Quasinormal bulk-edge characters of gravitons in Nariai geometry
Extracted source text · arXiv 2506.07556
\documentclass[a4paper,12pt]{article}
\usepackage{jheppub} % for details on the use of the package, please see the JINST-author-manual
%\usepackage{lineno}
%\linenumbers
%\usepackage{showkeys}
%\arxivnumber{1234.56789} % if you have one
%Logarithmic corrections to the near extremal black hole from quasinormal modes
\title{\boldmath Quasinormal bulk-edge characters of gravitons in Nariai geometry }
% Collaborations
%% [A] If main author
%% \collaboration{\includegraphics[height=17mm]{collabroation-logo}\\[6pt]
%% XXX collaboration}
\usepackage{comment}
%% or
%% [B] If "on behalf of"
%% \collaboration[c]{on behalf of XXX collaboration}
% Authors
% The "\note" macro will give a warning: "Ignoring empty anchor...", you can safely ignore it.
%% [A] simple case: 2 authors, same institution
%% \author[1]{A. Uthor\note{Corresponding author.}}
%% \author{and A. Nother Author}
%% \affiliation{Institution,\\Address, Country}
%% or, e.g.
%% [B] more complex case: 4 authors, 3 institutions, 2 footnotes
%% \author[a,b]{F. Irst,\note{Now at another university}}
%% \author[c]{S. Econd,}
%% \author[a,2]{T. Hird\note{Also at Some University.}}
%% \author[c,2]{and Fourth}
%% \affiliation[a]{Institution_1,\\Address, Country}
%% \affiliation[b]{Institution_2,\\Address, Country}
%% \affiliation[c]{Institution_3,\\Address, Country}
%\author{A. Uthor}
%\affiliation{One University,\\
%some-street, Country}
%\affiliation{Another University,\\
%different-address, Country}
% E-mail addresses: only for the corresponding author
%\emailAdd{first@one.univ}
\author{Jyotirmoy Mukherjee}
\affiliation[a]{ Department of Theoretical Physics\\
Tata Institute for Fundamental Research, Mumbai 400005, India.}
% e-mail addresses: one for each author, in the same order as the authors
%\emailAdd{budhadityab@iisc.ac.in}
\emailAdd{jyotirmoy.mukherjee\_119@tifr.res.in}
\abstract{In this paper, we evaluate the character partition function of gravitons in the Nariai geometry using quasinormal modes. Employing the Denef-Hartnoll-Sachdev (DHS) prescription, we compute the bulk partition function from the spectrum of quasinormal frequencies. We demonstrate that the quasinormal partition function agrees with the bulk one-loop determinant of gravitons on \( S^2 \times S^2 \), as obtained via the heat kernel method. Furthermore, we isolate the contribution of the edge character from the full one-loop partition function on \( S^2 \times S^2 \), and provide remarks on the possible origin of this edge contribution in the quasinormal mode framework.
}
\begin{document}
\maketitle
\flushbottom
\section{Introduction}
Quantum corrections to black hole spacetimes offer crucial insights into the microscopic origin of black hole entropy. At the semi-classical level, such corrections are encoded in the one-loop partition function of matter fields \cite{Banerjee:2010qc, Banerjee:2011jp}, around the black hole background. In the path integral approach, the one-loop contribution arises from integrating over these quadratic fluctuations, and is formally expressed as the functional determinant of the corresponding kinetic operator.
Although the one-loop partition function plays a central role in capturing quantum corrections to black hole physics, its explicit evaluation in black hole backgrounds is often technically challenging. This difficulty arises primarily from the complexity of determining the spectrum of the associated kinetic operator. The heat kernel method offers a powerful framework for computing one-loop determinants, particularly in Euclidean settings \cite{Vassilevich:2003xt}. Alternatively, a compelling approach involves expressing the one-loop determinant in terms of quasinormal modes, known as the Denef-Hartnoll-Sachdev formula \cite{Denef:2009kn}. These modes characterize the spectrum of quantum fluctuations and capture the inherently dissipative dynamics of black hole spacetimes. Denef-Hartnoll-Sachdev(DHS) formula has been quite useful in computing the one-loop determinant of matter fields in the background of black holes in de Sitter and anti-de Sitter spaces \cite{Datta:2011za, Castro:2017mfj,Grewal:2022hlo} and finding quantum corrections in the extremal and near-extremal black holes \cite{Mukherjee:2024nhx,Kapec:2024zdj,Arnaudo:2024bbd}.
In this work, we focus on evaluating the one-loop determinant of gravitons in the Nariai geometry. The Nariai geometry arises as a special limit of the Schwarzschild--de Sitter black hole, where the black hole and the cosmological horizons coincide \cite{1950SRToh..34..160N}. In this limit, the geometry of the near horizon takes the form of a direct product, \( dS_2 \times S^2 \). Interestingly, this structure closely resembles the near-horizon geometry of extremal black holes in asymptotically flat or $AdS$ spacetimes, where a decoupled throat region of the form \( AdS_2 \times S^2 \) emerges. However, unlike the extremal limit, the Nariai limit does not rely on the presence of charge or angular momentum. Instead, it stems purely from the existence of a positive cosmological constant and the degeneracy of the black hole and cosmological horizons.
Using the Denef-Hartnoll-Sachdev prescription, we evaluate the one-loop partition function of gravitons %\footnote{Including the scalar modes, the quasinormal character partition function can be written as
%\begin{align}
%\log\mathcal{Z}^{\rm{bulk}}_{\rm{ch}}&=\sum_{s=0}^{2}\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=s}^{\infty}d_{l,s}\frac{e^{-\frac{t}{2}+i\nu_s t}+e^{-\frac{t}{2}-i\nu_s t}}{1-e^{-t}}
%\end{align}, with $d_{l,0}=(2l+1)$.}
in Nariai geometry in $D=4$ dimension, which admits the following integral form:
\begin{align}
\log\mathcal{Z}^{\rm{bulk}}_{\rm{ch}}&=\sum_{s=1}^{2}\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=s}^{\infty}d_{l,s}(-1)^{s-1}\frac{e^{-\frac{t}{2}+i\nu_s t}+e^{-\frac{t}{2}-i\nu_s t}}{1-e^{-t}},\label{bulkch1}
\end{align}
where $d_{l,2}=5(2l+1)$ and $d_{l,1}=3(2l+1)$ are the degeneracy factors of the transverse spin-2 and spin-1 components (including the spin degeneracy) and $\nu_s=\sqrt{\left((l+2)(l-1)-\frac{1}{4}\right)}$ for gravitational perturbation. This particular integral form of the partition function obtained from the infinite product of quasinormal modes agrees with the heat-kernel evaluation of the bulk one-loop determinant of gravitons on $S^2\times S^2$.
However, there is an edge contribution to the one-loop determinant of graviton on $S^2\times S^2$ that is not captured in the infinite product over quasinormal modes. We present the edge contribution of the graviton one-loop determinant in section \ref{edge}. We wish to understand the origin of the edge character partition function from quasinormal modes and its connection to the entanglement entropy of edge states of gravitons in the near future.
The organization of the paper is as follows. In section, \ref{sec2}, we briefly review the Denef-Hartnoll-Sachdev prescription for a minimally coupled scalar field. Using the DHS prescription, we write the character partition function from quasinormal modes. In section \ref{sec3}, we present the quasinormal modes of gravitational perturbations and with those we derive the quasinormal bulk character in \ref{sec4}. Finally, in section \ref{sc6}, we show the bulk-edge split of the one-loop determinant of on $S^2\times S^2$ and offer some remarks on the edge characters.
\paragraph{Note:} While this paper was in preparation, the preprint \cite{Law:2025yec} appeared on arXiv, showing the bulk-edge split of one-loop partition function of the graviton in the Nariai geometry in arbitrary dimension.
