Leaver's QNM Solutions

• Leaver’s QNM solutions are a frequency-domain construction that compute black-hole modes through a continued-fraction quantization method with precise ingoing/outgoing boundary conditions.
• The method employs a Frobenius series expansion and recurrence relations to solve the Regge–Wheeler, Zerilli, and Teukolsky equations in Schwarzschild and Kerr spacetimes.
• Extensions include handling discontinuities, source terms, and analytic derivatives to address convergence issues and benchmark modern perturbation theory techniques.

Leaver’s quasinormal-mode (QNM) solutions are the canonical frequency-domain construction of black-hole normal modes that satisfy purely ingoing/outgoing boundary conditions at the event horizon and spatial infinity. In Schwarzschild, they arise as special solutions of the Regge–Wheeler or Zerilli master equations, obtained by factoring the asymptotics, expanding the regular remainder in a Frobenius series, and imposing a continued-fraction condition that quantizes the complex frequency $\omega$; in Kerr, the same logic is applied to the Teukolsky angular and radial equations, with the angular separation constant coupled to the radial spectral condition (Amicis et al., 15 May 2026, Tanay, 2022).

1. Foundational boundary-value problem

For Schwarzschild perturbations, the starting point is the $1+1$ wave equation

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$

with tortoise coordinate defined by $dr_*/dr=1/f(r)$, and, for Schwarzschild, $f(r)=1-r_h/r$ with $r_h=2M$. In frequency space one obtains

$$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$

For the Regge–Wheeler problem,

$$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$

QNMs impose the dissipative boundary conditions

$$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$

and the spectrum is equivalently characterized by zeros of the Wronskian $W(\omega)$ (Li et al., 6 Feb 2026).

For Schwarzschild even-parity perturbations, the Zerilli equation has the same Schrödinger-type structure after Fourier decomposition,

$1+1$0

with the tortoise coordinate

$1+1$1

The same QNM boundary conditions are used: purely ingoing at the horizon and purely outgoing at infinity (0708.0450).

For Kerr, the separated problem is governed by the angular and radial Teukolsky equations. In one standard form, the angular equation is

$1+1$2

while the radial equation is

$1+1$3

with $1+1$4. The QNM boundary conditions are purely ingoing at the outer horizon and purely outgoing at infinity (Tanay, 2022, 1908.10377).

2. Original continued-fraction construction

Leaver’s original Schwarzschild construction factors out the known asymptotics and expands a regular remainder. For the Schwarzschild Regge–Wheeler equation, one writes

$1+1$5

with

$1+1$6

Substitution yields a three-term recurrence,

$1+1$7

and, for $1+1$8,

$1+1$9

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$0

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$1

The minimal-solution requirement converts the recurrence into the continued-fraction quantization condition

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$2

and the inverted form is used numerically to target the $$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$3-th mode. Nollert’s improvement accelerates convergence by asymptotically approximating the continued-fraction tail (Li et al., 6 Feb 2026).

An equivalent first-order Schwarzschild presentation appears in the Regge–Wheeler form used for second-order perturbation theory. There the first-order solution is written as

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$4

with

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$5

and

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$6

This is the same three-term minimal-solution paradigm in different notation (0708.0450).

In the unperturbed Schwarzschild Regge–Wheeler problem, the high-overtone asymptotics are

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$7

so the high overtones asymptote to a line parallel to the imaginary axis (Li et al., 6 Feb 2026).

3. Kerr generalization and computational realizations

In Kerr, Leaver’s method is a coupled angular-radial construction. The physically admissible radial solution is written as a Frobenius-type series that enforces the correct ingoing/outgoing asymptotics, and the coefficients satisfy a three-term recurrence

$$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$8

where $$\frac{\partial^2}{\partial t^2}\Psi(t,r_*)+\left(-\frac{\partial^2}{\partial r_*^2}+V_{\mathrm{eff}}(r)\right)\Psi(t,r_*)=0,$$9 depend on $dr_*/dr=1/f(r)$0, $dr_*/dr=1/f(r)$1, $dr_*/dr=1/f(r)$2, and $dr_*/dr=1/f(r)$3. The QNM spectrum is then determined by the continued-fraction condition

$dr_*/dr=1/f(r)$4

coupled to the angular eigenvalue problem for $dr_*/dr=1/f(r)$5 (Gu et al., 26 Apr 2026).

