Introduction
The holographic duality with its most concrete realization via AdS/CFT correspondence [1 , 2 ] has been of great importance in contemporary theoretical physics for the past few decades. It relates a gravitational theory in a ( d + 1 ) (d+1) ( d + 1 ) -dimensional bulk to a d d d -dimensional boundary theory; and it has been extensively studied with a variety of approaches, from more traditional position & Mellin space techniques [3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ] and boundary CFT perspectives [13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 ] to more recently appreciated explicit bulk analyses with technologies borrowed from the scattering amplitudes program [23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 ] . This paper is a natural continuation of the recent progress in the last approach.1
Quite independently of these developments in the study of holography, stochastic quantization has also been an active area of research where the tentative idea is not so different from that of the holography: we have two field theories, a ( d + 1 ) (d+1) ( d + 1 ) -dimensional one in which the fields evolve diffusively in a fictitious time t t t according to a Langevin equation (hence the stochastic nature), and a d d d -dimensional Euclidean one, whose equilibrium action governs the late-time limit t → ∞ t \to\infty t → ∞ of the former [37 ] . However, unlike the curved space holography, this duality is more of an equilibrium-evolution relation than a geometric bulk-boundary relation. Nevertheless, the schematic resemblance did lead to a sizable interest in finding a more concrete relation between these two concepts, and the literature has seen quite an effort along this direction [38 , 39 , 40 , 41 ] .
The first major result in the establishment of such a relation can be taken as the conjectured identification of the bulk action on one side with the Fokker-Planck action on the other [39 ] , which is roughly the identification of the partition functions of the holographic and stochastic models in the classical limit. This was later sharpened into a dictionary between the radial flow of the holographic Wilsonian effective action [42 , 43 ] and stochastic time evolution [37 ] : the stochastic time is the radial coordinate, the Fokker–Planck Hamiltonian is the radial Hamiltonian, and the multi-trace couplings of the Wilsonian action are fixed by stochastic correlation functions.
This dictionary has been worked out in a growing number of examples2 and the conceptual grasp around it has certainly improved;3 however, no analysis of gauge fields beyond the free level has been conducted yet,4 despite it being an important gap in the literature.5 We will address this situation by explicitly analyzing gluon correlators in A d S 4 \mathrm{AdS}_4 AdS 4 , which have been shown to take remarkably simple forms if written in a judiciously chosen coordinate system [66 ] .6 Although some form of a reduction in complexity is expected as four-dimensional Yang–Mills theory on AdS enjoys the classical Weyl invariance, making its tree-level perturbation expansion identical to diagrams on a flat half-space,7 the form of the Witten diagrams that we will
to be utilizing benefits from far more simplifications, thanks to the collaborative effort on the holography program.8
In this paper we combine these developments. We extend the stochastic–holographic dictionary to S U ( N ) \mathrm{SU}(N) SU ( N ) Yang–Mills theory in A d S 4 \mathrm{AdS}_4 AdS 4 at tree level and use it to analyze the gauge sector of the holographic Wilsonian action. Our main points are as follows:
a. Holographic realization of stochastic gauge fixing: The ( d + 1 ) (d+1) ( d + 1 ) -dimensional formulation of stochastic gauge fixing of Baulieu and Zwanziger [85 ] is realized by the bulk gauge field: the radial component A z A_z A z is Zwanziger’s gauge-fixing field, radial gauge A z = 0 A_z=0 A z = 0 is the Parisi–Wu process, and other bulk gauges are other stochastic gauge fixings.
b. Closed-form multi-trace couplings of gluons at finite cutoff: By utilizing the operator of [66 ] that maps scalar to gluon diagrams as the functional calculus of the stochastic drift, we can construct a closed-form map between the triple-trace & quadruple-trace flows of gluons and scalar results.9 We also provide closed form results for the exchange seed at finite cutoff and its longitudinal limit.
c. Analysis of gauge dependence in Yang–Mills: We show that, on transverse data, the entire gauge dependence of the quadruple-trace couplings in a single longitudinal zero mode, which is absent when the Gauss law is imposed and is inconsequential in the flat-space limit.10
d. Further remarks: We also comment on the Chern–Simons solution of the zero-energy Hamilton–Jacobi equation (§ III B 2), on the form of the loop-level statement
(§ 1.3 C, § VII), and on the de Sitter counterpart of our results (§ VI).
The paper is organized as follows. In § 1.1 we review stochastic quantization of gauge fields, the holographic Wilsonian dictionary, and the Yang–Mills theory in AdS4 _4 4 . We then proceed to § 1.3 where we recast the relation of [66 ] as a functional calculus and present the all-order tree-level argument with which we derive the gauge-sector dictionary in § IV and the explicit multi-trace flows and their gauge dependence in § 2.2 . We discuss our results in the context of cosmology in § VI and conclude with a brief discussion of open directions in § VII. The appendices collect several conceptual and technical details and the attached Mathematica file re-derives the formulae of sections III–V from their definitions (the gluon statements mechanically from the Lagrangian) and contains every check quoted in the text.
Reviews
Stochastic quantization
Basics
Although many excellent discussions of stochastic quantization exist elsewhere,11 12 we would like to quickly review the fundamentals for the sake of completeness.
Consider the generalized Stokes theorem ∫ M d Ω = ∫ ∂ M Ω \int_M d\Omega= \int_{\partial M}\Omega ∫ M d Ω = ∫ ∂ M Ω for any d d d -form Ω \Omega Ω , and propose a generalization of it in field theory [95 ]
S d [ ϕ ] = ∫ ∂ M Ω [ ϕ ( x ) ] , S d + 1 [ ϕ ] = ∫ M d Ω [ ϕ ( x ) ] (1) S_d[\phi] = \int_{\partial M}\Omega[\phi(x)], \qquad S_{d+1}[\phi] = \int_M d\Omega[\phi(x)] \tag*{(1)} S d [ ϕ ] = ∫ ∂ M Ω [ ϕ ( x )] , S d + 1 [ ϕ ] = ∫ M d Ω [ ϕ ( x )] (1)
where we take the actions S d [ ϕ ] S_d[\phi] S d [ ϕ ] and S d + 1 [ ϕ ] S_{d+1}[\phi] S d + 1 [ ϕ ] to be schematically related to each other as
S d + 1 [ ϕ ] = ∫ M d t ∧ d ϕ d t δ S d [ ϕ ] δ ϕ . (2) S_{d+1}[\phi] = \int_M dt \wedge\frac{d\phi}{dt}\frac{\delta S_d[\phi]}{\delta\phi}. \tag*{(2)} S d + 1 [ ϕ ] = ∫ M d t ∧ d t d ϕ δ ϕ δ S d [ ϕ ] . (2)
for the additional dimension being parameterized as t t t , a fictitious time parameter. The invariance of the action against how the field is extended into the bulk is ensured by the existence of a topological invariant via the
constraint S d + 1 [ ϕ + δ ϕ ] = S d + 1 [ ϕ ] S_{d+1}[\phi+\delta\phi]=S_{d+1}[\phi] S d + 1 [ ϕ + δ ϕ ] = S d + 1 [ ϕ ] whenever δ ϕ ∣ ∂ M = 0 \delta\phi\big|_{\partial M}=0 δ ϕ ∂ M = 0 . This symmetry can be fixed by a gauge choice that we conveniently choose to be the Langevin equation: the standard follow-up procedure13 then leads to the white-noise field η \eta η with which we identify the final form of the model as the supersymmetric generalization of the Fokker–Planck action. In this paper however, we will work with the non-supersymmetric version.14
The aforementioned Langevin equation reads as
∂ ϕ ( t , x ) ∂ t = − κ δ S E [ ϕ ] δ ϕ ( t , x ) + η ( t , x ) , (3) \frac{\partial\phi(t,x)}{\partial t}=-\kappa\frac{\delta S_E[\phi]}{\delta\phi(t,x)}+\eta(t,x), \tag*{(3)} ∂ t ∂ ϕ ( t , x ) = − κ δ ϕ ( t , x ) δ S E [ ϕ ] + η ( t , x ) , (3)
which is interpreted as evolution of the field ϕ \phi ϕ that quantizes a d d d -dimensional Euclidean field theory with action S E [ ϕ ] S_E[\phi] S E [ ϕ ] . The white noise η \eta η is Gaussian,
⟨ η ( t , x ) η ( t ′ , x ′ ) ⟩ = 2 κ δ ( t − t ′ ) δ d ( x − x ′ ) , (4) \langle\eta(t,x)\eta(t',x')\rangle=2\kappa\,\delta(t-t')\delta^d(x-x'), \tag*{(4)} ⟨ η ( t , x ) η ( t ′ , x ′ )⟩ = 2 κ δ ( t − t ′ ) δ d ( x − x ′ ) , (4)
and it represents the coupling to a “thermal” reservoir.15
On the bulk side, the Fokker–Planck action has the form16
S F P = ∫ d t ∫ d d x [ 1 4 κ ( ∂ ϕ ∂ t + κ δ S E δ ϕ ) 2 − κ 2 δ 2 S E [ ϕ ] δ ϕ 2 ] . (5) S_{FP}=\int dt\int d^dx\left[\frac{1}{4\kappa}\left(\frac{\partial\phi}{\partial t}+\kappa\frac{\delta S_E}{\delta\phi}\right)^2-\frac{\kappa}{2}\frac{\delta^2S_E[\phi]}{\delta\phi^2}\right]. \tag*{(5)} S F P = ∫ d t ∫ d d x [ 4 κ 1 ( ∂ t ∂ ϕ + κ δ ϕ δ S E ) 2 − 2 κ δ ϕ 2 δ 2 S E [ ϕ ] ] . (5)
and the probability distribution P [ ϕ , t ] P[\phi,t] P [ ϕ , t ] of this action does relax to e − S E e^{-S_E} e − S E in the limit t → ∞ t\to\infty t → ∞ : this ensures that equal-time correlators of asymptote to the Euclidean correlation functions of S E S_E S E on the boundary. With a little calculation,17 we can turn the Fokker–Planck equation into a Euclidean Schrödinger equation,
∂ t ψ S = − H F P ψ S (6) \partial_t\psi_S=-H_{FP}\psi_S \tag*{(6)} ∂ t ψ S = − H F P ψ S (6)
with
H F P = ∫ d d x [ − 1 2 δ 2 δ ϕ 2 + 1 8 ( δ S E δ ϕ ) 2 − 1 4 δ 2 S E δ ϕ 2 ] . (7) H_{FP}=\int d^dx\left[-\frac{1}{2}\frac{\delta^2}{\delta\phi^2}+\frac{1}{8}\left(\frac{\delta S_E}{\delta\phi}\right)^2-\frac{1}{4}\frac{\delta^2S_E}{\delta\phi^2}\right]. \tag*{(7)} H F P = ∫ d d x [ − 2 1 δ ϕ 2 δ 2 + 8 1 ( δ ϕ δ S E ) 2 − 4 1 δ ϕ 2 δ 2 S E ] . (7)
and the probability distribution P P P identified as
P = ψ S e − S E / 2 (8) P=\psi_S e^{-S_E/2} \tag*{(8)} P = ψ S e − S E /2 (8)
Subtleties with gauge fields
For a gauge field A i a A_i^a A i a , gauge invariance of S E S_E S E implies D i δ S E / δ A i = 0 D_i\,\delta S_E/\delta A_i=0 D i δ S E / δ A i = 0 , i.e. the drift is orthogonal to the gauge orbits, while the noise is not. In the Parisi–Wu formulation [58 ] , gauge-invariant observables equilibrate, whereas the gauge modes perform an unrestricted random walk and P P P has no normalizable equilibrium along the gauge orbits. We can however damp the gauge modes without affecting gauge-invariant observables by modifying the drift,
∂ t A i = − κ δ S E δ A i + D i v + η i . (9) \partial_t A_i=-\kappa\frac{\delta S_E}{\delta A_i}+D_i v+\eta_i \, . \tag*{(9)} ∂ t A i = − κ δ A i δ S E + D i v + η i . (9)
with an arbitrary function v [ A ] v[A] v [ A ] , for instance v = α − 1 ∂ i A i v=\alpha^{-1}\partial_i A_i v = α − 1 ∂ i A i [59 ] . If we further choose v v v to be time component of ( d + 1 ) (d+1) ( d + 1 ) -dimensional gauge field, this damping can be realized in a bulk gauge covariant way [85 ] ,18 as
F t i ≡ ∂ t A i − D i A t = − κ δ S E δ A i + η i , A t = v . (10) F_{ti}\equiv\partial_t A_i-D_i A_t=-\kappa\frac{\delta S_E}{\delta A_i}+\eta_i \, ,\qquad A_t=v \, . \tag*{(10)} F t i ≡ ∂ t A i − D i A t = − κ δ A i δ S E + η i , A t = v . (10)
This motivation for the choice of v v v will have direct relevance from the holographic perspective because we will identify A t A_t A t as the radial component of the bulk gauge field.
Holographic Wilsonian renormalization group
It is straightforward to observe that the energy scale of a boundary theory corresponds to the radial coordinate of the AdS bulk [1 , 2 , 98 , 99 ] : next step from here is the postulation of a map between the radial cut-off in AdS and a Wilsonian cut-off in the boundary CFT [42 , 43 ] .19 For bulk fields ϕ ( z , x ) \phi(z,x) ϕ ( z , x ) and boundary operators Φ ( x ) \Phi(x) Φ ( x ) , this postulate takes the form
∫ D ϕ e − S b u l k ∣ z ≥ ϵ = ∫ D ϕ ∣ k < δ − 1 e − S b o u n d a r y + ∫ d d x ϕ ϵ ( x ) O ( x ) (11) \left.\int\mathcal{D}\phi\,e^{-S_{\mathrm{bulk}}}\right|_{z\ge\epsilon}
=
\left.\int\mathcal{D}\phi\right|_{k<\delta^{-1}}
e^{-S_{\mathrm{boundary}}+\int d^d x\,\phi_\epsilon(x)\mathcal{O}(x)} \tag*{(11)} ∫ D ϕ e − S bulk z ≥ ϵ = ∫ D ϕ k < δ − 1 e − S boundary + ∫ d d x ϕ ϵ ( x ) O ( x ) (11)
where ϕ ϵ ( x ) ≡ ϕ ( ϵ , x ) \phi_\epsilon(x)\equiv\phi(\epsilon,x) ϕ ϵ ( x ) ≡ ϕ ( ϵ , x ) and where O ( x ) \mathcal{O}(x) O ( x ) form a complete set of local single-trace operators built out of Φ ( x ) \Phi(x) Φ ( x ) .20
The cut-off’s in the map are the content of the postulate: from the bulk perspective, we effectively shift the boundary from z = 0 z=0 z = 0 to z = ϵ z=\epsilon z = ϵ ; from the boundary perspective, we put a UV cut-off 1 / δ ( ϵ ) 1/\delta(\epsilon) 1/ δ ( ϵ ) in the momentum space, which we will take as 1 / ϵ 1/\epsilon 1/ ϵ .21 We can now define the Wilsonian boundary action 22
Ψ U V [ ϕ ϵ ; ϵ ] ≡ e − S B [ ϕ ϵ ; ϵ ] (12) \Psi_{\mathrm{UV}}[\phi_\epsilon;\epsilon]\equiv e^{-S_B[\phi_\epsilon;\epsilon]} \tag*{(12)} Ψ UV [ ϕ ϵ ; ϵ ] ≡ e − S B [ ϕ ϵ ; ϵ ] (12)
which satisfies the radial Schrödinger equation, ∂ ϵ Ψ U V = − H R G Ψ U V \partial_\epsilon\Psi_{\mathrm{UV}}=-H_{\mathrm{RG}}\Psi_{\mathrm{UV}} ∂ ϵ Ψ UV = − H RG Ψ UV for the radial Hamiltonian H R G H_{\mathrm{RG}} H RG [112 ] ; indeed, S B S_B S B obeys the corresponding Hamilton–Jacobi equations in the classical limit [42 ] . This radial-evolution form of the holographic RG flow is precisely why a connection with stochastic quantization was conjectured;23 indeed, one can now propose that [37 , 44 , 48 , 49 , 50 ]
( t , S E , ψ ) ↔ ( ϵ , 2 S I R , Ψ U V ) (13) (t,S_E,\psi)\quad\leftrightarrow\quad(\epsilon,2S_{\mathrm{IR}},\Psi_{\mathrm{UV}}) \tag*{(13)} ( t , S E , ψ ) ↔ ( ϵ , 2 S IR , Ψ UV ) (13)
where S I R S_{\mathrm{IR}} S IR is the on-shell action of the IR region with regular boundary conditions.24
This dictionary with the identification in (8) gives a relation between Wilsonian boundary action of AdS and equilibrium Euclidean action with probability distribution associated to its stochastic evolution in fictitious time . This relation is best understood as a perturbative expansion around the Gaussian fixed point ϕ = 0 \phi=0 ϕ = 0 , which gives the schematic map
D ( n ) ( ϵ ) ∣ ϵ = t = P ( n ) ( t ) − 1 2 G ( n ) . (14) \left.D^{(n)}(\epsilon)\right|_{\epsilon=t}=P^{(n)}(t)-\frac{1}{2}G^{(n)} \, . \tag*{(14)} D ( n ) ( ϵ ) ϵ = t = P ( n ) ( t ) − 2 1 G ( n ) . (14)
for kernels D ( n ) D^{(n)} D ( n ) , P ( n ) P^{(n)} P ( n ) , and G ( n ) G^{(n)} G ( n ) defined via25
S B = ∑ n D ( n ) ϕ n , S E = ∑ n G ( n ) ϕ n , − ln P = ∑ n P ( n ) ϕ n (15) S_B=\sum_n D^{(n)}\phi^n,\qquad S_E=\sum_n G^{(n)}\phi^n,\qquad-\ln P=\sum_n P^{(n)}\phi^n \tag*{(15)} S B = n ∑ D ( n ) ϕ n , S E = n ∑ G ( n ) ϕ n , − ln P = n ∑ P ( n ) ϕ n (15)
Z = ∫ D ϕ ϵ Ψ U V [ ϕ ϵ ; ϵ ] Ψ I R [ ϕ ϵ ; ϵ ] (16) Z=\int\mathcal{D}\phi_\epsilon\,\Psi_{\mathrm{UV}}[\phi_\epsilon;\epsilon]\,\Psi_{\mathrm{IR}}[\phi_\epsilon;\epsilon] \tag*{(16)} Z = ∫ D ϕ ϵ Ψ UV [ ϕ ϵ ; ϵ ] Ψ IR [ ϕ ϵ ; ϵ ] (16)
for Ψ U V \Psi_{\mathrm{UV}} Ψ UV defined as before, and Ψ I R \Psi_{\mathrm{IR}} Ψ IR defined analogously: integration of the bulk path integral in IR region z > ϵ z>\epsilon z > ϵ at fixed ϕ ϵ \phi_\epsilon ϕ ϵ . Ψ I R \Psi_{\mathrm{IR}} Ψ IR satisfies an analogous (but “time-flipped”) radial Schrödinger equation ∂ ϵ Ψ I R = H R G Ψ I R \partial_\epsilon\Psi_{\mathrm{IR}}=H_{\mathrm{RG}}\Psi_{\mathrm{IR}} ∂ ϵ Ψ IR = H RG Ψ IR , and S I R = − log Ψ I R S_{\mathrm{IR}}=-\log\Psi_{\mathrm{IR}} S IR = − log Ψ IR satisfy the corresponding Hamilton–Jacobi equations.
where we are also stripping off a momentum-conserving delta function δ ^ ( ∑ r k r ) ≡ ( 2 π ) d δ ( d ) ( ∑ r k r ) \hat{\delta}(\sum_r k_r) \equiv(2\pi)^d\delta^{(d)}(\sum_r k_r) δ ^ ( ∑ r k r ) ≡ ( 2 π ) d δ ( d ) ( ∑ r k r ) .
