$$\mathrm {PV}\int_{\mathbb{R}^n} \frac{e^{i(b_1z_1+\cdots+b_nz_n)}}{\prod_{j<k}(z_k-z_j)}dz=?$$ Other than integrate this term by term? It is in fact the Fourier transform of the inverse Vandermonde determinant.

I also found this question: Fourier transform of the inverse of Vandermonde determinant which is asking for the $n=3$ case.

  • 1
    $\begingroup$ @Clayton I forgot to mention that it is a real integral. $\endgroup$ – Jiyuan Zhang Mar 6 '19 at 3:02
  • $\begingroup$ What do you mean with a real integral ? $F(z)=\frac{e^{i(b_1z_1+\cdots+b_nz_n)}}{\prod_{j<k}(z_k-z_j)}$ is holomorphic in several variables on simply connected domains $U$ where the denominator doesn't vanish, there for each $z_j$ there is an holomorphic function such that $F = \partial_{z_j} f_j$ and $\int_a^b F(z)dz_j= f_j(b)-f_j(a)$ doesn't depend on the path. $\endgroup$ – reuns Mar 6 '19 at 3:18
  • $\begingroup$ @reuns all $z_i$ are real variables.This integral is the principal value integral. $\endgroup$ – Jiyuan Zhang Mar 6 '19 at 4:26
  • $\begingroup$ @reuns $\int_{x_1\ne x_2}\frac{e^{ib_1x_1+ib_2x_2}}{x_2-x_1}\mathrm dx_1\mathrm d x_2$. The support is when the Vandermonde determinant is non-zero. $\endgroup$ – Jiyuan Zhang Mar 6 '19 at 4:34

So finally you meant the Fourier transform of the distribution $$F(x) = PV(\frac{1}{\prod_{j < k} (x_k-x_j)})$$ that is $$\hat{F}(\omega) = \int_{\mathbb{R}^n} F(x) e^{-i \langle \omega,x \rangle}d^n x$$ Then $$1=F(x)\prod_{j < k} (x_k-x_j)$$

implies $$(2\pi)^n i^{n(n-1)/2} \delta(\omega)=\prod_{j < k}(\partial_j-\partial_k) \widehat{F}(\omega) $$

so that, with $\ast$ the convolution and $T_{j,k}$ the Dirac delta distribution on the subspace orthogonal to $x_j-x_k$ $$\widehat{F}(\omega) = (2\pi)^n i^{n(n-1)/2} \quad T_{1,2} 1_{x_1-x_2 > 0} \ast \ldots \ast T_{j,k} 1_{x_j-x_k > 0} \ast \ldots \ast T_{n-1,n} 1_{x_{n-1}-x_n > 0} + P(\omega)$$

for some polynomial $P$ such that $(\prod_{j < k}(\partial_j-\partial_k))P = 0$, and we may deduce the coefficients of $P$ from symmetry considerations.

| cite | improve this answer | |

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.