No label defined (Q4345352)

Let \(P_1\), \(P_2\) be polynomials on \(\mathbb{R}^n\) satisfying the growth condition of hypoellipticity on \([1,\infty)^n\). It may then be assumed that each polynomial is positive outside a compact subset of \([1,\infty)\), and the ultimate goal is to describe the precise asymptotic behaviour of the counting function \[ N(t_1,t_2)= \#\mathbb{N}^n\cap \{P_1\leq t_1\}\cap\{P_2\leq t_2\}\quad\text{as }t_1,t_2\to \infty. \] Define the zeta function \[ D(s)= \sum_{m\in\mathbb{N}^n,\;P_1P_2(m)\neq 0}{1\over P_1(m)^{s_1}P_2(m)^{s_2}}. \] \(D(s)\) has an analytic continuation to \(\mathbb{C}^2\), and the real part of its domain of analyticity is bounded by a Newton polygon \(\Gamma\). Let \({\mathcal V}\), \({\mathcal L}\) denote respectively the sets of vertices and complex lines that contain a face of \(\Gamma\). Results on the growth of \(N(t_1,t_2)\) can be shown to follow if a certain non-vanishing hypothesis holds. Conjecture: For each vertex \(v\) in \({\mathcal V}\) and \(L\) in \({\mathcal L}\) with \(v\) on \(L\), \[ \text{res}_v\text{res}_L(t^{s_1}_1 t^{s_2}_2 D(s)/s_1s_2 ds_1ds_2)\neq 0. \] If the conjecture holds for a particular vertex \(v= (v_1,v_2)\), there is a region in which \(N(t_1,t_2)\) grows like \(A_vt^{v_1}_1 t^{v_2}_2\). In this paper, the conjecture is verified for the addditive polynomials \(P_1(x)= \sum^n_{i=1} x^{b_i}_i\), \(P_2(x)= \sum^n_{i=1} x^{c_i}_i\). The vertices of the Newton polygon \(\Gamma\) in this case can be given explicitly. If the coordinates are chosen so that \(b_1/c_1\leq\cdots\leq b_n/c_n\) and we set \((b_{n+1}, c_{n+1})= (1,0)\), then they are \[ \begin{aligned} v_0= \Biggl(\sum^n_{j= 1}{1\over b_j},0\Biggr),\;v_i & =\Biggl(\sum^n_{j= i+1}{1\over b_j},\;\sum^i_{j=1} {1\over c_j}\Biggr)\quad (1\leq i\leq n-1),\\ v_n & =\Biggl(0,\;\sum^n_{j= 1}{1\over c_j}\Biggr)\end{aligned} \] and \[ N(t_1, t_2)= B_it^{v_{1i}}_1 t^{v_{2i}}_2+ O(t^{v_{1i}-\theta}_1 t^{v_{2i}-\theta}_2) \] as \(t_1,t_2\to\infty\) provided \((t_1,t_2)\) diverges from the boundary of the set \(t^{c_1/b_1}_1< t_2\) \((i=0)\) or \(t^{c_{i+ 1}/b_{i+1}}_1< t_2< t^{c_i/b_i}_1\) \((1\leq i\leq n)\). The proof is geometric and does use the exact form of \(P_1\), \(P_2\), but it may be that an argument along the same lines would yield the conjecture if each polynomial were nondegenerate with respect to its polyhedron at infinity.