Beyond quasinormal modes: a complete mode decomposition of black hole perturbations

arXiv:2510.18956 · gr-qc, hep-th · Submitted 2025-10-21 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: General Relativity and Quantum Cosmology. The day in general relativity and quantum cosmology.

Kai: We're looking at the paper: "Beyond quasinormal modes".

Mira: Here's the gist of the paper. Beyond quasinormal modes: a complete mode decomposition of black hole perturbations shows that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime.

Kai: Core Finding on Mode Decomposition The central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum.

Mira: A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times.

Kai: The authors resolve the issue where QNM sums diverge in certain spacetime regions by showing that an additional set of modes which appear at Matsubara frequencies associated to the horizon temperature, which they call Matsubara modes (MMs), provide a convergent sum in regions where QNMs do not.

Paper summary: Kai: So to recap, the paper "Beyond quasinormal modes: a complete mode decomposition of black hole perturbations" shows that the retarded Green’s functions for black hole spacetimes can be written as a convergent mode sum everywhere in spacetime.

Mira: The main thesis is that they partition spacetime into distinct regions by lightrays bouncing off the black hole potential, and in each region, you use a different convergent mode sum.

Lev: They address the problem where QNM sums diverge in certain regions by introducing Matsubara modes at early times to provide a convergent representation there >

Kai: The methodology involves splitting the frequency space Green’s function into two terms, Ge plus and Ge minus, and showing that different contour closures lead to different convergent residue sums >

Mira: They define three distinct regions: Region I where only QNMs converge, Region II where only Matsubara modes converge, and Region III where the Green’s function is zero due to causality >

Lev: The paper shows the convergence of these mode sums analytically by studying the asymptotics of residue sums and also provides numerical demonstrations across several black hole spacetimes like Pöschl-Teller, BTZ, and Schwarzschild-de Sitter >

Kai: For Schwarzschild specifically, they express the branch cut contribution as a convergent sum of de Sitter quasinormal modes when the cosmological constant Lambda goes to zero from above >

Mira: They also use recent exact solutions to the Heun connection problem to handle those details in that case >

Lev: And for BTZ in asymptotically AdS three dimensional spacetime, they find that Region II doesn't intersect the boundary except at a lightcone singularity, limiting the holographic response there >

Kai: So what this means for us is that we have a complete picture of how to represent these fundamental quantities across all of spacetime >

Mira: It gives us a framework where we can use QNMs for late times and Matsubara modes for early times, which should help when fitting generic initial data >

Lev: The paper's limitation is that they focus on stationary spacetimes, and the future work suggests investigating more general stationary spacetimes like Kerr >

Conclusion: Kai: So, to wrap up this discussion on "Beyond quasinormal modes: a complete mode decomposition of black hole perturbations," the core finding is that we have a convergent mode expansion for black hole Green’s functions in all regions, using both QNMs and Matsubara modes.

Mira: It means the convergence isn't just a late-time thing anymore; it's everywhere, controlled by those lightrays that separate the regions >

Lev: The paper suggests that this decomposition allows you to determine how a mode decomposes by convolving it with initial data, which shows an overlap region where both QNMs and Matsubara modes are involved >

Kai: And the analysis also showed that in some cases, one expansion can be obtained as the analytic continuation of the other >

Mira: This gives us a more robust tool for fitting ringdown signals with superpositions of QNMs because we have to include those Matsubara contributions >

Lev: We need to extend those orthogonality relations to include these Matsubara modes too, and also remember that in some cases, you can just use the analytic continuation of the other expansion >

Paolo Arnaudo, Javier Carballo, Benjamin Withers

University of Southampton · STAG Research Centre

gr-qc, hep-th

Submitted: 2025-10-21

Updated: 2026-10-29

Comments: v2: updated to match the published version

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 79/100

The gist: Beyond quasinormal modes: a complete mode decomposition of black hole perturbations shows that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum…

Terminology

Summary

Beyond quasinormal modes: a complete mode decomposition of black hole perturbations shows that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime. p. 0

Where the paper says it

Beyond quasinormal modes: a complete mode decomposition of black hole perturbations Paolo Arnaudo,∗ Javier Carballo,† and Benjamin Withers‡ Mathematical Sciences and STAG Research Centre, University of Southampton, Highfield, Southampton SO17 1BJ, UK (Dated: 2025) We show that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime. At late times a quasinormal mode sum converges, while at early times a Matsubara (or, Euclidean) mode sum converges.

