Introduction
In holography [1 , 2 , 3 ] , the region behind the horizon of a black hole, including the singularity, can be described by boundary correlators continued analytically from Euclidean signature [4 ] . In the eternal AdS-Schwarzschild black hole, a null geodesic can leave one boundary, bounce off the singularity, and reach the other boundary. Correspondingly, the analytically continued thermal two-point function G ( t ) G(t) G ( t ) is singular at the complex time of this bouncing geodesic [5 , 6 , 7 , 8 ] . These bouncing singularities appear at leading order in 1 / N 1/N 1/ N , and they are smoothed out at finite N N N [9 , 10 , 11 , 12 ] or at finite ’t Hooft coupling [13 ] .
Recently, the bouncing geodesic has been identified in three kinds of boundary data. The first is the OPE of the thermal two-point function. Its stress-tensor sector, which contains the identity, the stress tensor and its products, can be computed from a near-boundary expansion of the bulk wave equation [14 ] , and the remaining double-trace part can then be fixed by the KMS condition [15 , 16 ] . The stress-tensor sector is singular at the complex time of the bouncing geodesic [17 ] . The second is the spectral function at large real frequency, which receives exponentially small corrections controlled by the bouncing geodesic [18 , 19 , 20 , 21 ] . The third is the QNM spectrum. The QNMs at large overtone number are sensitive to the geometry near the singularity, as shown for Kasner interiors [22 , 23 ] , quantum black holes in three dimensions [24 , 25 ] , and Einstein–Maxwell–dilaton black holes [26 , 27 , 28 ] , and their asymptotic spectrum is related to the bouncing singularities [12 , 29 , 30 , 31 ] . Related information is also carried by the bulk phase shift [32 ] .
When the singularity is spacelike, the geometry near it often takes the Kasner form [33 , 34 ] , as in black branes with scalar hair [35 , 36 , 37 ] . The Kasner exponents can be read off from the boundary data [22 , 23 , 35 , 38 , 39 ] . For the Reissner–Nordström black brane, the singularity is timelike and lies beyond the inner horizon, and the OPE [40 ] and the spectral function [41 ] still detect a null geodesic bouncing off it.
In the AdS-Schwarzschild and Reissner–Nordström black branes, the metric has a single singular point, which is the physical singularity in the spacetime, and all the signatures above come from it. In this paper, we consider black branes whose metric has several singular points where the curvature diverges. They are the black branes of the STU model [42 , 43 ] , which carry three R-charges of N = 4 \mathcal{N}=4 N = 4 SYM, with parameters q 1 ≥ q 2 ≥ q 3 ≥ 0 q_1 \ge q_2 \ge q_3 \ge0 q 1 ≥ q 2 ≥ q 3 ≥ 0 . The metric has curvature singular points at R = r 2 = − q i R=r^2=-q_i R = r 2 = − q i . For equal R-charges, they merge into one, and the solution is the Reissner–Nordström black brane. For unequal R-charges, there are three of them, but the spacetime ends at the first one, R = − q 3 R=-q_3 R = − q 3 , which is the physical singularity. The other two lie beyond it and are not part of the spacetime. Nevertheless, the bulk wave equation is regular there, so its solutions can be continued to them. Each singular point has an optical time, the complex time of a null ray from the boundary to the point and back. The three optical times T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 form a right triangle in the complex plane, as shown in Figure 1 . This naturally raises the question of which singular point each observable detects.
Figure 1. Optical times in the complex T T T -plane. The black lines are the image of the real R R R axis. The OPE converges inside the disk of radius ∣ T 1 ∣ |T_1| ∣ T 1 ∣ . The red line is the direction of T 2 T_2 T 2 , and a line of QNMs lies along 1 / T 2 1/T_2 1/ T 2 . The dotted lines are the Stokes lines of T 1 T_1 T 1 and T 2 T_2 T 2 when ω \omega ω is real (§ 3.3 ), which control the spectral function at large ω \omega ω . The insets zoom in on the right triangle formed by T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 .
To answer this question, we use three observables, all built from the retarded correlator of T x y T_{xy} T x y at zero spatial momentum: the spectral function, the OPE coefficients, and the QNMs. Unlike a probe operator with an arbitrarily chosen dimension, the stress tensor is present in any CFT, and the bulk dual of T x y T_{xy} T x y obeys the equation of a minimally coupled massless scalar [44 , 45 , 46 ] . We derive the signatures of the singular points from a WKB analysis at large frequency, in which a wave sent in from the boundary is reflected very slightly by each singular point, as in [18 , 41 ] , and we check them numerically.
We find that the singular points that contribute most are different for each observable. The exponentially small correction to the spectral function at large real frequency falls like e − ω Im T i e^{-\omega\operatorname{Im} T_i} e − ω Im T i , so it is dominated by the points with the smallest Im T i \operatorname{Im} T_i Im T i . These are T 1 T_1 T 1 and T 2 T_2 T 2 , which have the same imaginary part, so their two terms have the same size and beat. The radius of convergence of the OPE is the distance from t = 0 t=0 t = 0 to the nearest singularity of G R ( t ) G_R(t) G R ( t ) , which is ∣ T 1 ∣ |T_1| ∣ T 1 ∣ , the smallest ∣ T i ∣ |T_i| ∣ T i ∣ . We confirm this from the OPE coefficients alone, without using the optical times. The QNMs appear where e i ω T i e^{i\omega T_i} e iω T i first becomes of order one as the phase of ω \omega ω is rotated, so the complex QNMs lie along the direction of 1 / T 2 1/T_2 1/ T 2 , where T 2 T_2 T 2 has the smallest arg T i \arg T_i arg T i . In all three cases, the selected points lie beyond the physical singularity. The physical singularity T 3 T_3 T 3 enters the spectral function only through an additional factor on the T 2 T_2 T 2 reflection term, and it sets neither the OPE radius nor the direction of the complex QNMs. It appears at leading order only in the QNMs on the negative imaginary axis, where the terms of T 2 T_2 T 2 and T 3 T_3 T 3 have the same size and interfere. For q 3 = 0 q_3=0 q 3 = 0 , where the physical singularity is a Kasner singularity, these QNMs come in pairs, and T 1 T_1 T 1 governs their exponentially small splitting. The three observables therefore hear different singular points.
This paper is organized as follows. In § 2 , we review the STU black branes and find the optical times of the three curvature singular points. In § 3 , we introduce the T x y T_{xy} T x y correlator and the three observables, and derive the reflections from the singular points. We then study the spectral function in § 4 , the OPE in § 5 , and the QNMs in § 6 . We conclude with remarks and future directions in § 7 . Appendix § 9 explains how we compute A ( ω ) A(\omega) A ( ω ) and the QNMs, and Appendix § 10 explains how we compute the OPE coefficients.
STU black branes
The STU model is the U ( 1 ) 3 U(1)^3 U ( 1 ) 3 consistent truncation of five-dimensional N = 8 \mathcal{N}=8 N = 8 S O ( 6 ) SO(6) S O ( 6 ) gauged supergravity, which comes from the S 5 S^5 S 5 reduction of type IIB supergravity [43 ] . Its three U ( 1 ) U(1) U ( 1 ) charges correspond to the three R-charges U ( 1 ) 3 ⊂ S O ( 6 ) R U(1)^3 \subset SO(6)_R U ( 1 ) 3 ⊂ S O ( 6 ) R of N = 4 \mathcal{N}=4 N = 4 SYM.
Solution
The planar black brane with three U ( 1 ) U(1) U ( 1 ) charges [42 ] is usually written as
d s 2 = − ( H 1 H 2 H 3 ) − 2 / 3 f d t 2 + ( H 1 H 2 H 3 ) 1 / 3 ( d r 2 f + r 2 d x ⃗ 2 ) , (1) ds^2 = -(H_1H_2H_3)^{-2/3}f\,dt^2 + (H_1H_2H_3)^{1/3}\left(\frac{dr^2}{f}+r^2d\vec{x}^{\,2}\right), \tag*{(1)} d s 2 = − ( H 1 H 2 H 3 ) − 2/3 f d t 2 + ( H 1 H 2 H 3 ) 1/3 ( f d r 2 + r 2 d x 2 ) , (1)
with
H i = 1 + q i r 2 , f = r 2 H 1 H 2 H 3 − M r 2 , q i ≥ 0. (2) H_i=1+\frac{q_i}{r^2},\qquad f=r^2H_1H_2H_3-\frac{M}{r^2},\qquad q_i\geq0. \tag*{(2)} H i = 1 + r 2 q i , f = r 2 H 1 H 2 H 3 − r 2 M , q i ≥ 0. (2)
M M M is a mass parameter, and x ⃗ \vec{x} x denotes the three spatial coordinates of the boundary. Without loss of generality, we take
q 1 ≥ q 2 ≥ q 3 ≥ 0. (3) q_1\geq q_2\geq q_3\geq0. \tag*{(3)} q 1 ≥ q 2 ≥ q 3 ≥ 0. (3)
The solution also has three gauge fields and two scalars, but the correlator we study below sees only the metric. The metric depends on r r r only through R = r 2 R=r^2 R = r 2 . With
P ( R ) = ∏ i ( R + q i ) , D ( R ) = P ( R ) − M R , (4) P(R)=\prod_i(R+q_i),\qquad D(R)=P(R)-MR, \tag*{(4)} P ( R ) = i ∏ ( R + q i ) , D ( R ) = P ( R ) − M R , (4)
we have H 1 H 2 H 3 = P / R 3 H_1H_2H_3=P/R^3 H 1 H 2 H 3 = P / R 3 and f = D / R 2 f=D/R^2 f = D / R 2 , so the metric becomes
d s 2 = − D P 2 / 3 d t 2 + P 1 / 3 4 D d R 2 + P 1 / 3 d x ⃗ 2 . (5) ds^2=-\frac{D}{P^{2/3}}dt^2+\frac{P^{1/3}}{4D}dR^2+P^{1/3}d\vec{x}^{\,2}. \tag*{(5)} d s 2 = − P 2/3 D d t 2 + 4 D P 1/3 d R 2 + P 1/3 d x 2 . (5)
From now on, we use R R R instead of r r r . For q 3 > 0 q_3>0 q 3 > 0 , the metric is regular at R = 0 R=0 R = 0 , so the spacetime continues to R < 0 R<0 R < 0 , where r r r is imaginary, and ends at the singularity R = − q 3 R=-q_3 R = − q 3 . The boundary is at R = ∞ R=\infty R = ∞ . We use the scaling symmetry of the planar solution to place the outer horizon at R = 1 R=1 R = 1 , so that D ( 1 ) = 0 D(1)=0 D ( 1 ) = 0 gives M = P ( 1 ) M=P(1) M = P ( 1 ) . The inverse temperature is given by
β = 2 π P ( 1 ) D ′ ( 1 ) . (6) \beta=\frac{2\pi\sqrt{P(1)}}{D'(1)}. \tag*{(6)} β = D ′ ( 1 ) 2 π P ( 1 ) . (6)
The q i q_i q i are parameters of the metric, and the physical R-charge densities are proportional to M q i \sqrt{Mq_i} M q i . Equal or unequal q i q_i q i therefore mean equal or unequal R-charges.
For q 1 = q 2 = q 3 = q q_1=q_2=q_3=q q 1 = q 2 = q 3 = q , in terms of r ~ 2 = R + q \tilde r^2=R+q r ~ 2 = R + q , we have
P = r ~ 6 , D = r ~ 6 − M ( r ~ 2 − q ) . (7) P=\tilde r^6,\qquad D=\tilde r^6-M(\tilde r^2-q). \tag*{(7)} P = r ~ 6 , D = r ~ 6 − M ( r ~ 2 − q ) . (7)
Therefore, (5) becomes the Reissner–Nordström black brane,
d s 2 = − f ~ d t 2 + d r ~ 2 f ~ + r ~ 2 d x ⃗ 2 , f ~ = r ~ 2 − M r ~ 2 + M q r ~ 4 , (8) ds^2=-\tilde f\,dt^2+\frac{d\tilde r^2}{\tilde f}+\tilde r^2d\vec{x}^{\,2},\qquad\tilde f=\tilde r^2-\frac{M}{\tilde r^2}+\frac{Mq}{\tilde r^4}, \tag*{(8)} d s 2 = − f ~ d t 2 + f ~ d r ~ 2 + r ~ 2 d x 2 , f ~ = r ~ 2 − r ~ 2 M + r ~ 4 M q , (8)
with the timelike singularity at r ~ = 0 \tilde r=0 r ~ = 0 (R = − q R=-q R = − q ).
Local thermodynamic stability of the STU black brane [47 , 48 , 49 ] holds if and only if
2 − q 1 − q 2 − q 3 + q 1 q 2 q 3 > 0 , q 1 + q 2 + q 3 < 3. (9) 2-q_{1}-q_{2}-q_{3}+q_{1}q_{2}q_{3}>0,\qquad q_{1}+q_{2}+q_{3}<3. \tag*{(9)} 2 − q 1 − q 2 − q 3 + q 1 q 2 q 3 > 0 , q 1 + q 2 + q 3 < 3. (9)
For a single charge q 1 q_{1} q 1 with q 2 = q 3 = 0 q_{2}=q_{3}=0 q 2 = q 3 = 0 , the stability condition gives q 1 < 2 q_{1}<2 q 1 < 2 [50 , 51 ] . For three equal charges q 1 = q 2 = q 3 = q q_{1}=q_{2}=q_{3}=q q 1 = q 2 = q 3 = q , it gives q < 1 q<1 q < 1 . The wave equation and the analysis below apply to any charges, but the interpretation as N = 4 \mathcal{N}=4 N = 4 SYM in equilibrium applies only to stable states.
The two examples used throughout this paper, ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 1 5 ) (q_{1},q_{2},q_{3})=(3,1,\frac{1}{5}) ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 5 1 ) and ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , are locally unstable, and in such states R-charge diffusion modes grow at nonzero momentum [48 , 49 ] . Nevertheless, we use them because their singular points are well separated, and we find no growing modes in the T x y T_{xy} T x y channel at zero momentum. For the stable states ( 1 , 0 , 0 ) (1,0,0) ( 1 , 0 , 0 ) and ( 1.6 , 0.3 , 0.02 ) (1.6,0.3,0.02) ( 1.6 , 0.3 , 0.02 ) , we checked that the OPE and the QNMs follow the same rules, except that ( 1 , 0 , 0 ) (1,0,0) ( 1 , 0 , 0 ) has no QNM line on the imaginary axis, since T 2 = T 3 T_{2}=T_{3} T 2 = T 3 there.
Three singular points
The zeros of P P P are singular points, and the zeros of D D D are horizons. We explain their configuration.
The curvature diverges at each zero of P P P , that is at R = − q i R=-q_{i} R = − q i . For example, near a simple zero with q i > 0 q_{i}>0 q i > 0 , the Ricci scalar behaves as ( R + q i ) − 7 / 3 (R+q_{i})^{-7/3} ( R + q i ) − 7/3 . For q 1 > q 2 > q 3 q_{1}>q_{2}>q_{3} q 1 > q 2 > q 3 , the three zeros are distinct. Coming from the boundary, the spacetime ends at the first one, R = − q 3 R=-q_{3} R = − q 3 , so this is the physical singularity. The other two, R = − q 2 R=-q_{2} R = − q 2 and R = − q 1 R=-q_{1} R = − q 1 , lie beyond it and are not part of the spacetime. Nevertheless, they matter. They are singular points of the metric, but the wave equation of the next section is regular there, so its solutions continue past R = − q 3 R=-q_{3} R = − q 3 . We will see that the observables are controlled by different singular points.
Since D ( 1 ) = 0 D(1)=0 D ( 1 ) = 0 , D ′ ( 1 ) > 0 D'(1)>0 D ′ ( 1 ) > 0 from β > 0 \beta>0 β > 0 , and D ( 0 ) = q 1 q 2 q 3 ≥ 0 D(0)=q_{1}q_{2}q_{3}\geq0 D ( 0 ) = q 1 q 2 q 3 ≥ 0 , D ( R ) D(R) D ( R ) has a zero in 0 ≤ R < 1 0\leq R<1 0 ≤ R < 1 . We call this R − R_{-} R − . Also, since D ( − q 1 ) = M q 1 > 0 D(-q_{1})=Mq_{1}>0 D ( − q 1 ) = M q 1 > 0 and D ( − ∞ ) = − ∞ D(-\infty)=-\infty D ( − ∞ ) = − ∞ , D ( R ) D(R) D ( R ) has a zero in R < − q 1 R<-q_{1} R < − q 1 , which we call R 3 R_{3} R 3 . We can write
D = ( R − 1 ) ( R − R − ) ( R − R 3 ) , 0 ≤ R − < 1 , R 3 < − q 1 . (10) D=(R-1)(R-R_{-})(R-R_{3}),\qquad0\leq R_{-}<1,\qquad R_{3}<-q_{1}. \tag*{(10)} D = ( R − 1 ) ( R − R − ) ( R − R 3 ) , 0 ≤ R − < 1 , R 3 < − q 1 . (10)
R = 1 R=1 R = 1 is the outer horizon, and for q 3 > 0 q_{3}>0 q 3 > 0 , R = R − > 0 R=R_{-}>0 R = R − > 0 is the inner horizon. R 3 R_{3} R 3 lies beyond all the singular points, so it is not part of the spacetime.
When q 3 > 0 q_{3}>0 q 3 > 0 , D ( − q 3 ) = M q 3 > 0 D(-q_{3})=Mq_{3}>0 D ( − q 3 ) = M q 3 > 0 and P ( R ) > 0 P(R)>0 P ( R ) > 0 for R > − q 3 R>-q_{3} R > − q 3 . So, R R R is spacelike near the physical singularity and the physical singularity is timelike, as in the Reissner–Nordström case. In terms of r r r , the point R = − q 3 R=-q_{3} R = − q 3 is the pair r = ± i q 3 r=\pm i\sqrt{q_{3}} r = ± i q 3 . On the other hand, when q 3 = 0 q_{3}=0 q 3 = 0 , the pair becomes degenerate at r = 0 r=0 r = 0 , and R − = 0 R_{-}=0 R − = 0 as well, so the inner horizon disappears. Then D < 0 D<0 D < 0 for 0 < R < 1 0<R<1 0 < R < 1 , and the singularity at R = 0 R=0 R = 0 is a spacelike Kasner singularity.
Optical times
From the metric, a radial null ray satisfies d t = ± P 2 D d R dt = \pm\frac{\sqrt{P}}{2D}\,dR d t = ± 2 D P d R . The time it takes to go from the boundary to the singular point R = − q i R=-q_i R = − q i and back, is given by
T i ≡ 2 ∫ − q i ∞ P 2 D d R . (11) T_i \equiv2\int_{-q_i}^{\infty}\frac{\sqrt{P}}{2D}\,dR. \tag*{(11)} T i ≡ 2 ∫ − q i ∞ 2 D P d R . (11)
We call this the optical time of R = − q i R=-q_i R = − q i . The integrand has poles at the horizons where D = 0 D=0 D = 0 , and we pass them through the lower half R R R -plane. This makes T i T_i T i complex with Im T i > 0 \operatorname{Im} T_i>0 Im T i > 0 . Replacing − q i -q_i − q i by R R R in this integral gives a map R ↦ T ( R ) R\mapsto T(R) R ↦ T ( R ) , and the image of the real R R R axis is shown by the black lines in Figure 1 .
The exterior R ≥ 1 R\ge1 R ≥ 1 maps to the real T T T axis, and T → + ∞ T\to+\infty T → + ∞ as R → 1 R\to1 R → 1 . Each pole of the integrand contributes half of its residue. We define
β − = 2 π P ( R − ) ∣ D ′ ( R − ) ∣ , β 3 = 2 π P ( R 3 ) D ′ ( R 3 ) , (12) \beta_-=\frac{2\pi\sqrt{P(R_-)}}{\lvert D'(R_-)\rvert},\qquad\beta_3=\frac{2\pi\sqrt{P(R_3)}}{D'(R_3)}, \tag*{(12)} β − = ∣ D ′ ( R − )∣ 2 π P ( R − ) , β 3 = D ′ ( R 3 ) 2 π P ( R 3 ) , (12)
in analogy with β \beta β at the outer horizon R = 1 R=1 R = 1 . We take the absolute value in β − \beta_- β − because D ′ ( R − ) < 0 D'(R_-)<0 D ′ ( R − ) < 0 , so that β − ≥ 0 \beta_-\ge0 β − ≥ 0 , just like β \beta β and β 3 \beta_3 β 3 . Passing R = 1 R=1 R = 1 shifts T T T by i β / 2 i\beta/2 i β /2 , so $R_-<R<1 maps to the line Im T = β / 2 \operatorname{Im} T = \beta/2 Im T = β /2 . Passing R = R − R = R_{-} R = R − shifts T T T by − i β − / 2 -i\beta_{-}/2 − i β − /2 , so − q 3 < R < R − -q_{3} < R < R_{-} − q 3 < R < R − maps to the line Im T = 1 2 ( β − β − ) \operatorname{Im} T = \frac{1}{2}(\beta- \beta_{-}) Im T = 2 1 ( β − β − ) , which ends at T 3 T_{3} T 3 .
When − q 2 < R < − q 3 -q_{2} < R < -q_{3} − q 2 < R < − q 3 , the integrand P / 2 D \sqrt{P}/2D P /2 D is imaginary, so T 2 T_{2} T 2 and T 3 T_{3} T 3 are joined by a vertical edge of length G G G . When − q 1 < R < − q 2 -q_{1} < R < -q_{2} − q 1 < R < − q 2 , the integrand is real, so T 1 T_{1} T 1 and T 2 T_{2} T 2 are joined by a horizontal edge of length B B B . When R < − q 1 R < -q_{1} R < − q 1 , the integrand is imaginary again, so the black line continues vertically. Passing R = R 3 R = R_{3} R = R 3 shifts T T T by − β 3 / 2 -\beta_{3}/2 − β 3 /2 . The black line then returns to T = 0 T = 0 T = 0 at R = − ∞ R = -\infty R = − ∞ along the imaginary axis, since the integrand falls as ∣ R ∣ − 3 / 2 |R|^{-3/2} ∣ R ∣ − 3/2 and has no poles in the lower half R R R -plane. This gives Re T 1 = 1 2 β 3 \operatorname{Re} T_{1} = \frac{1}{2}\beta_{3} Re T 1 = 2 1 β 3 . Therefore,
T 1 = β 3 2 + i 2 ( β − β − ) − i G , T 2 = T 1 + B , T 3 = T 2 + i G , (13) T_{1} = \frac{\beta_{3}}{2} + \frac{i}{2}(\beta- \beta_{-}) - iG,\qquad T_{2} = T_{1} + B,\qquad T_{3} = T_{2} + iG, \tag*{(13)} T 1 = 2 β 3 + 2 i ( β − β − ) − i G , T 2 = T 1 + B , T 3 = T 2 + i G , (13)
and the three optical times form a right triangle. For a special case, if q 3 = 0 q_{3} = 0 q 3 = 0 , we have β − = 0 \beta_{-} = 0 β − = 0 and hence Im T 3 = β / 2 \operatorname{Im} T_{3} = \beta/2 Im T 3 = β /2 .
For ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 1 5 ) (q_{1}, q_{2}, q_{3}) = (3, 1, \frac{1}{5}) ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 5 1 ) , we obtain
T 1 = 0.6550170488 + 0.9748769088 i , ∣ T 1 ∣ = 1.1744923676 , T 2 = 0.7597022366 + 0.9748769088 i , ∣ T 2 ∣ = 1.2359338476 , T 3 = 0.7597022366 + 1.0585677062 i , ∣ T 3 ∣ = 1.3029631909 , (14) \begin{aligned}
T_{1} &= 0.6550170488 + 0.9748769088i,\qquad|T_{1}| = 1.1744923676,\\
T_{2} &= 0.7597022366 + 0.9748769088i,\qquad|T_{2}| = 1.2359338476,\\
T_{3} &= 0.7597022366 + 1.0585677062i,\qquad|T_{3}| = 1.3029631909, \tag*{(14)}
\end{aligned} T 1 T 2 T 3 = 0.6550170488 + 0.9748769088 i , ∣ T 1 ∣ = 1.1744923676 , = 0.7597022366 + 0.9748769088 i , ∣ T 2 ∣ = 1.2359338476 , = 0.7597022366 + 1.0585677062 i , ∣ T 3 ∣ = 1.3029631909 , (14)
with B = 0.1046851878 B = 0.1046851878 B = 0.1046851878 and G = 0.0836907974 G = 0.0836907974 G = 0.0836907974 .
For ( q 1 , q 2 , q 3 ) = ( 2 , 1 2 , 0 ) (q_{1}, q_{2}, q_{3}) = (2, \frac{1}{2}, 0) ( q 1 , q 2 , q 3 ) = ( 2 , 2 1 , 0 ) , we obtain
T 1 = 0.7916069411 + 1.1947240304 i , ∣ T 1 ∣ = 1.4331807486 , T 2 = 0.9588164034 + 1.1947240304 i , ∣ T 2 ∣ = 1.5318924258 , T 3 = 0.9588164034 + 1.4809609794 i , ∣ T 3 ∣ = 1.7642489388 , (15) \begin{aligned}
T_{1} &= 0.7916069411 + 1.1947240304i,\qquad|T_{1}| = 1.4331807486,\\
T_{2} &= 0.9588164034 + 1.1947240304i,\qquad|T_{2}| = 1.5318924258,\\
T_{3} &= 0.9588164034 + 1.4809609794i,\qquad|T_{3}| = 1.7642489388, \tag*{(15)}
\end{aligned} T 1 T 2 T 3 = 0.7916069411 + 1.1947240304 i , ∣ T 1 ∣ = 1.4331807486 , = 0.9588164034 + 1.1947240304 i , ∣ T 2 ∣ = 1.5318924258 , = 0.9588164034 + 1.4809609794 i , ∣ T 3 ∣ = 1.7642489388 , (15)
with B = 0.1672094623 B = 0.1672094623 B = 0.1672094623 and G = 0.2862369490 G = 0.2862369490 G = 0.2862369490 .