The map (14) has been checked for n = 2 n=2 n = 2 in free theories and for n = 3 , 4 n=3,4 n = 3 , 4 in the literature;26 the main output of this paper is essentially the first explicit realization of this relation for an interacting gauge theory, including the gauge sector.
Yang–Mills theory in AdS4
The Yang–Mills action in Euclidean AdS4 _4 4 reads as
S = 1 4 ∫ d 4 x g F μ ν a F a , μ ν (17) S=\frac{1}{4}\int d^4x\sqrt{g}\,F_{\mu\nu}^aF^{a,\mu\nu} \tag*{(17)} S = 4 1 ∫ d 4 x g F μν a F a , μν (17)
for F μ ν a = ∂ μ A ν a − ∂ ν A μ a + g f a b c A μ b A ν c F_{\mu\nu}^a=\partial_\mu A_\nu^a-\partial_\nu A_\mu^a+gf^{abc}A_\mu^bA_\nu^c F μν a = ∂ μ A ν a − ∂ ν A μ a + g f ab c A μ b A ν c , where we are explicitly working in the Poincaré patch d s 2 = ( d z 2 + d x 2 ) / z 2 ds^2=(dz^2+dx^2)/z^2 d s 2 = ( d z 2 + d x 2 ) / z 2 . Classical Weyl invariance rewrites this action as
S = 1 4 ∫ d z d 3 x F μ ν a F μ ν a (18) S=\frac{1}{4}\int dz\,d^3x\,F_{\mu\nu}^aF_{\mu\nu}^a \tag*{(18)} S = 4 1 ∫ d z d 3 x F μν a F μν a (18)
with indices contracted with the flat metric: Yang–Mills theory in AdS4 _4 4 is classically equivalent to Yang–Mills theory on the flat half-space z > 0 z>0 z > 0 , with the conformal boundary at z = 0 z=0 z = 0 .
We will explicitly work in radial/axial gauge in this paper, i.e. A z a = 0 A_z^a=0 A z a = 0 : the action then reads as
S = ∫ 0 ∞ d z ∫ d 3 x [ 1 2 ∂ z A i a ∂ z A i a + 1 2 B i a B i a ] (19) S=\int_0^\infty dz\int d^3x\left[\frac{1}{2}\partial_z A_i^a\partial_z A_i^a+\frac{1}{2}B_i^aB_i^a\right] \tag*{(19)} S = ∫ 0 ∞ d z ∫ d 3 x [ 2 1 ∂ z A i a ∂ z A i a + 2 1 B i a B i a ] (19)
with B i a = 1 2 ϵ i j k F j k a B_i^a=\frac{1}{2}\epsilon_{ijk}F_{jk}^a B i a = 2 1 ϵ ij k F j k a . The radial “time” z z z does not appear explicitly. The classical radial Hamiltonian, i.e. the Legendre transform of (19) , is
H R G = ∫ d 3 x [ 1 2 E i a E i a − 1 2 B i a B i a ] , E i a = ∂ z A i a (20) H_{RG}=\int d^3x\left[\frac{1}{2}E_i^aE_i^a-\frac{1}{2}B_i^aB_i^a\right],\qquad E_i^a=\partial_z A_i^a \tag*{(20)} H R G = ∫ d 3 x [ 2 1 E i a E i a − 2 1 B i a B i a ] , E i a = ∂ z A i a (20)
and the radial Schrödinger operator of § II B is H R G = ∫ d 3 x [ − 1 2 δ 2 δ A i a δ A i a + 1 2 B i a B i a ] H_{RG}=\int d^3x[-\frac{1}{2}\frac{\delta^2}{\delta A_i^a\delta A_i^a}+\frac{1}{2}B_i^aB_i^a] H R G = ∫ d 3 x [ − 2 1 δ A i a δ A i a δ 2 + 2 1 B i a B i a ] . The residual gauge transformations are z z z -independent; since the Dirichlet condition requires the gauge parameter to vanish at z = 0 z=0 z = 0 , only global colour rotations remain and the gauge is completely fixed [42 ] . The A z A_z A z equation of motion, i.e. the Gauss law D i E i = 0 D_iE_i=0 D i E i = 0 , is not imposed in radial gauge; we return to this point in § IV.
As commonly used in the relevant holography literature [24 , 25 , 26 , 36 , 66 , 67 , 113 , 114 , 115 , 116 ] , we Fourier transform along the boundary, A ( z , x ) = ∫ k e i k ⋅ x A ( z , k ) A(z,x)=\int_k e^{ik\cdot x}A(z,k) A ( z , x ) = ∫ k e ik ⋅ x A ( z , k ) with ∫ k ≡ ∫ d 3 k / ( 2 π ) 3 \int_k\equiv\int d^3k/(2\pi)^3 ∫ k ≡ ∫ d 3 k / ( 2 π ) 3 , write k ≡ ∣ k ∣ k\equiv|k| k ≡ ∣ k ∣ , and introduce the transverse and longitudinal projectors
Π i j ⊥ ( k ) = δ i j − k i k j k 2 , Π i j ∥ ( k ) = k i k j k 2 . (21) \Pi_{ij}^{\perp}(k)=\delta_{ij}-\frac{k_i k_j}{k^2},\qquad\Pi_{ij}^{\parallel}(k)=\frac{k_i k_j}{k^2}. \tag*{(21)} Π ij ⊥ ( k ) = δ ij − k 2 k i k j , Π ij ∥ ( k ) = k 2 k i k j . (21)
At quadratic order, 1 2 B 2 \frac{1}{2}B^2 2 1 B 2 becomes 1 2 A ( − k ) ⋅ k 2 Π ⊥ ( k ) ⋅ A ( k ) \frac{1}{2}A(-k)\cdot k^2\Pi^\perp(k)\cdot A(k) 2 1 A ( − k ) ⋅ k 2 Π ⊥ ( k ) ⋅ A ( k ) : transverse modes behave like conformally coupled scalars, while longitudinal modes are massless in the radial direction. With Dirichlet conditions at z = 0 z=0 z = 0 and regularity in the interior, the bulk-to-bulk propagator is [24 , 67 ]
G i j ( z , w ; k ) = Π i j ⊥ ( k ) G k ( z , w ) + Π i j ∥ ( k ) G 0 ( z , w ) (22) G_{ij}(z,w;k)=\Pi_{ij}^{\perp}(k)G_k(z,w)+\Pi_{ij}^{\parallel}(k)G_0(z,w) \tag*{(22)} G ij ( z , w ; k ) = Π ij ⊥ ( k ) G k ( z , w ) + Π ij ∥ ( k ) G 0 ( z , w ) (22)
where
G k ( z , w ) = e − k ∣ z − w ∣ − e − k ( z + w ) 2 k , G 0 ( z , w ) = min ( z , w ) , (23) G_k(z,w)=\frac{e^{-k|z-w|}-e^{-k(z+w)}}{2k},\qquad G_0(z,w)=\min(z,w), \tag*{(23)} G k ( z , w ) = 2 k e − k ∣ z − w ∣ − e − k ( z + w ) , G 0 ( z , w ) = min ( z , w ) , (23)
are the propagator of a (rescaled) conformally coupled scalar and its k → 0 k\to0 k → 0 limit.
Main tools
Functional calculus of AdS4 Yang–Mills
The form of the bulk-to-bulk propagator in (22) and how it decomposes into scalar propagators via (23) motivate the following definition. For a function W ( k ) W(k) W ( k ) of the norm of a momentum, let
K k [ W ] ≡ Π ⊥ ( k ) W ( k ) + Π ∥ ( k ) W ( 0 ) = W ( Ω k ) , (24) \mathcal{K}_k[W]\equiv\Pi^\perp(k)W(k)+\Pi^\parallel(k)W(0)=W(\Omega_k), \tag*{(24)} K k [ W ] ≡ Π ⊥ ( k ) W ( k ) + Π ∥ ( k ) W ( 0 ) = W ( Ω k ) , (24)
with Ω k = k Π ⊥ ( k ) \Omega_k=k\Pi^\perp(k) Ω k = k Π ⊥ ( k ) . In other words, K \mathcal{K} K is the functional calculus of the matrix Ω k \Omega_k Ω k , whose eigenvalues are k k k (transverse) and 0 0 0 (longitudinal).27 In this language (22) reads G = K k [ G k ] G=\mathcal{K}_k[G_k] G = K k [ G k ] . Up to a factor of − i -i − i and the Lorentzian boundary metric, K \mathcal{K} K is the operator
P k m n W ≡ − i η m n W + i k m k n k 2 ( W − W ∣ k → 0 ) (25) \mathcal{P}_k^{mn}W\equiv-i\eta^{mn}W+i\frac{k^m k^n}{k^2}\left(W-W\big|_{k\to0}\right) \tag*{(25)} P k mn W ≡ − i η mn W + i k 2 k m k n ( W − W k → 0 ) (25)
of [66 ] . Since a functional calculus is an algebra homomorphism,
K k [ W 1 W 2 ] = K k [ W 1 ] ⋅ K k [ W 2 ] , (26) \mathcal{K}_k[W_1W_2]=\mathcal{K}_k[W_1]\cdot\mathcal{K}_k[W_2], \tag*{(26)} K k [ W 1 W 2 ] = K k [ W 1 ] ⋅ K k [ W 2 ] , (26)
which is the identity P k m n ( A B ) = i P k m α ( A ) η α β P k β n ( B ) \mathcal{P}_k^{mn}(AB)=i\mathcal{P}_k^{m\alpha}(A)\eta_{\alpha\beta}\mathcal{P}_k^{\beta n}(B) P k mn ( A B ) = i P k m α ( A ) η α β P k β n ( B ) in Euclidean form, which follows from Π ⊥ Π ∥ = 0 \Pi^\perp\Pi^\parallel=0 Π ⊥ Π ∥ = 0 and ( Π ⊥ ) 2 = Π ⊥ (\Pi^\perp)^2=\Pi^\perp ( Π ⊥ ) 2 = Π ⊥ , ( Π ∥ ) 2 = Π ∥ (\Pi^\parallel)^2=\Pi^\parallel ( Π ∥ ) 2 = Π ∥ . K \mathcal{K} K also commutes with radial integrations and sums. We will use (26) repeatedly: any product of propagators or response functions carrying the same momentum is dressed by a single K \mathcal{K} K .
In radial gauge, the cubic and quartic terms in (19) contain only boundary derivatives,
1 2 B a B a ⊃ g f a b c ( ∂ i A j a ) A i b A j c + g 2 4 f a b e f c d e A i a A j b A i c A j d . (27) \frac{1}{2}B^aB^a\supset gf^{abc}(\partial_iA_j^a)A_i^bA_j^c+\frac{g^2}{4}f^{abe}f^{cde}A_i^aA_j^bA_i^cA_j^d. \tag*{(27)} 2 1 B a B a ⊃ g f ab c ( ∂ i A j a ) A i b A j c + 4 g 2 f ab e f c d e A i a A j b A i c A j d . (27)
so in momentum space the vertices are tensors V i j k a b c ( k 1 , k 2 , k 3 ) V^{abc}_{ijk}(k_1,k_2,k_3) V ij k ab c ( k 1 , k 2 , k 3 ) and V i j k ℓ a b c d V^{abcd}_{ijk\ell} V ij k ℓ ab c d that do not depend on z z z ; all radial dependence sits in the propagators [24 ] . This is why the expression for any gluon Witten diagram factorizes into vertex factors and a truncated diagram , which depends on the external momenta only through the sum of their norms at each vertex [66 ] . For instance, the Witten diagrams in Figure 1 read
Figure 1. Example Witten diagrams. Figure taken from [66 ] .
W 4 = V 12 m V 34 n M k 12 m n , W 5 = V 12 m V 345 n M k 12 m n (28) W_4=V^{12m}V^{34n}\mathcal{M}^{mn}_{k_{12}},\qquad W_5=V^{12m}V^{345n}\mathcal{M}^{mn}_{k_{12}} \tag*{(28)} W 4 = V 12 m V 34 n M k 12 mn , W 5 = V 12 m V 345 n M k 12 mn (28)
where28 M k 12 \mathcal{M}_{k_{12}} M k 12 is the truncated diagram with one bulk-to-bulk propagator of momentum k 12 k_{12} k 12 .29 The crucial observation of [66 ] is that M \mathcal{M} M is the corresponding conformally coupled scalar diagram, dressed:
M m n = − i K k 12 m n [ 1 k 1234 k 1212 k 3412 ] (29) \mathcal{M}^{mn}=-i\mathcal{K}^{mn}_{k_{12}}\left[\frac{1}{k_{1234}k_{1212}k_{3412}}\right] \tag*{(29)} M mn = − i K k 12 mn [ k 1234 k 1212 k 3412 1 ] (29)
where the bracket is the truncated exchange diagram of a conformally coupled scalar in AdS4 _4 4 ; more generally, every internal line is dressed by K \mathcal{K} K .30 Since the factorization relies on Weyl invariance, it does not hold in AdSd + 1 ≠ 4 _{d+1\ne4} d + 1 = 4 .
Equilibrium actions of Yang–Mills in AdS4 _4 4
The regular branch
Let S I R [ A ; ϵ ] S_{\mathrm{IR}}[A;\epsilon] S IR [ A ; ϵ ] be the on-shell action of the region z > ϵ z>\epsilon z > ϵ with A ( ϵ ) = A A(\epsilon)=A A ( ϵ ) = A and regular behaviour as z → ∞ z\to\infty z → ∞ . By regular we mean that the action is finite, so that A A A approaches a flat connection.31 Since the Lagrangian in (19) has no explicit z z z -dependence, the shift z → z − ϵ z\to z-\epsilon z → z − ϵ shows that S I R S_{\mathrm{IR}} S IR does not depend on ϵ \epsilon ϵ , i.e. ∂ ϵ S I R = 0 \partial_\epsilon S_{\mathrm{IR}}=0 ∂ ϵ S IR = 0 . Its Hamilton–Jacobi equation, ∂ ϵ S I R = H R G [ A , E = − δ S I R / δ A ] \partial_\epsilon S_{\mathrm{IR} }=\mathcal{H}_{\mathrm{RG}}[A,E=-\delta S_{\mathrm{IR}}/\delta A] ∂ ϵ S IR = H RG [ A , E = − δ S IR / δ A ] , then implies that the radial energy of a regular solution vanishes: with (20) , S I R S_{\mathrm{IR}} S IR obeys the stationarity condition
1 2 ∫ d 3 x ( δ S I R δ A i a ) 2 = 1 2 ∫ d 3 x B i a B i a (30) \frac{1}{2}\int d^3x\left(\frac{\delta S_{\mathrm{IR}}}{\delta A_i^a}\right)^2=\frac{1}{2}\int d^3x\,B_i^aB_i^a \tag*{(30)} 2 1 ∫ d 3 x ( δ A i a δ S IR ) 2 = 2 1 ∫ d 3 x B i a B i a (30)
for all A A A . Thus S I R S_{\mathrm{IR}} S IR is a zero-energy solution of the Hamilton–Jacobi equation, a “superpotential” in the sense of holographic RG flows [101 , 118 , 119 , 120 ] . The regular solutions obey the first-order flow ∂ z A = − δ S I R / δ A \partial_zA=-\delta S_{\mathrm{IR}}/\delta A ∂ z A = − δ S IR / δ A , which is the noiseless Langevin equation of § III C. At tree level, S I R = − log Z C F T [ A ] S_{\mathrm{IR}}=-\log Z_{\mathrm{CFT}}[A] S IR = − log Z CFT [ A ] is the generating functional of connected gluon correlators, i.e. of the Witten diagrams of [66 ] . The equilibrium action of the dictionary (13) is therefore
S E [ A ] = 2 S I R [ A ] = − 2 log Z C F T [ A ] (31) S_E[A]=2S_{\mathrm{IR}}[A]=-2\log Z_{\mathrm{CFT}}[A] \tag*{(31)} S E [ A ] = 2 S IR [ A ] = − 2 log Z CFT [ A ] (31)
and e − S E = Ψ I R 2 e^{-S_E}=\Psi_{\mathrm{IR}}^2 e − S E = Ψ IR 2 is the square of the regular radial wave-function. Explicitly, to quartic order,
G ( 2 ) = K k [ k ] = Ω k (32) G^{(2)}=\mathcal{K}_k[k]=\Omega_k \tag*{(32)} G ( 2 ) = K k [ k ] = Ω k (32)
G ( 3 ) = 1 3 V ∘ K [ 1 k 123 ] (33) G^{(3)}=\frac{1}{3}V\circ\mathcal{K}\left[\frac{1}{k_{123}}\right] \tag*{(33)} G ( 3 ) = 3 1 V ∘ K [ k 123 1 ] (33)
G ( 4 ) = − 1 4 Sym [ V V ∘ K [ 1 k 1234 k 1212 k 3412 ] ] + 1 12 V ( 4 ) ∘ K [ 1 k 1234 ] . (34) \begin{aligned}
G^{(4)}={}&-\frac{1}{4}\operatorname{Sym}\left[V V\circ\mathcal{K}\left[\frac{1}{k_{1234}k_{1212}k_{3412}}\right]\right]\\
&+\frac{1}{12}V^{(4)}\circ\mathcal{K}\left[\frac{1}{k_{1234}}\right]. \tag*{(34)}
\end{aligned} G ( 4 ) = − 4 1 Sym [ V V ∘ K [ k 1234 k 1212 k 3412 1 ] ] + 12 1 V ( 4 ) ∘ K [ k 1234 1 ] . (34)
where K \mathcal{K} K dresses every line: each external leg r r r by K r ≡ K k r \mathcal{K}_r\equiv\mathcal{K}_{k_r} K r ≡ K k r , acting on the argument k r k_r k r , and, in the exchange term, the internal line by K k 12 \mathcal{K}_{k_{12}} K k 12 acting on k 12 k_{12} k 12 . The symbol ∘ \circ ∘ contracts the vertex indices with the dressing operators of the corresponding lines, and Sym \operatorname{Sym} Sym averages over permutations of the external labels (momentum, colour and vector index). At quadratic order e − S E e^{-S_E} e − S E is Gaussian in the transverse modes and flat along the gauge orbits.
