Introduction
High energy partons propagating through the bulk matter produced in heavy-ion collisions undergo repeated interactions with this medium, acquiring transverse momentum and radiating additional partons. These processes modify the angular distribution of the hadronic final state energy flow, providing a way to probe the microscopic properties of the underlying matter. While substantial progress has been made in understanding medium modifications of jet substructure and energy correlators in collinear configurations in nuclear collisions [1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 ] , back-to-back topologies offer another setting in which medium modifications can be theoretically computed and experimentally tested through angular decorrelations, see e.g. [11 , 12 , 13 , 14 ] . These angular distributions are, however, not exclusively sensitive to medium effects, as they also receive significant contributions from perturbative and non-perturbative vacuum QCD processes, see discussions in [11 , 12 ] . Related studies have also explored forward dijet azimuthal correlations as a probe of medium-induced transverse-momentum broadening [15 ] . Therefore, developing a clear understanding of back-to-back configurations requires controlled observables that allow the relative size of the different physical effects to be assessed.
Transverse energy-energy correlators (TEECs) [16 , 17 , 18 ] provide a natural framework for this purpose. These observables measure azimuthal correlations between final-state hadronic transverse energy flows, and their inclusive definition does not require assigning the measured particles to individual jets, see Figure 1 . They can be understood as azimuthal projections of two-point functions of the energy flow detector operator [19 , 20 , 21 ] ,
Figure 1. Illustration of the event topology being considered. Two nearly back-to-back jets are produced from a large center of mass hard interaction, and their production and TEEC measurement are represented by Eq. (4) . Collinear hard radiation is indicated by pink lines, while the hadrons entering the measurement are identified by red lines with their respective transverse energies. Soft radiation is indicated by green gluon lines.
E ( n ⃗ ) = lim r → ∞ r 2 ∫ 0 ∞ d t n i T 0 i ( t , r n ⃗ ) , (1) \mathcal{E}(\vec{n}) = \lim_{r \to\infty} r^2 \int_0^\infty\mathrm{d}t\, n_i T^{0i}(t,r\vec{n}), \tag*{(1)} E ( n ) = r → ∞ lim r 2 ∫ 0 ∞ d t n i T 0 i ( t , r n ) , (1)
where T μ ν T^{\mu\nu} T μν is the energy momentum tensor, and n ⃗ \vec{n} n labels the detector position on a large r r r sphere. Its action on a state ∣ X ⟩ |X\rangle ∣ X ⟩ , E ( n ⃗ ) ∣ X ⟩ = ∑ a ∈ X E a δ ( 2 ) ( n ⃗ − n ⃗ a ) ∣ X ⟩ \mathcal{E}(\vec{n})|X\rangle= \sum_{a \in X} E_a \delta^{(2)}(\vec{n} - \vec{n}_a)|X\rangle E ( n ) ∣ X ⟩ = ∑ a ∈ X E a δ ( 2 ) ( n − n a ) ∣ X ⟩ , with the sum over single particle states, allows for clear experimental interpretation of the observable in terms of asymptotic calorimetric detectors, and as a result, the E ( n ⃗ ) \mathcal{E}(\vec{n}) E ( n ) operator provides a direct connection between quantum field theory and collider phenomenology. Taking the beam as the polar axis, one defines the total transverse-energy operator as E ^ T = ∫ d Ω sin θ E ( n ⃗ ) \widehat{E}_T = \int\mathrm{d}\Omega\,\sin\theta\,\mathcal{E}(\vec{n}) E T = ∫ d Ω sin θ E ( n ) , with E ^ T ( X ) ∣ X ⟩ = ∑ a ∈ X E a sin θ a ∣ X ⟩ \widehat{E}_T(X)|X\rangle= \sum_{a \in X} E_a\sin\theta_a|X\rangle E T ( X ) ∣ X ⟩ = ∑ a ∈ X E a sin θ a ∣ X ⟩ , such that the normalized TEEC distribution is then
1 σ d Σ d τ = ∫ d Ω 1 d Ω 2 sin θ 1 sin θ 2 ⟨ E ( n ⃗ 1 ) E ( n ⃗ 2 ) ⟩ σ ( ∑ i E T , i ) 2 δ ( τ − 1 + cos ϕ 12 2 ) = 1 σ ∑ X ∫ d σ X ∑ a , b ∈ X E T , a E T , b ( ∑ i E T , i ) 2 δ ( τ − 1 + cos ϕ a b 2 ) , (2) \begin{aligned}
\frac{1}{\sigma}\frac{\mathrm{d}\Sigma}{\mathrm{d}\tau} = \int\mathrm{d}\Omega_1\,\mathrm{d}\Omega_2\,\sin\theta_1\sin\theta_2\,\frac{\langle\mathcal{E}(\vec{n}_1)\mathcal{E}(\vec{n}_2)\rangle_\sigma}{\left(\sum_i E_{T,i}\right)^2}\,\delta\left(\tau-\frac{1+\cos\phi_{12}}{2}\right) \\
&= \frac{1}{\sigma}\sum_X\int\mathrm{d}\sigma_X\sum_{a,b\in X}\frac{E_{T,a}E_{T,b}}{\left(\sum_i E_{T,i}\right)^2}\,\delta\left(\tau-\frac{1+\cos\phi_{ab}}{2}\right), \tag*{(2)}
\end{aligned} σ 1 d τ d Σ = ∫ d Ω 1 d Ω 2 sin θ 1 sin θ 2 ( ∑ i E T , i ) 2 ⟨ E ( n 1 ) E ( n 2 ) ⟩ σ δ ( τ − 2 1 + cos ϕ 12 ) = σ 1 X ∑ ∫ d σ X a , b ∈ X ∑ ( ∑ i E T , i ) 2 E T , a E T , b δ ( τ − 2 1 + cos ϕ ab ) , (2)
where ϕ 12 ∈ [ 0 , π ] \phi_{12}\in[0,\pi] ϕ 12 ∈ [ 0 , π ] is the relative azimuthal angle, E T = ∑ i E T , i E_T=\sum_i E_{T,i} E T = ∑ i E T , i is the total transverse energy of the event, and we denote the expectation value as the cross section averaged over final states, i.e. ⟨ O ^ ⟩ σ ≡ 1 σ ∑ X ∫ d σ X O ( X ) \langle\widehat{O}\rangle_\sigma\equiv\frac{1}{\sigma}\sum_X\int\mathrm{d}\sigma_X\,O(X) ⟨ O ⟩ σ ≡ σ 1 ∑ X ∫ d σ X O ( X ) . Equation (2) provides three equivalent ways to compute the TEEC distribution, either from a cross section based formulation, from the correlation functions of flow operators, or from a particle based approach; we shall explore these different forms to compute distinct contributions to the observable, see [22 , 23 ] for further discussion on this aspect. Indeed, in the back-to-back limit, the cross section based formulation is particularly well suited to high energy QCD calculations, as it allows the separation of hard, collinear and soft dynamics within transverse momentum dependent (TMD) factorization and the resummation of large logarithms associated with small angular decorrelations [17 , 24 ] . This framework has been applied to EECs at leptonic colliders [24 ] and TEECs at hadronic colliders [17 ] and in deep-inelastic scattering [18 , 25 ] , reaching high perturbative accuracy [26 , 27 ] , with extensions to higher-point correlators [28 ] . The azimuthal dependence of back-to-back EECs has also been explored as a probe of TMD partonic structures [29 ] . Related developments have extended the study of back-to-back TEECs to the small-x x x regime at the EIC and to vector-boson-tagged hadron production in p p pp pp and p A pA p A collisions [30 , 31 ] . Phenomenological studies of transverse energy correlators in QCD have enabled the experimental determinations of the strong coupling and its scale dependence [32 ] .
In this work, we study the TEEC in the back-to-back limit, for a colour-singlet quark–antiquark pair produced in a hard process and propagating through a homogeneous background medium. This setup simplifies the case of realistic hadronic/nuclear scattering processes where nuclear initial state effects have to be considered, allowing us to isolate final state modifications. Under a set of additional assumptions specified below, we retain the vacuum factorized structure and incorporate medium effects through modifications to the functions entering the back-to-back TMD factorization formula. We combine the vacuum distribution at next-to-leading logarithmic prime (NLL′) accuracy with medium-induced corrections at first order in the strong coupling in the dense medium limit. Using the definition of the TEEC as a correlator of asymptotic detectors, we further estimate the non-perturbative contributions associated with the medium response to the dijet system, following [33 , 34 , 35 ] . For the TEEC, a response spread over a broad range of azimuthal angles is expected to vary slowly across a narrow back-to-back window. Its angular dependence can consequently differ from that of the perturbative contribution, which is concentrated near the back-to-back limit. Including both contributions allows their relative importance to be assessed as a function of the angular decorrelation, providing guidance for experimental studies. Numerical studies of the resulting TEEC distribution show that momentum broadening lowers the back-to-back peak, while radiative contributions produce a stronger suppression and enhance larger angular decorrelations. The broad medium response gives a slowly varying contribution whose relative importance grows away from the peak. These results illustrate the sensitivity of the TEEC to distinct components of the in-medium evolution. They also show that a quantitative interpretation of the peak requires control over medium-induced radiative corrections, which were neglected in previous studies of the back-to-back limit in heavy-ion collisions [11 , 12 ] .
The paper is organized as follows. Section II reviews the vacuum factorization and resummation of the back-to-back TEEC. Section III introduces the medium-modified jet functions and separates (classical) momentum broadening from the medium-induced radiative correction. The analytic treatment of these contributions is presented in Sec. IV. § 0.5 describes the contribution from the medium response. Numerical results are presented in Sec. VI. We summarize our findings and conclude in § 0.7 .
Back-to-Back limit of the TEEC in vacuum
In QCD, the TEEC can be defined in terms of the transverse energy weighted differential inclusive cross section for producing two outgoing hadrons from an initial hadronic collision [16 , 17 , 27 ] :
1 σ d Σ d τ = 1 σ ∑ a , b ∫ d σ A A → a + b + X E T , a E T , b ( ∑ i E T , i ) 2 δ ( τ − 1 + cos ϕ a b 2 ) (3) \frac{1}{\sigma}\frac{\mathrm{d}\Sigma}{\mathrm{d}\tau} = \frac{1}{\sigma}\sum_{a,b}\int\mathrm{d}\sigma_{AA\to a+b+X}\frac{E_{T,a}E_{T,b}}{\left(\sum_i E_{T,i}\right)^2}\delta\left(\tau-\frac{1+\cos\phi_{ab}}{2}\right) \tag*{(3)} σ 1 d τ d Σ = σ 1 a , b ∑ ∫ d σ AA → a + b + X ( ∑ i E T , i ) 2 E T , a E T , b δ ( τ − 2 1 + cos ϕ ab ) (3)
with ϕ a b \phi_{ab} ϕ ab the relative azimuthal angle between the measured hadrons, and E T , j E_{T,j} E T , j the transverse energy of hadrons j j j (see Figure 1 and Eq. (2) ). In the τ → 0 \tau\to0 τ → 0 back-to-back limit, and assuming a large center of mass energy, d σ A A → a + b + X \mathrm{d}\sigma_{AA\to a+b+X} d σ AA → a + b + X is described by the TMD factorization formula [17 , 27 , 36 ] , which at leading power gives the TEEC distribution
d Σ L P d τ ∝ tr [ H f 1 f 2 → f 3 f 4 S ] ⊗ B f 1 / N 1 ⊗ B f 2 / N 2 ⊗ J f 3 ⊗ J f 4 . (4) \frac{\mathrm{d}\Sigma^{\mathrm{LP}}}{\mathrm{d}\tau} \propto\operatorname{tr}\left[\mathbf{H}_{f_1f_2\to f_3f_4}\mathbf{S}\right]\otimes\mathbf{B}_{f_1/N_1}\otimes\mathbf{B}_{f_2/N_2}\otimes\mathbf{J}_{f_3}\otimes\mathbf{J}_{f_4}. \tag*{(4)} d τ d Σ LP ∝ tr [ H f 1 f 2 → f 3 f 4 S ] ⊗ B f 1 / N 1 ⊗ B f 2 / N 2 ⊗ J f 3 ⊗ J f 4 . (4)
The beam (B B B ), hard (H H H ), soft (S S S ), and jet (J J J ) functions depend on the factorization scale μ \mu μ , while the beam, soft, and jet functions also depend on the rapidity scale ν \nu ν , see [17 ] for further details. In the context of heavy ion collisions, Eq. (4) can be employed to describe the production of high energy hadrons/jets which survive the propagation through the plasma at a high intrinsic momentum scale [12 ] (see e.g. [37 , 38 , 39 ] for recent related works). Nonetheless, as the presence of the bulk matter introduces new scales, one is required to compute medium dependent corrections to the different functions entering the above factorization formula. Of course such a computation is only sensible assuming that Eq. (4) still holds in heavy ion collisions, which is a reasonable phenomenological assumption if we consider high p T p_T p T dijets.
The origins of the modifications to the different functions entering the factorization formula are rather distinct. Corrections to the beam and hard functions will enter via cold nuclear matter effects, and shall not be discussed further in this work. In contrast, the soft factor, due to the production of stimulated soft bremsstrahlung radiation, and the jet function, where hard/collinear medium-induced branching contributions change the jet fragmentation pattern, are modified due to the late time evolution of the jet cascade. These modifications can be computed within a perturbative QCD picture, and Sec. III specifies which of them are included in this work. The full calculation of all these corrections, in realistic heavy-ion environments, can only be performed using fully numerical methods (see e.g. [40 , 41 ] for recent examples). However, qualitative effects of bulk dependent physics can be explored using some simplifying assumptions, which we discuss below, while the structure of the corrections being computed remains generic.
In this work, as we focus on resolving the effects on final state fragmentation, we consider a simplified idealized scenario of an e + e − e^{+}e^{-} e + e − collision occurring in the presence of a bulk background state, representing the quark-gluon plasma produced in heavy-ion collisions. The vacuum process is e + e − → γ ∗ → q q ˉ + X e^{+}e^{-} \to\gamma^{*} \to q\bar{q} + X e + e − → γ ∗ → q q ˉ + X for massless quarks, with Q = s Q = \sqrt{s} Q = s . With transverse energies measured relative to the lepton beam axis z ^ \hat{z} z ^ , the TEEC is written as
1 σ d Σ d τ = 1 σ ∑ a , b ∫ d σ γ ∗ → a + b + X E T , a E T , b ( ∑ i E T , i ) 2 δ ( τ − 1 + cos ϕ a b 2 ) . (5) \frac{1}{\sigma}\frac{d\Sigma}{d\tau} = \frac{1}{\sigma}\sum_{a,b}\int d\sigma_{\gamma^{*}\to a+b+X}\frac{E_{T,a}E_{T,b}}{\left(\sum_i E_{T,i}\right)^2}\delta\left(\tau-\frac{1+\cos\phi_{ab}}{2}\right). \tag*{(5)} σ 1 d τ d Σ = σ 1 a , b ∑ ∫ d σ γ ∗ → a + b + X ( ∑ i E T , i ) 2 E T , a E T , b δ ( τ − 2 1 + cos ϕ ab ) . (5)
At Born level, we denote the momenta of the final state quark and antiquark by p ⃗ 0 \vec{p}_0 p 0 and − p ⃗ 0 -\vec{p}_0 − p 0 , with ∣ p ⃗ 0 ∣ = Q / 2 |\vec{p}_0| = Q/2 ∣ p 0 ∣ = Q /2 , write c = cos θ 0 c = \cos\theta_0 c = cos θ 0 for the polar angle of the quark with respect to the beam, and
p T = Q 2 sin θ 0 , n ^ = p ⃗ 0 ∣ p ⃗ 0 ∣ , y ^ = z ^ × n ^ ∣ z ^ × n ^ ∣ . (6) p_T = \frac{Q}{2}\sin\theta_0,\qquad\hat{n} = \frac{\vec{p}_0}{|\vec{p}_0|},\qquad\hat{y} = \frac{\hat{z}\times\hat{n}}{|\hat{z}\times\hat{n}|}. \tag*{(6)} p T = 2 Q sin θ 0 , n ^ = ∣ p 0 ∣ p 0 , y ^ = ∣ z ^ × n ^ ∣ z ^ × n ^ . (6)
The TEEC is defined with finite jet p T p_T p T where sin θ 0 \sin\theta_0 sin θ 0 is of order one in the power expansion. The quantity p T p_T p T labels the parton in leading order of the reconstructed jet transverse momentum, since to define the TEEC two leading jets are required in the event. For a photon exchange, d σ B o r n / d c = σ B o r n B ( c ) d\sigma_{\mathrm{Born}}/dc = \sigma_{\mathrm{Born}}B(c) d σ Born / d c = σ Born B ( c ) , with B ( c ) = 3 ( 1 + c 2 ) / 8 B(c) = 3(1+c^2)/8 B ( c ) = 3 ( 1 + c 2 ) /8 , and we normalize the distribution at fixed c c c to it,
1 σ d Σ d τ = ( d σ B o r n d c ) − 1 d 2 Σ L P d c d τ , \frac{1}{\sigma}\frac{d\Sigma}{d\tau} = \left(\frac{d\sigma_{\mathrm{Born}}}{dc}\right)^{-1}\frac{d^2\Sigma_{\mathrm{LP}}}{dc\,d\tau}, σ 1 d τ d Σ = ( d c d σ Born ) − 1 d c d τ d 2 Σ LP ,
that is, the TEEC of the events at fixed c c c , normalized to their leading power cross section d σ B o r n / d c d\sigma_{\mathrm{Born}}/dc d σ Born / d c . At c = 0 c = 0 c = 0 , hence p T = Q / 2 p_T = Q/2 p T = Q /2 , this conditional kernel is finite, while a density in p T 2 p_T^2 p T 2 instead contains the angular Jacobian. From here on, distributions are at fixed c c c , equivalently fixed p T p_T p T , and the figures use c = 0 c=0 c = 0 , that is p T = Q / 2 p_T=Q/2 p T = Q /2 with both jets at zero rapidity.
We define n μ = ( 1 , n ^ ) n^\mu=(1,\hat n) n μ = ( 1 , n ^ ) , n ˉ μ = ( 1 , − n ^ ) \bar n^\mu=(1,-\hat n) n ˉ μ = ( 1 , − n ^ ) along the direction of final state jets, so that n ⋅ n ˉ = 2 n\cdot\bar n=2 n ⋅ n ˉ = 2 , and a small expansion parameter λ ∼ q y / Q ∼ τ \lambda\sim q_y/Q\sim\sqrt{\tau} λ ∼ q y / Q ∼ τ with q y = 2 p T τ q_y=2p_T\sqrt{\tau} q y = 2 p T τ . The relevant vacuum modes have the SCETI I _{\mathrm{II}} II scaling [42 , 43 ]
( n ⋅ p , n ˉ ⋅ p , p ⊥ ) ∼ { Q ( λ 2 , 1 , λ ) , n -collinear , Q ( 1 , λ 2 , λ ) , n ˉ -collinear , Q ( λ , λ , λ ) , soft . (7) (n\cdot p,\bar n\cdot p,p_\perp)\sim
\begin{cases}
Q(\lambda^2,1,\lambda), & n\text{-collinear},\\
Q(1,\lambda^2,\lambda), & \bar n\text{-collinear},\\
Q(\lambda,\lambda,\lambda), & \text{soft}.
\end{cases} \tag*{(7)} ( n ⋅ p , n ˉ ⋅ p , p ⊥ ) ∼ ⎩ ⎨ ⎧ Q ( λ 2 , 1 , λ ) , Q ( 1 , λ 2 , λ ) , Q ( λ , λ , λ ) , n -collinear , n ˉ -collinear , soft . (7)
The collinear and soft modes therefore have the same characteristic virtuality q y 2 q_y^2 q y 2 but different rapidities. Their separation thus requires both virtuality and rapidity renormalization scales, denoted by μ \mu μ and ν \nu ν . For hadrons in the two collinear sectors, at Born level, the momentum fractions of final state hadrons with respect to their parent partons are x = 2 E a / Q x=2E_a/Q x = 2 E a / Q and x ˉ = 2 E b / Q \bar x=2E_b/Q x ˉ = 2 E b / Q . To leading power, E T , a = x p T E_{T,a}=xp_T E T , a = x p T , E T , b = x ˉ p T E_{T,b}=\bar x p_T E T , b = x ˉ p T and ∑ i E T , i = 2 p T \sum_i E_{T,i}=2p_T ∑ i E T , i = 2 p T , so their pair weight is x x ˉ / 2 x\bar x/2 x x ˉ /2 . Let ℓ \ell ℓ and ℓ ˉ \bar\ell ℓ ˉ be the fragmentation momenta relative to the collinear sectors after separating their total soft recoil, and let k s , y k_{s,y} k s , y be the total soft momentum in the measured direction, so that k a , y = ℓ y / x k_{a,y}=\ell_y/x k a , y = ℓ y / x and k b , y = ℓ ˉ y / x ˉ k_{b,y}=\bar\ell_y/\bar x k b , y = ℓ ˉ y / x ˉ . For ϕ a b = π − δ ϕ \phi_{ab}=\pi-\delta\phi ϕ ab = π − δ ϕ the measurement at leading power is
τ = δ ϕ 2 4 + O ( δ ϕ 4 ) , τ ∣ L P = ( k a , y + k b , y − k s , y ) 2 4 p T 2 . (8) \tau=\frac{\delta\phi^2}{4}+\mathcal{O}(\delta\phi^4),\qquad
\left.\tau\right|_{\mathrm{LP}}=\frac{(k_{a,y}+k_{b,y}-k_{s,y})^2}{4p_T^2}. \tag*{(8)} τ = 4 δ ϕ 2 + O ( δ ϕ 4 ) , τ ∣ LP = 4 p T 2 ( k a , y + k b , y − k s , y ) 2 . (8)
In the τ → 0 \tau\to0 τ → 0 limit, a TMD factorization analogous to the one alluded for hadronic collisions exists [24 ] :
1 σ d Σ d τ = p T τ ∫ − ∞ ∞ d b 2 π e − 2 i b p T τ H ( Q , μ ) J q ( b , Q ; μ , ν ) J q ˉ ( b , Q ; μ , ν ) S ( b ; μ , ν ) , (9) \frac{1}{\sigma}\frac{\mathrm{d}\Sigma}{\mathrm{d}\tau}
=\frac{p_T}{\sqrt{\tau}}\int_{-\infty}^{\infty}\frac{\mathrm{d}b}{2\pi}e^{-2ibp_T\sqrt{\tau}}H(Q,\mu)J_q(b,Q;\mu,\nu)J_{\bar q}(b,Q;\mu,\nu)S(b;\mu,\nu), \tag*{(9)} σ 1 d τ d Σ = τ p T ∫ − ∞ ∞ 2 π d b e − 2 ib p T τ H ( Q , μ ) J q ( b , Q ; μ , ν ) J q ˉ ( b , Q ; μ , ν ) S ( b ; μ , ν ) , (9)
where the functions of b b b are evaluated along y ^ \hat y y ^ , as in Eq. (14).1 As the soft function is not affected by the measurement, we shall focus on the energy weighted inclusive jet function, which can be written as
J q ( b , Q ; μ , ν ) = ∑ h ∫ 0 1 d x x ∫ ℓ e i b ⋅ ℓ / x F q → h ( x , ℓ ; μ , ν ) = ∑ h ∫ 0 1 d x x F ~ q → h ( x , b / x ; μ , ν ) , (10) J_q(b,Q;\mu,\nu)=\sum_h\int_0^1\mathrm{d}x\,x\int_{\ell}e^{ib\cdot\ell/x}F_{q\to h}(x,\ell;\mu,\nu)=\sum_h\int_0^1\mathrm{d}x\,x\,\widetilde F_{q\to h}(x,b/x;\mu,\nu), \tag*{(10)} J q ( b , Q ; μ , ν ) = h ∑ ∫ 0 1 d x x ∫ ℓ e ib ⋅ ℓ / x F q → h ( x , ℓ ; μ , ν ) = h ∑ ∫ 0 1 d x x F q → h ( x , b / x ; μ , ν ) , (10)
and by charge conjugation J q ˉ = J q J_{\bar q}=J_q J q ˉ = J q . Here we have defined the jet function in terms of the TMD fragmentation function (TMDFF) F i → h ( x , ℓ ; μ , ν ) F_{i\to h}(x,\ell;\mu,\nu) F i → h ( x , ℓ ; μ , ν ) , which describes a hadron carrying fraction x x x and transverse momentum ℓ \ell ℓ relative to its parent parton. This can be perturbatively matched to the collinear fragmentation function
F i → a ( x , ℓ ) = ∑ j ∫ d ζ ζ 3 I i j ( x ζ , ℓ ζ ) f j → a ( ζ , μ c ) + O ( Λ Q C D 2 ℓ 2 ) , (11) F_{i\to a}(x,\ell)=\sum_j\int\frac{d\zeta}{\zeta^3}I_{ij}\left(\frac{x}{\zeta},\frac{\ell}{\zeta}\right)f_{j\to a}(\zeta,\mu_c)+\mathcal{O}\left(\frac{\Lambda_{\mathrm{QCD}}^2}{\ell^2}\right), \tag*{(11)} F i → a ( x , ℓ ) = j ∑ ∫ ζ 3 d ζ I ij ( ζ x , ζ ℓ ) f j → a ( ζ , μ c ) + O ( ℓ 2 Λ QCD 2 ) , (11)
such that, using x = z ζ x=z\zeta x = z ζ , ℓ = ζ k \ell=\zeta k ℓ = ζ k and the momentum sum rule ∑ h ∫ 0 1 d ζ ζ f j → h ( ζ , μ ) = 1 \sum_h\int_0^1d\zeta\,\zeta f_{j\to h}(\zeta,\mu)=1 ∑ h ∫ 0 1 d ζ ζ f j → h ( ζ , μ ) = 1 , one finds
J q ( b , Q ; μ , ν ) = ∑ j ∫ 0 1 d z z ∫ k e i b ⋅ k / z I q j ( z , k ; μ , ν ) = ∑ j ∫ 0 1 d z z I ~ q j ( z , b / z ; μ , ν ) , (12) J_q(b,Q;\mu,\nu)=\sum_j\int_0^1 dz\,z\int_k e^{ib\cdot k/z}I_{qj}(z,k;\mu,\nu)=\sum_j\int_0^1 dz\,z\,\widetilde{I}_{qj}(z,b/z;\mu,\nu), \tag*{(12)} J q ( b , Q ; μ , ν ) = j ∑ ∫ 0 1 d z z ∫ k e ib ⋅ k / z I q j ( z , k ; μ , ν ) = j ∑ ∫ 0 1 d z z I q j ( z , b / z ; μ , ν ) , (12)
Thus, the jet function is a moment of the matching coefficient due to the momentum sum rule. This reflects the inclusive nature of the energy flow measurement, in contrast with the particle based observables, which would depend on the full distribution. Combining this with Eq. (10) , we find the simplified form2
1 σ d Σ d τ = H ( Q , μ ) 2 ∑ h , h ˉ ∫ 0 1 d x d x ˉ x x ˉ ∫ ℓ , ℓ ˉ , k s F q → h ( x , ℓ ; μ , ν ) F q ˉ → h ˉ ( x ˉ , ℓ ˉ ; μ , ν ) S ( k s ; μ , ν ) × δ ( τ − ( ℓ y / x + ℓ ˉ y / x ˉ − k s , y ) 2 4 p T 2 ) = H ( Q , μ ) 2 ∑ h , h ˉ ∫ 0 1 d x ∫ 0 1 d x ˉ x x ˉ ∫ − ∞ + ∞ d b 2 π 2 p T τ cos ( 2 b p T τ ) × F ~ q → h ( x , b x y ^ ; μ , ν ) F ~ q ˉ → h ˉ ( x ˉ , b x ˉ y ^ ; μ , ν ) S ~ ( − b y ^ ; μ , ν ) . (13) \begin{aligned}
\frac{1}{\sigma}\frac{d\Sigma}{d\tau}=\frac{H(Q,\mu)}{2}\sum_{h,\bar h}\int_0^1dx\,d\bar{x}\,x\bar{x}\int_{\ell,\bar{\ell},k_s}F_{q\to h}(x,\ell;\mu,\nu)F_{\bar q\to\bar h}(\bar{x},\bar{\ell};\mu,\nu)S(k_s;\mu,\nu) \\
&\times\delta\left(\tau-\frac{(\ell_y/x+\bar{\ell}_y/\bar{x}-k_{s,y})^2}{4p_T^2}\right) \\
&=\frac{H(Q,\mu)}{2}\sum_{h,\bar h}\int_0^1dx\int_0^1d\bar{x}\,x\bar{x}\int_{-\infty}^{+\infty}\frac{db}{2\pi}\frac{2p_T}{\sqrt{\tau}}\cos(2bp_T\sqrt{\tau}) \\
&\times\widetilde{F}_{q\to h}\left(x,\frac{b}{x}\hat{y};\mu,\nu\right)\widetilde{F}_{\bar q\to\bar h}\left(\bar{x},\frac{b}{\bar{x}}\hat{y};\mu,\nu\right)\widetilde{S}(-b\hat{y};\mu,\nu). \tag*{(13)}
\end{aligned} σ 1 d τ d Σ = 2 H ( Q , μ ) h , h ˉ ∑ ∫ 0 1 d x d x ˉ x x ˉ ∫ ℓ , ℓ ˉ , k s F q → h ( x , ℓ ; μ , ν ) F q ˉ → h ˉ ( x ˉ , ℓ ˉ ; μ , ν ) S ( k s ; μ , ν ) × δ ( τ − 4 p T 2 ( ℓ y / x + ℓ ˉ y / x ˉ − k s , y ) 2 ) = 2 H ( Q , μ ) h , h ˉ ∑ ∫ 0 1 d x ∫ 0 1 d x ˉ x x ˉ ∫ − ∞ + ∞ 2 π d b τ 2 p T cos ( 2 b p T τ ) × F q → h ( x , x b y ^ ; μ , ν ) F q ˉ → h ˉ ( x ˉ , x ˉ b y ^ ; μ , ν ) S ( − b y ^ ; μ , ν ) . (13)
The dependence on the transverse momenta ℓ \ell ℓ , ℓ ˉ \bar{\ell} ℓ ˉ , and k s k_s k s has been factorized into a product of Fourier-transformed TMD fragmentation functions and the soft function, while the hard factor has a trivial convolution structure. Since only the y ^ \hat{y} y ^ components are measured, the remaining integral is over the impact parameter b b b along y ^ \hat{y} y ^ , and its kernel cos ( 2 b p T τ ) \cos(2bp_T\sqrt{\tau}) cos ( 2 b p T τ ) implements the constraint on the measured variable τ \tau τ .