A widely used realization is the $dr_*/dr=1/f(r)$6 Python package, which implements a Leaver solver in the radial sector together with the Cook–Zalutskiy spectral approach in the angular sector. In the angular problem, the spin-weighted spheroidal harmonic is expanded in spin-weighted spherical harmonics,

$dr_*/dr=1/f(r)$7

leading to a banded matrix eigenvalue problem for $dr_*/dr=1/f(r)$8. The radial solver then evaluates the continued fraction numerically, using cache interpolation and root polishing for speed and robustness (1908.10377).

A later robust-algorithm refinement replaced numerical finite differencing by analytic first derivatives. In that formulation, the derivative of the radial continued-fraction residual satisfies

$dr_*/dr=1/f(r)$9

while the angular derivative is obtained from a complex-symmetric Hellmann–Feynman-type identity,

$f(r)=1-r_h/r$0

The reported effect is a speedup of about a factor of $f(r)=1-r_h/r$1 at fixed $f(r)=1-r_h/r$2, and the use of analytic $f(r)=1-r_h/r$3 and $f(r)=1-r_h/r$4 allows significantly larger safe steps in $f(r)=1-r_h/r$5 during continuation (Tanay, 2022).

Leaver’s method also functions as a benchmark standard. In a physics-informed operator-learning study for Kerr gravitational QNMs with $f(r)=1-r_h/r$6 and overtones $f(r)=1-r_h/r$7, high-precision Leaver eigenvalues from the public $f(r)=1-r_h/r$8 package were used as ground truth; a single trained model reached relative errors of $f(r)=1-r_h/r$9 for the fundamental mode and $r_h=2M$0 for higher overtones, benchmarked against Leaver’s method (Gu et al., 26 Apr 2026).

4. Extensions beyond the standard homogeneous problem

A direct extension of Leaver’s scheme to effective potentials with a discontinuity expands the wavefunction at the discontinuity $r_h=2M$1, rather than at the horizon, and incorporates the Israel–Lanczos–Sen junction conditions. For a finite jump, $r_h=2M$2, the matching condition reduces to continuity of the Wronskian,

$r_h=2M$3

The wavefunction is written piecewise as

$r_h=2M$4

with

$r_h=2M$5

The left series satisfies a six-term recurrence and the right series a five-term recurrence; the key subtlety is that $r_h=2M$6, so the junction condition supplies the missing relation for $r_h=2M$7. With this strategy, up to $r_h=2M$8 modes were computed with high precision, and the low-lying modes showed excellent agreement with the matrix and Prony methods (Li et al., 6 Feb 2026).

The same continued-fraction philosophy also extends to inhomogeneous problems. In second-order Schwarzschild perturbation theory, the regularized second-order Regge–Wheeler function is expanded as

$r_h=2M$9

and the coefficients obey an inhomogeneous three-term recurrence,

$$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$0

Here the same $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$1 are used as in the homogeneous problem but evaluated at $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$2, because the source drives at

$$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$3

The paper’s main finding is that the second-order QNM frequencies are twice the first-order frequencies, and the gravitational-wave amplitude is up to $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$4 that of the first order for binary black-hole mergers (0708.0450).

These extensions suggest that “Leaver’s QNM solutions” are best understood not only as a specific Schwarzschild or Kerr algorithm, but as a boundary-condition-preserving continued-fraction framework that can absorb matching conditions, source terms, and modified asymptotics without abandoning the minimal-solution principle.