\section{Denef-Hartnoll-Sachdev (DHS) \label{sec2}prescription}\label{sec2}
%\usepackage{amsmath}
%\usepackage{amssymb}
%\usepackage{hyperref}
%\usepackage{xcolor}
%\begin{document}
In this section, we briefly review the Denef-Hartnoll-Sachdev (DHS) prescription \cite{Denef:2009kn} to evaluate the one-loop determinant of a scalar field from an infinite product of quasinormal modes. In the DHS framework, one can think of the one-loop determinant as a meromorphic unction of the mass parameter or the conformal dimension $\Delta$. To understand this more explicitly, let us consider a complex scalar field in the black hole background. The one-loop determinant can be written as
\begin{align}
\mathcal{Z}^{(1)}&=\int D\phi\, e^{-\int \sqrt{g}\phi^{*}\left(-\nabla^2+m^2\right)\phi}\sim \frac{1}{\det\left(-\nabla^2+m^2\right) }
\end{align}
We analytically continue $\mathcal{Z}^{(1)}(m^2)$ in a complex $m^2$-plane. The one-loop partition function $\mathcal{Z}^{(1)}(m^2)$ can be totally characterized in terms of its poles, zeros and its value at the asymptotic infinity. In particular, we encounter a pole when the determinant is zero. This occurs when $m^2$ is tuned in such a way that there exists zero mode of the scalar field $\phi$. Let us denote the zero modes by $\phi_{*,k}$. The Euclidean zero modes $\phi_{*,k}$ correspond to solutions of the Klein-Gordon equation that are regular and periodic in Euclidean time. The index $k$ labels the Matsubara (thermal) modes associated with the compact Euclidean time circle, while $*$ represents all additional quantum numbers characterizing the spatial and internal structure of the mode.
The DHS prescription relies on the connection between the Euclidean zero modes and the Lorentzian quasinormal modes (related via Wick rotation). As it is well known that, the quasinormal modes are defined to be the solution of the equation of motion which satisfies ingoing boundary condition at the horizon. We denote their frequencies by $\omega_{*}(\Delta)$, which depend parametrically on the conformal dimension $\Delta$ (or the mass squared $m^2$). When $\Delta$ is adjusted so that $\omega_{*}(\Delta)$ coincides with a Matsubara frequency $\omega_k = 2\pi i k T$, the corresponding Lorentzian mode $\phi_{*,\omega}$ analytically continues to the Euclidean zero mode $\phi_{k,*}$. This matching of mode spectra underlies the spectral representation of the one-loop determinant in terms of quasinormal frequencies.
From the location of poles at $\omega_*(\Delta)=\omega_k$, the one-loop determinant can be written as \cite{Denef:2009kn,Datta:2011za,Castro:2017mfj}:
\begin{align}
\mathcal{Z}^{(1)}&=e^{\rm{Pol}(\Delta)}\prod_{k,*} \left(\omega_k-\omega_*(\Delta)\right)^{-1}\nonumber\\
&=\mathcal{Z}_G\mathcal{Z}_{\rm{ch}}\label{DHS},
\end{align}
where $\mathcal{Z}_{\rm{ch}}$ denotes the one-loop partition function computed from the infinite product over quasinormal modes, which we call as quasinormal character partition function. $\mathcal{Z}_G$ denotes the component of the one-loop partition function which does not depend on the quasinormal mode frequency. It can in general depend on the coupling constant or the volume of the gauge group when applied for higher spin theories.
\subsection{Character partition function from quasinormal modes}
We study the infinite product over quasinormal modes to evaluate the free energy. While this approach captures many important aspects—such as the ultraviolet structure of the one-loop determinant—it may not fully account for certain features of the theory, including the gauge group volume and the dependence on coupling constants. Keeping these limitations in mind, we now turn to the character partition function of a real minimally coupled free scalar, which is constructed from the infinite product over (anti-)quasinormal modes.
\begin{align}
\mathcal{Z}_{\rm{ch}}
&= \prod_{\omega_*, \bar{\omega}_*} \prod_{k=-\infty}^{\infty}
\left(| k| + \frac{i \omega_*}{2\pi T} \right)^{-\frac{1}{2}} \left( |k| - \frac{i \bar{\omega}_*}{2\pi T} \right)^{-\frac{1}{2}} ,
\end{align}
While taking the product over the quasinormal modes, we have to also take into account degeneracy of other quantum numbers present in the quasinormal modes.
To evaluate the free energy from the product over quasinormal modes, we take logarithm of the product and use the following integral representation of the logarithm \footnote{In principle, one should keep a regulator $\epsilon$ in the lower limit of the integral to keep track of the UV divergence.}:
\begin{align}
-\log x=\int_0^{\infty}\frac{dt}{t}\, e^{-tx}.
\end{align}
We evaluate the character free energy or the logarithm of the character partition function in the following way:
\begin{align}
\log Z_{\text{ch}}
&= \int_0^\infty \frac{dt}{2t} \sum_{\omega_*}\sum_{\bar{\omega_*}} \sum_{k=-\infty}^{\infty}\left(
\, e^{-\left( |k| + \frac{i \omega_*}{2\pi T} \right) t}+ \, e^{-\left( |k| - \frac{i \bar{\omega_*}}{2\pi T} \right) t}\right) \\
&= \int_0^\infty \frac{dt}{2t} \frac{1 + e^{-t}}{1 - e^{-t}} \sum_{\omega_*}\sum_{\bar{\omega_*}}\left(
\, e^{-\left( \frac{i \omega_*}{2\pi T} \right) t}+ \, e^{\left( + \frac{i \bar{\omega_*}}{2\pi T} \right) t}\right)\nonumber\\
&=\int_0^\infty \frac{dt}{2t} \frac{1 + e^{-t}}{1 - e^{-t}}\chi_{\rm{QNM}}.
\label{qnmdef}
\end{align}
In the second line, we have performed the sum over Matsubara modes.
The character partition function takes a nice integral formula in terms of quasinormal modes \cite{Grewal:2022hlo, Mukherjee:2024nhx}. This particular form of the integral will be very useful to match against the heat-kernel expressions. As we have already mentioned after the equation \eqref{DHS}, this character partition function do not capture about information of the coupling constant or the volume of the gauge group,
one requires to compute $\mathcal{Z}_G$ separately. In the following sections, we will focus on the bulk-edge split of the character partition function.
\section{Nariai geometry}\label{sec3}
The Nariai geometry emerges as a special limit of the Schwarzschild–de Sitter (SdS) spacetime when the black hole and cosmological horizons approach each other \cite{1950SRToh..34..160N}.
The Schwarzschild--de Sitter (SdS) metric in four dimensions is given by
\begin{equation}
ds^2 = -f(r)\, dt^2 + \frac{dr^2}{f(r)} + r^2 d\Omega_2^2,
\end{equation}
where
\begin{equation}
f(r) = 1 - \frac{2M}{r} - \frac{\Lambda}{3} r^2,
\end{equation}
and \( d\Omega_2^2 = d\theta^2 + \sin^2\theta\, d\phi^2 \) is the round unit 2-sphere.
When the black hole and cosmological horizons coincide, \( r_b \to r_c \), we approach the Nariai limit. Zooming in to the near-horizon region, this geometry takes a direct product form : $dS^2\times S^2$ \cite{1950SRToh..34..160N}. \begin{align}
ds^2=-f(r)dt^2+\frac{dr^2}{f(r)}+R^2d\Omega^2,\quad f(r)=1-\frac{r^2}{\ell_N^2}\label{metric}
\end{align}
Here $\ell_N$ is the de Sitter length of $dS^2$ and $R$ is the radius of the sphere $S^2$ and the Hawking temperature $T_N=\frac{1}{2\pi\ell_N}$.
The Nariai solution is notable for being an exact solution to the four-dimensional Einstein equations with a positive cosmological constant.
Moreover, the Nariai geometry bears a striking resemblance to the near-horizon geometry of extremal black holes in asymptotically flat or anti-de Sitter (AdS) spacetimes, which typically exhibit a \( \text{AdS}_2 \times S^2 \) throat. In contrast, the Nariai limit does not require electric/magnetic charge or angular momentum. The existence of this near-horizon geometry follows solely from the presence of a positive cosmological constant and the degeneracy of black hole and cosmological horizons.