The resummation in the TMD limit is achieved by solving the renormalization group (RG) equations of the hard, jet, and soft functions:
d ln H d ln μ = 2 Γ ln Q 2 μ 2 + 2 γ H , d ln J q d ln μ = − Γ ln Q 2 ν 2 + 2 γ J , d ln S d ln μ = 2 Γ ln μ 2 ν 2 − 2 γ S . (14) \begin{aligned}
\frac{d\ln H}{d\ln\mu}&=2\Gamma\ln\frac{Q^2}{\mu^2}+2\gamma_H,\\
\frac{d\ln J_q}{d\ln\mu}&=-\Gamma\ln\frac{Q^2}{\nu^2}+2\gamma_J,\\
\frac{d\ln S}{d\ln\mu}&=2\Gamma\ln\frac{\mu^2}{\nu^2}-2\gamma_S. \tag*{(14)}
\end{aligned} d ln μ d ln H d ln μ d ln J q d ln μ d ln S = 2Γ ln μ 2 Q 2 + 2 γ H , = − Γ ln ν 2 Q 2 + 2 γ J , = 2Γ ln ν 2 μ 2 − 2 γ S . (14)
RG consistency requires γ H + 2 γ J − γ s = 0 \gamma_H + 2\gamma_J - \gamma_s = 0 γ H + 2 γ J − γ s = 0 . The RG evolution in the rapidity scale reads
A ( b , μ ) = ∫ μ b 2 μ 2 d m 2 m 2 Γ ( α s ( m ) ) − γ r ( α s ( μ b ) ) , \mathcal{A}(b,\mu) = \int_{\mu_b^2}^{\mu^2} \frac{d m^2}{m^2} \Gamma(\alpha_s(m)) - \gamma_r(\alpha_s(\mu_b)), A ( b , μ ) = ∫ μ b 2 μ 2 m 2 d m 2 Γ ( α s ( m )) − γ r ( α s ( μ b )) ,
d ln J q d ln ν = A ( b , μ ) , d ln S d ln ν = − 2 A ( b , μ ) . (15) \frac{d\ln J_q}{d\ln\nu} = \mathcal{A}(b,\mu), \qquad\frac{d\ln S}{d\ln\nu} = -2\mathcal{A}(b,\mu). \tag*{(15)} d ln ν d ln J q = A ( b , μ ) , d ln ν d ln S = − 2 A ( b , μ ) . (15)
with μ b = b 0 / b \mu_b = b_0/b μ b = b 0 / b , b 0 = 2 e − γ E b_0 = 2e^{-\gamma_E} b 0 = 2 e − γ E . Since the primary goal of this work is to investigate medium-induced effects, we perform the vacuum resummation at next-to-leading logarithmic prime (NLL′ ^\prime ′ ) accuracy, retaining the cusp anomalous dimension through two loops, the noncusp anomalous dimensions at one loop, and the one-loop remainders in the functions. We evaluate the perturbative functions at the profiled transverse separation b ˉ \bar{b} b ˉ , defined by
b ∗ = b 1 + b 2 / b max 2 , b ˉ = b ∗ 2 + b 0 2 Q 2 , μ ˉ b = b 0 b ˉ . (16) b_* = \frac{b}{\sqrt{1+b^2/b_{\max}^2}}, \qquad\bar{b} = \sqrt{b_*^2+\frac{b_0^2}{Q^2}}, \qquad\bar{\mu}_b = \frac{b_0}{\bar{b}}. \tag*{(16)} b ∗ = 1 + b 2 / b m a x 2 b , b ˉ = b ∗ 2 + Q 2 b 0 2 , μ ˉ b = b ˉ b 0 . (16)
where b max = b 0 G e V − 1 b_{\max} = b_0\,\mathrm{GeV}^{-1} b m a x = b 0 GeV − 1 [45 ] . The b ∗ b_* b ∗ prescription freezes the perturbative scale at large b b b , and the additional modification at small b b b ensures μ ˉ b ≤ Q \bar{\mu}_b \le Q μ ˉ b ≤ Q . The Fourier phase retains the physical impact parameter b b b . Our central scale choices are
μ H = ν J = Q , μ J = μ s ≡ μ = μ ˉ b , ν s = μ b . (17) \mu_H = \nu_J = Q, \qquad\mu_J = \mu_s \equiv\mu= \bar{\mu}_b, \qquad\nu_s = \mu_b. \tag*{(17)} μ H = ν J = Q , μ J = μ s ≡ μ = μ ˉ b , ν s = μ b . (17)
The invariant mass Q Q Q sets the hard scale, and for Born level jets perpendicular to the beam, Q = 2 p T Q = 2p_T Q = 2 p T . We use fixed n f = 5 n_f = 5 n f = 5 and α s ( m Z ) = 0.118 \alpha_s(m_Z) = 0.118 α s ( m Z ) = 0.118 , with m Z = 91.1876 G e V m_Z = 91.1876\,\mathrm{GeV} m Z = 91.1876 GeV . The coupling is held constant on any part of an evolution interval below 1 GeV, while the scale logarithms and evolution intervals are retained. We estimate perturbative uncertainties by varying μ H \mu_H μ H , the common scale μ \mu μ , and ν s \nu_s ν s independently by factors of two about Eq. (17) . Each variation changes a single scale, with the remaining scales at their central values and ν J = Q \nu_J = Q ν J = Q fixed. The band is the envelope of these six variations and the central prediction.
Non-perturbative effects enter as a multiplicative factor in the impact parameter integrand. We adopt the functional form of Fit 1 in [45 ] ,
F N P ( b , Q ) = exp [ − 2 a 1 b a 2 − 4 D N P ( b ) ln Q μ ˉ b ] , D N P ( b ) = g 2 b b ∗ , (18) F_{\mathrm{NP}}(b,Q) = \exp\left[-2a_1b^{a_2} - 4D_{\mathrm{NP}}(b)\ln\frac{Q}{\bar{\mu}_b}\right], \qquad D_{\mathrm{NP}}(b) = g_2bb_*, \tag*{(18)} F NP ( b , Q ) = exp [ − 2 a 1 b a 2 − 4 D NP ( b ) ln μ ˉ b Q ] , D NP ( b ) = g 2 b b ∗ , (18)
with a 1 = 0.530 G e V a 2 a_1 = 0.530\,\mathrm{GeV}^{a_2} a 1 = 0.530 GeV a 2 , a 2 = 1.152 a_2 = 1.152 a 2 = 1.152 and g 2 = 0.076 G e V 2 g_2 = 0.076\,\mathrm{GeV}^2 g 2 = 0.076 GeV 2 . The first term in the exponent accounts for the two jet factors, each parametrized by exp [ − S N P ( b ) ] \exp[-S_{\mathrm{NP}}(b)] exp [ − S NP ( b )] with S N P ( b ) = a 1 b a 2 S_{\mathrm{NP}}(b) = a_1b^{a_2} S NP ( b ) = a 1 b a 2 . The second follows from Collins–Soper evolution between ζ i = μ ˉ b 2 \zeta_i = \bar{\mu}_b^2 ζ i = μ ˉ b 2 and ζ f = Q 2 \zeta_f = Q^2 ζ f = Q 2 , in the convention K N P = − 2 D N P K_{\mathrm{NP}} = -2D_{\mathrm{NP}} K NP = − 2 D NP . The parametrization D N P ∝ b b ∗ D_{\mathrm{NP}} \propto bb_* D NP ∝ b b ∗ is also used in [46 ] .
Final state medium modifications to the TEEC
In this section, we construct the medium-modified jet functions that enter the back-to-back TEEC, in the simplified e + e − e^+e^- e + e − configuration of Sec. II with a medium traversed by the outgoing quark and antiquark. In Sec. III A the jet function is matched onto collinear fragmentation functions. Section III B gives the classical broadening term, and Sec. III C gives the q → q g q \to qg q → q g cross section for dense matter. That subsection also splits the O ( α s ) \mathcal{O}(\alpha_s) O ( α s ) correction into time regions, separating a broadened vacuum term from a medium-induced radiative correction, discusses the large ∣ p ∣ |\mathbf{p}| ∣ p ∣ behaviour of the latter and introduces the numerical regulator used in Sec. IV.
We make the following assumptions on the medium dependence of the factorization in Eq. (13) , see, e.g., [37 , 38 , 39 , 47 , 48 ] for related approaches to in-medium factorization.
Medium dependence only in the final state TMDFFs. The hard and soft functions are evaluated in vacuum. A medium correction to the soft function would require the wide-angle radiation off a back-to-back colour dipole, which lies outside the collinear regime considered below. In particular, the recoil of the collinear partons against soft medium-induced gluons is not included. As we discuss in Sec. III C, the attribution of the medium modifications to the jet function has some degree of ambiguity, but the final distributions are independent of this choice at the order considered.
Medium geometry. The medium is transversely homogeneous, and its momentum-transfer distribution is isotropic in the plane transverse to the parent direction. Both parents traverse the same length L + L^+ L + , so that the quark and antiquark jet functions carry the same medium parameters. In particular, the medium is invariant under reflection through the plane spanned by z ^ \hat{\mathbf{z}} z ^ and p ⃗ 0 \vec{\mathbf{p}}_0 p 0 . This idealization sets aside effects of more general medium structure and evolution on jet propagation (see, e.g., [51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ] ).
To keep the transverse recoil of the collinear sector explicit in the TMD matching, we introduce a parent-resolved TMDFF F q → a ( x , ℓ , K ; p 0 ) F_{q\to a}(x,\ell,\mathbf{K};p_0) F q → a ( x , ℓ , K ; p 0 ) , where p 0 p_0 p 0 is the four-momentum of the parent quark, with three-momentum p ⃗ 0 \vec{p}_0 p 0 , and
K = P − p 0 , ℓ = p a − x P . (19) \mathbf{K}=\mathbf{P}-\mathbf{p}_0,\qquad\ell=\mathbf{p}_a-x\mathbf{P}. \tag*{(19)} K = P − p 0 , ℓ = p a − x P . (19)
Here P \mathbf{P} P is the final total transverse momentum of the collinear sector, K \mathbf{K} K its recoil relative to the original parent, and ℓ \ell ℓ the momentum of the measured particle relative to the final collinear sector. From here on, boldface vectors are transverse to the fixed light-cone direction n n n of Eq. (7) . All partons of the sector, including the parent, are collinear to n n n , ∣ p ⊥ ∣ ∼ λ p + |\mathbf{p}_{\perp}|\sim\lambda p^+ ∣ p ⊥ ∣ ∼ λ p + with λ ≪ 1 \lambda\ll1 λ ≪ 1 . This is the power counting of the BDMPS–Z formalism of Sec. III C, in which the medium is described along n n n . It fixes n n n only up to rotations by angles of order λ \lambda λ , which leave the jet function unchanged in a transversely homogeneous medium. Since y ^ \hat{y} y ^ is orthogonal to the beam and to n ^ \hat{n} n ^ , the y ^ \hat{y} y ^ components are those of Sec. II. In vacuum, P = p 0 \mathbf{P}=\mathbf{p}_0 P = p 0 , so the parent direction and the jet axis coincide and the parent-resolved TMDFF reduces to the usual one,
F q → a v a c ( x , ℓ , K ; p 0 ) = F q → a v a c ( x , ℓ ) ( 2 π ) 2 δ ( 2 ) ( K ) . (20) F^{\mathrm{vac}}_{q\to a}(x,\ell,\mathbf{K};p_0)=F^{\mathrm{vac}}_{q\to a}(x,\ell)(2\pi)^2\delta^{(2)}(\mathbf{K}). \tag*{(20)} F q → a vac ( x , ℓ , K ; p 0 ) = F q → a vac ( x , ℓ ) ( 2 π ) 2 δ ( 2 ) ( K ) . (20)
In the medium, the jet axis is deflected from the parent direction by K \mathbf{K} K . In terms of these variables the measured displacements are generalized such that Eq. (8) becomes
τ = 1 4 p T 2 ( K y + ℓ y x + K ˉ y + ℓ ˉ y x ˉ − k s , y ) 2 . (21) \tau=\frac{1}{4p_T^2}\left(K_y+\frac{\ell_y}{x}+\bar{K}_y+\frac{\bar{\ell}_y}{\bar{x}}-k_{s,y}\right)^2. \tag*{(21)} τ = 4 p T 2 1 ( K y + x ℓ y + K ˉ y + x ˉ ℓ ˉ y − k s , y ) 2 . (21)
Thus, the parent-resolved TEEC is Eq. (13) with the TMDFFs replaced by the parent-resolved ones, with the additional integrations over K \mathbf{K} K and K ˉ \bar{\mathbf{K}} K ˉ , and with the measurement of Eq. (21) . The jet function of Eq. (10) changes accordingly, with the additional integration over K \mathbf{K} K and the phase e i b ⋅ K e^{i\mathbf{b}\cdot\mathbf{K}} e i b ⋅ K . In the perturbative regime, the parent-resolved TMDFF is matched onto collinear fragmentation functions through Eq. (11) , with K \mathbf{K} K as an additional argument of F i → a F_{i\to a} F i → a and I i j I_{ij} I ij . Since hadronization takes place outside the medium, the f j → a f_{j\to a} f j → a are the vacuum fragmentation functions, and all medium effects are contained in I i j I_{ij} I ij . Therefore, the momentum sum rule still removes the fragmentation functions from the energy weighted jet function,4 and Eq. (12) becomes
J q ( b ) = ∑ i ∫ 0 1 d x x ∫ ℓ , K e i b ⋅ ( K + ℓ / x ) I q i ( x , ℓ , K ) , (22) J_q(b)=\sum_i\int_0^1 dx\,x\int_{\ell,\mathbf{K}}e^{i\mathbf{b}\cdot(\mathbf{K}+\ell/x)}I_{qi}(x,\ell,\mathbf{K}), \tag*{(22)} J q ( b ) = i ∑ ∫ 0 1 d x x ∫ ℓ , K e i b ⋅ ( K + ℓ / x ) I q i ( x , ℓ , K ) , (22)
with b = b y ^ b=b\hat{y} b = b y ^ , and analogously for J q ˉ J_{\bar{q}} J q ˉ . With b b b the impact parameter along y ^ \hat{y} y ^ of Eq. (14), the reflection symmetry of the medium geometry gives J q ( b ) = J q ( − b ) J_q(b)=J_q(-b) J q ( b ) = J q ( − b ) , as in vacuum, so that Eq. (9) holds with these jet functions. Their dependence on the parent energy and the medium parameters is left implicit. The matching coefficient follows from Eq. (11) evaluated for partonic final states, for which both sides can be computed in perturbation theory, I q i = I q i ( 0 ) + I q i ( 1 ) + O ( α s 2 ) I_{qi}=I_{qi}^{(0)}+I_{qi}^{(1)}+\mathcal{O}(\alpha_s^2) I q i = I q i ( 0 ) + I q i ( 1 ) + O ( α s 2 ) , which we discuss in Secs. III B and III C.
Medium-induced broadening of the jet function
This subsection evaluates the zeroth-order matching coefficient I q i ( 0 ) I_{qi}^{(0)} I q i ( 0 ) and the jet function of Eq. (22) that it gives. Let d σ ^ q ∣ p 0 d\hat{\sigma}_{q|p_0} d σ ^ q ∣ p 0 denote the partonic cross section conditional on producing a parent quark with momentum p 0 p_0 p 0 , with the production process absorbed into the hard factor.5 Without resolved radiation, f j → i ( 0 ) ( ζ ) = δ i j δ ( 1 − ζ ) f_{j\to i}^{(0)}(\zeta)=\delta_{ij}\delta(1-\zeta) f j → i ( 0 ) ( ζ ) = δ ij δ ( 1 − ζ ) and
I q i ( 0 ) ( x , ℓ , K ) = ( 2 π ) 2 d σ ^ q ∣ p 0 d 2 K δ i q δ ( x − 1 ) ( 2 π ) 2 δ ( 2 ) ( ℓ ) . (23) I_{qi}^{(0)}(x,\ell,\boldsymbol{K})=(2\pi)^2\frac{d\hat{\sigma}_{q|p_0}}{d^2\boldsymbol{K}}\delta_{iq}\delta(x-1)(2\pi)^2\delta^{(2)}(\ell). \tag*{(23)} I q i ( 0 ) ( x , ℓ , K ) = ( 2 π ) 2 d 2 K d σ ^ q ∣ p 0 δ i q δ ( x − 1 ) ( 2 π ) 2 δ ( 2 ) ( ℓ ) . (23)
In a transversely homogeneous medium occupying the light-cone time interval 0 < s + < L + 0<s^+<L^+ 0 < s + < L + , the conditional cross section without resolved radiation describes pure broadening of the parent quark,
( 2 π ) 2 d σ ^ q ∣ p 0 d 2 K = P F ( K ; L + , 0 ) . (24) (2\pi)^2\frac{d\hat{\sigma}_{q|p_0}}{d^2\boldsymbol{K}}=\mathcal{P}_{F}(\boldsymbol{K};L^+,0). \tag*{(24)} ( 2 π ) 2 d 2 K d σ ^ q ∣ p 0 = P F ( K ; L + , 0 ) . (24)
The broadening distribution of a parton in colour representation R R R is the Fourier transform
P R ( q ; x f + , x i + ) = ∫ d 2 u e i q ⋅ u D R [ u ; x f + , x i + ] (25) \mathcal{P}_{R}(\boldsymbol{q};x_f^+,x_i^+)=\int d^2\boldsymbol{u}\,e^{i\boldsymbol{q}\cdot\boldsymbol{u}}D_{R}[\boldsymbol{u};x_f^+,x_i^+] \tag*{(25)} P R ( q ; x f + , x i + ) = ∫ d 2 u e i q ⋅ u D R [ u ; x f + , x i + ] (25)
of the dipole function
D R [ x ; x f + , x i + ] = exp { − C R 2 ∫ x i + x f + d s + n ( s + ) σ ( x ( s + ) ) } . (26) D_{R}[\boldsymbol{x};x_f^+,x_i^+]=\exp\left\{-\frac{C_R}{2}\int_{x_i^+}^{x_f^+}ds^+\,n(s^+)\sigma(\boldsymbol{x}(s^+))\right\}. \tag*{(26)} D R [ x ; x f + , x i + ] = exp { − 2 C R ∫ x i + x f + d s + n ( s + ) σ ( x ( s + )) } . (26)
where n ( s + ) n(s^+) n ( s + ) is the density of scattering centres and σ ( x ) \sigma(\boldsymbol{x}) σ ( x ) the dipole cross section. A constant argument, as in Eq. (25) , denotes x ( s + ) = u \boldsymbol{x}(s^+)=\boldsymbol{u} x ( s + ) = u . Since D R [ 0 ] = 1 D_R[\boldsymbol{0}]=1 D R [ 0 ] = 1 , the distribution is normalized, ∫ q P R ( q ) = 1 \int_{\boldsymbol{q}}\mathcal{P}_R(\boldsymbol{q})=1 ∫ q P R ( q ) = 1 . Substituting this distribution into Eq. (23) and using Eq. (22) gives
J ( 0 ) ( b ) = D F [ b ; L + , 0 ] , (27) \mathcal{J}^{(0)}(b)=D_F[b;L^+,0], \tag*{(27)} J ( 0 ) ( b ) = D F [ b ; L + , 0 ] , (27)
with J ( 0 ) , v a c ( b ) = 1 J^{(0),\mathrm{vac}}(b)=1 J ( 0 ) , vac ( b ) = 1 . From Sec. IV onwards we evaluate the dipole function in the harmonic-oscillator approximation [63 , 64 ] . It keeps the small-u u u behaviour of the dipole cross section in quadratic form and drops the accompanying logarithm,
C R 2 n ( s + ) σ ( u ) ⟶ C R C F q ^ ( s + ) 4 u 2 , (28) \frac{C_R}{2}n(s^+)\sigma(\boldsymbol{u}) \longrightarrow\frac{C_R}{C_F}\frac{\hat{q}(s^+)}{4}\boldsymbol{u}^2, \tag*{(28)} 2 C R n ( s + ) σ ( u ) ⟶ C F C R 4 q ^ ( s + ) u 2 , (28)
D R [ u ; x f + , x i + ] = exp { − C R C F u 2 4 ∫ x i + x f + d s + q ^ ( s + ) } , D_R[\boldsymbol{u};x_f^+,x_i^+] = \exp\left\{-\frac{C_R}{C_F}\frac{\boldsymbol{u}^2}{4}\int_{x_i^+}^{x_f^+}\mathrm{d}s^+\,\hat{q}(s^+)\right\}, D R [ u ; x f + , x i + ] = exp { − C F C R 4 u 2 ∫ x i + x f + d s + q ^ ( s + ) } ,
which defines the quark transport coefficient q ^ ≡ q ^ F \hat{q}\equiv\hat{q}_F q ^ ≡ q ^ F , with q ^ A = ( C A / C F ) q ^ \hat{q}_A=(C_A/C_F)\hat{q} q ^ A = ( C A / C F ) q ^ . In the medium-induced terms we work at leading order in 1 / N c 1/N_c 1/ N c , with C F = N c / 2 C_F=N_c/2 C F = N c /2 and C A = N c C_A=N_c C A = N c . Then q ^ A = 2 q ^ \hat{q}_A=2\hat{q} q ^ A = 2 q ^ , and a gluon line is represented by a pair of fundamental lines, so that its dipole is D F 2 D_F^2 D F 2 (Sec. III C). Corrections of relative order 1 / N c 2 1/N_c^2 1/ N c 2 to the medium-induced terms are neglected, while the vacuum functions J v a c J^{\mathrm{vac}} J vac and S v a c S_{\mathrm{vac}} S vac keep their full colour dependence.
Medium-induced radiative correction to the jet function
At O ( α s ) O(\alpha_s) O ( α s ) , Eq. (11) gives I q i ( 1 ) = F q → i ( 1 ) − ∑ j I q j ( 0 ) ⊗ f j → i ( 1 ) I_{qi}^{(1)}=F_{q\to i}^{(1)}-\sum_j I_{qj}^{(0)}\otimes f_{j\to i}^{(1)} I q i ( 1 ) = F q → i ( 1 ) − ∑ j I q j ( 0 ) ⊗ f j → i ( 1 ) . The first term contains the real emission q → q g q\to qg q → q g , conditional on the parent, and the virtual correction. The second term is the broadening of the parent, contained in I ( 0 ) I^{(0)} I ( 0 ) , followed by a vacuum collinear splitting in f ( 1 ) f^{(1)} f ( 1 ) , which does not resolve the transverse momentum of the fragmenting parton. Let z z z be the energy fraction of the final quark in q → q g q\to qg q → q g , and let p = ( 1 − z ) p 1 − z p 2 \boldsymbol{p}=(1-z)\boldsymbol{p}_1-z\boldsymbol{p}_2 p = ( 1 − z ) p 1 − z p 2 be the relative transverse momentum of the daughters, where p 1 \boldsymbol{p}_1 p 1 and p 2 \boldsymbol{p}_2 p 2 are the final quark and gluon transverse momenta. We denote by d σ ^ q g ∣ p 0 \mathrm{d}\hat{\sigma}_{qg|p_0} d σ ^ q g ∣ p 0 the conditional cross section for gluon emission off the quark, and by d σ ^ q g ∣ p 0 v a c \mathrm{d}\hat{\sigma}^{\mathrm{vac}}_{qg|p_0} d σ ^ q g ∣ p 0 vac its vacuum value. With f ( 1 ) f^{(1)} f ( 1 ) defined by a sharp cutoff μ c \mu_c μ c on the relative momentum,
I q i ( 1 ) ( x , ℓ , K ) = ∫ 0 1 d z ∫ p ( 2 π ) 4 [ d σ ^ q g ∣ p 0 d z d 2 p d 2 K M r e s . − Θ ( μ c 2 − p 2 ) d σ ^ q ∣ p 0 d 2 K d σ ^ q g ∣ p 0 v a c d z d 2 p M l o c . ] + [ terms ∝ δ i q δ ( 1 − x ) ] , (29) I_{qi}^{(1)}(x,\ell,\boldsymbol{K})=\int_0^1\mathrm{d}z\int_{\boldsymbol{p}}(2\pi)^4\left[\frac{\mathrm{d}\hat{\sigma}_{qg|p_0}}{\mathrm{d}z\,\mathrm{d}^2\boldsymbol{p}\,\mathrm{d}^2\boldsymbol{K}}\mathcal{M}_{\mathrm{res.}}-\Theta(\mu_c^2-p^2)\frac{\mathrm{d}\hat{\sigma}_{q|p_0}}{\mathrm{d}^2\boldsymbol{K}}\frac{\mathrm{d}\hat{\sigma}^{\mathrm{vac}}_{qg|p_0}}{\mathrm{d}z\,\mathrm{d}^2\boldsymbol{p}}\mathcal{M}_{\mathrm{loc.}}\right]+\left[\text{terms}\propto\delta_{iq}\delta(1-x)\right], \tag*{(29)} I q i ( 1 ) ( x , ℓ , K ) = ∫ 0 1 d z ∫ p ( 2 π ) 4 [ d z d 2 p d 2 K d σ ^ q g ∣ p 0 M res. − Θ ( μ c 2 − p 2 ) d 2 K d σ ^ q ∣ p 0 d z d 2 p d σ ^ q g ∣ p 0 vac M loc. ] + [ terms ∝ δ i q δ ( 1 − x ) ] , (29)
where the first term may be evaluated in vacuum or in the medium, and the last term contains the virtual corrections. The resolved measurement retains the relative momentum of the two daughters, while the local one retains only their longitudinal fractions,
M r e s . = ( 2 π ) 2 [ δ i q δ ( x − z ) δ ( 2 ) ( ℓ − p ) + δ i g δ ( x − ( 1 − z ) ) δ ( 2 ) ( ℓ + p ) ] , \mathcal{M}_{\mathrm{res.}}=(2\pi)^2\left[\delta_{iq}\delta(x-z)\delta^{(2)}(\ell-p)+\delta_{ig}\delta(x-(1-z))\delta^{(2)}(\ell+p)\right], M res. = ( 2 π ) 2 [ δ i q δ ( x − z ) δ ( 2 ) ( ℓ − p ) + δ i g δ ( x − ( 1 − z )) δ ( 2 ) ( ℓ + p ) ] ,
M l o c . = ( 2 π ) 2 [ δ i q δ ( x − z ) δ ( 2 ) ( ℓ ) + δ i g δ ( x − ( 1 − z ) ) δ ( 2 ) ( ℓ ) ] . (30) \mathcal{M}_{\mathrm{loc.}}=(2\pi)^2\left[\delta_{iq}\delta(x-z)\delta^{(2)}(\ell)+\delta_{ig}\delta(x-(1-z))\delta^{(2)}(\ell)\right]. \tag*{(30)} M loc. = ( 2 π ) 2 [ δ i q δ ( x − z ) δ ( 2 ) ( ℓ ) + δ i g δ ( x − ( 1 − z )) δ ( 2 ) ( ℓ ) ] . (30)
As in vacuum, the soft singularity of the real emission at z → 1 z\to1 z → 1 is treated with a plus-prescription,
M r e s . [ + ] = M r e s . − ( 2 π ) 2 δ i q δ ( x − 1 ) δ ( 2 ) ( ℓ − p ) , \mathcal{M}_{\mathrm{res.}}^{[+]}=\mathcal{M}_{\mathrm{res.}}-(2\pi)^2\delta_{iq}\delta(x-1)\delta^{(2)}(\ell-p), M res. [ + ] = M res. − ( 2 π ) 2 δ i q δ ( x − 1 ) δ ( 2 ) ( ℓ − p ) ,
M l o c . [ + ] = M l o c . − ( 2 π ) 2 δ i q δ ( x − 1 ) δ ( 2 ) ( ℓ ) , (31) \mathcal{M}_{\mathrm{loc.}}^{[+]}=\mathcal{M}_{\mathrm{loc.}}-(2\pi)^2\delta_{iq}\delta(x-1)\delta^{(2)}(\ell), \tag*{(31)} M loc. [ + ] = M loc. − ( 2 π ) 2 δ i q δ ( x − 1 ) δ ( 2 ) ( ℓ ) , (31)
and the plus-subtracted forms are understood from this point on. In vacuum, the remaining coefficient of δ ( 1 − x ) \delta(1-x) δ ( 1 − x ) , with a rapidity regulator and renormalization, and combined with the soft function, gives the standard J v a c ( b , μ , ν ) J^{\mathrm{vac}}(b,\mu,\nu) J vac ( b , μ , ν ) and S v a c S_{\mathrm{vac}} S vac [24 ] , and the scales μ \mu μ and ν \nu ν enter the jet function only through this coefficient. In the medium, it also receives the recoil of the quark against soft medium-induced gluons, which is rapidity divergent as well, since the medium-induced spectrum grows like ( 1 − z ) − 3 / 2 (1-z)^{-3/2} ( 1 − z ) − 3/2 as z → 1 z\to1 z → 1 in the limit of dense matter; see [65 ] for related discussion in the dilute limit. This divergence is expected to cancel only against a medium-modified soft function, which lies outside the collinear BDMPS–Z formalism used here, and which we leave for future work. Consistently with the vacuum soft function, we omit this medium recoil term and use the plus-prescription without it for the medium-induced radiative correction. The double-logarithmic part of the omitted recoil is the radiative transverse momentum broadening of [66 , 67 , 68 ] .