5. High overtones, convergence, and spectral instability

In the unperturbed Schwarzschild problem, high overtones lie roughly parallel to the imaginary axis. With a discontinuity or truncation, the asymptotic spectrum tilts toward the real axis and becomes nearly evenly spaced in $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$5. For large $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$6, the empirical spacings reported for the discontinuous Regge–Wheeler problem are

$$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$7

and the product $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$8 approaches $$\left[-\omega^2-\frac{d^2}{dr_*^2}+V_{\mathrm{eff}}(r)\right]\widetilde{\Psi}(\omega,r_*)=0.$$9; the paper speculates that it tends to $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$0, consistent with a truncated Pöschl–Teller potential. The high-overtone deformation is described as universal, with equal spacing in $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$1 and weak dependence on the detailed shape of the discontinuity (Li et al., 6 Feb 2026).

A distinct high-overtone instability arises in parametrized black-hole potentials. Exact-WKB analysis, confirmed by Leaver numerics, shows that for $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$2 at the special values satisfying

$$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$3

the leading constant in the asymptotic quantization vanishes and the real part grows logarithmically with overtone number. For $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$4,

$$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$5

so $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$6 diverges as a power law. The paper states that the convergence of the real parts of high-overtone QNMs is a distinctive feature of general relativity, while parametrized corrections generically lead to divergent spectral behaviors (Miyachi et al., 21 Dec 2025).

The convergence of time-domain QNM reconstructions depends not only on the pole locations but also on the large-overtone residue phases. A source-driven Schwarzschild analysis identifies a “bounce radius” at $$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$7, corresponding to

$$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$8

and shows that the QNM contribution decays in the lower-half-plane contour closure only if

$$V_{\mathrm{RW}}(r;r_h)=f(r)\left[\frac{\ell(\ell+1)}{r^2}+(1-s^2)\frac{r_h}{r^3}\right].$$9

For sources right of the bounce, the QNM signal scatters off $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$0; for sources left of it, the QNMs propagate directly on the curved light cone (Amicis et al., 15 May 2026).

A complementary radial-scattering analysis recasts the same structure as a mirror problem. The reflected optical length is

$$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$1

and the QNM expansion converges only when

$$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$2

After folding the radial line, the direct and mirror terms become diagonal and off-diagonal propagation channels of a two-component half-line problem. This provides a spectral interpretation of the “second lightcone distance” that controls QNM convergence (Kehagias et al., 23 Jun 2026).

6. Limitations, controversies, and post-Leaver developments

The principal modern limitation of conventional continued-fraction methods is the negative imaginary axis (NIA). In Type-D black holes, the QNM spectrum has a branch cut along the NIA, and standard continued-fraction approaches suffer severe convergence failures for modes that cross or lie on that axis. This failure has been tied to two long-standing problems: the apparent discontinuity in highly damped Kerr QNMs as $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$3, and the unexplained spectral proximity between QNMs and algebraically special frequencies (Chen et al., 17 Jun 2025).

A Heun-based analytic-continuation method was developed to overcome that limitation. By mapping the angular and radial equations to confluent Heun form and imposing a radial spectral condition after rotating the $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$4-contour into the complex plane, the method eliminates the dependence on auxiliary parameters in the connection formulas and computes modes crossing or lying on the NIA. For Schwarzschild and Kerr, over the ranges $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$5 and $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$6, the resulting complete spectra were validated through scattering amplitudes with errors $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$7. The paper’s conclusions are that the apparent Schwarzschild–Kerr discontinuity is an artifact of missing NIA-crossing branches, and that when a QNM exactly coincides with the algebraically special frequency, an additional unconventional mode appears nearby (Chen et al., 17 Jun 2025).

This does not remove the historical and practical centrality of Leaver’s method. Leaver’s continued-fraction formulation remains the standard benchmark for Kerr QNMs, underlies packages such as $$g(\omega,r_*)\sim e^{-i\omega r_*}\quad (r_*\to-\infty),\qquad h(\omega,r_*)\sim e^{+i\omega r_*}\quad (r_*\to+\infty),$$8, supports modern derivative-based improvements, and has been generalized to discontinuous potentials and second-order perturbation theory (1908.10377, Tanay, 2022, Li et al., 6 Feb 2026). At the same time, the recent Heun and exact-WKB developments make clear that Leaver’s QNM solutions are part of a broader spectral framework in which boundary conditions, analytic continuation, Stokes geometry, and residue asymptotics are all essential to a complete understanding of black-hole ringdown.