Core Finding on Mode Decomposition

The central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum. A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times. p. 0

Where the paper says it

Our central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum. A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times.

Methodology for Mode Sum Convergence

The authors resolve the issue where QNM sums diverge in certain spacetime regions by showing that an additional set of modes which appear at Matsubara frequencies associated to the horizon temperature, which they call Matsubara modes (MMs), provide a convergent sum in regions where QNMs do not. The common steps to reach these conclusions involve splitting the frequency space Green’s function Ge into two terms using connection formulae, Ge+ and Ge−. p. 1

Where the paper says it

our findings using retarded Green’s function calculations for several black hole spacetimes. The common steps to reach these conclusions are as follows, and are also illustrated in FIG.

Spacetime Region Partitioning

The spacetime is partitioned into three distinct regions (I, II, and III) by lightrays emanating from the insertion point and reflecting from the black hole potential.

  1. Region I is where both terms must be closed in the LHP, leading to a convergent sum of QNMs only. p. 2

    Where the paper says it

    In region I) both terms must be closed in the LHP, leading to a cancellation of the MM contributions. Thus G(t, t′ , r, r′ ) is given by a convergent sum of QNMs only.

  2. Region II requires Ge+ to be closed in the UHP and Ge− in the LHP, resulting in a convergent sum of MMs only. p. 2

    Where the paper says it

    In region II) Ge+ requires UHP closure and Ge− LHP closure. There are no QNM contributions and G(t, t′ , r, r′ ) is given by a convergent sum of MMs only.

  3. Region III requires both terms to be closed in the UHP, leading to G = 0 due to causality. p. 2

    Where the paper says it

    There are no QNM contributions and G(t, t′ , r, r′ ) is given by a convergent sum of MMs only. In region III) both terms must be closed in the UHP, the residue contributions cancel leading to a zero as required by causality.

Analytic Structure of Mode Sums

The convergence of these mode sums is demonstrated analytically using ratio tests and the asymptotic form of residues of a two-term split of the frequency-space Green’s function. For instance, in Region I, the QNM sum converges absolutely because LQNM < 1 in that region. p. 5

Where the paper says it

(30) Recall that we took r ′ ∗ > 0 without loss of generality, and r∗ can take any sign in this region. The definition (19) is equivalent to t − t ′ − r ′ ∗ > |r∗| and thus LQNM < 1 in region I where the QNM sum converges absolutely.

In Region II, the Matsubara mode sum converges because L+MM and L-MM are both less than 1. p. 0
Where the paper says it

Beyond quasinormal modes: a complete mode decomposition of black hole perturbations Paolo Arnaudo,∗ Javier Carballo,† and Benjamin Withers‡ Mathematical Sciences and STAG Research Centre, University of Southampton, Highfield, Southampton SO17 1BJ, UK (Dated: 2025) We show that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime. At late times a quasinormal mode sum converges, while at early times a Matsubara (or, Euclidean) mode sum converges.

Application to Specific Spacetimes

The decomposition is illustrated for several black hole spacetimes, including Pöschl-Teller, BTZ, and Schwarzschild(-de Sitter). For Schwarzschild, the branch cut contribution is expressed as a convergent sum of de Sitter quasinormal modes as Λ → 0+. p. 0

Where the paper says it

We illustrate our results for P¨oschl-Teller, BTZ, and Schwarzschild. In the case of Schwarzschild, we express the branch cut contribution as a convergent sum of de Sitter quasinormal modes as Λ → 0 +, and exploit recent exact solutions to the Heun connection problem.

For BTZ in asymptotically AdS3 spacetime, region II does not intersect the boundary except at the lightcone singularity, restricting holographic response to region I. p. 7
Where the paper says it

In this section, we turn our attention to linear perturbations on the (non-rotating) BTZ black hole. We choose to study BTZ since, given that it is an asymptotically AdS3 spacetime, it allows us to demonstrate the universality of our results across different asymptotics, discussing the slightly different details that arise in AdS where needed.

Implications for Physical Signals

The inclusion of MM contributions is necessary in order to achieve a consistent fit for generic initial data when fitting ringdown signals with a superposition of QNMs. This suggests the need to extend recent QNM orthogonality relations [30–34] to include MMs too. p. 3

Where the paper says it

The study of QNMs is a rich and varied active research subject and we defer to reviews for a more complete treatment [25–29]. In particular, it is worth highlighting that despite a lack of a spectral theorem, there are known orthogonality relations among QNMs [30–34].