We evaluate the first-order matching coefficient of Eq. (29) using the in-medium q → q g q\to qg q → q g cross section. In the BDMPS–Z formalism [63 , 64 , 69 , 70 ] , the fixed-order cross section for q ( p 0 ) → q ( p 1 ) g ( p 2 ) q(p_0)\to q(p_1)g(p_2) q ( p 0 ) → q ( p 1 ) g ( p 2 ) at finite z z z reads [71 , 72 , 73 ]
( 2 π ) 2 d σ ^ q g ∣ p 0 d z d 2 p d 2 K = 2 C F α s ( 2 ω + ) 2 P q → q g v a c ( z ) 2 ℜ ∫ q 0 , q 1 , q ˉ 0 , q ˉ 1 , l 1 , x + < x ˉ + e − ϵ ( x ˉ + + x + ) ( κ ˉ ⋅ κ ) × P F ( q 0 − p 0 ; x + , 0 ) S q → q g ( 3 ) ( l 1 − z q ˉ 0 , κ , q 0 − q ˉ 0 ; x ˉ + , x + ) × S q → q g ( 4 ) ( x ∞ + , p 1 , p 2 ; x ˉ + , q ˉ 0 − l 1 , l 1 , q ˉ 0 − q ˉ 1 , q ˉ 1 ) , (32) (2\pi)^2\frac{\mathrm{d}\hat{\sigma}_{qg|p_0}}{\mathrm{d}z\,\mathrm{d}^2\boldsymbol{p}\,\mathrm{d}^2\boldsymbol{K}}
=\frac{2C_F\alpha_s}{(2\omega^+)^2}P^{\mathrm{vac}}_{q\to qg}(z)\,2\Re\int_{q_0,q_1,\bar q_0,\bar q_1,l_1,x^+<\bar x^+}
e^{-\epsilon(\bar x^++x^+)}(\bar{\boldsymbol{\kappa}}\cdot\boldsymbol{\kappa})
\times\mathcal{P}_F(q_0-p_0;x^+,0)\,S^{(3)}_{q\to qg}(l_1-z\bar q_0,\boldsymbol{\kappa},q_0-\bar q_0;\bar x^+,x^+)
\times S^{(4)}_{q\to qg}(x_\infty^+,\boldsymbol{p}_1,\boldsymbol{p}_2;\bar x^+,\bar q_0-l_1,l_1,\bar q_0-\bar q_1,\bar q_1), \tag*{(32)} ( 2 π ) 2 d z d 2 p d 2 K d σ ^ q g ∣ p 0 = ( 2 ω + ) 2 2 C F α s P q → q g vac ( z ) 2ℜ ∫ q 0 , q 1 , q ˉ 0 , q ˉ 1 , l 1 , x + < x ˉ + e − ϵ ( x ˉ + + x + ) ( κ ˉ ⋅ κ ) × P F ( q 0 − p 0 ; x + , 0 ) S q → q g ( 3 ) ( l 1 − z q ˉ 0 , κ , q 0 − q ˉ 0 ; x ˉ + , x + ) × S q → q g ( 4 ) ( x ∞ + , p 1 , p 2 ; x ˉ + , q ˉ 0 − l 1 , l 1 , q ˉ 0 − q ˉ 1 , q ˉ 1 ) , (32)
with z = p 1 + / p 0 + z=p_1^+/p_0^+ z = p 1 + / p 0 + , ω + = z ( 1 − z ) p 0 + \omega^+=z(1-z)p_0^+ ω + = z ( 1 − z ) p 0 + , P q → q g v a c ( z ) = ( 1 + z 2 ) / ( 1 − z ) P^{\mathrm{vac}}_{q\to qg}(z)=(1+z^2)/(1-z) P q → q g vac ( z ) = ( 1 + z 2 ) / ( 1 − z ) , the total transverse momentum P = p 1 + p 2 = p 0 + K \boldsymbol{P}=\boldsymbol{p}_1+\boldsymbol{p}_2=\boldsymbol{p}_0+\boldsymbol{K} P = p 1 + p 2 = p 0 + K and the relative transverse momentum p = ( 1 − z ) p 1 − z p 2 \boldsymbol{p}=(1-z)\boldsymbol{p}_1-z\boldsymbol{p}_2 p = ( 1 − z ) p 1 − z p 2 . Unbarred internal momenta belong to the amplitude and barred ones to the conjugate amplitude. The splitting vertices occur at x + x^+ x + in the amplitude and at x ˉ + > x + \bar{x}^+>x^+ x ˉ + > x + in the conjugate amplitude, and the time shorthand denotes ∫ 0 ∞ d x + ∫ x + ∞ d x ˉ + \int_0^\infty\mathrm{d}x^+\int_{x^+}^\infty\mathrm{d}\bar{x}^+ ∫ 0 ∞ d x + ∫ x + ∞ d x ˉ + . The parent and daughter-quark momenta adjacent to the amplitude vertex are q 0 \boldsymbol{q}_0 q 0 and q 1 \boldsymbol{q}_1 q 1 , with relative momentum κ = q 1 − z q 0 \boldsymbol{\kappa}=\boldsymbol{q}_1-z\boldsymbol{q}_0 κ = q 1 − z q 0 . On the conjugate side they are q ˉ 0 \bar{\boldsymbol{q}}_0 q ˉ 0 and q ˉ 1 \bar{\boldsymbol{q}}_1 q ˉ 1 , with κ ˉ = q ˉ 1 − z q ˉ 0 \bar{\boldsymbol{\kappa}}=\bar{\boldsymbol{q}}_1-z\bar{\boldsymbol{q}}_0 κ ˉ = q ˉ 1 − z q ˉ 0 . The momentum l 1 \boldsymbol{l}_1 l 1 is that of the amplitude daughter quark at x ˉ + \bar{x}^+ x ˉ + . The common final time x ∞ + x_\infty^+ x ∞ + lies after the medium, where p 1 \boldsymbol{p}_1 p 1 and p 2 \boldsymbol{p}_2 p 2 are specified on both sides of the cut. The convergence factor is removed after the late-time integration by taking ϵ → 0 + \epsilon\to0^+ ϵ → 0 + . The three factors describe the successive stages of the process. Between 0 0 0 and x + x^+ x + the parent propagates on both sides of the cut and broadens with P F \mathcal{P}_F P F . Between x + x^+ x + and x ˉ + \bar{x}^+ x ˉ + the amplitude contains the quark-gluon pair and the conjugate amplitude the parent, which gives the three-point function. After x ˉ + \bar{x}^+ x ˉ + the pair propagates on both sides, which gives the four-point function. As a check of the normalization, in vacuum P F → ( 2 π ) 2 δ ( 2 ) \mathcal{P}_F\to(2\pi)^2\delta^{(2)} P F → ( 2 π ) 2 δ ( 2 ) , the correlators reduce to free propagation, κ = κ ˉ = p \boldsymbol{\kappa}=\bar{\boldsymbol{\kappa}}=\boldsymbol{p} κ = κ ˉ = p , and the time integrals give ( 2 ω + / p 2 ) 2 (2\omega^+/\boldsymbol{p}^2)^2 ( 2 ω + / p 2 ) 2 . Equation (32) then reduces to ( 2 π ) 2 δ ( 2 ) ( K ) 2 C F α s P q → q g v a c ( z ) / p 2 (2\pi)^2\delta^{(2)}(\boldsymbol{K})\,2C_F\alpha_sP^{\mathrm{vac}}_{q\to qg}(z)/\boldsymbol{p}^2 ( 2 π ) 2 δ ( 2 ) ( K ) 2 C F α s P q → q g vac ( z ) / p 2 , which is the leading-order splitting ( 2 π ) 2 d σ ^ q g ∣ p 0 / ( d z d 2 p ) (2\pi)^2\mathrm{d}\hat{\sigma}_{qg|p_0}/(\mathrm{d}z\,\mathrm{d}^2\boldsymbol{p}) ( 2 π ) 2 d σ ^ q g ∣ p 0 / ( d z d 2 p ) .
In the large N c N_c N c approximation, the momentum space three-point correlator is
S q → q g ( 3 ) ( l 1 − z q ˉ 0 , κ , q 0 − q ˉ 0 ; x f + , x i + ) = ∫ d 2 b 3 d 2 r f d 2 r i e − i r f ⋅ ( l 1 − z q ˉ 0 ) e i r i ⋅ κ e i b 3 ⋅ ( q 0 − q ˉ 0 ) ∫ r i r f D r exp { i ω + 2 ∫ x i + x f + d s + r ˙ 2 } D F [ b 3 − z r ; x f + , x i + ] D F [ r ; x f + , x i + ] . (33) \mathcal{S}^{(3)}_{q\to qg}(l_1-z\bar{q}_0,\boldsymbol{\kappa},\mathbf{q}_0-\bar{\mathbf{q}}_0;x_f^+,x_i^+)=\int d^2\mathbf{b}_3\,d^2\mathbf{r}_f\,d^2\mathbf{r}_i\,e^{-i\mathbf{r}_f\cdot(l_1-z\bar{\mathbf{q}}_0)}e^{i\mathbf{r}_i\cdot\boldsymbol{\kappa}}e^{i\mathbf{b}_3\cdot(\mathbf{q}_0-\bar{\mathbf{q}}_0)}
\int_{\mathbf{r}_i}^{\mathbf{r}_f}\mathcal{D}\mathbf{r}\exp\left\{i\frac{\omega^+}{2}\int_{x_i^+}^{x_f^+}ds^+\,\dot{\mathbf{r}}^2\right\}D_F[\mathbf{b}_3-z\mathbf{r};x_f^+,x_i^+]D_F[\mathbf{r};x_f^+,x_i^+]. \tag*{(33)} S q → q g ( 3 ) ( l 1 − z q ˉ 0 , κ , q 0 − q ˉ 0 ; x f + , x i + ) = ∫ d 2 b 3 d 2 r f d 2 r i e − i r f ⋅ ( l 1 − z q ˉ 0 ) e i r i ⋅ κ e i b 3 ⋅ ( q 0 − q ˉ 0 ) ∫ r i r f D r exp { i 2 ω + ∫ x i + x f + d s + r ˙ 2 } D F [ b 3 − z r ; x f + , x i + ] D F [ r ; x f + , x i + ] . (33)
Here r ( s + ) \mathbf{r}(s^+) r ( s + ) is the daughter relative coordinate with endpoints r i \mathbf{r}_i r i and r f \mathbf{r}_f r f , and b 3 \mathbf{b}_3 b 3 is the separation of the daughter transverse centre from the conjugate-amplitude parent. The dipole with argument r \mathbf{r} r connects the amplitude quark and gluon, and the dipole with argument b 3 − z r \mathbf{b}_3-z\mathbf{r} b 3 − z r connects the amplitude gluon and the conjugate parent. Transverse translation invariance of the four-point function allows its common centre coordinate to be integrated out. This gives ( 2 π ) 2 δ ( 2 ) ( k 1 + k 2 − k ˉ 1 − k ˉ 2 ) (2\pi)^2\delta^{(2)}(\mathbf{k}_1+\mathbf{k}_2-\bar{\mathbf{k}}_1-\bar{\mathbf{k}}_2) ( 2 π ) 2 δ ( 2 ) ( k 1 + k 2 − k ˉ 1 − k ˉ 2 ) , which is satisfied in Eq. (32) because both incoming total momenta are q ˉ 0 \bar{\mathbf{q}}_0 q ˉ 0 and both outgoing total momenta are P \mathbf{P} P . After this factor is removed, the reduced correlator is
S q → q g ( 4 ) ( x ∞ + , p 1 , p 2 ; x ˉ + , q ˉ 0 − l 1 , l 1 , q ˉ 0 − q ˉ 1 , q ˉ 1 ) = ∫ d 2 b 4 d 2 Δ d 2 u d 2 w e − i ( P − q ˉ 0 ) ⋅ b 4 e − i p ⋅ Δ e i ( l 1 − z q ˉ 0 ) ⋅ u e − i ( q ˉ 1 − z q ˉ 0 ) ⋅ w × S ~ q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b 4 , z ) . (34) \begin{aligned}
\mathcal{S}^{(4)}_{q\to qg}(x_\infty^+,\mathbf{p}_1,\mathbf{p}_2;\bar{x}^+,\bar{\mathbf{q}}_0-l_1,l_1,\bar{\mathbf{q}}_0-\bar{\mathbf{q}}_1,\bar{\mathbf{q}}_1)
&=\int d^2\mathbf{b}_4\,d^2\boldsymbol{\Delta}\,d^2\mathbf{u}\,d^2\mathbf{w}\,e^{-i(\mathbf{P}-\bar{\mathbf{q}}_0)\cdot\mathbf{b}_4}e^{-i\mathbf{p}\cdot\boldsymbol{\Delta}}e^{i(l_1-z\bar{\mathbf{q}}_0)\cdot\mathbf{u}}e^{-i(\bar{\mathbf{q}}_1-z\bar{\mathbf{q}}_0)\cdot\mathbf{w}}\\
&\quad\times\widetilde{\mathcal{S}}^{(4)}_{q\to qg}(x_\infty^+,\boldsymbol{\Delta};\bar{x}^+,\mathbf{u},\mathbf{w};\mathbf{b}_4,z). \tag*{(34)}
\end{aligned} S q → q g ( 4 ) ( x ∞ + , p 1 , p 2 ; x ˉ + , q ˉ 0 − l 1 , l 1 , q ˉ 0 − q ˉ 1 , q ˉ 1 ) = ∫ d 2 b 4 d 2 Δ d 2 u d 2 w e − i ( P − q ˉ 0 ) ⋅ b 4 e − i p ⋅ Δ e i ( l 1 − z q ˉ 0 ) ⋅ u e − i ( q ˉ 1 − z q ˉ 0 ) ⋅ w × S q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b 4 , z ) . (34)
Here u \mathbf{u} u and w \mathbf{w} w are the initial daughter relative coordinates in the amplitude and conjugate amplitude, Δ \boldsymbol{\Delta} Δ is the difference of the final ones, and b 4 \mathbf{b}_4 b 4 is the difference of the two transverse centre coordinates. The coordinate b 4 \mathbf{b}_4 b 4 is conjugate to the total momentum transferred after x ˉ + \bar{x}^+ x ˉ + , and Δ \boldsymbol{\Delta} Δ to the final relative momentum. The coordinate-space kernel is
S ~ q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b 4 , z ) = ∫ d 2 r ∫ v ( x ˉ + ) = u v ( x ∞ + ) = r D v ∫ v ˉ ( x ˉ + ) = w v ˉ ( x ∞ + ) = r − Δ D v ˉ exp { i ω + 2 ∫ x ˉ + x ∞ + d s + ( v ˙ 2 − v ˉ ˙ 2 ) } C q → q g ( 4 ) [ v , v ˉ , b 4 ] . (35) \begin{aligned}
\widetilde{\mathcal{S}}^{(4)}_{q\to qg}(x_\infty^+,\boldsymbol{\Delta};\bar{x}^+,\mathbf{u},\mathbf{w};\mathbf{b}_4,z)
&=\int d^2\mathbf{r}\int_{\mathbf{v}(\bar{x}^+)=\mathbf{u}}^{\mathbf{v}(x_\infty^+)=\mathbf{r}}\mathcal{D}\mathbf{v}\int_{\bar{\mathbf{v}}(\bar{x}^+)=\mathbf{w}}^{\bar{\mathbf{v}}(x_\infty^+)=\mathbf{r}-\boldsymbol{\Delta}}\mathcal{D}\bar{\mathbf{v}}\exp\left\{i\frac{\omega^+}{2}\int_{\bar{x}^+}^{x_\infty^+}ds^+\,(\dot{\mathbf{v}}^2-\dot{\bar{\mathbf{v}}}^2)\right\}\mathcal{C}^{(4)}_{q\to qg}[\mathbf{v},\bar{\mathbf{v}},\mathbf{b}_4]. \tag*{(35)}
\end{aligned} S q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b 4 , z ) = ∫ d 2 r ∫ v ( x ˉ + ) = u v ( x ∞ + ) = r D v ∫ v ˉ ( x ˉ + ) = w v ˉ ( x ∞ + ) = r − Δ D v ˉ exp { i 2 ω + ∫ x ˉ + x ∞ + d s + ( v ˙ 2 − v ˉ ˙ 2 ) } C q → q g ( 4 ) [ v , v ˉ , b 4 ] . (35)
where the r \mathbf{r} r integration sums over one final relative coordinate at fixed Δ \boldsymbol{\Delta} Δ . In the same large N c N_c N c representation the gluon is a pair of fundamental Wilson lines, and the six lines of the amplitude and conjugate amplitude organize into a dipole and a quadrupole,
C q → q g ( 4 ) [ v , v ˉ , b 4 ] = D F [ b 4 − z ( v − v ˉ ) ; x ∞ + , x ˉ + ] 1 N c ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ , (36) \mathcal{C}^{(4)}_{q\to qg}[\mathbf{v},\bar{\mathbf{v}},\mathbf{b}_4]=D_F[\mathbf{b}_4-z(\mathbf{v}-\bar{\mathbf{v}});x_\infty^+,\bar{x}^+]\frac{1}{N_c}\left\langle\operatorname{Tr}\left[U_F(\mathbf{r}_1)U_F^\dagger(\mathbf{r}_2)U_F(\bar{\mathbf{r}}_2)U_F^\dagger(\bar{\mathbf{r}}_1)\right]\right\rangle, \tag*{(36)} C q → q g ( 4 ) [ v , v ˉ , b 4 ] = D F [ b 4 − z ( v − v ˉ ) ; x ∞ + , x ˉ + ] N c 1 ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ , (36)
with Wilson lines from x ˉ + \bar{x}^+ x ˉ + to x ∞ + x_\infty^+ x ∞ + along the paths
v = r 1 − r 2 , v ˉ = r ˉ 1 − r ˉ 2 , b 4 = z ( r 1 − r ˉ 1 ) + ( 1 − z ) ( r 2 − r ˉ 2 ) . (37) \mathbf{v}=\mathbf{r}_1-\mathbf{r}_2,\qquad\bar{\mathbf{v}}=\bar{\mathbf{r}}_1-\bar{\mathbf{r}}_2,\qquad\mathbf{b}_4=z(\mathbf{r}_1-\bar{\mathbf{r}}_1)+(1-z)(\mathbf{r}_2-\bar{\mathbf{r}}_2). \tag*{(37)} v = r 1 − r 2 , v ˉ = r ˉ 1 − r ˉ 2 , b 4 = z ( r 1 − r ˉ 1 ) + ( 1 − z ) ( r 2 − r ˉ 2 ) . (37)
Since b 4 − z ( v − v ˉ ) = r 2 − r ˉ 2 b_{4}-z(v-\bar{v})=r_{2}-\bar{r}_{2} b 4 − z ( v − v ˉ ) = r 2 − r ˉ 2 , the dipole in Eq. (36) connects the gluon in the amplitude and in the conjugate amplitude. To express the cross section in the variables of the jet function, we substitute Eqs. (33) and (34) into Eq. (32) and transform in K = P − p 0 K=P-p_{0} K = P − p 0 with the phase e i b ⋅ K e^{i\boldsymbol{b}\cdot\boldsymbol{K}} e i b ⋅ K of Eq. (22) . The momentum integrations set b 3 = b 4 = b b_{3}=b_{4}=b b 3 = b 4 = b , give the parent broadening D F [ b ; x + , 0 ] D_{F}[b;x^{+},0] D F [ b ; x + , 0 ] , and cancel the p 0 p_{0} p 0 dependence of the phase. Therefore, all medium factors depend on the impact parameter b \boldsymbol{b} b . The vertex momenta become derivatives acting on the kernels, and the remaining integrations are carried out in Sec. IV.
The vertex times x + < x ˉ + x^{+}<\bar{x}^{+} x + < x ˉ + in Eq. (32) fall into three regions relative to the medium: in–in with x ˉ + < L + \bar{x}^{+}<L^{+} x ˉ + < L + , in–out with x + < L + < x ˉ + x^{+}<L^{+}<\bar{x}^{+} x + < L + < x ˉ + , and out–out with L + < x + L^{+}<x^{+} L + < x + . In the out–out region the parent broadens over [ 0 , L + ] [0,L^{+}] [ 0 , L + ] and splits during the subsequent vacuum propagation. Shifting both vertex times by L + L^{+} L + changes only the convergence factor e − ϵ ( x ˉ + + x + ) e^{-\epsilon(\bar{x}^{+}+x^{+})} e − ϵ ( x ˉ + + x + ) , whose effect disappears for ϵ → 0 + \epsilon\to0^{+} ϵ → 0 + . Therefore, the out–out contribution is the broadening of the parent followed by the vacuum splitting,
( 2 π ) 4 d σ ^ o u t − o u t d z d 2 p d 2 K = ( 2 π ) 4 d σ ^ q ∣ p 0 d 2 K d σ ^ q g ∣ p 0 v a c d z d 2 p = P F ( K ; L + , 0 ) ( 2 π ) 2 d σ ^ q g ∣ p 0 v a c d z d 2 p , (38) (2\pi)^{4}\frac{d\hat{\sigma}_{\mathrm{out-out}}}{dz\,d^{2}p\,d^{2}K}=(2\pi)^{4}\frac{d\hat{\sigma}_{q|p_{0}}}{d^{2}K}\frac{d\hat{\sigma}^{\mathrm{vac}}_{qg|p_{0}}}{dz\,d^{2}p}=\mathcal{P}_{F}(K;L^{+},0)(2\pi)^{2}\frac{d\hat{\sigma}^{\mathrm{vac}}_{qg|p_{0}}}{dz\,d^{2}p}, \tag*{(38)} ( 2 π ) 4 d z d 2 p d 2 K d σ ^ out − out = ( 2 π ) 4 d 2 K d σ ^ q ∣ p 0 d z d 2 p d σ ^ q g ∣ p 0 vac = P F ( K ; L + , 0 ) ( 2 π ) 2 d z d 2 p d σ ^ q g ∣ p 0 vac , (38)
which also multiplies the local subtraction in Eq. (29) . We call the sum of the other two regions the medium-induced radiative cross section,
d σ ^ q g ∣ p 0 m e d d z d 2 p ≡ d σ ^ i n − i n d z d 2 p + d σ ^ i n − o u t d z d 2 p , (39) \frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz\,d^{2}p} \equiv\frac{d\hat{\sigma}_{\mathrm{in-in}}}{dz\,d^{2}p}+\frac{d\hat{\sigma}_{\mathrm{in-out}}}{dz\,d^{2}p}, \tag*{(39)} d z d 2 p d σ ^ q g ∣ p 0 med ≡ d z d 2 p d σ ^ in − in + d z d 2 p d σ ^ in − out , (39)
which vanishes in vacuum. Substituting this decomposition and Eq. (38) into Eq. (29) , the out–out term and the local subtraction share the factor P F ( K ; L + , 0 ) \mathcal{P}_{F}(K;L^{+},0) P F ( K ; L + , 0 ) and leave the vacuum combination M r e s . [ + ] − Θ ( μ c 2 − p 2 ) M l o c . [ + ] \mathcal{M}_{\mathrm{res.}}^{[+]}-\Theta(\mu_{c}^{2}-p^{2})\mathcal{M}_{\mathrm{loc.}}^{[+]} M res. [ + ] − Θ ( μ c 2 − p 2 ) M loc. [ + ] . Together with the virtual term retained above, this gives
I q i ( 1 ) ( x , ℓ , K ) = P F ( K ; L + , 0 ) I q i ( 1 ) , v a c ( x , ℓ ) + ∫ 0 1 d z ∫ p ( 2 π ) 4 d σ ^ q g ∣ p 0 m e d d z d 2 p d 2 K M r e s . [ + ] . (40) I_{q_i}^{(1)}(x,\ell,K)=\mathcal{P}_{F}(K;L^{+},0)I_{q_i}^{(1),\mathrm{vac}}(x,\ell)+\int_{0}^{1}dz\int_{\boldsymbol{p}}(2\pi)^{4}\frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz\,d^{2}p\,d^{2}K}\mathcal{M}_{\mathrm{res.}}^{[+]}. \tag*{(40)} I q i ( 1 ) ( x , ℓ , K ) = P F ( K ; L + , 0 ) I q i ( 1 ) , vac ( x , ℓ ) + ∫ 0 1 d z ∫ p ( 2 π ) 4 d z d 2 p d 2 K d σ ^ q g ∣ p 0 med M res. [ + ] . (40)
Here I q i ( 1 ) , v a c ( x , ℓ ) I_{q_i}^{(1),\mathrm{vac}}(x,\ell) I q i ( 1 ) , vac ( x , ℓ ) is the usual vacuum matching coefficient with the factor ( 2 π ) 2 δ ( 2 ) ( K ) (2\pi)^{2}\delta^{(2)}(K) ( 2 π ) 2 δ ( 2 ) ( K ) removed. Its dependence on μ \mu μ and ν \nu ν is implicit and is inherited by J v a c ( b , μ , ν ) J^{\mathrm{vac}}(b,\mu,\nu) J vac ( b , μ , ν ) . Taking the energy-weighted moment of Eq. (22) , the Fourier transform of P F \mathcal{P}_{F} P F gives D F D_{F} D F . Including the zeroth-order term of Eq. (27) , we obtain
J ( b , μ , ν ) = D F [ b ; L + , 0 ] J v a c ( b , μ , ν ) + J ( 1 ) , m e d . r a d . ( b ) + O ( α s 2 ) , (41) J(b,\mu,\nu)=D_{F}[b;L^{+},0]J^{\mathrm{vac}}(b,\mu,\nu)+J^{(1),\mathrm{med.\ rad.}}(b)+\mathcal{O}(\alpha_{s}^{2}), \tag*{(41)} J ( b , μ , ν ) = D F [ b ; L + , 0 ] J vac ( b , μ , ν ) + J ( 1 ) , med. rad. ( b ) + O ( α s 2 ) , (41)
with
J ( 1 ) , m e d . r a d . ( b ) = ∫ 0 1 d z ∫ K , p ( 2 π ) 4 d σ ^ q g ∣ p 0 m e d d z d 2 p d 2 K ∑ i = q , g , u h i ( z ) e i b ⋅ ( K + a i p ) , (42) J^{(1),\mathrm{med.\ rad.}}(b)=\int_{0}^{1}dz\int_{K,p}(2\pi)^{4}\frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz\,d^{2}p\,d^{2}K}\sum_{i=q,g,u}h_{i}(z)e^{i\boldsymbol{b}\cdot(K+a_{i}p)}, \tag*{(42)} J ( 1 ) , med. rad. ( b ) = ∫ 0 1 d z ∫ K , p ( 2 π ) 4 d z d 2 p d 2 K d σ ^ q g ∣ p 0 med i = q , g , u ∑ h i ( z ) e i b ⋅ ( K + a i p ) , (42)
where the energy-weighted, plus-subtracted measurement of Eq. (31) gives the three channels
( h q , a q ) = ( z , 1 z ) , ( h g , a g ) = ( 1 − z , − 1 1 − z ) , ( h u , a u ) = ( − 1 , 1 ) , (43) (h_q,a_q)=\left(z,\frac{1}{z}\right),\qquad(h_g,a_g)=\left(1-z,-\frac{1}{1-z}\right),\qquad(h_u,a_u)=(-1,1), \tag*{(43)} ( h q , a q ) = ( z , z 1 ) , ( h g , a g ) = ( 1 − z , − 1 − z 1 ) , ( h u , a u ) = ( − 1 , 1 ) , (43)
with u u u denoting the unresolved subtraction channel. At b = 0 b=0 b = 0 all phases of the channel sum coincide and h q + h g + h u = 0 h_q+h_g+h_u=0 h q + h g + h u = 0 , so that
J ( 1 ) , m e d . r a d . ( 0 ) = 0 , (44) J^{(1),\mathrm{med.\ rad.}}(0)=0, \tag*{(44)} J ( 1 ) , med. rad. ( 0 ) = 0 , (44)
which is the energy sum rule for the medium-induced radiative correction. Integrated over p p p and K K K , the medium-induced cross section is the harmonic-oscillator BDMPS–Z spectrum [69 , 70 ] , which grows like ( 1 − z ) − 3 / 2 (1-z)^{-3/2} ( 1 − z ) − 3/2 as z → 1 z\to1 z → 1 and like z − 1 / 2 z^{-1/2} z − 1/2 as z → 0 z\to0 z → 0 (App. C). At z → 1 z\to1 z → 1 the energy-weighted measurement summed over channels is suppressed by one power of 1 − z 1-z 1 − z , since h g = 1 − z h_g=1-z h g = 1 − z and z e i b ⋅ p / z − e i b ⋅ p = O ( 1 − z ) z e^{i\boldsymbol{b}\cdot\boldsymbol{p}/z}-e^{i\boldsymbol{b}\cdot\boldsymbol{p}}=O(1-z) z e i b ⋅ p / z − e i b ⋅ p = O ( 1 − z ) , so that the z z z integrand of the jet function behaves at most as ( 1 − z ) − 1 / 2 (1-z)^{-1/2} ( 1 − z ) − 1/2 . Therefore, the radiative correction introduces no soft divergence into the jet function. At z → 0 z\to0 z → 0 the gluon and unresolved weights tend to 1 1 1 and − 1 -1 − 1 , and the integrable z − 1 / 2 z^{-1/2} z − 1/2 behaviour remains.