The T x y T_{xy} T x y correlator
We study the retarded correlator of the transverse stress-tensor component T x y T_{xy} T x y at zero spatial momentum, where x x x and y y y are two components of x ⃗ \vec{x} x in (5) . In any four-dimensional CFT, the stress tensor is a universal operator of dimension Δ = 4 \Delta= 4 Δ = 4 . Thus, unlike a probe operator with an arbitrarily chosen mass or dimension, T x y T_{xy} T x y is an operator that is intrinsically present in N = 4 \mathcal{N} = 4 N = 4 SYM.
This channel is particularly simple in the charged STU background. At zero spatial momentum, its bulk dual, the tensor metric fluctuation h y x h^{x}_{y} h y x , decouples from the gauge fields and scalars and obeys the equation of a minimally coupled massless scalar [44 , 45 , 46 ] . Other channels generally involve coupled fluctuations at finite density. The T x y T_{xy} T x y channel has been extensively studied through shear viscosity, quasinormal modes, and spectral functions [44 , 52 , 53 , 54 , 55 , 56 , 57 ] , and more recently in studies of boundary signatures of black-hole singularities [18 , 19 , 41 ] . Here, the T x y T_{xy} T x y channel provides a clean setting in which to compare how the different singular points of the STU geometry appear in the spectral function, the OPE, and the QNM spectrum.
Wave equation
At zero spatial momentum, the transverse graviton h y x = ϕ ( R ) e − i ω t h^x_y=\phi(R)e^{-i\omega t} h y x = ϕ ( R ) e − iω t has spin 2 under rotations of x ⃗ \vec{x} x , so it decouples from the gauge fields and the scalars, and it is a minimally coupled massless scalar, so its wave equation is given by
∂ μ ( − g g μ ν ∂ ν h y x ) = 0. (16) \partial_\mu\left(\sqrt{-g}g^{\mu\nu}\partial_\nu h^x_y\right)=0. \tag*{(16)} ∂ μ ( − g g μν ∂ ν h y x ) = 0. (16)
Since the metric is diagonal and depends only on R R R , and h y x h^x_y h y x does not depend on x ⃗ \vec{x} x , this becomes
∂ R ( − g g R R ∂ R ϕ ) + ω 2 − g ( − g t t ) ϕ = 0. (17) \partial_R\left(\sqrt{-g}g^{RR}\partial_R\phi\right)+\omega^2\sqrt{-g}\left(-g^{tt}\right)\phi=0. \tag*{(17)} ∂ R ( − g g R R ∂ R ϕ ) + ω 2 − g ( − g tt ) ϕ = 0. (17)
From (5) , we have
− g = D P 2 / 3 ⋅ P 1 / 3 4 D ⋅ ( P 1 / 3 ) 3 = P 2 / 3 4 , (18) -g=\frac{D}{P^{2/3}}\cdot\frac{P^{1/3}}{4D}\cdot\left(P^{1/3}\right)^3=\frac{P^{2/3}}{4}, \tag*{(18)} − g = P 2/3 D ⋅ 4 D P 1/3 ⋅ ( P 1/3 ) 3 = 4 P 2/3 , (18)
since x ⃗ \vec{x} x is three dimensional, and
− g g R R = 2 D , − g ( − g t t ) = P 2 D . (19) \sqrt{-g}g^{RR}=2D,\qquad\sqrt{-g}\left(-g^{tt}\right)=\frac{P}{2D}. \tag*{(19)} − g g R R = 2 D , − g ( − g tt ) = 2 D P . (19)
So, the wave equation becomes
∂ R ( D ∂ R ϕ ) = − ω 2 P 4 D ϕ . (20) \partial_R\left(D\partial_R\phi\right)=-\frac{\omega^2P}{4D}\phi. \tag*{(20)} ∂ R ( D ∂ R ϕ ) = − 4 D ω 2 P ϕ . (20)
We introduce the coordinate x x x by
d x = − d R 2 D , x = − ∫ ∞ R d R ′ 2 D ( R ′ ) , (21) dx=-\frac{dR}{2D},\qquad x=-\int_{\infty}^{R}\frac{dR'}{2D(R')}, \tag*{(21)} d x = − 2 D d R , x = − ∫ ∞ R 2 D ( R ′ ) d R ′ , (21)
which vanishes at the boundary, increases inwards and diverges logarithmically at the horizon. In terms of x x x , the wave equation is written in the Schrödinger form,
ϕ ′ ′ + ω 2 P ϕ = 0 , (22) \phi''+\omega^2P\phi=0, \tag*{(22)} ϕ ′′ + ω 2 P ϕ = 0 , (22)
where ′ = d d x \prime=\frac{d}{dx} ′ = d x d .
Observables
Near the boundary, the two solutions of the wave equation behave as ϕ A ≃ 1 \phi_A \simeq1 ϕ A ≃ 1 and ϕ B ≃ R − 2 \phi_B \simeq R^{-2} ϕ B ≃ R − 2 . They are the non-normalizable and the normalizable mode. We write the solution that is infalling at the horizon as
ϕ i n = A ( ω ) ϕ A + B ( ω ) ϕ B . (23) \phi_{\mathrm{in}} = \mathcal{A}(\omega)\phi_A+\mathcal{B}(\omega)\phi_B. \tag*{(23)} ϕ in = A ( ω ) ϕ A + B ( ω ) ϕ B . (23)
Since A \mathcal{A} A is the source and B \mathcal{B} B is the response, up to normalization and contact terms, the retarded correlator of T x y T_{xy} T x y at zero spatial momentum,
G R ( t ) = i θ ( t ) ⟨ [ T x y ( t ) , T x y ( 0 ) ] ⟩ β , G R ( ω ) = ∫ d t e i ω t G R ( t ) , (24) G_R(t)=i\theta(t)\left\langle[T_{xy}(t),T_{xy}(0)]\right\rangle_\beta,\qquad G_R(\omega)=\int dt\,e^{i\omega t}G_R(t), \tag*{(24)} G R ( t ) = i θ ( t ) ⟨ [ T x y ( t ) , T x y ( 0 )] ⟩ β , G R ( ω ) = ∫ d t e iω t G R ( t ) , (24)
is given by [58 ]
G R ( ω ) = B ( ω ) A ( ω ) . (25) G_R(\omega)=\frac{\mathcal{B}(\omega)}{\mathcal{A}(\omega)}. \tag*{(25)} G R ( ω ) = A ( ω ) B ( ω ) . (25)
We study three observables built from it. The first is the spectral function ρ ( ω ) = Im G R ( ω ) \rho(\omega)=\operatorname{Im}G_R(\omega) ρ ( ω ) = Im G R ( ω ) for real ω \omega ω . In Appendix § 9 , when ϕ i n \phi_{\mathrm{in}} ϕ in is normalized as ∣ ϕ i n ∣ → 1 |\phi_{\mathrm{in}}|\to1 ∣ ϕ in ∣ → 1 at the horizon, we show that
ρ = ω P ( 1 ) 4 ∣ A ( ω ) ∣ 2 , (26) \rho=\frac{\omega\sqrt{P(1)}}{4|\mathcal{A}(\omega)|^2}, \tag*{(26)} ρ = 4∣ A ( ω ) ∣ 2 ω P ( 1 ) , (26)
and that the vacuum value is ρ v a c = π ω 4 / 32 \rho_{\mathrm{vac}}=\pi\omega^4/32 ρ vac = π ω 4 /32 . At large real ω \omega ω , we study the difference Δ ρ = ρ − ρ v a c \Delta\rho=\rho-\rho_{\mathrm{vac}} Δ ρ = ρ − ρ vac . The second is the set of OPE coefficients, the coefficients of the expansion of G R ( t ) G_R(t) G R ( t ) near t = 0 t=0 t = 0 . The third is the set of QNMs, the poles of G R ( ω ) G_R(\omega) G R ( ω ) . They are the solutions of A ( ω ) = 0 \mathcal{A}(\omega)=0 A ( ω ) = 0 .
Reflection from the singular points
We now find A \mathcal{A} A and B \mathcal{B} B at large ω \omega ω . The idea is that a wave sent in from the boundary is reflected very slightly by the singular points, and the reflected wave modifies A \mathcal{A} A and B \mathcal{B} B . The same idea was used for the AdS-Schwarzschild black brane in [18 ] and for the Reissner–Nordström black brane in [41 ] , each with a single singular point.
At large ω \omega ω we solve (22) by the WKB method. Writing ϕ = e Φ \phi=e^\Phi ϕ = e Φ , the equation becomes Φ ′ ′ + Φ ′ 2 + ω 2 P = 0 \Phi''+\Phi'^2+\omega^2P=0 Φ ′′ + Φ ′2 + ω 2 P = 0 . We expand
Φ ′ = i ω y 0 + y 1 + O ( 1 / ω ) (27) \Phi'=i\omega y_0+y_1+O(1/\omega) \tag*{(27)} Φ ′ = iω y 0 + y 1 + O ( 1/ ω ) (27)
and collect powers of ω \omega ω :
y 0 = ± P , y 1 = − P ′ 4 P . (28) y_0=\pm\sqrt{P},\qquad y_1=-\frac{P'}{4P}. \tag*{(28)} y 0 = ± P , y 1 = − 4 P P ′ . (28)
The two WKB solutions are therefore
f ± = P − 1 / 4 e ± i ω S , S ( x ) = ∫ 0 x P d x ′ . (29) f_{\pm}=P^{-1/4}e^{\pm i\omega S},\qquad S(x)=\int_{0}^{x}\sqrt{P}\,dx'. \tag*{(29)} f ± = P − 1/4 e ± iω S , S ( x ) = ∫ 0 x P d x ′ . (29)
Since d S = P d x = − P 2 D d R dS=\sqrt{P}\,dx=-\frac{\sqrt{P}}{2D}\,dR d S = P d x = − 2 D P d R , S S S is the tortoise coordinate measured from the boundary. It vanishes at the boundary and grows inward. With the time dependence e − i ω t e^{-i\omega t} e − iω t , f + f_{+} f + moves inward, so near the horizon
ϕ i n ∝ f + . (30) \phi_{\mathrm{in}}\propto f_{+}. \tag*{(30)} ϕ in ∝ f + . (30)
The WKB approximation fails near the zeros of P P P and near the boundary. On the real section R ≥ 1 R\geq1 R ≥ 1 , between the horizon and the boundary, we have P > 0 P>0 P > 0 , so WKB fails there only near the boundary, which we treat separately below. The WKB series of f + f_{+} f + never produces f − f_{-} f − : to any order in 1 / ω 1/\omega 1/ ω , an infalling wave stays infalling. Any reflection is therefore exponentially small, and it must come from the zeros of P P P , where WKB fails.
The zeros of P P P are not on the real section R ≥ 1 R\geq1 R ≥ 1 , but those with q c > 0 q_c>0 q c > 0 are regular points of the wave equation, so we can analytically continue ϕ \phi ϕ to them through the complex plane. In the complex S S S plane, the singular point R = − q c R=-q_c R = − q c sits at
S c = ∫ 0 x c P d x = ∫ − q c ∞ P 2 D d R = T c 2 , Im S c > 0 , (31) S_c=\int_{0}^{x_c}\sqrt{P}\,dx=\int_{-q_c}^{\infty}\frac{\sqrt{P}}{2D}\,dR=\frac{T_c}{2},\qquad\operatorname{Im}S_c>0, \tag*{(31)} S c = ∫ 0 x c P d x = ∫ − q c ∞ 2 D P d R = 2 T c , Im S c > 0 , (31)
where T c T_c T c is its optical time (11) . The physical path, the real S S S axis, passes below S c S_c S c since Im S c > 0 \operatorname{Im}S_c>0 Im S c > 0 . We show next that near S c S_c S c the infalling wave acquires a reflected component of size e 2 i ω S c = e i ω T c e^{2i\omega S_c}=e^{i\omega T_c} e 2 iω S c = e iω T c .
At x = x c x=x_c x = x c , we have P = 0 P=0 P = 0 . Near x c x_c x c , we write P ≃ C X ℓ P\simeq CX^\ell P ≃ C X ℓ with X = x − x c X=x-x_c X = x − x c . The local equation ∂ X 2 ϕ + ω 2 C X ℓ ϕ = 0 \partial_X^2\phi+\omega^2CX^\ell\phi=0 ∂ X 2 ϕ + ω 2 C X ℓ ϕ = 0 is solved exactly by
ϕ = X Z ν ( ξ ) , ν = 1 ℓ + 2 , ξ = 2 ω C ℓ + 2 X ℓ + 2 2 . (32) \phi=\sqrt{X}Z_\nu(\xi),\qquad\nu=\frac{1}{\ell+2},\qquad\xi=\frac{2\omega\sqrt{C}}{\ell+2}X^{\frac{\ell+2}{2}}. \tag*{(32)} ϕ = X Z ν ( ξ ) , ν = ℓ + 2 1 , ξ = ℓ + 2 2 ω C X 2 ℓ + 2 . (32)
where Z ν Z_\nu Z ν is any Bessel function of order ν \nu ν . For a simple zero with q c > 0 q_c>0 q c > 0 , we have D ( R = − q c ) = M q c ≠ 0 D(R=-q_c)=Mq_c\neq0 D ( R = − q c ) = M q c = 0 , so d x / d R = − 1 / ( 2 D ) dx/dR=-1/(2D) d x / d R = − 1/ ( 2 D ) is finite and nonzero there. Hence X ≃ − R + q c 2 M q c X\simeq-\frac{R+q_c}{2Mq_c} X ≃ − 2 M q c R + q c near x c x_c x c , so ℓ = 1 \ell=1 ℓ = 1 and ν = 1 3 \nu=\frac{1}{3} ν = 3 1 . In this case the solutions (32) are Airy functions.
Near x c x_c x c , d S = P d x ≃ C X ℓ / 2 d X dS=\sqrt{P}\,dx\simeq\sqrt{C}X^{\ell/2}dX d S = P d x ≃ C X ℓ /2 d X , so ξ ≃ ω ( S − S c ) \xi\simeq\omega(S-S_c) ξ ≃ ω ( S − S c ) . For ∣ ξ ∣ ≫ 1 |\xi|\gg1 ∣ ξ ∣ ≫ 1 ,
H ν ( 1 , 2 ) ( ξ ) ≃ 2 π ξ e ± i ( ξ − ν π 2 − π 4 ) . (33) H_\nu^{(1,2)}(\xi)\simeq\sqrt{\frac{2}{\pi\xi}}e^{\pm i\left(\xi-\frac{\nu\pi}{2}-\frac{\pi}{4}\right)}. \tag*{(33)} H ν ( 1 , 2 ) ( ξ ) ≃ π ξ 2 e ± i ( ξ − 2 ν π − 4 π ) . (33)
and since X ξ − 1 / 2 ∝ X − ℓ / 4 ∝ P − 1 / 4 \sqrt{X}\,\xi^{-1/2}\propto X^{-\ell/4}\propto P^{-1/4} X ξ − 1/2 ∝ X − ℓ /4 ∝ P − 1/4 , the solutions (32) reduce to the WKB waves:
X H ν ( 1 ) ( ξ ) ∝ f + , X H ν ( 2 ) ( ξ ) ∝ f − . (34) \sqrt{X}H_\nu^{(1)}(\xi)\propto f_{+},\qquad\sqrt{X}H_\nu^{(2)}(\xi)\propto f_{-}. \tag*{(34)} X H ν ( 1 ) ( ξ ) ∝ f + , X H ν ( 2 ) ( ξ ) ∝ f − . (34)
The WKB waves hold far from the singular point and the local solutions hold near it. At large ω \omega ω these two regions overlap, and the two descriptions are matched there.
On the horizon side of S c S_c S c , ϕ i n \phi_{\mathrm{in}} ϕ in is a pure f + f_+ f + up to exponentially small terms. Near x c x_c x c it is therefore X H ν ( 1 ) ( ξ ) \sqrt{X}H_\nu^{(1)}(\xi) X H ν ( 1 ) ( ξ ) with arg ξ ≈ 0 \arg\xi\approx0 arg ξ ≈ 0 . Following the real axis toward the boundary, the path passes below S c S_c S c , and arg ( S − S c ) \arg(S-S_c) arg ( S − S c ) changes from 0 to − π -\pi − π . On the boundary side we therefore write
ξ ≃ ξ 0 e − i π , ξ 0 = ω ( S c − S ) , (35) \xi\simeq\xi_0 e^{-i\pi}, \qquad\xi_0=\omega(S_c-S), \tag*{(35)} ξ ≃ ξ 0 e − iπ , ξ 0 = ω ( S c − S ) , (35)
with ξ 0 \xi_0 ξ 0 approximately real and positive, and use the continuation formula [59 §10.11]
H ν ( 1 ) ( ξ 0 e − i π ) = 2 cos ( ν π ) H ν ( 1 ) ( ξ 0 ) + e − i ν π H ν ( 2 ) ( ξ 0 ) = 2 π ξ 0 [ 2 cos ( ν π ) e − i ( ν π 2 + π 4 ) e i ξ 0 + e − i ν π e i ( ν π 2 + π 4 ) e − i ξ 0 ] . (36) \begin{aligned}
H_\nu^{(1)}(\xi_0e^{-i\pi})&=2\cos(\nu\pi)H_\nu^{(1)}(\xi_0)+e^{-i\nu\pi}H_\nu^{(2)}(\xi_0)\\
&=\sqrt{\frac{2}{\pi\xi_0}}\left[2\cos(\nu\pi)e^{-i\left(\frac{\nu\pi}{2}+\frac{\pi}{4}\right)}e^{i\xi_0}+e^{-i\nu\pi}e^{i\left(\frac{\nu\pi}{2}+\frac{\pi}{4}\right)}e^{-i\xi_0}\right]. \tag*{(36)}
\end{aligned} H ν ( 1 ) ( ξ 0 e − iπ ) = 2 cos ( ν π ) H ν ( 1 ) ( ξ 0 ) + e − i ν π H ν ( 2 ) ( ξ 0 ) = π ξ 0 2 [ 2 cos ( ν π ) e − i ( 2 ν π + 4 π ) e i ξ 0 + e − i ν π e i ( 2 ν π + 4 π ) e − i ξ 0 ] . (36)
Since ξ 0 = ω ( S c − S ) \xi_0=\omega(S_c-S) ξ 0 = ω ( S c − S ) ,
e i ξ 0 = e i ω S c e − i ω S = e i ω S c P 1 / 4 f − , e − i ξ 0 = e − i ω S c e i ω S = e − i ω S c P 1 / 4 f + . (37) e^{i\xi_0}=e^{i\omega S_c}e^{-i\omega S}=e^{i\omega S_c}P^{1/4}f_-, \qquad e^{-i\xi_0}=e^{-i\omega S_c}e^{i\omega S}=e^{-i\omega S_c}P^{1/4}f_+. \tag*{(37)} e i ξ 0 = e iω S c e − iω S = e iω S c P 1/4 f − , e − i ξ 0 = e − iω S c e iω S = e − iω S c P 1/4 f + . (37)
The ratio of the two coefficients of f − f_- f − , f + f_+ f + on the boundary side is
( f − c o e f f i c i e n t ) ( f + c o e f f i c i e n t ) = 2 cos ( ν π ) e − i ( ν π 2 + π 4 ) e i ω S c e − i ν π e i ( ν π 2 + π 4 ) e − i ω S c = − 2 i cos ( π ν ) e i ω T c . (38) \begin{aligned}
\frac{(f_-\ \mathrm{coefficient})}{(f_+\ \mathrm{coefficient})}
&=\frac{2\cos(\nu\pi)e^{-i\left(\frac{\nu\pi}{2}+\frac{\pi}{4}\right)}e^{i\omega S_c}}{e^{-i\nu\pi}e^{i\left(\frac{\nu\pi}{2}+\frac{\pi}{4}\right)}e^{-i\omega S_c}}\\
&=-2i\cos(\pi\nu)e^{i\omega T_c}. \tag*{(38)}
\end{aligned} ( f + coefficient ) ( f − coefficient ) = e − i ν π e i ( 2 ν π + 4 π ) e − iω S c 2 cos ( ν π ) e − i ( 2 ν π + 4 π ) e iω S c = − 2 i cos ( π ν ) e iω T c . (38)
On the boundary side of S c S_c S c , the infalling solution is therefore
ϕ i n ∝ f + − 2 i cos ( π ν c ) e i ω T c f − , (39) \phi_{\mathrm{in}}\propto f_+-2i\cos(\pi\nu_c)e^{i\omega T_c}f_-, \tag*{(39)} ϕ in ∝ f + − 2 i cos ( π ν c ) e iω T c f − , (39)
where ν c \nu_c ν c is the value of ν \nu ν at R = − q c R=-q_c R = − q c . The reflected wave is switched on where the path crosses the vertical line Re S = Re S c \operatorname{Re}S=\operatorname{Re}S_c Re S = Re S c below S c S_c S c . On this line e 2 i ω ( S − S c ) e^{2i\omega(S-S_c)} e 2 iω ( S − S c ) is real, so one WKB wave is maximally dominant over the other. This line is the Stokes line of the singular point. In terms of T = 2 S T=2S T = 2 S , it is the vertical line Re T = Re T c \operatorname{Re}T=\operatorname{Re}T_c Re T = Re T c below T c T_c T c . Figure 2 (a) shows the Stokes lines of T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 together with the path. The reflected wave is switched on only where the path crosses a Stokes line. So the result depends only on which Stokes lines the path crosses, and in which direction it crosses each of them. It does not depend on the shape of the path, as long as the path does not cross the black lines of Figure 1 .
Figure 2. Stokes lines (dotted) of T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , for (a) real ω \omega ω , (b) arg ω = − arg T 2 \arg\omega=-\arg T_2 arg ω = − arg T 2 and (c) ω = − i y \omega=-iy ω = − i y , and (d) for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) and ω = − i y \omega=-iy ω = − i y . The black lines are those of Figure 1 , and the insets in (a)–(c) zoom in on T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 . The gray arrow is the path along which ϕ i n \phi_{\rm in} ϕ in in (a) and (b), or ϕ B \phi_B ϕ B in (c) and (d), is followed from where it is fixed, and a reflected wave is added at each dot, where the path crosses a Stokes line. In (a), the Stokes line of T 3 T_3 T 3 runs along the vertical edge into T 2 T_2 T 2 . In (b), the red line is where ∣ e i ω T ∣ = 1 |e^{i\omega T}|=1 ∣ e iω T ∣ = 1 , and it passes through T 2 T_2 T 2 . In (c) and (d), the Stokes line of T 1 T_1 T 1 runs along the horizontal edge into T 2 T_2 T 2 . In (d), T 3 T_3 T 3 lies on the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 , so the path ends on the Stokes line of T 3 T_3 T 3 (half-filled dot) instead of crossing it.
Each singular point produces its own reflection. Every reflection is exponentially small, so terms with two reflections are of order e − ω ( Im T c + Im T c ′ ) e^{-\omega(\operatorname{Im}T_c+\operatorname{Im}T_{c'})} e − ω ( Im T c + Im T c ′ ) and can be dropped:
ϕ i n ∝ f + − ( ∑ c 2 i cos ( π ν c ) e i ω T c ) f − . (40) \phi_{\mathrm{in}}\propto f_+-\left(\sum_c 2i\cos(\pi\nu_c)e^{i\omega T_c}\right)f_-. \tag*{(40)} ϕ in ∝ f + − ( c ∑ 2 i cos ( π ν c ) e iω T c ) f − . (40)
The local computation above assumes that the Stokes line reaches the real axis without meeting another singular point. We call a singular point visible if the vertical segment from T c T_c T c down to the real T T T axis does not meet another black line of Figure 1 . Only visible points contribute their own term to (40) . T 1 T_1 T 1 and T 2 T_2 T 2 are visible, since the black lines other than the real T T T axis lie at Im T ≥ Im T 2 \operatorname{Im}T\geq\operatorname{Im}T_2 Im T ≥ Im T 2 or on Re T = 0 \operatorname{Re}T=0 Re T = 0 . Since T 2 − T 1 = B T_2-T_1=B T 2 − T 1 = B is real, they contribute at the same order e − ω Im T 1 e^{-\omega\operatorname{Im}T_1} e − ω Im T 1 . T 3 T_3 T 3 is not visible, since the segment below T 3 T_3 T 3 is the vertical edge of length G G G itself, which ends at T 2 T_2 T 2 . We say that T 3 T_3 T 3 is shadowed. It oscillates with the same frequency as T 2 T_2 T 2 , since Re T 3 = Re T 2 \operatorname{Re} T_3=\operatorname{Re} T_2 Re T 3 = Re T 2 , so it cannot appear as a separate term. It enters only through the T 2 T_2 T 2 term, as discussed in § 4 .