\subsection{Quasinormal modes in Nariai geometry}
Quasinormal modes in Nariai geometry can be studied analytically. Since geometry is a direct product of $dS^2\times S^2$, wave equations take a simple form in this background. The spectrum of scalar, vector, and tensor type perturbations on the angular part ($S^2$) are exactly known \cite{Camporesi:1994ga}, which simplifies the computation by a large amount. Essentially, the wave equation becomes a radial equation on $dS^2$ with the additional mass coming from the modes on $S^2$.
The master equation in the tortoise coordinate takes a simple form:
\begin{align}
\left[\partial_{r_*}^2+\omega^2-V_s(r_*)\right]\psi_s(r)=0
\end{align}
The form of the potential is different for different type of perturbations. We present the solution of the wave equation with the appropriate boundary conditions in the appendix \ref{app1}.
The quasinormal modes for the massless spin-s ($s\leq 2$) perturbation in the Nariai background can also be computed analytically \cite{Cardoso:2003sw,Venancio:2020ttw}. Here we just present the quasinormal modes, the details of the derivations are presented in \ref{app1}. For minimally coupled free massless scalar on the Nariai geometry, the quasinomal modes are given by \cite{Cardoso:2003sw,Venancio:2020ttw}
\begin{align}
\, \frac{\omega_*}{2\pi T_N}&=-i\left(n+\frac{1}{2}\right)+\nu_0,\label{QNMs}
%\frac{ \bar{\omega}_*}{2\pi T_N}&=-i\left(n+\frac{1}{2}\right)+\nu_0, \quad n\geq 0,\label{QNMs}
\end{align}
where $\nu_0$ is given by $
\nu_0=\sqrt{l(l+1)-\frac{1}{4}}.$ For electromagnetic perturbations, one obtains the same set of quasinormal modes as given in \eqref{QNMs}. However, for a gravitational perturbation, the factor $\nu_0$ changes. The quasinormal modes for the gravitational perturbation is given by\cite{Cardoso:2003sw,Venancio:2020ttw}
\begin{align}
\, \frac{\omega_*}{2\pi T_N}&=-i\left(n+\frac{1}{2}\right)+\nu_2,\label{QNMmodes}%\nonumber\\
%\frac{ \bar{\omega}_*}{2\pi T_N}&=-i\left(n+\frac{1}{2}\right)+\nu_2, \quad n\geq 0\label{QNMmodes}
\end{align}
where $\nu_2$ is given by
\begin{align}
\nu_2=\sqrt{(l+2)(l-1)-\frac{1}{4}}.\label{nus}
\end{align}
\section{Quasinormal bulk-partition function}\label{sec4}
The DHS prescription is quite useful in computing the one-loop partition function in black hole background. However, it is not always possible to find a nice analytic formula for the quasinormal frequency in the black hole geometry. The story is also similar from the Euclidean perspective, where the eigen spectrum of the kinetic operator takes a complicated form in a black hole background. Since the Einstein metric $S^2\times S^2$ can be interpreted as the Euclidean continuation of the Nariai geometry, we can evaluate the partition function analytically from the quasinormal modes.
In this section, we will study the one-loop character partition function of gravitons in Nariai geometry.
From the quasinormal modes given in \eqref{QNMmodes}, we can derive the character partition function of these tensor modes. We plug these (anti)quasinormal modes into the formula \eqref{qnmdef} and obtain
\begin{align}
\log\mathcal{Z}^{(s=2)}_{\rm{ch}}&=\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{n=0,l=2}^{\infty}5(2l+1)\left(e^{-t(n+\frac{1}{2}+i\nu_2)}+e^{-t(n+\frac{1}{2}-i\nu_2)}\right)\nonumber\\
&=\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=2}^{\infty}5(2l+1)\frac{e^{-\frac{t}{2}-i\nu_2t}+e^{-\frac{t}{2}+i\nu_2 t}}{1-e^{-t}}.\label{tensorch}
\end{align}
Note that quasinormal modes do not depend on the spin degeneracy and the azimuthal quantum number $l\leq m\leq l$. For spin-2 field, sum over the azimuthal quantum number together with the spin degenracy gives $5(2l+1)$ factor.
In the second line of \eqref{tensorch}, we perform the sum over $n$.
Similarly, with the vector (anti)quasinormal modes, we compute the quasinormal character partition function:
\begin{align}
\log\mathcal{Z}^{(s=1)}_{\rm{ch}}&=\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=1}^{\infty}3(2l+1)\frac{e^{-\frac{t}{2}-i\nu_1t}+e^{-\frac{t}{2}+i\nu_1t}}{1-e^{-t}}.\label{vecch}
\end{align}
Combining, the tensor, and the vector quasinormal character partition function, we can write
\begin{align}
\log\mathcal{Z}^{\rm{bulk}}_{\rm{ch}}&=\sum_{s=1}^{2}\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=s}^{\infty}(-1)^{s-1}d_{l,s}\frac{e^{-\frac{t}{2}+i\nu_s t}+e^{-\frac{t}{2}-i\nu_s t}}{1-e^{-t}}.\label{bulkch}
\end{align}
Here $d_{l,2}=5(2l+1)$ and $d_{l,1}=3(2l+1)$.
The character bulk-partition function admits a nice form in terms of its tensor and vector components.
However, the complete bulk partition function also includes the volume of the gauge group. The quasinormal product does not give us the full bulk partition function.
For completeness, let us also present the quasinormal character partition function of a minimally coupled massless scalar in Nariai geometry. Using the quasinormal modes of scalars, we find
\begin{align}
\log\mathcal{Z}^{s=0}_{\rm{ch}}&=\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=0}^{\infty}(2l+1)\frac{e^{-\frac{t}{2}+i\nu_0 t}+e^{-\frac{t}{2}-i\nu_0 t}}{1-e^{-t}}\label{scch}
\end{align}
The character partition function of the minimally coupled massless scalar agrees with \cite{Law:2022zdq}.
For propagating modes of graviton in the bulk geometry, we should consider a transverse traceless degree of freedom. We will compute the character partition function from the Euclidean method using the eigen spectrum of the Laplacian in the following section \ref{sc6}.
\section{One-loop determinant of a scalar on $S^2\times S^2$}
In this section, we focus in evaluating one-loop determinant of a scalar using the heat-kernel approach. We will follow closely \cite{Anninos:2020hfj} to obtain an integral representation of the kernel. Finally, we will match the one-loop determinant of a massless scalar with the quasinormal bulk partition function.
The one-loop partition function of a minimally coupled free massless scalar field on $S^2\times S^2$ is given by
\begin{align}
\log\mathcal{Z}&=-\frac{1}{2}\log\det\Delta_0.
\end{align}
Here, $\Delta_0$ is the determinant of a scalar Laplacian on $S^2\times S^2$.
The eigen spectrum of the scalar Laplacian on $S^2\times S^2$ is given by \cite{Volkov:2000ih}
\begin{align}
\lambda^0&=n(n+1)+l(l+1),\quad n,l\geq 0\nonumber\\
d^0_{n,l}&=(2n+1)(2l+1).
\end{align}
The heat-kernel corresponding to the scalar eigen spectrum can be evaluated
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(0)}_{{n,l}}s}\right)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{n=0,l=0}^{\infty}d_{n,l}\, e^{-\lambda^{(0)}_{{n,l}}s}\label{scalrkernel}
\end{align}
In general, the sum over $n$ or $l$ is difficult to perform, because the eigenvalues are quadratic in $n$ and $l$. Following \cite{Anninos:2020hfj}, we use Hubbard-Stratonovich trick to linearise the sum
\begin{equation}
\sum_{n=0}^{\infty} (2n+1) e^{-s (n + \tfrac{1}{2})^2}
= \int_{C} \frac{du}{\sqrt{4\pi s}} e^{-\frac{u^2}{4s}} f(u).