Equation (41) is the initial condition of the jet function at its natural scales, μ = μ ˉ b \mu=\bar{\mu}_b μ = μ ˉ b of Eq. (16) and ν = ν J \nu=\nu_J ν = ν J . The hard and soft functions are those of vacuum, so the full jet function evolves with the vacuum anomalous dimensions, which is consistent at this order because the radiative correction has no scale dependence at O ( α s ) O(\alpha_s) O ( α s ) . In the logarithmic counting, D F [ b ; L + , 0 ] D_F[b;L^+,0] D F [ b ; L + , 0 ] carries no power of α s \alpha_s α s and enters from LL accuracy onwards, while the radiative correction is an O ( α s ) O(\alpha_s) O ( α s ) term without large logarithms, of the order of the one-loop constants of the hard, soft and jet functions, and is included with them at NLL′ ^\prime ′ . The product of hard, soft and jet functions at their natural scales becomes
D F [ b ; L + , 0 ] 2 [ H S J q v a c J q ˉ v a c ] O ( α s ) + 2 D F [ b ; L + , 0 ] J ( 1 ) , m e d . r a d . ( b ) , (45) D_F[b;L^+,0]^2\left[H S J_q^{\mathrm{vac}}J_{\bar q}^{\mathrm{vac}}\right]_{O(\alpha_s)}+2D_F[b;L^+,0]J^{(1),\mathrm{med.\ rad.}}(b), \tag*{(45)} D F [ b ; L + , 0 ] 2 [ H S J q vac J q ˉ vac ] O ( α s ) + 2 D F [ b ; L + , 0 ] J ( 1 ) , med. rad. ( b ) , (45)
with the vacuum product expanded to first order in α s \alpha_s α s . Since the broadening factor and the radiative correction carry no scale dependence, assigning them to the jet or to the soft functions leaves the product in Eq. (45) unchanged at this order, and both assignments acquire the same vacuum evolution factor. With the broadening factors assigned to the soft function, the broadening term has the form used in earlier studies of dijet decorrelations in heavy ion collisions [12 ] . Terms of order α s 2 \alpha_s^2 α s 2 , among them ( J ( 1 ) , m e d . r a d . ) 2 (J^{(1),\mathrm{med.\ rad.}})^2 ( J ( 1 ) , med. rad. ) 2 and its products with the one-loop vacuum functions, are of the order of the two-loop constants that are not included. The radiative correction takes its coupling at the scale of the one-loop vacuum jet function, α s ( μ ˉ b ) \alpha_s(\bar{\mu}_b) α s ( μ ˉ b ) , which stays perturbative at large b b b , and follows this scale in the variations. A medium scale such as q ^ L \sqrt{\hat{q}L} q ^ L would differ from μ ˉ b \bar{\mu}_b μ ˉ b by a factor of order one for the parameters considered, a change of order α s 2 \alpha_s^2 α s 2 . Finally, unlike the one-loop vacuum functions, which are evaluated at b ˉ \bar{b} b ˉ , the broadening factor and the radiative correction are evaluated at the physical impact parameter b b b .
The evaluation of Eq. (42) requires care at large relative momentum, where the medium-induced cross section falls slowly. There, the leading 1 / p 2 1/p^{2} 1/ p 2 tail corresponds to an early splitting followed by independent broadening of the daughters, with the combined factor D F D A = D F 3 D_{F}D_{A}=D_{F}^{3} D F D A = D F 3 at large N c N_{c} N c . Subtracting the out–out contribution gives, up to terms that oscillate in p 2 p^{2} p 2 or fall faster,
∫ d 2 K ( 2 π ) 2 e i b ⋅ K ( 2 π ) 4 d σ ^ q g ∣ p 0 m e d d z d 2 p d 2 K ⟶ [ D F 3 ( b ; L + , 0 ) − D F ( b ; L + , 0 ) ] ( 2 π ) 2 d σ ^ q g ∣ p 0 v a c d z d 2 p . (46) \int\frac{d^{2}K}{(2\pi)^{2}} e^{i\boldsymbol{b}\cdot\boldsymbol{K}} (2\pi)^{4}\frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz\,d^{2}p\,d^{2}K}
\longrightarrow
\left[D_{F}^{3}(b;L^{+},0)-D_{F}(b;L^{+},0)\right](2\pi)^{2}\frac{d\hat{\sigma}^{\mathrm{vac}}_{qg|p_{0}}}{dz\,d^{2}p}. \tag*{(46)} ∫ ( 2 π ) 2 d 2 K e i b ⋅ K ( 2 π ) 4 d z d 2 p d 2 K d σ ^ q g ∣ p 0 med ⟶ [ D F 3 ( b ; L + , 0 ) − D F ( b ; L + , 0 ) ] ( 2 π ) 2 d z d 2 p d σ ^ q g ∣ p 0 vac . (46)
The separate time regions have larger tails, which cancel in this sum (App. B). The coefficient in Eq. (46) vanishes at b = 0 b=0 b = 0 . For b ≠ 0 b\ne0 b = 0 the p p p integral converges only through the measurement phases e i a i ⋅ b ⋅ p e^{i\boldsymbol{a}_{i}\cdot\boldsymbol{b}\cdot\boldsymbol{p}} e i a i ⋅ b ⋅ p , which after the angular integration give radial integrands of order J 0 ( ∣ a i ∣ b ∣ p ∣ ) / ∣ p ∣ J_{0}(|a_{i}|b|p|)/|p| J 0 ( ∣ a i ∣ b ∣ p ∣ ) /∣ p ∣ . Therefore, the integrand is not absolutely integrable in p p p and the vertex times jointly, and the order of integration matters. The jet function is defined with the time integrals done first, at fixed p p p , and the p p p integral last. To perform the p p p integral analytically before the time integrals (App. B), we include the regulator e − η U V p 2 e^{-\eta_{\mathrm{UV}}p^{2}} e − η UV p 2 , η U V > 0 \eta_{\mathrm{UV}}>0 η UV > 0 , the same in every contribution. For η U V > 0 \eta_{\mathrm{UV}}>0 η UV > 0 the integrand is absolutely integrable, so that the two orders agree, and the regulated result tends to the jet function as η U V → 0 + \eta_{\mathrm{UV}}\to0^{+} η UV → 0 + , with the limit taken after the time integrations. The radiative correction is then
J ( 1 ) , m e d . r a d . ( b ) = lim η U V → 0 + ∑ i = q , g , u ∫ 0 1 d z h i ( z ) ∫ R 2 d 2 p ( 2 π ) 2 e − η U V p 2 e i a i ⋅ b ⋅ p × ∫ d 2 K ( 2 π ) 2 e i b ⋅ K ( 2 π ) 4 [ d σ ^ i n − i n d z d 2 p d 2 K + d σ ^ i n − o u t d z d 2 p d 2 K ] . (47) J^{(1),\mathrm{med.rad.}}(b)=\lim_{\eta_{\mathrm{UV}}\to0^{+}}\sum_{i=q,g,u}\int_{0}^{1} dz\,h_{i}(z)\int_{\mathbb{R}^{2}}\frac{d^{2}p}{(2\pi)^{2}}e^{-\eta_{\mathrm{UV}}p^{2}}e^{i\boldsymbol{a}_{i}\cdot\boldsymbol{b}\cdot\boldsymbol{p}}
\times\int\frac{d^{2}K}{(2\pi)^{2}}e^{i\boldsymbol{b}\cdot\boldsymbol{K}}(2\pi)^{4}\left[\frac{d\hat{\sigma}_{\mathrm{in-in}}}{dz\,d^{2}p\,d^{2}K}+\frac{d\hat{\sigma}_{\mathrm{in-out}}}{dz\,d^{2}p\,d^{2}K}\right]. \tag*{(47)} J ( 1 ) , med.rad. ( b ) = η UV → 0 + lim i = q , g , u ∑ ∫ 0 1 d z h i ( z ) ∫ R 2 ( 2 π ) 2 d 2 p e − η UV p 2 e i a i ⋅ b ⋅ p × ∫ ( 2 π ) 2 d 2 K e i b ⋅ K ( 2 π ) 4 [ d z d 2 p d 2 K d σ ^ in − in + d z d 2 p d 2 K d σ ^ in − out ] . (47)
The analytic p p p integration and a subtraction that improves the numerical convergence are given in App. B. Section IV computes the in–in and in–out contributions at fixed relative momentum.
Analytic evaluation of the medium-induced radiative correction
In this section we analytically evaluate the in–in and in–out contributions to Eq. (47) at fixed relative momentum p p p , for a homogeneous medium in the large N c N_{c} N c harmonic-oscillator approximation of Eq. (28) . The remaining p p p integration is carried out in App. B. Following Eqs. (33) and (34) , the K K K transform of the medium-induced radiative cross section at fixed p p p , i.e., the second line of Eq. (47) , is
∫ d 2 K ( 2 π ) 2 e i K ⋅ b ( 2 π ) 4 d σ ^ q g ∣ p 0 m e d d z d 2 p d 2 K = 2 C F α s ( 2 ω + ) 2 P q → q g v a c ( z ) 2 Re ∫ x + < L + , x + < x ˉ + e − ( x ˉ + + x + ) × ∇ v ∫ d 2 u D F [ b ; x + , 0 ] S ~ q → q g ( 3 ) ( u , x ˉ + ; v , x + ; b , z ) ∣ v = 0 ⋅ ∇ w ∫ d 2 Δ e − i Δ ⋅ p S ~ q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b , z ) ∣ w = 0 . (48) \int\frac{d^{2}K}{(2\pi)^{2}}e^{i\boldsymbol{K}\cdot\boldsymbol{b}}(2\pi)^{4}\frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz\,d^{2}p\,d^{2}K}
=
\frac{2C_{F}\alpha_{s}}{(2\omega^{+})^{2}}P_{q\to qg}^{\mathrm{vac}}(z)\,2\operatorname{Re}\int_{x^{+}<L^{+},\,x^{+}<\bar{x}^{+}}e^{-(\bar{x}^{+}+x^{+})}
\times\nabla_{v}\int d^{2}u\,D_{F}[b;x^{+},0]\,\widetilde{S}^{(3)}_{q\to qg}(u,\bar{x}^{+};v,x^{+};b,z)\bigg|_{v=0}
\cdot\nabla_{w}\int d^{2}\Delta\,e^{-i\Delta\cdot p}\widetilde{S}^{(4)}_{q\to qg}(x_{\infty}^{+},\Delta;\bar{x}^{+},u,w;b,z)\bigg|_{w=0}. \tag*{(48)} ∫ ( 2 π ) 2 d 2 K e i K ⋅ b ( 2 π ) 4 d z d 2 p d 2 K d σ ^ q g ∣ p 0 med = ( 2 ω + ) 2 2 C F α s P q → q g vac ( z ) 2 Re ∫ x + < L + , x + < x ˉ + e − ( x ˉ + + x + ) × ∇ v ∫ d 2 u D F [ b ; x + , 0 ] S q → q g ( 3 ) ( u , x ˉ + ; v , x + ; b , z ) v = 0 ⋅ ∇ w ∫ d 2 Δ e − i Δ ⋅ p S q → q g ( 4 ) ( x ∞ + , Δ ; x ˉ + , u , w ; b , z ) w = 0 . (48)
Here S ~ q → q g ( 3 ) \widetilde{\mathcal{S}}^{(3)}_{q\to qg} S q → q g ( 3 ) is the coordinate-space representation of the three-point function in Eq. (33) , with initial relative coordinate r i = v \boldsymbol{r}_i=\boldsymbol{v} r i = v at x + x^+ x + and final relative coordinate r f = u \boldsymbol{r}_f=\boldsymbol{u} r f = u at x ˉ + \bar{x}^+ x ˉ + . The colour structure of the four-point function S ~ q → q g ( 4 ) \widetilde{\mathcal{S}}^{(4)}_{q\to qg} S q → q g ( 4 ) matters only in the in–in region, x ˉ + < L + \bar{x}^+<L^+ x ˉ + < L + , since in the in–out region it is trivially given by its vacuum counterpart. Its colour structure is spanned by the dipole–dipole and quadrupole correlators
C 1 ( 4 ) = 1 N c 2 ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) ] Tr [ U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ , C 2 ( 4 ) = 1 N c ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ . (49) \begin{aligned}
\mathcal{C}^{(4)}_1 &= \frac{1}{N_c^2}\left\langle\operatorname{Tr}\left[U_F(\boldsymbol{r}_1)U_F^\dagger(\boldsymbol{r}_2)\right]\operatorname{Tr}\left[U_F(\bar{\boldsymbol{r}}_2)U_F^\dagger(\bar{\boldsymbol{r}}_1)\right]\right\rangle,\\
\mathcal{C}^{(4)}_2 &= \frac{1}{N_c}\left\langle\operatorname{Tr}\left[U_F(\boldsymbol{r}_1)U_F^\dagger(\boldsymbol{r}_2)U_F(\bar{\boldsymbol{r}}_2)U_F^\dagger(\bar{\boldsymbol{r}}_1)\right]\right\rangle. \tag*{(49)}
\end{aligned} C 1 ( 4 ) C 2 ( 4 ) = N c 2 1 ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) ] Tr [ U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ , = N c 1 ⟨ Tr [ U F ( r 1 ) U F † ( r 2 ) U F ( r ˉ 2 ) U F † ( r ˉ 1 ) ] ⟩ . (49)
All Wilson lines run from x ˉ + \bar{x}^+ x ˉ + to the time t t t along the four paths of Eq. (37) . The quadrupole in Eq. (36) is C 2 ( 4 ) \mathcal{C}^{(4)}_2 C 2 ( 4 ) , so that C q → q g ( 4 ) = D F [ b − z ( v − v ˉ ) ] C 2 ( 4 ) \mathcal{C}^{(4)}_{q\to qg}=D_F[\boldsymbol{b}-z(\boldsymbol{v}-\bar{\boldsymbol{v}})]\mathcal{C}^{(4)}_2 C q → q g ( 4 ) = D F [ b − z ( v − v ˉ )] C 2 ( 4 ) . At large N c N_c N c the dipole–dipole component evolves autonomously and sources the quadrupole, so that the evolution is triangular. In the harmonic-oscillator approximation it reads [73 , 74 ]
∂ t ( C 1 ( 4 ) C 2 ( 4 ) ) = − q ^ ( t ) 4 ( V 1 0 T V 2 ) ( C 1 ( 4 ) C 2 ( 4 ) ) , (50) \partial_t
\begin{pmatrix}
\mathcal{C}^{(4)}_1\\
\mathcal{C}^{(4)}_2
\end{pmatrix}
=
-\frac{\hat{q}(t)}{4}
\begin{pmatrix}
\mathcal{V}_1 & 0\\
\mathcal{T} & \mathcal{V}_2
\end{pmatrix}
\begin{pmatrix}
\mathcal{C}^{(4)}_1\\
\mathcal{C}^{(4)}_2
\end{pmatrix}, \tag*{(50)} ∂ t ( C 1 ( 4 ) C 2 ( 4 ) ) = − 4 q ^ ( t ) ( V 1 T 0 V 2 ) ( C 1 ( 4 ) C 2 ( 4 ) ) , (50)
with C 1 ( 4 ) = C 2 ( 4 ) = 1 \mathcal{C}^{(4)}_1=\mathcal{C}^{(4)}_2=1 C 1 ( 4 ) = C 2 ( 4 ) = 1 at t = x ˉ + t=\bar{x}^+ t = x ˉ + , where the Wilson lines have zero extent. In terms of the instantaneous relative coordinate Δ ( t ) = v ( t ) − v ˉ ( t ) \boldsymbol{\Delta}(t)=\boldsymbol{v}(t)-\bar{\boldsymbol{v}}(t) Δ ( t ) = v ( t ) − v ˉ ( t ) , the potentials are
V 1 = v 2 + v ˉ 2 , V 2 = 2 b 2 + 2 ( 1 − 2 z ) b ⋅ Δ + [ z 2 + ( 1 − z ) 2 ] Δ 2 , T ( Δ , b ) = 2 z ( 1 − z ) Δ 2 − 2 b 2 − 2 ( 1 − 2 z ) b ⋅ Δ . (51) \begin{aligned}
\mathcal{V}_1 &= \boldsymbol{v}^{\,2}+\bar{\boldsymbol{v}}^{\,2},\\
\mathcal{V}_2 &= 2\boldsymbol{b}^{\,2}+2(1-2z)\boldsymbol{b}\cdot\boldsymbol{\Delta}+\left[z^2+(1-z)^2\right]\boldsymbol{\Delta}^{\,2},\\
\mathcal{T}(\boldsymbol{\Delta},\boldsymbol{b}) &= 2z(1-z)\boldsymbol{\Delta}^{\,2}-2\boldsymbol{b}^{\,2}-2(1-2z)\boldsymbol{b}\cdot\boldsymbol{\Delta}. \tag*{(51)}
\end{aligned} V 1 V 2 T ( Δ , b ) = v 2 + v ˉ 2 , = 2 b 2 + 2 ( 1 − 2 z ) b ⋅ Δ + [ z 2 + ( 1 − z ) 2 ] Δ 2 , = 2 z ( 1 − z ) Δ 2 − 2 b 2 − 2 ( 1 − 2 z ) b ⋅ Δ . (51)
The diagonal potential V 2 = ( r 1 − r ˉ 1 ) 2 + ( r 2 − r ˉ 2 ) 2 \mathcal{V}_2=(\boldsymbol{r}_1-\bar{\boldsymbol{r}}_1)^2+(\boldsymbol{r}_2-\bar{\boldsymbol{r}}_2)^2 V 2 = ( r 1 − r ˉ 1 ) 2 + ( r 2 − r ˉ 2 ) 2 is the sum of the squared sizes of the quark and gluon dipoles that connect the amplitude and the conjugate amplitude, and the transition potential is T = Δ 2 − V 2 \mathcal{T}=\boldsymbol{\Delta}^{\,2}-\mathcal{V}_2 T = Δ 2 − V 2 . As a check, when the amplitude or the conjugate-amplitude pair is unresolved, v = 0 \boldsymbol{v}=0 v = 0 or v ˉ = 0 \bar{\boldsymbol{v}}=0 v ˉ = 0 , one has T + V 2 = V 1 \mathcal{T}+\mathcal{V}_2=\mathcal{V}_1 T + V 2 = V 1 , and Eq. (50) gives C 2 ( 4 ) = C 1 ( 4 ) \mathcal{C}^{(4)}_2=\mathcal{C}^{(4)}_1 C 2 ( 4 ) = C 1 ( 4 ) . The four-point function then reduces to two dipoles, which is the structure of the three-point function in Eq. (33) with the roles of amplitude and conjugate amplitude exchanged. Since the evolution is triangular, its solution separates exactly into diagonal propagation and a single off-diagonal colour transition,
S ~ q → q g ( 4 ) = S ~ q → q g ( 4 ) , f a c + S ~ q → q g ( 4 ) , n o n f a c . (52) \widetilde{\mathcal{S}}^{(4)}_{q\to qg}
=
\widetilde{\mathcal{S}}^{(4),\mathrm{fac}}_{q\to qg}
+
\widetilde{\mathcal{S}}^{(4),\mathrm{nonfac}}_{q\to qg}. \tag*{(52)} S q → q g ( 4 ) = S q → q g ( 4 ) , fac + S q → q g ( 4 ) , nonfac . (52)
The first term is the homogeneous solution of Eq. (50) . The second contains one insertion of T \mathcal{T} T at an intermediate time, with diagonal propagation on either side. Since V 2 \mathcal{V}_2 V 2 is a sum of two dipole potentials, the first term is a product of a quark and a gluon dipole, and we refer to the two terms as the factorizable and nonfactorizable parts.
Factorizable contribution
At large N c N_c N c , the factorizable correlator is
C q → q g ( 4 ) , f a c ( v , v ˉ , b ; L + , x ˉ + ) = D F ( b + ( 1 − z ) ( v − v ˉ ) ; L + , x ˉ + ) D F 2 ( b − z ( v − v ˉ ) ; L + , x ˉ + ) . (53) \mathcal{C}^{(4),\mathrm{fac}}_{q\to qg}(\boldsymbol{v},\bar{\boldsymbol{v}},\boldsymbol{b};L^+,\bar{x}^+) = D_F\bigl(\boldsymbol{b}+(1-z)(\boldsymbol{v}-\bar{\boldsymbol{v}});L^+,\bar{x}^+\bigr)D_F^2\bigl(\boldsymbol{b}-z(\boldsymbol{v}-\bar{\boldsymbol{v}});L^+,\bar{x}^+\bigr). \tag*{(53)} C q → q g ( 4 ) , fac ( v , v ˉ , b ; L + , x ˉ + ) = D F ( b + ( 1 − z ) ( v − v ˉ ) ; L + , x ˉ + ) D F 2 ( b − z ( v − v ˉ ) ; L + , x ˉ + ) . (53)
where x ∞ + x^+_\infty x ∞ + was replaced by L + L^+ L + since we take x ˉ + < L + \bar{x}^+<L^+ x ˉ + < L + for the in–in region. It depends on the paths only through Δ \boldsymbol{\Delta} Δ , which the integral over the common final coordinate keeps constant after x ˉ + \bar{x}^+ x ˉ + , equal to u − w \boldsymbol{u}-\boldsymbol{w} u − w . Therefore, at fixed p \boldsymbol{p} p the four-point function is
∫ d 2 Δ e − i Δ ⋅ p S ~ q → q g ( 4 ) , f a c ( L + , Δ ; x ˉ + , u , w ; b , z ) = e − i ( u − w ) ⋅ p D F ( b + ( 1 − z ) ( u − w ) ; L + , x ˉ + ) D F 2 ( b − z ( u − w ) ; L + , x ˉ + ) . (54) \begin{aligned}
\int d^2\boldsymbol{\Delta}\,e^{-i\boldsymbol{\Delta}\cdot\boldsymbol{p}}\widetilde{\mathcal{S}}^{(4),\mathrm{fac}}_{q\to qg}(L^+,\boldsymbol{\Delta};\bar{x}^+,\boldsymbol{u},\boldsymbol{w};\boldsymbol{b},z) \\
&= e^{-i(\boldsymbol{u}-\boldsymbol{w})\cdot\boldsymbol{p}}D_F\bigl(\boldsymbol{b}+(1-z)(\boldsymbol{u}-\boldsymbol{w});L^+,\bar{x}^+\bigr)D_F^2\bigl(\boldsymbol{b}-z(\boldsymbol{u}-\boldsymbol{w});L^+,\bar{x}^+\bigr). \tag*{(54)}
\end{aligned} ∫ d 2 Δ e − i Δ ⋅ p S q → q g ( 4 ) , fac ( L + , Δ ; x ˉ + , u , w ; b , z ) = e − i ( u − w ) ⋅ p D F ( b + ( 1 − z ) ( u − w ) ; L + , x ˉ + ) D F 2 ( b − z ( u − w ) ; L + , x ˉ + ) . (54)
Since Eq. (54) depends on u \boldsymbol{u} u and w \boldsymbol{w} w only through u − w \boldsymbol{u}-\boldsymbol{w} u − w , the derivative ∇ w \nabla_{\boldsymbol{w}} ∇ w can be traded for − ∇ u -\nabla_{\boldsymbol{u}} − ∇ u , and integration by parts in u \boldsymbol{u} u gives
∇ w ∫ d 2 u S ~ q → q g ( 3 ) , i n − i n ∣ v = 0 ⋅ ∇ w ∫ d 2 Δ e − i Δ ⋅ p S ~ q → q g ( 4 ) , f a c ∣ w = 0 = ∫ d 2 u e − i u ⋅ p D F ( b + ( 1 − z ) u ; L + , x ˉ + ) D F 2 ( b − z u ; L + , x ˉ + ) ( ∇ u ⋅ ∇ v ) S ~ q → q g ( 3 ) , i n − i n ∣ v = 0 . (55) \begin{aligned}
\left.\nabla_{\boldsymbol{w}}\int d^2\boldsymbol{u}\,\widetilde{\mathcal{S}}^{(3),\mathrm{in-in}}_{q\to qg}\right|_{\boldsymbol{v}=0}\cdot\nabla_{\boldsymbol{w}}\left.\int d^2\boldsymbol{\Delta}\,e^{-i\boldsymbol{\Delta}\cdot\boldsymbol{p}}\widetilde{\mathcal{S}}^{(4),\mathrm{fac}}_{q\to qg}\right|_{\boldsymbol{w}=0} \\
&= \int d^2\boldsymbol{u}\,e^{-i\boldsymbol{u}\cdot\boldsymbol{p}}D_F\bigl(\boldsymbol{b}+(1-z)\boldsymbol{u};L^+,\bar{x}^+\bigr)D_F^2\bigl(\boldsymbol{b}-z\boldsymbol{u};L^+,\bar{x}^+\bigr)\left.\left(\nabla_{\boldsymbol{u}}\cdot\nabla_{\boldsymbol{v}}\right)\widetilde{\mathcal{S}}^{(3),\mathrm{in-in}}_{q\to qg}\right|_{\boldsymbol{v}=0}. \tag*{(55)}
\end{aligned} ∇ w ∫ d 2 u S q → q g ( 3 ) , in − in v = 0 ⋅ ∇ w ∫ d 2 Δ e − i Δ ⋅ p S q → q g ( 4 ) , fac w = 0 = ∫ d 2 u e − i u ⋅ p D F ( b + ( 1 − z ) u ; L + , x ˉ + ) D F 2 ( b − z u ; L + , x ˉ + ) ( ∇ u ⋅ ∇ v ) S q → q g ( 3 ) , in − in v = 0 . (55)
The boundary term vanishes for x ˉ + < L + \bar{x}^+<L^+ x ˉ + < L + , where the final state broadening factors decay at large ∣ u ∣ |\boldsymbol{u}| ∣ u ∣ , and the endpoint x ˉ + = L + \bar{x}^+=L^+ x ˉ + = L + follows by continuity. For a time-independent and homogeneous medium, we denote the two-dimensional harmonic-oscillator propagator with constant frequency Ω \Omega Ω by
K Ω ( x f , t f ; x i , t i ) = ∫ x ( t i ) = x i x ( t f ) = x f D x exp { i ω + 2 ∫ t i t f d s + [ x ˙ 2 − Ω 2 x 2 ] } = ω + c 1 π exp { i ω + 2 c 2 ( x f 2 + x i 2 ) + 2 ω + c 1 x f ⋅ x i } , \begin{aligned}
\mathcal{K}_{\Omega}(\boldsymbol{x}_f,t_f;\boldsymbol{x}_i,t_i)=\int_{\boldsymbol{x}(t_i)=\boldsymbol{x}_i}^{\boldsymbol{x}(t_f)=\boldsymbol{x}_f}\mathcal{D}\boldsymbol{x}\exp\left\{\frac{i\omega^+}{2}\int_{t_i}^{t_f}ds^+\left[\dot{\boldsymbol{x}}^2-\Omega^2\boldsymbol{x}^2\right]\right\} \\
&=\frac{\omega^+c_1}{\pi}\exp\left\{\frac{i\omega^+}{2}c_2\left(\boldsymbol{x}_f^2+\boldsymbol{x}_i^2\right)+2\omega^+c_1\boldsymbol{x}_f\cdot\boldsymbol{x}_i\right\},
\end{aligned} K Ω ( x f , t f ; x i , t i ) = ∫ x ( t i ) = x i x ( t f ) = x f D x exp { 2 i ω + ∫ t i t f d s + [ x ˙ 2 − Ω 2 x 2 ] } = π ω + c 1 exp { 2 i ω + c 2 ( x f 2 + x i 2 ) + 2 ω + c 1 x f ⋅ x i } ,
with
c 1 = Ω 2 i sin [ Ω ( t f − t i ) ] , c 2 = Ω tan [ Ω ( t f − t i ) ] . c_1=\frac{\Omega}{2i\sin[\Omega(t_f-t_i)]},\qquad c_2=\frac{\Omega}{\tan[\Omega(t_f-t_i)]}. c 1 = 2 i sin [ Ω ( t f − t i )] Ω , c 2 = tan [ Ω ( t f − t i )] Ω .