The kernels (32) are finite whenever at least one external leg has a transverse component, which covers all the couplings computed in § V.32 Although not relevant for this paper, one does get divergent pieces in the middle steps of other computations; however, they do cancel among each other and lead to a finite final result. For
k i 1 i 2 … i 1 n 1 i 2 i 2 … i 2 n 2 … i m 1 i m 2 … i m n m j 1 j 2 … j p ≡ ∑ a = 1 m ∑ b = 1 n a k i a b + ∑ c = 1 p ∣ k j c ∣ (35) k_{i_1i_2\ldots i_{1n_1}\,i_{2\,i_{2}\ldots i_{2n_2}\ldots i_{m_1}i_{m_2}\ldots i_{mn_m}j_1j_2\ldots j_p}
\equiv\sum_{a=1}^{m}\sum_{b=1}^{n_a}k_{i_{ab}}+\sum_{c=1}^{p}|k_{j_c}|} \tag*{(35)} k i 1 i 2 … i 1 n 1 i 2 i 2 … i 2 n 2 … i m 1 i m 2 … i m n m j 1 j 2 … j p ≡ ∑ a = 1 m ∑ b = 1 n a k i ab + ∑ c = 1 p ∣ k j c ∣ (35)
and
k i 1 i 2 … i n ≡ k i 1 + k i 2 + ⋯ + k i n . (36) k_{i_1i_2\ldots i_n}\equiv k_{i_1}+k_{i_2}+\cdots+k_{i_n}. \tag*{(36)} k i 1 i 2 … i n ≡ k i 1 + k i 2 + ⋯ + k i n . (36)
E.g., k 12345 ≡ ∣ k 1 + k 2 ∣ + ∣ k 3 ∣ + ∣ k 4 + k 5 ∣ k_{12345}\equiv|k_1+k_2|+|k_3|+|k_4+k_5| k 12345 ≡ ∣ k 1 + k 2 ∣ + ∣ k 3 ∣ + ∣ k 4 + k 5 ∣ and k 12 ≡ k 1 + k 2 k_{12}\equiv k_1+k_2 k 12 ≡ k 1 + k 2 . instance, one such computation occurs in star-triangle topology of six-point gluon amplitude;[33 ] a similar cancellation happens between the exchange and contact diagrams of purely-longitudinal four-point Witten diagram: schematically, the second-order magnetic field of a linearized pure gauge obeys q ⋅ B ( 2 ) = 0 \boldsymbol{q}\cdot\boldsymbol{B}^{(2)}=0 q ⋅ B ( 2 ) = 0 , so the exchange term, which involves ∣ q × B ( 2 ) ∣ 2 / q 2 |\boldsymbol{q}\times\boldsymbol{B}^{(2)}|^2/q^2 ∣ q × B ( 2 ) ∣ 2 / q 2 , cancels the contact term ∣ B ( 2 ) ∣ 2 |\boldsymbol{B}^{(2)}|^2 ∣ B ( 2 ) ∣ 2 .\33
The Chern–Simons branch
The stationarity condition (30) has solutions other than S I R S_{\mathrm{IR}} S IR . Since δ S C S / δ A i a = B i a \delta S_{\mathrm{CS}}/\delta A_i^a=B_i^a δ S CS / δ A i a = B i a for the Chern–Simons functional
S C S [ A ] = 1 2 ∫ d 3 x ϵ i j k ( A i a ∂ j A k a + g 3 f a b c A i a A j b A k c ) , (37) S_{\mathrm{CS}}[A]=\frac{1}{2}\int d^3x\,\epsilon_{ijk}\left(A_i^a\partial_j A_k^a+\frac{g}{3}f^{abc}A_i^a A_j^b A_k^c\right), \tag*{(37)} S CS [ A ] = 2 1 ∫ d 3 x ϵ ij k ( A i a ∂ j A k a + 3 g f ab c A i a A j b A k c ) , (37)
± S C S \pm S_{\mathrm{CS}} ± S CS solves (30) exactly. This is the familiar statement that the Chern–Simons state is a zero-energy solution of the Yang–Mills Schrödinger equation [121 ] . At quadratic order the stationary solutions are labelled by a sign for each helicity, S ( 2 ) = 1 2 ∑ h s h k ∣ A h ∣ 2 S^{(2)}=\frac{1}{2}\sum_h s_h k|A_h|^2 S ( 2 ) = 2 1 ∑ h s h k ∣ A h ∣ 2 : S I R S_{\mathrm{IR}} S IR has s ± = + 1 s_{\pm}=+1 s ± = + 1 , while ± S C S \pm S_{\mathrm{CS}} ± S CS has opposite signs for the two helicities. Holographically, this formal solution corresponds to boundary conditions that are regular in the IR for one helicity only. Taking S E = ± 2 S C S S_E=\pm2S_{\mathrm{CS}} S E = ± 2 S CS in the dictionary of (13) , the drift becomes − 1 2 δ S E / δ A i = ∓ B i -\frac{1}{2}\delta S_E/\delta A_i=\mp B_i − 2 1 δ S E / δ A i = ∓ B i . The noiseless Langevin flow is then the (anti-)self-duality equation F t i = ∓ B i F_{ti}=\mp B_i F t i = ∓ B i , i.e. instantons are the gradient flow of the Chern–Simons functional, and the supersymmetric (Parisi–Sourlas) form of the Fokker–Planck action [93 , 95 , 122 ] is Witten’s topological Yang–Mills theory, viewed as supersymmetric quantum mechanics on the space of connections with
residue of the 1 / k 1234 1/k_{1234} 1/ k 1234 pole,
Res = − ∑ ( i j ) ( k ℓ ) V i j ⋅ Π ⊥ ( q ) ⋅ V k ℓ q 2 + V ( 4 ) , (38) \operatorname{Res}=-\sum_{(ij)(k\ell)}\frac{V_{ij}\cdot\Pi^\perp(q)\cdot V_{k\ell}}{q^2}+V^{(4)}, \tag*{(38)} Res = − ( ij ) ( k ℓ ) ∑ q 2 V ij ⋅ Π ⊥ ( q ) ⋅ V k ℓ + V ( 4 ) , (38)
vanishes. Here the sum runs over the three pairings, q = k i + k j q=k_i+k_j q = k i + k j , and V i j V_{ij} V ij is the cubic vertex contracted with the polarizations of legs i i i and j j j , with the internal colour summed over. This is the Ward identity of S E S_E S E .
ϵ 1 ⋯ ϵ 4 ⋅ δ 4 S I R ∣ l o n g = ∑ ( i j ) ( k ℓ ) V i j ⋅ Π ⊥ ( q ) ⋅ V k ℓ ∣ q ∣ 3 , (39) \epsilon_1\cdots\epsilon_4\cdot\left.\delta^4S_{\mathrm{IR}}\right|_{\mathrm{long}}=\sum_{(ij)(k\ell)}\frac{V_{ij}\cdot\Pi^\perp(q)\cdot V_{k\ell}}{|q|^3}, \tag*{(39)} ϵ 1 ⋯ ϵ 4 ⋅ δ 4 S IR long = ( ij ) ( k ℓ ) ∑ ∣ q ∣ 3 V ij ⋅ Π ⊥ ( q ) ⋅ V k ℓ , (39)
is the action 1 2 A ~ ⋅ ∣ q ∣ ⋅ A ~ \frac{1}{2}\widetilde A\cdot|q|\cdot\widetilde A 2 1 A ⋅ ∣ q ∣ ⋅ A of the transverse field A ~ = O ( A 2 ) \widetilde A=\mathcal{O}(A^2) A = O ( A 2 ) that completes a linearized pure gauge into a flat connection. It does not enter any of the transverse couplings below. We have verified these statements with SU(2) and SU(3) colours, in exact arithmetic, in the attached Mathematica file: the vanishing of V V V on three longitudinal legs and of the longitudinal-line numerator, the cancellation (38) , and the convergence of the regulated sum to (39) .
the Chern–Simons functional as Morse function [123 ] . Since e − S E e^{-S_E} e − S E is then not normalizable, it has no stochastic equilibrium. Moreover the linearized flow ∂ t A = ∓ ∇ × A \partial_t A=\mp\nabla\times A ∂ t A = ∓ ∇ × A amplifies one helicity at all momenta, so the Langevin process is not well posed; the statements of this subsection are statements about the zero-energy state e ∓ S C S e^{\mp S_{\mathrm{CS}}} e ∓ S CS and about the noiseless flow. In the rest of the paper we take the regular branch, hence S E = 2 S I R S_E=2S_{\mathrm{IR}} S E = 2 S IR .
Tree-level correspondence to all orders
Statement and its proof
The relation (14) has been verified explicitly one multi-trace coupling at a time, and derived for general n n n , for scalar theories without explicit radial dependence, in section 4.1 of [48 ] by completing the square in the bulk action. We restate that argument in a form that applies verbatim to gauge fields and makes its probabilistic content explicit.
Proposition 1. {35 } Consider a bulk theory whose action (for gauge fields in radial gauge) is S = ∫ 0 ∞ d z ∫ d d x [ 1 2 ( ∂ z Φ ) 2 + V [ Φ ] ] S=\int_0^\infty dz\int d^d x\left[\frac{1}{2}(\partial_z\Phi)^2+\mathcal{V}[\Phi]\right] S = ∫ 0 ∞ d z ∫ d d x [ 2 1 ( ∂ z Φ ) 2 + V [ Φ ] ] with vanishing boundary condition at z z z , and Dirichlet conditions Φ ( 0 ) = 0 \Phi(0)=0 Φ ( 0 ) = 0 . Let S E = 2 S I R S_E=2S_{\mathrm{IR}} S E = 2 S IR , and let P [ Φ , t ] P[\Phi,t] P [ Φ , t ] be the distribution at time t t t of the Langevin process ∂ t Φ = − 1 2 δ S E / δ Φ + η \partial_t\Phi=-\frac{1}{2}\delta S_E/\delta\Phi+\eta ∂ t Φ = − 2 1 δ S E / δ Φ + η , with ⟨ η ⟩ = δ \langle\eta\rangle=\delta ⟨ η ⟩ = δ and Φ ( 0 ) = 0 \Phi(0)=0 Φ ( 0 ) = 0 . Then, at tree level and up to a Φ \Phi Φ -independent constant c c c ,
− log P [ Φ , t ] = S B [ Φ ; t ] + 1 2 S E [ Φ ] + c . (40) -\log P[\Phi,t]=S_B[\Phi;t]+\frac{1}{2}S_E[\Phi]+c. \tag*{(40)} − log P [ Φ , t ] = S B [ Φ ; t ] + 2 1 S E [ Φ ] + c . (40) In particular, (14) holds for all n n n .
Proof. As in (30) , the z z z -independence of V \mathcal{V} V implies 1 8 ∫ ( δ S E / δ Φ ) 2 = ∫ V [ Φ ] \frac{1}{8}\int(\delta S_E/\delta\Phi)^2=\int\mathcal{V}[\Phi] 8 1 ∫ ( δ S E / δ Φ ) 2 = ∫ V [ Φ ] . The Onsager–Machlup representation of P P P reads
P [ Φ , t ] = ∫ Φ ( 0 ) = 0 Φ ( t ) = Φ D Φ exp { − 1 2 ∫ 0 t d t ′ ∫ d d x ( ∂ t ′ Φ + 1 2 δ S E δ Φ ) 2 } . (41) P[\Phi,t]=\int_{\Phi(0)=0}^{\Phi(t)=\Phi}\mathcal{D}\Phi\exp\left\{-\frac{1}{2}\int_0^t dt'\int d^d x\left(\partial_{t'}\Phi+\frac{1}{2}\frac{\delta S_E}{\delta\Phi}\right)^2\right\}. \tag*{(41)} P [ Φ , t ] = ∫ Φ ( 0 ) = 0 Φ ( t ) = Φ D Φ exp { − 2 1 ∫ 0 t d t ′ ∫ d d x ( ∂ t ′ Φ + 2 1 δ Φ δ S E ) 2 } . (41)
which is same as (5) upto a term irrelevant at the tree level.{36 } Expanding the square, the cross term integrates to 1 2 S E [ Φ ] − 1 2 S E [ 0 ] \frac{1}{2}S_E[\Phi]-\frac{1}{2}S_E[0] 2 1 S E [ Φ ] − 2 1 S E [ 0 ] , and the remaining terms give ∫ 0 t d t ′ ∫ d d x [ 1 2 ( ∂ t ′ Φ ) 2 + V [ Φ ] ] \int_0^t dt'\int d^d x\left[\frac{1}{2}(\partial_{t'}\Phi)^2+\mathcal{V}[\Phi]\right] ∫ 0 t d t ′ ∫ d d x [ 2 1 ( ∂ t ′ Φ ) 2 + V [ Φ ] ] , i.e. the bulk action of the region 0 < z < t 0<z<t 0 < z < t . Hence P e S E / 2 = ∫ Φ ( 0 ) = 0 Φ ( t ) = Φ D Φ e − S 0 < z < t = e − S B [ Φ ; t ] P\,e^{S_E/2}=\int_{\Phi(0)=0}^{\Phi(t)=\Phi}\mathcal{D}\Phi\,e^{-S_{0<z<t}}=e^{-S_B[\Phi;t]} P e S E /2 = ∫ Φ ( 0 ) = 0 Φ ( t ) = Φ D Φ e − S 0 < z < t = e − S B [ Φ ; t ] at tree level. □ \square □
Interpretation
Exponentiating (40) and using (12) and (13) gives P [ Φ , t ] = e − c Ψ U V [ Φ ; t ] Ψ I R [ Φ ] P[\Phi,t] = e^{-c}\Psi_{\mathrm{UV}}[\Phi;t]\Psi_{\mathrm{IR}}[\Phi] P [ Φ , t ] = e − c Ψ UV [ Φ ; t ] Ψ IR [ Φ ] . Normalizing P P P fixes e c = ∫ D Φ Ψ U V Ψ I R = Z e^{c} = \int D\Phi\,\Psi_{\mathrm{UV}}\Psi_{\mathrm{IR}} = Z e c = ∫ D Φ Ψ UV Ψ IR = Z , the bulk partition function in the Heemskerk–Polchinski split (16) . Hence
P [ Φ , t ] = Ψ U V [ Φ ; t ] Ψ I R [ Φ ] Z (42) P[\Phi,t] = \frac{\Psi_{\mathrm{UV}}[\Phi;t]\Psi_{\mathrm{IR}}[\Phi]}{Z} \tag*{(42)} P [ Φ , t ] = Z Ψ UV [ Φ ; t ] Ψ IR [ Φ ] (42)
The Langevin process samples the bulk field on the slice z = t z=t z = t with the Heemskerk–Polchinski weight. In other words, it is the ground-state (Doob) transform of the radial evolution by Ψ I R \Psi_{\mathrm{IR}} Ψ IR [86 , 124 , 125 , 126 ] : since H R G Ψ I R = 0 H_{\mathrm{RG}}\Psi_{\mathrm{IR}}=0 H RG Ψ IR = 0 , the drift is ∇ log Ψ I R = − 1 2 ∇ S E \nabla\log\Psi_{\mathrm{IR}}=-\frac{1}{2}\nabla S_{\mathrm{E}} ∇ log Ψ IR = − 2 1 ∇ S E . In Hamiltonian language, ψ S = P e S E / 2 = Ψ U V / Z \psi_{\mathrm{S}}=P e^{S_{\mathrm{E}}/2}=\Psi_{\mathrm{UV}}/Z ψ S = P e S E /2 = Ψ UV / Z , so ψ S \psi_{\mathrm{S}} ψ S and Ψ U V \Psi_{\mathrm{UV}} Ψ UV obey the same equation with the same initial condition, i.e. H F P = H R G H_{\mathrm{FP}}=H_{\mathrm{RG}} H FP = H RG at the classical level [37 , 48 ] .
At the non-perturbative level, the Doob-transform relation is an identity rather than a conjecture. If Ψ I R \Psi_{\mathrm{IR}} Ψ IR is the exact positive ground state of H R G H_{\mathrm{RG}} H RG , with energy E 0 E_0 E 0 , and S E ≡ − 2 log Ψ I R S_{\mathrm{E}}\equiv-2\log\Psi_{\mathrm{IR}} S E ≡ − 2 log Ψ IR , then H F P = H R G − E 0 H_{\mathrm{FP}}=H_{\mathrm{RG}}-E_0 H FP = H RG − E 0 holds identically, with H F P H_{\mathrm{FP}} H FP as in (7) .37 A loop-level dictionary is therefore a matter of defining Ψ I R \Psi_{\mathrm{IR}} Ψ IR and E 0 E_0 E 0 (§ VII); what goes wrong if S E S_{\mathrm{E}} S E is instead kept at its tree-level value is the subject of remark (d) below.
(a) Scope. The proposition applies to conformally coupled scalars in A d S d + 1 \mathrm{AdS}_{d+1} AdS d + 1 with the Weyl-invariant interaction ϕ 2 ( d + 1 ) / ( d − 1 ) \phi^{2(d+1)/(d-1)} ϕ 2 ( d + 1 ) / ( d − 1 ) after the rescaling ϕ = z ( d − 1 ) / 2 f \phi=z^{(d-1)/2}f ϕ = z ( d − 1 ) /2 f , and to Maxwell and Yang–Mills theory in A d S 4 \mathrm{AdS}_4 AdS 4 . The ϕ 3 / A d S 6 \phi^3/\mathrm{AdS}_6 ϕ 3 / AdS 6 and ϕ 4 / A d S 4 \phi^4/\mathrm{AdS}_4 ϕ 4 / AdS 4 checks of [48 , 49 ] are instances. For fields not conformally coupled, S I R S_{\mathrm{IR}} S IR depends on ϵ \epsilon ϵ ; see remark (c).