It remains to match to the boundary. Near the boundary,
D ( R ) ≃ R 3 , x = − ∫ ∞ R d R ′ 2 D ( R ′ ) ≃ 1 4 R 2 , R ≃ ( 4 x ) − 1 / 2 , (41) D(R)\simeq R^3,\qquad x=-\int_{\infty}^{R}\frac{\mathrm{d}R'}{2D(R')}\simeq\frac{1}{4R^2},\qquad R\simeq(4x)^{-1/2}, \tag*{(41)} D ( R ) ≃ R 3 , x = − ∫ ∞ R 2 D ( R ′ ) d R ′ ≃ 4 R 2 1 , R ≃ ( 4 x ) − 1/2 , (41)
so P ≃ R 3 ≃ ( 4 x ) − 3 / 2 P\simeq R^3\simeq(4x)^{-3/2} P ≃ R 3 ≃ ( 4 x ) − 3/2 . With z = 1 / R z=1/\sqrt{R} z = 1/ R , we have
S ≃ ( 4 x ) 1 / 4 = z , P − 1 / 4 ≃ z 3 / 2 . (42) S\simeq(4x)^{1/4}=z,\qquad P^{-1/4}\simeq z^{3/2}. \tag*{(42)} S ≃ ( 4 x ) 1/4 = z , P − 1/4 ≃ z 3/2 . (42)
For ψ = P 1 / 4 ϕ \psi=P^{1/4}\phi ψ = P 1/4 ϕ , (3.5) then becomes
ψ z z + ( ω 2 − 15 4 z 2 ) ψ = 0. (43) \psi_{zz}+\left(\omega^2-\frac{15}{4z^2}\right)\psi=0. \tag*{(43)} ψ z z + ( ω 2 − 4 z 2 15 ) ψ = 0. (43)
The WKB solutions f ± f_{\pm} f ± are ψ = e ± i ω z \psi=e^{\pm i\omega z} ψ = e ± iω z , which neglect the 1 / z 2 1/z^2 1/ z 2 term. So WKB fails for ω z ≲ 1 \omega z\lesssim1 ω z ≲ 1 , but the solutions are the vacuum ones, z 2 H 2 ( 1 , 2 ) ( ω z ) = R − 1 H 2 ( 1 , 2 ) ( ω / R ) z^2H_2^{(1,2)}(\omega z)=R^{-1}H_2^{(1,2)}(\omega/\sqrt{R}) z 2 H 2 ( 1 , 2 ) ( ω z ) = R − 1 H 2 ( 1 , 2 ) ( ω / R ) . WKB holds for ω z ≫ 1 \omega z\gg1 ω z ≫ 1 and the vacuum form holds for ω z 3 ≪ 1 \omega z^3\ll1 ω z 3 ≪ 1 , since S = z + O ( z 3 ) S=z+O(z^3) S = z + O ( z 3 ) , so at large ω \omega ω the two can be matched for 1 / ω ≪ z ≪ ω − 1 / 3 1/\omega\ll z\ll\omega^{-1/3} 1/ ω ≪ z ≪ ω − 1/3 . Using
z 2 H 2 ( 1 , 2 ) ( ω z ) ≃ 2 π ω z 3 / 2 e ± i ( ω z − 5 π 4 ) , (44) z^2H_2^{(1,2)}(\omega z)\simeq\sqrt{\frac{2}{\pi\omega}}z^{3/2}e^{\pm i(\omega z-\frac{5\pi}{4})}, \tag*{(44)} z 2 H 2 ( 1 , 2 ) ( ω z ) ≃ π ω 2 z 3/2 e ± i ( ω z − 4 5 π ) , (44)
we have
f + = π ω 2 e 5 π i 4 z 2 H 2 ( 1 ) ( ω z ) , f − = π ω 2 e − 5 π i 4 z 2 H 2 ( 2 ) ( ω z ) , (45) f_+=\sqrt{\frac{\pi\omega}{2}}e^{\frac{5\pi i}{4}}z^2H_2^{(1)}(\omega z),\qquad f_-=\sqrt{\frac{\pi\omega}{2}}e^{-\frac{5\pi i}{4}}z^2H_2^{(2)}(\omega z), \tag*{(45)} f + = 2 π ω e 4 5 π i z 2 H 2 ( 1 ) ( ω z ) , f − = 2 π ω e − 4 5 π i z 2 H 2 ( 2 ) ( ω z ) , (45)
and (3.25) becomes
ϕ i n ∝ z 2 [ H 2 ( 1 ) ( ω z ) + ρ ~ H 2 ( 2 ) ( ω z ) ] , ρ ~ = − 2 ∑ c cos ( π ν c ) e i ω T c . (46) \phi_{\mathrm{in}}\propto z^2\left[H_2^{(1)}(\omega z)+\widetilde{\rho}H_2^{(2)}(\omega z)\right],\qquad\widetilde{\rho}=-2\sum_c\cos(\pi\nu_c)e^{i\omega T_c}. \tag*{(46)} ϕ in ∝ z 2 [ H 2 ( 1 ) ( ω z ) + ρ H 2 ( 2 ) ( ω z ) ] , ρ = − 2 c ∑ cos ( π ν c ) e iω T c . (46)
For 1 / ω ≪ z ≪ ω − 1 / 3 1/\omega\ll z\ll\omega^{-1/3} 1/ ω ≪ z ≪ ω − 1/3 , ϕ A \phi_A ϕ A and ϕ B \phi_B ϕ B also take the vacuum form of Appendix A, z 2 J 2 ( ω z ) = ω 2 8 ϕ B z^2J_2(\omega z)=\frac{\omega^2}{8}\phi_B z 2 J 2 ( ω z ) = 8 ω 2 ϕ B and z 2 Y 2 ( ω z ) = − 4 π ω 2 ϕ A z^2Y_2(\omega z)=-\frac{4}{\pi\omega^2}\phi_A z 2 Y 2 ( ω z ) = − π ω 2 4 ϕ A . Hence
z 2 H 2 ( 1 , 2 ) ( ω z ) = ω 2 8 ϕ B ∓ 4 i π ω 2 ϕ A , A ∝ 1 − ρ ~ , B ∝ 1 + ρ ~ , (47) z^2H_2^{(1,2)}(\omega z)=\frac{\omega^2}{8}\phi_B\mp\frac{4i}{\pi\omega^2}\phi_A,\qquad\mathcal{A}\propto1-\widetilde{\rho},\qquad\mathcal{B}\propto1+\widetilde{\rho}, \tag*{(47)} z 2 H 2 ( 1 , 2 ) ( ω z ) = 8 ω 2 ϕ B ∓ π ω 2 4 i ϕ A , A ∝ 1 − ρ , B ∝ 1 + ρ , (47)
and
B A = i π ω 4 32 1 + ρ ~ 1 − ρ ~ . (48) \frac{\mathcal{B}}{\mathcal{A}}=\frac{i\pi\omega^4}{32}\frac{1+\widetilde{\rho}}{1-\widetilde{\rho}}. \tag*{(48)} A B = 32 iπ ω 4 1 − ρ 1 + ρ . (48)
Spectral function
From (48) , we have
ρ ρ v a c = Im ( B / A ) π ω 4 / 32 = Re 1 + ρ ~ 1 − ρ ~ . (49) \frac{\rho}{\rho_{\mathrm{vac}}}=\frac{\operatorname{Im}(\mathcal{B}/\mathcal{A})}{\pi\omega^4/32}=\operatorname{Re}\frac{1+\widetilde{\rho}}{1-\widetilde{\rho}}. \tag*{(49)} ρ vac ρ = π ω 4 /32 Im ( B / A ) = Re 1 − ρ 1 + ρ . (49)
Since ρ ~ \widetilde{\rho} ρ of (46) is exponentially small,
Δ ρ ρ v a c = ρ − ρ v a c ρ v a c = Re 2 ρ ~ 1 − ρ ~ ≃ 2 Re ρ ~ , (50) \frac{\Delta\rho}{\rho_{\mathrm{vac}}}=\frac{\rho-\rho_{\mathrm{vac}}}{\rho_{\mathrm{vac}}}=\operatorname{Re}\frac{2\widetilde{\rho}}{1-\widetilde{\rho}}\simeq2\operatorname{Re}\widetilde{\rho}, \tag*{(50)} ρ vac Δ ρ = ρ vac ρ − ρ vac = Re 1 − ρ 2 ρ ≃ 2 Re ρ , (50)
and inserting ρ ~ \widetilde{\rho} ρ gives
Δ ρ ρ v a c ≃ Re ∑ c A c e i ω T c , A c = − 4 cos ( π ν c ) , (51) \frac{\Delta\rho}{\rho_{\mathrm{vac}}}\simeq\operatorname{Re}\sum_c A_c e^{i\omega T_c},\qquad A_c=-4\cos(\pi\nu_c), \tag*{(51)} ρ vac Δ ρ ≃ Re c ∑ A c e iω T c , A c = − 4 cos ( π ν c ) , (51)
where the sum runs over the visible points T 1 T_1 T 1 and T 2 T_2 T 2 . We call this exponentially small correction of ρ \rho ρ to ρ v a c \rho_{\mathrm{vac}} ρ vac the spectral tail. The points T 1 T_1 T 1 and T 2 T_2 T 2 have the smallest Im T c \operatorname{Im}T_c Im T c , so the tail falls like e − ω Im T 1 e^{-\omega\operatorname{Im}T_1} e − ω Im T 1 .
The amplitude A c A_c A c depends only on the order ℓ \ell ℓ of the zero of P P P , through ν c = 1 / ( ℓ + 2 ) \nu_c=1/(\ell+2) ν c = 1/ ( ℓ + 2 ) . A simple zero has ν = 1 3 \nu=\frac{1}{3} ν = 3 1 and A = − 2 A=-2 A = − 2 , a double zero has ν = 1 4 \nu=\frac{1}{4} ν = 4 1 and A = − 2 2 A=-2\sqrt{2} A = − 2 2 , and a triple zero has ν = 1 5 \nu=\frac{1}{5} ν = 5 1 and A = − 4 cos ( π / 5 ) A=-4\cos(\pi/5) A = − 4 cos ( π /5 ) . Since R = − q 1 R=-q_1 R = − q 1 and R = − q 2 R=-q_2 R = − q 2 are simple zeros, A 1 = A 2 = − 2 A_1=A_2=-2 A 1 = A 2 = − 2 . For the Reissner–Nordström black brane, the three zeros merge into a triple zero, and its amplitude agrees with [41 ] .
Since T 2 − T 1 = B T_2-T_1=B T 2 − T 1 = B is real, the two terms from T 1 , T 2 T_1,T_2 T 1 , T 2 have the same size and they beat as
Δ ρ ρ v a c ≃ Re [ − 2 e i ω T 1 − 2 e i ω T 2 ] = − 4 e − ω Im T 1 cos ( ω Re T 1 + Re T 2 2 ) cos ( ω B 2 ) . (52) \begin{aligned}
\frac{\Delta\rho}{\rho_{\mathrm{vac}}}&\simeq\operatorname{Re}\left[-2e^{i\omega T_1}-2e^{i\omega T_2}\right]\\
&=-4e^{-\omega\operatorname{Im}T_1}\cos\left(\omega\frac{\operatorname{Re}T_1+\operatorname{Re}T_2}{2}\right)\cos\left(\frac{\omega B}{2}\right). \tag*{(52)}
\end{aligned} ρ vac Δ ρ ≃ Re [ − 2 e iω T 1 − 2 e iω T 2 ] = − 4 e − ω Im T 1 cos ( ω 2 Re T 1 + Re T 2 ) cos ( 2 ω B ) . (52)
The first cosine is the fast oscillation, with period 4 π / ( Re T 1 + Re T 2 ) 4\pi/(\operatorname{Re}T_1+\operatorname{Re}T_2) 4 π / ( Re T 1 + Re T 2 ) , which is 8.9 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and 7.2 for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) . Since B ≪ Re T 1 + Re T 2 B\ll\operatorname{Re}T_1+\operatorname{Re}T_2 B ≪ Re T 1 + Re T 2 , the second cosine varies slowly and gives its envelope, with period 2 π / B 2\pi/B 2 π / B , which is 60.0 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and 37.6 for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) .
The shadowed point T 3 T_3 T 3 enters through the T 2 T_2 T 2 term, since Re T 2 = Re T 3 \operatorname{Re}T_2=\operatorname{Re}T_3 Re T 2 = Re T 3 . In § 3.3 , we treated each zero of P P P separately. In the T T T -plane, the local solutions around T 2 T_2 T 2 are needed in a region of size of order 1 / ω 1/\omega 1/ ω , and T 3 T_3 T 3 is at distance G G G from T 2 T_2 T 2 , so this requires ω G ≫ 1 \omega G\gg1 ω G ≫ 1 . For ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , we have G = 0.286 G=0.286 G = 0.286 , so in the range 16 ≤ ω ≤ 120 16\leq\omega\leq120 16 ≤ ω ≤ 120 of our numerics ω G \omega G ω G runs from 4.6 to 34. Then T 3 T_3 T 3 lies well outside the region of size 1 / ω 1/\omega 1/ ω around T 2 T_2 T 2 , so we use (52) . For ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , we have G = 0.0837 G=0.0837 G = 0.0837 , so in this range ω G \omega G ω G only runs from 1.3 to 10, and we have to treat the two zeros together.
Near them, P P P is quadratic in x x x . With X = x − 1 2 ( x 2 + x 3 ) X=x-\frac{1}{2}(x_2+x_3) X = x − 2 1 ( x 2 + x 3 ) and δ = 1 2 ∣ x 2 − x 3 ∣ \delta=\frac{1}{2}|x_2-x_3| δ = 2 1 ∣ x 2 − x 3 ∣ , we write P ≃ C ( X 2 − δ 2 ) P\simeq C(X^2-\delta^2) P ≃ C ( X 2 − δ 2 ) . Integrating P \sqrt{P} P from x 2 x_2 x 2 to x 3 x_3 x 3 gives S 3 − S 2 = i G / 2 S_3-S_2=iG/2 S 3 − S 2 = i G /2 , so G = π C δ 2 G=\pi\sqrt{C}\,\delta^2 G = π C δ 2 . In terms of ζ = ( 4 ω 2 C ) 1 / 4 X \zeta=(4\omega^2C)^{1/4}X ζ = ( 4 ω 2 C ) 1/4 X , (3.7) becomes
∂ ζ 2 ϕ + ( ζ 2 4 − a ) ϕ = 0 , a = ω G 2 π . (53) \partial_\zeta^2\phi+\left(\frac{\zeta^2}{4}-a\right)\phi=0,\qquad a=\frac{\omega G}{2\pi}. \tag*{(53)} ∂ ζ 2 ϕ + ( 4 ζ 2 − a ) ϕ = 0 , a = 2 π ω G . (53)
This local equation is solved exactly by parabolic cylinder functions [59 §12.14] , which replace the Bessel functions of (32) . Repeating the calculation of § 3.3 , the reflected wave of T 2 T_2 T 2 in (40) is multiplied by a factor W ( ω G ) W(\omega G) W ( ω G ) , and we obtain
Δ ρ ρ v a c ≃ Re [ − 2 e i ω T 1 − 2 W ( ω G ) e i ω T 2 ] , (54) \frac{\Delta\rho}{\rho_{\mathrm{vac}}}\simeq\operatorname{Re}\left[-2e^{i\omega T_1}-2W(\omega G)e^{i\omega T_2}\right], \tag*{(54)} ρ vac Δ ρ ≃ Re [ − 2 e iω T 1 − 2 W ( ω G ) e iω T 2 ] , (54)
where
W ( g ) : = 2 π e − g / 4 ( g / 2 π e ) − i a Γ ( 1 2 − i g 2 π ) , a = g 2 π . (55) W(g):=\frac{\sqrt{2\pi}\,e^{-g/4}(g/2\pi e)^{-ia}}{\Gamma\left(\frac{1}{2}-\frac{ig}{2\pi}\right)},\qquad a=\frac{g}{2\pi}. \tag*{(55)} W ( g ) := Γ ( 2 1 − 2 π i g ) 2 π e − g /4 ( g /2 π e ) − ia , a = 2 π g . (55)
By Stirling’s formula, W ( ω G ) → 1 W(\omega G)\to1 W ( ω G ) → 1 for ω G ≫ 1 \omega G\gg1 ω G ≫ 1 , so T 2 T_2 T 2 reflects alone, as we explained. For G → 0 G\to0 G → 0 , the two zeros merge into a double zero with ℓ = 2 \ell=2 ℓ = 2 and ν = 1 4 \nu=\frac{1}{4} ν = 4 1 , and − 2 W → − 2 2 -2W\to-2\sqrt{2} − 2 W → − 2 2 reproduces its amplitude.
We now compare these formulas (52) and (54) with numerics. From (26) ,
Δ ρ ρ v a c = 8 P ( 1 ) π ω 3 ∣ A ( ω ) ∣ 2 − 1 , (56) \frac{\Delta\rho}{\rho_{\mathrm{vac}}}=\frac{8\sqrt{P(1)}}{\pi\omega^3\lvert A(\omega)\rvert^2}-1, \tag*{(56)} ρ vac Δ ρ = π ω 3 ∣ A ( ω ) ∣ 2 8 P ( 1 ) − 1 , (56)
so we only need A ( ω ) A(\omega) A ( ω ) , which we compute as explained in Appendix § 9 . Since Δ ρ / ρ v a c \Delta\rho/\rho_{\mathrm{vac}} Δ ρ / ρ vac from (52) and (54) falls like e − ω Im T 1 e^{-\omega\operatorname{Im}T_1} e − ω Im T 1 , we plot e ω Im T 1 Δ ρ / ρ v a c e^{\omega\operatorname{Im}T_1}\Delta\rho/\rho_{\mathrm{vac}} e ω Im T 1 Δ ρ / ρ vac , which stays of order one.
Figure 3 compares the numerics for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) with (54) , and Figure 4 compares the numerics for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) with (52) . These formulas have no fitted parameter, and the two agree, including the beats. To test the amplitudes, we also tried fitting the data with ω ≥ 55 \omega\ge55 ω ≥ 55 , leaving A 1 A_1 A 1 and A 2 A_2 A 2 in (51) free and allowing corrections up to 1 / ω 3 1/\omega^3 1/ ω 3 to each term. For ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , with the factor W W W , we obtain A 1 = − 2.0004 + 0.0005 i A_1=-2.0004+0.0005i A 1 = − 2.0004 + 0.0005 i and A 2 = − 1.9998 − 0.0004 i A_2=-1.9998-0.0004i A 2 = − 1.9998 − 0.0004 i . For ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , we obtain A 1 = − 2.0000 + 0.0001 i A_1=-2.0000+0.0001i A 1 = − 2.0000 + 0.0001 i and A 2 = − 2.0001 − 0.0001 i A_2=-2.0001-0.0001i A 2 = − 2.0001 − 0.0001 i .
Figure 3. e ω Im T 1 Δ ρ / ρ v a c e^{\omega\operatorname{Im}T_1}\Delta\rho/\rho_{\mathrm{vac}} e ω Im T 1 Δ ρ / ρ vac for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) . The points are the numerics and the line is (4.6), with no fitted parameter.
Figure 4. e ω Im T 1 Δ ρ / ρ v a c e^{\omega}\operatorname{Im}T_1\,\Delta\rho/\rho_{\mathrm{vac}} e ω Im T 1 Δ ρ / ρ vac for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) . The points are the numerics and the line is (52) , with no fitted parameter.
Therefore, the spectral function at large real ω \omega ω is controlled by T 1 T_1 T 1 and T 2 T_2 T 2 .
OPE convergence radius
At short times, the retarded correlator is a vacuum term plus a power series,
G R ( t ) = 3 2 t − 5 + ∑ m ≥ 1 γ m t 2 m − 1 , (57) G_R(t)=\frac{3}{2}t^{-5}+\sum_{m\ge1}\gamma_m t^{2m-1}, \tag*{(57)} G R ( t ) = 2 3 t − 5 + m ≥ 1 ∑ γ m t 2 m − 1 , (57)
as we show in Appendix § 10 , where we also explain how to compute γ m \gamma_m γ m . The radius of convergence of the series is the distance from t = 0 t=0 t = 0 to the nearest singularity of G R ( t ) − 3 2 t − 5 G_R(t)-\frac{3}{2}t^{-5} G R ( t ) − 2 3 t − 5 . The spectral function tells us about these singularities. Since G R ( t ) G_R(t) G R ( t ) vanishes for t < 0 t<0 t < 0 and is real, the spectral function is
ρ ( ω ) = Im ∫ 0 ∞ d t e i ω t G R ( t ) = ∫ 0 ∞ d t sin ( ω t ) G R ( t ) , (58) \rho(\omega)=\operatorname{Im}\int_0^\infty dt\,e^{i\omega t}G_R(t)=\int_0^\infty dt\,\sin(\omega t)G_R(t), \tag*{(58)} ρ ( ω ) = Im ∫ 0 ∞ d t e iω t G R ( t ) = ∫ 0 ∞ d t sin ( ω t ) G R ( t ) , (58)
and inverting this sine transform gives
G R ( t ) = 2 π ∫ 0 ∞ d ω sin ( ω t ) ρ ( ω ) , t > 0. (59) G_R(t)=\frac{2}{\pi}\int_0^\infty d\omega\,\sin(\omega t)\rho(\omega),\qquad t>0. \tag*{(59)} G R ( t ) = π 2 ∫ 0 ∞ d ω sin ( ω t ) ρ ( ω ) , t > 0. (59)
Using
∫ 0 ∞ d ω ω 4 e i ω τ = 24 i τ 5 , Im τ > 0 , (60) \int_{0}^{\infty} d\omega\,\omega^4 e^{i\omega\tau}=\frac{24i}{\tau^5},\qquad\operatorname{Im}\tau>0, \tag*{(60)} ∫ 0 ∞ d ω ω 4 e iω τ = τ 5 24 i , Im τ > 0 , (60)
the vacuum part ρ v a c = π ω 4 / 32 \rho_{\mathrm{vac}}=\pi\omega^4/32 ρ vac = π ω 4 /32 gives the vacuum term 3 2 t − 5 \frac{3}{2}t^{-5} 2 3 t − 5 , so
G R ( t ) − 3 2 t − 5 = 2 π ∫ 0 ∞ d ω sin ( ω t ) Δ ρ ( ω ) , t > 0. (61) G_R(t)-\frac{3}{2}t^{-5}=\frac{2}{\pi}\int_{0}^{\infty}d\omega\,\sin(\omega t)\,\Delta\rho(\omega),\qquad t>0. \tag*{(61)} G R ( t ) − 2 3 t − 5 = π 2 ∫ 0 ∞ d ω sin ( ω t ) Δ ρ ( ω ) , t > 0. (61)
Since Δ ρ \Delta\rho Δ ρ falls like e − ω Im T 1 e^{-\omega\operatorname{Im}T_1} e − ω Im T 1 , this integral converges and is analytic for ∣ Im t ∣ < Im T 1 |\operatorname{Im}t|<\operatorname{Im}T_1 ∣ Im t ∣ < Im T 1 . Both T 1 T_1 T 1 and T 2 T_2 T 2 lie on the edge of this strip (Im T 1 = Im T 2 \operatorname{Im}T_1=\operatorname{Im}T_2 Im T 1 = Im T 2 ), and the two terms of the spectral tail (51) produce singularities there. Since ρ ~ \widetilde{\rho} ρ in (46) has no power of ω \omega ω in front of e i ω T c e^{i\omega T_c} e iω T c , these terms keep the ω 4 \omega^4 ω 4 of ρ v a c \rho_{\mathrm{vac}} ρ vac ,
Δ ρ ≃ π ω 4 32 Re ∑ c = 1 , 2 A c e i ω T c . (62) \Delta\rho\simeq\frac{\pi\omega^4}{32}\operatorname{Re}\sum_{c=1,2}A_c e^{i\omega T_c}. \tag*{(62)} Δ ρ ≃ 32 π ω 4 Re c = 1 , 2 ∑ A c e iω T c . (62)
Writing sin ( ω t ) \sin(\omega t) sin ( ω t ) and the real part in exponentials, and using that A c A_c A c is real,
sin ( ω t ) Re [ A c e i ω T c ] = A c 4 i ( e i ω ( T c + t ) − e i ω ( T c − t ) + e i ω ( t − T ‾ c ) − e − i ω ( t + T ‾ c ) ) . (63) \sin(\omega t)\operatorname{Re}\left[A_c e^{i\omega T_c}\right]=\frac{A_c}{4i}\left(e^{i\omega(T_c+t)}-e^{i\omega(T_c-t)}+e^{i\omega(t-\overline{T}_c)}-e^{-i\omega(t+\overline{T}_c)}\right). \tag*{(63)} sin ( ω t ) Re [ A c e iω T c ] = 4 i A c ( e iω ( T c + t ) − e iω ( T c − t ) + e iω ( t − T c ) − e − iω ( t + T c ) ) . (63)
Inside the strip, each exponential is e i ω τ e^{i\omega\tau} e iω τ with Im τ > 0 \operatorname{Im}\tau>0 Im τ > 0 , so (60) applies to each of them, and inserting the term of T c T_c T c into (61) gives
C c [ ( t − T c ) − 5 + ( t + T c ) − 5 + ( t − T ‾ c ) − 5 + ( t + T ‾ c ) − 5 ] , C c = 3 A c 8 . (64) C_c\left[(t-T_c)^{-5}+(t+T_c)^{-5}+(t-\overline{T}_c)^{-5}+(t+\overline{T}_c)^{-5}\right],\qquad C_c=\frac{3A_c}{8}. \tag*{(64)} C c [ ( t − T c ) − 5 + ( t + T c ) − 5 + ( t − T c ) − 5 + ( t + T c ) − 5 ] , C c = 8 3 A c . (64)
So T 1 T_1 T 1 and T 2 T_2 T 2 are singularities of G R ( t ) G_R(t) G R ( t ) , where the vacuum singularity 3 2 t − 5 \frac{3}{2}t^{-5} 2 3 t − 5 reappears with the coefficient C c = − 3 4 C_c=-\frac{3}{4} C c = − 4 3 . The straight lines from t = 0 t=0 t = 0 to T 1 T_1 T 1 and T 2 T_2 T 2 lie inside the strip, so both are on the principal sheet, even though R = − q 1 R=-q_1 R = − q 1 and R = − q 2 R=-q_2 R = − q 2 lie beyond the physical singularity.