\end{equation}
The contour $C$ runs from $-\infty$ to $\infty$, slightly above the real axis (small positive imaginary constant value). $f(u)$ is also given by
\begin{align}
f(u)=\sum_{n=0}^{\infty}(2n+1)e^{iu(n+\frac{1}{2})}=\frac{e^{\frac{i u}{2}} \left(1+e^{i u}\right)}{\left(-1+e^{i u}\right)^2}\label{fdef}
\end{align}
We plug $f(u)$ into \eqref{scalrkernel} and perform the integral over $s$
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(0)}_{{n,l}}s}\right)= \sum_{l=0}^{\infty}(2l+1)\Big(\int_{C} \frac{du}{2\sqrt{u^2 + \epsilon^2}} \left( e^{-\nu_0 \sqrt{u^2 + \epsilon^2}} f(u) \right)
\end{align}
Note that, the above integral has a branch cut from $u=i\epsilon$ along the imaginary axis. We deform the contour $C$ to $C'$ that runs on both sides of the
branch cut on the imaginary axis and substitute $u=it$. We take $\epsilon\rightarrow 0$ at the end to obtain
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(0)}_{{n,l}}s}\right)&=\sum_{l=0}^{\infty}(2l+1)\int_{0}^{\infty} \frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\frac{e^{-\frac{t}{2}+i\nu_0t}+e^{-\frac{t}{2}-i\nu_0 t}}{1-e^{-t}}\label{scdet}
\end{align}
We find that the one-loop heat-kernel of a massless scalar on $S^2\times S^2$ agrees with the quasinormal chracter partition function presented in \eqref{scch}.
\section{One-loop determinant of graviton on $S^2\times S^2$}\label{sc6}
The gauge fixed one-loop path integral of graviton on $S^2\times S^2$ is given by \cite{Volkov:2000ih}
\begin{align}
\mathcal{Z}^{1-\rm{loop}}_{\rm{grav}}&=\frac{\Omega_2}{\Omega_1}\frac{\det(\Delta_{(1)})^{\frac{1}{2}}}{\det(\Delta_{(2)})^{\frac{1}{2}}},\label{grone}
\end{align}
where $\Delta_2$ is operator for the transverse traceless tensor fluctuation:
\begin{align}
\Delta_2 h_{\mu\nu}=-\nabla^2h_{\mu\nu}-2R_{\mu\nu\alpha\beta}h^{\alpha\beta}.
\end{align}
The vector operator acts on the co-exact vector $\xi_{\mu}$:
\begin{align}
\Delta_1:=-\nabla_{\mu}\nabla^{\mu}-1.
\end{align}
Here, $\Omega_2$ is the contribution from a negative tensor mode on $S^2\times S^2$. $\Omega_1$ is the isometry factor which in this case would be $\rm{vol}(SO(3))\times \rm{vol}(SO(3))$. The details of the gauge fixing procedure along with the counting zero modes and determining path integral measure has been shown in \cite{Volkov:2000ih}. The complete bulk-edge partition function includes the character partition function along with the isometry factor $\Omega_1$. In the following section, we will focus on evaluating the heat-kernels of the determinants isolating the bulk and the edge parts of the character partition function.
\subsection{Eigen spectrum of Laplacians}
From \eqref{grone}, the spectrum of the operator $(\Delta_{(2)})$ acting on the transverse traceless graviton on $S^2\times S^2$ are given by \cite{Volkov:2000ih}
\paragraph{Tensor spectrum:}
\begin{align}
\lambda^{(2a)}_{n,l} &= n(n+1) + l(l +1)-2, \quad n\geq 2, l\geq 2\nonumber\\
d_{n,l}&=5(2n+1)(2l+1)
\label{tens1}
\end{align}
In addition to this, we have
\begin{align}
\lambda^{(2b)}_{n}&=n(n+1),\quad n\geq 2\nonumber\\
d_n&=18(2n+1)\label{tens2}
\end{align}
And finally, we have another set of eigen spectrum
\begin{align}
\lambda^{(2c)}_{n}&=n(n+1)-2,\quad n\geq 2\nonumber\\
d_n&=2(2n+1)\label{tens3}
\end{align}
In addition to the three different set of eigen spectrum of the tensor Laplacian, we have nine tensor eigen modes with eigenvalue $2$ and a single tensor mode with a negative eigenvalue $-2$.
\paragraph{Vector spectrum}:
Let us also write down the eigen spectrum of the operator $(\Delta_{(1)})$ operator \cite{Volkov:2000ih}
\begin{align}
\lambda^{(1a)}_{n,l} &= (n(n+1) - 1) + (l(l + 1) - 1),\quad n\geq 1, l\geq 1\nonumber\\
d_{n,l}&=3(2n+1)(2l+1),
\label{vecs1}.
\end{align}
In addition to this spectrum, we also have
\begin{align}
\lambda^{(1b)}_{n} &= (n(n+1) - 2) ,\quad n\geq 2, \nonumber\\
d_n&=2(2n+1)\label{vecs2}
\end{align}
It is important to note that the tensor spectrum $\lambda^{(2,c)}_{n}$ is the same as the vector spectrum $\lambda^{(1b)}_{n}$, that is, $\lambda^{(2,c)}_{n}=\lambda^{(1b)}_{n}$. Due to this isospectral condition, we get a cancelation between these terms in the heat kernel.
\subsection{One-loop determinant on $S^2\times S^2$}\label{edge}
Let us examine the one-loop determinant from the tensor eigen spectrum in \eqref{tens1}. The heat-kernel corresponding to the tensor eigen modes is given by
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2a)}_{{n,l}}s}\right)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{n=2,l=2}^{\infty}d_{n,l}\, e^{-\lambda^{(2a)}_{{n,l}}s}\nonumber\\
&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{l=2}^{\infty}5(2l+1)\sum_{n=2}^{\infty}(2n+1) \, e^{-s\left((n+\frac{1}{2})^2-\nu_2^2\right)}\nonumber\\
&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{l=2}^{\infty}5(2l+1)\Big(\sum_{n=0}^{\infty}(2n+1) \, e^{-s\left((n+\frac{1}{2})^2-\nu_2^2\right)}\nonumber\\
&~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-\sum_{n=0}^{1}(2n+1) \, e^{-s\left((n+\frac{1}{2})^2-\nu_2^2\right)}\Big),
\label{tensordet}
\end{align}
where $\nu_2$ is defined in \eqref{nus}.
In the third line, we added the modes with $n=0$ and $n=1$ and subtracted it from the main kernel. We would like to pay special attention to the first term (with all modes from $n=0$ to $\infty$). Finally, we will subtract the $n=0$ and $n=1$ modes from the kernel.
In general, it is difficult to perform the sum over $n$, since it is quadratic in the exponential. However, using the Hubbard-Stratonovich trick, one can linearize the exponential and perform the sum. The linearized sum over $n$ is evaluated in \eqref{fdef}. Following the same trick as shown for the scalar,
we plug the $f(u)$ into \eqref{tensordet} and perform the integral over $s$ to obtain
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2a)}_{{n,l}}s}\right)&=\sum_{l=2}^{\infty}5(2l+1)\Big(\int_{0}^{\infty} \frac{ds}{2s} \int_{C} \frac{du}{\sqrt{4\pi s}} e^{-\frac{u^2}{4s}} e^{s \nu_2^2} f(u)\nonumber\\
&- \int_0^{\infty}\frac{dt}{2t}\sum_{n=0}^{1}(2n+1) e^{-t\left((n+\frac{1}{2})^2-\nu_2^2\right)}\Big)\nonumber\\
&=\sum_{l=2}^{\infty}5(2l+1)\Big(\int_{C} \frac{du}{2\sqrt{u^2 + \epsilon^2}} \left( e^{-\nu_s \sqrt{u^2 + \epsilon^2}} f(u) \right)\nonumber\\
&- \int_0^{\infty}\frac{dt}{2t}\sum_{n=0}^{1}(2n+1) e^{-t\left((n+\frac{1}{2})^2-\nu_2^2\right)}\Big)
\end{align}
In the second line, we performed the integral over $s$. Let us also deform the contour from $C$ to $C'$ and collect the contribution through the branch cut (with the branch cut point $u=i\epsilon$) and substitute $u=it$. Finally, we take $\epsilon\rightarrow 0$ limit to obtain
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2a)}_{{n,l}}s}\right)&=\sum_{l=2}^{\infty}5(2l+1)\Big(\int_{0}^{\infty} \frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\frac{e^{-\frac{t}{2}+i\nu_2t}+e^{-\frac{t}{2}-i\nu_2 t}}{1-e^{-t}}\nonumber\\&-\int_0^{\infty}\frac{dt}{2t}\sum_{n=0}^{1}(2n+1) e^{-t\left((n+\frac{1}{2})^2-\nu_2^2\right)}\Big)\label{finaldettens}
\end{align}
We observe that the first term of the one-loop determinant of the tensor modes on $S^2\times S^2$ in \eqref{finaldettens} agrees with the quasinormal partition function in \eqref{tensorch}. However, the second term is not present in the quasinormal partition function.