(b) Initial and boundary conditions. Dirichlet conditions at z = 0 z=0 z = 0 correspond to Φ ( 0 ) = 0 \Phi(0)=0 Φ ( 0 ) = 0 , i.e. P [ Φ , 0 ] = δ [ Φ ] P[\Phi,0]=\delta[\Phi] P [ Φ , 0 ] = δ [ Φ ] . This is the setting of the Dirichlet analyses [48 , 49 ] , and it implies that all multi-trace couplings with n ≥ 3 n\ge3 n ≥ 3 vanish at the boundary, D ( n ) ( ϵ ) → 0 D^{(n)}(\epsilon)\to0 D ( n ) ( ϵ ) → 0 as ϵ → 0 \epsilon\to0 ϵ → 0 ; see § V for n = 3 , 4 n=3,4 n = 3 , 4 .38 Nothing in (42) singles out this initial state. A general UV boundary condition Ψ U V [ Φ ; 0 ] \Psi_{\mathrm{UV}}[\Phi;0] Ψ UV [ Φ ; 0 ] , such as the Neumann and mixed conditions available for gauge fields in A d S 4 \mathrm{AdS}_4 AdS 4 [127 , 128 ] , corresponds to the initial distribution P [ Φ , 0 ] ∝ Ψ U V [ Φ ; 0 ] Ψ I R [ Φ ] P[\Phi,0]\propto\Psi_{\mathrm{UV}}[\Phi;0]\Psi_{\mathrm{IR}}[\Phi] P [ Φ , 0 ] ∝ Ψ UV [ Φ ; 0 ] Ψ IR [ Φ ] , provided this product is normalizable; Dirichlet conditions are the special case Ψ U V [ Φ ; 0 ] = δ [ Φ ] \Psi_{\mathrm{UV}}[\Phi;0]=\delta[\Phi] Ψ UV [ Φ ; 0 ] = δ [ Φ ] . Massive scalars with multi-trace boundary deformations were analyzed in [50 , 51 ] by matching lower multi-trace couplings to the stochastic data at the initial time. Relating those matching conditions to this initial distribution is left to future work [129 ] .
(c) Radial inhomogeneity. For bulk theories with explicit radial dependence, Ψ I R \Psi_{\mathrm{IR}} Ψ IR depends on the cutoff. Then, by the Doob-transform argument of § 1.3 C 2, P = Ψ U V Ψ I R / Z P=\Psi_{\mathrm{UV}}\Psi_{\mathrm{IR}}/Z P = Ψ UV Ψ IR / Z still solves a Fokker–Planck equation, now with the time-dependent drift ∇ log Ψ I R [ Φ ; t ] \nabla\log\Psi_{\mathrm{IR}}[\Phi;t] ∇ log Ψ IR [ Φ ; t ] .39 Nevertheless, − log P = S B [ Φ ; t ] + 1 2 S E [ Φ ; t ] + c -\log P=S_{\mathrm{B}}[\Phi;t]+\frac{1}{2}S_{\mathrm{E}}[\Phi;t]+c − log P = S B [ Φ ; t ] + 2 1 S E [ Φ ; t ] + c still holds, with S E [ Φ ; t ] = 2 S I R [ Φ ; t ] S_{\mathrm{E}}[\Phi;t]=2S_{\mathrm{IR}}[\Phi;t] S E [ Φ ; t ] = 2 S IR [ Φ ; t ] , and so does D ( n ) = 1 2 G ( n ) ( t ) D^{(n)}=\frac{1}{2}G^{(n)}(t) D ( n ) = 2 1 G ( n ) ( t ) . Therefore, radial homogeneity is not a condition for the existence of the dictionary; it is the condition under which the drift is time independent, so that the Langevin process is the stochastic quantization of a fixed action.
(d) Beyond tree level. The identity H F P = H R G − E 0 H_{\mathrm{FP}}=H_{\mathrm{RG}}-E_0 H FP = H RG − E 0 of § 1.3 C 2 needs S E S_{\mathrm{E}} S E to be the exact − 2 log Ψ I R -2\log\Psi_{\mathrm{IR}} − 2 log Ψ IR . Suppose instead that S E S_{\mathrm{E}} S E is kept at its tree-level value S E ( 0 ) = 2 S I R S_{\mathrm{E}}^{(0)}=2S_{\mathrm{IR}} S E ( 0 ) = 2 S IR . Stationarity gives 1 8 ( δ S E / δ Φ ) 2 = V \frac{1}{8}(\delta S_{\mathrm{E}}/\delta\Phi)^2=\mathcal{V} 8 1 ( δ S E / δ Φ ) 2 = V exactly, and hence
H F P = H R G − 1 4 ∫ d d x δ 2 S E ( 0 ) δ Φ ( x ) δ Φ ( x ) (43) H_{\mathrm{FP}}=H_{\mathrm{RG}}-\frac{1}{4}\int d^d x\,\frac{\delta^2 S_{\mathrm{E}}^{(0)}}{\delta\Phi(x)\delta\Phi(x)} \tag*{(43)} H FP = H RG − 4 1 ∫ d d x δ Φ ( x ) δ Φ ( x ) δ 2 S E ( 0 ) (43)
For free fields the last term is a field-independent constant, and the dictionary holds to all orders. For interacting fields its field-dependent part acts as an extra one-loop potential, so the stochastic quantization of the tree-level action does not reproduce the one-loop holographic flow. For Yang–Mills theory the cubic term of S E ( 0 ) S_{\mathrm{E}}^{(0)} S E ( 0 ) does not contribute, since f a a c = 0 f^{aac}=0 f aa c = 0 . The mismatch first appears in the double-trace coupling at O ( g 2 ) \mathcal{O}(g^2) O ( g 2 ) , through the trace of the quartic kernel G ( 4 ) G^{(4)} G ( 4 ) over one pair of legs. The dictionary is restored by the one-loop correction S E ( 1 ) S_{\mathrm{E}}^{(1)} S E ( 1 ) of − 2 log Ψ I R -2\log\Psi_{\mathrm{IR}} − 2 log Ψ IR . Expanding H F P = H R G − E 0 H_{\mathrm{FP}}=H_{\mathrm{RG}}-E_0 H FP = H RG − E 0 shows that it obeys the transport equation40
∫ d d x δ S ( 0 ) δ Φ δ S ( 1 ) δ Φ = ∫ d d x δ 2 S ( 0 ) δ Φ ( x ) δ Φ ( x ) − 4 E 0 . (44) \int d^d x\,\frac{\delta S^{(0)}}{\delta\Phi}\frac{\delta S^{(1)}}{\delta\Phi}=\int d^d x\,\frac{\delta^2 S^{(0)}}{\delta\Phi(x)\,\delta\Phi(x)}-4E_0 . \tag*{(44)} ∫ d d x δ Φ δ S ( 0 ) δ Φ δ S ( 1 ) = ∫ d d x δ Φ ( x ) δ Φ ( x ) δ 2 S ( 0 ) − 4 E 0 . (44)
A one-loop comparison therefore tests the computation of S E ( 1 ) S_E^{(1)} S E ( 1 ) , i.e. of the one-loop ground state, rather than the dictionary itself. For Yang–Mills theory in AdS4 _4 4 , the trace anomaly makes S E ( 1 ) S_E^{(1)} S E ( 1 ) depend on the cutoff;41 however, the transport equation (44) actually makes this a tree-level calculation: we leave details of this computation and its consequences to future work [129 ] .
IV The gauge sector
A Radial gauge is Parisi–Wu
We now combine the tools we have discussed in the previous section, § III. At quadratic order G ( 2 ) = Ω k G^{(2)}=\Omega_k G ( 2 ) = Ω k , so the free Langevin equation is
∂ t A i ( t , k ) = − k Π i j ⊥ ( k ) A j ( t , k ) + η i ( t , k ) , (45) \partial_t A_i(t,\mathbf{k})=-k\Pi_{ij}^{\perp}(\mathbf{k})A_j(t,\mathbf{k})+\eta_i(t,\mathbf{k}) , \tag*{(45)} ∂ t A i ( t , k ) = − k Π ij ⊥ ( k ) A j ( t , k ) + η i ( t , k ) , (45)
with noise in all three components. Writing the solution with A ( 0 ) = 0 A(0)=0 A ( 0 ) = 0 as A ( t ) = ∫ 0 t d s R ( t − s ) η ( s ) A(t)=\int_0^t ds\,R(t-s)\eta(s) A ( t ) = ∫ 0 t d s R ( t − s ) η ( s ) defines the response function R R R , the retarded Green’s function of (41), ( ∂ τ + Ω k ) R = δ ( τ ) (\partial_\tau+\Omega_k)R=\delta(\tau) ( ∂ τ + Ω k ) R = δ ( τ ) ; the correlator is then ∫ 0 min ( t , t ′ ) d s R ( t − s ) R ( t ′ − s ) T \int_0^{\min(t,t')}ds\,R(t-s)R(t'-s)^T ∫ 0 m i n ( t , t ′ ) d s R ( t − s ) R ( t ′ − s ) T . By the functional calculus of § III A, the response function is
R i j ( τ ; k ) = K k [ e − k τ ] i j θ ( τ ) , (46) R_{ij}(\tau;k)=\mathcal{K}_k[e^{-k\tau}]_{ij}\,\theta(\tau) , \tag*{(46)} R ij ( τ ; k ) = K k [ e − k τ ] ij θ ( τ ) , (46)
for the Heaviside-theta distribution θ \theta θ , and the correlator reads
⟨ A i ( t , k ) A j ( t ′ , k ′ ) ⟩ = δ ^ ( k + k ′ ) K k [ G k ( t , t ′ ) ] i j , (47) \langle A_i(t,\mathbf{k})A_j(t',\mathbf{k}')\rangle=\hat{\delta}(\mathbf{k}+\mathbf{k}')\,\mathcal{K}_k[G_k(t,t')]_{ij} , \tag*{(47)} ⟨ A i ( t , k ) A j ( t ′ , k ′ )⟩ = δ ^ ( k + k ′ ) K k [ G k ( t , t ′ ) ] ij , (47)
i.e. Π ⊥ G k ( t , t ′ ) + Π ∥ min ( t , t ′ ) \Pi^\perp G_k(t,t')+\Pi^\parallel\min(t,t') Π ⊥ G k ( t , t ′ ) + Π ∥ min ( t , t ′ ) . Transverse modes perform Ornstein–Uhlenbeck processes, while longitudinal modes perform Brownian motion with covariance min ( t , t ′ ) = G 0 ( t , t ′ ) \min(t,t')=\mathcal{G}_0(t,t') min ( t , t ′ ) = G 0 ( t , t ′ ) . Comparing with (22) , the stochastic propagator coincides with the bulk-to-bulk propagator in radial gauge, including its longitudinal part. In other words, the term W − W ∣ k → 0 W-W|_{k\to0} W − W ∣ k → 0 of (25) is the random walk of the gauge modes in Parisi–Wu stochastic quantization. The same structure appears in the double-trace couplings. From the UV region one finds
D ( 2 ) ( ϵ ) = 1 2 K k [ k coth k ϵ ] = 1 2 [ Π ⊥ k coth k ϵ + Π ∥ ϵ ] , (48) D^{(2)}(\epsilon)=\frac{1}{2}\mathcal{K}_k[k\coth k\epsilon]=\frac{1}{2}\left[\Pi^\perp k\coth k\epsilon+\frac{\Pi^\parallel}{\epsilon}\right], \tag*{(48)} D ( 2 ) ( ϵ ) = 2 1 K k [ k coth k ϵ ] = 2 1 [ Π ⊥ k coth k ϵ + ϵ Π ∥ ] , (48)
while the Langevin process gives P ( 2 ) ( t ) = 1 2 ⟨ A A ⟩ − 1 = K k [ k / ( 1 − e − 2 k t ) ] P^{(2)}(t)=\frac{1}{2}\langle AA\rangle^{-1}=\mathcal{K}_k[k/(1-e^{-2kt})] P ( 2 ) ( t ) = 2 1 ⟨ AA ⟩ − 1 = K k [ k / ( 1 − e − 2 k t )] . Hence D ( 2 ) = P ( 2 ) − 1 2 G ( 2 ) D^{(2)}=P^{(2)}-\frac{1}{2}G^{(2)} D ( 2 ) = P ( 2 ) − 2 1 G ( 2 ) holds separately in both the transverse and longitudinal coupling. The longitudinal coupling 1 / ( 2 ϵ ) 1/(2\epsilon) 1/ ( 2 ϵ ) is half the inverse variance of a Brownian path, and it is absent from the gauge-invariant equilibrium distribution.
B The radial component of the gauge field is Zwanziger's gauge-fixing field
Now we relax the gauge condition. With A z ≠ 0 A_z\ne0 A z = 0 the radial derivative is replaced by F z i = ∂ z A i − D i A z F_{zi}=\partial_zA_i-D_iA_z F z i = ∂ z A i − D i A z , and we consider the Langevin equation in the Baulieu–Zwanziger form (10), F i = − 1 2 δ S E / δ A i + η i F_i=-\frac{1}{2}\delta S_E/\delta A_i+\eta_i F i = − 2 1 δ S E / δ A i + η i , with t = z t=z t = z and A t = A z A_t=A_z A t = A z . The Onsager–Machlup action becomes
1 2 ∫ ( F t i + 1 2 δ S E δ A i ) 2 = ∫ [ 1 2 F t i F t i + 1 8 ( δ S E δ A i ) 2 ] + 1 2 ∫ F t i δ S E δ A i . (49) \frac{1}{2}\int\left(F_{ti}+\frac{1}{2}\frac{\delta S_E}{\delta A_i}\right)^2=\int\left[\frac{1}{2}F_{ti}F_{ti}+\frac{1}{8}\left(\frac{\delta S_E}{\delta A_i}\right)^2\right]
+\frac{1}{2}\int F_{ti}\frac{\delta S_E}{\delta A_i}. \tag*{(49)} 2 1 ∫ ( F t i + 2 1 δ A i δ S E ) 2 = ∫ [ 2 1 F t i F t i + 8 1 ( δ A i δ S E ) 2 ] + 2 1 ∫ F t i δ A i δ S E . (49)
In the last term, ∫ ∂ t A i δ S E / δ A i = d S E / d t \int\partial_tA_i\,\delta S_E/\delta A_i=dS_E/dt ∫ ∂ t A i δ S E / δ A i = d S E / d t as before, while ∫ ( D i A t ) a δ S E / δ A i a = − ∫ F i a ( D i δ S E / δ A i ) a = 0 \int(D_iA_t)^a\delta S_E/\delta A_i^a=-\int F_i^a(D_i\delta S_E/\delta A_i)^a=0 ∫ ( D i A t ) a δ S E / δ A i a = − ∫ F i a ( D i δ S E / δ A i ) a = 0 by gauge invariance of S E S_E S E . Hence the proof of Proposition 1 goes through for any A t A_t A t , and the Onsager–Machlup action is the bulk Yang–Mills action 1 4 ∫ F μ ν F μ ν \frac{1}{4}\int F_{\mu\nu}F_{\mu\nu} 4 1 ∫ F μν F μν with A z = A t A_z=A_t A z = A t . This gives the dictionary
A z ↔ ν , A_z\ \leftrightarrow\ \nu, A z ↔ ν ,
bulk gauge condition \ \leftrightarrow\ stochastic gauge fixing , {#eq:45}
under which bulk four-dimensional gauge invariance is the ( d + 1 ) (d+1) ( d + 1 ) -dimensional gauge invariance of [85 ] , radial gauge A z = 0 A_z=0 A z = 0 is Parisi–Wu, and the “parabolic” gauge A z = α − 1 ∂ i A i A_z=\alpha^{-1}\partial_i A_i A z = α − 1 ∂ i A i , with α > 0 \alpha>0 α > 0 , is Zwanziger’s gauge.[44 ] In the latter the longitudinal modes acquire a restoring drift and become Ornstein–Uhlenbeck processes with rate k 2 / α k^2/\alpha k 2 / α , so that the equilibrium distribution is normalizable; we work this out in § IV D.
The dictionary has, formally, a third entry. If A z A_z A z is integrated over rather than fixed, its equation of motion, the Gauss law D i F z i = 0 D_iF_{zi}=0 D i F z i = 0 , reads D i η i = 0 D_i\eta_i=0 D i η i = 0 on the Langevin side.[45 ] In radial gauge the Gauss law is not imposed, and correspondingly neither S B S_B S B nor P [ A , t ] P[A,t] P [ A , t ] is gauge invariant at finite t t t ; only gauge-invariant data equilibrate. With A z A_z A z integrated over, the holographic Wilsonian action is gauge invariant, and we compare the two prescriptions at quartic order in § 2.2 E. This extends to the stochastic side, and to the interacting non-abelian theory, the treatment of the gauge sector of [42 , 43 ] , which was left out of the U ( 1 ) U(1) U ( 1 ) analysis of [37 ] .
Faddeev–Popov determinant and Gribov copies
The classical argument of § IV B survives in the path integral because the gauge-fixing measure is trivial, as in the five-dimensional formulation of [85 ] , whose arguments we adapt to the slab with a Dirichlet condition. In the bulk path integral, the parabolic gauge comes with the Faddeev–Popov operator M F P = D z − α − 1 ∂ i D i ≡ ∂ z + K M_{\mathrm{FP}}=D_z-\alpha^{-1}\partial_iD_i\equiv\partial_z+K M FP = D z − α − 1 ∂ i D i ≡ ∂ z + K , with K = g ad ( A z ) − α − 1 ∂ i D i K=g\,\operatorname{ad}(A_z)-\alpha^{-1}\partial_iD_i K = g ad ( A z ) − α − 1 ∂ i D i . It acts on gauge parameters with ω ( 0 ) = 0 \omega(0)=0 ω ( 0 ) = 0 , which are the transformations that preserve the Dirichlet condition. These are not required to vanish at z = ϵ z=\epsilon z = ϵ , so they act on the cutoff data, consistent with the fact that S B S_B S B is not gauge invariant (§ 2.2 E). Since M F P M_{\mathrm{FP}} M FP is first order in z z z with an initial condition, it is a Volterra operator, and
det M F P = exp ( θ ( 0 ) ∫ d z Tr K ) . \det M_{\mathrm{FP}}=\exp\left(\theta(0)\int dz\,\operatorname{Tr}K\right). det M FP = exp ( θ ( 0 ) ∫ d z Tr K ) .
Moreover, Tr K \operatorname{Tr}K Tr K is field independent, because Tr ad ( X ) = f a a c X c = 0 \operatorname{Tr}\operatorname{ad}(X)=f^{aac}X^c=0 Tr ad ( X ) = f aa c X c = 0 for both X = A z X=A_z X = A z and the g ∂ i ad ( A i ) g\,\partial_i\operatorname{ad}(A_i) g ∂ i ad ( A i ) part of ∂ i D i \partial_iD_i ∂ i D i . The determinant is therefore a (formally divergent) constant, and no ghosts are needed, in agreement with stochastic gauge fixing, where no determinant arises in the first place [59 , 60 ] .