These singularities determine γ m \gamma_m γ m at large m m m . Expanding around t = 0 t=0 t = 0 ,
( t − T c ) − 5 + ( t + T c ) − 5 = − 2 T c 5 ∑ m ≥ 1 ( 2 m + 3 4 ) ( t T c ) 2 m − 1 , (65) (t-T_c)^{-5}+(t+T_c)^{-5}=-\frac{2}{T_c^5}\sum_{m\geq1}\binom{2m+3}{4}\left(\frac{t}{T_c}\right)^{2m-1}, \tag*{(65)} ( t − T c ) − 5 + ( t + T c ) − 5 = − T c 5 2 m ≥ 1 ∑ ( 4 2 m + 3 ) ( T c t ) 2 m − 1 , (65)
and the terms with T ‾ c \overline{T}_c T c give the complex conjugate. So (64) contributes
− 4 C c ( 2 m + 3 4 ) Re [ T c − 2 m − 4 ] (66) -4C_c\binom{2m+3}{4}\operatorname{Re}\left[T_c^{-2m-4}\right] \tag*{(66)} − 4 C c ( 4 2 m + 3 ) Re [ T c − 2 m − 4 ] (66)
to the coefficient of t 2 m − 1 t^{2m-1} t 2 m − 1 . At large m m m , the coefficients of a power series are given by the sum of such contributions from its singularities. Therefore,
γ m ≃ − 3 2 ( 2 m + 3 4 ) ∑ c = 1 , 2 A c Re [ T c − 2 m − 4 ( 1 + O ( m − 1 ) ) ] . (67) \gamma_m\simeq-\frac{3}{2}\binom{2m+3}{4}\sum_{c=1,2}A_c\operatorname{Re}\left[T_c^{-2m-4}\left(1+O(m^{-1})\right)\right]. \tag*{(67)} γ m ≃ − 2 3 ( 4 2 m + 3 ) c = 1 , 2 ∑ A c Re [ T c − 2 m − 4 ( 1 + O ( m − 1 ) ) ] . (67)
The O ( m − 1 ) O(m^{-1}) O ( m − 1 ) corrections come from the 1 / ω 1/\omega 1/ ω corrections to the tail, which give the weaker singularities ( t − T c ) − 4 (t-T_c)^{-4} ( t − T c ) − 4 , ( t − T c ) − 3 (t-T_c)^{-3} ( t − T c ) − 3 , ⋯ \cdots ⋯ .
Since T 1 = T 2 − B T_1=T_2-B T 1 = T 2 − B , T 1 T_1 T 1 is closer to t = 0 t=0 t = 0 than T 2 T_2 T 2 , so we expect the radius of convergence to be ∣ T 1 ∣ |T_1| ∣ T 1 ∣ . The spectral tail alone does not prove this. The integral (61) converges only inside the strip, but since ∣ T 1 ∣ > I m T 1 |T_1|>\mathop{\rm Im} T_1 ∣ T 1 ∣ > Im T 1 , the disk ∣ t ∣ < ∣ T 1 ∣ |t|<|T_1| ∣ t ∣ < ∣ T 1 ∣ also contains points outside the strip, near the imaginary axis, shown in blue in Figure 5 . The spectral tail does not tell us whether G R ( t ) G_R(t) G R ( t ) is singular there. If it were, the radius of convergence would be smaller than ∣ T 1 ∣ |T_1| ∣ T 1 ∣ . The coefficients γ m \gamma_m γ m , on the other hand, are controlled at large m m m by the singularity or singularities nearest to t = 0 t=0 t = 0 , regardless of their direction in the complex t t t -plane. We therefore locate the nearest singularity directly from γ m \gamma_m γ m , which we have computed up to m = 397 m=397 m = 397 .
Figure 5. The complex t t t -plane for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) . The integral (61) converges in the gray strip ∣ I m t ∣ < I m T 1 |\mathop{\rm Im} t|<\mathop{\rm Im} T_1 ∣ Im t ∣ < Im T 1 , and the spectral tail gives singularities on its edges, at ± T 1 \pm T_1 ± T 1 , ± T ‾ 1 \pm\overline{T}_1 ± T 1 and ± T 2 \pm T_2 ± T 2 , ± T ‾ 2 \pm\overline{T}_2 ± T 2 . The blue circle is ∣ t ∣ = ∣ T 1 ∣ |t|=|T_1| ∣ t ∣ = ∣ T 1 ∣ , and the blue parts of the disk lie outside the strip.
We define
s m ≡ γ m ( 2 m + 3 4 ) . (68) s_m \equiv\frac{\gamma_m}{\binom{2m+3}{4}}. \tag*{(68)} s m ≡ ( 4 2 m + 3 ) γ m . (68)
At large m m m , the nearest singularity dominates. We call it T c T_c T c , without assuming which point it is or what its amplitude is. With z c = T c − 2 z_c=T_c^{-2} z c = T c − 2 , s m s_m s m is then a linear combination of z c m + 2 z_c^{m+2} z c m + 2 and z ‾ c m + 2 \overline{z}_c^{m+2} z c m + 2 with constant coefficients, up to the O ( m − 1 ) O(m^{-1}) O ( m − 1 ) corrections. Since z c z_c z c and z ‾ c \overline{z}_c z c are the roots of x 2 − a x + b x^2-ax+b x 2 − a x + b with a = z c + z ‾ c a=z_c+\overline{z}_c a = z c + z c and b = ∣ z c ∣ 2 b=|z_c|^2 b = ∣ z c ∣ 2 , the sequence s m s_m s m satisfies
s m + 1 = a s m − b s m − 1 . (69) s_{m+1}=a s_m-b s_{m-1}. \tag*{(69)} s m + 1 = a s m − b s m − 1 . (69)
We treat a a a and b b b as unknowns. Imposing (69) at two consecutive values of m m m fixes a , b a,b a , b from four consecutive coefficients, and the nearest singularity follows from
∣ T c ∣ = b − 1 / 4 , cos ( 2 arg T c ) = a 2 b , 0 < arg T c < π 2 . (70) |T_c| = b^{-1/4}, \qquad\cos(2\arg T_c) = \frac{a}{2\sqrt{b}}, \qquad0 < \arg T_c < \frac{\pi}{2}. \tag*{(70)} ∣ T c ∣ = b − 1/4 , cos ( 2 arg T c ) = 2 b a , 0 < arg T c < 2 π . (70)
Because of the O ( m − 1 ) O(m^{-1}) O ( m − 1 ) corrections, (69) holds only approximately. Even so, at m = 395 m=395 m = 395 the estimate agrees with T 1 T_1 T 1 within 2 × 10 − 6 2 \times10^{-6} 2 × 1 0 − 6 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and 5 × 10 − 8 5 \times10^{-8} 5 × 1 0 − 8 for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , as shown in Table 1 . The difference is much smaller than the distance B = 0.105 B=0.105 B = 0.105 and 0.167 0.167 0.167 between T 1 T_1 T 1 and T 2 T_2 T 2 . The optical times T 1 , 2 , 3 T_{1,2,3} T 1 , 2 , 3 do not enter this estimate, so γ m \gamma_m γ m alone locates the nearest singularity at T 1 T_1 T 1 .
Table 1. The nearest singularity of G R ( t ) G_R(t) G R ( t ) , located from γ m \gamma_m γ m alone by (69) at m = 395 m=395 m = 395 , compared with the optical time T 1 T_1 T 1 . ∣ Δ T c ∣ |\Delta T_c| ∣Δ T c ∣ is the distance between the two. For comparison, T 2 = 0.75970 + 0.97488 i T_2=0.75970+0.97488\,i T 2 = 0.75970 + 0.97488 i for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and T 2 = 0.95882 + 1.19472 i T_2=0.95882+1.19472\,i T 2 = 0.95882 + 1.19472 i for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) .
Conversely, we fix T 1 T_1 T 1 and T 2 T_2 T 2 to the optical times and test the amplitudes A c A_c A c in (67) . Fitting A 1 A_1 A 1 , A 2 A_2 A 2 and the 1 / m 1/m 1/ m corrections, we obtain A 1 = − 2 A_1=-2 A 1 = − 2 within 2 × 10 − 14 2 \times10^{-14} 2 × 1 0 − 14 and A 2 = − 2 A_2=-2 A 2 = − 2 within 2 × 10 − 3 2 \times10^{-3} 2 × 1 0 − 3 in both examples.
Therefore, the OPE convergence radius is controlled by T 1 T_1 T 1 . It is ∣ T 1 ∣ = 1.1745 |T_1|=1.1745 ∣ T 1 ∣ = 1.1745 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and 1.4332 1.4332 1.4332 for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , smaller than ∣ T 3 ∣ = 1.3030 |T_3|=1.3030 ∣ T 3 ∣ = 1.3030 and 1.7642 1.7642 1.7642 of the physical singularity.
QNM lines
The QNMs are the solutions of A ( ω ) = 0 \mathcal{A}(\omega)=0 A ( ω ) = 0 . If ω \omega ω is a QNM, so is − ω ˉ -\bar{\omega} − ω ˉ , so we consider Re ω ≥ 0 \operatorname{Re}\omega\ge0 Re ω ≥ 0 and write ω = ∣ ω ∣ e − i θ \omega=|\omega|e^{-i\theta} ω = ∣ ω ∣ e − i θ with 0 < θ ≤ π / 2 0<\theta\le\pi/2 0 < θ ≤ π /2 .
For this ω \omega ω , e i ω ( T − T c ) e^{i\omega(T-T_c)} e iω ( T − T c ) is real and larger than 1 on the ray T = T c − i r e i θ T=T_c-ire^{i\theta} T = T c − i r e i θ with r > 0 r>0 r > 0 . This is the Stokes line of T c T_c T c , which is the vertical line of § 3.3 tilted to the right by the angle θ \theta θ , as in Figure 2 (b). For 0 < θ < π / 2 0<\theta<\pi/2 0 < θ < π /2 , the Stokes lines of T 1 T_1 T 1 , T 2 T_2 T 2 and T 3 T_3 T 3 all reach the real T T T axis. In particular, the Stokes line of T 3 T_3 T 3 no longer runs along the vertical edge, so T 3 T_3 T 3 is not shadowed. The calculation of § 3.3 then gives
A ∝ 1 − ρ ~ = 1 + 2 ∑ c = 1 3 cos ( π ν c ) e i ω T c , (71) \mathcal{A} \propto1-\tilde{\rho}=1+2\sum_{c=1}^{3}\cos(\pi\nu_c)e^{i\omega T_c}, \tag*{(71)} A ∝ 1 − ρ ~ = 1 + 2 c = 1 ∑ 3 cos ( π ν c ) e iω T c , (71)
from (46) and (47) . Since
∣ e i ω T c ∣ = ∣ exp ( i ∣ ω ∣ e − i θ ∣ T c ∣ e i arg T c ) ∣ = exp ( − ∣ ω ∣ ∣ T c ∣ sin ( arg T c − θ ) ) , (72) \left|e^{i\omega T_c}\right|=\left|\exp\left(i|\omega|e^{-i\theta}|T_c|e^{i\arg T_c}\right)\right|
=\exp\left(-|\omega||T_c|\sin(\arg T_c-\theta)\right), \tag*{(72)} e iω T c = exp ( i ∣ ω ∣ e − i θ ∣ T c ∣ e i a r g T c ) = exp ( − ∣ ω ∣∣ T c ∣ sin ( arg T c − θ ) ) , (72)
e i ω T c e^{i\omega T_c} e iω T c is exponentially small for θ < arg T c \theta<\arg T_c θ < arg T c and exponentially large for θ > arg T c \theta>\arg T_c θ > arg T c . Also, it has absolute value 1 at θ = arg T c \theta=\arg T_c θ = arg T c .
So, as θ \theta θ grows from 0, A = 0 \mathcal{A}=0 A = 0 first becomes possible at the smallest arg T c \arg T_c arg T c . Since T 1 = T 2 − B T_1=T_2-B T 1 = T 2 − B and T 3 = T 2 + i G T_3=T_2+iG T 3 = T 2 + i G , this is always arg T 2 \arg T_2 arg T 2 . At θ = arg T 2 \theta=\arg T_2 θ = arg T 2 , the terms of T 1 T_1 T 1 and T 3 T_3 T 3 are exponentially small and the term of T 2 T_2 T 2 has absolute value 1. With ν 2 = 1 3 \nu_2=\frac{1}{3} ν 2 = 3 1 , A = 0 \mathcal{A}=0 A = 0 becomes 1 + e i ω T 2 = 0 1+e^{i\omega T_2}=0 1 + e iω T 2 = 0 , so
ω ∼ ( 2 k + 1 ) π T 2 (73) \omega\sim\frac{(2k+1)\pi}{T_2} \tag*{(73)} ω ∼ T 2 ( 2 k + 1 ) π (73)
for large integers k k k . This is a line of QNMs along the direction of 1 / T 2 1/T_2 1/ T 2 , that is, at arg ω = − arg T 2 \arg\omega=-\arg T_2 arg ω = − arg T 2 , where arg T 2 \arg T_2 arg T 2 is the angle of the red line in Figure 1 . For arg T 2 < θ < π / 2 \arg T_2<\theta<\pi/2 arg T 2 < θ < π /2 , the term of T 2 T_2 T 2 in (71) is exponentially larger than all the other terms, including 1, so we expect no QNM there. In particular, there is no QNM line along 1 / T 1 1/T_1 1/ T 1 or 1 / T 3 1/T_3 1/ T 3 .
We compute the QNMs with ∣ ω ∣ < 125 |\omega|<125 ∣ ω ∣ < 125 as the zeros of J ( ω ) = A ( ω ) Γ ( 1 − i ω β / 2 π ) J(\omega)=\frac{A(\omega)}{\Gamma(1-i\omega\beta/2\pi)} J ( ω ) = Γ ( 1 − iω β /2 π ) A ( ω ) of (A.39), following the method explained in Appendix A.4. Figure 6 shows the numerical results. The complex QNMs approach the direction − arg T 2 -\arg T_2 − arg T 2 , which is − 52.07 ∘ -52.07^\circ − 52.0 7 ∘ for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and − 51.25 ∘ -51.25^\circ − 51.2 5 ∘ for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) . The QNM with the largest ∣ ω ∣ |\omega| ∣ ω ∣ in Figure 6 is ω = 76.5170 − 98.2487 i \omega=76.5170-98.2487i ω = 76.5170 − 98.2487 i for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and ω = 75.7098 − 94.3654 i \omega=75.7098-94.3654i ω = 75.7098 − 94.3654 i for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , which give ω T 2 / π = 48.991 − 0.014 i \omega T_2/\pi=48.991-0.014i ω T 2 / π = 48.991 − 0.014 i and 58.993 − 0.009 i 58.993-0.009i 58.993 − 0.009 i . These are close to the odd integers 49 and 59, that is, to (73) with k = 24 k=24 k = 24 and k = 29 k=29 k = 29 . For these QNMs, the distance ∣ ω T 2 / π − ( 2 k + 1 ) ∣ |\omega T_2/\pi-(2k+1)| ∣ ω T 2 / π − ( 2 k + 1 ) ∣ is 0.017 and 0.011. For the complex QNMs with ∣ ω ∣ > 30 |\omega|>30 ∣ ω ∣ > 30 , this distance decreases as ∣ ω ∣ |\omega| ∣ ω ∣ grows.
Figure 6. arg ω \operatorname{arg}\omega arg ω against ∣ ω ∣ |\omega| ∣ ω ∣ for the QNMs with Re ω ≥ 0 \operatorname{Re}\omega\ge0 Re ω ≥ 0 . The filled circles are the complex QNMs and the open circles are the QNMs on the negative imaginary axis. The horizontal lines are the directions − arg T c -\operatorname{arg}T_c − arg T c . There is no zero of J ( ω ) J(\omega) J ( ω ) in the shaded regions.
To check that no QNM is missed, we count the zeros of J J J in a region Ω \Omega Ω of the complex ω \omega ω -plane with the argument principle,
N Ω = 1 2 π i ∮ ∂ Ω J ′ ( ω ) J ( ω ) d ω . (74) N_\Omega=\frac{1}{2\pi i}\oint_{\partial\Omega}\frac{J'(\omega)}{J(\omega)}\,d\omega. \tag*{(74)} N Ω = 2 π i 1 ∮ ∂ Ω J ( ω ) J ′ ( ω ) d ω . (74)
In general, N Ω N_\Omega N Ω is the number of zeros of J J J in Ω \Omega Ω minus the number of poles of J J J in Ω \Omega Ω . Since J J J has no poles, N Ω N_\Omega N Ω is the number of zeros of J J J in Ω \Omega Ω , that is, the number of QNMs in Ω \Omega Ω . We take regions r 1 < ∣ ω ∣ < r 2 r_1<|\omega|<r_2 r 1 < ∣ ω ∣ < r 2 , α 1 < arg ω < α 2 \alpha_1<\arg\omega<\alpha_2 α 1 < arg ω < α 2 , which together cover the disk ∣ ω ∣ < 125 |\omega|<125 ∣ ω ∣ < 125 .
For ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , there is no zero in − 88 ∘ < arg ω < − 53 ∘ -88^\circ< \operatorname{arg}\omega< -53^\circ − 8 8 ∘ < arg ω < − 5 3 ∘ , which contains − arg T 1 = − 56.10 ∘ -\operatorname{arg}T_1=-56.10^\circ − arg T 1 = − 56.1 0 ∘ and − arg T 3 = − 54.33 ∘ -\operatorname{arg}T_3=-54.33^\circ − arg T 3 = − 54.3 3 ∘ . The region − 53 ∘ < arg ω < − 49.5 ∘ -53^\circ< \operatorname{arg}\omega< -49.5^\circ − 5 3 ∘ < arg ω < − 49. 5 ∘ around − arg T 2 -\operatorname{arg}T_2 − arg T 2 contains 24 zeros, and − 92 ∘ < arg ω < − 88 ∘ -92^\circ< \operatorname{arg}\omega< -88^\circ − 9 2 ∘ < arg ω < − 8 8 ∘ contains 31 zeros.
For ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , there is again no zero in − 88 ∘ < arg ω < − 53 ∘ -88^\circ< \operatorname{arg}\omega< -53^\circ − 8 8 ∘ < arg ω < − 5 3 ∘ , which contains − arg T 1 = − 56.47 ∘ -\operatorname{arg}T_1=-56.47^\circ − arg T 1 = − 56.4 7 ∘ and − arg T 3 = − 57.08 ∘ -\operatorname{arg}T_3=-57.08^\circ − arg T 3 = − 57.0 8 ∘ . The region − 53 ∘ < arg ω < − 48 ∘ -53^\circ< \operatorname{arg}\omega< -48^\circ − 5 3 ∘ < arg ω < − 4 8 ∘ around − arg T 2 -\operatorname{arg}T_2 − arg T 2 contains 29 zeros, and − 92 ∘ < arg ω < − 88 ∘ -92^\circ< \operatorname{arg}\omega< -88^\circ − 9 2 ∘ < arg ω < − 8 8 ∘ contains 12 zeros.
In both examples, there is also no zero between the line of QNMs and the positive real axis, and no zero in the upper half-plane. Each count equals the number of QNMs that we found in the same region, so no QNM is missed.
We now turn to the QNMs on the negative imaginary axis, ω = − i y \omega=-iy ω = − i y with y > 0 y>0 y > 0 , shown as open red circles in Figure 6 . For ω = − i y \omega=-iy ω = − i y , the term of T 3 T_3 T 3 in (71) has the same size as the term of T 2 T_2 T 2 , since ∣ e i ω ( T 3 − T 2 ) ∣ = ∣ e − ω G ∣ = 1 \lvert e^{i\omega(T_3-T_2)}\rvert=\lvert e^{-\omega G}\rvert=1 ∣ e iω ( T 3 − T 2 ) ∣ = ∣ e − ω G ∣ = 1 . The calculation of § 3.3 does not apply here. There, ϕ i n \phi_{\mathrm{in}} ϕ in is followed from the horizon to the boundary along the real T T T axis, which is the image of R > 1 R>1 R > 1 , and a reflected wave is added at each point where the real T T T axis crosses a Stokes line. For θ = π / 2 \theta=\pi/2 θ = π /2 , the Stokes line of each T c T_c T c is the horizontal ray T = T c + r T=T_c+r T = T c + r with r > 0 r>0 r > 0 , as in Figure 2 (c), and none of these rays crosses the real T T T axis, so ϕ i n \phi_{\mathrm{in}} ϕ in picks up no reflected wave. Instead, we use the fact that A = 0 \mathcal{A}=0 A = 0 means ϕ i n ∝ ϕ B \phi_{\mathrm{in}}\propto\phi_B ϕ in ∝ ϕ B , and compare the two solutions on the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 , which is the image of R − < R < 1 R_-<R<1 R − < R < 1 . Since the wave equation (20) has real coefficients for ω = − i y \omega=-iy ω = − i y , ϕ i n \phi_{\mathrm{in}} ϕ in and ϕ B \phi_B ϕ B can both be taken real for R > 1 R>1 R > 1 , so ϕ i n ∝ ϕ B \phi_{\mathrm{in}}\propto\phi_B ϕ in ∝ ϕ B can hold only with a real constant. Two solutions that are proportional for R > 1 R>1 R > 1 stay proportional, with the same constant, for all R R R and T T T . Therefore, on the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 , ϕ B \phi_B ϕ B must have the same phase as ϕ i n \phi_{\mathrm{in}} ϕ in .