We can write \eqref{finaldettens}, in a compact form by performing a single sum over $n=2$ to $\infty$, which is given by
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2a)}_{{n,l}}s}\right)&=\int_0^{\infty}\frac{dt}{2t}\sum_{l=2}^{\infty}5(2l+1)\left(e^{-i\nu_2 t}+e^{i\nu_2 t}\right) \frac{ e^{-\frac{3 t}{2}} \left(5 e^t-3\right)}{\left(e^t-1\right)^2} \label{1ht}
\end{align}
The complete integral representation of the of the heat-kernel from the eigen spectrum of the tensor modes in \eqref{1ht} agrees with the equation (C.7) of \cite{Anninos:2025ltd}. However, we wish to separate the terms with modes $n=0$ and $n=1$, which can contribute to the quasinormal edge characters.
Similarly, we can evaluate the heat-kernel formula for the transverse vector spectrum in \eqref{vecs1}.
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(1a)}_{{n,l}}s}\right)&=\sum_{l=1}^{\infty}3(2l+1)\Big(\int_{0}^{\infty} \frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\frac{e^{-\frac{t}{2}+i\nu_1t}+e^{-\frac{t}{2}-i\nu_1 t}}{1-e^{-t}}\nonumber\\&-\int_0^{\infty}\frac{dt}{2t} e^{-t\left((\frac{1}{2})^2-\nu_1^2\right)}\Big).\label{finaldetvecs}
\end{align}
Here $\nu_1=\sqrt{l(l+1)-\frac{9}{4}}$
and again, we added the $n=0$ mode in the kernel and subtracted off. The first term in \eqref{finaldetvecs} agrees with the quasinormal bulk characters in \eqref{vecch}. It seems that the $n=0$ mode in the vector spectrum contributes to the quasinormal edge character.
From the first terms in \eqref{finaldettens} and \eqref{finaldetvecs}, we can write the one-loop determinant of the tensor of the vector sector in the combined form:
\begin{align}
\log\mathcal{Z}^{\rm{Bulk}}_{\rm{ch}}&=\sum_{s=1}^{2}\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\sum_{l=s}^{\infty}(-1)^{s-1}d_{l,s}\frac{e^{-\frac{t}{2}+i\nu_s t}+e^{-\frac{t}{2}-i\nu_s t}}{1-e^{-t}}, \quad s\in(1,2)\label{bpart}
\end{align}
To obtain the complete one-loop determinant, we have to combine both the tensor and the vector heat-kernel with the appropriate sign \eqref{grone}.
We observe that \eqref{bpart} agrees with the quasinormal bulk partition function in the Nariai geometry. We would like to interpret \eqref{bpart} as a bulk character partition function.
As we have already mentioned, the isospectral condition of the tensor and vector modes $\lambda^{(2,c)}_{n}=\lambda^{(1b)}_{n}$ simplifies the rest of the computation of the heat kernel. It is clear from the eigen modes and the degeneracy, in \eqref{tens2} and \eqref{vecs2}, we get a cancelation of the heat kernels of the tensor and vector eigen modes mentioned above in the heat kernel.
We are now left with the eigenspectra in \eqref{tens2}. The heat-kernel corresponding to this eigen spectrum is given by
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2b)}_{{n}}s}\right)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{n=2}^{\infty}18(2n+1)\, e^{-\lambda^{(2b)}_{{n}}s}
\end{align}
Using the Hubbard-Stratonovich trick, we can perform the sum over $n$-modes.
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2b)}_{{n}}s}\right)&=\int_0^{\infty}\frac{dt}{2t} \frac{18 e^{-\frac{3 t}{2}} \left(5 e^t-3\right)}{\left(e^t-1\right)^2}(e^{-i\frac{t}{2}}+e^{i\frac{t}{2}})
\end{align}
In addition , to these kernels we have nine constant eigen modes of $2$ and one negative eigen mode $-2$. In general, the path integral does not converge with the negative eigen mode, but one can incorporate this mode by taking its absolute value. Combining all these constant modes, we can construct the heat-kernel and denote the partition function from these modes as $\mathcal{Z}_0$.
Let us now write the one-loop partition function that does not belong to the bulk partition function or the quasinormal bulk character partition function. We call this as edge partition function of graviton
\begin{align}
\log\mathcal{Z}^{\rm{edge}}_{\rm{ch}}&=\int_0^{\infty}\frac{ds}{2s}\,e^{-\frac{\epsilon^2}{4s}}\Big(-\sum_{l=2}^{\infty} 5(2l+1)\sum_{n=0}^1(2n+1)\,e^{-\lambda^{(1a)}_{n,l}s}+\sum_{n=2}^{\infty}18(2n+1)\,e^{-s\lambda^{(2b)}_{n}}\nonumber\\
&+\sum_{l=1}^{\infty}3(2l+1)\,e^{-s\lambda^{(1a)}_{0,l}}\Big)\label{edge1}
\end{align}
The first term in \eqref{edge1}, comes from the $n=0$ and $n=1$ modes in \eqref{tensordet}. The second term in \eqref{edge1} arises from the tesnor eigen spectrum in \eqref{tens2} which does not have the quasinormal analogue presented in \eqref{tensorch}. The third term comes from the $n=0$ mode subtraction from the vector eigenspectrum \eqref{vecs1} which is also not present in the quasinormal mode character partition function.
From a direct computation of the one-loop determinant for gravitons using the heat kernel method on \(S^2 \times S^2\), we are able to isolate the edge character partition function. The complete edge partition function can be written by incorporating the isometry factor $\Omega_1$ in \eqref{grone} and can be written as a sum over two terms
\begin{align}
\log\mathcal{Z}_{\rm{edge}}&=\log\mathcal{Z}^{\rm{edge}}_{\rm{ch}}+\log\Omega_1.
\end{align}
Unlike the case of a minimally coupled scalar field, the one-loop partition function obtained from quasinormal modes does not exactly match the result from the heat-kernel approach. At present, it remains unclear how to incorporate the edge contribution within the quasi-normal mode framework. However, this discrepancy suggests that the boundary conditions for higher-spin fields must be treated with greater care near the horizon. It is conceivable that the edge modes can be interpreted as localized scalar and vector degrees of freedom.
%\textcolor{blue}{Check all $\lambda^{1,1}$ typo errors, be sure, we have included evrything, what is the physical origin of the QNMs of the edge, change bc? missing QNMs?}
\begin{comment}
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(2)}_{{n,l}}s}+e^{-\lambda^{(1)}_{{n,l}}s}+e^{-\lambda^{(0)}_{{n,l}}s}\right) &=\sum_{\alpha=0}^{2}\sum_{l=2}^{\infty}d_l\int_{0}^{\infty} \frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\frac{e^{-\frac{t}{2}+i\nu_\alpha t}+e^{-\frac{t}{2}-i\nu_\alpha t}}{1-e^{-t}}
\end{align}
Following a similar procedure, we compute the log of the determinants of vectors and scalars (from the transverse traceless part) using the spectrum given in \eqref{vecs} and \eqref{scalars} respectively.