and with K 0 \mathcal{K}_0 K 0 the free propagator. The in–in three-point function is an oscillator with frequency Ω 3 2 = − i ( 1 + z 2 ) q ^ / ( 2 ω + ) \Omega_3^2=-i(1+z^2)\hat{q}/(2\omega^+) Ω 3 2 = − i ( 1 + z 2 ) q ^ / ( 2 ω + ) , centered at z b / ( 1 + z 2 ) z\boldsymbol{b}/(1+z^2) z b / ( 1 + z 2 ) ,
S ~ q → q g ( 3 ) , i n − i n ( u , x ˉ + ; v , x + ; b , z ) = exp { − q ^ b 2 4 ( 1 + z 2 ) τ } K Ω 3 ( u − z 1 + z 2 b , x ˉ + ; v − z 1 + z 2 b , x + ) . (56) \widetilde{\mathcal{S}}^{(3),\mathrm{in-in}}_{q\to qg}(\boldsymbol{u},\bar{x}^+;\boldsymbol{v},x^+;\boldsymbol{b},z)=\exp\left\{-\frac{\hat{q}\boldsymbol{b}^2}{4(1+z^2)}\tau\right\}\mathcal{K}_{\Omega_3}\left(\boldsymbol{u}-\frac{z}{1+z^2}\boldsymbol{b},\bar{x}^+;\boldsymbol{v}-\frac{z}{1+z^2}\boldsymbol{b},x^+\right). \tag*{(56)} S q → q g ( 3 ) , in − in ( u , x ˉ + ; v , x + ; b , z ) = exp { − 4 ( 1 + z 2 ) q ^ b 2 τ } K Ω 3 ( u − 1 + z 2 z b , x ˉ + ; v − 1 + z 2 z b , x + ) . (56)
with τ = x ˉ + − x + \tau=\bar{x}^{+}-x^{+} τ = x ˉ + − x + . With the mixed derivative of Eq. (56) inserted, the remaining u \mathbf{u} u integral in Eq. (55) is Gaussian and we display its result through the two scalar coefficients
A G ( x ˉ + , x + ) = q ^ 4 ( L + − x ˉ + ) ( ( 1 − z ) 2 + 2 z 2 ) − i ω + 2 c 2 ( x ˉ + − x + ) , A_{\mathrm{G}}(\bar{x}^{+},x^{+})=\frac{\hat{q}}{4}(L^{+}-\bar{x}^{+})\left((1-z)^{2}+2z^{2}\right)-\frac{i\omega^{+}}{2}c_{2}(\bar{x}^{+}-x^{+}), A G ( x ˉ + , x + ) = 4 q ^ ( L + − x ˉ + ) ( ( 1 − z ) 2 + 2 z 2 ) − 2 i ω + c 2 ( x ˉ + − x + ) ,
B G ( x ˉ + , x + ) = q ^ 2 ( L + − x ˉ + ) ( 1 − 3 z ) + z ω + 1 + z 2 [ 2 c 1 ( x ˉ + − x + ) + i c 2 ( x ˉ + − x + ) ] , (57) B_{\mathrm{G}}(\bar{x}^{+},x^{+})=\frac{\hat{q}}{2}(L^{+}-\bar{x}^{+})(1-3z)+\frac{z\omega^{+}}{1+z^{2}}\left[2c_{1}(\bar{x}^{+}-x^{+})+ic_{2}(\bar{x}^{+}-x^{+})\right], \tag*{(57)} B G ( x ˉ + , x + ) = 2 q ^ ( L + − x ˉ + ) ( 1 − 3 z ) + 1 + z 2 z ω + [ 2 c 1 ( x ˉ + − x + ) + i c 2 ( x ˉ + − x + ) ] , (57)
which contain no transverse variables. The transverse dependence is then explicit in
∫ d 2 u e − i u ⋅ p D F ( b + ( 1 − z ) u ; L + , x ˉ + ) D F 2 ( b − z u ; L + , x ˉ + ) ( ∇ u ⋅ ∇ v ) S ~ q → q g ( 3 ) , i n − i n ∣ v = 0 = ω + c 1 ( x ˉ + − x + ) A G ( x ˉ + , x + ) exp { − q ^ b 2 4 ( 1 + z 2 ) ( x ˉ + − x + ) − 3 q ^ b 2 4 ( L + − x ˉ + ) + z 2 ω + b 2 ( 1 + z 2 ) 2 [ 2 c 1 ( x ˉ + − x + ) + i c 2 ( x ˉ + − x + ) ] − p 2 − 2 i B G ( x ˉ + , x + ) b ⋅ p − B G 2 ( x ˉ + , x + ) b 2 4 A G ( x ˉ + , x + ) } × { 4 ω + c 1 + z 2 ( ω + ) 2 b 2 ( 1 + z 2 ) 2 ( 4 i c 1 c 2 + 4 c 1 2 − c 2 2 ) + 2 i ( ω + ) 2 c 1 c 2 ( 1 A G + − p 2 + 2 i B G b ⋅ p + B G 2 b 2 4 A G 2 ) + z ( ω + ) 2 1 + z 2 ( 4 i c 1 c 2 + 4 c 1 2 − c 2 2 ) i b ⋅ p + B G b 2 2 A G } . (58) \begin{aligned}
\int d^{2}\mathbf{u}\,e^{-i\mathbf{u}\cdot\mathbf{p}}D_{\mathrm{F}}(\mathbf{b}+(1-z)\mathbf{u};L^{+},\bar{x}^{+})D_{\mathrm{F}}^{2}(\mathbf{b}-z\mathbf{u};L^{+},\bar{x}^{+})\left.(\nabla_{\mathbf{u}}\cdot\nabla_{\mathbf{v}})\widetilde{\mathcal{S}}_{q\to qg}^{(3),\mathrm{in-in}}\right|_{\mathbf{v}=0} \\
&=\frac{\omega^{+}c_{1}(\bar{x}^{+}-x^{+})}{A_{\mathrm{G}}(\bar{x}^{+},x^{+})}\exp\left\{-\frac{\hat{q}\mathbf{b}^{2}}{4(1+z^{2})}(\bar{x}^{+}-x^{+})-\frac{3\hat{q}\mathbf{b}^{2}}{4}(L^{+}-\bar{x}^{+})+\frac{z^{2}\omega^{+}\mathbf{b}^{2}}{(1+z^{2})^{2}}\left[2c_{1}(\bar{x}^{+}-x^{+})+ic_{2}(\bar{x}^{+}-x^{+})\right]-\frac{\mathbf{p}^{2}-2iB_{\mathrm{G}}(\bar{x}^{+},x^{+})\mathbf{b}\cdot\mathbf{p}-B_{\mathrm{G}}^{2}(\bar{x}^{+},x^{+})\mathbf{b}^{2}}{4A_{\mathrm{G}}(\bar{x}^{+},x^{+})}\right\} \\
&\times\left\{4\omega^{+}c_{1}+\frac{z^{2}(\omega^{+})^{2}\mathbf{b}^{2}}{(1+z^{2})^{2}}\left(4ic_{1}c_{2}+4c_{1}^{2}-c_{2}^{2}\right)+2i(\omega^{+})^{2}c_{1}c_{2}\left(\frac{1}{A_{\mathrm{G}}}+\frac{-\mathbf{p}^{2}+2iB_{\mathrm{G}}\mathbf{b}\cdot\mathbf{p}+B_{\mathrm{G}}^{2}\mathbf{b}^{2}}{4A_{\mathrm{G}}^{2}}\right)+\frac{z(\omega^{+})^{2}}{1+z^{2}}\left(4ic_{1}c_{2}+4c_{1}^{2}-c_{2}^{2}\right)\frac{i\mathbf{b}\cdot\mathbf{p}+B_{\mathrm{G}}\mathbf{b}^{2}}{2A_{\mathrm{G}}}\right\}. \tag*{(58)}
\end{aligned} ∫ d 2 u e − i u ⋅ p D F ( b + ( 1 − z ) u ; L + , x ˉ + ) D F 2 ( b − z u ; L + , x ˉ + ) ( ∇ u ⋅ ∇ v ) S q → q g ( 3 ) , in − in v = 0 = A G ( x ˉ + , x + ) ω + c 1 ( x ˉ + − x + ) exp { − 4 ( 1 + z 2 ) q ^ b 2 ( x ˉ + − x + ) − 4 3 q ^ b 2 ( L + − x ˉ + ) + ( 1 + z 2 ) 2 z 2 ω + b 2 [ 2 c 1 ( x ˉ + − x + ) + i c 2 ( x ˉ + − x + ) ] − 4 A G ( x ˉ + , x + ) p 2 − 2 i B G ( x ˉ + , x + ) b ⋅ p − B G 2 ( x ˉ + , x + ) b 2 } × { 4 ω + c 1 + ( 1 + z 2 ) 2 z 2 ( ω + ) 2 b 2 ( 4 i c 1 c 2 + 4 c 1 2 − c 2 2 ) + 2 i ( ω + ) 2 c 1 c 2 ( A G 1 + 4 A G 2 − p 2 + 2 i B G b ⋅ p + B G 2 b 2 ) + 1 + z 2 z ( ω + ) 2 ( 4 i c 1 c 2 + 4 c 1 2 − c 2 2 ) 2 A G i b ⋅ p + B G b 2 } . (58)
In the last two lines c 1 c_{1} c 1 , c 2 c_{2} c 2 , A G A_{\mathrm{G}} A G and B G B_{\mathrm{G}} B G carry the arguments displayed above. Before the relative-momentum integration, the factorizable in–in contribution is
J i n − i n ( 1 ) , m e d . r a d . f a c ( b ) = 2 C F α s ∑ i = q , g , u ∫ 0 1 d z P q → q g v a c ( z ) h i ( z ) ( 2 ω + ) 2 ∫ R 2 d 2 p ( 2 π ) 2 e − η U V p 2 e i a ⋅ b p J_{\mathrm{in-in}}^{(1),\mathrm{med.\ rad.\ fac}}(\mathbf{b})=2C_{\mathrm{F}}\alpha_{s}\sum_{i=q,g,u}\int_{0}^{1}dz\,\frac{P_{q\to qg}^{\mathrm{vac}}(z)h_{i}(z)}{(2\omega^{+})^{2}}\int_{\mathbb{R}^{2}}\frac{d^{2}\mathbf{p}}{(2\pi)^{2}}e^{-\eta_{\mathrm{UV}}\mathbf{p}^{2}}e^{i\mathbf{a}\cdot\mathbf{b}\mathbf{p}} J in − in ( 1 ) , med. rad. fac ( b ) = 2 C F α s i = q , g , u ∑ ∫ 0 1 d z ( 2 ω + ) 2 P q → q g vac ( z ) h i ( z ) ∫ R 2 ( 2 π ) 2 d 2 p e − η UV p 2 e i a ⋅ bp
× 2 R e ∫ 0 L + d τ ∫ 0 L + − τ d x + e − q ^ b 2 4 x + [ R H S o f E q . ( 59 ) ] x ˉ + = x + + τ . (59) \times2\mathop{\mathrm{Re}}\int_{0}^{L^{+}}d\tau\int_{0}^{L^{+}-\tau}dx^{+}\,e^{-\frac{\hat{q}\mathbf{b}^{2}}{4}x^{+}}\left[\mathrm{RHS\ of\ Eq.\ }(59)\right]_{\bar{x}^{+}=x^{+}+\tau}. \tag*{(59)} × 2 Re ∫ 0 L + d τ ∫ 0 L + − τ d x + e − 4 q ^ b 2 x + [ RHS of Eq. ( 59 ) ] x ˉ + = x + + τ . (59)
The p \mathbf{p} p integral is performed analytically in App. B, which leaves the two time integrations.
Nonfactorizable contribution
Nonfactorizable contributions account for medium induced correlations between the daughter partons (see, e.g., [71 , 73 , 75 ] for recent related calculations). The off-diagonal solution of Eq. (50) has one insertion of − q ^ T / 4 -\hat{q}T/4 − q ^ T /4 at an intermediate time y + y^{+} y + , with C 2 ( 4 ) \mathcal{C}_{2}^{(4)} C 2 ( 4 ) evolving before y + y^{+} y + and C 2 ( 4 ) \mathcal{C}_{2}^{(4)} C 2 ( 4 ) after it. Between x ˉ + \bar{x}^{+} x ˉ + and y + y^{+} y + the amplitude and conjugate-amplitude relative coordinates are coupled through the additional dipole D F [ b − z ( v − v ˉ ) ] D_{\mathrm{F}}[\mathbf{b}-z(\mathbf{v}-\bar{\mathbf{v}})] D F [ b − z ( v − v ˉ )] of Eq. (36) . Since the kinetic terms of v \mathbf{v} v and v ˉ \bar{\mathbf{v}} v ˉ enter with opposite signs, the coupled system is diagonalized by a hyperbolic transformation. In the transformed variables, the kernel factorizes into a harmonic-oscillator propagator and its complex conjugate. The nonfactorizable four-point function can then be written as
S ~ q → q g ( 4 ) , n o n f a c ( L + , Δ ; x ˉ + , u , w ; b , z ) = − 1 4 T ( Δ , b ) ∫ x ˉ + L + d y + q ^ ( y + ) D F ( b + ( 1 − z ) Δ ; L + , y + ) D F 2 ( b − z Δ ; L + , y + ) × exp { − b 2 4 ( 1 + 2 z 2 ) ∫ x ˉ + y + d s + q ^ ( s + ) } ∫ d 2 x K Ω ( X f , y + ; X i , x ˉ + ) K Ω ∗ ( X ˉ f , y + ; X ˉ i , x ˉ + ) , (60) \widetilde{\mathcal{S}}_{q\to qg}^{(4),\mathrm{nonfac}}(L^+,\Delta;\bar{x}^+,u,w;b,z)
= -\frac{1}{4}\mathcal{T}(\Delta,b)\int_{\bar{x}^+}^{L^+} \mathrm{d}y^+\,\hat{q}(y^+)D_F(b+(1-z)\Delta;L^+,y^+)D_F^2(b-z\Delta;L^+,y^+)
\times\exp\left\{-\frac{b^2}{4(1+2z^2)}\int_{\bar{x}^+}^{y^+}\mathrm{d}s^+\,\hat{q}(s^+)\right\}\int\mathrm{d}^2x\,K_\Omega(X_f,y^+;X_i,\bar{x}^+)K_\Omega^*(\bar{X}_f,y^+;\bar{X}_i,\bar{x}^+), \tag*{(60)} S q → q g ( 4 ) , nonfac ( L + , Δ ; x ˉ + , u , w ; b , z ) = − 4 1 T ( Δ , b ) ∫ x ˉ + L + d y + q ^ ( y + ) D F ( b + ( 1 − z ) Δ ; L + , y + ) D F 2 ( b − z Δ ; L + , y + ) × exp { − 4 ( 1 + 2 z 2 ) b 2 ∫ x ˉ + y + d s + q ^ ( s + ) } ∫ d 2 x K Ω ( X f , y + ; X i , x ˉ + ) K Ω ∗ ( X ˉ f , y + ; X ˉ i , x ˉ + ) , (60)
where
λ = 1 + 2 z 2 , β = λ − 1 λ + 1 , γ = 1 1 − β 2 , δ = z λ 3 / 2 , \lambda=\sqrt{1+2z^2},\qquad\beta=\frac{\lambda-1}{\lambda+1},\qquad\gamma=\frac{1}{\sqrt{1-\beta^2}},\qquad\delta=\frac{z}{\lambda^{3/2}}, λ = 1 + 2 z 2 , β = λ + 1 λ − 1 , γ = 1 − β 2 1 , δ = λ 3/2 z ,
X i = γ ( u − β w ) − δ b , X ˉ i = γ ( w − β u ) + δ b , X_i=\gamma(u-\beta w)-\delta b,\qquad\bar{X}_i=\gamma(w-\beta u)+\delta b, X i = γ ( u − β w ) − δ b , X ˉ i = γ ( w − β u ) + δ b ,
X f = ( 1 − β ) γ x + γ β Δ − δ b , X ˉ f = ( 1 − β ) γ x − γ Δ + δ b , X_f=(1-\beta)\gamma x+\gamma\beta\Delta-\delta b,\qquad\bar{X}_f=(1-\beta)\gamma x-\gamma\Delta+\delta b, X f = ( 1 − β ) γ x + γ β Δ − δ b , X ˉ f = ( 1 − β ) γ x − γ Δ + δ b ,
Ω 2 = − i λ q ^ 2 ω + . (61) \Omega^2=-\frac{i\lambda\hat{q}}{2\omega^+}. \tag*{(61)} Ω 2 = − 2 ω + iλ q ^ . (61)
The transformation ( v , v ˉ ) → ( X , X ˉ ) (v,\bar{v})\to(X,\bar{X}) ( v , v ˉ ) → ( X , X ˉ ) leaves the kinetic combination v ˙ 2 − v ˉ ˙ 2 \dot{v}^{\,2}-\dot{\bar{v}}^{\,2} v ˙ 2 − v ˉ ˙ 2 invariant, and in the shifted variables the potential v 2 + v ˉ 2 + ( b − z ( v − v ˉ ) ) 2 v^2+\bar{v}^2+(b-z(v-\bar{v}))^2 v 2 + v ˉ 2 + ( b − z ( v − v ˉ ) ) 2 becomes λ ( X 2 + X ˉ 2 ) + b 2 / ( 1 + 2 z 2 ) \lambda(X^2+\bar{X}^2)+b^2/(1+2z^2) λ ( X 2 + X ˉ 2 ) + b 2 / ( 1 + 2 z 2 ) . This fixes β \beta β , γ \gamma γ , δ \delta δ and the frequency Ω \Omega Ω , and produces the b 2 b^2 b 2 term in the exponent of Eq. (60) . To further analytically evaluate this contribution, we define the intervals τ 3 = x ˉ + − x + \tau_3=\bar{x}^+-x^+ τ 3 = x ˉ + − x + and τ 4 = y + − x ˉ + \tau_4=y^+-\bar{x}^+ τ 4 = y + − x ˉ + , such that L + − y + = L + − x + − τ 3 − τ 4 L^+-y^+=L^+-x^+-\tau_3-\tau_4 L + − y + = L + − x + − τ 3 − τ 4 and the domains read
0 < τ 3 < L + , 0 < τ 4 < L + − τ 3 , 0 < x + < L + − τ 3 − τ 4 . (62) 0<\tau_3<L^+,\qquad0<\tau_4<L^+-\tau_3,\qquad0<x^+<L^+-\tau_3-\tau_4. \tag*{(62)} 0 < τ 3 < L + , 0 < τ 4 < L + − τ 3 , 0 < x + < L + − τ 3 − τ 4 . (62)
For a time-independent, homogeneous medium the oscillator coefficients on the three- and four-point intervals are
c 1 , 3 = Ω 3 2 i sin ( Ω 3 τ 3 ) , c 2 , 3 = Ω 3 tan ( Ω 3 τ 3 ) , c_{1,3}=\frac{\Omega_3}{2i\sin(\Omega_3\tau_3)},\qquad c_{2,3}=\frac{\Omega_3}{\tan(\Omega_3\tau_3)}, c 1 , 3 = 2 i sin ( Ω 3 τ 3 ) Ω 3 , c 2 , 3 = tan ( Ω 3 τ 3 ) Ω 3 ,
c 1 , 4 = Ω 2 i sin ( Ω τ 4 ) , c 2 , 4 = Ω tan ( Ω τ 4 ) , D 4 = c 2 , 4 − c 2 , 4 ∗ , c_{1,4}=\frac{\Omega}{2i\sin(\Omega\tau_4)},\qquad c_{2,4}=\frac{\Omega}{\tan(\Omega\tau_4)},\qquad D_4=c_{2,4}-c_{2,4}^*, c 1 , 4 = 2 i sin ( Ω τ 4 ) Ω , c 2 , 4 = tan ( Ω τ 4 ) Ω , D 4 = c 2 , 4 − c 2 , 4 ∗ ,
Ω 3 2 = − i ( 1 + z 2 ) q ^ 2 ω + , Ω 2 = − i λ q ^ 2 ω + . (63) \Omega_3^2=-\frac{i(1+z^2)\hat{q}}{2\omega^+},\qquad\Omega^2=-\frac{i\lambda\hat{q}}{2\omega^+}. \tag*{(63)} Ω 3 2 = − 2 ω + i ( 1 + z 2 ) q ^ , Ω 2 = − 2 ω + iλ q ^ . (63)
At w = 0 w=0 w = 0 the four endpoints of the kernels in Eq. (60) are
X i = γ u − δ b , X ˉ i = − γ β u + δ b , X_i=\gamma u-\delta b,\qquad\bar{X}_i=-\gamma\beta u+\delta b, X i = γ u − δ b , X ˉ i = − γ β u + δ b ,
X f = ( 1 − β ) γ x + γ β Δ − δ b , X ˉ f = ( 1 − β ) γ x − γ Δ + δ b . (64) X_f=(1-\beta)\gamma x+\gamma\beta\Delta-\delta b,\qquad\bar{X}_f=(1-\beta)\gamma x-\gamma\Delta+\delta b. \tag*{(64)} X f = ( 1 − β ) γ x + γ β Δ − δ b , X ˉ f = ( 1 − β ) γ x − γ Δ + δ b . (64)
The Gaussian integral over the common transverse coordinate, including the derivative ∇ w \nabla_w ∇ w that acts on both kernels, gives
∇ w ∫ d 2 x K Ω ( X f , y + ; X i , x ˉ + ) K Ω ∗ ( X ˉ f , y + ; X ˉ i , x ˉ + ) ∣ w = 0 = − 2 i ( ω + ) 2 c 1 , 4 c 1 , 4 ∗ π D 4 ( 1 − β ) 2 γ exp { E 4 ( u , Δ ) } W ( u , Δ ) , (65) \left.\nabla_w\int d^2x\,\mathcal{K}_{\Omega}(X_f,y^+;X_i,\bar{x}^+)\mathcal{K}_{\Omega}^{*}(\bar{X}_f,y^+;\bar{X}_i,\bar{x}^+)\right|_{w=0}
=-\frac{2i(\omega^+)^2c_{1,4}c_{1,4}^{*}}{\pi D_4(1-\beta)^2\gamma}\exp\{\mathcal{E}_4(u,\Delta)\}\mathcal{W}(u,\Delta), \tag*{(65)} ∇ w ∫ d 2 x K Ω ( X f , y + ; X i , x ˉ + ) K Ω ∗ ( X ˉ f , y + ; X ˉ i , x ˉ + ) w = 0 = − π D 4 ( 1 − β ) 2 γ 2 i ( ω + ) 2 c 1 , 4 c 1 , 4 ∗ exp { E 4 ( u , Δ )} W ( u , Δ ) , (65)
where the exponent E 4 \mathcal{E}_4 E 4 is quadratic and the prefactor W \mathcal{W} W is linear in u u u , Δ \Delta Δ and b b b . The remaining u u u and Δ \Delta Δ integrations form a four-dimensional Gaussian. Combining Eqs. (56) , (60) and (65) , the exponent before the u u u and Δ \Delta Δ integrations is
− 1 2 ( u Δ ) M ( u Δ ) + b ⋅ ( s u s Δ ) ( u Δ ) + C 0 b 2 − i Δ ⋅ p , (66) -\frac{1}{2}\begin{pmatrix}u & \Delta\end{pmatrix}M\begin{pmatrix}u\\\Delta\end{pmatrix}
+b\cdot\begin{pmatrix}s_u & s_\Delta\end{pmatrix}\begin{pmatrix}u\\\Delta\end{pmatrix}
+C_0b^2-i\Delta\cdot p, \tag*{(66)} − 2 1 ( u Δ ) M ( u Δ ) + b ⋅ ( s u s Δ ) ( u Δ ) + C 0 b 2 − i Δ ⋅ p , (66)
where M M M , ( s u , s Δ ) (s_u,s_\Delta) ( s u , s Δ ) and C 0 C_0 C 0 are given in Eq. (A6). We write the inverse matrix as
M − 1 = ( C u u C u Δ C u Δ C Δ Δ ) . (67) M^{-1}=\begin{pmatrix}C_{uu} & C_{u\Delta}\\C_{u\Delta} & C_{\Delta\Delta}\end{pmatrix}. \tag*{(67)} M − 1 = ( C uu C u Δ C u Δ C ΔΔ ) . (67)
The Gaussian means of u u u and Δ \Delta Δ , with the momentum dependence displayed, are
U = ( C u u s u + C u Δ s Δ ) b − i C u Δ p , V = ( C u Δ s u + C Δ Δ s Δ ) b − i C Δ Δ p . (68) \begin{aligned}
U&=(C_{uu}s_u+C_{u\Delta}s_\Delta)b-iC_{u\Delta}p,\\
V&=(C_{u\Delta}s_u+C_{\Delta\Delta}s_\Delta)b-iC_{\Delta\Delta}p. \tag*{(68)}
\end{aligned} U V = ( C uu s u + C u Δ s Δ ) b − i C u Δ p , = ( C u Δ s u + C ΔΔ s Δ ) b − i C ΔΔ p . (68)
The polynomial prefactor comes from the transition vertex and the two kernel derivatives. Its Gaussian average is
P ( p 2 , b ⋅ p ) = ⟨ T ( Δ , b ) [ 2 ω + c 1 , 3 u − z ω + 1 + z 2 ( 2 c 1 , 3 + i c 2 , 3 ) b ] ⋅ W ⟩ , (69) \mathcal{P}(p^2,b\cdot p)=\left\langle\mathcal{T}(\Delta,b)\left[2\omega^+c_{1,3}u-\frac{z\omega^+}{1+z^2}(2c_{1,3}+ic_{2,3})b\right]\cdot\mathcal{W}\right\rangle, \tag*{(69)} P ( p 2 , b ⋅ p ) = ⟨ T ( Δ , b ) [ 2 ω + c 1 , 3 u − 1 + z 2 z ω + ( 2 c 1 , 3 + i c 2 , 3 ) b ] ⋅ W ⟩ , (69)
where the average has the means of Eq. (68) and covariance M − 1 M^{-1} M − 1 . The coefficients of W \mathcal{W} W and the expanded polynomial are given in App. A. Since U U U and V V V are given explicitly in Eq. (68) , P \mathcal{P} P is a polynomial in p 2 p^2 p 2 and b ⋅ p b\cdot p b ⋅ p of degree at most four in the components of p p p . The u u u and Δ \Delta Δ integrations give
∫ d 2 u d 2 Δ e − i Δ ⋅ p T ( Δ , b ) [ 2 ω + c 1 , 3 u − z ω + 1 + z 2 ( 2 c 1 , 3 + i c 2 , 3 ) b ] ⋅ W × exp { − 1 2 ( u Δ ) M ( u Δ ) + b ⋅ ( s u s Δ ) ( u Δ ) + C 0 b 2 } = ( 2 π ) 2 det M exp [ Φ b 2 − α p 2 + i κ 0 b b ⋅ p ] P ( p 2 , b ⋅ p ) , (70) \begin{aligned}
&\int d^2u\,d^2\Delta\,e^{-i\Delta\cdot p}\mathcal{T}(\Delta,b)\left[2\omega^+c_{1,3}u-\frac{z\omega^+}{1+z^2}(2c_{1,3}+ic_{2,3})b\right]\cdot\mathcal{W}\\
&\quad\times\exp\left\{-\frac{1}{2}\begin{pmatrix}u & \Delta\end{pmatrix}M\begin{pmatrix}u\\\Delta\end{pmatrix}
+b\cdot\begin{pmatrix}s_u & s_\Delta\end{pmatrix}\begin{pmatrix}u\\\Delta\end{pmatrix}
+C_0b^2\right\}\\
&=\frac{(2\pi)^2}{\det M}\exp\left[\Phi b^2-\alpha p^2+i\frac{\kappa_0}{b}b\cdot p\right]\mathcal{P}(p^2,b\cdot p), \tag*{(70)}
\end{aligned} ∫ d 2 u d 2 Δ e − i Δ ⋅ p T ( Δ , b ) [ 2 ω + c 1 , 3 u − 1 + z 2 z ω + ( 2 c 1 , 3 + i c 2 , 3 ) b ] ⋅ W × exp { − 2 1 ( u Δ ) M ( u Δ ) + b ⋅ ( s u s Δ ) ( u Δ ) + C 0 b 2 } = det M ( 2 π ) 2 exp [ Φ b 2 − α p 2 + i b κ 0 b ⋅ p ] P ( p 2 , b ⋅ p ) , (70)
where
α = 1 2 C Δ Δ , \alpha= \frac{1}{2}C_{\Delta\Delta}, α = 2 1 C ΔΔ ,
κ 0 = − b [ s u C u Δ + s Δ C Δ Δ ] , \kappa_0 = -b\left[s_u C_{u\Delta}+s_\Delta C_{\Delta\Delta}\right], κ 0 = − b [ s u C u Δ + s Δ C ΔΔ ] ,
κ i = b [ a i − s u C u Δ − s Δ C Δ Δ ] , \kappa_i = b\left[a_i-s_u C_{u\Delta}-s_\Delta C_{\Delta\Delta}\right], κ i = b [ a i − s u C u Δ − s Δ C ΔΔ ] ,
Φ = C 0 + 1 2 ( s u 2 C u u + 2 s u s Δ C u Δ + s Δ 2 C Δ Δ ) . (71) \Phi= C_0+\frac{1}{2}\left(s_u^2 C_{uu}+2s_u s_\Delta C_{u\Delta}+s_\Delta^2 C_{\Delta\Delta}\right). \tag*{(71)} Φ = C 0 + 2 1 ( s u 2 C uu + 2 s u s Δ C u Δ + s Δ 2 C ΔΔ ) . (71)
Equation (70) does not yet include the measurement factor e i a i b ⋅ p e^{i a_i\boldsymbol{b}\cdot\boldsymbol{p}} e i a i b ⋅ p , which replaces κ 0 \kappa_0 κ 0 by κ i = κ 0 + a i b \kappa_i=\kappa_0+a_i b κ i = κ 0 + a i b . The quantities α \alpha α and Φ \Phi Φ do not depend on the channel. Collecting the factor − q ^ / 4 -\hat{q}/4 − q ^ /4 of Eq. (60) , the normalization ω + c 1 , 3 / π \omega^+c_{1,3}/\pi ω + c 1 , 3 / π of the three-point function, and the prefactor of Eq. (65) , the nonfactorizable in–in contribution before the relative-momentum integration is