We first find the phase of ϕ i n \phi_{\mathrm{in}} ϕ in . Near the horizon, ϕ i n = ϵ s ( 1 + a 1 ϵ + ⋯ ) \phi_{\mathrm{in}}=\epsilon^s(1+a_1\epsilon+\cdots) ϕ in = ϵ s ( 1 + a 1 ϵ + ⋯ ) with ϵ = R − 1 \epsilon=R-1 ϵ = R − 1 , as in Appendix § 9 . For ω = − i y \omega=-iy ω = − i y , the exponent s = − i ω β 4 π = − y β 4 π s=-\frac{i\omega\beta}{4\pi}=-\frac{y\beta}{4\pi} s = − 4 π iω β = − 4 π y β and the coefficients a n a_n a n are real, so ϕ i n \phi_{\mathrm{in}} ϕ in is real for R > 1 R>1 R > 1 . To reach R < 1 R<1 R < 1 , we pass R = 1 R=1 R = 1 through the lower half R R R -plane, as for the optical times. Then ϵ = ∣ ϵ ∣ e − i π \epsilon=\lvert\epsilon\rvert e^{-i\pi} ϵ = ∣ ϵ ∣ e − iπ , and ϵ s = ∣ ϵ ∣ s e i y β / 4 \epsilon^s=\lvert\epsilon\rvert^s e^{iy\beta/4} ϵ s = ∣ ϵ ∣ s e i y β /4 acquires a phase. Since the equation is real, this phase stays the same for all R − < R < 1 R_-<R<1 R − < R < 1 , so on the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 , i.e., for R − < R < 1 R_-<R<1 R − < R < 1 , we have
ϕ i n = e i y β / 4 × ( real ) . (75) \phi_{\mathrm{in}}=e^{iy\beta/4}\times(\text{real}). \tag*{(75)} ϕ in = e i y β /4 × ( real ) . (75)
Next, we bring ϕ B \phi_B ϕ B to the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 . For ω = − i y \omega=-iy ω = − i y , the WKB waves are f ± = P − 1 / 4 e ± y T / 2 f_\pm=P^{-1/4}e^{\pm yT/2} f ± = P − 1/4 e ± y T /2 . Near the boundary, ϕ B ∝ z 2 I 2 ( y z ) \phi_B\propto z^2I_2(yz) ϕ B ∝ z 2 I 2 ( y z ) , which grows inward, so ϕ B ∝ f + \phi_B\propto f_+ ϕ B ∝ f + on the real T T T axis, since T T T increases as R R R decreases from the boundary. We follow it along the path of Figure 2 (c): along the real T T T axis to T = p T=p T = p , and then straight up to T = p + i β / 2 T=p+i\beta/2 T = p + i β /2 . We take p > Re T 2 p>\operatorname{Re}T_2 p > Re T 2 , so that the path crosses none of the black lines in Figure 2 (c), which are the image of the real R R R axis. The path then reaches R − < R < 1 R_-<R<1 R − < R < 1 through the lower half R R R -plane, as for ϕ i n \phi_{\mathrm{in}} ϕ in in (6.5). On the way up, the path crosses the Stokes lines of T 2 T_2 T 2 and T 3 T_3 T 3 . Each crossing adds a reflected wave as in (39) , with the opposite sign since the path goes around T c T_c T c in the opposite direction:
ϕ B ∝ f + + 2 i ∑ c = 2 , 3 cos ( π ν c ) e − y T c f − . (76) \phi_B\propto f_+ + 2i\sum_{c=2,3}\cos(\pi\nu_c)e^{-yT_c}f_-. \tag*{(76)} ϕ B ∝ f + + 2 i c = 2 , 3 ∑ cos ( π ν c ) e − y T c f − . (76)
On the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 , T = Re T + i β / 2 T=\operatorname{Re}T+i\beta/2 T = Re T + i β /2 and P > 0 P>0 P > 0 , so f ± = e ± i y β / 4 × ( real ) f_\pm=e^{\pm iy\beta/4}\times(\text{real}) f ± = e ± i y β /4 × ( real ) . f + f_+ f + already acquires the same phase as (6.5), and hence ϕ i n ∝ ϕ B \phi_{\mathrm{in}}\propto\phi_B ϕ in ∝ ϕ B requires the second term of (6.6) to have the same phase. Since f − f_{-} f − acquires the opposite phase e − i y β / 4 e^{-iy\beta/4} e − i y β /4 , its coefficient must be e i y β / 2 e^{iy\beta/2} e i y β /2 times a real number. Since Re T 2 = Re T 3 \operatorname{Re} T_{2}=\operatorname{Re} T_{3} Re T 2 = Re T 3 and
e y T c = e y Re T 2 e i y Im T c = e y Re T 2 e i y β / 2 e − i h c y , ( c = 2 , 3 ) , (77) e^{yT_{c}}=e^{y\operatorname{Re} T_{2}}e^{iy\operatorname{Im} T_{c}}
= e^{y\operatorname{Re} T_{2}}e^{iy\beta/2}e^{-ih_{c}y},\qquad(c=2,3), \tag*{(77)} e y T c = e y Re T 2 e i y Im T c = e y Re T 2 e i y β /2 e − i h c y , ( c = 2 , 3 ) , (77)
with h c : = β 2 − Im T c h_{c}:=\frac{\beta}{2}-\operatorname{Im} T_{c} h c := 2 β − Im T c , we have to require
i ∑ c = 2 , 3 cos ( π ν c ) e − i h c y ∈ R , (78) i\sum_{c=2,3}\cos(\pi\nu_{c})e^{-ih_{c}y}\in\mathbb{R}, \tag*{(78)} i c = 2 , 3 ∑ cos ( π ν c ) e − i h c y ∈ R , (78)
that is,
E ( y ) : = ∑ c = 2 , 3 cos ( π ν c ) cos ( h c y ) = 0. (79) E(y):=\sum_{c=2,3}\cos(\pi\nu_{c})\cos(h_{c}y)=0. \tag*{(79)} E ( y ) := c = 2 , 3 ∑ cos ( π ν c ) cos ( h c y ) = 0. (79)
For q 3 > 0 q_{3}>0 q 3 > 0 , both R = − q 2 R=-q_{2} R = − q 2 and R = − q 3 R=-q_{3} R = − q 3 are simple zeros of P P P , so ν 2 = ν 3 = 1 3 \nu_{2}=\nu_{3}=\frac{1}{3} ν 2 = ν 3 = 3 1 and we obtain
E ( y ) ∝ cos h 2 y + cos h 3 y = 2 cos ( h 2 + h 3 ) y 2 cos G y 2 , (80) E(y)\propto\cos h_{2}y+\cos h_{3}y=2\cos\frac{(h_{2}+h_{3})y}{2}\cos\frac{Gy}{2}, \tag*{(80)} E ( y ) ∝ cos h 2 y + cos h 3 y = 2 cos 2 ( h 2 + h 3 ) y cos 2 G y , (80)
where we used h 2 − h 3 = Im ( T 3 − T 2 ) = G h_{2}-h_{3}=\operatorname{Im}(T_{3}-T_{2})=G h 2 − h 3 = Im ( T 3 − T 2 ) = G . This gives two families of QNMs on the negative imaginary axis,
y ≃ ( 2 k + 1 ) π h 2 + h 3 , y ≃ ( 2 k + 1 ) π G . (81) y\simeq\frac{(2k+1)\pi}{h_{2}+h_{3}},\qquad y\simeq\frac{(2k+1)\pi}{G}. \tag*{(81)} y ≃ h 2 + h 3 ( 2 k + 1 ) π , y ≃ G ( 2 k + 1 ) π . (81)
For q 3 = 0 q_{3}=0 q 3 = 0 , T 3 T_{3} T 3 is the Kasner singularity. Near R = 0 R=0 R = 0 , both P ∝ R P\propto R P ∝ R and D ∝ R D\propto R D ∝ R vanish, so x x x diverges logarithmically and P P P vanishes exponentially in x x x . The local solutions are Bessel functions of order zero, so ν 3 = 0 \nu_{3}=0 ν 3 = 0 . Moreover, β − = 0 \beta_{-}=0 β − = 0 gives h 3 = 0 h_{3}=0 h 3 = 0 and h 2 = G h_{2}=G h 2 = G , so T 3 T_{3} T 3 lies on the line Im T = β / 2 \operatorname{Im}T=\beta/2 Im T = β /2 itself. For ω = − i y \omega=-iy ω = − i y , the Stokes line of T 3 T_{3} T 3 is the horizontal ray to its right, so our path ends on the Stokes line of T 3 T_{3} T 3 instead of crossing it, as in Figure 2 (d). On the Stokes line itself, the Stokes multiplier of the reflected component is one half of its fully switched-on value [60 ] , so the c = 3 c=3 c = 3 term of (79) has an extra factor 1 2 \frac{1}{2} 2 1 . With cos ( π ν 2 ) = 1 2 \cos(\pi\nu_{2})=\frac{1}{2} cos ( π ν 2 ) = 2 1 and cos ( π ν 3 ) = 1 \cos(\pi\nu_{3})=1 cos ( π ν 3 ) = 1 , we obtain
E ( y ) ∝ cos G y + 1 = 2 cos 2 G y 2 , (82) E(y)\propto\cos Gy+1=2\cos^{2}\frac{Gy}{2}, \tag*{(82)} E ( y ) ∝ cos G y + 1 = 2 cos 2 2 G y , (82)
which has double zeros at y = ( 2 k + 1 ) π / G y=(2k+1)\pi/G y = ( 2 k + 1 ) π / G .
We now include the effect of T 1 T_{1} T 1 and determine how these double zeros split. For ω = − i y \omega=-iy ω = − i y , the Stokes line of T 1 T_{1} T 1 runs along the horizontal edge into T 2 T_{2} T 2 , so T 1 T_{1} T 1 is shadowed, as T 3 T_{3} T 3 is for real ω \omega ω . Since ω ( T 1 − T 2 ) = i y B \omega(T_{1}-T_{2})=iyB ω ( T 1 − T 2 ) = i y B has the same form as ω ( T 3 − T 2 ) = i ω G \omega(T_{3}-T_{2})=i\omega G ω ( T 3 − T 2 ) = iω G for real ω \omega ω of § 4 , the term of T 2 T_{2} T 2 is multiplied by W ( y B ) W(yB) W ( y B ) as in (54) . Using ∣ Γ ( 1 2 − i g 2 π ) ∣ 2 = π / cosh ( g / 2 ) \left|\Gamma\left(\frac{1}{2}-\frac{ig}{2\pi}\right)\right|^{2}=\pi/\cosh(g/2) Γ ( 2 1 − 2 π i g ) 2 = π / cosh ( g /2 ) , we have ∣ W ( g ) ∣ 2 = 1 + e − g \lvert W(g)\rvert^{2}=1+e^{-g} ∣ W ( g ) ∣ 2 = 1 + e − g from (55) , and
E ( y ) ∝ 1 + ∣ W ( y B ) ∣ cos ( G y + arg W ) , (83) E(y)\propto1+\lvert W(yB)\rvert\cos(Gy+\arg W), \tag*{(83)} E ( y ) ∝ 1 + ∣ W ( y B )∣ cos ( G y + arg W ) , (83)
with ∣ W ( y B ) ∣ = 1 + e − y B |W(yB)| = \sqrt{1 + e^{-yB}} ∣ W ( y B ) ∣ = 1 + e − y B . The phase of W W W shifts the center of each pair of zeros. At large y y y , its variation is subleading, while the splitting is controlled at leading order by the excess ∣ W ∣ − 1 |W|-1 ∣ W ∣ − 1 . The minima of cos G y + 1 \cos Gy+1 cos G y + 1 only touch zero, and the slightly larger amplitude ∣ W ∣ > 1 |W|>1 ∣ W ∣ > 1 of the cosine pushes them below zero. Near y 0 = ( 2 k + 1 ) π / G y_0=(2k+1)\pi/G y 0 = ( 2 k + 1 ) π / G , we have cos G y ≃ − 1 + 1 2 G 2 ( y − y 0 ) 2 \cos Gy \simeq-1+\frac{1}{2}G^2(y-y_0)^2 cos G y ≃ − 1 + 2 1 G 2 ( y − y 0 ) 2 and ∣ W ∣ ≃ 1 + 1 2 e − B y 0 |W| \simeq1+\frac{1}{2}e^{-By_0} ∣ W ∣ ≃ 1 + 2 1 e − B y 0 , so
E ∝ − 1 2 e − B y 0 + 1 2 G 2 ( y − y 0 ) 2 , (84) E \propto-\frac{1}{2}e^{-By_0}+\frac{1}{2}G^2(y-y_0)^2, \tag*{(84)} E ∝ − 2 1 e − B y 0 + 2 1 G 2 ( y − y 0 ) 2 , (84)
and each double zero splits into two simple zeros separated by
Δ y ≃ 2 G e − B y / 2 . (85) \Delta y \simeq\frac{2}{G}e^{-By/2}. \tag*{(85)} Δ y ≃ G 2 e − B y /2 . (85)
Figure 7 compares the QNMs on the negative imaginary axis with the zeros of E ( y ) E(y) E ( y ) . For ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , where h 2 = 0.7633 h_2=0.7633 h 2 = 0.7633 and h 3 = 0.6796 h_3=0.6796 h 3 = 0.6796 , the 31 QNMs with y < 125 y<125 y < 125 correspond one to one to the 31 zeros of E ( y ) E(y) E ( y ) . For the first family of (81) , the difference is below 0.013 for y > 40 y > 40 y > 40 . The second family is set by the distance G G G between T 2 T_2 T 2 and T 3 T_3 T 3 . It predicts y = 37.54 y = 37.54 y = 37.54 and 112.61 112.61 112.61 , and the QNMs are at y = 37.49 y = 37.49 y = 37.49 and 112.53 112.53 112.53 .
Figure 7. QNMs on the negative imaginary axis, ω = − i y \omega=-iy ω = − i y , compared with E ( y ) E(y) E ( y ) . (a) ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) , with the envelope ± 2 cos ( G y / 2 ) \pm2\cos(Gy/2) ± 2 cos ( G y /2 ) . (b) ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , with the splitting Δ y \Delta y Δ y of each pair.
For ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) , the 12 QNMs form six pairs, one around each double zero y = ( 2 k + 1 ) π / G y = (2k + 1)\pi/G y = ( 2 k + 1 ) π / G , which are y = 10.98 y = 10.98 y = 10.98 , 32.93 32.93 32.93 , 54.88 54.88 54.88 , 76.83 76.83 76.83 , 98.78 98.78 98.78 and 120.73 120.73 120.73 . The pairs are split by Δ y = 3.10 \Delta y = 3.10 Δ y = 3.10 , 0.475 0.475 0.475 , 0.0738 0.0738 0.0738 , 0.0117 0.0117 0.0117 , 0.00185 0.00185 0.00185 and 0.000294 0.000294 0.000294 . These are 1.11, 1.07, 1.04, 1.03, 1.02 and 1.02 times (85) , so the ratio approaches 1 as y y y grows.
Therefore, the complex QNMs are controlled by T 2 T_2 T 2 , and the QNMs on the negative imaginary axis are controlled by T 2 T_2 T 2 and T 3 T_3 T 3 . The physical singularity T 3 T_3 T 3 appears at leading order only in the latter. T 1 T_1 T 1 has no QNM line of its own. For q 3 = 0 q_3 = 0 q 3 = 0 , it appears instead through the exponentially small splitting of the double zeros on the negative imaginary axis.
Conclusion and discussions
We studied the retarded correlator of T x y T_{xy} T x y at zero spatial momentum in the STU black branes with unequal R-charges. Their metric has three singular points at R = − q i R = -q_i R = − q i , and only R = − q 3 R = -q_3 R = − q 3 is the physical singularity. The optical times form a right triangle with T 1 = T 2 − B T_1 = T_2 - B T 1 = T 2 − B and T 3 = T 2 + i G T_3 = T_2 + iG T 3 = T 2 + i G , and the three observables detect different points of this triangle.
The spectral function at large real ω \omega ω is controlled by T 1 T_1 T 1 and T 2 T_2 T 2 , which have the smallest Im T i \operatorname{Im} T_i Im T i . Their two reflection terms have the same amplitude A 1 = A 2 = − 2 A_1 = A_2 = -2 A 1 = A 2 = − 2 in (51) , and they beat with the period 2 π / B 2\pi/B 2 π / B . The physical singularity T 3 T_3 T 3 is shadowed by T 2 T_2 T 2 , and it contributes only through the additional factor W ( ω G ) W(\omega G) W ( ω G ) of the T 2 T_2 T 2 reflection term.
The OPE converges up to ∣ T 1 ∣ |T_1| ∣ T 1 ∣ , the smallest ∣ T i ∣ |T_i| ∣ T i ∣ , which is smaller than ∣ T 3 ∣ |T_3| ∣ T 3 ∣ of the physical singularity. The OPE coefficients alone locate the nearest singularity at T 1 T_1 T 1 , without using the optical times. Although R = − q 1 R = -q_1 R = − q 1 lies beyond the physical singularity, T 1 T_1 T 1 is on the principal sheet of G R ( t ) G_R(t) G R ( t ) .
The complex QNMs lie along the direction of 1 / T 2 1/T_2 1/ T 2 , since T 2 T_2 T 2 has the smallest arg T i \arg T_i arg T i . The QNMs on the negative imaginary axis ω = − i y \omega= -iy ω = − i y are set by T 2 T_2 T 2 and T 3 T_3 T 3 , so the physical singularity appears at leading order only there. For q 3 = 0 q_3 = 0 q 3 = 0 , the physical singularity is a Kasner singularity, and its optical time T 3 T_3 T 3 lies on the line Im T = β / 2 \operatorname{Im} T = \beta/2 Im T = β /2 . The QNMs on the negative imaginary axis then come in pairs near the double zeros of E ( y ) E(y) E ( y ) (82) , and T 1 T_1 T 1 splits each pair by an exponentially small amount Δ y \Delta y Δ y (85) .
Conversely, the spectral function, the OPE coefficients, and the QNMs are all data of the dual CFT. From them, one can read off the positions of several singular points of the geometry, including those beyond the physical singularity, which are not part of the spacetime. One can even read off the order of each zero of P P P . The amplitude A c = − 4 cos ( π ν c ) A_c = -4\cos(\pi\nu_c) A c = − 4 cos ( π ν c ) with ν c = 1 / ( ℓ + 2 ) \nu_c = 1/(\ell+2) ν c = 1/ ( ℓ + 2 ) fixes the order ℓ \ell ℓ , and our fits of the spectral function and of the OPE coefficients recover A 1 = A 2 = − 2 A_1=A_2=-2 A 1 = A 2 = − 2 , that is, simple zeros. The QNM line carries the same information. If R = − q 2 R=-q_2 R = − q 2 were a zero of another order, the condition 1 + 2 cos ( π ν 2 ) e i ω T 2 = 0 1+2\cos(\pi\nu_2)e^{i\omega T_2}=0 1 + 2 cos ( π ν 2 ) e iω T 2 = 0 would give
ω T 2 = ( 2 k + 1 ) π + i ln ( 2 cos ( π ν 2 ) ) , \omega T_2=(2k+1)\pi+i\ln\left(2\cos(\pi\nu_2)\right), ω T 2 = ( 2 k + 1 ) π + i ln ( 2 cos ( π ν 2 ) ) ,
so the offset of ω T 2 / π \omega T_2/\pi ω T 2 / π from the odd integers measures ν 2 \nu_2 ν 2 . For a simple zero, 2 cos ( π ν 2 ) = 1 2\cos(\pi\nu_2)=1 2 cos ( π ν 2 ) = 1 and the offset vanishes, consistent with § 6 .
There are several interesting open directions. The singular points beyond the physical singularity are not part of the spacetime, but they control the boundary data. It would be interesting to understand what they correspond to in the dual field theory, and how their signatures change at finite N N N , where the bouncing singularities are smoothed out [9 , 10 , 11 , 12 ] . The charged black holes in A d S 4 \mathrm{AdS}_4 AdS 4 and A d S 7 \mathrm{AdS}_7 AdS 7 carry four and two charges, respectively [43 ] , and their metrics also have several singular points, so the same question can be asked there. Finally, other channels, such as the R-current correlators, involve coupled fluctuations at finite density, and they may detect the singular points differently.
Acknowledgements
I thank Seok Kim for helpful discussions. This work is supported in part by the NRF grant RS-2026-25477643. I used Claude Code for numerical calculations and figures.
Computing A ( ω ) \mathcal{A}(\omega) A ( ω )
In this section, we explain how to compute A ( ω ) \mathcal{A}(\omega) A ( ω ) for a given ω \omega ω . We need it for the spectral function through (26) , and the QNMs are its zeros. The difficulty is that ϕ i n \phi_{\mathrm{in}} ϕ in is defined by its behavior at the horizon, while A \mathcal{A} A is read off at the boundary, and both are singular points of the wave equation.
Wronskian
For two solutions ϕ 1 , ϕ 2 \phi_1,\phi_2 ϕ 1 , ϕ 2 of the wave equation (20) , we define the Wronskian
W [ ϕ 1 , ϕ 2 ] = ϕ 1 ∂ R ϕ 2 − ϕ 2 ∂ R ϕ 1 . (86) W[\phi_1,\phi_2]=\phi_1\partial_R\phi_2-\phi_2\partial_R\phi_1. \tag*{(86)} W [ ϕ 1 , ϕ 2 ] = ϕ 1 ∂ R ϕ 2 − ϕ 2 ∂ R ϕ 1 . (86)
Then, D W DW D W is independent of R R R :
∂ R ( D W ) = ϕ 1 ∂ R ( D ∂ R ϕ 2 ) − ϕ 2 ∂ R ( D ∂ R ϕ 1 ) = − ω 2 P 4 D ( ϕ 1 ϕ 2 − ϕ 2 ϕ 1 ) = 0. (87) \partial_R(DW)=\phi_1\partial_R(D\partial_R\phi_2)-\phi_2\partial_R(D\partial_R\phi_1)
=-\frac{\omega^2P}{4D}(\phi_1\phi_2-\phi_2\phi_1)=0. \tag*{(87)} ∂ R ( D W ) = ϕ 1 ∂ R ( D ∂ R ϕ 2 ) − ϕ 2 ∂ R ( D ∂ R ϕ 1 ) = − 4 D ω 2 P ( ϕ 1 ϕ 2 − ϕ 2 ϕ 1 ) = 0. (87)
At the boundary, the solutions ϕ A \phi_A ϕ A and ϕ B \phi_B ϕ B of § 3.2 behave as ϕ A ≃ 1 \phi_A \simeq1 ϕ A ≃ 1 and ϕ B ≃ R − 2 \phi_B \simeq R^{-2} ϕ B ≃ R − 2 , and D ≃ R 3 D \simeq R^3 D ≃ R 3 , so W [ ϕ A , ϕ B ] ≃ − 2 R − 3 W[\phi_A,\phi_B] \simeq-2R^{-3} W [ ϕ A , ϕ B ] ≃ − 2 R − 3 and
D W [ ϕ A , ϕ B ] = − 2. (88) D W[\phi_A,\phi_B] = -2. \tag*{(88)} D W [ ϕ A , ϕ B ] = − 2. (88)
Using ϕ i n = A ϕ A + B ϕ B \phi_{\mathrm{in}} = A\phi_A + B\phi_B ϕ in = A ϕ A + B ϕ B and W [ ϕ B , ϕ B ] = 0 W[\phi_B,\phi_B] = 0 W [ ϕ B , ϕ B ] = 0 , we have
W [ ϕ i n , ϕ B ] = A W [ ϕ A , ϕ B ] = − 2 D A , (89) W[\phi_{\mathrm{in}},\phi_B] = A W[\phi_A,\phi_B] = -\frac{2}{D}A, \tag*{(89)} W [ ϕ in , ϕ B ] = A W [ ϕ A , ϕ B ] = − D 2 A , (89)
so we can find A ( ω ) A(\omega) A ( ω ) as
A ( ω ) = − D 2 ( ϕ i n ∂ R ϕ B − ϕ B ∂ R ϕ i n ) . (90) A(\omega) = -\frac{D}{2}\left(\phi_{\mathrm{in}} \partial_R\phi_B - \phi_B \partial_R\phi_{\mathrm{in}}\right). \tag*{(90)} A ( ω ) = − 2 D ( ϕ in ∂ R ϕ B − ϕ B ∂ R ϕ in ) . (90)
This can be computed at any R R R . Therefore, we only need to know ϕ i n \phi_{\mathrm{in}} ϕ in , ϕ B \phi_B ϕ B and their derivatives, at the same point R R R .
Frobenius expansions
The wave equation (20) is
D 2 ∂ R 2 ϕ + D ( ∂ R D ) ∂ R ϕ + ω 2 P 4 ϕ = 0 , (91) D^2\partial_R^2\phi+ D\left(\partial_R D\right)\partial_R\phi+ \frac{\omega^2P}{4}\phi= 0, \tag*{(91)} D 2 ∂ R 2 ϕ + D ( ∂ R D ) ∂ R ϕ + 4 ω 2 P ϕ = 0 , (91)
and dividing this by D 2 D^2 D 2 , we can write
∂ R 2 ϕ + ∂ R D D ∂ R ϕ + ω 2 P 4 D 2 ϕ = 0. (92) \partial_R^2\phi+ \frac{\partial_R D}{D}\partial_R\phi+ \frac{\omega^2P}{4D^2}\phi= 0. \tag*{(92)} ∂ R 2 ϕ + D ∂ R D ∂ R ϕ + 4 D 2 ω 2 P ϕ = 0. (92)
At finite R R R , the coefficients diverge only at the zeros of D D D , that is at R = 1 R = 1 R = 1 , R − R_- R − , R 3 R_3 R 3 . There, the coefficient of ∂ R ϕ \partial_R\phi ∂ R ϕ has at most a simple pole and the coefficient of ϕ \phi ϕ has at most a double pole, so these are regular singular points. The boundary R = ∞ R = \infty R = ∞ is also a regular singular point, as we see from (A.24) below. Therefore, we can apply the Frobenius expansion around each of them. An expansion around a singular point converges up to the nearest other singular point. For example, the expansion around the outer horizon R = 1 R = 1 R = 1 converges in the region ∣ R − 1 ∣ < 1 − R − \lvert R - 1\rvert< 1 - R_- ∣ R − 1 ∣ < 1 − R − .
We need the expansions around the horizon and the boundary, which give ϕ i n \phi_{\mathrm{in}} ϕ in and ϕ B \phi_B ϕ B . These are singular points of the wave equation, so a numerical integration cannot start there. The infalling condition at the horizon and the normalizable condition at the boundary are statements about the behavior at these points. The Frobenius expansions turn them into values of ϕ \phi ϕ and ∂ R ϕ \partial_R\phi ∂ R ϕ at regular points, with any desired precision. When the convergence regions of the two expansions do not overlap, we connect them by solving the wave equation numerically.
We start with the horizon. With ϵ = R − 1 \epsilon= R - 1 ϵ = R − 1 , we can write
D = d 1 ϵ + d 2 ϵ 2 + d 3 ϵ 3 , P = p 0 + p 1 ϵ + p 2 ϵ 2 + p 3 ϵ 3 . (93) D = d_1\epsilon+ d_2\epsilon^2 + d_3\epsilon^3,\qquad P = p_0 + p_1\epsilon+ p_2\epsilon^2 + p_3\epsilon^3. \tag*{(93)} D = d 1 ϵ + d 2 ϵ 2 + d 3 ϵ 3 , P = p 0 + p 1 ϵ + p 2 ϵ 2 + p 3 ϵ 3 . (93)
At ϵ → 0 \epsilon\to0 ϵ → 0 , the leading order of the wave equation (A.6) is
( d 1 ϵ ) 2 ∂ ϵ 2 ϕ + d 1 2 ϵ ⋅ ∂ ϵ ϕ + ω 2 p 0 4 ⋅ ϕ = 0. (94) \left(d_1\epsilon\right)^2\partial_{\epsilon}^2\phi+d_1^2\epsilon\cdot\partial_{\epsilon}\phi+\frac{\omega^2p_0}{4}\cdot\phi=0. \tag*{(94)} ( d 1 ϵ ) 2 ∂ ϵ 2 ϕ + d 1 2 ϵ ⋅ ∂ ϵ ϕ + 4 ω 2 p 0 ⋅ ϕ = 0. (94)
If we insert ϕ = ϵ s \phi=\epsilon^s ϕ = ϵ s ,
d 1 2 ⋅ s ( s − 1 ) + d 1 2 ⋅ s + ω 2 p 0 4 = 0 → d 1 2 ⋅ s 2 + ω 2 p 0 4 = 0 , (95) d_1^2\cdot s(s-1)+d_1^2\cdot s+\frac{\omega^2p_0}{4}=0\ \to\ d_1^2\cdot s^2+\frac{\omega^2p_0}{4}=0, \tag*{(95)} d 1 2 ⋅ s ( s − 1 ) + d 1 2 ⋅ s + 4 ω 2 p 0 = 0 → d 1 2 ⋅ s 2 + 4 ω 2 p 0 = 0 , (95)
so we obtain
s = ± i ω p 0 2 d 1 = ± i ω β 4 π , (96) s=\pm\frac{i\omega\sqrt{p_0}}{2d_1}=\pm\frac{i\omega\beta}{4\pi}, \tag*{(96)} s = ± 2 d 1 iω p 0 = ± 4 π iω β , (96)
where we used d 1 = D ′ ( 1 ) d_1=D'(1) d 1 = D ′ ( 1 ) , p 0 = P ( 1 ) p_0=P(1) p 0 = P ( 1 ) and β = 2 π P ( 1 ) D ′ ( 1 ) \beta=\frac{2\pi\sqrt{P(1)}}{D'(1)} β = D ′ ( 1 ) 2 π P ( 1 ) . We have to find which one is the infalling solution. Near the horizon, the tortoise coordinate S S S of § 3.3 satisfies
d S = − P 2 D d R ≃ − p 0 2 d 1 ϵ d ϵ = − β 4 π d ϵ ϵ , (97) dS=-\frac{\sqrt{P}}{2D}dR\simeq-\frac{\sqrt{p_0}}{2d_1\epsilon}d\epsilon=-\frac{\beta}{4\pi}\frac{d\epsilon}{\epsilon}, \tag*{(97)} d S = − 2 D P d R ≃ − 2 d 1 ϵ p 0 d ϵ = − 4 π β ϵ d ϵ , (97)
and
S ≃ − β 4 π ln ϵ → + ∞ . (98) S\simeq-\frac{\beta}{4\pi}\ln\epsilon\to+\infty. \tag*{(98)} S ≃ − 4 π β ln ϵ → + ∞. (98)
The infalling wave is e − i ω ( t − S ) e^{-i\omega(t-S)} e − iω ( t − S ) and
ϕ ∝ e i ω S = ϵ − i ω β / 4 π . (99) \phi\propto e^{i\omega S}=\epsilon^{-i\omega\beta/4\pi}. \tag*{(99)} ϕ ∝ e iω S = ϵ − iω β /4 π . (99)
Therefore, s = − i ω β 4 π s=-\frac{i\omega\beta}{4\pi} s = − 4 π iω β corresponds to the infalling wave.
