pi} \frac{\hbar}{u^2+\hbar^2}$.

So far everything is clear. Obviously a constant function is an eigenfunction fo $K_+$ with eigenvalue 1 and so $(1-K_+)$ ist invertible. It is the said, that an operator $K_-$ can be definied on functions vanishing at $\infty$ by on functions vanishing at $\infty$


This is still quite clear although I don't really understand, why the functions must vanish at $\infty$.

Then it ist said, that due to the degeneracy of the operator $K_+$, $K_-$ ist just defined up to a constant. What ist meant by this? Is it just due to the fact that every constant function is an eigenfunction?

Now without any calculations it is said, that the kernel $k_-(u)$ is given by

$k_-(u)=\frac{1}{\pi\hbar} \Psi(1+i\frac{u}{\hbar})+\Psi(1-i\frac{u}{\hbar})+\Delta$

where $\Psi(u)=\Gamma\,'(u)/\Gamma(u)$ is the digamma-function.

It is only said, that this can easily obtaind by fourier transform. I don't have really much knowledge about integral-equations and do not see how this result can be found. Could please someone show me?

I have some other questions concerning the following calculations, but maybe I stop here until this problem is solved.

  • 1
    $\begingroup$ Which paper are you reading? $\endgroup$ – Qmechanic Nov 2 '13 at 12:38
  • 1
    $\begingroup$ I'm reading a paper by Sklyanin about the quantization of the toda lattice. I don't know if I'm allowd to attach the pdf, I'll do it anyway. You can download it here: dropbox.com/s/tju18r0jvb80zdk/Skly85_Toda.pdf The part I'm struggling with starts at page 221. $\endgroup$ – TheoPhysicae Nov 2 '13 at 12:55
  • $\begingroup$ I tried to use $(1-K_-)(1-K_+)\rho=\rho$. This gives one integral which contains only $k_-$, one with onliy $k_+$ and one double integral. I don't see, how using fouriertransform and inverse fouriertransform would lead to isolating $k_-$. I knwo how to use fouriertransform to solve partial differential equations, but I have no idea how it could be used here. $\endgroup$ – TheoPhysicae Nov 2 '13 at 16:43
  • $\begingroup$ Note that $\hbar\equiv\eta$. $\endgroup$ – Qmechanic Nov 5 '13 at 15:49

We start with the Lorentzian distribution

$$ \tag{4.11} k_{+}(u)~:=~ \frac{1}{\pi} \frac{\eta}{u^2+\eta^2}; $$

with Fourier transform

$$ \tag{4.11'} \tilde{k}_{+}(x) ~:=~\int_{\mathbb{R}}\!du ~e^{-ixu} k_{+}(u)~=~e^{-\eta|x|}; $$

with integral operator

$$ \tag{4.10} (K_{\pm}\rho)(u)~:=~\int_{\mathbb{R}}\!dv~ k_{\pm}(u-v)\rho(v) ~=~(k_{\pm}\ast\rho)(u). $$

The Fourier transformed operator is a multiplication operator

$$ \tag{4.10'} (K_{\pm}\rho)^{\sim}(x)~:=~\tilde{k}_{\pm}(x) \tilde{\rho}(x). $$

Repeated application of the integral operator leads to repeated convolutions

$$ \tag{A} (K_{\pm}^n\rho)(u) ~=~(\underbrace{k_{\pm}\ast\ldots \ast k_{\pm}}_{n\text{ factors}}\ast\rho)(u), \qquad (K_{\pm}^n\rho)^{\sim}(x)~:=~\tilde{k}_{\pm}^n(x) \tilde{\rho}(x). $$

Note that for a constant function $\rho \propto 1$ is an eigenfunction with eigenvalue 1 for the integral operator $K_{+}$:

$$\tag{B} (K_{+} 1)(u) ~=~\int_{\mathbb{R}}\!dv~ k_{+}(u-v)~=~1. $$ Since we want the operator $1-K_{+}$ to be invertible, we will from now on only consider integrable functions $\rho:\mathbb{R}\to \mathbb{C}$, such that $\int_{\mathbb{R}}\!du~\rho(u)=0$. In particular, we will exclude constant functions $\rho \propto 1$. This means that the kernels $k_{\pm}(u)\to k_{\pm}(u)+\Delta_{\pm}$ are only defined modulo additive constants $\Delta_{\pm}$.


$$\tag{C} 1-K_{-} ~=~ \frac{1}{1-K_{+}}~=~\sum_{n=0}^{\infty}K_{+}^n ,$$

and therefore, naively,

$$\tag{D} \tilde{k}_{-}(x)~=~\left(1- \frac{1}{\tilde{k}_{+}(x)}\right)^{-1} ~=~\frac{1}{1-e^{\eta|x|}} . $$

However the expression (D) is not integrable at $x=0$. The solution is to regularize $k_{-}(u)$ with an infinite additive constant $\Delta_{-}=\int_{\mathbb{R}}\!\frac{dx}{2\pi}~\frac{e^{-\eta|x|}}{\eta|x|}=\infty$, so that

$$ k_{-}(u)~=~ \int_{\mathbb{R}}\!\frac{dx}{2\pi} \left[\frac{e^{-\eta|x|}}{\eta|x|}+ \frac{e^{ixu}}{1-e^{\eta|x|}}\right] ~=~\int_{0}^{\infty}\!\frac{dx}{\pi} \left[\frac{e^{-\eta x}}{\eta x} +\frac{\cos(xu)}{1-e^{\eta x}} \right]$$

$$\tag{E} ~\stackrel{t=\eta x}{=}~\sum_{\pm}\int_{0}^{\infty}\!\frac{dt}{2\pi\eta} \left[\frac{e^{-t}}{t}+\frac{\exp\left(\pm\frac{ iut}{\eta}\right) }{1-e^{t}}\right]~=~\frac{1}{2\pi\eta}\sum_{\pm}\psi(1\pm\frac{iu}{\eta}), $$

cf. the Digamma function and Abramowitz & Stegun eq. (6.3.21). Note that the final formula (E) differs from Sklyanin's formula (4.12) by a factor 2 in the overall normalization.


  1. E.K. Sklyanin, The quantum Toda chain, Lecture Notes in Physics, 226 (1985) 196.
  • $\begingroup$ Perfect, I talked to one of my professors about that problem and he also gave the adivice, that I just need to use the convulition theorem and an integral representation of the digamma function. Nevertheless I didn't really succeed, thanks a lot for this answer! $\endgroup$ – TheoPhysicae Nov 5 '13 at 15:20
  • $\begingroup$ Hello, I have another problem with the paper. Right after the calculation the asymptotic conditions for $\rho_-$ are derived. I'don't see, where they come from. I thought they would be trivial, but now when I look closer at them, that's no longer the case and I'm quite unsure how Sklyanin found them. Could you maybe help me again? $\endgroup$ – TheoPhysicae Dec 5 '13 at 0:22

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.