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(1)}_{{n,l}}s}\right)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\left(\sum_{n=1,l=1}^{\infty}d_nd_l\,e^{-s\lambda^{(1a)}_{n,l}}+\sum_{n=1,l=2}^{\infty}d_nd_l\,e^{-s\lambda^{(1b)}}\right)\label{trv}\\
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{(0)}_{{n,l}}s}\right)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\left(\sum_{n=2,l=2}^{\infty}d_nd_l\,e^{-s\lambda^{(0a)}_{n,l}}+\sum_{n=2,l=1}^{\infty}d_nd_l\,e^{-s\lambda^{(0b)}}+\sum_{n=0,l=0}^{\infty}d_nd_l\,e^{-s\lambda^{(0c)}_{n,l}}\right)\label{trs}
\end{align}
\paragraph{Ghost determinants:}
Let us also evaluate the ghost determinants in a similar way. The spectrum for the ghosts are given in \eqref{vecgh} and \eqref{ghscalar1} and \eqref{ghscalar2}.
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{'(1)}_{{n,l}}s}\right)&= \int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\left(\sum_{n=1,l=1}^{\infty}d_nd_l\,e^{-s\lambda'^{(1)}_{n,l}}\right)\label{ghv}\\
\int_0^{\infty}\frac{ds}{2s}\rm{Tr}\left( e^{-\lambda^{'(0)}_{{n,l}}s}\right)& =\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\left(\sum_{n=1,l=0}^{\infty}d_nd_l\,e^{-s\lambda'^{(0a)}_{n,l}}+\sum_{n=1,l=1}^{\infty}d_nd_l\,e^{-s\lambda'^{(0b)}_{n,l}}\right)\label{ghs}
\end{align}
From the condition $\lambda^{(1,b)}_{n,l}=\lambda'^{(1)}_{n,l}$, $\lambda'^{(0b)}=\lambda^{(0c)}$ and $\lambda^{'(0a)}=\lambda^{(0b)}$ we can see a lot of cancellations between ghosts and the transverse part of the determinants.
We now combine, all the terms from transverse, traceless part and the ghost determinants in \eqref{finaldettens}, \eqref{trv}, \eqref{trs},\eqref{ghv} and \eqref{ghs}, and obtain:
\begin{align}
\int_0^{\infty}\frac{ds}{2s}\,e^{-\frac{\epsilon^2}{4s}}K(s)&=\int_0^{\infty}\frac{ds}{2s}e^{-\frac{\epsilon^2}{4s}}\sum_{n=0,l=2}^{\infty}d_n\,d_l\, \left(e^{-s\lambda^{(2)}_{n,l}}+e^{-s\lambda^{(1)}_{n,l}}+e^{-s\lambda^{(0)}_{n,l}}\right)\nonumber\\
&+
\end{align}
\textcolor{red}{Complete this}
\end{comment}
\section{Discussion}
In this paper, we evaluate the bulk character partition function of scalars, vectors, and gravitons from their quasinormal mode spectra using the Denef-Hartnoll-Sachdev (DHS) prescription. For minimally coupled scalar fields, we demonstrate that the quasinormal mode character partition function agrees precisely with the one-loop determinant computed via the heat kernel method on the $S^2 \times S^2$ background. In contrast, for gravitons, the full one-loop determinant obtained from the heat kernel approach does not coincide with the quasinormal mode character partition function. We find that only the bulk contribution matches, while the discrepancy arises from edge modes. We isolate the edge-character partition function from the heat-kernel expression to account for this difference.
In \(\mathrm{AdS}\), the Denef-Hartnoll-Sachdev (DHS) prescription \cite{Castro:2017mfj} requires modification for fields with spin \(s > 0\), as certain higher-spin modes become singular at the horizon. Specifically, it has been shown that for a spin-\(s\) field, the modes with the Matsubara frequency number \(|k| < s\) are singular and must be excluded from the DHS product formula. This suggests that, the boundary conditions for higher-spin fields should be treated with greater care near the horizon. It would be interesting to understand the origin of the edge character partition function from the perspective of quasinormal modes in de Sitter black holes.
We aim to interpret the edge character partition function of gravitons in terms of a putative entropy associated with gravitational edge modes \cite{David:2022jfd,David:2021wrw, Mukherjee:2023ihb, Mukherjee:2021rri}. In four dimensions (\(D = 4\)), it has been demonstrated that the entanglement entropy of the gravitational edge modes \cite{David:2022jfd} matches the edge character partition function computed on \(S^4\) \cite{Anninos:2020hfj}. However, for the case at hand, no analogous independent computation is currently available. We hope to pursue this direction in the near future, which may offer deeper insight into the physical interpretation of the edge character partition function.
\section*{Acknowledgement}
The author thanks Shiraz Minwalla and Justin David for discussions and encouragement.
\appendix
\section{Quasinormal modes in Nariai geometry}\label{app1}
The metric of the Nariai space is given by
\begin{align}
ds^2=-f(r)dt^2+\frac{dr^2}{f(r)}+R^2d\Omega^2,\quad f(r)=1-\frac{r^2}{\ell_N^2}.\label{metric}
\end{align}
Let us consider a minimally coupled massless scalar in this background. The wave equation obeyed by the minimally coupled scalar is given by
\begin{align}
\frac{1}{\sqrt{-g}}\partial_{\mu}\left(\sqrt{-g}g^{\mu\nu}\partial_{\nu}\right)\Phi=0
\end{align}
The wave equation takes a simple Schr\"odinger like form in the tortoise coordinate:
\begin{align}
\left[\frac{1}{f(r)}\left(\partial_{r_*}^2-\partial_{t}^2\right)+\nabla_{\Omega}^2\right]\Phi=0,\label{waveform}
\end{align}
where $\nabla^2_{\Omega}$ is the scalar Laplacian on $S^2$. For simplicity, we will consider $R=1$.
To solve the wave equation, we can take the following mode ansatz:
\begin{align}
\Phi(t,r,\Omega)=\psi(r)e^{-i\omega t}Y_{l,m}(\Omega),
\end{align}
where $Y_{l,m}(\Omega)$ is the scalar spherical harmonics on $S^2$. The wave equation \eqref{waveform} can be written in a simple equation of the radial coordinate.
\begin{align}
\left[\partial_{r_*}^2+\omega^2-V_0(r_*)\right]\psi(r)=0\label{mastereq}
\end{align}
The scalar potential $V_0(r_*)$ is given by
\begin{align}
V_0(r_*)=\frac{l(l+1)}{\cosh^2(\frac{r_*}{\ell_N})}
\end{align}
The above potential admits a well known Pöschl-Teller potential \cite{Cardoso:2003sw,Venancio:2020ttw}. The standard technique to solve the equation \eqref{mastereq} is the following:
We first define $\xi$
\begin{align}
\xi=\frac{1}{2}+\frac{1}{2}\tanh(\frac{r_*}{\ell_N}).
\end{align}
The asymptotic boundary $r_*\rightarrow \infty$ corresponds to $\xi\rightarrow 1$. Let us also define the following parameters
\begin{align}
a&=\frac{1}{2}+i\sqrt{l(l+1)-\frac{1}{4}},\\
b&=\frac{1}{2}-i\sqrt{l(l+1)-\frac{1}{4}}\\
c&=i\omega\ell_N+1.\label{abc}
\end{align}
With the parameters defined in \eqref{abc}, the wave equation takes a simple hypergeometric form in $\xi$ variable under the following redefinition of $\psi(r_*)$:
\begin{align}
\psi(r_*) = \xi^{(c - 1)/2} (1 - \xi)^{\frac{1}{2}(a + b - c)} R(\xi)
\end{align}
The equation in $R(\xi)$ is given by
\begin{align}
\xi(1 - \xi) \frac{d^2 R(\xi)}{d\xi^2} + [c - \xi(a + b + 1)] \frac{dR(\xi)}{d\xi} - ab\, R (\xi)= 0.