In perturbation theory, the gauge condition is reached by solving a forward parabolic equation for ω \omega ω from ω ( 0 ) = 0 \omega(0)=0 ω ( 0 ) = 0 ; this is well posed for α > 0 \alpha>0 α > 0 . Its solution is unique, order by order in g g g , so there are no Gribov copies [59 , 85 ] . As in radial gauge, the condition together with ω ( 0 ) = 0 \omega(0)=0 ω ( 0 ) = 0 fixes ω \omega ω completely, in particular ω ( ϵ ) \omega(\epsilon) ω ( ϵ ) , so there is no residual gauge
freedom. Beyond perturbation theory, the question is existence rather than uniqueness: smooth solutions of the nonlinear flow for finite gauge transformations are unique whenever they exist, but the flow must exist for all z z z for every configuration to be brought to the gauge. For the group-valued gauge transformation U ( z , x ) U(z,x) U ( z , x ) its principal part, U − 1 ∂ z U = α − 1 ∂ i ( U − 1 ∂ i U ) U^{-1}\partial_zU=\alpha^{-1}\partial_i(U^{-1}\partial_iU) U − 1 ∂ z U = α − 1 ∂ i ( U − 1 ∂ i U ) , is the harmonic-map heat flow into the gauge group, which can develop finite-time singularities in three dimensions [135 ] . We do not address this question, which plays no role at tree level. The Langevin Jacobian is likewise field independent up to the term − 1 4 δ 2 S E / δ A 2 -\frac{1}{4}\delta^2S_E/\delta A^2 − 4 1 δ 2 S E / δ A 2 of (7) , which matters only at loop level (§ VII).
Zwanziger's gauge at quadratic and cubic order
At quadratic order the drift in Zwanziger’s gauge is
Ω k ( α ) = k Π ⊥ + α − 1 k Π ∥ , \Omega_k^{(\alpha)}=k\Pi^\perp+\alpha^{-1}k\Pi^\parallel, Ω k ( α ) = k Π ⊥ + α − 1 k Π ∥ ,
and the dressing operator becomes the functional calculus of Ω ( α ) \Omega^{(\alpha)} Ω ( α ) ,
K k ( α ) [ W ] = Π ⊥ ( k ) W ( k ) + Π ∥ ( k ) W ( k 2 / α ) . (50) \mathcal{K}_k^{(\alpha)}[W]=\Pi^\perp(k)W(k)+\Pi^\parallel(k)W(k^2/\alpha). \tag*{(50)} K k ( α ) [ W ] = Π ⊥ ( k ) W ( k ) + Π ∥ ( k ) W ( k 2 / α ) . (50)
This interpolates between the radial-gauge operator of [66 ] (α → ∞ \alpha\to\infty α → ∞ ) and a gauge in which the longitudinal modes are frozen (α → 0 \alpha\to0 α → 0 ). The equilibrium distribution is e − S E − α − 1 ∫ ( ∂ i A i ) 2 e^{-S_E-\alpha^{-1}\int(\partial_iA_i)^2} e − S E − α − 1 ∫ ( ∂ i A i ) 2 at quadratic order. In the bulk, the longitudinal part of 1 2 F z i F z i \frac{1}{2}F_{zi}F_{zi} 2 1 F z i F z i becomes 1 2 ( ∂ z + γ ) A ∥ ∥ 2 \frac{1}{2}(\partial_z+\gamma)A^\parallel\Vert^2 2 1 ( ∂ z + γ ) A ∥ ∥ 2 with γ = k 2 / α \gamma=k^2/\alpha γ = k 2 / α . Its cross term is a total derivative, which contributes a boundary term to S B S_B S B :
D α ( 2 ) ( ϵ ) = 1 2 K k ( α ) [ k coth k ] + k 2 2 α Π ∥ , D_\alpha^{(2)}(\epsilon)=\frac{1}{2}\mathcal{K}_k^{(\alpha)}[k\coth k]+\frac{k^2}{2\alpha}\Pi^\parallel, D α ( 2 ) ( ϵ ) = 2 1 K k ( α ) [ k coth k ] + 2 α k 2 Π ∥ ,
P α ( 2 ) ( t ) = K k ( α ) [ k 1 − e − 2 k t ] . (51) P_\alpha^{(2)}(t)=\mathcal{K}_k^{(\alpha)}\left[\frac{k}{1-e^{-2kt}}\right]. \tag*{(51)} P α ( 2 ) ( t ) = K k ( α ) [ 1 − e − 2 k t k ] . (51)
These again satisfy D α ( 2 ) = P α ( 2 ) − 1 2 G ( 2 ) D_\alpha^{(2)}=P_\alpha^{(2)}-\frac{1}{2}G^{(2)} D α ( 2 ) = P α ( 2 ) − 2 1 G ( 2 ) , with the gauge-invariant G ( 2 ) = Ω k G^{(2)}=\Omega_k G ( 2 ) = Ω k . The longitudinal double-trace coupling now flows from the Brownian value 1 / ( 2 ϵ ) 1/(2\epsilon) 1/ ( 2 ϵ ) in the UV to the gauge-fixing term k 2 / α k^2/\alpha k 2 / α in the IR, exponentially fast: the Zwanziger damping removes the power-law tails of radial gauge, please see (67) and § 2.2 E.
At cubic order, for non-abelian fields, both sides acquire an extra term. On the Langevin side it is the quadratic part g f a b c A i b v c g f^{abc}A_i^b v^c g f ab c A i b v c of the Zwanziger drift D i v D_i v D i v . In the bulk it is the vertex g f a b c X i a v b A i c g f^{abc}X_i^a v^b A_i^c g f ab c X i a v b A i c in 1 2 F z i F z i \frac{1}{2}F_{zi}F_{zi} 2 1 F z i F z i , with X i = ∂ z A i − ∂ i v X_i=\partial_zA_i-\partial_iv X i = ∂ z A i − ∂ i v the linear part of F z i F_{zi} F z i . Neither term is a gradient, but the Onsager–Machlup argument of § IV B does not require one anyway. We can proceed by defining D 3 ≡ ϵ 1 ϵ 2 ϵ 3 ⋅ δ 3 S B / δ A 1 δ A 2 δ A 3 \mathcal{D}_3\equiv\epsilon_1\epsilon_2\epsilon_3\cdot\delta^3S_B/\delta A_1\delta A_2\delta A_3 D 3 ≡ ϵ 1 ϵ 2 ϵ 3 ⋅ δ 3 S B / δ A 1 δ A 2 δ A 3 , and similarly P 3 \mathcal{P}_3 P 3 and G 3 \mathcal{G}_3 G 3 . With the profiles u r ( z ) = K k r ( α ) [ S k r ( z ) ] ϵ r u_r(z)=\mathcal{K}_{k_r}^{(\alpha)}[S_{k_r}(z)]\epsilon_r u r ( z ) = K k r ( α ) [ S k r ( z )] ϵ r , see (54) , the holographic side gives
D 3 ( α ) ( ϵ ) = ∫ 0 ϵ d z [ V ( u 1 , u 2 , u 3 ) − g ∑ σ ∈ S 3 f a σ 1 a σ 2 a σ 3 ( X σ 1 ⋅ u σ 2 ) v σ 3 ] . (52) \mathcal{D}_3^{(\alpha)}(\epsilon)=\int_0^\epsilon dz\left[V(u_1,u_2,u_3)-g\sum_{\sigma\in S_3}f^{a_{\sigma1}a_{\sigma2}a_{\sigma3}}(X_{\sigma1}\cdot u_{\sigma2})v_{\sigma3}\right]. \tag*{(52)} D 3 ( α ) ( ϵ ) = ∫ 0 ϵ d z [ V ( u 1 , u 2 , u 3 ) − g σ ∈ S 3 ∑ f a σ 1 a σ 2 a σ 3 ( X σ 1 ⋅ u σ 2 ) v σ 3 ] . (52)
where V ( u 1 , u 2 , u 3 ) V(u_1,u_2,u_3) V ( u 1 , u 2 , u 3 ) is the cubic vertex of (27) evaluated on the profiles, X r = ( ∂ z + k r 2 α I r ) u r X_r=(\partial_z+\frac{k_r^2}{\alpha}\mathbb{I}_r)u_r X r = ( ∂ z + α k r 2 I r ) u r and v r = α − 1 i k r ⋅ u r v_r=\alpha^{-1}i k_r\cdot u_r v r = α − 1 i k r ⋅ u r . The Langevin side gives
D 3 ( α ) ( t ) = ∑ r = 1 3 ∫ 0 t d s [ 1 2 G 3 ( x r , u r ′ , u r ′ ′ ) − Z r ] , (53) \mathcal{D}_3^{(\alpha)}(t)=\sum_{r=1}^{3}\int_0^t ds\left[\frac{1}{2}\mathcal{G}_3(x_r,u_{r'},u_{r''})-Z_r\right], \tag*{(53)} D 3 ( α ) ( t ) = r = 1 ∑ 3 ∫ 0 t d s [ 2 1 G 3 ( x r , u r ′ , u r ′′ ) − Z r ] , (53)
where x r = ( ∂ s + Ω k ( α ) ) u r x_r=(\partial_s+\Omega_k^{(\alpha)})u_r x r = ( ∂ s + Ω k ( α ) ) u r is the amputated response function, { r , r ′ , r ′ ′ } = { 1 , 2 , 3 } \{r,r',r''\}=\{1,2,3\} { r , r ′ , r ′′ } = { 1 , 2 , 3 } , and Z r = g f a r a r ′ a r ′ ′ ( x r ⋅ u r ′ ) v r ′ ′ + ( r ′ ↔ r ′ ′ ) Z_r=g f^{a_ra_{r'}a_{r''}}(x_r\cdot u_{r'})v_{r''}+(r'\leftrightarrow r'') Z r = g f a r a r ′ a r ′′ ( x r ⋅ u r ′ ) v r ′′ + ( r ′ ↔ r ′′ ) comes from the Zwanziger drift.46 For transverse external data v r = 0 v_r=0 v r = 0 , and D 3 ( α ) \mathcal{D}_3^{(\alpha)} D 3 ( α ) coincides with (55) for every α \alpha α , whereas the couplings with longitudinal legs depend on α \alpha α . This does not extend to higher orders: transverse linear data are not gauge invariant beyond linear order, and at quartic order the internal line of the exchange diagrams can be longitudinal. The quadruple-trace couplings on transverse data then depend on α \alpha α through a single zero mode (§ 2.2 E).
Multi-trace flows of gluons
Reduction to scalar seeds
Proposition 1 already implies (14) for gluons at tree level. It is nevertheless instructive, and useful for explicit results, to see how the gluon couplings arise from scalar ones. On the holographic side, the UV region 0 < z < ϵ 0<z<\epsilon 0 < z < ϵ has the bulk-to-cutoff propagator K k [ S k ( z ) ] \mathcal{K}_k[S_k(z)] K k [ S k ( z )] and the bulk-to-bulk propagator K k [ G k U V ] \mathcal{K}_k[G_k^{\mathrm{UV}}] K k [ G k UV ] , where
S k ( z ) = sinh k z sinh k ϵ , S_k(z)=\frac{\sinh kz}{\sinh k\epsilon}, S k ( z ) = sinh k ϵ sinh k z ,
G k U V ( z 1 , z 2 ) = sinh ( k z < ) sinh ( k ( ϵ − z > ) ) k sinh k ϵ . (54) G_k^{\mathrm{UV}}(z_1,z_2)=\frac{\sinh(kz_<)\sinh(k(\epsilon-z_>))}{k\sinh k\epsilon}. \tag*{(54)} G k UV ( z 1 , z 2 ) = k sinh k ϵ sinh ( k z < ) sinh ( k ( ϵ − z > )) . (54)
for z < : = min ( z 1 , z 2 ) z_<:=\min(z_1,z_2) z < := min ( z 1 , z 2 ) and z > : = max ( z 1 , z 2 ) z_>:=\max(z_1,z_2) z > := max ( z 1 , z 2 ) . The longitudinal limit G 0 U V = min ( z , w ) − z w / ϵ = z < ( ϵ − z > ) / ϵ G_0^{\mathrm{UV}}=\min(z,w)-zw/\epsilon=z_<(\epsilon-z_>)/\epsilon G 0 UV = min ( z , w ) − z w / ϵ = z < ( ϵ − z > ) / ϵ is the covariance of a Brownian bridge.47 On the stochastic side, tree diagrams are built from the response function and correlator (42) and from the vertices of S E S_E S E . On both sides the vertices are z z z - (or t t t -)independent tensors, so by (26) every tree diagram equals the vertex tensors contracted with K \mathcal{K} K -dressed radial integrals of a scalar theory with the same diagram topology. The relevant scalar seeds are the cubic contact and exchange diagrams, whose radial
integrals coincide with those of the conformally coupled ϕ 3 \phi^3 ϕ 3 theory in AdS6 _6 6 [48 ] , and the quartic contact diagram of the conformally coupled ϕ 4 \phi^4 ϕ 4 theory in AdS4 _4 4 [49 ] . Since the scalar relations hold as identities in the norms of all momenta, including their k → 0 k\to0 k → 0 limits, they lift to gluons leg by leg.
Triple-trace couplings
To first order in g g g , the UV region gives
D ( 3 ) ( ϵ ) = 1 3 ! V ∘ K 1 K 2 K 3 [ I 3 ( ϵ ) ] , (55) D^{(3)}(\epsilon)=\frac{1}{3!}V\circ\mathcal{K}_1\mathcal{K}_2\mathcal{K}_3[I_3(\epsilon)], \tag*{(55)} D ( 3 ) ( ϵ ) = 3 ! 1 V ∘ K 1 K 2 K 3 [ I 3 ( ϵ )] , (55)
I 3 ( ϵ ) = ∫ 0 ϵ d z ∏ r = 1 3 S k r ( z ) . I_3(\epsilon)=\int_0^\epsilon dz\prod_{r=1}^{3}S_{k_r}(z). I 3 ( ϵ ) = ∫ 0 ϵ d z r = 1 ∏ 3 S k r ( z ) .
with the closed form48
I 3 ( ϵ ) = ∑ η ∈ { ± 1 } 2 η 1 η 2 σ η ( cosh σ η ϵ − 1 ) 4 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ ∣ σ η = k 1 + η 1 k 2 + η 2 k 3 , (56) I_3(\epsilon)=\frac{\sum_{\eta\in\{\pm1\}^2}\frac{\eta_1\eta_2}{\sigma_\eta}(\cosh\sigma_\eta\epsilon-1)}{4\sinh k_1\epsilon\sinh k_2\epsilon\sinh k_3\epsilon}\bigg|_{\sigma_\eta=k_1+\eta_1k_2+\eta_2k_3}, \tag*{(56)} I 3 ( ϵ ) = 4 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ ∑ η ∈ { ± 1 } 2 σ η η 1 η 2 ( cosh σ η ϵ − 1 ) σ η = k 1 + η 1 k 2 + η 2 k 3 , (56)
As ϵ → ∞ \epsilon\to\infty ϵ → ∞ ,49 I 3 → 1 / k 123 I_3\to1/k_{123} I 3 → 1/ k 123 , so that D ( 3 ) → 1 2 G ( 3 ) D^{(3)}\to\frac{1}{2}G^{(3)} D ( 3 ) → 2 1 G ( 3 ) . On the stochastic side, the connected three-point function at first order in g g g gives50
P ( 3 ) ( t ) = 1 3 ! V ∘ K 1 K 2 K 3 [ I 3 ( t ) + 1 k 123 ] . (57) P^{(3)}(t)=\frac{1}{3!}V\circ\mathcal{K}_1\mathcal{K}_2\mathcal{K}_3\left[I_3(t)+\frac{1}{k_{123}}\right]. \tag*{(57)} P ( 3 ) ( t ) = 3 ! 1 V ∘ K 1 K 2 K 3 [ I 3 ( t ) + k 123 1 ] . (57)
This follows from the scalar identity
∑ r = 1 3 ∫ 0 t d s R k r ( t − s ) g k r ( t ) ∏ r ′ ≠ r G k r ′ ( s , t ) g k r ′ ( t ) = 1 + k 123 I 3 ( t ) , (58) \sum_{r=1}^{3}\int_0^t ds\,\frac{R_{k_r}(t-s)}{g_{k_r}(t)}\prod_{r'\ne r}\frac{G_{k_{r'}}(s,t)}{g_{k_{r'}}(t)}=1+k_{123}I_3(t), \tag*{(58)} r = 1 ∑ 3 ∫ 0 t d s g k r ( t ) R k r ( t − s ) r ′ = r ∏ g k r ′ ( t ) G k r ′ ( s , t ) = 1 + k 123 I 3 ( t ) , (58)
where R k ( τ ) = e − k τ R_k(\tau)=e^{-k\tau} R k ( τ ) = e − k τ and g k ( t ) = G k ( t , t ) g_k(t)=G_k(t,t) g k ( t ) = G k ( t , t ) . The identity holds because R k ( t − s ) / g k ( t ) = ( ∂ s + k ) S k ( s ) R_k(t-s)/g_k(t)=(\partial_s+k)S_k(s) R k ( t − s ) / g k ( t ) = ( ∂ s + k ) S k ( s ) and G k ( s , t ) / g k ( t ) = S k ( s ) G_k(s,t)/g_k(t)=S_k(s) G k ( s , t ) / g k ( t ) = S k ( s ) with ϵ → t \epsilon\to t ϵ → t (in S k S_k S k ), so that the left-hand side integrates to a total derivative.51 Thus we arrive at D ( 3 ) = P ( 3 ) − 1 2 G ( 3 ) D^{(3)}=P^{(3)}-\frac{1}{2}G^{(3)} D ( 3 ) = P ( 3 ) − 2 1 G ( 3 ) , consistent with Proposition 1. Note that P ( 3 ) P^{(3)} P ( 3 ) is the cubic on-shell action of the
P ( 3 ) = − ⟨ A A A ⟩ c / ( 3 ! ∏ ⟨ A A ⟩ ) . (59) P^{(3)}=-\langle AAA\rangle_c/(3!\prod\langle AA\rangle). \tag*{(59)} P ( 3 ) = − ⟨ AAA ⟩ c / ( 3 ! ∏ ⟨ AA ⟩) . (59)
∑ r = 1 n ∫ 0 t d s R k r ( t − s ) g k r ( t ) ∏ r ′ ≠ r S k r ′ ( s ) = 1 + k 1 … n I n ( t ) \sum_{r=1}^{n}\int_0^t ds\,\frac{R_{k_r}(t-s)}{g_{k_r}(t)}\prod_{r'\ne r}S_{k_{r'}}(s)=1+k_{1\ldots n}I_n(t) r = 1 ∑ n ∫ 0 t d s g k r ( t ) R k r ( t − s ) r ′ = r ∏ S k r ′ ( s ) = 1 + k 1 … n I n ( t )
with I n ( t ) = ∫ 0 t d s ∏ r S k r ( s ) I_n(t)=\int_0^t ds\prod_r S_{k_r}(s) I n ( t ) = ∫ 0 t d s ∏ r S k r ( s ) , which is the contact identity used at quartic order below. full half-space with the field pinned at z = t z=t z = t , in line with § 1.3 C 2.52
Quadruple-trace couplings: seeds
At second order the UV region gives exchange and contact terms,
D ( 4 ) ( ϵ ) = − 1 8 Sym [ V V ∘ K [ F ( ϵ ) ] ] + 1 4 ! V ( 4 ) ∘ K [ I 4 ( ϵ ) ] , (60) D^{(4)}(\epsilon)=-\frac{1}{8}\operatorname{Sym}\left[V V\circ\mathcal{K}[F(\epsilon)]\right]+\frac{1}{4!}V^{(4)}\circ\mathcal{K}[I_4(\epsilon)], \tag*{(60)} D ( 4 ) ( ϵ ) = − 8 1 Sym [ V V ∘ K [ F ( ϵ )] ] + 4 ! 1 V ( 4 ) ∘ K [ I 4 ( ϵ )] , (60)
with
I 4 ( ϵ ) = ∫ 0 ϵ d z ∏ r = 1 4 S k r ( z ) (61) I_4(\epsilon)=\int_0^\epsilon dz\prod_{r=1}^{4}S_{k_r}(z) \tag*{(61)} I 4 ( ϵ ) = ∫ 0 ϵ d z r = 1 ∏ 4 S k r ( z ) (61)
and
F ( ϵ ) = ∫ 0 ϵ ∫ 0 ϵ d z d w S k 1 ( z ) S k 2 ( z ) G k 12 U V ( z , w ) S k 3 ( w ) S k 4 ( w ) . (62) F(\epsilon)=\int_0^\epsilon\int_0^\epsilon dz\,dw\,S_{k_1}(z)S_{k_2}(z)G_{k_{12}}^{\mathrm{UV}}(z,w)S_{k_3}(w)S_{k_4}(w). \tag*{(62)} F ( ϵ ) = ∫ 0 ϵ ∫ 0 ϵ d z d w S k 1 ( z ) S k 2 ( z ) G k 12 UV ( z , w ) S k 3 ( w ) S k 4 ( w ) . (62)
The stochastic quadruple-trace weight is extracted from the connected four-point function via
P ( 4 ) = 1 4 ! [ ∑ c h a n n e l s ( 3 ! ) 2 P ( 3 ) ⟨ A A ⟩ P ( 3 ) − ⟨ A A A A ⟩ c ∏ ⟨ A A ⟩ ] . (63) P^{(4)}=\frac{1}{4!}\left[\sum_{\mathrm{channels}}(3!)^2P^{(3)}\langle AA\rangle P^{(3)}-\frac{\langle AAAA\rangle_{\mathrm{c}}}{\prod\langle AA\rangle}\right]. \tag*{(63)} P ( 4 ) = 4 ! 1 [ channels ∑ ( 3 ! ) 2 P ( 3 ) ⟨ AA ⟩ P ( 3 ) − ∏ ⟨ AA ⟩ ⟨ AAAA ⟩ c ] . (63)
At order g 2 g^2 g 2 the four-point function receives three kinds of contributions: diagrams with a response function as internal line (from iterating the cubic drift), diagrams with a correlator as internal line, and the quartic term of S E S_E S E . For the scalar seeds we have verified that they combine into
P ( 4 ) ( t ) = − 1 8 Sym [ V V ∘ K [ F ( t ) + 1 k 1234 k 1212 k 3412 ] ] + 1 4 ! V ( 4 ) ∘ K [ I 4 ( t ) + 1 k 1234 ] , (64) \begin{aligned}
P^{(4)}(t)={}&-\frac{1}{8}\operatorname{Sym}\left[V V\circ\mathcal{K}\left[F(t)+\frac{1}{k_{1234}k_{1212}k_{3412}}\right]\right]\\
&+\frac{1}{4!}V^{(4)}\circ\mathcal{K}\left[I_4(t)+\frac{1}{k_{1234}}\right], \tag*{(64)}
\end{aligned} P ( 4 ) ( t ) = − 8 1 Sym [ V V ∘ K [ F ( t ) + k 1234 k 1212 k 3412 1 ] ] + 4 ! 1 V ( 4 ) ∘ K [ I 4 ( t ) + k 1234 1 ] , (64)
i.e. D ( 4 ) = P ( 4 ) − 1 2 G ( 4 ) D^{(4)}=P^{(4)}-\frac{1}{2}G^{(4)} D ( 4 ) = P ( 4 ) − 2 1 G ( 4 ) via (32). The contact part follows from the four-leg version of (58) ; this is the identity behind [49 ] . The exchange part is guaranteed by Proposition 1;53 at the fixed point, the O ( λ 2 ) O(\lambda^2) O ( λ 2 ) exchange contributions of a bulk cubic coupling were already considered in [48 , 136 ] . As far as we are aware, for an exchange generated by a bulk cubic vertex, this comparison has not been made at finite cutoff before (see [50 , 51 ] for exchange-type contributions of multi-trace deformations).