The analysis also implies that the representation of G in region II as a convergent sum of MMs can be obtained from an analytic continuation of the Euclidean Green’s function. p. 0
Where the paper says it

Our central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum. A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times.

Further Research Directions

Future work could investigate the Green’s function in more general stationary spacetimes such as Kerr. It is also suggested to explicitly construct an example of a QNM completeness counterexample using Matsubara mode contributions. p. 1

Where the paper says it

It is also well-known that QNMs do not provide a complete set of modes everywhere in spacetime, even in the absence of branch cuts. This is sharply illustrated by the ‘QNM completeness counterexample’ [11].

The representation of G in region II as a convergent sum of MMs can be obtained from an analytic continuation of the Euclidean Green’s function. p. 0
Where the paper says it

Our central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum. A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times.

The Regge poles decomposition of the Euclidean correlator has access to the decomposition in different regions we studied. p. 4
Where the paper says it

Here the Matsubara modes appear in a systematic treatment of the spectral decomposition of the black hole Green’s function, and as far as we are aware this is the first such treatment. We also note that the appearance of different regions has also arisen, [9, 19, 24, 41–44].

Conclusion on Convergence

The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and a new ingredient, MMs. p. 3

Where the paper says it

In region III, both Ge+ and Ge− are closed in the UHP, the MM residues cancel, and one is left with the usual result that G vanishes due to analyticity in the UHP. The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and a new ingredient, MMs.

The convergence of each geometric series is set by lightrays that border the region of interest. p. 5
Where the paper says it

The definition (19) is equivalent to t − t ′ − r ′ ∗ > |r∗| and thus LQNM < 1 in region I where the QNM sum converges absolutely. On the boundary of the region where L = 1 the ratio test is inconclusive, however, a numerical investigation of the coefficients indicates that the terms in the series are alternating and monotonically decreasing, implying convergence via the alternating series test.

The spacetime dependence appears only in the form of e t-t'+r+r' with a logarithmic branch point singularity at e t-t'+r+r'=1, which is responsible for the radius of convergence of the regional expansions.

Initial Data Dependence

The mode decomposition allows one to determine the mode decomposition of f by convolving it with initial data. This shows that there is an overlap region where the Green’s function is expressed as a convergent sum over both QNMs and MMs. p. 6

Where the paper says it

In region I (blue) the Green’s function is a convergent sum of QNMs, see section II A. In region II (green) it is a convergent sum of MMs, see section II B.

The analysis also shows that in some cases, one expansion can be obtained as the analytic continuation of the other. p. 8
Where the paper says it

Let us now study the analytic structure in the complex ω plane of the Ge+ and Ge− integrands in (2) at finite r∗, 10 At first sight, one may think that the 4−iω factor in (62) could contribute to a constant shift in the exponent t − t ′ − |r∗ ± r ′ ∗ |, hence changing the location of the lightrays delineating the three spacetime regions. However, a careful analysis of the large ω asymptotics for this term including the contributions coming from Cn.

p. 8
Where the paper says it

Let us now study the analytic structure in the complex ω plane of the Ge+ and Ge− integrands in (2) at finite r∗, 10 At first sight, one may think that the 4−iω factor in (62) could contribute to a constant shift in the exponent t − t ′ − |r∗ ± r ′ ∗ |, hence changing the location of the lightrays delineating the three spacetime regions. However, a careful analysis of the large ω asymptotics for this term including the contributions coming from Cn.

p. 8
Where the paper says it

Let us now study the analytic structure in the complex ω plane of the Ge+ and Ge− integrands in (2) at finite r∗, 10 At first sight, one may think that the 4−iω factor in (62) could contribute to a constant shift in the exponent t − t ′ − |r∗ ± r ′ ∗ |, hence changing the location of the lightrays delineating the three spacetime regions. However, a careful analysis of the large ω asymptotics for this term including the contributions coming from Cn.

Appendix Details

Appendix A discusses the Rindler patch of AdS2, where G can be expressed as a convergent mode sum after the lightray has bounced from the boundary. Appendix D provides technical details for Schwarzschild-de Sitter, including Heun equation parameters and gauge theory quantities. The analysis of Heun poles in the Schwarzschild-de Sitter case is detailed through residue calculations. p. 0

Where the paper says it

In the case of Schwarzschild, we express the branch cut contribution as a convergent sum of de Sitter quasinormal modes as Λ → 0 +, and exploit recent exact solutions to the Heun connection problem. In each case we analytically show convergence by studying the asymptotics of residue sums and also provide numerical demonstrations.