\end{align}
The solution of the above equation is given by
\begin{align}
R(\xi) = A\, F(a, b, c; \xi) + B\, \xi^{1 - c} F(1 + a - c, 1 + b - c, 2 - c; \xi).
\end{align}
where $F(a,b,c,\xi)$ is the usual Hypergeometric function.
We should impose appropriate boundary conditions to obtain the quasinormal modes. Let us investigate the behavior of the mode functions at $r_*\rightarrow \pm \infty$.
We impose the following boundary conditions: at $r_* \rightarrow -\infty$, the modes should behave like $\sim e^{-i\omega (t+r_*)}$ and at $r_*\rightarrow \infty$, the modes should go like $e^{-i\omega(t-r_*)}$. These set of boundary conditions imply the outgoing waves at the aymptotic infinity and the ingoing waves at the horizon \cite{PhysRevD.30.295}
\begin{align}
\left. \psi(r_*) \right|_{r_* \to +\infty} \rightarrow
&\ B\, \frac{\Gamma(c - a - b)\Gamma(2 - c)}{\Gamma(1 - a)\Gamma(1 - b)}\, e^{-i\omega(t - r_*)} \notag \\
&+ B\, \frac{\Gamma(a + b - c)\Gamma(2 - c)}{\Gamma(a - c + 1)\Gamma(b - c + 1)}\, e^{-i\omega(t + r_*)}.
\end{align}
Therefore, we should reduce the coefficient of $e^{-i\omega(t+r_*)}$ which can be done by setting $a-c+1=-n$ or $b-c+1=-n$, where $n$ is a non-negative integer. This boundary condition is analogous to the boundary condition in the horizon. Choosing the former condition with the definitions in \eqref{abc}, we find
\begin{align}
\frac{ i\omega_*}{2\pi T_N}=n+\frac{1}{2}+i\nu_0,
\end{align}
where $\nu_0=\sqrt{l(l+1)-\frac{1}{4}}.$ These set of quasinormal mode frequencies has been computed in \cite{Cardoso:2003sw, Venancio:2020ttw}.
For gravitational perturbation, the form of the Pöschl-Teller potential type changes, which is mainly due to the different eigen spectrum of vector and tensor modes on $S^2$.
For the form of the potential are given by
\begin{align}
%V_1(r_*)&=\frac{l(l+1)}{\cosh^2(\frac{r_*}{\ell_N})}\\
V_2(r_*)&=\frac{(l+2)(l-1)}{\cosh^2(\frac{r_*}{\ell_N})}\label{pots}
\end{align}
Following the same procedure as scalar, we can compute the quasinormal modes of the gravitational perturbations. The quasinormal mode frequencies for gravitational perturbations take the following form
\begin{align}
\frac{ i\omega_*}{2\pi T_N}=n+\frac{1}{2}+i\nu_2,\quad n\geq 0
\end{align}
where $\nu_2=\sqrt{(l+2)(l-1)-\frac{1}{4}}$. Here, the quantum number $l$ runs from $2$ to $\infty$.
\bibliographystyle{JHEP}
\bibliography{biblio.bib}
\end{document}
\subsection*{ \quad Tensor spectrum}
The transverse-traceless spectrum of the operator $\Delta_{(2)} - 2\Lambda$ on $S^2 \times S^2$ is given by
\begin{align}
\tilde{\lambda}^{(2)}_n &= \left(n(n+1) - 2\right)\Lambda \,, &
\tilde{d}^{(21)}_n &= 2(2n+1) \,,
\end{align}
with $n \in \mathbb{Z}^+ + 1$. Moreover, we have
\begin{align}
\tilde{\gamma}^{(2)}_n &= n(n+1)\Lambda \,, &
\tilde{d}^{(22)}_n &= 18(2n+1) \,
\end{align}
with $n \in \mathbb{Z}^+ + 1$, and finally
\begin{align}
\tilde{\delta}^{(2)}_n &= \left(n_1(n_1+1) + n_2(n_2+1) - 2\right)\Lambda \,, &
\tilde{d}^{(23)}_{n_1 n_2} &= 5(2n_1 + 1)(2n_2 + 1) \,, \label{sp2}
\end{align}
with $n_1, n_2 \in \mathbb{Z}^+ + 1$. In addition there is a single negative tensor eigenmode with eigenvalue $-2\Lambda$, and nine tensor eigenmodes with eigenvalue $2\Lambda$. The negative mode corresponds to changing the relative size of the two-spheres without breaking any spherical symmetry.
One loop determinant with the spectrum \eqref{sp2}
\begin{align}
\mathcal{I}_{a,\mu}& = \sum_{n_1, n_2 \geq 2} (2n_1 + 1)(2n_2 + 1) \int_{\mathbb{R}^{+}} \frac{d\tau}{2\tau}
\, e^{-\frac{\varepsilon^2}{4\tau}}
\, e^{-(n_1(n_1+1)\Lambda -2)\tau}
\, e^{-(n_2(n_2+1)\Lambda -2)\tau} \nonumber\\
&=\sum_{n_2\geq 2}(2n_2+1)\int_0^{\infty}\frac{dt}{2t}\frac{1+e^{-t}}{1-e^{-t}}\left(e^{-i\nu_l t}\frac{e^{-\frac{3 t}{2}} \left(5 e^t-3\right)}{\left(e^t-1\right)^2}+e^{i\nu_l t}\frac{e^{\frac{5 t}{2}} \left(5-3 e^t\right)}{\left(e^t-1\right)^2}\right)
\end{align}
\subsection{QNM modes}
The quasinormal modes for the tensor perturbation is given by:
\begin{align}
\omega_*&=\left(\sqrt{l(l+1)-\frac{9}{4}}-i\left(n+\frac{1}{2}\right)\right)\nonumber\\
&=\left(\nu_l-i\left(n+\frac{1}{2}\right)\right)
\end{align}
With these QNMs, let us evaluate the character
\begin{align}
\chi_{\rm{QNM}}&=\sum_{l=2}^{\infty}(2l+1)\sum_{n=2}^{\infty}(2n+1)\left(e^{-i\omega_* t}+e^{-i\bar{\omega_* }t}\right),\quad \bar{\omega_* }=-\omega_*.\nonumber\\
&=\sum_{l=2}^{\infty}(2l+1)\left(e^{-i\nu_l t}\frac{e^{-\frac{3 t}{2}} \left(5 e^t-3\right)}{\left(e^t-1\right)^2}+e^{i\nu_l t}\frac{e^{\frac{5 t}{2}} \left(5-3 e^t\right)}{\left(e^t-1\right)^2}\right)
\end{align}
\begin{align}
\log\mathcal{Z}_{\rm{QNM}}&=\int_0^\infty \frac{dt}{2t} \frac{1 + e^{-2\pi t/\beta_H}}{1 - e^{-2\pi t/\beta_H}}\chi_{\rm{QNM}}.
\end{align}
\end{document}
\end{document}
\end{document}
\providecommand{\href}[2]{#2}\begingroup\raggedright\begin{thebibliography}{10}
\bibitem{Banerjee:2010qc}
S.~Banerjee, R.K.~Gupta and A.~Sen, \emph{{Logarithmic Corrections to Extremal Black Hole Entropy from Quantum Entropy Function}}, \href{https://doi.org/10.1007/JHEP03(2011)147}{\emph{JHEP} {\bfseries 03} (2011) 147} [\href{https://arxiv.org/abs/1005.3044}{{\ttfamily 1005.3044}}].
\bibitem{Banerjee:2011jp}
S.~Banerjee, R.K.~Gupta, I.~Mandal and A.~Sen, \emph{{Logarithmic Corrections to N=4 and N=8 Black Hole Entropy: A One Loop Test of Quantum Gravity}}, \href{https://doi.org/10.1007/JHEP11(2011)143}{\emph{JHEP} {\bfseries 11} (2011) 143} [\href{https://arxiv.org/abs/1106.0080}{{\ttfamily 1106.0080}}].