We can now proceed to obtaining closed form expressions as both seeds are elementary. Expanding ∏ r sinh k r z \prod_r\sinh k_r z ∏ r sinh k r z into hyperbolic cosines gives
I 4 ( ϵ ) = ∑ η ∈ { ± 1 } 3 η 2 η 3 η 4 sinh ( σ η ϵ ) σ η 8 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ sinh k 4 ϵ , σ η = k 1 + η 2 k 2 + η 3 k 3 + η 4 k 4 (65) I_4(\epsilon)=\frac{\sum_{\eta\in\{\pm1\}^3}\eta_2\eta_3\eta_4\frac{\sinh(\sigma_\eta\epsilon)}{\sigma_\eta}}{8\sinh k_1\epsilon\sinh k_2\epsilon\sinh k_3\epsilon\sinh k_4\epsilon},
\qquad\sigma_\eta=k_1+\eta_2k_2+\eta_3k_3+\eta_4k_4 \tag*{(65)} I 4 ( ϵ ) = 8 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ sinh k 4 ϵ ∑ η ∈ { ± 1 } 3 η 2 η 3 η 4 σ η s i n h ( σ η ϵ ) , σ η = k 1 + η 2 k 2 + η 3 k 3 + η 4 k 4 (65)
For the exchange seed, with s ≡ k 12 s\equiv k_{12} s ≡ k 12 ,
F ( ϵ ) = ∑ η 2 , η 4 = ± 1 η 2 η 4 J s ( k 1 + η 2 k 2 , k 3 + η 4 k 4 ) 4 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ sinh k 4 ϵ , (66) F(\epsilon)=\frac{\sum_{\eta_2,\eta_4=\pm1}\eta_2\eta_4J_s(k_1+\eta_2k_2,k_3+\eta_4k_4)}{4\sinh k_1\epsilon\sinh k_2\epsilon\sinh k_3\epsilon\sinh k_4\epsilon}, \tag*{(66)} F ( ϵ ) = 4 sinh k 1 ϵ sinh k 2 ϵ sinh k 3 ϵ sinh k 4 ϵ ∑ η 2 , η 4 = ± 1 η 2 η 4 J s ( k 1 + η 2 k 2 , k 3 + η 4 k 4 ) , (66)
whereas the longitudinal seed F ( 0 ) F^{(0)} F ( 0 ) follows from F ( 0 ) ≡ F ∣ s → 0 F^{(0)}\equiv F|_{s\to0} F ( 0 ) ≡ F ∣ s → 0 . Here J s ( a , b ) J_s(a,b) J s ( a , b ) is an elementary function: see § A 3 for its details, along with those of the longitudinal seed F ( 0 ) F^{(0)} F ( 0 ) .
In the IR, we have I 4 → 1 / k 1234 I_4\to1/k_{1234} I 4 → 1/ k 1234 and F → 1 / ( k 1234 k 1212 k 3412 ) F\to1/(k_{1234}k_{1212}k_{3412}) F → 1/ ( k 1234 k 1212 k 3412 ) exponentially fast, whereas the longitudinal seed approaches its limit only as a power,
F ( 0 ) ( ϵ ) = 1 k 1234 k 12 k 34 − 1 ϵ k 12 2 k 34 2 + ⋯ , (67) F^{(0)}(\epsilon)=\frac{1}{k_{1234}k_{12}k_{34}}-\frac{1}{\epsilon k_{12}^2k_{34}^2}+\cdots, \tag*{(67)} F ( 0 ) ( ϵ ) = k 1234 k 12 k 34 1 − ϵ k 12 2 k 34 2 1 + ⋯ , (67)
where the dots are exponentially small. We derive (67) in § 2.2 E.54
In the UV, I 4 = ϵ / 5 + O ( ϵ 3 ) I_4=\epsilon/5+O(\epsilon^3) I 4 = ϵ /5 + O ( ϵ 3 ) and F , F ( 0 ) = ϵ 3 / 112 + O ( ϵ 5 ) F,F^{(0)}=\epsilon^3/112+O(\epsilon^5) F , F ( 0 ) = ϵ 3 /112 + O ( ϵ 5 ) , so all quadruple-trace couplings vanish at the boundary, as required by the Dirichlet condition, and the contact term dominates near it.
Colour-ordered quadruple-trace flows
Expressions
We can now contract quadruple-trace couplings with transverse polarizations, ϵ r ⋅ k r = 0 \epsilon_r\cdot k_r=0 ϵ r ⋅ k r = 0 , and define the quadruple-trace vertex D 4 ≡ ϵ 1 ϵ 2 ϵ 3 ϵ 4 ⋅ δ 4 S B / δ A 1 ⋯ δ A 4 = 4 ! D ( 4 ) ⋅ ϵ 1 ⋯ ϵ 4 \mathfrak{D}_4\equiv\epsilon_1\epsilon_2\epsilon_3\epsilon_4\cdot\delta^4S_B/\delta A_1\cdots\delta A_4=4!\,D^{(4)}\cdot\epsilon_1\cdots\epsilon_4 D 4 ≡ ϵ 1 ϵ 2 ϵ 3 ϵ 4 ⋅ δ 4 S B / δ A 1 ⋯ δ A 4 = 4 ! D ( 4 ) ⋅ ϵ 1 ⋯ ϵ 4 . In the DDM basis of § A 2,
D 4 = D 4 ( 1 , 2 , 3 , 4 ; ϵ ) f a 1 a 2 c f c a 3 a 4 + D 4 ( 1 , 3 , 2 , 4 ; ϵ ) f a 1 a 3 c f c a 2 a 4 , (68) \mathfrak{D}_4=\mathfrak{D}_4(1,2,3,4;\epsilon)f^{a_1a_2c}f^{ca_3a_4}+\mathfrak{D}_4(1,3,2,4;\epsilon)f^{a_1a_3c}f^{ca_2a_4}, \tag*{(68)} D 4 = D 4 ( 1 , 2 , 3 , 4 ; ϵ ) f a 1 a 2 c f c a 3 a 4 + D 4 ( 1 , 3 , 2 , 4 ; ϵ ) f a 1 a 3 c f c a 2 a 4 , (68)
with the colour-ordered quadruple-trace flows
D 4 ( 1 , 2 , 3 , 4 ; ϵ ) = g 2 [ N s ( ϵ ) − N t ( ϵ ) + C 1234 I 4 ( ϵ ) ] , (69) \mathfrak{D}_4(1,2,3,4;\epsilon)=g^2\left[N_s(\epsilon)-N_t(\epsilon)+C_{1234}I_4(\epsilon)\right], \tag*{(69)} D 4 ( 1 , 2 , 3 , 4 ; ϵ ) = g 2 [ N s ( ϵ ) − N t ( ϵ ) + C 1234 I 4 ( ϵ ) ] , (69)
D 4 ( 1 , 3 , 2 , 4 ; ϵ ) = g 2 [ N t ( ϵ ) − N u ( ϵ ) + C 1324 I 4 ( ϵ ) ] . \mathfrak{D}_4(1,3,2,4;\epsilon)=g^2\left[N_t(\epsilon)-N_u(\epsilon)+C_{1324}I_4(\epsilon)\right]. D 4 ( 1 , 3 , 2 , 4 ; ϵ ) = g 2 [ N t ( ϵ ) − N u ( ϵ ) + C 1324 I 4 ( ϵ ) ] .
The s s s -channel function is
N s ( ϵ ) = n s ⊥ F s ( ϵ ) + n s ∥ F s ( 0 ) ( ϵ ) , n s ⊥ = v 12 ⋅ Π ⊥ ( k 1 + k 2 ) ⋅ v 34 , (70) N_s(\epsilon)=n_s^{\perp}F_s(\epsilon)+n_s^{\parallel}F_s^{(0)}(\epsilon),
\qquad
n_s^{\perp}=v_{12}\cdot\Pi^{\perp}(k_1+k_2)\cdot v_{34}, \tag*{(70)} N s ( ϵ ) = n s ⊥ F s ( ϵ ) + n s ∥ F s ( 0 ) ( ϵ ) , n s ⊥ = v 12 ⋅ Π ⊥ ( k 1 + k 2 ) ⋅ v 34 , (70)
n s ∥ = − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ( k 1 2 − k 2 2 ) ( k 3 2 − k 4 2 ) ∣ k 1 + k 2 ∣ 2 , n_s^{\parallel}=-\frac{(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)(k_1^2-k_2^2)(k_3^2-k_4^2)}{\lvert k_1+k_2\rvert^2}, n s ∥ = − ∣ k 1 + k 2 ∣ 2 ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ( k 1 2 − k 2 2 ) ( k 3 2 − k 4 2 ) ,
with F s F_s F s the exchange seed for the pairing ( 12 ) ( 34 ) (12)(34) ( 12 ) ( 34 ) and
v 12 = ( ϵ 1 ⋅ ϵ 2 ) ( k 1 − k 2 ) + 2 ( ϵ 1 ⋅ k 2 ) ϵ 2 − 2 ( ϵ 2 ⋅ k 1 ) ϵ 1 , (71) v_{12}=(\epsilon_1\cdot\epsilon_2)(k_1-k_2)+2(\epsilon_1\cdot k_2)\epsilon_2-2(\epsilon_2\cdot k_1)\epsilon_1, \tag*{(71)} v 12 = ( ϵ 1 ⋅ ϵ 2 ) ( k 1 − k 2 ) + 2 ( ϵ 1 ⋅ k 2 ) ϵ 2 − 2 ( ϵ 2 ⋅ k 1 ) ϵ 1 , (71)
and similarly for v 34 v_{34} v 34 . The t t t - and u u u -channel functions follow by successive ( 243 ) (243) ( 243 ) relabellings, i.e. the pairings ( 14 ) ( 23 ) (14)(23) ( 14 ) ( 23 ) and ( 13 ) ( 42 ) (13)(42) ( 13 ) ( 42 ) . The contact polynomials are
C 1234 = 2 ( ϵ 1 ⋅ ϵ 3 ) ( ϵ 2 ⋅ ϵ 4 ) − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) − ( ϵ 1 ⋅ ϵ 4 ) ( ϵ 2 ⋅ ϵ 3 ) , (72) C_{1234}=2(\epsilon_1\cdot\epsilon_3)(\epsilon_2\cdot\epsilon_4)-(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)-(\epsilon_1\cdot\epsilon_4)(\epsilon_2\cdot\epsilon_3), \tag*{(72)} C 1234 = 2 ( ϵ 1 ⋅ ϵ 3 ) ( ϵ 2 ⋅ ϵ 4 ) − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) − ( ϵ 1 ⋅ ϵ 4 ) ( ϵ 2 ⋅ ϵ 3 ) , (72)
and C 1324 C_{1324} C 1324 is obtained by exchanging 2 ↔ 3 2\leftrightarrow3 2 ↔ 3 . The longitudinal part n s ∥ n_s^{\parallel} n s ∥ is the AdS-specific Coulomb-like term: it involves k r 2 = ∣ k r ∣ 2 k_r^2=\lvert k_r\rvert^2 k r 2 = ∣ k r ∣ 2 and would vanish for null external momenta. In the IR it has only the total-energy pole, n s ∥ / ( k 1234 k 12 k 34 ) = − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ( k 1 − k 2 ) ( k 3 − k 4 ) / ( ∣ k 1 + k 2 ∣ 2 k 1234 ) n_s^{\parallel}/(k_{1234}k_{12}k_{34})=-(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)(k_1-k_2)(k_3-k_4)/(\lvert k_1+k_2\rvert^2k_{1234}) n s ∥ / ( k 1234 k 12 k 34 ) = − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ( k 1 − k 2 ) ( k 3 − k 4 ) / (∣ k 1 + k 2 ∣ 2 k 1234 ) , as befits an instantaneous interaction (§ 2.2 E).
We can interpret the results in the light of Proposition 1:
P 4 ( ⋯ ; t ) = D 4 ( ⋯ ; t ) + 1 2 G 4 = D 4 ( ⋯ ; t ) + D 4 ( ⋯ ; ∞ ) (73) \mathcal{P}_4(\cdots;t)=\mathcal{D}_4(\cdots;t)+\frac{1}{2}\mathcal{G}_4=\mathcal{D}_4(\cdots;t)+\mathcal{D}_4(\cdots;\infty) \tag*{(73)} P 4 ( ⋯ ; t ) = D 4 ( ⋯ ; t ) + 2 1 G 4 = D 4 ( ⋯ ; t ) + D 4 ( ⋯ ; ∞ ) (73)
where the second equality uses that the slab couplings approach the equilibrium kernel in the infrared, {(xref)}(70). As for P ( 3 ) \mathcal{P}^{(3)} P ( 3 ) , this is the quartic on-shell action of the half-space with the field pinned at z = t z=t z = t : the slab 0 < z < t 0<z<t 0 < z < t contributes D 4 ( t ) \mathcal{D}_4(t) D 4 ( t ) and the region z > t z>t z > t contributes 1 2 G 4 \frac{1}{2}\mathcal{G}_4 2 1 G 4 , see {(xref)}(38).