The connection coefficients of the Heun differential equation were obtained from connection formulae for semiclassical Virasoro conformal blocks. p. 4
Where the paper says it

These are normalised such that ψdown(z) ∼ e −iωr∗ , ψup(z) ∼ e iωr∗ as r∗ → ∞, and ψin(z) ∼ e −iωr∗ , ψout(z) ∼ e iωr∗ as r∗ → −∞. As (9) is a second order ODE there are only two independent solutions, and so there are two connection formulae, ψin(z) = Cin,upψup(z) + Cin,downψdown(z), (14) ψout(z) = Cout,upψup(z) + Cout,downψdown(z), (15) with connection coefficients given in (B7)-(B10).

Summary of Results for BTZ

For BTZ, region II did not intersect the boundary except at the lightcone singularity, so holographic response is restricted to region I. In this case, each residue sum is a geometric series with a finite radius of convergence determined by lightrays. p. 6

Where the paper says it

The spacetime dependence appears only in the form of e t−t ′−r∗−r ′ ∗ with a logarithmic branch point singularity at e t−t ′−r∗−r ′ ∗ = −1, which is responsible for the radius of convergence of the regional expansions. 8 Again, we focus on the case of Im ν ̸= 0 for simplicity.

The mode sums are compared to direct numerical integration and show excellent agreement in regions where they converge. p. 6
Where the paper says it

The lines show sums of the first 30 modes in each sum – in the region where they apply (solid), and their extension to the regions where they do not (dashed). Black points show the Green’s function obtained by direct numerical integration for comparison.

Final Convergence Criterion

The residue sum is convergent in the region t < t'+r+r'−2r bounce∗, where r bounce∗ is a constant independent of t, t', r+, r- and corresponds to the radius at which the lightray bounces from the black hole potential. The bounce location is given by r bounce∗ ≡ Rh log Rh/√Λ + Rh γt. p. 8

Where the paper says it

,out = 2 Γ(c)Γ(1 + a + b − c) Γ(a)Γ(b) . (57) The timelike boundary acts as a reflective barrier where lightrays bounce off, which gives rise to the region subdivision of spacetime shown in FIG.

The QNM sum converges absolutely in region I where t−t' > r∗ and thus LQNM 1 in region I, implying a divergent MM sum. The convergence criterion is shown to be t < t'+r+r'−2r bounce∗.

Initial Data Overlap

The analysis of initial data shows that each delta function comes with its own region I and region II, and these then overlap with each other for multiple sources. This results in an overlap region where the Green’s function is expressed as a convergent sum over both QNMs and MMs. p. 6

Where the paper says it

In region I (blue) the Green’s function is a convergent sum of QNMs, see section II A. In region II (green) it is a convergent sum of MMs, see section II B.

The analysis also shows that in some cases, one expansion can be obtained as the analytic continuation of the other.

Final Summary of Results

Green’s function, Ge. We also showed in some cases that one expansion can be obtained as the analytic continuation of the other. p. 6

Where the paper says it

the LHP arc for I−. Finally, we note that in both region I and region II, the mode sum can be performed exactly at large r∗, r′ ∗ , which gives the same closed form expression upon analytic continuation, G(t, t′ , z, z′ ) ∼ 1 2 2F1 1 2 − ν, 1 2 + ν, 1, −e t−t ′−r∗−r ′ ∗ , (36) as r∗, r′ ∗ → ∞.

In some black holes, the frequency space expression Ge contains branch point singularities. p. 1
Where the paper says it

Computation of a retarded black hole Green’s function G(t, t′ , r, r′ ) by Fourier transforming the frequency-space expression Ge(ω, r, r′ ) using a contour integral. The analytic structure of Ge(ω, r, r′ ) consists of poles corresponding to QNM modes, and, depending on the black hole, branch cut contributions (not shown here).

This occurs in Schwarzschild at ω = 0, for example. p. 0
Where the paper says it

When there are branch cut discontinuities of Ge in the complex ω plane, we choose to resolve them into a set of poles. For example, in the case of Schwarzschild we resolve its cut into a line of de 1 For simplicity we focus on static, spherically symmetric black holes.