J i n − i n ( 1 ) , m e d . r a d . n o n f a c ( b ) = 2 C F α s ∑ i = q , g , u ∫ 0 1 d z P q → q g v a c ( z ) h i ( z ) ( 2 ω + ) 2 ∫ R 2 d 2 p ( 2 π ) 2 e − η U V p 2 e i a i b ⋅ p J_{\mathrm{in-in}}^{(1),\mathrm{med.\ rad.\ nonfac}}(b)=2C_F\alpha_s\sum_{i=q,g,u}\int_0^1 dz\,\frac{P_{q\to qg}^{\mathrm{vac}}(z)h_i(z)}{(2\omega^+)^2}\int_{\mathbb{R}^2}\frac{d^2p}{(2\pi)^2}e^{-\eta_{\mathrm{UV}}p^2}e^{i a_i\boldsymbol{b}\cdot\boldsymbol{p}} J in − in ( 1 ) , med. rad. nonfac ( b ) = 2 C F α s i = q , g , u ∑ ∫ 0 1 d z ( 2 ω + ) 2 P q → q g vac ( z ) h i ( z ) ∫ R 2 ( 2 π ) 2 d 2 p e − η UV p 2 e i a i b ⋅ p
× 2 ℜ ∫ 0 L + d τ 3 ∫ 0 L + − τ 3 d τ 4 ∫ 0 L + − τ 3 − τ 4 d x + 2 i q ^ ( ω + ) 3 c 1 , 3 c 1 , 4 c 1 , 4 ∗ ( 2 π ) 2 D 4 ( 1 − β ) 2 γ [ R H S o f E q . ( 71 ) ] . (72) \times2\Re\int_0^{L^+}d\tau_3\int_0^{L^+-\tau_3}d\tau_4\int_0^{L^+-\tau_3-\tau_4}dx^+\frac{2i\hat{q}(\omega^+)^3c_{1,3}c_{1,4}c_{1,4}^*}{(2\pi)^2D_4(1-\beta)^2\gamma}\left[\mathrm{RHS\ of\ Eq.\ }(71)\right]. \tag*{(72)} × 2ℜ ∫ 0 L + d τ 3 ∫ 0 L + − τ 3 d τ 4 ∫ 0 L + − τ 3 − τ 4 d x + ( 2 π ) 2 D 4 ( 1 − β ) 2 γ 2 i q ^ ( ω + ) 3 c 1 , 3 c 1 , 4 c 1 , 4 ∗ [ RHS of Eq. ( 71 ) ] . (72)
In–out interference
The in–out region is 0 < x + < L + < x ˉ + < ∞ 0<x^+<L^+<\bar{x}^+<\infty 0 < x + < L + < x ˉ + < ∞ . The three-point function propagates through the medium to L + L^+ L + and then freely to x ˉ + \bar{x}^+ x ˉ + . With r L r_L r L its relative coordinate at the medium boundary,
S ~ q → q g ( 3 ) , i n − o u t ( r f , r i , b ; x ˉ + , x + ) = exp { − L + − x + 4 ( 1 + z 2 ) q ^ b 2 } ∫ d 2 r L K 0 ( r f − z 1 + z 2 b − r L ; x ˉ + , L + ) K Ω 3 ( r L , L + ; r i − z 1 + z 2 b , x + ) . (73) \widetilde{S}_{q\to qg}^{(3),\mathrm{in-out}}(r_f,r_i,b;\bar{x}^+,x^+)
=\exp\left\{-\frac{L^+-x^+}{4(1+z^2)}\hat{q}b^2\right\}
\int d^2r_L\,K_0\left(r_f-\frac{z}{1+z^2}b-r_L;\bar{x}^+,L^+\right)K_{\Omega_3}\left(r_L,L^+;r_i-\frac{z}{1+z^2}b,x^+\right). \tag*{(73)} S q → q g ( 3 ) , in − out ( r f , r i , b ; x ˉ + , x + ) = exp { − 4 ( 1 + z 2 ) L + − x + q ^ b 2 } ∫ d 2 r L K 0 ( r f − 1 + z 2 z b − r L ; x ˉ + , L + ) K Ω 3 ( r L , L + ; r i − 1 + z 2 z b , x + ) . (73)
Since the conjugate splitting occurs outside the medium, there is no medium evolution after x ˉ + \bar{x}^+ x ˉ + , and the fixed-p p p four-point function is equal to its vacuum counterpart, e − i ( u − w ) ⋅ p e^{-i(u-w)\cdot p} e − i ( u − w ) ⋅ p . After integration by parts in u u u , as in Eq. (55) , the free propagation between L + L^+ L + and x ˉ + \bar{x}^+ x ˉ + is Fourier transformed directly, and the Gaussian integral over r L r_L r L gives
∫ d 2 u e − i u ⋅ p ( ∇ u ⋅ ∇ v ) S ~ q → q g ( 3 ) , i n − o u t ( u , x ˉ + ; v , x + ; b , z ) ∣ v = 0 = [ 2 i c 1 ( τ ) c 2 ( τ ) p 2 + i z q ^ 2 c 2 ( τ ) b ⋅ p ] × 2 i c 1 ( τ ) c 2 ( τ ) exp { − q ^ b 2 4 ( 1 + z 2 ) τ − i p 2 2 ω + ( x ˉ + − L + ) − i p 2 2 ω + c 2 ( τ ) − q ^ z 2 b 2 4 ( 1 + z 2 ) c 2 ( τ ) + i z b ⋅ p 1 + z 2 ( 2 i c 1 ( τ ) c 2 ( τ ) − 1 ) } . (74) \begin{aligned}
\left.\int d^2u\,e^{-iu\cdot p}\left(\nabla_u\cdot\nabla_v\right)\widetilde{S}_{q\to qg}^{(3),\mathrm{in-out}}(u,\bar{x}^+;v,x^+;b,z)\right|_{v=0}
=\left[\frac{2ic_1(\tau)}{c_2(\tau)}p^2+\frac{iz\hat{q}}{2c_2(\tau)}b\cdot p\right] \\
&\times\frac{2ic_1(\tau)}{c_2(\tau)}\exp\left\{-\frac{\hat{q}b^2}{4(1+z^2)}\tau-\frac{ip^2}{2\omega^+}(\bar{x}^+-L^+)-\frac{ip^2}{2\omega^+c_2(\tau)}
-\frac{\hat{q}z^2b^2}{4(1+z^2)c_2(\tau)}+\frac{iz\,b\cdot p}{1+z^2}\left(\frac{2ic_1(\tau)}{c_2(\tau)}-1\right)\right\}. \tag*{(74)}
\end{aligned} ∫ d 2 u e − i u ⋅ p ( ∇ u ⋅ ∇ v ) S q → q g ( 3 ) , in − out ( u , x ˉ + ; v , x + ; b , z ) v = 0 = [ c 2 ( τ ) 2 i c 1 ( τ ) p 2 + 2 c 2 ( τ ) i z q ^ b ⋅ p ] × c 2 ( τ ) 2 i c 1 ( τ ) exp { − 4 ( 1 + z 2 ) q ^ b 2 τ − 2 ω + i p 2 ( x ˉ + − L + ) − 2 ω + c 2 ( τ ) i p 2 − 4 ( 1 + z 2 ) c 2 ( τ ) q ^ z 2 b 2 + 1 + z 2 i z b ⋅ p ( c 2 ( τ ) 2 i c 1 ( τ ) − 1 ) } . (74)
with τ = L + − x + \tau= L^{+} - x^{+} τ = L + − x + . The integrand depends on x ˉ + \bar{x}^{+} x ˉ + only through the phase e − i p 2 ( x ˉ + − L + ) / ( 2 ω + ) e^{-ip^{2}(\bar{x}^{+}-L^{+})/(2\omega^{+})} e − i p 2 ( x ˉ + − L + ) / ( 2 ω + ) of the free propagation after the medium. With the factor e − ϵ x ˉ + e^{-\epsilon\bar{x}^{+}} e − ϵ x ˉ + of Eq. (48) , its integral over L + < x ˉ + < ∞ L^{+} < \bar{x}^{+} < \infty L + < x ˉ + < ∞ is, for p ≠ 0 p \ne0 p = 0 ,
lim ϵ → 0 + ∫ L + ∞ d x ˉ + e − ϵ ( x ˉ + − L + ) e − i p 2 2 ω + ( x ˉ + − L + ) = − 2 i ω + p 2 . (75) \lim_{\epsilon\to0^{+}} \int_{L^{+}}^{\infty} d\bar{x}^{+}\,e^{-\epsilon(\bar{x}^{+}-L^{+})}e^{-\frac{ip^{2}}{2\omega^{+}}(\bar{x}^{+}-L^{+})} = -\frac{2i\omega^{+}}{p^{2}}. \tag*{(75)} ϵ → 0 + lim ∫ L + ∞ d x ˉ + e − ϵ ( x ˉ + − L + ) e − 2 ω + i p 2 ( x ˉ + − L + ) = − p 2 2 i ω + . (75)
Including the parent broadening D F [ b ; x + , 0 ] = e − q ^ b 2 ( L + − τ ) / 4 D_{\mathrm{F}}[b;x^{+},0] = e^{-\hat{q}b^{2}(L^{+}-\tau)/4} D F [ b ; x + , 0 ] = e − q ^ b 2 ( L + − τ ) /4 of Eq. (48) , the fixed-p p p in–out contribution is
J i n − o u t ( 1 ) , m e d . r a d . ( b ) = 2 C F α s ∑ i = q , q ˉ , g ∫ 0 1 d z P q → q g v a c ( z ) h i ( z ) ( 2 ω + ) 2 ∫ R 2 d 2 p ( 2 π ) 2 e − η U V p 2 e i a i ⋅ b ⋅ p 2 ℜ ∫ 0 L + d τ J_{\mathrm{in-out}}^{(1),\mathrm{med.\ rad.}}(b) = 2C_{\mathrm{F}}\alpha_{\mathrm{s}}\sum_{i=q,\bar{q},g}\int_{0}^{1} dz\,\frac{P_{q\to qg}^{\mathrm{vac}}(z)h_i(z)}{(2\omega^{+})^{2}}\int_{\mathbb{R}^{2}}\frac{d^{2}p}{(2\pi)^{2}}e^{-\eta_{\mathrm{UV}}p^{2}}e^{i a_i\cdot b\cdot p^{2}}\Re\int_{0}^{L^{+}}d\tau J in − out ( 1 ) , med. rad. ( b ) = 2 C F α s i = q , q ˉ , g ∑ ∫ 0 1 d z ( 2 ω + ) 2 P q → q g vac ( z ) h i ( z ) ∫ R 2 ( 2 π ) 2 d 2 p e − η UV p 2 e i a i ⋅ b ⋅ p 2 ℜ ∫ 0 L + d τ
× ( − 2 i ω + p 2 ) 2 i c 1 ( τ ) c 2 ( τ ) exp { − q ^ b 2 4 ( L + − τ ) − q ^ b 2 4 ( 1 + z 2 ) τ − i p 2 2 ω + c 2 ( τ ) − q ^ z 2 b 2 4 ( 1 + z 2 ) c 2 ( τ ) + i z b ⋅ p 1 + z 2 ( 2 i c 1 ( τ ) c 2 ( τ ) − 1 ) } [ 2 i c 1 ( τ ) c 2 ( τ ) p 2 + i z q ^ 2 c 2 ( τ ) b ⋅ p ] . (76) {}\times\left(-\frac{2i\omega^{+}}{p^{2}}\right)\frac{2ic_{1}(\tau)}{c_{2}(\tau)}\exp\left\{-\frac{\hat{q}b^{2}}{4}(L^{+}-\tau)-\frac{\hat{q}b^{2}}{4(1+z^{2})}\tau-\frac{ip^{2}}{2\omega^{+}c_{2}(\tau)}-\frac{\hat{q}z^{2}b^{2}}{4(1+z^{2})c_{2}(\tau)}+\frac{iz\,b\cdot p}{1+z^{2}}\left(\frac{2ic_{1}(\tau)}{c_{2}(\tau)}-1\right)\right\}
\left[\frac{2ic_{1}(\tau)}{c_{2}(\tau)}p^{2}+\frac{iz\hat{q}}{2c_{2}(\tau)}b\cdot p\right]. \tag*{(76)} × ( − p 2 2 i ω + ) c 2 ( τ ) 2 i c 1 ( τ ) exp { − 4 q ^ b 2 ( L + − τ ) − 4 ( 1 + z 2 ) q ^ b 2 τ − 2 ω + c 2 ( τ ) i p 2 − 4 ( 1 + z 2 ) c 2 ( τ ) q ^ z 2 b 2 + 1 + z 2 i z b ⋅ p ( c 2 ( τ ) 2 i c 1 ( τ ) − 1 ) } [ c 2 ( τ ) 2 i c 1 ( τ ) p 2 + 2 c 2 ( τ ) i z q ^ b ⋅ p ] . (76)
Appendix B performs the common p p p integration for this expression and for the two in–in contributions.
Medium response
We consider the contribution of the medium response to the TEEC, following recent studies of energy flow and EECs in many body states, see e.g. [33 , 34 , 76 , 77 , 78 , 79 , 80 , 81 ] . Throughout this section, expectation values are taken at fixed dijet geometry, the average over the dijet geometry being performed afterwards. We decompose the energy flow operator of Eq. (1) into jet and medium response (wake [ W ] [W] [ W ] ) contributions, E = E J + E W \mathcal{E} = \mathcal{E}_{J} + \mathcal{E}_{W} E = E J + E W . We treat the response through its classical mean flow (see, e.g., [82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 ] for phenomenological studies) and neglect connected correlations involving it, so that the correlator in Eq. (2) becomes
⟨ E ( n ⃗ 1 ) E ( n ⃗ 2 ) ⟩ ≃ ⟨ E J ( n ⃗ 1 ) E J ( n ⃗ 2 ) ⟩ + ⟨ E J ( n ⃗ 1 ) ⟩ ⟨ E W ( n ⃗ 2 ) ⟩ + ⟨ E W ( n ⃗ 1 ) ⟩ ⟨ E J ( n ⃗ 2 ) ⟩ + ⟨ E W ( n ⃗ 1 ) ⟩ ⟨ E W ( n ⃗ 2 ) ⟩ , (77) \langle\mathcal{E}(\vec{n}_{1})\mathcal{E}(\vec{n}_{2})\rangle\simeq\langle\mathcal{E}_{J}(\vec{n}_{1})\mathcal{E}_{J}(\vec{n}_{2})\rangle+ \langle\mathcal{E}_{J}(\vec{n}_{1})\rangle\langle\mathcal{E}_{W}(\vec{n}_{2})\rangle
+ \langle\mathcal{E}_{W}(\vec{n}_{1})\rangle\langle\mathcal{E}_{J}(\vec{n}_{2})\rangle+ \langle\mathcal{E}_{W}(\vec{n}_{1})\rangle\langle\mathcal{E}_{W}(\vec{n}_{2})\rangle, \tag*{(77)} ⟨ E ( n 1 ) E ( n 2 )⟩ ≃ ⟨ E J ( n 1 ) E J ( n 2 )⟩ + ⟨ E J ( n 1 )⟩ ⟨ E W ( n 2 )⟩ + ⟨ E W ( n 1 )⟩ ⟨ E J ( n 2 )⟩ + ⟨ E W ( n 1 )⟩ ⟨ E W ( n 2 )⟩ , (77)
where the first term is the perturbative correlator computed in Secs. II–IV and the remaining ones are products of mean flows. We use the distribution in ϕ \phi ϕ , σ − 1 d Σ / d ϕ \sigma^{-1}d\Sigma/d\phi σ − 1 d Σ/ d ϕ , obtained from Eq. (2) by replacing δ ( τ − ( 1 + cos ϕ 12 ) / 2 ) \delta(\tau-(1+\cos\phi_{12})/2) δ ( τ − ( 1 + cos ϕ 12 ) /2 ) with δ ( ϕ − ϕ 12 ) \delta(\phi-\phi_{12}) δ ( ϕ − ϕ 12 ) . For the response estimate we take the two jet axes in the transverse plane, so that E T = 2 p T E_{T}=2p_{T} E T = 2 p T at leading power. Since the measurement depends on the azimuthal separation, we first integrate over the polar angles,
e A ( φ ) = ∫ 0 π d θ sin 2 θ ⟨ E A ( θ , φ ) ⟩ , A = J , W , (78) e_{A}(\varphi) = \int_{0}^{\pi} d\theta\,\sin^{2}\theta\,\langle\mathcal{E}_{A}(\theta,\varphi)\rangle,\qquad A=J,W, \tag*{(78)} e A ( φ ) = ∫ 0 π d θ sin 2 θ ⟨ E A ( θ , φ )⟩ , A = J , W , (78)
the transverse energy per unit beam azimuth. The two separations selected by the measurement, φ 2 = φ 1 ± ϕ \varphi_2 = \varphi_1 \pm\phi φ 2 = φ 1 ± ϕ , give
1 σ d Σ d ϕ = 1 σ ( d Σ p e r t d ϕ + d Σ J W d ϕ + d Σ W W d ϕ ) , (79) \frac{1}{\sigma}\frac{d\Sigma}{d\phi} = \frac{1}{\sigma}\left(\frac{d\Sigma_{\mathrm{pert}}}{d\phi} + \frac{d\Sigma_{JW}}{d\phi} + \frac{d\Sigma_{WW}}{d\phi}\right), \tag*{(79)} σ 1 d ϕ d Σ = σ 1 ( d ϕ d Σ pert + d ϕ d Σ J W + d ϕ d Σ W W ) , (79)
with d Σ p e r t / d ϕ d\Sigma_{\mathrm{pert}}/d\phi d Σ pert / d ϕ the first contribution in Eq. (77) and
1 σ d Σ J W d ϕ = 2 E T 2 ∫ 0 2 π d φ [ e J ( φ ) e W ( φ + ϕ ) + e W ( φ ) e J ( φ + ϕ ) ] , 1 σ d Σ W W d ϕ = 2 E T 2 ∫ 0 2 π d φ e W ( φ ) e W ( φ + ϕ ) . (80) \begin{aligned}
\frac{1}{\sigma}\frac{d\Sigma_{JW}}{d\phi} &= \frac{2}{E_T^2}\int_0^{2\pi}d\varphi\left[e_J(\varphi)e_W(\varphi+\phi)+e_W(\varphi)e_J(\varphi+\phi)\right],\\
\frac{1}{\sigma}\frac{d\Sigma_{WW}}{d\phi} &= \frac{2}{E_T^2}\int_0^{2\pi}d\varphi\,e_W(\varphi)e_W(\varphi+\phi). \tag*{(80)}
\end{aligned} σ 1 d ϕ d Σ J W σ 1 d ϕ d Σ W W = E T 2 2 ∫ 0 2 π d φ [ e J ( φ ) e W ( φ + ϕ ) + e W ( φ ) e J ( φ + ϕ ) ] , = E T 2 2 ∫ 0 2 π d φ e W ( φ ) e W ( φ + ϕ ) . (80)
For identical jets and wakes with profiles symmetric about each jet axis, the total mean flows satisfy e A ( φ + π ) = e A ( φ ) e_A(\varphi+\pi)=e_A(\varphi) e A ( φ + π ) = e A ( φ ) . We assign the transverse energy Δ p T = ε p T \Delta p_T=\varepsilon p_T Δ p T = ε p T to each wake and ( 1 − ε ) p T (1-\varepsilon)p_T ( 1 − ε ) p T to each surviving jet, so that with E T = 2 p T E_T=2p_T E T = 2 p T
∫ 0 2 π d φ e J ( φ ) = ( 1 − ε ) E T , ∫ 0 2 π d φ e W ( φ ) = ε E T . (81) \int_0^{2\pi}d\varphi\,e_J(\varphi)=(1-\varepsilon)E_T,\qquad\int_0^{2\pi}d\varphi\,e_W(\varphi)=\varepsilon E_T. \tag*{(81)} ∫ 0 2 π d φ e J ( φ ) = ( 1 − ε ) E T , ∫ 0 2 π d φ e W ( φ ) = ε E T . (81)
Let σ J \sigma_J σ J and σ W \sigma_W σ W be the angular scales on which the jet and wake profiles vary. For σ J ≪ σ W \sigma_J\ll\sigma_W σ J ≪ σ W the mixed contribution can be expanded across each jet,
∫ 0 2 π d φ e J ( φ ) e W ( φ + ϕ ) = ( 1 − ε ) E T [ e W ( ϕ ) + O ( σ J 2 e W ′ ′ ( ϕ ) ) ] , (82) \int_0^{2\pi}d\varphi\,e_J(\varphi)e_W(\varphi+\phi)=(1-\varepsilon)E_T\left[e_W(\phi)+\mathcal{O}\left(\sigma_J^2 e_W^{\prime\prime}(\phi)\right)\right], \tag*{(82)} ∫ 0 2 π d φ e J ( φ ) e W ( φ + ϕ ) = ( 1 − ε ) E T [ e W ( ϕ ) + O ( σ J 2 e W ′′ ( ϕ ) ) ] , (82)
where reflection symmetry removes the linear term, so that
1 σ d Σ J W d ϕ = 4 ( 1 − ε ) E T e W ( ϕ ) [ 1 + O ( σ J 2 σ W 2 ) ] , (83) \frac{1}{\sigma}\frac{d\Sigma_{JW}}{d\phi}=\frac{4(1-\varepsilon)}{E_T}e_W(\phi)\left[1+\mathcal{O}\left(\frac{\sigma_J^2}{\sigma_W^2}\right)\right], \tag*{(83)} σ 1 d ϕ d Σ J W = E T 4 ( 1 − ε ) e W ( ϕ ) [ 1 + O ( σ W 2 σ J 2 ) ] , (83)
i.e., a broad wake probes the integrated jet energy, and the sensitivity to the angular distribution of the jet is suppressed by the squared ratio of widths. In this limit, the same property of the profile controls its variation across the back-to-back window. With ϕ = π − δ \phi=\pi-\delta ϕ = π − δ , both terms are even about π \pi π and
d Σ X d ϕ ∣ π − δ = d Σ X d ϕ ∣ π [ 1 + O ( δ 2 σ W 2 ) ] , X = J W , W W , (84) \left.\frac{d\Sigma_X}{d\phi}\right|_{\pi-\delta}=\left.\frac{d\Sigma_X}{d\phi}\right|_{\pi}\left[1+\mathcal{O}\left(\frac{\delta^2}{\sigma_W^2}\right)\right],\qquad X=JW,WW, \tag*{(84)} d ϕ d Σ X π − δ = d ϕ d Σ X π [ 1 + O ( σ W 2 δ 2 ) ] , X = J W , W W , (84)
so that for δ ≪ σ W \delta\ll\sigma_W δ ≪ σ W the response is approximately constant across the window.6 Therefore, its relative importance can grow toward the edge of the window, where the perturbative distribution is smaller. At small ε \varepsilon ε this contribution is linear in Δ p T / p T \Delta p_T/p_T Δ p T / p T , with its height and residual angular dependence set by the wake profile. Since the perturbative medium correction retains the soft partons that may thermalize, their overlap with the response is not subtracted, and Δ p T \Delta p_T Δ p T is an effective response strength whose relation to the deposited energy is model dependent. A more complete treatment would exclude thermalized partons from the perturbative final state and assign their energy to the wake. This requires a consistent separation between the perturbative radiation and the response, see e.g. [33 ] , and lies beyond the scope of this work.
For the numerical results of Sec. VI we specialize to a Gaussian wake [34 ] about each jet direction ± n ^ \pm\hat n ± n ^ , with n ^ \hat n n ^ the Born quark direction of Sec. II,
⟨ E W ( n ⃗ 1 ) ⟩ ∝ exp [ − arccos 2 ( n ⃗ 1 ⋅ n ^ ) θ W 2 ] + exp [ − arccos 2 ( − n ⃗ 1 ⋅ n ^ ) θ W 2 ] , (85) \langle\mathcal{E}_W(\vec n_1)\rangle\propto\exp\left[-\frac{\arccos^2(\vec n_1\cdot\hat n)}{\theta_W^2}\right]+\exp\left[-\frac{\arccos^2(-\vec n_1\cdot\hat n)}{\theta_W^2}\right], \tag*{(85)} ⟨ E W ( n 1 )⟩ ∝ exp [ − θ W 2 arccos 2 ( n 1 ⋅ n ^ ) ] + exp [ − θ W 2 arccos 2 ( − n 1 ⋅ n ^ ) ] , (85)
of width θ W \theta_W θ W , perform the polar projection, and fix the normalization through Eq. (81) . In the limit θ W → ∞ \theta_W\to\infty θ W → ∞ , where the expansions above apply, the projected profile is uniform and the two terms are the constants 2 ε ( 1 − ε ) / π 2\varepsilon(1-\varepsilon)/\pi 2 ε ( 1 − ε ) / π and ε 2 / π \varepsilon^2/\pi ε 2 / π . In the opposite limit, θ W ≪ 1 \theta_W\ll1 θ W ≪ 1 , retaining σ J ≪ θ W \sigma_J\ll\theta_W σ J ≪ θ W , the polar projection is itself Gaussian in beam azimuth about each jet, and the two terms take the explicit form
1 σ d Σ J W d ϕ ≃ 2 ε ( 1 − ε ) π θ W [ e − ( π − ϕ ) 2 / θ W 2 + e − ϕ 2 / θ W 2 ] , 1 σ d Σ W W d ϕ ≃ ε 2 2 π θ W [ e − ( π − ϕ ) 2 / 2 θ W 2 + e − ϕ 2 / 2 θ W 2 ] . (86) \begin{aligned}
\frac{1}{\sigma}\frac{d\Sigma_{JW}}{d\phi}&\simeq\frac{2\varepsilon(1-\varepsilon)}{\sqrt{\pi}\theta_W}\left[e^{-(\pi-\phi)^2/\theta_W^2}+e^{-\phi^2/\theta_W^2}\right],\\
\frac{1}{\sigma}\frac{d\Sigma_{WW}}{d\phi}&\simeq\frac{\varepsilon^2}{\sqrt{2\pi}\theta_W}\left[e^{-(\pi-\phi)^2/2\theta_W^2}+e^{-\phi^2/2\theta_W^2}\right]. \tag*{(86)}
\end{aligned} σ 1 d ϕ d Σ J W σ 1 d ϕ d Σ W W ≃ π θ W 2 ε ( 1 − ε ) [ e − ( π − ϕ ) 2 / θ W 2 + e − ϕ 2 / θ W 2 ] , ≃ 2 π θ W ε 2 [ e − ( π − ϕ ) 2 /2 θ W 2 + e − ϕ 2 /2 θ W 2 ] . (86)
i.e., Gaussians about ϕ = π \phi=\pi ϕ = π and ϕ = 0 \phi=0 ϕ = 0 of widths θ W \theta_W θ W and 2 θ W \sqrt{2}\theta_W 2 θ W , whose expansion about π \pi π reproduces Eq. (84) with corrections − δ 2 / θ W 2 -\delta^2/\theta_W^2 − δ 2 / θ W 2 and − δ 2 / 2 θ W 2 -\delta^2/2\theta_W^2 − δ 2 /2 θ W 2 .
Numerical results
Figures 2 and 3 show the back-to-back TEEC for transverse jets with parent energy E 0 = p T E_0=p_T E 0 = p T and Q = 2 p T Q=2p_T Q = 2 p T , with p T = 35 p_T=35 p T = 35 , 50 and 100 GeV, in a medium of physical length L = 5 L=5 L = 5 fm and q ^ L = 3 \hat qL=3 q ^ L = 3 , 5 and 8 GeV2 ^2 2 .7 We compare the vacuum distribution at LL and NLL′ ' ′ , computed with the numerical prescription of Sec. II, with the medium-modified distribution at NLL′ ' ′ , first with broadening alone and then with the radiative correction added as in Eq. (45) . The bands are the scale variations of Sec. II, with the same variation in the numerator and denominator of the ratios. The propagated numerical uncertainty of the medium-induced radiative correction, below 1% of the vacuum distribution throughout the plotted range, is smaller than these bands and is not shown.