We can also comment on the infrared and flat-space limits.55 In radial gauge, as ϵ → ∞ \epsilon\to\infty ϵ → ∞ ,
D 4 ( 1 , 2 , 3 , 4 ; ∞ ) = g 2 [ n s ⊥ k 1234 k 12 2 k 34 2 + n s ∥ k 1234 k 12 k 34 − ( s → t ) + C 1234 k 1234 ] , (74) \mathcal{D}_4(1,2,3,4;\infty)=g^2\left[\frac{n_s^{\perp}}{k_{1234}k_{12}^2k_{34}^2}+\frac{n_s^{\parallel}}{k_{1234}k_{12}k_{34}}-(s\to t)+\frac{C_{1234}}{k_{1234}}\right], \tag*{(74)} D 4 ( 1 , 2 , 3 , 4 ; ∞ ) = g 2 [ k 1234 k 12 2 k 34 2 n s ⊥ + k 1234 k 12 k 34 n s ∥ − ( s → t ) + k 1234 C 1234 ] , (74)
which is the colour-ordered combination W s − W t W_s-W_t W s − W t of the gluon Witten diagrams {(xref)}(29) of [66 ] plus the contact diagram, up to the overall sign of δ 4 S I R = − ⟨ J J J J ⟩ c \delta^4S_{\rm IR}=-\langle JJJJ\rangle_c δ 4 S IR = − ⟨ J J J J ⟩ c and the factor of − i -i − i in their conventions. As an independent check of all relative signs, the residue at the total-energy pole k 1234 → 0 k_{1234}\to0 k 1234 → 0 must be the flat-space colour-ordered amplitude for the null momenta p r = ( k r , i ∣ k r ∣ ) p_r=(k_r,i\lvert k_r\rvert) p r = ( k r , i ∣ k r ∣) [25 , 26 , 113 ] . Indeed, at k 1234 = 0 k_{1234}=0 k 1234 = 0 the transverse and longitudinal parts combine into the spatial part of the flat-space axial-gauge propagator, Π ⊥ / ( k 12 2 k 34 2 ) + Π ∥ / ( k 12 k 34 ) → ( 1 − q q / k 12 2 ) / ( q 2 − k 12 2 ) \Pi^{\perp}/(k_{12}^2k_{34}^2)+\Pi^{\parallel}/(k_{12}k_{34})\to(1-qq/k_{12}^2)/(q^2-k_{12}^2) Π ⊥ / ( k 12 2 k 34 2 ) + Π ∥ / ( k 12 k 34 ) → ( 1 − q q / k 12 2 ) / ( q 2 − k 12 2 ) with q = k 1 + k 2 q=k_1+k_2 q = k 1 + k 2 . At random complex kinematics with k 1234 = 0 k_{1234}=0 k 1234 = 0 and helicity polarizations, we numerically verified to high precision that the residue vanishes for the all-plus and single-minus configurations, as it must for tree-level Yang–Mills amplitudes. For MHV configurations, the ratio of the two colour orderings reproduces the Parke–Taylor ratio − ⟨ 13 ⟩ ⟨ 24 ⟩ / ( ⟨ 12 ⟩ ⟨ 34 ⟩ ) -\langle13\rangle\langle24\rangle/(\langle12\rangle\langle34\rangle) − ⟨ 13 ⟩ ⟨ 24 ⟩ / (⟨ 12 ⟩ ⟨ 34 ⟩) [137 ] .
Gauge dependence and the Gauss law
The longitudinal channel of the quadruple-trace flows isolates the gauge dependence, and we can understand it in four steps:
In radial gauge , the longitudinal exchange is the Coulomb energy of the radial colour charges minus a zero mode of the Brownian bridge; the Coulomb term is instantaneous and carries the total-energy pole, the zero mode is what relaxes as a power.
In Zwanziger’s gauge , the Coulomb term is unchanged and only the zero mode changes: it relaxes exponentially, but to an α \alpha α -dependent infrared value, so the equilibrium distribution of Zwanziger’s process is not the naive gauge-fixed weight beyond quadratic order, while the flat-space limit is unaffected.
The same infrared shift follows from the stationary Fokker–Planck equation of Zwanziger’s process alone, which provides a non-trivial consistency check of our quartic formulae for a drift that is not a gradient.
Finally , the zero mode is the remnant of the constraint that the gauge condition imposes on E ∥ E^{\parallel} E ∥ ; it is absent when the Gauss law is imposed, and it is a redundant coupling that a gauge rotation of the cutoff data removes. The content of the flows independent of the bulk gauge condition is therefore the transverse exchange, the contact term and the Coulomb term.
Coulomb term and zero mode
In radial gauge the longitudinal exchange splits into a local Coulomb term and a zero mode. For the pairing (12)(34), let
W 12 ( z ) ≡ ∂ z S k 1 ( z ) S k 2 ( z ) − S k 1 ( z ) ∂ z S k 2 ( z ) . (75) W_{12}(z) \equiv\partial_z S_{k_1}(z)S_{k_2}(z)-S_{k_1}(z)\partial_z S_{k_2}(z). \tag*{(75)} W 12 ( z ) ≡ ∂ z S k 1 ( z ) S k 2 ( z ) − S k 1 ( z ) ∂ z S k 2 ( z ) . (75)
Up to normalization, ( ϵ 1 ⋅ ϵ 2 ) W 12 (\epsilon_1\cdot\epsilon_2)W_{12} ( ϵ 1 ⋅ ϵ 2 ) W 12 is the colour-charge density g f a b c A i b ∂ z A i c g f^{abc}A_i^b\partial_z A_i^c g f ab c A i b ∂ z A i c of the two transverse modes, i.e. the source of the Gauss law D i E i = 0 D_iE_i=0 D i E i = 0 .56 Since ∂ z 2 S k = k 2 S k \partial_z^2S_k=k^2S_k ∂ z 2 S k = k 2 S k , we have ∂ z W 12 = ( k 1 2 − k 2 2 ) S k 1 S k 2 \partial_zW_{12}=(k_1^2-k_2^2)S_{k_1}S_{k_2} ∂ z W 12 = ( k 1 2 − k 2 2 ) S k 1 S k 2 . The longitudinal projection of the cubic vertex is therefore ( ϵ 1 ⋅ ϵ 2 ) ∂ z W 12 / ∣ k 1 + k 2 ∣ (\epsilon_1\cdot\epsilon_2)\partial_zW_{12}/|k_1+k_2| ( ϵ 1 ⋅ ϵ 2 ) ∂ z W 12 /∣ k 1 + k 2 ∣ , and n s ∥ n_s^{\parallel} n s ∥ in (70) is the product of two such factors. Integrating by parts against the Brownian bridge and using (A5), we find
n s ∥ F s ( 0 ) = − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ∣ k 1 + k 2 ∣ 2 [ ∫ 0 ϵ d z W 12 W 34 − 1 ϵ ∫ 0 ϵ d z W 12 ∫ 0 ϵ d w W 34 ] . (76) n_s^{\parallel}F_s^{(0)}=-\frac{(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)}{|k_1+k_2|^2}\left[\int_0^\epsilon dz\,W_{12}W_{34}-\frac{1}{\epsilon}\int_0^\epsilon dz\,W_{12}\int_0^\epsilon dw\,W_{34}\right]. \tag*{(76)} n s ∥ F s ( 0 ) = − ∣ k 1 + k 2 ∣ 2 ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) [ ∫ 0 ϵ d z W 12 W 34 − ϵ 1 ∫ 0 ϵ d z W 12 ∫ 0 ϵ d w W 34 ] . (76)
The first term is the Coulomb energy of the radial colour charges. It is local in z z z , i.e. instantaneous in stochastic time, and it carries the total-energy pole. The second term is the − z w / ϵ -zw/\epsilon − z w / ϵ part of the bridge covariance, its zero mode. The internal longitudinal field vanishes at both ends of the slab, so in radial gauge E ∥ = ∂ z A ∥ E^{\parallel}=\partial_zA^{\parallel} E ∥ = ∂ z A ∥ has zero mean, ∫ 0 ϵ E ∥ = 0 \int_0^\epsilon E^{\parallel}=0 ∫ 0 ϵ E ∥ = 0 , and cannot follow the total charge ∫ 0 ϵ W 12 \int_0^\epsilon W_{12} ∫ 0 ϵ W 12 . As ϵ → ∞ \epsilon\to\infty ϵ → ∞ the two integrals approach ( k 1 − k 2 ) ( k 3 − k 4 ) / k 1234 (k_1-k_2)(k_3-k_4)/k_{1234} ( k 1 − k 2 ) ( k 3 − k 4 ) / k 1234 and ( k 1 − k 2 ) / k 12 (k_1-k_2)/k_{12} ( k 1 − k 2 ) / k 12 exponentially fast, so the power law (67) comes entirely from the zero mode.
Zwanziger’s gauge
In Zwanziger’s gauge the Coulomb term is unchanged and the entire gauge dependence sits in the zero mode. With A z = α − 1 ∂ i A i A_z=\alpha^{-1}\partial_iA_i A z = α − 1 ∂ i A i two things change. The internal longitudinal line becomes the Ornstein–Uhlenbeck bridge G γ O U \mathcal{G}^{\mathrm{OU}}_\gamma G γ OU with γ = ∣ k 1 + k 2 ∣ 2 / α \gamma=|k_1+k_2|^2/\alpha γ = ∣ k 1 + k 2 ∣ 2 / α , see (50) . In addition, the vertex g f a b c X i a ∂ i A i b g f^{abc}X_i^a\partial_iA_i^b g f ab c X i a ∂ i A i b of § IV D couples to this line through its v v v , which turns the longitudinal source into ( ∂ z − γ ) W 12 (\partial_z-\gamma)W_{12} ( ∂ z − γ ) W 12 . By using (A5), we can see that the longitudinal channel function becomes
N s , α ∥ = − ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) ∣ k 1 + k 2 ∣ 2 [ ∫ 0 ϵ d z W 12 W 34 − γ sinh γ ϵ W ^ 12 W ^ 34 ] , (77) N_{s,\alpha}^{\parallel}=-\frac{(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)}{|k_1+k_2|^2}\left[\int_0^\epsilon dz\,W_{12}W_{34}-\frac{\gamma}{\sinh\gamma\epsilon}\widehat{W}_{12}\widehat{W}_{34}\right], \tag*{(77)} N s , α ∥ = − ∣ k 1 + k 2 ∣ 2 ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) [ ∫ 0 ϵ d z W 12 W 34 − sinh γ ϵ γ W 12 W 34 ] , (77)
with W ^ 12 ≡ ∫ 0 ϵ d z e γ ( z − ϵ / 2 ) W 12 ( z ) \widehat{W}_{12}\equiv\int_0^\epsilon dz\,e^{\gamma(z-\epsilon/2)}W_{12}(z) W 12 ≡ ∫ 0 ϵ d z e γ ( z − ϵ /2 ) W 12 ( z ) . The transverse internal line and the contact term are unchanged, since the Zwanziger vertices need a longitudinal leg. Hence D 4 ( α ) ( 1 , 2 , 3 , 4 ; ϵ ) \mathcal{D}_4^{(\alpha)}(1,2,3,4;\epsilon) D 4 ( α ) ( 1 , 2 , 3 , 4 ; ϵ ) is given by (69) , with n s ∥ F s ( 0 ) n_s^{\parallel}F_s^{(0)} n s ∥ F s ( 0 ) replaced by N s , α ∥ N_{s,\alpha}^{\parallel} N s , α ∥ and similarly in the t t t channel. As α → ∞ \alpha\to\infty α → ∞ the zero-mode term reduces to the radial one. For finite α \alpha α it approaches its limit exponentially fast, so Zwanziger’s gauge has no power-law tails. Its infrared value, however, differs from that of radial gauge:
D 4 ( α ) ( 1 , 2 , 3 , 4 ; ∞ ) = D 4 ( 1 , 2 , 3 , 4 ; ∞ ) + g 2 ( Z s − Z t ) , (78) \mathcal{D}_4^{(\alpha)}(1,2,3,4;\infty)=\mathcal{D}_4(1,2,3,4;\infty)+g^2(Z_s-Z_t), \tag*{(78)} D 4 ( α ) ( 1 , 2 , 3 , 4 ; ∞ ) = D 4 ( 1 , 2 , 3 , 4 ; ∞ ) + g 2 ( Z s − Z t ) , (78)
where
Z s = 2 ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) α k 1 − k 2 k 12 + γ s k 3 − k 4 k 34 + γ s , (79) Z_s=\frac{2(\epsilon_1\cdot\epsilon_2)(\epsilon_3\cdot\epsilon_4)}{\alpha}\frac{k_1-k_2}{k_{12}+\gamma_s}\frac{k_3-k_4}{k_{34}+\gamma_s}, \tag*{(79)} Z s = α 2 ( ϵ 1 ⋅ ϵ 2 ) ( ϵ 3 ⋅ ϵ 4 ) k 12 + γ s k 1 − k 2 k 34 + γ s k 3 − k 4 , (79)
with γ s = ∣ k 1 + k 2 ∣ 2 / α \gamma_s=|k_1+k_2|^2/\alpha γ s = ∣ k 1 + k 2 ∣ 2 / α , and Z t Z_t Z t is obtained by the relabelling used in (70) . Z s Z_s Z s has no total-energy pole, so the flat-space limit of § 2.2 D does not depend on α \alpha α . By [15 ] the same shift appears in the equilibrium weights,
P 4 ( α ) ( 1 , 2 , 3 , 4 ; ∞ ) = 2 P 4 ( 1 , 2 , 3 , 4 ; ∞ ) + g 2 ( Z s − Z t ) , \mathcal{P}_4^{(\alpha)}(1,2,3,4;\infty)=2\mathcal{P}_4(1,2,3,4;\infty)+g^2(Z_s-Z_t), P 4 ( α ) ( 1 , 2 , 3 , 4 ; ∞ ) = 2 P 4 ( 1 , 2 , 3 , 4 ; ∞ ) + g 2 ( Z s − Z t ) ,
where the first term is the colour-ordered part of the gauge-invariant kernel G 4 \mathcal{G}_4 G 4 . The equilibrium distribution of Zwanziger’s process is therefore not e − S E − α − 1 ∫ ( ∂ i A i ) 2 e^{-S_E-\alpha^{-1}\int(\partial_iA_i)^2} e − S E − α − 1 ∫ ( ∂ i A i ) 2 beyond quadratic order.57 Transverse data see the nonlinear completion of the gauge-fixing weight through the internal longitudinal line.58
The shift from the stationary Fokker–Planck equation
The shift Z s − Z t Z_s-Z_t Z s − Z t can be derived on the stochastic side alone, without reference to the bulk. We first write the equilibrium distribution of Zwanziger’s process as e − S E − Δ e^{-S_E-\Delta} e − S E − Δ and use the fact that gauge invariance implies ∫ ( D i v ) a δ S E / δ A i a = 0 \int(D_i v)^a\delta S_E/\delta A_i^a=0 ∫ ( D i v ) a δ S E / δ A i a = 0 ; then, at tree level, Δ \Delta Δ obeys
( 1 2 δ S E δ A + D v ) ⋅ δ Δ δ A + 1 2 ∣ δ Δ δ A ∣ 2 = 0. (80) \left(\frac{1}{2}\frac{\delta S_E}{\delta A}+Dv\right)\cdot\frac{\delta\Delta}{\delta A}+\frac{1}{2}\left|\frac{\delta\Delta}{\delta A}\right|^2=0. \tag*{(80)} ( 2 1 δ A δ S E + D v ) ⋅ δ A δ Δ + 2 1 δ A δ Δ 2 = 0. (80)
with the dot denoting the integral over x x x and the sums over colour and vector indices. At quadratic order Δ 2 = α − 1 ∫ ( ∂ i A i ) 2 \Delta_2=\alpha^{-1}\int(\partial_iA_i)^2 Δ 2 = α − 1 ∫ ( ∂ i A i ) 2 . At higher orders, the linear part of 1 2 δ S E / δ A + D v \frac{1}{2}\delta S_E/\delta A+Dv 2 1 δ S E / δ A + D v together with δ Δ 2 / δ A \delta\Delta_2/\delta A δ Δ 2 / δ A multiplies each monomial by the sum of the relaxation rates of its legs, k k k for a transverse and γ \gamma γ for a longitudinal one, and the equation is solved order by order.59 On transverse data the result at quartic order is
ϵ 1 ⋯ ϵ 4 ⋅ δ 4 Δ = ∑ ( a b ) ( c d ) 2 γ a b j a b ⋅ j c d ( k a b + γ a b ) ( k c d + γ a b ) , (81) \epsilon_1 \cdots\epsilon_4 \cdot\delta^4 \Delta= \sum_{(ab)(cd)} \frac{2\gamma_{ab}j_{ab}\cdot j_{cd}}{(k_{ab}+\gamma_{ab})(k_{cd}+\gamma_{ab})}, \tag*{(81)} ϵ 1 ⋯ ϵ 4 ⋅ δ 4 Δ = ( ab ) ( c d ) ∑ ( k ab + γ ab ) ( k c d + γ ab ) 2 γ ab j ab ⋅ j c d , (81)
with γ a b = ∣ k a + k b ∣ 2 / α \gamma_{ab} = \lvert k_a+k_b\rvert^2/\alpha γ ab = ∣ k a + k b ∣ 2 / α and j a b j_{ab} j ab the longitudinal projection of 1 2 δ S E ( 3 ) δ A \frac{1}{2}\frac{\delta S_E^{(3)}}{\delta A} 2 1 δ A δ S E ( 3 ) on legs a a a and b b b . This reproduces g 2 ( Z s − Z t ) g^2(Z_s-Z_t) g 2 ( Z s − Z t ) for the colour ordering ( 1 , 2 , 3 , 4 ) (1,2,3,4) ( 1 , 2 , 3 , 4 ) , as we have checked with SU(2) colours in the Mathematica file. The cancellation of k 1234 k_{1234} k 1234 is the stochastic counterpart of the absence of a total-energy pole. Only the cubic kernel of S E S_E S E enters, and the agreement with (74)–(75) checks the gauge-sector dictionary of § IV at quartic order, for a drift that is not a gradient.
Gauss law and redundancy
The zero mode depends on the bulk gauge condition: it vanishes when the Gauss law is imposed, and it can be removed by a gauge rotation of the cutoff data. As α → 0 \alpha\to0 α → 0 , the gauge condition becomes the Coulomb gauge ∂ i A i = 0 \partial_i A_i = 0 ∂ i A i = 0 , with A z A_z A z fixed by the Gauss law, and the zero-mode term vanishes: Z s = O ( α ) Z_s = O(\alpha) Z s = O ( α ) , and so is the zero-mode term in (77) at finite ϵ \epsilon ϵ . The same result follows if A z A_z A z is integrated over instead of fixed. The longitudinal field E ∥ = ∂ z A ∥ − ∂ A z E^\parallel= \partial_z A^\parallel-\partial A_z E ∥ = ∂ z A ∥ − ∂ A z is then unconstrained, and minimizing 1 2 ∫ ∣ E ∥ ∣ 2 \frac{1}{2}\int|E^\parallel|^2 2 1 ∫ ∣ E ∥ ∣ 2 against the charge source leaves only the Coulomb term. The zero-mode terms in (76) and (77) are what remains of the constraints ∫ 0 ϵ E ∥ = 0 \int_0^\epsilon E^\parallel= 0 ∫ 0 ϵ E ∥ = 0 and ∫ 0 ϵ e γ z E ∥ = 0 \int_0^\epsilon e^{\gamma z}E^\parallel= 0 ∫ 0 ϵ e γ z E ∥ = 0 . These are imposed by the two gauge conditions, under which E ∥ = ∂ z A ∥ E^\parallel= \partial_z A^\parallel E ∥ = ∂ z A ∥ and E ∥ = ( ∂ z + γ ) A ∥ E^\parallel= (\partial_z+\gamma)A^\parallel E ∥ = ( ∂ z + γ ) A ∥ respectively. With the Gauss law imposed, the holographic Wilsonian action is independent of the bulk gauge condition. Moreover, minimizing the Wilsonian action of any bulk gauge over gauge rotations V V V of the cutoff data gives the Gauss-law action, S B G [ A ] = min V S B [ A V ] S_B^{G}[A] = \min_V S_B[ A^V ] S B G [ A ] = min V S B [ A V ] .