Our approach in this work was to resolve the branch point into a set of poles along the negative imaginary axis, by adding a small cosmological constant, Λ > 0. p. 1
Where the paper says it

Our work addresses this issue. Sitter QNM poles by adding a small cosmological constant, Λ > 0 (i.

This provides a mode sum description of the physics of Schwarzschild (in particular the power-law tail). p. 0
Where the paper says it

Our central new result is that black hole spacetimes are partitioned into distinct regions by certain lightrays, and in each region the Green’s function is represented by a different convergent mode sum. A quasinormal mode (QNM) sum provides a convergent representation at late times, while a new ingredient in this context, a Matsubara mode (MM) sum, provides a convergent representation at early times.

At strictly Λ = 0, region I would contain a QNM branch cut discontinuity integral, and region II would contain MM branch cut discontinuity integrals too. p. 3
Where the paper says it

2, region II intersects the right horizon, and the MMs are those with the temperature of that horizon. In the case of asymptotically flat spacetimes with region II intersecting null infinity, the relevant temperature is zero, and the line of MMs forms a cut instead, and the Matsubara contribution takes the form of cut discontinuity integrals.

Resolving these discontinuity integrals with a small Λ > 0 gives the same outcome as Λ → 0+ [3] and is easier to compute. p. 1
Where the paper says it

2 In other words, for all intents and purposes, the contributions coming from poles in this paper may be regarded as also including contributions from cut discontinuity integrals. Such cut resolutions are easier to work with and do not affect our conclusions.

The results of our analysis have also implications when it comes to fitting ringdown signals with a superposition of QNMs. p. 3
Where the paper says it

There is no fitting, this is directly evaluated by summing residues of Ge±. In region I (blue) the Green’s function is a convergent sum of QNMs.

Indeed, as is evident from the discussion in section V, the inclusion of MM contributions is necessary in order to achieve a consistent fit for generic initial data. p. 0
Where the paper says it

General initial data 16 VI. Discussion 16 Acknowledgments 18 I.

This also suggests the need to extend recent QNM orthogonality relations [30–34] to include MMs too. p. 3
Where the paper says it

The study of QNMs is a rich and varied active research subject and we defer to reviews for a more complete treatment [25–29]. In particular, it is worth highlighting that despite a lack of a spectral theorem, there are known orthogonality relations among QNMs [30–34].

The analysis also shows that in some cases, one expansion can be obtained as the analytic continuation of the other. The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and a new ingredient, MMs.

Acknowledgments

It is a pleasure to thank Nils Andersson, Robin Karlsson, Christiana Pantelidou and Kostas Skenderis for useful discussions. PA is supported by the Royal Society grant URF231002, ‘Dynamics of holographic field theories’. p. 7

Where the paper says it

In particular, we restrict our attention to cases where ∆ is the largest root of this equation, i. The dynamics of Φ is governed by the massive Klein-Gordon equation on this background, (□ − m2 Φ)Φ = 0.

JC is supported by the Royal Society grant RFERE210267. BW is supported by a Royal Society University Research Fellowship and in part by the STFC consolidated grant ‘New Frontiers In Particle Physics, Cosmology And Gravity’. p. 0
Where the paper says it

Beyond quasinormal modes: a complete mode decomposition of black hole perturbations Paolo Arnaudo,∗ Javier Carballo,† and Benjamin Withers‡ Mathematical Sciences and STAG Research Centre, University of Southampton, Highfield, Southampton SO17 1BJ, UK (Dated: 2025) We show that retarded Green’s functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime. At late times a quasinormal mode sum converges, while at early times a Matsubara (or, Euclidean) mode sum converges.

Appendix A Details

In Appendix A, we revisit this illustrative example, and also analyse the analogue of region II in this case. The Rindler patch of AdS2 is given by ds squared = -sinh 2(rho dt squared + dρ 2). p. 4

Where the paper says it

We also note that the appearance of different regions has also arisen, [9, 19, 24, 41–44]. Perhaps the closest existing work in the literature is [9] in which different requirements for closing contour integrals in different regions of spacetime is mentioned, though not pursued, as well as [24] in which the QNM region for scalar correlators in the Rindler patch of AdS2 was identified (this example is discussed in Appendix A).

The massless scalar equation is given by -∂2t + ∂2r∗ Φ(t, r) = 0. p. 1
Where the paper says it

In this work we resolve these questions by showing that an additional set of modes which appear at Matsubara frequencies associated to the horizon temperature, which we call Matsubara modes (MMs), provide a convergent sum in regions where QNMs do not. We will illustrate 2 This is analogous to the finite temperature resolution of the cut for charged scalar correlators in RNAdS5 [4] (both cases are a confluence limit of the Heun equation).