For the parameters studied, broadening lowers the distribution at ϕ = π \phi=\pi ϕ = π by 5–17%, while the radiative correction is several times larger there, suppressing the peak by 20–70% of the vacuum value, and dominates the enhancement toward the edge of the plotted window. Both modifications grow with q ^ L \hat qL q ^ L and with decreasing p T p_T p T , the radiative correction more strongly with p T p_T p T and broadening more strongly with q ^ L \hat qL q ^ L , as expected from the medium scales that control each of them. From p T = 100 p_T = 100 p T = 100 to 35 35 35 GeV the suppression of the peak grows by a factor of about 2 for the radiative correction and 1.7–1.8 for broadening, while from q ^ L = 3 \hat q L = 3 q ^ L = 3 to 8 G e V 2 8\ \mathrm{GeV}^2 8 GeV 2 it grows by 1.5–1.7 and 2.0–2.2, respectively. Broadening suppresses the b b b integrand beyond b ∼ ( q ^ L ) − 1 / 2 b \sim(\hat q L)^{-1/2} b ∼ ( q ^ L ) − 1/2 , which smears the peak by a transverse momentum of order q ^ L \sqrt{\hat q L} q ^ L , about 2 GeV for q ^ L = 5 G e V 2 \hat q L = 5\ \mathrm{GeV}^2 q ^ L = 5 GeV 2 . The radiative correction redistributes energy flow toward larger decorrelations. Since J ( 1 ) , m e d . r a d . ( b ) J^{(1),\mathrm{med.\ rad.}}(b) J ( 1 ) , med. rad. ( b ) is negative over the relevant range of b b b , it lowers the distribution at ϕ = π \phi= \pi ϕ = π , and its contribution changes sign at π − ϕ ≈ 0.13 \pi- \phi\approx0.13 π − ϕ ≈ 0.13 –0.27 0.27 0.27 , beyond which it enhances the distribution. The soft BDMPS–Z spectrum [69 ] scales as d σ ^ q g ∣ p 0 m e d / d z ∼ α s / ( 1 − z ) q ^ L 2 / ( z ( 1 − z ) p T ) d\hat{\sigma}^{\mathrm{med}}_{qg|p_0}/dz \sim\alpha_s/(1-z)\sqrt{\hat q L^2/(z(1-z)p_T)} d σ ^ q g ∣ p 0 med / d z ∼ α s / ( 1 − z ) q ^ L 2 / ( z ( 1 − z ) p T ) for ( 1 − z ) p T ≪ q ^ L 2 (1-z)p_T \ll\hat q L^2 ( 1 − z ) p T ≪ q ^ L 2 , as follows from Eq. (112) in App. C. This suggests a larger redistribution as q ^ L \hat q L q ^ L increases or p T p_T p T decreases at fixed L L L , and, since a gluon that broadens over a distance of order L L L gives π − ϕ ∼ q ^ L / [ ( 1 − z ) p T ] \pi- \phi\sim\sqrt{\hat q L}/[(1-z)p_T] π − ϕ ∼ q ^ L / [( 1 − z ) p T ] , that softer emissions can populate larger decorrelations. The radiative correction is proportional to the coupling evaluated at μ ˉ b \bar{\mu}_b μ ˉ b , whereas broadening at fixed q ^ L \hat q L q ^ L contains no explicit factor of α s \alpha_s α s . A different coupling prescription, such as a fixed value or a medium scale (Sec. III C), changes the size of the correction roughly by the ratio of the couplings in the region of b b b that dominates the peak, where α s ( μ ˉ b ) \alpha_s(\bar{\mu}_b) α s ( μ ˉ b ) lies between 0.2 and 0.3, and, since the coupling enters at each b b b , it also affects the shape in ϕ \phi ϕ . Near ϕ = π \phi= \pi ϕ = π the scale bands are driven mainly by μ → μ / 2 \mu\to\mu/2 μ → μ /2 , which raises the coupling at large b b b . At lower p T p_T p T and larger q ^ L \hat q L q ^ L , where the correction becomes comparable to the vacuum distribution and some scale variations give negative values near the peak, sub-eikonal corrections to the medium-induced radiative correction can become important.
Figures 4 and 5 add the medium response of § 0.5 , Eq. (79) , to the distribution with broadening and medium-induced radiation. In the harmonic-oscillator spectrum, the energy carried by soft gluons below a fixed thermalization scale scales as q ^ L 2 \sqrt{\hat q L^2} q ^ L 2 , independently of p T p_T p T . Motivated by this behavior, we adopt the response prescription Δ p T = Δ p T ( 5 ) q ^ L / 5 G e V 2 \Delta p_T = \Delta p_T^{(5)}\sqrt{\hat q L/5\ \mathrm{GeV}^2} Δ p T = Δ p T ( 5 ) q ^ L /5 GeV 2 , with Δ p T ( 5 ) = 3 \Delta p_T^{(5)} = 3 Δ p T ( 5 ) = 3 GeV and θ W = 1 \theta_W = 1 θ W = 1 for the central curve, and vary Δ p T ( 5 ) \Delta p_T^{(5)} Δ p T ( 5 ) between 1 GeV and 10 GeV, and θ W \theta_W θ W between 0.3, a representative jet angular scale, and ∞ \infty ∞ , with the full polar projection of Eq. (85) . The 1 GeV benchmark is guided by the wake energy inside the jet cone estimated in [34 ] , while 10 GeV is an illustrative upper benchmark. The central value of ϵ = Δ p T / p T \epsilon= \Delta p_T/p_T ϵ = Δ p T / p T ranges from 0.023 to 0.11, and over the Δ p T ( 5 ) \Delta p_T^{(5)} Δ p T ( 5 ) band from 0.008 to 0.36. At small ϵ \epsilon ϵ , the response scales as 1 / p T 1/p_T 1/ p T and as q ^ L \sqrt{\hat q L} q ^ L at fixed L L L under this prescription. The response bands, which vary Δ p T ( 5 ) \Delta p_T^{(5)} Δ p T ( 5 ) or θ W \theta_W θ W separately at central perturbative scales, describe the parameter dependence of the response model and are not combined with the scale variation bands. The energy variation changes the response by roughly a factor of one third to three relative to the central choice, and for the largest energy fractions the quadratic terms in ϵ \epsilon ϵ also matter. A broad wake gives a slowly varying addition across the window, so its relative importance grows as the perturbative distribution falls. Within the chosen parameter ranges, the wake energy has the larger effect away from the peak, while narrower wakes concentrate more energy near ϕ = π \phi= \pi ϕ = π . At p T = 100 p_T = 100 p T = 100 GeV and q ^ L = 5 G e V 2 \hat q L = 5\ \mathrm{GeV}^2 q ^ L = 5 GeV 2 , the central response raises the ratio to vacuum at ϕ = 2.64 \phi= 2.64 ϕ = 2.64 from about 1.4 to 1.6. The response becomes more significant at lower p T p_T p T and larger q ^ L \hat{q}L q ^ L , although its quantitative interpretation is limited by the unsubtracted overlap with perturbative radiation (§ 0.5 ). A broad response may become dominant at larger decorrelations beyond the back-to-back region.
Conclusions
Back-to-back topologies in heavy ion collisions offer a setting, complementary to collinear configurations, in which medium modifications of the final state energy flow can be computed and tested through angular decorrelations. In this work we studied the TEEC in the back-to-back limit in a simplified scenario of an e + e − e^+e^- e + e − collision occurring in the presence of a homogeneous background medium, in which a q q ˉ q\bar{q} q q ˉ pair produced in a colour-singlet hard process propagates through the medium. This setup discards the nuclear initial state effects of realistic heavy ion collisions, thus isolating final state modifications. We assumed that the vacuum factorized structure holds in the medium, with medium effects entering only through the jet functions of the back-to-back TMD factorization formula. We combined the vacuum distribution at NLL′ ^\prime ′ with momentum broadening and the medium-induced radiative correction at first order in the strong coupling, computed for partons propagating through a dense medium, in the large N c N_c N c harmonic-oscillator approximation. We further estimated the contribution of the medium response through its mean flow contribution to the energy flow operator.
Figure 3. Ratio of the medium-modified distributions of Figure 2 to the vacuum distribution at NLL′ ^\prime ′ , for the same values of p T p_T p T and q ^ L \hat{q}L q ^ L . The scale bands use the same variations in the numerator and denominator.
Figure 2. Back-to-back TEEC distribution σ − 1 d Σ / d ϕ \sigma^{-1}d\Sigma/d\phi σ − 1 d Σ/ d ϕ at fixed p T p_T p T , for Q = 2 p T Q = 2p_T Q = 2 p T and L = 5 L = 5 L = 5 fm. Rows: p T = 100 p_T = 100 p T = 100 , 50 50 50 and 35 35 35 GeV. Columns: q ^ L = 3 \hat{q}L = 3 q ^ L = 3 , 5 5 5 and 8 8 8 GeV2 ^2 2 . Shown are the vacuum distribution at LL and NLL′ ^\prime ′ , and the medium-modified distribution at NLL′ ^\prime ′ with broadening only and with broadening and medium-induced radiation. The bands denote the scale variation discussed in the main text.
For jets with p T = 35 p_T = 35 p T = 35 –100 100 100 GeV in a medium with q ^ L = 3 \hat{q}L = 3 q ^ L = 3 –8 8 8 GeV2 ^2 2 , broadening lowers the back-to-back peak by 5–17%. The radiative correction suppresses the peak by 20–70% of the vacuum value and enhances the distribution at larger angular decorrelations. Both modifications grow with q ^ L \hat{q}L q ^ L and with decreasing p T p_T p T . The radiative correction depends more strongly on p T p_T p T , as expected from the scaling of the soft BDMPS–Z spectrum with q ^ L 2 / p T \sqrt{\hat{q}L^2/p_T} q ^ L 2 / p T , whereas the broadening factor is mostly sensitive to q ^ L \hat{q}L q ^ L . Therefore, a quantitative description of the peak requires the medium-induced radiative correction in addition to broadening. Its size is proportional to the coupling evaluated at the scale μ ˉ b \bar{\mu}_b μ ˉ b , so that it depends directly on the coupling prescription. For a wake of transverse energy Δ p T \Delta p_T Δ p T and angular width θ W \theta_W θ W about each jet, the medium response adds a nearly constant contribution when the wake is broader than the back-to-back window, linear in Δ p T / p T \Delta p_T/p_T Δ p T / p T when this ratio is small. In the numerical results the response populates larger angular decorrelations, where its relative importance grows as the perturbative distribution falls, and its size is set mainly by the wake energy, while narrower wakes concentrate more of it near ϕ = π \phi= \pi ϕ = π .
Together, these results provide a starting point for the study of back-to-back configurations in heavy ion collisions with energy correlators, and show that momentum broadening and the medium response leave distinct imprints on the azimuthal decorrelation. Several extensions are left for future work. A direct application to phenomenology requires initial state effects and a realistic expanding medium in which the two jets traverse different path lengths. Beyond momentum broadening, the assignment of medium modifications to the jet and soft functions requires a consistent treatment of the medium-dependent soft and collinear sectors, and relaxing the independent propagation of the quark and antiquark requires accounting for medium-dependent correlations between them. At low p T p_T p T and large q ^ L \hat{q}L q ^ L , where the radiative correction becomes comparable to the vacuum distribution, sub-eikonal corrections may become relevant. Finally, a more sophisticated modeling of the background would allow for a better separation between the perturbative effects and the medium response.
Figure 5. Ratio of the distributions of Figure 4 to the vacuum distribution at NLL′ ^\prime ′ , for the same values of p T p_T p T and q ^ L \hat{q}L q ^ L . Scale bands use the same variations in the numerator and denominator. Response bands vary the model parameters at central perturbative scales.
Figure 4. Back-to-back TEEC distribution as in Figure 2 , with the medium response of Eq. (79) added to the NLL′ ^{\prime} ′ distribution with broadening and medium-induced radiation, at fixed L = 5 L = 5 L = 5 fm, for Δ p T = Δ p T ( 5 ) q ^ L / 5 G e V 2 \Delta p_T = \Delta p_T^{(5)}\sqrt{\hat{q}_L/5\ \mathrm{GeV}^2} Δ p T = Δ p T ( 5 ) q ^ L /5 GeV 2 with Δ p T ( 5 ) = 3 \Delta p_T^{(5)} = 3 Δ p T ( 5 ) = 3 GeV and θ W = 1 \theta_W = 1 θ W = 1 . Bands: scale variation for the vacuum and medium curves, Δ p T ( 5 ) = 1 \Delta p_T^{(5)} = 1 Δ p T ( 5 ) = 1 –10 10 10 GeV at θ W = 1 \theta_W = 1 θ W = 1 , and θ W = 0.3 \theta_W = 0.3 θ W = 0.3 –∞ \infty ∞ at Δ p T ( 5 ) = 3 \Delta p_T^{(5)} = 3 Δ p T ( 5 ) = 3 GeV. Response bands vary these parameters at central perturbative scales.
Acknowledgments
JB was supported by the U.S. Department of Energy under Contract No. DE-SC0012704. HTL is supported by the National Natural Science Foundation of China under Grants No. 12275156 and No. 12321005, and the Shandong Provincial Department of Science and Technology under Project No. TQ012025001. The work of AVS is supported by the Spanish Ministry of Science, Innovation and Universities through grant PID2025-174203NB-I00 (MICIU/AEI/10.13039/501100011033) and by the Basque Government through grant IT1977-26. AVS also acknowledges support from Ikerbasque, Basque Foundation for Science. JMS is supported by the National Science Foundation under grant No. PHY-2515057.
Coefficients of the non-factorizable contribution
This appendix collects the coefficients entering the Gaussian integrals of Sec. IV B. We use a 2 × 2 2 \times2 2 × 2 notation on the pair ( u , Δ ) (\mathbf{u},\boldsymbol{\Delta}) ( u , Δ ) , in which every entry multiplies the identity in transverse space, and introduce the coefficient space vectors
ρ f = ( 0 , γ β ) , ρ i = ( γ , 0 ) , ρ ˉ f = ( 0 , − γ ) , ρ ˉ i = ( − γ β , 0 ) , σ f = σ i = − δ , σ ˉ f = σ ˉ i = δ . (87) \begin{aligned}
\rho_f &= (0,\gamma\beta), \qquad\rho_i = (\gamma,0), \qquad\bar{\rho}_f = (0,-\gamma), \qquad\bar{\rho}_i = (-\gamma\beta,0),\\
\sigma_f &= \sigma_i = -\delta, \qquad\bar{\sigma}_f = \bar{\sigma}_i = \delta. \tag*{(87)}
\end{aligned} ρ f σ f = ( 0 , γ β ) , ρ i = ( γ , 0 ) , ρ ˉ f = ( 0 , − γ ) , ρ ˉ i = ( − γ β , 0 ) , = σ i = − δ , σ ˉ f = σ ˉ i = δ . (87)
For example, γ β Δ − δ b = ρ f ⋅ ( u , Δ ) + σ f b \gamma\beta\boldsymbol{\Delta}-\delta b=\rho_f\cdot(\mathbf{u},\boldsymbol{\Delta})+\sigma_f b γ β Δ − δ b = ρ f ⋅ ( u , Δ ) + σ f b . With
χ = c 2 , 4 ρ f − c 2 , 4 ∗ ρ ˉ f − 2 i ( c 1 , 4 ρ i + c 1 , 4 ∗ ρ ˉ i ) , σ χ = c 2 , 4 σ f − c 2 , 4 ∗ σ ˉ f − 2 i ( c 1 , 4 σ i + c 1 , 4 ∗ σ ˉ i ) , (88) \begin{aligned}
\chi&= c_{2,4}\rho_f-c_{2,4}^{*}\bar{\rho}_f-2i(c_{1,4}\rho_i+c_{1,4}^{*}\bar{\rho}_i),\\
\sigma_\chi&= c_{2,4}\sigma_f-c_{2,4}^{*}\bar{\sigma}_f-2i(c_{1,4}\sigma_i+c_{1,4}^{*}\bar{\sigma}_i), \tag*{(88)}
\end{aligned} χ σ χ = c 2 , 4 ρ f − c 2 , 4 ∗ ρ ˉ f − 2 i ( c 1 , 4 ρ i + c 1 , 4 ∗ ρ ˉ i ) , = c 2 , 4 σ f − c 2 , 4 ∗ σ ˉ f − 2 i ( c 1 , 4 σ i + c 1 , 4 ∗ σ ˉ i ) , (88)
the quadratic, linear and constant parts of E 4 \mathcal{E}_4 E 4 are
K 4 = i ω + 2 c 2 , 4 ( ρ f ρ f T + ρ i ρ i T ) + ω + c 1 , 4 ( ρ f ρ i T + ρ i ρ f T ) − i ω + 2 c 2 , 4 ∗ ( ρ ˉ f ρ ˉ f T + ρ ˉ i ρ ˉ i T ) + ω + c 1 , 4 ∗ ( ρ ˉ f ρ ˉ i T + ρ ˉ i ρ ˉ f T ) − i ω + 2 D 4 χ χ T , (89) \begin{aligned}
K_4={}&\frac{i\omega^+}{2}c_{2,4}(\rho_f\rho_f^T+\rho_i\rho_i^T)+\omega^+c_{1,4}(\rho_f\rho_i^T+\rho_i\rho_f^T)\\
&-\frac{i\omega^+}{2}c_{2,4}^{*}(\bar{\rho}_f\bar{\rho}_f^T+\bar{\rho}_i\bar{\rho}_i^T)+\omega^+c_{1,4}^{*}(\bar{\rho}_f\bar{\rho}_i^T+\bar{\rho}_i\bar{\rho}_f^T)-\frac{i\omega^+}{2D_4}\chi\chi^T, \tag*{(89)}
\end{aligned} K 4 = 2 i ω + c 2 , 4 ( ρ f ρ f T + ρ i ρ i T ) + ω + c 1 , 4 ( ρ f ρ i T + ρ i ρ f T ) − 2 i ω + c 2 , 4 ∗ ( ρ ˉ f ρ ˉ f T + ρ ˉ i ρ ˉ i T ) + ω + c 1 , 4 ∗ ( ρ ˉ f ρ ˉ i T + ρ ˉ i ρ ˉ f T ) − 2 D 4 i ω + χ χ T , (89)
ℓ 4 = i ω + c 2 , 4 ( σ f ρ f + σ i ρ i ) + 2 ω + c 1 , 4 ( σ f ρ i + σ i ρ f ) − i ω + c 2 , 4 ∗ ( σ ˉ f ρ ˉ f + σ ˉ i ρ ˉ i ) + 2 ω + c 1 , 4 ∗ ( σ ˉ f ρ ˉ i + σ ˉ i ρ ˉ f ) − i ω + D 4 σ χ χ , (90) \begin{aligned}
\ell_4={}&i\omega^+c_{2,4}(\sigma_f\rho_f+\sigma_i\rho_i)+2\omega^+c_{1,4}(\sigma_f\rho_i+\sigma_i\rho_f)\\
&-i\omega^+c_{2,4}^{*}(\bar{\sigma}_f\bar{\rho}_f+\bar{\sigma}_i\bar{\rho}_i)+2\omega^+c_{1,4}^{*}(\bar{\sigma}_f\bar{\rho}_i+\bar{\sigma}_i\bar{\rho}_f)-\frac{i\omega^+}{D_4}\sigma_\chi\chi, \tag*{(90)}
\end{aligned} ℓ 4 = i ω + c 2 , 4 ( σ f ρ f + σ i ρ i ) + 2 ω + c 1 , 4 ( σ f ρ i + σ i ρ f ) − i ω + c 2 , 4 ∗ ( σ ˉ f ρ ˉ f + σ ˉ i ρ ˉ i ) + 2 ω + c 1 , 4 ∗ ( σ ˉ f ρ ˉ i + σ ˉ i ρ ˉ f ) − D 4 i ω + σ χ χ , (90)
k 4 = i ω + 2 c 2 , 4 ( σ f 2 + σ i 2 ) + 2 ω + c 1 , 4 σ f σ i − i ω + 2 c 2 , 4 ∗ ( σ ˉ f 2 + σ ˉ i 2 ) + 2 ω + c 1 , 4 ∗ σ ˉ f σ ˉ i − i ω + 2 D 4 σ χ 2 . (91) \begin{aligned}
k_4={}&\frac{i\omega^+}{2}c_{2,4}(\sigma_f^2+\sigma_i^2)+2\omega^+c_{1,4}\sigma_f\sigma_i\\
&-\frac{i\omega^+}{2}c_{2,4}^{*}(\bar{\sigma}_f^2+\bar{\sigma}_i^2)+2\omega^+c_{1,4}^{*}\bar{\sigma}_f\bar{\sigma}_i-\frac{i\omega^+}{2D_4}\sigma_\chi^2. \tag*{(91)}
\end{aligned} k 4 = 2 i ω + c 2 , 4 ( σ f 2 + σ i 2 ) + 2 ω + c 1 , 4 σ f σ i − 2 i ω + c 2 , 4 ∗ ( σ ˉ f 2 + σ ˉ i 2 ) + 2 ω + c 1 , 4 ∗ σ ˉ f σ ˉ i − 2 D 4 i ω + σ χ 2 . (91)
The coefficients of the combined exponent in Eq. (66) are
M = − 2 [ K 4 + ( i ω + 2 c 2 , 3 0 0 − q ^ 4 ( L + − x + − τ 3 − τ 4 ) ( ( 1 − z ) 2 + 2 z 2 ) ) ] , M=-2\left[K_4+
\begin{pmatrix}
\dfrac{i\omega^+}{2}c_{2,3} & 0\\
0 & -\dfrac{\hat q}{4}(L^+-x^+-\tau_3-\tau_4)((1-z)^2+2z^2)
\end{pmatrix}\right], M = − 2 K 4 + 2 i ω + c 2 , 3 0 0 − 4 q ^ ( L + − x + − τ 3 − τ 4 ) (( 1 − z ) 2 + 2 z 2 ) ,
( s u s Δ ) = ℓ 4 + ( − z ω + 1 + z 2 ( 2 c 1 , 3 + i c 2 , 3 ) − q ^ 2 ( L + − x + − τ 3 − τ 4 ) ( 1 − 3 z ) ) , \begin{pmatrix}s_u\\s_\Delta\end{pmatrix}
=\ell_4+
\begin{pmatrix}
-\dfrac{z\omega^+}{1+z^2}(2c_{1,3}+ic_{2,3})\\
-\dfrac{\hat q}{2}(L^+-x^+-\tau_3-\tau_4)(1-3z)
\end{pmatrix}, ( s u s Δ ) = ℓ 4 + − 1 + z 2 z ω + ( 2 c 1 , 3 + i c 2 , 3 ) − 2 q ^ ( L + − x + − τ 3 − τ 4 ) ( 1 − 3 z ) ,
C 0 = k 4 − q ^ x + 4 − q ^ τ 3 4 ( 1 + z 2 ) − q ^ τ 4 4 λ 2 − 3 q ^ 4 ( L + − x + − τ 3 − τ 4 ) (92) C_0=k_4-\frac{\hat q x^+}{4}-\frac{\hat q\tau_3}{4(1+z^2)}-\frac{\hat q\tau_4}{4\lambda^2}-\frac{3\hat q}{4}(L^+-x^+-\tau_3-\tau_4) \tag*{(92)} C 0 = k 4 − 4 q ^ x + − 4 ( 1 + z 2 ) q ^ τ 3 − 4 λ 2 q ^ τ 4 − 4 3 q ^ ( L + − x + − τ 3 − τ 4 ) (92)
+ z 2 ω + ( 1 + z 2 ) 2 ( 2 c 1 , 3 + i c 2 , 3 ) . (93) +\frac{z^2\omega^+}{(1+z^2)^2}(2c_{1,3}+ic_{2,3}). \tag*{(93)} + ( 1 + z 2 ) 2 z 2 ω + ( 2 c 1 , 3 + i c 2 , 3 ) . (93)
The polynomial prefactor arises from T ( Δ , b ) ∇ v S ~ ( 3 ) ⋅ ∇ w S ~ ( 4 ) \mathcal{T}(\Delta,\mathbf{b})\,\nabla_{\mathbf{v}}\widetilde{\mathcal{S}}^{(3)}\cdot\nabla_{\mathbf{w}}\widetilde{\mathcal{S}}^{(4)} T ( Δ , b ) ∇ v S ( 3 ) ⋅ ∇ w S ( 4 ) . We write the two linear forms as
2 ω + c 1 , 3 u − z ω + 1 + z 2 ( 2 c 1 , 3 + i c 2 , 3 ) b = L u u + L b b , W = W u u + W Δ Δ + W b b , (94) 2\omega^+c_{1,3}\mathbf{u}-\frac{z\omega^+}{1+z^2}(2c_{1,3}+ic_{2,3})\mathbf{b}=L_u\mathbf{u}+L_b\mathbf{b},
\qquad
\mathcal{W}=W_u\mathbf{u}+W_\Delta\Delta+W_b\mathbf{b}, \tag*{(94)} 2 ω + c 1 , 3 u − 1 + z 2 z ω + ( 2 c 1 , 3 + i c 2 , 3 ) b = L u u + L b b , W = W u u + W Δ Δ + W b b , (94)
with
L u = 2 ω + c 1 , 3 , L b = − z ω + 1 + z 2 ( 2 c 1 , 3 + i c 2 , 3 ) , (95) L_u=2\omega^+c_{1,3},\qquad
L_b=-\frac{z\omega^+}{1+z^2}(2c_{1,3}+ic_{2,3}), \tag*{(95)} L u = 2 ω + c 1 , 3 , L b = − 1 + z 2 z ω + ( 2 c 1 , 3 + i c 2 , 3 ) , (95)
and
( W u W Δ ) = 2 ( β c 1 , 4 ρ f − c 1 , 4 ∗ ρ ˉ f ) − 2 β c 1 , 4 − c 1 , 4 ∗ D 4 χ + i ( β c 2 , 4 ρ i + c 2 , 4 ∗ ρ ˉ i ) , W b = 2 ( β c 1 , 4 σ f − c 1 , 4 ∗ σ ˉ f ) − 2 β c 1 , 4 − c 1 , 4 ∗ D 4 σ χ + i ( β c 2 , 4 σ i + c 2 , 4 ∗ σ ˉ i ) . (96) \begin{aligned}
(W_u\ W_\Delta)&=2(\beta c_{1,4}\rho_f-c_{1,4}^{*}\bar{\rho}_f)-2\frac{\beta c_{1,4}-c_{1,4}^{*}}{D_4}\chi+i(\beta c_{2,4}\rho_i+c_{2,4}^{*}\bar{\rho}_i),\\
W_b&=2(\beta c_{1,4}\sigma_f-c_{1,4}^{*}\bar{\sigma}_f)-2\frac{\beta c_{1,4}-c_{1,4}^{*}}{D_4}\sigma_\chi+i(\beta c_{2,4}\sigma_i+c_{2,4}^{*}\bar{\sigma}_i). \tag*{(96)}
\end{aligned} ( W u W Δ ) W b = 2 ( β c 1 , 4 ρ f − c 1 , 4 ∗ ρ ˉ f ) − 2 D 4 β c 1 , 4 − c 1 , 4 ∗ χ + i ( β c 2 , 4 ρ i + c 2 , 4 ∗ ρ ˉ i ) , = 2 ( β c 1 , 4 σ f − c 1 , 4 ∗ σ ˉ f ) − 2 D 4 β c 1 , 4 − c 1 , 4 ∗ σ χ + i ( β c 2 , 4 σ i + c 2 , 4 ∗ σ ˉ i ) . (96)
Their product is
( L u u + L b b ) ⋅ ( W u u + W Δ Δ + W b b ) = q 1 u 2 + q 2 u ⋅ Δ + q 3 b ⋅ u + q 4 b ⋅ Δ + q 5 b 2 , (97) \begin{aligned}
(L_u\mathbf{u}+L_b\mathbf{b})\cdot(W_u\mathbf{u}+W_\Delta\Delta+W_b\mathbf{b})
&=q_1\mathbf{u}^2+q_2\mathbf{u}\cdot\Delta+q_3\mathbf{b}\cdot\mathbf{u}+q_4\mathbf{b}\cdot\Delta+q_5b^2, \tag*{(97)}
\end{aligned} ( L u u + L b b ) ⋅ ( W u u + W Δ Δ + W b b ) = q 1 u 2 + q 2 u ⋅ Δ + q 3 b ⋅ u + q 4 b ⋅ Δ + q 5 b 2 , (97)
with
q 1 = L u W u , q 2 = L u W Δ , q 3 = L u W b + L b W u , q 4 = L b W Δ , q 5 = L b W b . (98) q_1=L_uW_u,\qquad q_2=L_uW_\Delta,\qquad q_3=L_uW_b+L_bW_u,\qquad q_4=L_bW_\Delta,\qquad q_5=L_bW_b. \tag*{(98)} q 1 = L u W u , q 2 = L u W Δ , q 3 = L u W b + L b W u , q 4 = L b W Δ , q 5 = L b W b . (98)
Wick’s theorem then fixes all required moments of u \mathbf{u} u and Δ \Delta Δ . Denoting Gaussian expectation values by angle brackets, the connected second moments are ⟨ u j u k ⟩ − U j U k = C u u δ j k \langle u_j u_k\rangle-U_jU_k=C_{uu}\delta_{jk} ⟨ u j u k ⟩ − U j U k = C uu δ j k , ⟨ u j Δ k ⟩ − U j V k = C u Δ δ j k \langle u_j\Delta_k\rangle-U_jV_k=C_{u\Delta}\delta_{jk} ⟨ u j Δ k ⟩ − U j V k = C u Δ δ j k and ⟨ Δ j Δ k ⟩ − V j V k = C Δ Δ δ j k \langle\Delta_j\Delta_k\rangle-V_jV_k=C_{\Delta\Delta}\delta_{jk} ⟨ Δ j Δ k ⟩ − V j V k = C ΔΔ δ j k , so that, for example,
⟨ Δ 2 u 2 ⟩ = ( V 2 + 2 C Δ Δ ) ( U 2 + 2 C u u ) + 4 C u Δ U ⋅ V + 4 C u Δ 2 . (99) \langle\Delta^2u^2\rangle=(V^2+2C_{\Delta\Delta})(U^2+2C_{uu})+4C_{u\Delta}U\cdot V+4C_{u\Delta}^2. \tag*{(99)} ⟨ Δ 2 u 2 ⟩ = ( V 2 + 2 C ΔΔ ) ( U 2 + 2 C uu ) + 4 C u Δ U ⋅ V + 4 C u Δ 2 . (99)
The complete momentum polynomial is
P ( p 2 , b ⋅ p ) = 2 z ( 1 − z ) [ q 1 ⟨ Δ 2 u 2 ⟩ + q 2 ⟨ Δ 2 u ⋅ Δ ⟩ + q 3 ⟨ Δ 2 b ⋅ u ⟩ + q 4 ⟨ Δ 2 b ⋅ Δ ⟩ + q 5 b 2 ⟨ Δ 2 ⟩ ] − 2 ( 1 − 2 z ) [ q 1 ⟨ ( b ⋅ Δ ) u 2 ⟩ + q 2 ⟨ ( b ⋅ Δ ) ( u ⋅ Δ ) ⟩ + q 3 ⟨ ( b ⋅ Δ ) ( b ⋅ u ) ⟩ + q 4 ⟨ ( b ⋅ Δ ) 2 ⟩ + q 5 b 2 ⟨ b ⋅ Δ ⟩ \begin{aligned}
\mathcal{P}(p^2,\mathbf{b}\cdot\mathbf{p})={}&2z(1-z)\left[q_1\langle\Delta^2u^2\rangle+q_2\langle\Delta^2\mathbf{u}\cdot\Delta\rangle+q_3\langle\Delta^2\mathbf{b}\cdot\mathbf{u}\rangle\right.\\
&\left.+q_4\langle\Delta^2\mathbf{b}\cdot\Delta\rangle+q_5b^2\langle\Delta^2\rangle\right]\\
&-2(1-2z)\left[q_1\langle(\mathbf{b}\cdot\Delta)u^2\rangle+q_2\langle(\mathbf{b}\cdot\Delta)(\mathbf{u}\cdot\Delta)\rangle\right.\\
&\left.+q_3\langle(\mathbf{b}\cdot\Delta)(\mathbf{b}\cdot\mathbf{u})\rangle+q_4\langle(\mathbf{b}\cdot\Delta)^2\rangle+q_5b^2\langle\mathbf{b}\cdot\Delta\rangle\right.
