Finite stationary phase expansions (Q819598)

The motivation of this article is Bott's question: Why does the height function \(z\) on \(S^{2n}\) have the property that \(e^{tz}\) has a finite stationary phase expansion? The main result of the paper is Theorem 3.2: If \(M\) has constant curvature with Euler class \(\chi\) given by \(\chi = \text{Pfaff} (\frac{R}{2\pi})\) and if \(f = \text{Pfaff} (\frac{L}{2\pi})\) with \(L(Y) = {\mathcal L}_{\tilde{Y}} - \nabla_{\tilde{Y}}\) and curvature \(R\), then the stationary phase expansion of \(\int_M e^{tf}\chi\) relative to \(t\) has only finite many non-zero terms. The Duistermaat-Heckman theorem gives the finite stationary phase property for moment maps \(f\) of Hamiltonian actions, \(\int_Me^{tf}\frac{\omega^n}{n!} = \sum_{p\in C_f}\frac{e^{t f(p)}}{c_p t^n}\), where the summation is over the critical points of \(f\). For the equivariant Euler class \(\tilde{\chi} =\text{Pfaff}(\frac{R+uL}{2\pi})\) of \(M\), one has \[ \int_Me^{\tilde{\chi}} = \int_Me^{fu^2}(\chi_n + \frac{u^n}{2!} \sum_{j_1+j_2=n}\chi_{j_1}\chi_{j_2} + \frac{u^{2n}}{3!} \sum_{j_1+j_2+j_3=n}\chi_{j_1}\chi_{j_2}\chi_{j_3} + \cdots + \frac{u^{n(n-1)}}{n!} \chi_1^n), \] where \(\tilde{\chi} = \chi_n + \chi_{n-1}u+ \cdots + \chi_1 u^{n-1} + fu^n\) and \(\chi_k \cdot \chi_l\) is related to \(f\chi_{k+l}\) for the manifold \(M\) of constant curvature by Lemma 3.1. On the other hand, by the equivariant localization theorem, \[ \int_Me^{\tilde{\chi}} = \sum_{p\in C_f}\frac{(2\pi)^n}{f(p)u^n}e^{f(p)u^n} =P(t, \frac{\partial}{\partial t}) \int_M e^{ft}\chi, \] where \(u(t^{1/n}X) = t^{1/n}\) and \(P(t, \frac{\partial}{\partial t})= 1+b_1 t\frac{\partial}{\partial t} + \cdots + b_{n-1}t^{n-1} \frac{\partial^{n-1}}{\partial t^{n-1}}\) with \(b_i\) determined by some combinatorical formula. Then, by using \(\int_Me^{tf}\chi = \sum_{p\in C_f} e^{f(p)t}\sum_{j=0}^{\infty} c_{p, j}t^{-j}\), and by comparing coefficients on both sides of the equation for \(\int_M e^{\tilde{\chi}}\), all the constants \(c_{p, j}\) must vanish for suitably large \(j\). Therefore, the main result follows. This follows from the straightforward calculation for the relations among the equivariant Euler classes. But the result lacks of geometric and/or topological explanation. The paper starts with the Duistermaat-Heckman Theorem in section 1 and its generalization in section 2. The main result, Theorem 3.2, is proved in the section 3 and the key Lemma 3.1 is computed in section 4. Explicit computation on \(S^{2n}\) is given in section 5. Further questions are raised in the last section (6). It is interesting to see that the result in this paper is true also for critical submanifolds in the Bott sense.