The homogeneous solutions that are normalisable in the right and left wedge respectively are given by phiRn. p. 1
Where the paper says it

The solution to (1) in frequency space is given by standard techniques as follows Ge(ω, r, r′ ) = 1 W × ( ϕin(r) ϕup(r ′ ), r < r′ ϕin(r ′ ) ϕup(r), r > r′ , (3) where the Wronskian W = ϕup∂r∗ ϕin − ϕin∂r∗ ϕup. The fields ϕin and ϕup are the homogeneous solutions ingoing through H+ and outgoing through H+ c /I +.

(r) = e-iωr∗ − e iωr∗. p. 4
Where the paper says it

The homogeneous form of (1) in frequency space is given by ∂ 2 r∗ + ω 2 − V ψ(z) = 0, (9) whose solutions are associated Legendre functions, ψdown(z) = cd P −iω ν− 1 2 (z), (10) ψup(z) = cu P iω ν− 1 2 (z), (11) ψin(z) = c + i P iω ν− 1 2 (z) + c − i P −iω ν− 1 2 (z), (12) ψout(z) = c + o P iω ν− 1 2 (z) + c − o P −iω ν− 1 2 (z), (13) where the coefficients are functions of ω found in (B1)- (B6). These are normalised such that ψdown(z) ∼ e −iωr∗ , ψup(z) ∼ e iωr∗ as r∗ → ∞, and ψin(z) ∼ e −iωr∗ , ψout(z) ∼ e iωr∗ as r∗ → −∞.

The Wronskian is W = -4iω sinh(πω). p. 5
Where the paper says it

This split of terms and the associated regions also apply to finite r∗, r′ ∗ , which we verify a posteriori using convergence tests below. Let us turn to analytic structure for the integrand of (2) separated into the two terms corresponding to the Ge+ integral and Ge− integral respectively, I+ = − i 4π csch(πω) cos(πν)Γ 1 2 − ν − iω (22) × Γ 1 2 + ν − iω P iω ν− 1 2 (z<)P iω ν− 1 2 (z>)e −iω(t−t ′ ) , I− = i 4 csch(πω)P −iω ν− 1 2 (z<)P iω ν− 1 2 (z>)e −iω(t−t ′ ) .

The position space result G(t, t', r, r') can be computed as a convergent mode sum following the techniques outlined in the introduction.

Appendix B Details

In Appendix B, we provide additional technical details for section II. The coefficients appearing in the homogeneous solutions (10)-(13) are given. p. 4

Where the paper says it

It is convenient to work with a new coordinate z = r = tanh r∗. The homogeneous form of (1) in frequency space is given by ∂ 2 r∗ + ω 2 − V ψ(z) = 0, (9) whose solutions are associated Legendre functions, ψdown(z) = cd P −iω ν− 1 2 (z), (10) ψup(z) = cu P iω ν− 1 2 (z), (11) ψin(z) = c + i P iω ν− 1 2 (z) + c − i P −iω ν− 1 2 (z), (12) ψout(z) = c + o P iω ν− 1 2 (z) + c − o P −iω ν− 1 2 (z), (13) where the coefficients are functions of ω found in (B1)- (B6).

The connection coefficients appearing in (14), (15) are given. p. 4
Where the paper says it

These are normalised such that ψdown(z) ∼ e −iωr∗ , ψup(z) ∼ e iωr∗ as r∗ → ∞, and ψin(z) ∼ e −iωr∗ , ψout(z) ∼ e iωr∗ as r∗ → −∞. As (9) is a second order ODE there are only two independent solutions, and so there are two connection formulae, ψin(z) = Cin,upψup(z) + Cin,downψdown(z), (14) ψout(z) = Cout,upψup(z) + Cout,downψdown(z), (15) with connection coefficients given in (B7)-(B10).

In the large r∗, r′∗ asymptotics of each term, Ge±(ω, z, z′) ∼ α±e iωr∗±r' as r∗, r'→∞. The calculation proceeds essentially according to Jordan’s lemma but adapted to the analytic structure of α±. p. 9
Where the paper says it

Knowing the analytic structure of the integrands, it is now straightforward to construct the retarded G in each region as a convergent sum of the corresponding modes using the residue theorem as sketched in FIG. 2, adapted to the AdS regions in FIG.