\end{aligned} P ( p 2 , b ⋅ p ) = 2 z ( 1 − z ) [ q 1 ⟨ Δ 2 u 2 ⟩ + q 2 ⟨ Δ 2 u ⋅ Δ ⟩ + q 3 ⟨ Δ 2 b ⋅ u ⟩ + q 4 ⟨ Δ 2 b ⋅ Δ ⟩ + q 5 b 2 ⟨ Δ 2 ⟩ ] − 2 ( 1 − 2 z ) [ q 1 ⟨( b ⋅ Δ ) u 2 ⟩ + q 2 ⟨( b ⋅ Δ ) ( u ⋅ Δ )⟩ + q 3 ⟨( b ⋅ Δ ) ( b ⋅ u )⟩ + q 4 ⟨( b ⋅ Δ ) 2 ⟩ + q 5 b 2 ⟨ b ⋅ Δ ⟩
− 2 b 2 [ q 1 ⟨ u 2 ⟩ + q 2 ⟨ u ⋅ Δ ⟩ + q 3 ⟨ b ⋅ u ⟩ + q 4 ⟨ b ⋅ Δ ⟩ + q 5 b 2 ] . (100) -2b^2\left[q_1\langle u^2\rangle+q_2\langle u\cdot\Delta\rangle+q_3\langle b\cdot u\rangle+q_4\langle b\cdot\Delta\rangle+q_5b^2\right]. \tag*{(100)} − 2 b 2 [ q 1 ⟨ u 2 ⟩ + q 2 ⟨ u ⋅ Δ ⟩ + q 3 ⟨ b ⋅ u ⟩ + q 4 ⟨ b ⋅ Δ ⟩ + q 5 b 2 ] . (100)
Relative momentum integration and short time subtraction
a. Gaussian moments. Each fixed-p p p result of Sec. IV is a polynomial P ( p 2 , b ⋅ p ) P(p^2,b\cdot p) P ( p 2 , b ⋅ p ) multiplying e − α p 2 + i ( κ / b ) b ⋅ p e^{-\alpha p^2+i(\kappa/b)b\cdot p} e − α p 2 + i ( κ / b ) b ⋅ p , with complex α \alpha α and κ \kappa κ that depend on the time variables, on z z z and on the measurement channel. With the regulator of Eq. (47) ,
I η U V ( α , κ ) ≡ ∫ R 2 d 2 p ( 2 π ) 2 exp [ − ( α + η U V ) p 2 + i κ b b ⋅ p ] = 1 4 π ( α + η U V ) exp [ − κ 2 4 ( α + η U V ) ] , (101) \mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha,\kappa)\equiv\int_{\mathbb{R}^2}\frac{\mathrm{d}^2p}{(2\pi)^2}\exp\left[-(\alpha+\eta_{\mathrm{UV}})p^2+i\frac{\kappa}{b}b\cdot p\right]
=\frac{1}{4\pi(\alpha+\eta_{\mathrm{UV}})}\exp\left[-\frac{\kappa^2}{4(\alpha+\eta_{\mathrm{UV}})}\right], \tag*{(101)} I η UV ( α , κ ) ≡ ∫ R 2 ( 2 π ) 2 d 2 p exp [ − ( α + η UV ) p 2 + i b κ b ⋅ p ] = 4 π ( α + η UV ) 1 exp [ − 4 ( α + η UV ) κ 2 ] , (101)
and the polynomial moments follow by differentiation,
∫ R 2 d 2 p ( 2 π ) 2 e − ( α + η U V ) p 2 + i ( κ / b ) b ⋅ p P ( p 2 , b ⋅ p ) = P ( − ∂ α , − i b ∂ κ ) I η U V ( α , κ ) . (102) \int_{\mathbb{R}^2}\frac{\mathrm{d}^2p}{(2\pi)^2}e^{-(\alpha+\eta_{\mathrm{UV}})p^2+i(\kappa/b)b\cdot p}P(p^2,b\cdot p)=P(-\partial_\alpha,-ib\,\partial_\kappa)\mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha,\kappa). \tag*{(102)} ∫ R 2 ( 2 π ) 2 d 2 p e − ( α + η UV ) p 2 + i ( κ / b ) b ⋅ p P ( p 2 , b ⋅ p ) = P ( − ∂ α , − ib ∂ κ ) I η UV ( α , κ ) . (102)
The right hand sides are the analytic continuation from the domain in which the integrals converge absolutely. The in–out contribution also contains the inverse moment
H η U V ( α , κ ) ≡ ∫ R 2 d 2 p ( 2 π ) 2 b ⋅ p p 2 e − ( α + η U V ) p 2 + i ( κ / b ) b ⋅ p = i b 2 π κ [ 1 − e − κ 2 / [ 4 ( α + η U V ) ] ] . (103) \mathcal{H}_{\eta_{\mathrm{UV}}}(\alpha,\kappa)\equiv\int_{\mathbb{R}^2}\frac{\mathrm{d}^2p}{(2\pi)^2}\frac{b\cdot p}{p^2}e^{-(\alpha+\eta_{\mathrm{UV}})p^2+i(\kappa/b)b\cdot p}=\frac{ib}{2\pi\kappa}\left[1-e^{-\kappa^2/[4(\alpha+\eta_{\mathrm{UV}})]}\right]. \tag*{(103)} H η UV ( α , κ ) ≡ ∫ R 2 ( 2 π ) 2 d 2 p p 2 b ⋅ p e − ( α + η UV ) p 2 + i ( κ / b ) b ⋅ p = 2 π κ ib [ 1 − e − κ 2 / [ 4 ( α + η UV )] ] . (103)
α f a c = 1 4 A G ( x + + τ , x + ) , κ f a c , i = b [ a i + B G ( x + + τ , x + ) 2 A G ( x + + τ , x + ) ] . (104) \alpha_{\mathrm{fac}}=\frac{1}{4A_G(x^++\tau,x^+)},\qquad\kappa_{\mathrm{fac},i}=b\left[a_i+\frac{B_G(x^++\tau,x^+)}{2A_G(x^++\tau,x^+)}\right]. \tag*{(104)} α fac = 4 A G ( x + + τ , x + ) 1 , κ fac , i = b [ a i + 2 A G ( x + + τ , x + ) B G ( x + + τ , x + ) ] . (104)
In the non-factorizable contribution they are α \alpha α and κ i \kappa_i κ i of Eq. (71) , with the polynomial P P P of Eq. (70) and the prefactor in Eq. (72) . In both cases Eq. (102) gives the p p p integral in closed form. In the in–out contribution, Eq. (76) , we set
α o u t = i 2 ω + c 2 ( τ ) , κ o u t , i = b [ a i + z 1 + z 2 ( r ( τ ) − 1 ) ] , r ( τ ) = 2 i c 1 ( τ ) c 2 ( τ ) . (105) \alpha_{\mathrm{out}}=\frac{i}{2\omega^+c_2(\tau)},\qquad\kappa_{\mathrm{out},i}=b\left[a_i+\frac{z}{1+z^2}(r(\tau)-1)\right],\qquad r(\tau)=\frac{2ic_1(\tau)}{c_2(\tau)}. \tag*{(105)} α out = 2 ω + c 2 ( τ ) i , κ out , i = b [ a i + 1 + z 2 z ( r ( τ ) − 1 ) ] , r ( τ ) = c 2 ( τ ) 2 i c 1 ( τ ) . (105)
and the p p p integral of the integrand, without the common normalization, is
exp [ − q ^ b 2 4 ( L + − τ ) − q ^ b 2 4 ( 1 + z 2 ) τ − q ^ z 2 b 2 4 ( 1 + z 2 ) c 2 ( τ ) ] [ − 2 i ω + r 2 ( τ ) I η U V ( α o u t , κ o u t , i ) + i q ^ z ω + c 2 ( τ ) r ( τ ) H η U V ( α o u t , κ o u t , i ) ] . (106) \exp\left[-\frac{\widehat{q}b^2}{4}(L^+-\tau)-\frac{\widehat{q}b^2}{4(1+z^2)}\tau-\frac{\widehat{q}z^2b^2}{4(1+z^2)c_2(\tau)}\right]
\left[-2i\omega^+r^2(\tau)\mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha_{\mathrm{out}},\kappa_{\mathrm{out},i})+\frac{i\widehat{q}z\omega^+}{c_2(\tau)}r(\tau)\mathcal{H}_{\eta_{\mathrm{UV}}}(\alpha_{\mathrm{out}},\kappa_{\mathrm{out},i})\right]. \tag*{(106)} exp [ − 4 q b 2 ( L + − τ ) − 4 ( 1 + z 2 ) q b 2 τ − 4 ( 1 + z 2 ) c 2 ( τ ) q z 2 b 2 ] [ − 2 i ω + r 2 ( τ ) I η UV ( α out , κ out , i ) + c 2 ( τ ) i q z ω + r ( τ ) H η UV ( α out , κ out , i ) ] . (106)
After these steps only the time integrations, the z z z integral and the channel sum remain.
c. Short-time subtraction. The regulator matters only where the Gaussian coefficient of p 2 p^2 p 2 is small, that is for τ → 0 \tau\to0 τ → 0 in the factorizable and in–out contributions and for τ 3 , τ 4 → 0 \tau_3,\tau_4 \to0 τ 3 , τ 4 → 0 in the nonfactorizable one. These limits carry the large ∣ p ∣ |\mathbf{p}| ∣ p ∣ tails of the separate contributions. At fixed p \mathbf{p} p each contribution is a time integral of the vacuum phase e − i p 2 ( x ˉ + − x + ) / ( 2 ω + ) e^{-i\mathbf{p}^2(\bar{x}^+-x^+)/(2\omega^+)} e − i p 2 ( x ˉ + − x + ) / ( 2 ω + ) times a weight w ( x + , x ˉ + ) w(x^+,\bar{x}^+) w ( x + , x ˉ + ) , which varies on the medium time scales and on the resolution time ( ω + 2 / q ^ p 2 ) 1 / 3 (\omega^{+2}/\hat{q}\mathbf{p}^2)^{1/3} ( ω + 2 / q ^ p 2 ) 1/3 . When the formation time 2 ω + / p 2 2\omega^+/\mathbf{p}^2 2 ω + / p 2 is short compared with these times, integration by parts in x ˉ + − x + \bar{x}^+-x^+ x ˉ + − x + gives non-oscillating tails of order ( 2 ω + / p 2 ) 2 (2\omega^+/\mathbf{p}^2)^2 ( 2 ω + / p 2 ) 2 , from the boundaries of the time domains and from the in–in bulk. At large ∣ p ∣ |\mathbf{p}| ∣ p ∣ the pair is resolved immediately after the splitting, so that the equal-time weight is w ( x ˉ + , x ˉ + ) = D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) w(\bar{x}^+,\bar{x}^+)=D_F(\mathbf{b};\bar{x}^+,0)D_F^3(\mathbf{b};L^+,\bar{x}^+) w ( x ˉ + , x ˉ + ) = D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) . The in–in region then gives a tail proportional to w ( 0 , 0 ) + w ( L + , L + ) w(0,0)+w(L^+,L^+) w ( 0 , 0 ) + w ( L + , L + ) , whose bulk part w ( L + , L + ) − w ( 0 , 0 ) w(L^+,L^+)-w(0,0) w ( L + , L + ) − w ( 0 , 0 ) comes from the nonfactorizable contribution, and the in–out region gives − 2 w ( L + , L + ) -2w(L^+,L^+) − 2 w ( L + , L + ) with the same coefficient. The separate tails are of order one, while their sum, D F 3 ( b ; L + , 0 ) − D F ( b ; L + , 0 ) D_F^3(\mathbf{b};L^+,0)-D_F(\mathbf{b};L^+,0) D F 3 ( b ; L + , 0 ) − D F ( b ; L + , 0 ) , is the tail of Eq. (46) and vanishes at b = 0 b=0 b = 0 . Therefore, we subtract the short-time limits of the factorizable and in–out integrands, in which the pair is compact and the three-point correlator reduces to the parent dipole,
lim ∣ r ∣ → 0 D F [ b − z r ] D F [ r ] = D F [ b ] . (107) \lim_{|\mathbf{r}|\to0}D_F[\mathbf{b}-z\mathbf{r}]D_F[\mathbf{r}]=D_F[\mathbf{b}]. \tag*{(107)} ∣ r ∣ → 0 lim D F [ b − z r ] D F [ r ] = D F [ b ] . (107)
With α = i τ / ( 2 ω + ) \alpha=i\tau/(2\omega^+) α = i τ / ( 2 ω + ) , the limit of the factorizable integrand, where τ = x ˉ + − x + \tau=\bar{x}^+-x^+ τ = x ˉ + − x + , is
D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) I η U V ( α , a i b ) α + η U V [ 1 − a i 2 b 2 4 ( α + η U V ) ] . (108) D_F(\mathbf{b};\bar{x}^+,0)D_F^3(\mathbf{b};L^+,\bar{x}^+)\frac{\mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha,a_i b)}{\alpha+\eta_{\mathrm{UV}}}\left[1-\frac{a_i^2b^2}{4(\alpha+\eta_{\mathrm{UV}})}\right]. \tag*{(108)} D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) α + η UV I η UV ( α , a i b ) [ 1 − 4 ( α + η UV ) a i 2 b 2 ] . (108)
and that of the in–out integrand, where τ = L + − x + \tau=L^+-x^+ τ = L + − x + , is
− 2 i ω + D F ( b ; L + , 0 ) I η U V ( α , a i b ) . (109) -2i\omega^+D_F(\mathbf{b};L^+,0)\mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha,a_i b). \tag*{(109)} − 2 i ω + D F ( b ; L + , 0 ) I η UV ( α , a i b ) . (109)
They are subtracted at the same values of z z z , of the time variables, of the channel and of η U V \eta_{\mathrm{UV}} η UV , and their integral is added back. At fixed x ˉ + \bar{x}^+ x ˉ + the factorizable kernel is 2 i ω + ∂ τ 2i\omega^+\partial_\tau 2 i ω + ∂ τ of the equal-time weight times I η U V ( α , a i b ) \mathcal{I}_{\eta_{\mathrm{UV}}}(\alpha,a_i b) I η UV ( α , a i b ) , so that the added term is elementary. At η U V = 0 \eta_{\mathrm{UV}}=0 η UV = 0 , with the common normalization and real-part projection of Sec. IV, it reads
2 ω + 2 π ∫ 0 L + d x ˉ + x ˉ + [ D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) − D F ( b ; L + , 0 ) ] cos a i 2 b 2 ω + 2 x ˉ + , (110) \frac{2\omega^{+2}}{\pi}\int_0^{L^+}\frac{d\bar{x}^+}{\bar{x}^+}\left[D_F(\mathbf{b};\bar{x}^+,0)D_F^3(\mathbf{b};L^+,\bar{x}^+)-D_F(\mathbf{b};L^+,0)\right]\cos\frac{a_i^2b^2\omega^+}{2\bar{x}^+}, \tag*{(110)} π 2 ω + 2 ∫ 0 L + x ˉ + d x ˉ + [ D F ( b ; x ˉ + , 0 ) D F 3 ( b ; L + , x ˉ + ) − D F ( b ; L + , 0 ) ] cos 2 x ˉ + a i 2 b 2 ω + , (110)
which converges at x ˉ + → 0 \bar{x}^+\to0 x ˉ + → 0 through the oscillation and vanishes at b = 0 b=0 b = 0 . The nonfactorizable integrand is not subtracted. It takes the compact-pair form only for τ 4 ≲ ( ω + τ 3 / q ^ ) 1 / 3 \tau_4\lesssim(\omega^+\tau_3/\hat{q})^{1/3} τ 4 ≲ ( ω + τ 3 / q ^ ) 1/3 , and its tail, the bulk term above, is of the size of the tail of the sum. After the subtraction the factorizable integrand has no residual logarithm of η U V \eta_{\mathrm{UV}} η UV , and the subtracted and unsubtracted evaluations of the jet function agree within the uncertainty of the regulator removal.
Endpoint treatment and numerical implementation
a. Quadrature. The time quadrature resolves the regulator time scale, t U V ∼ 2 ω + η U V t_{\mathrm{UV}} \sim2\omega^{+}\eta_{\mathrm{UV}} t UV ∼ 2 ω + η UV . Ordered times are mapped to triangular domains, and long oscillatory intervals are divided into phase blocks. Exact and subtraction kernels are evaluated on the same nodes and subtracted before integration. Each contribution is converged separately, with a tolerance set by the final jet function, and so is their sum.
b. Longitudinal endpoints. As ω + = z ( 1 − z ) p 0 + → 0 \omega^{+}=z(1-z)p_{0}^{+}\to0 ω + = z ( 1 − z ) p 0 + → 0 , the oscillator frequency of the three-point function, ∣ Ω 3 ∣ = [ ( 1 + z 2 ) q ^ / ( 2 ω + ) ] 1 / 2 |\Omega_{3}|=[(1+z^{2})\hat{q}/(2\omega^{+})]^{1/2} ∣ Ω 3 ∣ = [( 1 + z 2 ) q ^ / ( 2 ω + ) ] 1/2 , grows and the formation time 1 / ∣ Ω 3 ∣ 1/|\Omega_{3}| 1/∣ Ω 3 ∣ becomes short compared with L + L^{+} L + . Integrated over p \mathbf{p} p and K \mathbf{K} K , the medium-induced cross section is the harmonic-oscillator BDMPS–Z spectrum [69 , 70 ] ,
d σ ^ q g ∣ p 0 m e d d z = α s C F π P q → q g v a c ( z ) Re ln cos ( Ω 3 L + ) , (111) \frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz}=\frac{\alpha_{s}C_{F}}{\pi}P^{\mathrm{vac}}_{q\to qg}(z)\operatorname{Re}\ln\cos(\Omega_{3}L^{+}) , \tag*{(111)} d z d σ ^ q g ∣ p 0 med = π α s C F P q → q g vac ( z ) Re ln cos ( Ω 3 L + ) , (111)
which for ∣ Ω 3 ∣ L + ≫ 1 |\Omega_{3}|L^{+}\gg1 ∣ Ω 3 ∣ L + ≫ 1 becomes
d σ ^ q g ∣ p 0 m e d d z ≃ α s C F π L + P q → q g v a c ( z ) ( 1 + z 2 ) q ^ 4 z ( 1 − z ) p 0 + , (112) \frac{d\hat{\sigma}^{\mathrm{med}}_{qg|p_{0}}}{dz}\simeq\frac{\alpha_{s}C_{F}}{\pi}L^{+}P^{\mathrm{vac}}_{q\to qg}(z)\sqrt{\frac{(1+z^{2})\hat{q}}{4z(1-z)p_{0}^{+}}} , \tag*{(112)} d z d σ ^ q g ∣ p 0 med ≃ π α s C F L + P q → q g vac ( z ) 4 z ( 1 − z ) p 0 + ( 1 + z 2 ) q ^ , (112)
up to relative corrections of order ( ∣ Ω 3 ∣ L + ) − 1 (|\Omega_{3}|L^{+})^{-1} ( ∣ Ω 3 ∣ L + ) − 1 . The z z z integral is evaluated on three bulk panels, [ 0.02 , 0.2 ] [0.02,0.2] [ 0.02 , 0.2 ] in the variable z \sqrt{z} z , [ 0.2 , 0.8 ] [0.2,0.8] [ 0.2 , 0.8 ] in z z z , and [ 0.8 , 0.98 ] [0.8,0.98] [ 0.8 , 0.98 ] in 1 − z \sqrt{1-z} 1 − z , and on the strips 10 − 3 ≤ z ≤ 2 × 10 − 2 10^{-3}\leq z\leq2\times10^{-2} 1 0 − 3 ≤ z ≤ 2 × 1 0 − 2 and 10 − 3 ≤ 1 − z ≤ 2 × 10 − 2 10^{-3}\leq1-z\leq2\times10^{-2} 1 0 − 3 ≤ 1 − z ≤ 2 × 1 0 − 2 . Below 10 − 3 10^{-3} 1 0 − 3 each contribution is completed from its asymptotic form. In units of the z z z integrand of Eq. (47) , the factorizable contribution minus its short-time kernel behaves as
C r i g h t ( 1 − z ) − 1 / 2 + d f a c ln 1 1 − z , C_{\mathrm{right}}(1-z)^{-1/2}+d_{\mathrm{fac}}\ln\frac{1}{1-z}, C right ( 1 − z ) − 1/2 + d fac ln 1 − z 1 ,
C r i g h t = − 2 C F α s π D F ( b ; L + , 0 ) ℓ η q ^ p 0 + , ℓ η = ln ( 1 + 2 q ^ L + η U V ) 2 q ^ η U V , (113) C_{\mathrm{right}}=-\frac{\sqrt{2}C_{F}\alpha_{s}}{\pi}D_{F}(\mathbf{b};L^{+},0)\ell_{\eta}\sqrt{\frac{\hat{q}}{p_{0}^{+}}},\qquad\ell_{\eta}=\frac{\ln(1+2\hat{q}L^{+}\eta_{\mathrm{UV}})}{2\hat{q}\eta_{\mathrm{UV}}}, \tag*{(113)} C right = − π 2 C F α s D F ( b ; L + , 0 ) ℓ η p 0 + q ^ , ℓ η = 2 q ^ η UV ln ( 1 + 2 q ^ L + η UV ) , (113)
with d f a c = − 2 C F α s π [ D F ( b ; L + , 0 ) 2 ( 1 + 2 q ^ L + η U V ) − D F 3 ( b ; L + , 0 ) ] d_{\mathrm{fac}}=-\frac{2C_{F}\alpha_{s}}{\pi}\left[\frac{D_{F}(\mathbf{b};L^{+},0)}{2(1+2\hat{q}L^{+}\eta_{\mathrm{UV}})}-D_{F}^{3}(\mathbf{b};L^{+},0)\right] d fac = − π 2 C F α s [ 2 ( 1 + 2 q ^ L + η UV ) D F ( b ; L + , 0 ) − D F 3 ( b ; L + , 0 ) ] , and the in–out contribution minus its kernel as − 2 C F α s π D F ( b ; L + , 0 ) ln ( ∣ Ω 3 ∣ L + ) -\frac{2C_{F}\alpha_{s}}{\pi}D_{F}(\mathbf{b};L^{+},0)\ln(|\Omega_{3}|L^{+}) − π 2 C F α s D F ( b ; L + , 0 ) ln ( ∣ Ω 3 ∣ L + ) . For η U V → 0 \eta_{\mathrm{UV}}\to0 η UV → 0 , ℓ η → L + \ell_{\eta}\to L^{+} ℓ η → L + and C r i g h t C_{\mathrm{right}} C right follows from Eq. (112) with the weight h q + h u = − ( 1 − z ) h_{q}+h_{u}=-(1-z) h q + h u = − ( 1 − z ) , the gluon channel being suppressed at b ≠ 0 b\ne0 b = 0 by its phase b ⋅ p / ( 1 − z ) \mathbf{b}\cdot\mathbf{p}/(1-z) b ⋅ p / ( 1 − z ) . The nonfactorizable contribution behaves as A + B ln [ 1 / ( 1 − z ) ] A+B\ln[1/(1-z)] A + B ln [ 1/ ( 1 − z )] , with A A A and B B B fitted. At z → 0 z\to0 z → 0 the factorizable contribution behaves as C l e f t z − 1 / 2 C_{\mathrm{left}}z^{-1/2} C left z − 1/2 , with C l e f t C_{\mathrm{left}} C left obtained by the argument that gives C r i g h t C_{\mathrm{right}} C right and checked numerically to better than 0.5 % 0.5\% 0.5% , the nonfactorizable one as A + B ln ( 1 / z ) A+B\ln(1/z) A + B ln ( 1/ z ) with fitted coefficients, and the in–out contribution vanishes. The terms not fixed analytically are fitted on the strip nodes in bases containing ln [ 1 / ( 1 − z ) ] \ln[1/(1-z)] ln [ 1/ ( 1 − z )] , 1 1 1 , 1 − z \sqrt{1-z} 1 − z and 1 − z 1-z 1 − z , and likewise in z z z , and the spread over the bases is the endpoint uncertainty.
The endpoint completion is performed at fixed η U V \eta_{\mathrm{UV}} η UV , before the regulator is removed, following the order of integrations specified in Sec. III C.
References
c. Removal of the regulator. For each b b b the jet function is computed, after the endpoint completion, at η U V = 1 / Λ 2 \eta_{\mathrm{UV}}=1/\Lambda^2 η UV = 1/ Λ 2 with Λ b = 10 \Lambda b=10 Λ b = 10 , 12 12 12 and 14 14 14 , and with Λ = 10 \Lambda=10 Λ = 10 , 12 12 12 and 14 14 14 GeV for b > 1 G e V − 1 b>1\ \mathrm{GeV}^{-1} b > 1 GeV − 1 . These values satisfy η U V ≪ b 2 / 4 \eta_{\mathrm{UV}}\ll b^2/4 η UV ≪ b 2 /4 , so that the measurement phases cut off the large ∣ p ∣ |\mathbf{p}| ∣ p ∣ tail of Eq. (46) before the regulator acts and the remaining dependence on η U V \eta_{\mathrm{UV}} η UV is polynomial. The η U V → 0 + \eta_{\mathrm{UV}}\to0^+ η UV → 0 + value is the envelope of a quadratic fit through the three points and a linear fit through the two smallest, each enlarged by the endpoint uncertainties propagated with the fit weights. A fourth point, Λ b = 17 \Lambda b=17 Λ b = 17 or Λ = 17 \Lambda=17 Λ = 17 GeV, is held out, and the regulator uncertainty is the larger of the difference between the two fits and the shift of the extrapolation when this point is included. The extrapolation is checked for each time region as well as for the sum, so that an unconverged cancellation between regions cannot hide behind a small total uncertainty.
d. Impact parameter grid. The only value imposed on the curve is J ( 1 ) , m e d . r a d . ( 0 ) = 0 J^{(1),\mathrm{med.rad.}}(0)=0 J ( 1 ) , med.rad. ( 0 ) = 0 , Eq. (44) . The nodes are uniform in ln b \ln b ln b and are refined until the interpolation error is below 0.1 % 0.1\% 0.1% of the largest ∣ J ( 1 ) , m e d . r a d . ∣ |J^{(1),\mathrm{med.rad.}}| ∣ J ( 1 ) , med.rad. ∣ . At small b b b the gluon channel loses the energy of soft gluons, 1 − z ≲ b q ^ L + / 2 1-z\lesssim b\sqrt{\hat q L^+}/2 1 − z ≲ b q ^ L + /2 , whose measurement phase b ⋅ p / ( 1 − z ) \mathbf{b}\cdot\mathbf{p}/(1-z) b ⋅ p / ( 1 − z ) decoheres. With the spectrum of Eq. (112) this gives J ( 1 ) , m e d . r a d . ( b ) ≃ − C b J^{(1),\mathrm{med.rad.}}(b)\simeq-C\sqrt{b} J ( 1 ) , med.rad. ( b ) ≃ − C b , which the nodes confirm at the smallest b b b . Accordingly, we interpolate J ( 1 ) , m e d . r a d . / b J^{(1),\mathrm{med.rad.}}/\sqrt{b} J ( 1 ) , med.rad. / b , and below the smallest node, b ≈ 10 − 2 G e V − 1 b\approx10^{-2}\ \mathrm{GeV}^{-1} b ≈ 1 0 − 2 GeV − 1 , we use − C b + D b 3 / 2 -C\sqrt{b}+Db^{3/2} − C b + D b 3/2 matched to the smallest nodes. At large b b b every contribution decays through its broadening factors. Beyond the last node, b = 3.16 G e V − 1 b=3.16\ \mathrm{GeV}^{-1} b = 3.16 GeV − 1 , where ∣ J ( 1 ) , m e d . r a d . ∣ < 3 × 10 − 5 |J^{(1),\mathrm{med.rad.}}|<3\times10^{-5} ∣ J ( 1 ) , med.rad. ∣ < 3 × 1 0 − 5 , the correction is set to zero, and the slowest Gaussian decay of the broadening factors, e − q ^ L + b 2 / 12 e^{-\hat q L^+b^2/12} e − q ^ L + b 2 /12 , makes the omitted tail negligible. The quoted uncertainty combines quadrature, endpoint completion, regulator removal, interpolation and the small-b b b form as correlated envelopes, so that the tabulated correction can be folded with another impact-parameter kernel. arXiv:2609.10679.