We set the Frobenius expansion:
ϕ i n = ∑ n ≥ 0 a n ϵ n + s , a 0 = 1. (100) \phi_{\mathrm{in}}=\sum_{n\geq0}a_n\epsilon^{n+s},\qquad a_0=1. \tag*{(100)} ϕ in = n ≥ 0 ∑ a n ϵ n + s , a 0 = 1. (100)
For brevity, we write
D 2 = ∑ k = 2 6 e k ϵ k , D ( ∂ R D ) = ∑ k = 1 5 g k ϵ k , P = ∑ k = 0 3 p k ϵ k . (101) D^2=\sum_{k=2}^{6}e_k\epsilon^k,\qquad D\left(\partial_R D\right)=\sum_{k=1}^{5}g_k\epsilon^k,\qquad P=\sum_{k=0}^{3}p_k\epsilon^k. \tag*{(101)} D 2 = k = 2 ∑ 6 e k ϵ k , D ( ∂ R D ) = k = 1 ∑ 5 g k ϵ k , P = k = 0 ∑ 3 p k ϵ k . (101)
Each term of (A.6) becomes
D 2 ∂ R 2 ϕ = ∑ k , n e k a n ( n + s ) ( n + s − 1 ) ϵ n + s + k − 2 , D ( ∂ R D ) ∂ R ϕ = ∑ k , n g k a n ( n + s ) ϵ n + s + k − 1 , ω 2 P 4 ϕ = ω 2 4 ∑ k , n p k a n ϵ n + s + k . \begin{aligned}
D^2\partial_R^2\phi&=\sum_{k,n}e_ka_n(n+s)(n+s-1)\epsilon^{n+s+k-2},\\
D\left(\partial_R D\right)\partial_R\phi&=\sum_{k,n}g_ka_n(n+s)\epsilon^{n+s+k-1},\\
\frac{\omega^2P}{4}\phi&=\frac{\omega^2}{4}\sum_{k,n}p_ka_n\epsilon^{n+s+k}.
\end{aligned} D 2 ∂ R 2 ϕ D ( ∂ R D ) ∂ R ϕ 4 ω 2 P ϕ = k , n ∑ e k a n ( n + s ) ( n + s − 1 ) ϵ n + s + k − 2 , = k , n ∑ g k a n ( n + s ) ϵ n + s + k − 1 , = 4 ω 2 k , n ∑ p k a n ϵ n + s + k .
By collecting the coefficient of ϵ N + s \epsilon^{N+s} ϵ N + s , we obtain
∑ n = 0 N a n [ ( n + s ) ( n + s − 1 ) e N − n + 2 + ( n + s ) g N − n + 1 + ω 2 4 p N − n ] = 0. (102) \sum_{n=0}^{N} a_n \left[(n+s)(n+s-1)e_{N-n+2}+(n+s)g_{N-n+1}+\frac{\omega^2}{4}p_{N-n}\right]=0. \tag*{(102)} n = 0 ∑ N a n [ ( n + s ) ( n + s − 1 ) e N − n + 2 + ( n + s ) g N − n + 1 + 4 ω 2 p N − n ] = 0. (102)
When n = N n=N n = N in the summation, the coefficient of a n a_n a n is
( N + s ) ( N + s − 1 ) e 2 + ( N + s ) g 1 + ω 2 p 0 4 = ( N + s ) 2 d 1 2 + ω 2 p 0 4 = ( N + s ) 2 d 1 2 − d 1 2 s 2 = d 1 2 N ( N + 2 s ) , (103) \begin{aligned}
(N+s)(N+s-1)e_2+(N+s)g_1+\frac{\omega^2p_0}{4}
&=(N+s)^2d_1^2+\frac{\omega^2p_0}{4}\\
&=(N+s)^2d_1^2-d_1^2s^2\\
&=d_1^2N(N+2s), \tag*{(103)}
\end{aligned} ( N + s ) ( N + s − 1 ) e 2 + ( N + s ) g 1 + 4 ω 2 p 0 = ( N + s ) 2 d 1 2 + 4 ω 2 p 0 = ( N + s ) 2 d 1 2 − d 1 2 s 2 = d 1 2 N ( N + 2 s ) , (103)
where we used e 2 = g 1 = d 1 2 e_2=g_1=d_1^2 e 2 = g 1 = d 1 2 and ω 2 p 0 = − 4 d 1 2 s 2 \omega^2p_0=-4d_1^2s^2 ω 2 p 0 = − 4 d 1 2 s 2 . Therefore, when N = 0 N=0 N = 0 , (A.18) automatically holds. This is the freedom of the overall constant factor and we have already fixed it by a 0 = 1 a_0=1 a 0 = 1 . For N > 0 N>0 N > 0 , we can find a N a_N a N recursively from (A.18):
a N = − 1 d 1 2 N ( N + 2 s ) ∑ n = 0 N − 1 a n [ ( n + s ) ( n + s − 1 ) e N − n + 2 + ( n + s ) g N − n + 1 + ω 2 4 p N − n ] . (104) a_N=-\frac{1}{d_1^2N(N+2s)}\sum_{n=0}^{N-1}a_n\left[(n+s)(n+s-1)e_{N-n+2}+(n+s)g_{N-n+1}+\frac{\omega^2}{4}p_{N-n}\right]. \tag*{(104)} a N = − d 1 2 N ( N + 2 s ) 1 n = 0 ∑ N − 1 a n [ ( n + s ) ( n + s − 1 ) e N − n + 2 + ( n + s ) g N − n + 1 + 4 ω 2 p N − n ] . (104)
As we already mentioned, this expansion converges only for ∣ ϵ ∣ < 1 − R − |\epsilon|<1-R_- ∣ ϵ ∣ < 1 − R − . Note that the denominator becomes zero when N + 2 s = 0 N+2s=0 N + 2 s = 0 . Since s = − i β ω 4 π s=-\frac{i\beta\omega}{4\pi} s = − 4 π i β ω , it means
ω = 2 π N i β . (105) \omega=\frac{2\pi N}{i\beta}. \tag*{(105)} ω = i β 2 π N . (105)
These are the Matsubara frequencies in the lower half-plane. The infalling solution ϕ i n \phi_{\mathrm{in}} ϕ in has poles at these frequencies, and they also appear in A ( ω ) \mathcal{A}(\omega) A ( ω ) . We can remove these poles from A ( ω ) \mathcal{A}(\omega) A ( ω ) by dividing it by Γ ( 1 + 2 s ) = Γ ( 1 − i ω β / 2 π ) \Gamma(1+2s)=\Gamma(1-i\omega\beta/2\pi) Γ ( 1 + 2 s ) = Γ ( 1 − iω β /2 π ) , since Γ ( n ) \Gamma(n) Γ ( n ) has poles at n = 0 , − 1 , − 2 , … n=0,-1,-2,\ldots n = 0 , − 1 , − 2 , … .
Next, we consider the boundary. With u = 1 / R u=1/R u = 1/ R , we define
H ( u ) ≡ H 1 H 2 H 3 = ∏ i ( 1 + q i u ) = u 3 P , F ( u ) ≡ H ( u ) − M u 2 = u 3 D . (106) H(u)\equiv H_1H_2H_3=\prod_i(1+q_i u)=u^3P,\qquad F(u)\equiv H(u)-Mu^2=u^3D. \tag*{(106)} H ( u ) ≡ H 1 H 2 H 3 = i ∏ ( 1 + q i u ) = u 3 P , F ( u ) ≡ H ( u ) − M u 2 = u 3 D . (106)
Using ∂ R = − u 2 ∂ u \partial_R=-u^2\partial_u ∂ R = − u 2 ∂ u , we obtain
D ∂ R ϕ = F u 3 ( − u 2 ) ∂ u ϕ = − F u ∂ u ϕ , (107) D\partial_R\phi=\frac{F}{u^3}(-u^2)\partial_u\phi=-\frac{F}{u}\partial_u\phi, \tag*{(107)} D ∂ R ϕ = u 3 F ( − u 2 ) ∂ u ϕ = − u F ∂ u ϕ , (107)
∂ R ( D ∂ R ϕ ) = u 2 ∂ u [ F u ∂ u ϕ ] = u F ϕ u u + ( u F u − F ) ϕ u , \partial_R(D\partial_R\phi)=u^2\partial_u\left[\frac{F}{u}\partial_u\phi\right]=uF\phi_{uu}+(uF_u-F)\phi_u, ∂ R ( D ∂ R ϕ ) = u 2 ∂ u [ u F ∂ u ϕ ] = u F ϕ uu + ( u F u − F ) ϕ u ,
where a subscript u u u denotes ∂ u \partial_u ∂ u . The wave equation ∂ R ( D ∂ R ϕ ) = − ω 2 P 4 D ϕ \partial_R(D\partial_R\phi)=-\frac{\omega^2P}{4D}\phi ∂ R ( D ∂ R ϕ ) = − 4 D ω 2 P ϕ becomes, after multiplying by F F F ,
u F 2 ϕ u u + ( u F F u − F 2 ) ϕ u + ω 2 H 4 ϕ = 0. (108) uF^2\phi_{uu}+(uFF_u-F^2)\phi_u+\frac{\omega^2H}{4}\phi=0. \tag*{(108)} u F 2 ϕ uu + ( u F F u − F 2 ) ϕ u + 4 ω 2 H ϕ = 0. (108)
Since F ( 0 ) = H ( 0 ) = 1 F(0)=H(0)=1 F ( 0 ) = H ( 0 ) = 1 and F u ( 0 ) = ∑ i q i F_u(0)=\sum_i q_i F u ( 0 ) = ∑ i q i , the leading equation is
u ϕ u u − ϕ u = 0. (109) u\phi_{uu}-\phi_u=0. \tag*{(109)} u ϕ uu − ϕ u = 0. (109)
By inserting ϕ = u α \phi=u^\alpha ϕ = u α , we obtain
α ( α − 1 ) − α = α ( α − 2 ) = 0. (110) \alpha(\alpha-1)-\alpha=\alpha(\alpha-2)=0. \tag*{(110)} α ( α − 1 ) − α = α ( α − 2 ) = 0. (110)
So, α = 0 \alpha=0 α = 0 or α = 2 \alpha=2 α = 2 . Near the boundary, a scalar dual to an operator of dimension Δ \Delta Δ in d d d dimensions has the non-normalizable mode r Δ − d r^{\Delta-d} r Δ − d , which is the source, and the normalizable mode r − Δ r^{-\Delta} r − Δ , which is the response. For our massless ϕ \phi ϕ , Δ = d = 4 \Delta=d=4 Δ = d = 4 , so these are r 0 = 1 r^0=1 r 0 = 1 and r − 4 = R − 2 = u 2 r^{-4}=R^{-2}=u^2 r − 4 = R − 2 = u 2 , which correspond to α = 0 \alpha=0 α = 0 and α = 2 \alpha=2 α = 2 , respectively.
So, ϕ A \phi_A ϕ A is the non-normalizable solution and ϕ B \phi_B ϕ B is the normalizable one. We only need ϕ B \phi_B ϕ B , and we write it as
ϕ B = ∑ n ≥ 0 b n u n + 2 , b 0 = 1. (111) \phi_B=\sum_{n\geq0}b_nu^{n+2},\qquad b_0=1. \tag*{(111)} ϕ B = n ≥ 0 ∑ b n u n + 2 , b 0 = 1. (111)
Just as in the near-horizon case, we write
F 2 = ∑ k = 0 6 e k u k , F F u = ∑ k = 0 5 g k u k , H = ∑ k = 0 3 p k u k . (112) F^2=\sum_{k=0}^{6}e_ku^k,\qquad FF_u=\sum_{k=0}^{5}g_ku^k,\qquad H=\sum_{k=0}^{3}p_ku^k. \tag*{(112)} F 2 = k = 0 ∑ 6 e k u k , F F u = k = 0 ∑ 5 g k u k , H = k = 0 ∑ 3 p k u k . (112)
The next steps are almost the same as near the horizon. By collecting the coefficient of u N + 1 u^{N+1} u N + 1 , we obtain
∑ n = 0 N b n [ ( n + 2 ) ( n + 1 ) e N − n + ( n + 2 ) ( g N − n − 1 − e N − n ) + ω 2 4 p N − n − 1 ] = ∑ n = 0 N b n [ ( n + 2 ) n e N − n + ( n + 2 ) g N − n − 1 + ω 2 4 p N − n − 1 ] = 0. (113) \begin{aligned}
\sum_{n=0}^{N}b_n\left[(n+2)(n+1)e_{N-n}+(n+2)\left(g_{N-n-1}-e_{N-n}\right)+\frac{\omega^2}{4}p_{N-n-1}\right]\\
=\sum_{n=0}^{N}b_n\left[(n+2)ne_{N-n}+(n+2)g_{N-n-1}+\frac{\omega^2}{4}p_{N-n-1}\right]=0. \tag*{(113)}
\end{aligned} n = 0 ∑ N b n [ ( n + 2 ) ( n + 1 ) e N − n + ( n + 2 ) ( g N − n − 1 − e N − n ) + 4 ω 2 p N − n − 1 ] = n = 0 ∑ N b n [ ( n + 2 ) n e N − n + ( n + 2 ) g N − n − 1 + 4 ω 2 p N − n − 1 ] = 0. (113)
When n = N n=N n = N , using e 0 = 1 e_0=1 e 0 = 1 , g − 1 = p − 1 = 0 g_{-1}=p_{-1}=0 g − 1 = p − 1 = 0 , we find that the coefficient of b n b_n b n in the summation is ( N + 2 ) N (N+2)N ( N + 2 ) N . Therefore, when N = 0 N=0 N = 0 , Eq. (A.29) automatically holds. Just as near the horizon, this is the freedom of the overall constant factor and we have already fixed it by b 0 = 1 b_0=1 b 0 = 1 . For N > 0 N>0 N > 0 , we can find b N b_N b N recursively from Eq. (A.29):
b N = − 1 ( N + 2 ) N ∑ n = 0 N − 1 b n [ ( n + 2 ) n e N − n + ( n + 2 ) g N − n − 1 + ω 2 4 p N − n − 1 ] . (114) b_N=-\frac{1}{(N+2)N}\sum_{n=0}^{N-1}b_n\left[(n+2)ne_{N-n}+(n+2)g_{N-n-1}+\frac{\omega^2}{4}p_{N-n-1}\right]. \tag*{(114)} b N = − ( N + 2 ) N 1 n = 0 ∑ N − 1 b n [ ( n + 2 ) n e N − n + ( n + 2 ) g N − n − 1 + 4 ω 2 p N − n − 1 ] . (114)
For example, b 1 = − 1 3 [ 2 g 0 + ω 2 4 p 0 ] b_1=-\frac{1}{3}\left[2g_0+\frac{\omega^2}{4}p_0\right] b 1 = − 3 1 [ 2 g 0 + 4 ω 2 p 0 ] where g 0 = q 1 + q 2 + q 3 g_0=q_1+q_2+q_3 g 0 = q 1 + q 2 + q 3 , p 0 = 1 p_0=1 p 0 = 1 . This series converges in the region ∣ u ∣ < min ( 1 , 1 ∣ R 3 ∣ ) |u|<\min\left(1,\frac{1}{|R_3|}\right) ∣ u ∣ < min ( 1 , ∣ R 3 ∣ 1 ) , since the singular points are at u = 1 u=1 u = 1 , u = 1 / R 3 u=1/R_3 u = 1/ R 3 and u = 1 / R − > 1 u=1/R_->1 u = 1/ R − > 1 .
Spectral function in terms of A \mathcal{A} A
We now derive (26) . We do not need to calculate B \mathcal{B} B . For real ω \omega ω , ϕ i n ∗ \phi_{\rm in}^{*} ϕ in ∗ is also a solution, so D W [ ϕ i n , ϕ i n ∗ ] D W[\phi_{\rm in},\phi_{\rm in}^{*}] D W [ ϕ in , ϕ in ∗ ] is independent of R R R . Near the horizon, ϕ i n ≃ ϵ s \phi_{\rm in}\simeq\epsilon^{s} ϕ in ≃ ϵ s and ϕ i n ∗ ≃ ϵ − s \phi_{\rm in}^{*}\simeq\epsilon^{-s} ϕ in ∗ ≃ ϵ − s because s = − i ω β 4 π s=-\frac{i\omega\beta}{4\pi} s = − 4 π iω β , so
D W [ ϕ i n , ϕ i n ∗ ] ≃ d 1 ϵ ⋅ ( − 2 s ϵ ) = i ω β d 1 2 π = i ω P ( 1 ) . (115) D W[\phi_{\rm in},\phi_{\rm in}^{*}]\simeq d_{1}\epsilon\cdot\left(-\frac{2s}{\epsilon}\right)=\frac{i\omega\beta d_{1}}{2\pi}=i\omega\sqrt{P(1)}. \tag*{(115)} D W [ ϕ in , ϕ in ∗ ] ≃ d 1 ϵ ⋅ ( − ϵ 2 s ) = 2 π iω β d 1 = iω P ( 1 ) . (115)
On the other hand, ϕ A \phi_{A} ϕ A and ϕ B \phi_{B} ϕ B are real for real ω \omega ω , so ϕ i n ∗ = A ∗ ϕ A + B ∗ ϕ B \phi_{\rm in}^{*}=\mathcal{A}^{*}\phi_{A}+\mathcal{B}^{*}\phi_{B} ϕ in ∗ = A ∗ ϕ A + B ∗ ϕ B and
D W [ ϕ i n , ϕ i n ∗ ] = ( A B ∗ − A ∗ B ) D W [ ϕ A , ϕ B ] = 4 i Im ( A ∗ B ) , (116) D W[\phi_{\rm in},\phi_{\rm in}^{*}]=(\mathcal{A}\mathcal{B}^{*}-\mathcal{A}^{*}\mathcal{B})D W[\phi_{A},\phi_{B}]=4i\operatorname{Im}(\mathcal{A}^{*}\mathcal{B}), \tag*{(116)} D W [ ϕ in , ϕ in ∗ ] = ( A B ∗ − A ∗ B ) D W [ ϕ A , ϕ B ] = 4 i Im ( A ∗ B ) , (116)
where we used (A.3). Therefore, Im ( A ∗ B ) = ω P ( 1 ) / 4 \operatorname{Im}(\mathcal{A}^{*}\mathcal{B})=\omega\sqrt{P(1)}/4 Im ( A ∗ B ) = ω P ( 1 ) /4 , and
ρ = Im B A = Im ( A ∗ B ) ∣ A ∣ 2 = ω P ( 1 ) 4 ∣ A ∣ 2 . (117) \rho=\operatorname{Im}\frac{\mathcal{B}}{\mathcal{A}}=\frac{\operatorname{Im}(\mathcal{A}^{*}\mathcal{B})}{|\mathcal{A}|^{2}}=\frac{\omega\sqrt{P(1)}}{4|\mathcal{A}|^{2}}. \tag*{(117)} ρ = Im A B = ∣ A ∣ 2 Im ( A ∗ B ) = 4∣ A ∣ 2 ω P ( 1 ) . (117)
This formula holds for the normalization ∣ ϕ i n ∣ = 1 |\phi_{\rm in}|=1 ∣ ϕ in ∣ = 1 at the horizon, which follows from a 0 = 1 a_{0}=1 a 0 = 1 .
Next, we calculate ρ v a c \rho_{\rm vac} ρ vac . In the vacuum, we have q i = 0 q_i=0 q i = 0 and M = 0 M=0 M = 0 , so P = D = R 3 P=D=R^{3} P = D = R 3 and (20) becomes
∂ R ( R 3 ∂ R ϕ ) = − ω 2 4 ϕ . (118) \partial_{R}\left(R^{3}\partial_{R}\phi\right)=-\frac{\omega^{2}}{4}\phi. \tag*{(118)} ∂ R ( R 3 ∂ R ϕ ) = − 4 ω 2 ϕ . (118)
With z = 1 / R z=1/\sqrt{R} z = 1/ R , its solutions are z 2 J 2 ( ω z ) z^{2}J_{2}(\omega z) z 2 J 2 ( ω z ) and z 2 Y 2 ( ω z ) z^{2}Y_{2}(\omega z) z 2 Y 2 ( ω z ) , or equivalently z 2 H 2 ( 1 , 2 ) ( ω z ) z^{2}H_{2}^{(1,2)}(\omega z) z 2 H 2 ( 1 , 2 ) ( ω z ) . Using
J 2 ( ζ ) ≃ ζ 2 8 , Y 2 ( ζ ) ≃ − 4 π ζ 2 , ( ζ → 0 ) , (119) J_{2}(\zeta)\simeq\frac{\zeta^{2}}{8},\qquad Y_{2}(\zeta)\simeq-\frac{4}{\pi\zeta^{2}},\qquad(\zeta\to0), \tag*{(119)} J 2 ( ζ ) ≃ 8 ζ 2 , Y 2 ( ζ ) ≃ − π ζ 2 4 , ( ζ → 0 ) , (119)
we have z 2 J 2 ( ω z ) ≃ ω 2 8 z 4 = ω 2 8 R − 2 z^{2}J_{2}(\omega z)\simeq\frac{\omega^{2}}{8}z^{4}=\frac{\omega^{2}}{8}R^{-2} z 2 J 2 ( ω z ) ≃ 8 ω 2 z 4 = 8 ω 2 R − 2 and z 2 Y 2 ( ω z ) ≃ − 4 π ω 2 z^{2}Y_{2}(\omega z)\simeq-\frac{4}{\pi\omega^{2}} z 2 Y 2 ( ω z ) ≃ − π ω 2 4 near the boundary. So,
ϕ B = 8 ω 2 z 2 J 2 ( ω z ) , ϕ A = − π ω 2 4 z 2 Y 2 ( ω z ) . (120) \phi_{B}=\frac{8}{\omega^{2}}z^{2}J_{2}(\omega z),\qquad\phi_{A}=-\frac{\pi\omega^{2}}{4}z^{2}Y_{2}(\omega z). \tag*{(120)} ϕ B = ω 2 8 z 2 J 2 ( ω z ) , ϕ A = − 4 π ω 2 z 2 Y 2 ( ω z ) . (120)
In the vacuum, the tortoise coordinate is S = z S=z S = z , and z 2 H 2 ( 1 ) ( ω z ) ∝ z 3 / 2 e i ω z z^{2}H_{2}^{(1)}(\omega z)\propto z^{3/2}e^{i\omega z} z 2 H 2 ( 1 ) ( ω z ) ∝ z 3/2 e iω z at large ω z \omega z ω z , so this is the infalling solution. Its overall factor drops out of B / A \mathcal{B}/\mathcal{A} B / A , so we may take
ϕ i n = z 2 H 2 ( 1 ) ( ω z ) = z 2 J 2 ( ω z ) + i z 2 Y 2 ( ω z ) = − 4 i π ω 2 ϕ A + ω 2 8 ϕ B . (121) \phi_{\rm in}=z^{2}H_{2}^{(1)}(\omega z)=z^{2}J_{2}(\omega z)+iz^{2}Y_{2}(\omega z)=-\frac{4i}{\pi\omega^{2}}\phi_{A}+\frac{\omega^{2}}{8}\phi_{B}. \tag*{(121)} ϕ in = z 2 H 2 ( 1 ) ( ω z ) = z 2 J 2 ( ω z ) + i z 2 Y 2 ( ω z ) = − π ω 2 4 i ϕ A + 8 ω 2 ϕ B . (121)
Therefore, we obtain
A v a c = − 4 i π ω 2 , B v a c = ω 2 8 , ρ v a c = Im B v a c A v a c = π ω 4 32 . (122) \mathcal{A}_{\rm vac}=-\frac{4i}{\pi\omega^{2}},\qquad\mathcal{B}_{\rm vac}=\frac{\omega^{2}}{8},\qquad\rho_{\rm vac}=\operatorname{Im}\frac{\mathcal{B}_{\rm vac}}{\mathcal{A}_{\rm vac}}=\frac{\pi\omega^{4}}{32}. \tag*{(122)} A vac = − π ω 2 4 i , B vac = 8 ω 2 , ρ vac = Im A vac B vac = 32 π ω 4 . (122)
Numerics and QNMs
For a given ω \omega ω , we choose a point R 0 R_0 R 0 inside the region of convergence of the Frobenius series near the horizon, and evaluate ϕ i n \phi_{\mathrm{in}} ϕ in and ∂ R ϕ i n \partial_R\phi_{\mathrm{in}} ∂ R ϕ in there from the series. We then solve the wave equation numerically from R 0 R_0 R 0 to a point R 1 R_1 R 1 , where R 1 R_1 R 1 is inside the region of convergence of the series of ϕ B \phi_B ϕ B near the boundary, as shown in Figure 8 .