The result is that when κ > 0 we close in the UHP to obtain IC+ ≤ πMCκ/ (1 − e-κR). p. 3
Where the paper says it

Thus, G is expressed as a convergent sum over only MMs in this region. In region III, both Ge+ and Ge− are closed in the UHP, the MM residues cancel, and one is left with the usual result that G vanishes due to analyticity in the UHP.

Appendix C Details

In Appendix C, we provide additional technical details for section II. The connection coefficients appearing in (54), (55) are given. p. 4

Where the paper says it

These are normalised such that ψdown(z) ∼ e −iωr∗ , ψup(z) ∼ e iωr∗ as r∗ → ∞, and ψin(z) ∼ e −iωr∗ , ψout(z) ∼ e iωr∗ as r∗ → −∞. As (9) is a second order ODE there are only two independent solutions, and so there are two connection formulae, ψin(z) = Cin,upψup(z) + Cin,downψdown(z), (14) ψout(z) = Cout,upψup(z) + Cout,downψdown(z), (15) with connection coefficients given in (B7)-(B10).

The closed form expression at r∗, r′→−∞ for the mode sum in region I reads G(t, t', z, z') ∼ π csch(πm) m Γ(∆) 2-2im-2∆ e-(t-t'+r+r')(im+∆) × "2/4im e 2im(t-t'+r+r')Γ(-im) Γ(1+im−∆) 4F3 - im squared + ∆ squared, - im squared + ∆ squared, 1/2 − im squared + ∆ squared, 1/2 + im squared + ∆ squared; 1+im, ∆, im+∆; e-2(t-t'+r+r'). The closed form expression at r∗, r'→−∞ for the mode sum in region II reads G(t, t', z, z') ∼ 1/2 "4F3 1 - im squared - ∆ squared, 1 + im squared - ∆ squared, -im squared + ∆ squared, im squared + ∆ squared; 1/2, 1/4; e(t-t'+r+r')! −e(t-t'+r+r')m s2 + (-1+∆) squared 4F3 3/2 - im squared - ∆ squared, 3/2 + im squared - ∆ squared, 1/2 - im squared + ∆ squared, 1/4; e(t-t'+r+r')! The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and MMs. p. 3
Where the paper says it

In region III, both Ge+ and Ge− are closed in the UHP, the MM residues cancel, and one is left with the usual result that G vanishes due to analyticity in the UHP. The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and a new ingredient, MMs.

Appendix D Details

Appendix D provides technical details for Schwarzschild-de Sitter, Heun equation, and gauge theory. It defines the tortoise coordinate r∗ as a complex expression. p. 0

Where the paper says it

The retarded Green’s function G(t, t′ , r, r′ ) is the solution to the following PDE, −∂ 2 t + ∂ 2 r∗ − V (r) G(t, t′ , r, r′ ) = −δ(t − t ′ )δ(r∗ − r ′ ∗ ), (1) which obeys causal response, vanishing for t < t′ . Here, r∗ is a tortoise coordinate, we have decomposed the perturbation into transverse spherical harmonics1 , and V is the potential whose details depend on the spacetime, species of perturbation, and the orbital quantum number, ℓ.

It defines the local solution satisfying ingoing boundary condition at the cosmological horizon as ψdown(z) = z γ/2 (1 − z) δ/2 (z − x) ϵ/2 Heun(x, q, α, β, γ, δ, z). It defines the connection coefficients Cin down and Cin up using Gamma functions. p. 4
Where the paper says it

These are normalised such that ψdown(z) ∼ e −iωr∗ , ψup(z) ∼ e iωr∗ as r∗ → ∞, and ψin(z) ∼ e −iωr∗ , ψout(z) ∼ e iωr∗ as r∗ → −∞. As (9) is a second order ODE there are only two independent solutions, and so there are two connection formulae, ψin(z) = Cin,upψup(z) + Cin,downψdown(z), (14) ψout(z) = Cout,upψup(z) + Cout,downψdown(z), (15) with connection coefficients given in (B7)-(B10).

