BackgroundBlack holes can be characterized from far away by their spectroscopic gravitational-wave "fingerprints," in analogy to electromagnetic spectroscopy of atoms, ions, and molecules. The idea of using the quasi-normal modes (QNMs) of black holes (BHs) for gravitational-wave (GW) spectroscopy was first made explicit by Detweiler (1980). QNMs of rotating Kerr BHs in general relativity (GR) depend only on the mass and spin of the BH. Thus GWs containing QNMs can be used to infer the remnant BH properties in a binary merger, or as a test of GR by checking the consistency between the inspiral and ringdown portions of a GW signal (Abbott and others 2016;Isi et al. 2019).For a review of QNMs see Berti, Cardoso, and Starinets (2009). A Kerr BH's QNMs are the homogeneous (source-free) solutions to the Teukolsky equation (Teukolsky 1973) subject to certain physical conditions. The Teukolsky equation can apply to different physical fields based on their spin-weight s; for gravitational perturbations, we are interested in s = −2 (describing the Newman-Penrose scalar ψ 4 ). The physical conditions for a QNM are quasi-periodicity in time, of the form ∝ e −iωt with complex ω; conditions of regularity, and that the solution has waves that are only going down the horizon and out at spatial infinity. Separating the radial/angular Teukolsky equations and imposing these conditions gives an eigenvalue problem where the frequency ω and separation constant A must be found simultaneously. This eigenvalue problem has a countably infinite, discrete spectrum labeled by angular harmonic numbers ( , m) with ≥ 2 (or ≥ |s| for fields of other spin weight), − ≤ m ≤ + , and overtone number n ≥ 0.There are several analytic techniques, e.g. one presented by Dolan and Ottewill (2009), to approximate the desired complex frequency and separation constant (ω ,m,n (a), A ,m,n (a)) as a function of spin parameter 0 ≤ a < M (we follow the convention of using units where the total mass is M = 1). These analytic techniques are useful as starting guesses before applying the numerical method of Leaver (1985) for root-polishing. Leaver's method uses Frobenius expansions of the radial and angular Teukolsky equations to find 3-term recurrence relations that must be satisfied at a complex frequency ω and separation constant A. The recurrence relations are made numerically stable to find so-called minimal solutions by being turned into infinite continued fractions. In Leaver's approach, there are thus two "error" functions E r (ω, A) and E a (ω, A) (each depending on a, , m, n) which are given as infinite continued fractions, and the goal is to find a pair of complex numbers (ω, A) which are simultaneous roots of both functions. This is typically accomplished by complex root-polishing, alternating between the radial and angular continued fractions.A refinement of this method was put forth by Cook and Zalutskiy (2014) (see also Appendix A of Hughes (2000)). Instead of solving the angular Teukolsky equation "from the endpoint" using Leaver's approach, one can...