\bibitem{Vassilevich:2003xt}
D.V.~Vassilevich, \emph{{Heat kernel expansion: User's manual}}, \href{https://doi.org/10.1016/j.physrep.2003.09.002}{\emph{Phys. Rept.} {\bfseries 388} (2003) 279} [\href{https://arxiv.org/abs/hep-th/0306138}{{\ttfamily hep-th/0306138}}].
\bibitem{Denef:2009kn}
F.~Denef, S.A.~Hartnoll and S.~Sachdev, \emph{{Black hole determinants and quasinormal modes}}, \href{https://doi.org/10.1088/0264-9381/27/12/125001}{\emph{Class. Quant. Grav.} {\bfseries 27} (2010) 125001} [\href{https://arxiv.org/abs/0908.2657}{{\ttfamily 0908.2657}}].
\bibitem{Datta:2011za}
S.~Datta and J.R.~David, \emph{{Higher Spin Quasinormal Modes and One-Loop Determinants in the BTZ black Hole}}, \href{https://doi.org/10.1007/JHEP03(2012)079}{\emph{JHEP} {\bfseries 03} (2012) 079} [\href{https://arxiv.org/abs/1112.4619}{{\ttfamily 1112.4619}}].
\bibitem{Castro:2017mfj}
A.~Castro, C.~Keeler and P.~Szepietowski, \emph{{Tweaking one-loop determinants in AdS$_{3}$}}, \href{https://doi.org/10.1007/JHEP10(2017)070}{\emph{JHEP} {\bfseries 10} (2017) 070} [\href{https://arxiv.org/abs/1707.06245}{{\ttfamily 1707.06245}}].
\bibitem{Grewal:2022hlo}
M.~Grewal, Y.T.A.~Law and K.~Parmentier, \emph{{Black hole horizon edge partition functions}}, \href{https://doi.org/10.1007/JHEP06(2023)025}{\emph{JHEP} {\bfseries 06} (2023) 025} [\href{https://arxiv.org/abs/2211.16644}{{\ttfamily 2211.16644}}].
\bibitem{Mukherjee:2024nhx}
J.~Mukherjee, \emph{{One loop determinant in the extremal black hole from quasinormal modes}}, \href{https://arxiv.org/abs/2408.06165}{{\ttfamily 2408.06165}}.
\bibitem{Kapec:2024zdj}
D.~Kapec, Y.T.A.~Law and C.~Toldo, \emph{{Quasinormal Corrections to Near-Extremal Black Hole Thermodynamics}}, \href{https://arxiv.org/abs/2409.14928}{{\ttfamily 2409.14928}}.
\bibitem{Arnaudo:2024bbd}
P.~Arnaudo, G.~Bonelli and A.~Tanzini, \emph{{One-loop corrections to near extremal Kerr thermodynamics from semiclassical Virasoro blocks}}, \href{https://arxiv.org/abs/2412.16057}{{\ttfamily 2412.16057}}.
\bibitem{1950SRToh..34..160N}
H.~{Nariai}, \emph{{On some static solutions of Einstein's gravitational field equations in a spherically symmetric case}}, {\emph{Sci. Rep. Tohoku Univ. Eighth Ser.} {\bfseries 34} (1950) 160}.
\bibitem{Law:2025yec}
Y.T.A.~Law and V.~Lochab, \emph{{Gravitons on Nariai Edges}}, \href{https://arxiv.org/abs/2506.02142}{{\ttfamily 2506.02142}}.
\bibitem{Camporesi:1994ga}
R.~Camporesi and A.~Higuchi, \emph{{Spectral functions and zeta functions in hyperbolic spaces}}, \href{https://doi.org/10.1063/1.530850}{\emph{J. Math. Phys.} {\bfseries 35} (1994) 4217}.
\bibitem{Cardoso:2003sw}
V.~Cardoso and J.P.S.~Lemos, \emph{{Quasinormal modes of the near extremal Schwarzschild-de Sitter black hole}}, \href{https://doi.org/10.1103/PhysRevD.67.084020}{\emph{Phys. Rev. D} {\bfseries 67} (2003) 084020} [\href{https://arxiv.org/abs/gr-qc/0301078}{{\ttfamily gr-qc/0301078}}].
\bibitem{Venancio:2020ttw}
J.~Ven\^ancio and C.~Batista, \emph{{Spin-2 Quasinormal Modes in Generalized Nariai Spacetimes}}, \href{https://doi.org/10.1103/PhysRevD.101.084037}{\emph{Phys. Rev. D} {\bfseries 101} (2020) 084037} [\href{https://arxiv.org/abs/2002.08449}{{\ttfamily 2002.08449}}].
\bibitem{Law:2022zdq}
Y.T.A.~Law and K.~Parmentier, \emph{{Black hole scattering and partition functions}}, \href{https://doi.org/10.1007/JHEP10(2022)039}{\emph{JHEP} {\bfseries 10} (2022) 039} [\href{https://arxiv.org/abs/2207.07024}{{\ttfamily 2207.07024}}].
\bibitem{Anninos:2020hfj}
D.~Anninos, F.~Denef, Y.T.A.~Law and Z.~Sun, \emph{{Quantum de Sitter horizon entropy from quasicanonical bulk, edge, sphere and topological string partition functions}}, \href{https://doi.org/10.1007/JHEP01(2022)088}{\emph{JHEP} {\bfseries 01} (2022) 088} [\href{https://arxiv.org/abs/2009.12464}{{\ttfamily 2009.12464}}].
\bibitem{Volkov:2000ih}
M.S.~Volkov and A.~Wipf, \emph{{Black hole pair creation in de Sitter space: A Complete one loop analysis}}, \href{https://doi.org/10.1016/S0550-3213(00)00287-X}{\emph{Nucl. Phys. B} {\bfseries 582} (2000) 313} [\href{https://arxiv.org/abs/hep-th/0003081}{{\ttfamily hep-th/0003081}}].
\bibitem{Anninos:2025ltd}
D.~Anninos, C.~Baracco, S.~Brian and F.~Denef, \emph{{Features of the Partition Function of a $\Lambda>0$ Universe}}, \href{https://arxiv.org/abs/2505.11330}{{\ttfamily 2505.11330}}.
\bibitem{David:2022jfd}
J.R.~David and J.~Mukherjee, \emph{{Entanglement entropy of gravitational edge modes}}, \href{https://doi.org/10.1007/JHEP08(2022)065}{\emph{JHEP} {\bfseries 08} (2022) 065} [\href{https://arxiv.org/abs/2201.06043}{{\ttfamily 2201.06043}}].
\bibitem{David:2021wrw}
J.R.~David and J.~Mukherjee, \emph{{Partition functions of p-forms from Harish-Chandra characters}}, \href{https://doi.org/10.1007/JHEP09(2021)094}{\emph{JHEP} {\bfseries 09} (2021) 094} [\href{https://arxiv.org/abs/2105.03662}{{\ttfamily 2105.03662}}].
\bibitem{Mukherjee:2023ihb}
J.~Mukherjee, \emph{{Entanglement entropy and the boundary action of edge modes}}, \href{https://doi.org/10.1007/JHEP06(2024)113}{\emph{JHEP} {\bfseries 06} (2024) 113} [\href{https://arxiv.org/abs/2310.14690}{{\ttfamily 2310.14690}}].
\bibitem{Mukherjee:2021rri}
J.~Mukherjee, \emph{{Partition functions and entanglement entropy: Weyl graviton and conformal higher spin fields}}, \href{https://doi.org/10.1007/JHEP04(2022)071}{\emph{JHEP} {\bfseries 04} (2022) 071} [\href{https://arxiv.org/abs/2112.15461}{{\ttfamily 2112.15461}}].
\bibitem{PhysRevD.30.295}
V.~Ferrari and B.~Mashhoon, \emph{New approach to the quasinormal modes of a black hole}, \href{https://doi.org/10.1103/PhysRevD.30.295}{\emph{Phys. Rev. D} {\bfseries 30} (1984) 295}.
\end{thebibliography}\endgroup