Figure 8. (a) The order of the zeros of D D D and of P P P on the real R R R axis. (b) The upper half of the complex R R R -plane for ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 1 5 ) (q_1,q_2,q_3) = (3,1,\frac{1}{5}) ( q 1 , q 2 , q 3 ) = ( 3 , 1 , 5 1 ) . The regions of convergence of the horizon series and of the boundary series do not overlap, so we carry the solution from R 0 R_0 R 0 to R 1 R_1 R 1 along the real axis.
Using the values of ϕ i n \phi_{\mathrm{in}} ϕ in , ∂ R ϕ i n \partial_R\phi_{\mathrm{in}} ∂ R ϕ in , ϕ B \phi_B ϕ B , ∂ R ϕ B \partial_R\phi_B ∂ R ϕ B at R 1 R_1 R 1 , the formula (A.5) gives A ( ω ) \mathcal{A}(\omega) A ( ω ) . From this A ( ω ) \mathcal{A}(\omega) A ( ω ) , we can obtain ρ ( ω ) \rho(\omega) ρ ( ω ) using (26) . In Sec. 4, we compare e ω Im T 1 Δ ρ / ρ v a c e^{\omega\operatorname{Im} T_1}\Delta\rho/\rho_{\mathrm{vac}} e ω Im T 1 Δ ρ / ρ vac with the WKB expectation. Since e ω Im T 1 e^{\omega\operatorname{Im} T_1} e ω Im T 1 is very large, we need Δ ρ / ρ v a c \Delta\rho/\rho_{\mathrm{vac}} Δ ρ / ρ vac to high accuracy: at ω = 120 \omega=120 ω = 120 , it is as small as 10 − 50 10^{-50} 1 0 − 50 for ( 3 , 1 , 1 5 ) (3,1,\frac{1}{5}) ( 3 , 1 , 5 1 ) and 10 − 62 10^{-62} 1 0 − 62 for ( 2 , 1 2 , 0 ) (2,\frac{1}{2},0) ( 2 , 2 1 , 0 ) . We therefore use 100 digits for real ω \omega ω in 16 ≤ ω ≤ 120 16\leq\omega\leq120 16 ≤ ω ≤ 120 .
The QNMs are the solutions of A ( ω ) = 0 \mathcal{A}(\omega)=0 A ( ω ) = 0 . To remove the Matsubara poles of A ( ω ) \mathcal{A}(\omega) A ( ω ) , we define
J ( ω ) = A ( ω ) Γ ( 1 − i ω β / 2 π ) . (123) J(\omega)=\frac{\mathcal{A}(\omega)}{\Gamma(1-i\omega\beta/2\pi)}. \tag*{(123)} J ( ω ) = Γ ( 1 − iω β /2 π ) A ( ω ) . (123)
which has no poles. We evaluate J ( ω ) J(\omega) J ( ω ) in the same way, and find its zeros with the secant method,
ω k + 1 = ω k − J ( ω k ) ω k − ω k − 1 J ( ω k ) − J ( ω k − 1 ) . (124) \omega_{k+1}=\omega_k-J(\omega_k)\frac{\omega_k-\omega_{k-1}}{J(\omega_k)-J(\omega_{k-1})}. \tag*{(124)} ω k + 1 = ω k − J ( ω k ) J ( ω k ) − J ( ω k − 1 ) ω k − ω k − 1 . (124)
J J J also vanishes at a Matsubara frequency if the sum in (A.20) vanishes there, since then A \mathcal{A} A stays finite. We checked in exact arithmetic that this does not happen for N ≤ 6 N\leq6 N ≤ 6 in our two examples. We compute each zero with 0.7 ∣ ω ∣ + 25 0.7|\omega|+25 0.7∣ ω ∣ + 25 digits.
OPE coefficients
In this section, we show that, for t > 0 t>0 t > 0 and in the normalization G R ( ω ) = B / A G_R(\omega)=\mathcal{B}/\mathcal{A} G R ( ω ) = B / A ,
G R ( t ) = 3 2 t − 5 + ∑ m ≥ 1 γ m t 2 m − 1 , (125) G_R(t)=\frac{3}{2}t^{-5}+\sum_{m\geq1}\gamma_m t^{2m-1}, \tag*{(125)} G R ( t ) = 2 3 t − 5 + m ≥ 1 ∑ γ m t 2 m − 1 , (125)
for the metric (5) , and explain how to compute γ m \gamma_m γ m from the large-ω \omega ω expansion of G R ( ω ) G_R(\omega) G R ( ω ) . Unlike A ( ω ) \mathcal{A}(\omega) A ( ω ) in Appendix A, this does not require solving the wave equation from the horizon to the boundary. We solve the wave equation only around the boundary, order by order in 1 / ω 2 1/\omega^2 1/ ω 2 . We then read off G R ( ω ) G_R(\omega) G R ( ω ) , and transform it to G R ( t ) G_R(t) G R ( t ) .
Near-boundary expansion in 1 / ω 2 1/\omega^2 1/ ω 2
Dividing (A.24) by F 2 F^2 F 2 gives
u ϕ u u + ( u F u F − 1 ) ϕ u + ω 2 H 4 F 2 ϕ = 0. (126) u\phi_{uu}+\left(\frac{uF_u}{F}-1\right)\phi_u+\frac{\omega^2H}{4F^2}\phi=0. \tag*{(126)} u ϕ uu + ( F u F u − 1 ) ϕ u + 4 F 2 ω 2 H ϕ = 0. (126)
The vacuum solutions of Appendix A depend on u u u through ω 2 u \omega^2u ω 2 u , so at large ω \omega ω we introduce
v = − ω 2 u . (127) v=-\omega^2u. \tag*{(127)} v = − ω 2 u . (127)
With ∂ u = − ω 2 ∂ v \partial_u=-\omega^2\partial_v ∂ u = − ω 2 ∂ v and after dividing by − ω 2 -\omega^2 − ω 2 , the equation becomes
L 0 ϕ : = v ϕ v v − ϕ v − 1 4 ϕ = Q 1 ( u ) ϕ v + Q 2 ( u ) ϕ , u = − v ω 2 . (128) L_0\phi:= v\phi_{vv}-\phi_v-\frac{1}{4}\phi=Q_1(u)\phi_v+Q_2(u)\phi,\qquad u=-\frac{v}{\omega^2}. \tag*{(128)} L 0 ϕ := v ϕ v v − ϕ v − 4 1 ϕ = Q 1 ( u ) ϕ v + Q 2 ( u ) ϕ , u = − ω 2 v . (128)
where
Q 1 ( u ) = − u F u F = ∑ k ≥ 1 Q 1 , k u k , Q 2 ( u ) = 1 4 ( H F 2 − 1 ) = ∑ k ≥ 1 Q 2 , k u k . (129) Q_1(u)=-\frac{uF_u}{F}=\sum_{k\geq1}Q_{1,k}u^k,\qquad Q_2(u)=\frac{1}{4}\left(\frac{H}{F^2}-1\right)=\sum_{k\geq1}Q_{2,k}u^k. \tag*{(129)} Q 1 ( u ) = − F u F u = k ≥ 1 ∑ Q 1 , k u k , Q 2 ( u ) = 4 1 ( F 2 H − 1 ) = k ≥ 1 ∑ Q 2 , k u k . (129)
Both Q 1 Q_1 Q 1 and Q 2 Q_2 Q 2 vanish at u = 0 u=0 u = 0 because F ( 0 ) = H ( 0 ) = 1 F(0)=H(0)=1 F ( 0 ) = H ( 0 ) = 1 . The operator L 0 L_0 L 0 does not contain ω \omega ω , and ω \omega ω enters only through u = − v / ω 2 u=-v/\omega^2 u = − v / ω 2 in Q 1 Q_1 Q 1 and Q 2 Q_2 Q 2 .
In the vacuum, Q 1 = Q 2 = 0 Q_1=Q_2=0 Q 1 = Q 2 = 0 , and the two solutions of L 0 ϕ = 0 L_0\phi=0 L 0 ϕ = 0 are v K 2 ( v ) vK_2(\sqrt{v}) v K 2 ( v ) and v I 2 ( v ) vI_2(\sqrt{v}) v I 2 ( v ) . The one that decays into the bulk for Im ω > 0 \operatorname{Im}\omega>0 Im ω > 0 is
χ ( v ) = v K 2 ( v ) , v = − i ω u , (130) \chi(v)=vK_2(\sqrt{v}),\qquad\sqrt{v}=-i\omega\sqrt{u}, \tag*{(130)} χ ( v ) = v K 2 ( v ) , v = − iω u , (130)
which is proportional to the vacuum infalling solution z 2 H 2 ( 1 ) ( ω z ) z^2H_2^{(1)}(\omega z) z 2 H 2 ( 1 ) ( ω z ) of Appendix § 9 , with z = u z=\sqrt{u} z = u .
Since u = − v / ω 2 u=-v/\omega^2 u = − v / ω 2 , the right-hand side of (B.4) is a power series in ω − 2 \omega^{-2} ω − 2 ,
Q 1 ϕ v + Q 2 ϕ = ∑ k ≥ 1 ω − 2 k ( − v ) k ( Q 1 , k ϕ v + Q 2 , k ϕ ) . (131) Q_1\phi_v+Q_2\phi=\sum_{k\geq1}\omega^{-2k}(-v)^k\left(Q_{1,k}\phi_v+Q_{2,k}\phi\right). \tag*{(131)} Q 1 ϕ v + Q 2 ϕ = k ≥ 1 ∑ ω − 2 k ( − v ) k ( Q 1 , k ϕ v + Q 2 , k ϕ ) . (131)
We therefore expand
ϕ = ∑ n ≥ 0 ω − 2 n ϕ n , ϕ 0 = χ ( v ) , (132) \phi=\sum_{n\geq0}\omega^{-2n}\phi_n,\qquad\phi_0=\chi(v), \tag*{(132)} ϕ = n ≥ 0 ∑ ω − 2 n ϕ n , ϕ 0 = χ ( v ) , (132)
and solve, order by order,
L 0 ϕ n = ∑ k = 1 n ( − v ) k ( Q 1 , k ∂ v ϕ n − k + Q 2 , k ϕ n − k ) . (133) L_0\phi_n=\sum_{k=1}^{n}(-v)^k\left(Q_{1,k}\partial_v\phi_{n-k}+Q_{2,k}\phi_{n-k}\right). \tag*{(133)} L 0 ϕ n = k = 1 ∑ n ( − v ) k ( Q 1 , k ∂ v ϕ n − k + Q 2 , k ϕ n − k ) . (133)
We look for each order in the form
ϕ n = λ n ( v ) χ + μ n ( v ) χ ′ , (134) \phi_n=\lambda_n(v)\chi+\mu_n(v)\chi', \tag*{(134)} ϕ n = λ n ( v ) χ + μ n ( v ) χ ′ , (134)
with polynomials λ n \lambda_n λ n and μ n \mu_n μ n , where ′ = d / d v '=d/dv ′ = d / d v . This form closes under ∂ v \partial_v ∂ v and L 0 L_0 L 0 , because χ ′ ′ \chi'' χ ′′ is a combination of χ \chi χ and χ ′ \chi' χ ′ . Since ϕ 0 = χ \phi_0=\chi ϕ 0 = χ , we have λ 0 = 1 \lambda_0=1 λ 0 = 1 and μ 0 = 0 \mu_0=0 μ 0 = 0 .
From L 0 χ = 0 L_0\chi=0 L 0 χ = 0 , we have χ ′ ′ = ( χ ′ + χ / 4 ) / v \chi''=(\chi'+\chi/4)/v χ ′′ = ( χ ′ + χ /4 ) / v . Hence, for any polynomials λ \lambda λ and μ \mu μ , we obtain
( λ χ + μ χ ′ ) ′ = λ ′ χ + ( λ + μ ′ ) χ ′ + μ χ ′ ′ = λ ′ χ + ( λ + μ ′ ) χ ′ + μ v ( χ ′ + χ 4 ) = ( λ ′ + μ 4 v ) χ + ( λ + μ ′ + μ v ) χ ′ . (135) \begin{aligned}
(\lambda\chi+\mu\chi')'&=\lambda'\chi+(\lambda+\mu')\chi'+\mu\chi''\\
&=\lambda'\chi+(\lambda+\mu')\chi'+\frac{\mu}{v}\left(\chi'+\frac{\chi}{4}\right)\\
&=\left(\lambda'+\frac{\mu}{4v}\right)\chi+\left(\lambda+\mu'+\frac{\mu}{v}\right)\chi'. \tag*{(135)}
\end{aligned} ( λ χ + μ χ ′ ) ′ = λ ′ χ + ( λ + μ ′ ) χ ′ + μ χ ′′ = λ ′ χ + ( λ + μ ′ ) χ ′ + v μ ( χ ′ + 4 χ ) = ( λ ′ + 4 v μ ) χ + ( λ + μ ′ + v μ ) χ ′ . (135)
and therefore,
L 0 [ λ χ + μ χ ′ ] = v ( λ χ + μ χ ′ ) ′ ′ − ( λ χ + μ χ ′ ) ′ − 1 4 ( λ χ + μ χ ′ ) = v [ ( λ ′ + μ 4 v ) χ + ( λ + μ ′ + μ v ) χ ′ ] ′ − [ ( λ ′ + μ 4 v ) χ + ( λ + μ ′ + μ v ) χ ′ ] − 1 4 ( λ χ + μ χ ′ ) = v [ ( λ ′ ′ + μ ′ 4 v − μ 4 v 2 ) χ + ( λ ′ + μ 4 v + λ ′ + μ ′ ′ + μ ′ v − μ v 2 ) χ ′ ] + ( λ + μ ′ + μ v ) 1 v ( χ ′ + χ 4 ) − [ ( λ ′ + μ 4 v ) χ + ( λ + μ ′ + μ v ) χ ′ ] − 1 4 ( λ χ + μ χ ′ ) = ( v λ ′ ′ − λ ′ + μ ′ 2 − μ 4 v ) χ + ( 2 v λ ′ + v μ ′ ′ + μ ′ − μ v ) χ ′ . (136) \begin{aligned}
L_0\left[\lambda\chi+\mu\chi'\right]
&=v(\lambda\chi+\mu\chi')''-(\lambda\chi+\mu\chi')'-\frac{1}{4}(\lambda\chi+\mu\chi')\\
&=v\left[\left(\lambda'+\frac{\mu}{4v}\right)\chi+\left(\lambda+\mu'+\frac{\mu}{v}\right)\chi'\right]'-
\left[\left(\lambda'+\frac{\mu}{4v}\right)\chi+\left(\lambda+\mu'+\frac{\mu}{v}\right)\chi'\right]-\frac{1}{4}(\lambda\chi+\mu\chi')\\
&=v\left[\left(\lambda''+\frac{\mu'}{4v}-\frac{\mu}{4v^2}\right)\chi+
\left(\lambda'+\frac{\mu}{4v}+\lambda'+\mu''+\frac{\mu'}{v}-\frac{\mu}{v^2}\right)\chi'\right]\\
&\quad+\left(\lambda+\mu'+\frac{\mu}{v}\right)\frac{1}{v}\left(\chi'+\frac{\chi}{4}\right)-
\left[\left(\lambda'+\frac{\mu}{4v}\right)\chi+
\left(\lambda+\mu'+\frac{\mu}{v}\right)\chi'\right]-\frac{1}{4}(\lambda\chi+\mu\chi')\\
&=\left(v\lambda''-\lambda'+\frac{\mu'}{2}-\frac{\mu}{4v}\right)\chi+
\left(2v\lambda'+v\mu''+\mu'-\frac{\mu}{v}\right)\chi'. \tag*{(136)}
\end{aligned} L 0 [ λ χ + μ χ ′ ] = v ( λ χ + μ χ ′ ) ′′ − ( λ χ + μ χ ′ ) ′ − 4 1 ( λ χ + μ χ ′ ) = v [ ( λ ′ + 4 v μ ) χ + ( λ + μ ′ + v μ ) χ ′ ] ′ − [ ( λ ′ + 4 v μ ) χ + ( λ + μ ′ + v μ ) χ ′ ] − 4 1 ( λ χ + μ χ ′ ) = v [ ( λ ′′ + 4 v μ ′ − 4 v 2 μ ) χ + ( λ ′ + 4 v μ + λ ′ + μ ′′ + v μ ′ − v 2 μ ) χ ′ ] + ( λ + μ ′ + v μ ) v 1 ( χ ′ + 4 χ ) − [ ( λ ′ + 4 v μ ) χ + ( λ + μ ′ + v μ ) χ ′ ] − 4 1 ( λ χ + μ χ ′ ) = ( v λ ′′ − λ ′ + 2 μ ′ − 4 v μ ) χ + ( 2 v λ ′ + v μ ′′ + μ ′ − v μ ) χ ′ . (136)
All coefficients in (B.11) and (B.12) are polynomials only if μ ( 0 ) = 0 \mu(0)=0 μ ( 0 ) = 0 , so we require μ n ( 0 ) = 0 \mu_n(0)=0 μ n ( 0 ) = 0 for all n n n .
The polynomial λ n \lambda_n λ n has one more freedom. Equation (B.9) fixes ϕ n \phi_n ϕ n only up to a solution of L 0 ϕ = 0 L_0\phi=0 L 0 ϕ = 0 , and only χ \chi χ is infalling, so ϕ n \phi_n ϕ n is fixed up to c n χ c_n\chi c n χ with a constant c n c_n c n . This corresponds to multiplying ϕ \phi ϕ by an overall factor that depends on ω \omega ω , which cancels in B / A \mathcal{B}/\mathcal{A} B / A . We use this freedom to set λ n ( 0 ) = 0 \lambda_n(0)=0 λ n ( 0 ) = 0 for n ≥ 1 n\ge1 n ≥ 1 .
At order n n n , the right-hand side of (B.9) contains only ϕ 0 , … , ϕ n − 1 \phi_0,\ldots,\phi_{n-1} ϕ 0 , … , ϕ n − 1 , which are already known. Using (B.11) for ∂ v ϕ n − k \partial_v\phi_{n-k} ∂ v ϕ n − k , it takes the form S n ( v ) χ + T n ( v ) χ ′ S_n(v)\chi+T_n(v)\chi' S n ( v ) χ + T n ( v ) χ ′ with the known polynomials
S n = ∑ k = 1 n ( − v ) k [ Q 1 , k ( λ n − k ′ + μ n − k 4 v ) + Q 2 , k λ n − k ] , S_n=\sum_{k=1}^{n}(-v)^k\left[Q_{1,k}\left(\lambda'_{n-k}+\frac{\mu_{n-k}}{4v}\right)+Q_{2,k}\lambda_{n-k}\right], S n = k = 1 ∑ n ( − v ) k [ Q 1 , k ( λ n − k ′ + 4 v μ n − k ) + Q 2 , k λ n − k ] ,
T n = ∑ k = 1 n ( − v ) k [ Q 1 , k ( λ n − k + μ n − k ′ + μ n − k v ) + Q 2 , k μ n − k ] . (137) T_n=\sum_{k=1}^{n}(-v)^k\left[Q_{1,k}\left(\lambda_{n-k}+\mu'_{n-k}+\frac{\mu_{n-k}}{v}\right)+Q_{2,k}\mu_{n-k}\right]. \tag*{(137)} T n = k = 1 ∑ n ( − v ) k [ Q 1 , k ( λ n − k + μ n − k ′ + v μ n − k ) + Q 2 , k μ n − k ] . (137)
For instance, λ 0 = 1 \lambda_0=1 λ 0 = 1 and μ 0 = 0 \mu_0=0 μ 0 = 0 give
S 1 = − Q 2 , 1 v , T 1 = − Q 1 , 1 v . (138) S_1=-Q_{2,1}v,\qquad T_1=-Q_{1,1}v. \tag*{(138)} S 1 = − Q 2 , 1 v , T 1 = − Q 1 , 1 v . (138)
Since χ \chi χ and χ ′ \chi' χ ′ are independent, (B.9) and (B.12) give
v λ n ′ ′ − λ n ′ + μ n ′ 2 − μ n 4 v = S n , 2 v λ n ′ + v μ n ′ ′ + μ n ′ − μ n v = T n . (139) v\lambda_n''-\lambda_n'+\frac{\mu_n'}{2}-\frac{\mu_n}{4v}=S_n,\qquad
2v\lambda_n'+v\mu_n''+\mu_n'-\frac{\mu_n}{v}=T_n. \tag*{(139)} v λ n ′′ − λ n ′ + 2 μ n ′ − 4 v μ n = S n , 2 v λ n ′ + v μ n ′′ + μ n ′ − v μ n = T n . (139)
We write
λ n = ∑ d ≥ 0 λ n , d v d , μ n = ∑ d ≥ 0 μ n , d v d , S n = ∑ d ≥ 0 S n , d v d , T n = ∑ d ≥ 0 T n , d v d . (140) \lambda_n=\sum_{d\ge0}\lambda_{n,d}v^d,\qquad
\mu_n=\sum_{d\ge0}\mu_{n,d}v^d,\qquad
S_n=\sum_{d\ge0}S_{n,d}v^d,\qquad
T_n=\sum_{d\ge0}T_{n,d}v^d. \tag*{(140)} λ n = d ≥ 0 ∑ λ n , d v d , μ n = d ≥ 0 ∑ μ n , d v d , S n = d ≥ 0 ∑ S n , d v d , T n = d ≥ 0 ∑ T n , d v d . (140)
where λ 0 , 0 = 1 \lambda_{0,0}=1 λ 0 , 0 = 1 , λ n , 0 = 0 \lambda_{n,0}=0 λ n , 0 = 0 for n ≥ 1 n\ge1 n ≥ 1 , μ n , 0 = 0 \mu_{n,0}=0 μ n , 0 = 0 for all n n n , and S n , 0 = T n , 0 = 0 S_{n,0}=T_{n,0}=0 S n , 0 = T n , 0 = 0 since every term of S n S_n S n and T n T_n T n carries a factor v k v^k v k with k ≥ 1 k\ge1 k ≥ 1 . The coefficients of v d v^d v d give
( d + 1 ) d λ n , d + 1 − ( d + 1 ) λ n , d + 1 + d + 1 2 μ n , d + 1 − 1 4 μ n , d + 1 = S n , d , 2 d λ n , d + ( d + 1 ) d μ n , d + 1 + ( d + 1 ) μ n , d + 1 − μ n , d + 1 = T n , d . (141) \begin{aligned}
(d+1)d\lambda_{n,d+1}-(d+1)\lambda_{n,d+1}
+\frac{d+1}{2}\mu_{n,d+1}-\frac{1}{4}\mu_{n,d+1}
&=S_{n,d},\\
2d\lambda_{n,d}+(d+1)d\mu_{n,d+1}
+(d+1)\mu_{n,d+1}-\mu_{n,d+1}
&=T_{n,d}. \tag*{(141)}
\end{aligned} ( d + 1 ) d λ n , d + 1 − ( d + 1 ) λ n , d + 1 + 2 d + 1 μ n , d + 1 − 4 1 μ n , d + 1 2 d λ n , d + ( d + 1 ) d μ n , d + 1 + ( d + 1 ) μ n , d + 1 − μ n , d + 1 = S n , d , = T n , d . (141)
that is,
( d + 1 ) ( d − 1 ) λ n , d + 1 + 2 d + 1 4 μ n , d + 1 = S n , d , 2 d λ n , d + d ( d + 2 ) μ n , d + 1 = T n , d . (142) \begin{aligned}
(d+1)(d-1)\lambda_{n,d+1}+\frac{2d+1}{4}\mu_{n,d+1}&=S_{n,d},\\
2d\lambda_{n,d}+d(d+2)\mu_{n,d+1}&=T_{n,d}. \tag*{(142)}
\end{aligned} ( d + 1 ) ( d − 1 ) λ n , d + 1 + 4 2 d + 1 μ n , d + 1 2 d λ n , d + d ( d + 2 ) μ n , d + 1 = S n , d , = T n , d . (142)
When we know S n , T n S_n,T_n S n , T n , we can find λ n , d \lambda_{n,d} λ n , d and μ n , d \mu_{n,d} μ n , d for all d d d , as follows. The first equation contains λ n , d + 1 \lambda_{n,d+1} λ n , d + 1 and μ n , d + 1 \mu_{n,d+1} μ n , d + 1 , and the second contains λ n , d \lambda_{n,d} λ n , d and μ n , d + 1 \mu_{n,d+1} μ n , d + 1 . Let v K v^K v K be the highest power appearing in S n S_n S n or T n T_n T n . We take λ n , d = 0 \lambda_{n,d}=0 λ n , d = 0 for d > K d>K d > K and μ n , d = 0 \mu_{n,d}=0 μ n , d = 0 for d > K + 1 d>K+1 d > K + 1 , and solve one unknown at a time starting from the highest d d d . The first equation at d = K d=K d = K gives μ n , K + 1 \mu_{n,K+1} μ n , K + 1 , the second at d = K d=K d = K then gives λ n , K \lambda_{n,K} λ n , K , the first at d = K − 1 d=K-1 d = K − 1 gives μ n , K \mu_{n,K} μ n , K , and so on down to μ n , 1 \mu_{n,1} μ n , 1 . At d = 0 d=0 d = 0 , the second equation reads 0 = T n , 0 0=T_{n,0} 0 = T n , 0 , which holds automatically. The first equation at d = 0 d=0 d = 0 gives
μ n , 1 = 4 λ n , 1 . (143) \mu_{n,1}=4\lambda_{n,1}. \tag*{(143)} μ n , 1 = 4 λ n , 1 . (143)
This determines λ n \lambda_n λ n and μ n \mu_n μ n from S n S_n S n and T n T_n T n .