The instanton part of the NS free energy is given as a power series in x by F(x) = lim epsilon2→0 epsilon2 log"(1 − x)-2epsilon-1/2 (1/2 +a1)(1/2 +ax)X Y⃗ xY⃗ zvec a, Yvec Y 4 i=1 zhyp a, Y, m vec i. The instanton expansion of a reads a = ± (r-1/4-u+a 2x+a 20+1/2+u-a 2x-a 20-a 21+a 2inf 1/2+u- 2a ax - 1 + x(2a1 - 1) + (1 - x) u. p. 3
Where the paper says it

In region III, both Ge+ and Ge− are closed in the UHP, the MM residues cancel, and one is left with the usual result that G vanishes due to analyticity in the UHP. The final result is thus a convergent mode expansion for black hole Green’s function in all regions, using QNMs and a new ingredient, MMs.

The parameter u appearing in the differential equation (93) is the complex moduli parametrizing the corresponding Seiberg-Witten curve. p. 7
Where the paper says it

Without loss of generality, we consider the radial problem for Φ by performing the following decomposition, Φ(t, r, φ) = X∞ m=−∞ Z dω e−iωt 1 √ r ϕ(r)e imφ , (40) where we have introduced a factor of 1/ √ r such that ϕ obeys the homogeneous form of (1) in frequency space, ∂ 2 r∗ + ω 2 − V ϕ(r) = 0, (41) with effective potential V (r) = ∆(∆ − 2)f + m2f r 2 − f 2 4r 2 + ff′ 2r , (42) and where r∗ corresponds to the standard tortoise radial coordinate r∗ = Z dr f(r) (43) = 1 2 log r − 1 r + 1 , (44) where we have fixed the integration constant such that r∗(r → +∞) = 0. In order to find the solutions to (41), it is convenient to change coordinates to z = (r 2 − 1)/r2 – where now z = 0 corresponds to the black hole horizon and z = 1 to the conformal boundary – and make the following field redefinition ϕ(r(z)) = z −iω/2 (1 − z) ∆/2−1/4ψ(z), (45) under which (41) takes the form of the hypergeometric differential equation [49] z(1 − z) d 2ψ dz2 + (c − (a + b + 1)z) dψ dz − ab ψ = 0, (46) with parameters a = − im 2 + ∆ 2 − iω 2 , (47) b = im 2 + ∆ 2 − iω 2 , (48) c = 1 − iω.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved ability to predict late-time black hole ringdown signals by incorporating Matsubara mode sums (MMs). The improved system can now distinguish between QNM convergence at late times and MM convergence at early times, leading to more accurate fits for generic initial data, as noted in Section VI: the inclusion of MM contributions is necessary in order to achieve a consistent fit for generic initial data.

  2. Enhanced analysis of spacetime geometry by resolving branch cut discontinuities via small cosmological constants. The system can now provide a mode sum description of Schwarzschild physics even at strictly zero cosmological constant by using a small cosmological constant, Lambda greater than zero, providing a mode sum description of the physics of Schwarzschild (in particular the power-law tail).

  3. More precise determination of spacetime regions for perturbation analysis. The AI can accurately partition spacetime into distinct regions I, II, and III based on lightrays emanating from the source and reflecting off potential barriers. This allows for region specific mode expansions, such as a convergent sum of QNMs only in Region I or a convergent sum of MMs only in Region II.

  4. Superior handling of asymptotic limits in asymptotically AdS spacetimes. The system can utilize the Heun equation framework and its connection coefficients to solve for Green's functions on BTZ black holes by mapping coordinates to the Rindler patch and conformal boundary, providing a complete mode decomposition that respects AdS boundary conditions.

  5. Better modeling of holographic thermal correlators in AdS/CFT contexts. The improved system can use the analytic continuation of Euclidean Green's functions, expressed as a Fourier sum of retarded Green's functions evaluated at Matsubara frequencies, to derive position-space correlators via a sum over the Regge poles decomposition.

Abstract

We show that retarded Green's functions of black hole spacetimes can be expressed as a convergent mode sum everywhere in spacetime. At late times a quasinormal mode sum converges, while at early times a Matsubara (or, Euclidean) mode sum converges. The two regions are separated by a lightcone which scatters from the black hole potential. The Matsubara sum is a Fourier series on the Euclidean thermal circle associated to the early time region. We illustrate our results for Pöschl-Teller, BTZ, and Schwarzschild. In the case of Schwarzschild, we express the branch cut contribution as a convergent sum of de Sitter quasinormal modes as Λ to 0+, and exploit recent exact solutions to the Heun connection problem. In each case we analytically show convergence by studying the asymptotics of residue sums and also provide numerical demonstrations.

Sources

Related papers