Integral transform related to Dedekind Zeta function

31 Views Asked by At

Im currently reading this paper about the Dedekind zeta function and I'm interested in its functional equation. Despite not being very proficient in algebraic number theory I believe everything is clear besides this one integral transform. Though it regards algebraic number theory, it's nothing more than the calculation of a Jacobian.

So we have two integers $r_1,r_2$ and $r=r_1+r_2 -1$ where $r$ is the rank of $\mathcal{O}^x_K$ and $r_1+2r_2=n=[K:Q]$ is the dimension of the field over the rationals. The integral is of the form $\int_{\mathbb{R}^{(r+1)}}f(X)d^*X$ where $d^*X=\prod_{i=1}^{r+1}\frac{dx_i}{x_i}$.

We are interested in making the substitution $t=\prod_{i=1}^{r+1}x_i^{N_i}=||X||$ where $N_i=1$ if $i$ corresponds to a real embedding (equivalently $I\leq r_1$), and $N_i=2$ otherwise. We then define $c_i$ as the variable satisfying $x_i =t^{\frac{1}{n}} c_i$ so that $||C||=1$, so this transform just integrates around the shells of these $r+1$ dimensional "balls", kind of like transforming to polar.

Now the Jacobian must be calculated: the first $r$ rows/columns are simple: we see $\frac{\partial x_i}{\partial c_j}=\delta_{i,j}t^{\frac{1}{n}}$. Since we have the relation $||C||=1$, $C$ has only $r$ degrees of freedom so one $c_j$ (which we have arbitrarily taken to be $r+1$) is dependent on the other elements in the tuple. We have $c_{r+1}^{N_{r+1}}=\prod_{i=1}^{r}c_i^{-N_i}$ so $N_{r+1}c_{r+1}^{N_{r+1}-1}\frac{\partial c_{r+1}}{\partial c_j}=-\frac{N_j}{c_j}\prod_{i=1}^{r}c_i^{-N_i}$ and $\frac{\partial c_{r+1}}{\partial c_j}=-\frac{N_jc_{r+1}}{N_{r+1}c_j}$. Now we compute $\frac{\partial x_{r+1}}{\partial c_j}=t^{\frac{1}{n}}\frac{\partial c_r+1}{\partial c_j}=t^{\frac{1}{n}}\frac{N_jc_{r+1}}{N_{r+1}c_j}$. We also have $\frac{\partial x_{i}}{\partial t}=J_{I,r+1}=\frac{t^{\frac{1}{n}-1}}{n}c_i$. Now we have something close to an upper triangular matrix with a nonzero bottom row, we we can row reduce by adding each row above $i=r+1$ multiplied by $\frac{N_jc_{r+1}}{N_{r+1}c_j}$ (respectively) to the bottom row ($i=r+1$). This leaves us with a upper triangular matrix with a diagonal of $t^{\frac{1}{n}}$ in every entry besides the last, where we now have $J_{r+1,r+1}=\frac{\partial x_{r+1}}{\partial t}+\sum_{i=1}^{r}\frac{\partial x_{i}}{\partial t}=\frac{t^{\frac{1}{n}}c_{r+1}}{ntN_{r+1}}(\sum_{i=1}^{r}{N_i}+1)=\frac{t^{\frac{1}{n}}c_{r+1}}{tN_{r+1}}$. Finally we have $det(J)=\frac{t^{\frac{r+1}{n}}c_{r+1}}{tN_{r+1}}$, and $d^*X=\prod_{i=1}^{r+1}\frac{1}{t^{\frac{1}{n}}c_i}det(J)dCdt=\prod_{i=1}^{r}\frac{dc_i}{c_i}\frac{dt}{tN_{r+1}}$.

In the paper, this was almost the result--what was mentioned was the exact same without the $N_{r+1}$ in the denominator. I wanted to post this as a sanity check, I've redone the calculations several times and always arrive at this. Any help would be greatly appreciated.