Starting from λ 0 = 1 \lambda_0=1 λ 0 = 1 and μ 0 = 0 \mu_0=0 μ 0 = 0 , we can obtain S 1 , T 1 S_1,T_1 S 1 , T 1 using (B.13), which gives (B.14). From S 1 , T 1 S_1,T_1 S 1 , T 1 , we find λ 1 , μ 1 \lambda_1,\mu_1 λ 1 , μ 1 using the process above. From λ 0 , μ 0 , λ 1 , μ 1 \lambda_0,\mu_0,\lambda_1,\mu_1 λ 0 , μ 0 , λ 1 , μ 1 , we can obtain S 2 , T 2 S_2,T_2 S 2 , T 2 using (B.13). We repeat this order by order: λ n , ⋯ , λ 0 \lambda_n,\cdots,\lambda_0 λ n , ⋯ , λ 0 and μ n , ⋯ , μ 0 \mu_n,\cdots,\mu_0 μ n , ⋯ , μ 0 give S n + 1 S_{n+1} S n + 1 and T n + 1 T_{n+1} T n + 1 , which give λ n + 1 \lambda_{n+1} λ n + 1 and μ n + 1 \mu_{n+1} μ n + 1 .
Reading off G R ( ω ) G_R(\omega) G R ( ω )
Near v = 0 v=0 v = 0 , χ ( v ) = v K 2 ( v ) \chi(v)=vK_2(\sqrt{v}) χ ( v ) = v K 2 ( v ) has the series expansion
χ ( v ) = 2 − v 2 + v 2 8 ( k E − 1 2 log v ) + O ( v 3 log v ) , k E = log 2 − γ E + 3 4 . (144) \chi(v)=2-\frac{v}{2}+\frac{v^2}{8}\left(k_E-\frac{1}{2}\log v\right)+O(v^3\log v),\qquad k_E=\log2-\gamma_E+\frac{3}{4}. \tag*{(144)} χ ( v ) = 2 − 2 v + 8 v 2 ( k E − 2 1 log v ) + O ( v 3 log v ) , k E = log 2 − γ E + 4 3 . (144)
where γ E \gamma_E γ E is Euler’s constant. Hence,
ϕ n = λ n χ + μ n χ ′ = ( λ n , 0 + λ n , 1 v + λ n , 2 v 2 + O ( v 3 ) ) ⋅ [ 2 − v 2 + v 2 8 ( k E − 1 2 log v ) + O ( v 3 log v ) ] + ( μ n , 1 v + μ n , 2 v 2 + O ( v 3 ) ) ⋅ [ − 1 2 + v 4 ( k E − 1 4 ) − v 8 log v + O ( v 2 log v ) ] = 2 λ n , 0 + v ( 2 λ n , 1 − λ n , 0 2 − μ n , 1 2 ) + v 2 ( 2 λ n , 2 − λ n , 1 2 + λ n , 0 8 k E − μ n , 2 2 + μ n , 1 4 ( k E − 1 4 ) ) + v 2 log v ( − λ n , 0 16 − μ n , 1 8 ) + O ( v 3 log v ) . (145) \begin{aligned}
\phi_n&=\lambda_n\chi+\mu_n\chi'\\
&=(\lambda_{n,0}+\lambda_{n,1}v+\lambda_{n,2}v^2+O(v^3))\cdot\left[2-\frac{v}{2}+\frac{v^2}{8}\left(k_E-\frac{1}{2}\log v\right)+O(v^3\log v)\right]\\
&\quad+(\mu_{n,1}v+\mu_{n,2}v^2+O(v^3))\cdot\left[-\frac{1}{2}+\frac{v}{4}\left(k_E-\frac{1}{4}\right)-\frac{v}{8}\log v+O(v^2\log v)\right]\\
&=2\lambda_{n,0}+v\left(2\lambda_{n,1}-\frac{\lambda_{n,0}}{2}-\frac{\mu_{n,1}}{2}\right)\\
&\quad+v^2\left(2\lambda_{n,2}-\frac{\lambda_{n,1}}{2}+\frac{\lambda_{n,0}}{8}k_E-\frac{\mu_{n,2}}{2}+\frac{\mu_{n,1}}{4}\left(k_E-\frac{1}{4}\right)\right)\\
&\quad+v^2\log v\left(-\frac{\lambda_{n,0}}{16}-\frac{\mu_{n,1}}{8}\right)+O(v^3\log v). \tag*{(145)}
\end{aligned} ϕ n = λ n χ + μ n χ ′ = ( λ n , 0 + λ n , 1 v + λ n , 2 v 2 + O ( v 3 )) ⋅ [ 2 − 2 v + 8 v 2 ( k E − 2 1 log v ) + O ( v 3 log v ) ] + ( μ n , 1 v + μ n , 2 v 2 + O ( v 3 )) ⋅ [ − 2 1 + 4 v ( k E − 4 1 ) − 8 v log v + O ( v 2 log v ) ] = 2 λ n , 0 + v ( 2 λ n , 1 − 2 λ n , 0 − 2 μ n , 1 ) + v 2 ( 2 λ n , 2 − 2 λ n , 1 + 8 λ n , 0 k E − 2 μ n , 2 + 4 μ n , 1 ( k E − 4 1 ) ) + v 2 log v ( − 16 λ n , 0 − 8 μ n , 1 ) + O ( v 3 log v ) . (145)
Since λ n , 0 = δ n , 0 \lambda_{n,0}=\delta_{n,0} λ n , 0 = δ n , 0 , the coefficient of v 0 v^{0} v 0 is 2 δ n , 0 2\delta_{n,0} 2 δ n , 0 , and by (B.19) the coefficient of v v v is − 1 2 δ n , 0 -\frac{1}{2}\delta_{n,0} − 2 1 δ n , 0 . We define X n X_n X n and X ~ n \widetilde{X}_n X n as the coefficients of v 2 v^2 v 2 and v 2 log v v^2\log v v 2 log v :
X n ≡ k E 8 λ n , 0 − 1 2 λ n , 1 + 2 λ n , 2 + ( k E 4 − 1 16 ) μ n , 1 − 1 2 μ n , 2 , (146) X_n \equiv\frac{k_E}{8}\lambda_{n,0}-\frac{1}{2}\lambda_{n,1}+2\lambda_{n,2}+\left(\frac{k_E}{4}-\frac{1}{16}\right)\mu_{n,1}-\frac{1}{2}\mu_{n,2}, \tag*{(146)} X n ≡ 8 k E λ n , 0 − 2 1 λ n , 1 + 2 λ n , 2 + ( 4 k E − 16 1 ) μ n , 1 − 2 1 μ n , 2 , (146)
X ~ n = − 1 16 λ n , 0 − 1 8 μ n , 1 . (147) \widetilde{X}_n=-\frac{1}{16}\lambda_{n,0}-\frac{1}{8}\mu_{n,1}. \tag*{(147)} X n = − 16 1 λ n , 0 − 8 1 μ n , 1 . (147)
Summing over n n n , we obtain
ϕ = 2 − 1 2 v 2 + v 2 ∑ n ≥ 0 ω − 2 n X n + v 2 log v ∑ n ≥ 0 ω − 2 n X ~ n + O ( v 3 log v ) . (148) \phi=2-\frac{1}{2}v^2+v^2\sum_{n\geq0}\omega^{-2n}X_n+v^2\log v\sum_{n\geq0}\omega^{-2n}\widetilde{X}_n+O(v^3\log v). \tag*{(148)} ϕ = 2 − 2 1 v 2 + v 2 n ≥ 0 ∑ ω − 2 n X n + v 2 log v n ≥ 0 ∑ ω − 2 n X n + O ( v 3 log v ) . (148)
With v = − ω 2 u v=-\omega^2u v = − ω 2 u and log v = log ( − ω 2 ) + log u \log v=\log(-\omega^2)+\log u log v = log ( − ω 2 ) + log u , this becomes
ϕ = 2 + 1 2 ω 2 u + ω 4 u 2 ∑ n ≥ 0 ω − 2 n [ X n + X ~ n log ( − ω 2 ) ] + ω 4 u 2 log u ∑ n ≥ 0 ω − 2 n X ~ n + O ( u 3 log u ) . (149) \phi=2+\frac{1}{2}\omega^2u+\omega^4u^2\sum_{n\geq0}\omega^{-2n}\left[X_n+\widetilde{X}_n\log(-\omega^2)\right]+\omega^4u^2\log u\sum_{n\geq0}\omega^{-2n}\widetilde{X}_n+O(u^3\log u). \tag*{(149)} ϕ = 2 + 2 1 ω 2 u + ω 4 u 2 n ≥ 0 ∑ ω − 2 n [ X n + X n log ( − ω 2 ) ] + ω 4 u 2 log u n ≥ 0 ∑ ω − 2 n X n + O ( u 3 log u ) . (149)
We compare this with ϕ = A ϕ A + B ϕ B \phi=\mathcal{A}\phi_A+\mathcal{B}\phi_B ϕ = A ϕ A + B ϕ B of § 3.2 , where ϕ A ≃ 1 \phi_A\simeq1 ϕ A ≃ 1 and ϕ B ≃ u 2 \phi_B\simeq u^2 ϕ B ≃ u 2 . The u 0 u^0 u 0 coefficient gives A = 2 \mathcal{A}=2 A = 2 . The terms in u u u and u 2 log u u^2\log u u 2 log u come only from ϕ A \phi_A ϕ A and are fixed by A \mathcal{A} A . The u 2 u^2 u 2 coefficient is B + A a 2 \mathcal{B}+\mathcal{A}a_2 B + A a 2 , where a 2 a_2 a 2 is the u 2 u^2 u 2 coefficient of ϕ A \phi_A ϕ A .
The condition ϕ A ≃ 1 \phi_A\simeq1 ϕ A ≃ 1 does not fix ϕ A \phi_A ϕ A uniquely, since ϕ A + c ϕ B \phi_A+c\phi_B ϕ A + c ϕ B also starts with 1. This changes a 2 a_2 a 2 by c c c and B \mathcal{B} B by − c A -c\mathcal{A} − c A , so G R = B / A G_R=\mathcal{B}/\mathcal{A} G R = B / A changes by − c -c − c . We choose a 2 a_2 a 2 to be a real polynomial in ω 2 \omega^2 ω 2 , so that G R G_R G R is fixed up to such polynomials. These are contact terms: since ∫ d t e i ω t δ ( k ) ( t ) = ( − i ω ) k \int dt\,e^{i\omega t}\delta^{(k)}(t)=(-i\omega)^k ∫ d t e iω t δ ( k ) ( t ) = ( − iω ) k , a polynomial in ω \omega ω is the Fourier transform of a combination of δ ( t ) \delta(t) δ ( t ) and its derivatives, which vanishes for t > 0 t>0 t > 0 . A real polynomial also has no imaginary part on the real axis. Hence, up to contact terms,
G R ( ω ) = B A = ω 4 2 ∑ n ≥ 0 ω − 2 n [ X n + X ~ n log ( − ω 2 ) ] . (150) G_R(\omega)=\frac{\mathcal{B}}{\mathcal{A}}=\frac{\omega^4}{2}\sum_{n\geq0}\omega^{-2n}\left[X_n+\widetilde{X}_n\log(-\omega^2)\right]. \tag*{(150)} G R ( ω ) = A B = 2 ω 4 n ≥ 0 ∑ ω − 2 n [ X n + X n log ( − ω 2 ) ] . (150)
Here X n X_n X n and X ~ n \widetilde{X}_n X n are independent of ω \omega ω . The X n X_n X n terms with n = 0 , 1 , 2 n=0,1,2 n = 0 , 1 , 2 are polynomials in ω 2 \omega^2 ω 2 , so they are also contact terms.
In the vacuum, H = F = 1 H=F=1 H = F = 1 and Q 1 = Q 2 = 0 Q_1=Q_2=0 Q 1 = Q 2 = 0 . The right-hand side of (B.9) then vanishes at every order, and ϕ n = 0 \phi_n=0 ϕ n = 0 for n ≥ 1 n\geq1 n ≥ 1 , so the vacuum correlator is the n = 0 n=0 n = 0 term of (B.25) alone. In the black brane, the n = 0 n=0 n = 0 term is the same, and the terms with n ≥ 1 n\geq1 n ≥ 1 are the corrections due to the background. For n = 0 n=0 n = 0 , we have λ 0 , 0 = 1 \lambda_{0,0}=1 λ 0 , 0 = 1 and μ 0 , 1 = 0 \mu_{0,1}=0 μ 0 , 1 = 0 , hence X ~ 0 = − 1 16 \widetilde{X}_0=-\frac{1}{16} X 0 = − 16 1 , and (B.25) contains − ω 4 32 log ( − ω 2 ) -\frac{\omega^4}{32}\log(-\omega^2) − 32 ω 4 log ( − ω 2 ) . For real ω > 0 \omega>0 ω > 0 , approached from Im ω > 0 \operatorname{Im}\omega>0 Im ω > 0 , we have log ( − ω 2 ) = log ω 2 − i π \log(-\omega^2)=\log\omega^2-i\pi log ( − ω 2 ) = log ω 2 − iπ , so Im G R = π ω 4 / 32 \operatorname{Im}G_R=\pi\omega^4/32 Im G R = π ω 4 /32 , which is ρ v a c \rho_{\mathrm{vac}} ρ vac of § 3.2 .
We now prove that X ~ n = 0 \widetilde{X}_n=0 X n = 0 for all n ≥ 1 n\geq1 n ≥ 1 and any charges. For n ≥ 1 n\geq1 n ≥ 1 , λ n , 0 = 0 \lambda_{n,0}=0 λ n , 0 = 0 and (B.19) give
X ~ n = − 1 8 μ n , 1 = − 1 2 λ n , 1 . (151) \widetilde{X}_n=-\frac{1}{8}\mu_{n,1}=-\frac{1}{2}\lambda_{n,1}. \tag*{(151)} X n = − 8 1 μ n , 1 = − 2 1 λ n , 1 . (151)
At d = 1 d=1 d = 1 , (B.18) reads 3 4 μ n , 2 = S n , 1 \frac{3}{4}\mu_{n,2}=S_{n,1} 4 3 μ n , 2 = S n , 1 and 2 λ n , 1 + 3 μ n , 2 = T n , 1 2\lambda_{n,1}+3\mu_{n,2}=T_{n,1} 2 λ n , 1 + 3 μ n , 2 = T n , 1 , so
λ n , 1 = 1 2 ( T n , 1 − 4 S n , 1 ) . (152) \lambda_{n,1}=\frac{1}{2}\left(T_{n,1}-4S_{n,1}\right). \tag*{(152)} λ n , 1 = 2 1 ( T n , 1 − 4 S n , 1 ) . (152)
The k k k -th term in S n S_n S n and T n T_n T n (B.13) carries the factor ( − v ) k (-v)^k ( − v ) k , so only k = 1 k=1 k = 1 contributes to S n , 1 S_{n,1} S n , 1 and T n , 1 T_{n,1} T n , 1 :
S n , 1 = − Q 1 , 1 ( λ n − 1 , 1 + 1 4 μ n − 1 , 1 ) − Q 2 , 1 λ n − 1 , 0 , T n , 1 = − Q 1 , 1 ( λ n − 1 , 0 + 2 μ n − 1 , 1 ) . (153) \begin{aligned}
S_{n,1}&=-Q_{1,1}\left(\lambda_{n-1,1}+\frac{1}{4}\mu_{n-1,1}\right)-Q_{2,1}\lambda_{n-1,0},\\
T_{n,1}&=-Q_{1,1}\left(\lambda_{n-1,0}+2\mu_{n-1,1}\right). \tag*{(153)}
\end{aligned} S n , 1 T n , 1 = − Q 1 , 1 ( λ n − 1 , 1 + 4 1 μ n − 1 , 1 ) − Q 2 , 1 λ n − 1 , 0 , = − Q 1 , 1 ( λ n − 1 , 0 + 2 μ n − 1 , 1 ) . (153)
Using (B.19) at order n − 1 n-1 n − 1 , the terms with λ n − 1 , 1 \lambda_{n-1,1} λ n − 1 , 1 and μ n − 1 , 1 \mu_{n-1,1} μ n − 1 , 1 cancel in T n , 1 − 4 S n , 1 T_{n,1}-4S_{n,1} T n , 1 − 4 S n , 1 , and (B.27) becomes
λ n , 1 = 1 2 ( 4 Q 2 , 1 − Q 1 , 1 ) λ n − 1 , 0 . (154) \lambda_{n,1}=\frac{1}{2}\left(4Q_{2,1}-Q_{1,1}\right)\lambda_{n-1,0}. \tag*{(154)} λ n , 1 = 2 1 ( 4 Q 2 , 1 − Q 1 , 1 ) λ n − 1 , 0 . (154)
For n ≥ 2 n\geq2 n ≥ 2 , this vanishes because λ n − 1 , 0 = 0 \lambda_{n-1,0}=0 λ n − 1 , 0 = 0 . For n = 1 n=1 n = 1 , we have λ 0 , 0 = 1 \lambda_{0,0}=1 λ 0 , 0 = 1 , but F ( 0 ) = H ( 0 ) = 1 F(0)=H(0)=1 F ( 0 ) = H ( 0 ) = 1 and F u ( 0 ) = H u ( 0 ) = ∑ i q i F_u(0)=H_u(0)=\sum_i q_i F u ( 0 ) = H u ( 0 ) = ∑ i q i give Q 1 , 1 = − ∑ i q i Q_{1,1}=-\sum_i q_i Q 1 , 1 = − ∑ i q i and Q 2 , 1 = − 1 4 ∑ i q i Q_{2,1}=-\frac{1}{4}\sum_i q_i Q 2 , 1 = − 4 1 ∑ i q i , so 4 Q 2 , 1 − Q 1 , 1 = 0 4Q_{2,1}-Q_{1,1}=0 4 Q 2 , 1 − Q 1 , 1 = 0 .
In conclusion, λ n , 1 = μ n , 1 = 0 \lambda_{n,1}=\mu_{n,1}=0 λ n , 1 = μ n , 1 = 0 and X ~ n = 0 \widetilde{X}_n=0 X n = 0 by (B.26) for all n ≥ 1 n\geq1 n ≥ 1 . No logarithm appears beyond the vacuum one. The terms with n = 1 , 2 n=1,2 n = 1 , 2 are then only contact terms, and the terms with n ≥ 3 n\geq3 n ≥ 3 are negative powers of ω \omega ω , which give the OPE coefficients through (B.31) below.
From G R ( ω ) G_R(\omega) G R ( ω ) to G R ( t ) G_R(t) G R ( t )
Contact terms do not affect G R ( t ) G_R(t) G R ( t ) at t > 0 t>0 t > 0 . The remaining terms of (B.25) are fixed by the behavior of G R ( t ) G_R(t) G R ( t ) near t = 0 t=0 t = 0 : by Watson’s lemma, the large-ω \omega ω expansion of G R ( ω ) = ∫ 0 ∞ d t e i ω t G R ( t ) G_R(\omega)=\int_0^\infty\mathrm{d}t\,e^{i\omega t}G_R(t) G R ( ω ) = ∫ 0 ∞ d t e iω t G R ( t ) is obtained by transforming the small-t t t expansion of G R ( t ) G_R(t) G R ( t ) term by term. For Im ω > 0 \operatorname{Im}\omega>0 Im ω > 0 and m ≥ 1 m\geq1 m ≥ 1 ,
∫ 0 ∞ d t e i ω t t 2 m − 1 = ( 2 m − 1 ) ! ( − i ω ) − 2 m = ( − 1 ) m ( 2 m − 1 ) ! ω − 2 m . (155) \int_0^\infty\mathrm{d}t\,e^{i\omega t}t^{2m-1}=(2m-1)!(-i\omega)^{-2m}=(-1)^m(2m-1)!\,\omega^{-2m}. \tag*{(155)} ∫ 0 ∞ d t e iω t t 2 m − 1 = ( 2 m − 1 )! ( − iω ) − 2 m = ( − 1 ) m ( 2 m − 1 )! ω − 2 m . (155)
so a term γ m t 2 m − 1 \gamma_m t^{2m-1} γ m t 2 m − 1 of G R ( t ) G_R(t) G R ( t ) gives ( − 1 ) m ( 2 m − 1 ) ! γ m ω − 2 m (-1)^m(2m-1)!\,\gamma_m\omega^{-2m} ( − 1 ) m ( 2 m − 1 )! γ m ω − 2 m . Since the term of (B.25) with n = m + 2 n=m+2 n = m + 2 is 1 2 X m + 2 ω − 2 m \frac{1}{2}X_{m+2}\omega^{-2m} 2 1 X m + 2 ω − 2 m , we can obtain γ m \gamma_m γ m from X m + 2 X_{m+2} X m + 2 as
γ m = ( − 1 ) m X m + 2 2 ( 2 m − 1 ) ! . (156) \gamma_m=\frac{(-1)^mX_{m+2}}{2(2m-1)!}. \tag*{(156)} γ m = 2 ( 2 m − 1 )! ( − 1 ) m X m + 2 . (156)
The integral of t − 5 t^{-5} t − 5 diverges at t = 0 t=0 t = 0 , so we shift the power of t t t by a small ε \varepsilon ε :
∫ 0 ∞ d t e i ω t t − 5 + ε = Γ ( − 4 + ε ) ( − i ω ) 4 − ε = ω 4 24 ε − ω 4 48 log ( − ω 2 ) + c o n s t ⋅ ω 4 + O ( ε ) . (157) \int_0^\infty\mathrm{d}t\,e^{i\omega t}t^{-5+\varepsilon}=\Gamma(-4+\varepsilon)(-i\omega)^{4-\varepsilon}=\frac{\omega^4}{24\varepsilon}-\frac{\omega^4}{48}\log(-\omega^2)+\mathrm{const}\cdot\omega^4+O(\varepsilon). \tag*{(157)} ∫ 0 ∞ d t e iω t t − 5 + ε = Γ ( − 4 + ε ) ( − iω ) 4 − ε = 24 ε ω 4 − 48 ω 4 log ( − ω 2 ) + const ⋅ ω 4 + O ( ε ) . (157)
The divergent term and the term c o n s t ⋅ ω 4 \mathrm{const}\cdot\omega^4 const ⋅ ω 4 are polynomials in ω \omega ω , so they are contact terms. A term c t − 5 ct^{-5} c t − 5 therefore gives − c 48 ω 4 log ( − ω 2 ) -\frac{c}{48}\omega^4\log(-\omega^2) − 48 c ω 4 log ( − ω 2 ) up to contact terms, and the vacuum logarithm − ω 4 32 log ( − ω 2 ) -\frac{\omega^{4}}{32}\log(-\omega^{2}) − 32 ω 4 log ( − ω 2 ) requires c = 3 2 c=\frac{3}{2} c = 2 3 . In the same way, t − 3 t^{-3} t − 3 and t − 1 t^{-1} t − 1 would give ω 2 log ( − ω 2 ) \omega^{2}\log(-\omega^{2}) ω 2 log ( − ω 2 ) and log ( − ω 2 ) \log(-\omega^{2}) log ( − ω 2 ) , the logarithms of n = 1 n=1 n = 1 and n = 2 n=2 n = 2 , and t 2 m − 1 log t t^{2m-1}\log t t 2 m − 1 log t would give ω − 2 m log ( − ω 2 ) \omega^{-2m}\log(-\omega^{2}) ω − 2 m log ( − ω 2 ) , the logarithm of n = m + 2 n=m+2 n = m + 2 . Since we proved X ~ n = 0 \widetilde{X}_{n}=0 X n = 0 for n ≥ 1 n\geq1 n ≥ 1 , none of these terms appear. Even powers of t t t would give odd powers of ω \omega ω , which do not appear either. Altogether, for t > 0 t>0 t > 0 ,
G R ( t ) = 3 2 t − 5 + ∑ m ≥ 1 γ m t 2 m − 1 . (158) G_R(t)=\frac{3}{2}t^{-5}+\sum_{m\geq1}\gamma_m t^{2m-1}. \tag*{(158)} G R ( t ) = 2 3 t − 5 + m ≥ 1 ∑ γ m t 2 m − 1 . (158)
For ( 3 , 1 , 1 5 ) \left(3,1,\frac{1}{5}\right) ( 3 , 1 , 5 1 ) the first orders are X 3 = − 116352 4375 X_3=-\frac{116352}{4375} X 3 = − 4375 116352 , X 4 = 5973504 21875 X_4=\frac{5973504}{21875} X 4 = 21875 5973504 and X 5 = 3523608576 109375 X_5=\frac{3523608576}{109375} X 5 = 109375 3523608576 , which give γ 1 = 58176 4375 \gamma_1=\frac{58176}{4375} γ 1 = 4375 58176 , γ 2 = 497792 21875 \gamma_2=\frac{497792}{21875} γ 2 = 21875 497792 and γ 3 = − 73408512 546875 \gamma_3=-\frac{73408512}{546875} γ 3 = − 546875 73408512 . For ( 2 , 1 2 , 0 ) \left(2,\frac{1}{2},0\right) ( 2 , 2 1 , 0 ) they are X 3 = − 458 63 X_3=-\frac{458}{63} X 3 = − 63 458 , X 4 = 5522 105 X_4=\frac{5522}{105} X 4 = 105 5522 and X 5 = 897472 231 X_5=\frac{897472}{231} X 5 = 231 897472 , which give γ 1 = 229 63 \gamma_1=\frac{229}{63} γ 1 = 63 229 , γ 2 = 2761 630 \gamma_2=\frac{2761}{630} γ 2 = 630 2761 and γ 3 = − 56092 3465 \gamma_3=-\frac{56092}{3465} γ 3 = − 3465 56092 .