This follows from a simple observation: every slab configuration with A z A_z A z unconstrained is a gauge transformation, with U ( 0 ) = 1 U(0)=1 U ( 0 ) = 1 , of a configuration in the chosen gauge whose cutoff data are A V A^V A V for some V V V . The zero-mode couplings can therefore be removed by the redefinition A → A V ∗ [ A ] A \to A_{V_*[A]} A → A V ∗ [ A ] of the cutoff data, with V ∗ V_* V ∗ the minimizing rotation. This redefinition is non-local, and so is the resulting Gauss-law action: already at quadratic order only the sum Π ⊥ k coth k ϵ + Π ∥ / ϵ \Pi_\perp k\coth k\epsilon+\Pi_\parallel/\epsilon Π ⊥ k coth k ϵ + Π ∥ / ϵ of (48) is analytic at k = 0 k=0 k = 0 . This is the non-abelian counterpart of the observation of [42 , 43 ] that the constrained amplitude is not a local Wilsonian action and that the longitudinal (Wilson-line) mode has to be retained in it; the radial-gauge kernels, including the zero mode and its power-law relaxation, are the ones with the expected locality. The quadruple-trace couplings of the Gauss-law action relax exponentially, and their infrared limit is that of radial gauge. In this sense the part of the quadruple-trace flows that does not depend on the bulk gauge condition is the transverse exchange, the contact term and the Coulomb term, while the power-law relaxation in radial gauge reflects the unrestricted random walk of the gauge modes in Parisi–Wu stochastic quantization.
The de Sitter reading
Although the dictionary of this paper is Euclidean, it has a Lorentzian counterpart in de Sitter space and there is a well established and widely studied analytical continuation between them [138 , 139 , 140 , 141 ] . It is then natural to expect that our results can actually be useful in the context of cosmology, and we review that perspective in this section.
Wavefunction, in-in correlators and stochastic inflation
Euclidean AdS4 _4 4 in the Poincaré patch continues to the expanding patch of dS4 _4 4 by z → − i η z \to-i\eta z → − i η together with the continuation of the radius, and for Weyl-invariant fields both are the flat half-space in conformal time. Under this continuation the regular solution e − k z e^{-kz} e − k z becomes the Bunch–Davies mode [142 ] , so the regular radial wavefunction Ψ I R [ Φ ; ϵ ] \Psi_{\mathrm{IR}}[\Phi;\epsilon] Ψ IR [ Φ ; ϵ ] becomes the Bunch–Davies wavefunction Ψ B D [ Φ ; η ] \Psi_{\mathrm{BD}}[\Phi;\eta] Ψ BD [ Φ ; η ] , and S I R S_{\mathrm{IR}} S IR becomes the generating functional of its coefficients ψ n \psi_n ψ n . The equilibrium weight of the Langevin process,
e − S E = Ψ I R 2 ↔ ∣ Ψ B D ∣ 2 , (82) e^{-S_E} = \Psi_{\mathrm{IR}}^2 \quad\leftrightarrow\quad\lvert\Psi_{\mathrm{BD}}\rvert^2, \tag*{(82)} e − S E = Ψ IR 2 ↔ ∣ Ψ BD ∣ 2 , (82)
is the Euclidean form of the Born-rule weight whose moments are the equal-time in-in correlators. The exact transformation formulas (59) and (63) , which convert the kernels of − log P -\log P − log P into connected correlators, are the Euclidean form of the familiar relations between wavefunction coefficients and in-in correlators [133 , 138 , 143 , 144 ] , ⟨ ϕ ⟩ = 1 / ( 2 Re ψ 2 ) \langle\phi\rangle= 1/(2\operatorname{Re}\psi_2) ⟨ ϕ ⟩ = 1/ ( 2 Re ψ 2 ) , ⟨ ϕ ϕ ⟩ c = − 2 Re ψ 3 / ∏ 2 Re ψ 2 \langle\phi\phi\rangle_c = -2\operatorname{Re}\psi_3/\prod2\operatorname{Re}\psi_2 ⟨ ϕϕ ⟩ c = − 2 Re ψ 3 / ∏ 2 Re ψ 2 , and so on; here all quantities are real and the real parts are absent. Equilibrium correlators of the Langevin process are thus the Euclidean counterparts of in-in correlators.
The finite-time distribution P = Ψ U V Ψ I R / Z P = \Psi_{\mathrm{UV}}\Psi_{\mathrm{IR}}/Z P = Ψ UV Ψ IR / Z of (42) continues to the product of Ψ B D [ Φ ; η ] \Psi_{\mathrm{BD}}[\Phi;\eta] Ψ BD [ Φ ; η ] with the amplitude to evolve from the configuration Φ \Phi Φ at time η \eta η to a vanishing field on the future boundary η = 0 \eta=0 η = 0 , which is what the Dirichlet condition at z = 0 z=0 z = 0 becomes. This is a bridge, not a quantity usually computed in cosmology, and the finite-cutoff couplings D ( n ) ( ϵ ) D^{(n)}(\epsilon) D ( n ) ( ϵ ) are its kernels: they describe the relaxation of the process from the sharp initial state δ [ Φ ] \delta[\Phi] δ [ Φ ] to the in-in equilibrium, not the wavefunction coefficients at finite time, which are ϵ \epsilon ϵ -independent for Weyl-invariant fields. The in-in weight itself corresponds to the equilibrium initial condition of remark (b), Ψ U V [ Φ ; 0 ] ∝ Ψ I R [ Φ ] \Psi_{\mathrm{UV}}[\Phi;0] \propto\Psi_{\mathrm{IR}}[\Phi] Ψ UV [ Φ ; 0 ] ∝ Ψ IR [ Φ ] , for which P = e − S E P=e^{-S_E} P = e − S E at all t t t . A Lorentzian construction that prepares the bra from the future, by extending the Poincaré patch, has recently been given in [145 ] ; the Heemskerk–Polchinski split is its Euclidean analogue, with the slab in the role of the future preparation.
The ground-state transform of § 1.3 C 2 also has a Lorentzian counterpart. For a real-time wavefunction Ψ ( η ) \Psi(\eta) Ψ ( η ) , Nelson’s stochastic mechanics [124 ] constructs a real diffusion whose density is ∣ Ψ ( η ) ∣ 2 |\Psi(\eta)|^2 ∣Ψ ( η ) ∣ 2 , with drift ∇ log ∣ Ψ ∣ \nabla\log|\Psi| ∇ log ∣Ψ∣ plus the current velocity ∇ arg Ψ \nabla\arg\Psi ∇ arg Ψ ; the Doob transform used here is the special case of a real wavefunction, for which the current vanishes. The relation between Nelson’s diffusion for the Bunch–Davies wavefunction and the Langevin equation of stochastic inflation, whose noise comes from modes crossing the horizon, will be discussed in [129 ] .
Stochastic inflation is where this structure is already in use. For a light scalar in de Sitter space, the long-wavelength modes perform a Langevin process in cosmic time, and their distribution obeys a Fokker–Planck equation whose equilibrium distribution is P e q ∝ e − 8 π 2 V / 3 H 4 P_{\mathrm{eq}} \propto e^{-8\pi^2 V/3H^4} P eq ∝ e − 8 π 2 V /3 H 4 [146 , 147 ] , the exponential of the Euclidean action on the four-sphere [148 ] . In recent years the Fokker–Planck equation and its corrections have been derived from quantum field theory in de Sitter space: from the Bunch–Davies wavefunction, by integrating out the short modes of P = Ψ B D ∗ Ψ B D P = \Psi_{\mathrm{BD}}^* \Psi_{\mathrm{BD}} P = Ψ BD ∗ Ψ BD [149 ] ; in the static patch, by tracing out the environment of a subhorizon observer [148 , 150 ] ; and as the dynamical renormalization group of Soft de Sitter Effective Theory [151 , 152 ] .
The identification of the holographic Wilsonian flow with a Langevin process is the Euclidean, anti-de Sitter version of the statement above, with two differences. In de Sitter space the Fokker–Planck equation is an effective description of the long-wavelength modes, obtained by splitting the field at a comoving cutoff, integrating out the short modes and approximating the result; the real-time evolution itself is oscillatory. In Euclidean signature the radial evolution e − z H R G e^{-zH_{\mathrm{RG}}} e − z H RG is a heat kernel, so the radial Schrödinger equation is a diffusion equation for all modes and at every z z z , and P = Ψ U V Ψ I R / Z P = \Psi_{\mathrm{UV}}\Psi_{\mathrm{IR}}/Z P = Ψ UV Ψ IR / Z obeys the Fokker–Planck equation exactly, with nothing integrated out. The second difference concerns the time variable. The continuation z → − i η z \to-i\eta z → − i η produces conformal time, in which the Bunch–Davies wavefunction of a light scalar depends on η \eta η , so that in the variables of this paper its drift is time dependent, the regime of remark (c); stochastic inflation obtains a fixed drift and a fixed equilibrium action by a different route, using the number of e-folds as time and keeping only super-horizon modes. Weyl invariance makes the drift time independent in conformal time for all modes, which is Proposition 1, and it also removes secular infrared growth; for conformally coupled scalars and gluons in dS4 _4 4 the stochastic approach therefore has no infrared content. Its content for gluons is the gauge sector, to which we now turn.
Gluons in dS4 _4 4
Yang–Mills theory in dS4 _4 4 is flat-space Yang–Mills theory in conformal time, and its late-time wavefunction coefficients have been computed in the same axial-gauge momentum-space formalism as the anti-de Sitter correlators of [66 ] ; see [36 ] and references therein. Up to the continuation factors, the equilibrium kernels (32) are twice these wavefunction coefficients, since e − S E = Ψ I R 2 e^{-S_{\mathrm{E}}} = \Psi_{\mathrm{IR}}^2 e − S E = Ψ IR 2 . The equilibrium correlators of the Langevin process are then the Euclidean counterparts of the in-in gluon correlators. The colour-ordered infrared couplings (74) , their total-energy pole and the flat-space amplitude at the pole are the familiar statements of the cosmological bootstrap [28 , 34 , 153 , 154 ] in this language.
The gauge-sector results translate as follows. Temporal gauge, A η = 0 A_\eta= 0 A η = 0 , which is the standard choice in cosmological gluon computations, is radial gauge, and by § IV A it is Parisi–Wu stochastic quantization: the longitudinal propagator, min ( z , z ′ ) \min(z,z') min ( z , z ′ ) up to the continuation factors, of these computations is the covariance of the Brownian motion of the gauge modes, which do not equilibrate. Gauge-invariant equilibrium data are unaffected by this, as in Parisi–Wu, but the finite-time distribution is not gauge invariant, and the entire gauge dependence of the quartic couplings on transverse data sits in the longitudinal zero mode of § 2.2 E. The continuation of Zwanziger’s gauge, A η = ( i α ) − 1 ∂ i A i A_\eta= (i\alpha)^{-1}\partial_i A_i A η = ( i α ) − 1 ∂ i A i , damps the gauge modes of the Euclidean process without ghosts and without Gribov copies in perturbation theory (§ IV C); the price is that the equilibrium distribution on transverse data differs from the naive gauge-fixed one by the shift Z s − Z t Z_s - Z_t Z s − Z t of (79) , which reflects the fact that transverse modes are not gauge invariant beyond linear order. Imposing the Gauss law, or minimizing over gauge rotations of the data, removes the zero mode and leaves the gauge-invariant content of § 2.2 E. We expect the same bookkeeping to be needed in a stochastic treatment of the gauge potentials, as opposed to the physical field strengths used in [155 , 156 ] , in particular for the non-abelian gauge fields of the inflationary models reviewed in [157 ] (gauge-flation, chromo-natural inflation), whose fluctuations propagate on a time-dependent homogeneous background and therefore belong to the time-dependent-drift regime of remark (c).
Finally, the Chern–Simons branch of § IIIB2 is parity odd, with opposite signs for the two helicities, and its gravitational cousin is the Kodama state [158 ] discussed alongside the Chern–Simons state in [121 ] ; we comment more on this in the next section.
Outlook
De Sitter space. The Lorentzian version of the dictionary is partly in place and partly not. The part that is in place is § VIA: P = ∣ Ψ B D ∣ 2 P = |\Psi_{\mathrm{BD}}|^2 P = ∣ Ψ BD ∣ 2 is the Born rule, and for light scalars the Fokker–Planck description of its long-wavelength modes, with its corrections, has been derived from quantum field theory in de Sitter space [148 , 149 , 150 , 151 , 152 ] . What this paper adds on the Euclidean side, the all-order tree-level proof for radially homogeneous bulk theories and the treatment of the gauge sector, has no counterpart there in this form. Although stochastic formalisms for gauge fields during inflation exist [155 , 156 , 159 ] , they are written for the long-wavelength physical electric and magnetic fields of abelian models: a treatment along the lines of this paper would in addition have to handle the random walk of the gauge modes in temporal gauge, the Gauss law, and a damping gauge of Zwanziger’s type, which is what § IV and § V E do in the Euclidean setting, and it would have to do so with a time-dependent drift: it would be interesting to see this correspondence made quantitative.
Loops. Proposition 1 is a tree-level statement, and we have not attempted a loop analysis in this paper. Such an analysis would have to address three points.
It would have to define the exact Ψ I R \Psi_{\mathrm{IR}} Ψ IR , i.e. the operator ordering and measure of the bulk radial path integral, including the Faddeev–Popov sector of the chosen bulk gauge. Given Ψ I R \Psi_{\mathrm{IR}} Ψ IR , the Jacobian term − 1 4 δ 2 S E δ A 2 -\frac{1}{4}\frac{\delta^{2}S_{\mathrm{E}}}{\delta A^{2}} − 4 1 δ A 2 δ 2 S E of (7) , i.e. the Itô–Stratonovich choice, is fixed by the identity H F P = H R G − E 0 H_{\mathrm{FP}}=H_{\mathrm{RG}}-E_{0} H FP = H RG − E 0 of § III C 2.
It would have to regularize the ultraviolet-divergent functional traces, such as Tr Ω \operatorname{Tr}\Omega Tr Ω and Tr δ 2 S E δ A 2 \operatorname{Tr}\frac{\delta^{2}S_{\mathrm{E}}}{\delta A^{2}} Tr δ A 2 δ 2 S E , which determine E 0 E_{0} E 0 , compatibly on both sides.
Beyond tree level, Yang–Mills theory in A d S 4 \mathrm{AdS}_{4} AdS 4 is in any case not equivalent to its flat-space counterpart: its phase structure depends on the coupling and on the boundary conditions [160 ] . It would have to account for the one-loop running of the Yang–Mills coupling. As the trace anomaly breaks the Weyl invariance that made S I R S_{\mathrm{IR}} S IR independent of ϵ \epsilon ϵ , this would mean that Ψ I R \Psi_{\mathrm{IR}} Ψ IR would acquire a logarithmic dependence on the cutoff, hence one should consider Langevin processes beyond time-homogeneous domain.
Gravitons. Analysis of time-inhomogeneous Langevin processes is already a prerequisite even at tree level for the analysis of gravitons, as Einstein gravity in A d S 4 \mathrm{AdS}_{4} AdS 4 is not Weyl invariant with the graviton bulk-to-boundary propagator in axial gauge being ( 1 + k z ) e − k z (1+kz)e^{-kz} ( 1 + k z ) e − k z [68 , 70 ] . Therefore, a stochastic description of graviton multi-trace flows requires the time-dependent construction of remark (c) together with a treatment of the gravitational constraints analogous to the Gauss law. Since graviton Witten diagrams in A d S 4 \mathrm{AdS}_{4} AdS 4 have been computed in the same momentum-space, axial-gauge formalism [68 , 70 ] , we expect explicit tree-level checks of the graviton analogue of (14) to be actually feasible: we hope to address this in the future.
Chern–Simons branch. It would be interesting to relate the Chern–Simons branch of § III B 2 to the self-dual limit of Yang–Mills theory in A d S 4 \mathrm{AdS}_{4} AdS 4 studied in [84 ] . There the θ \theta θ -angle enters through the boundary conditions [127 , 128 ] ; in our language, it leads to complex weights and complex Langevin dynamics [161 ] . On the de Sitter side, the branch has the Kodama state as its gravitational cousin [121 ] , which ties it to the de Sitter and graviton items above: it would be interesting to see such progress at the intersection of cosmology, gravity, and stochastic quantization.
Acknowledgments
The author would like to thank Savan Kharel for insightful discussions. Claude LLM Opus 5.5 made partial contributions at late stages of this work.60 The author takes full responsibility for the final result. In the early stages of this research, the author was supported by a VIDI grant of the Netherlands Organisation for Scientific Research (NWO) that is funded by the Dutch Ministry of Education, Culture and Science (OCW). In the later stages, he was supported by the BAGEP Award of the Science Academy in Türkiye and by the TUBITAK (The Scientific and Technological Research Council of Türkiye)⁶¹ 2232-b International Fellowship for Early Stage Researchers program, project number 122c153.
Appendices
The all-order correspondence in earlier work
Proposition 1 builds on several earlier works, and we want to be precise about what it adds.
Previous literature
Mansi, Mauri and Petkou conjectured that the bulk action is the Fokker–Planck action of a stochastic process in radial time, with the classical boundary action equal to twice the boundary effective action [39 ] . They observed that this can hold only for special actions. They verified the identification for the conformally coupled ϕ 4 \phi^{4} ϕ 4 theory in A d S 4 \mathrm{AdS}_{4} AdS 4 , using its exact bulk solutions and invoking Stokes-theorem structure of stochastic quantization [95 ] .
Oh and Jatkar recast the conjecture as an equality between the Fokker–Planck and radial Hamiltonians, with Ψ S / Ψ E = P E S / 2 \Psi_{\mathrm{S}}/\Psi_{\mathrm{E}}=P_{\mathrm{E}}^{S/2} Ψ S / Ψ E = P E S /2 identified with Ψ U V = e − S B \Psi_{\mathrm{UV}}=e^{-S_{\mathrm{B}}} Ψ UV = e − S B and S E = − 2 I O S S_{\mathrm{E}}=-2I_{\mathrm{OS}} S E = − 2 I OS , and described it as partially proved by their examples [37 ] .
The general argument is due to Oh [48 ] : for a scalar bulk action whose Lagrangian has no explicit radial dependence (the conformally coupled scalar in AdS6 _6 6 being the worked example), the on-shell boundary action obeys the zero-energy Hamilton–Jacobi equation, the bulk action is a perfect square up to a total derivative, it coincides with the Fokker–Planck action in the classical limit once t = r t=r t = r and S E = 2 I o s S_E=2I_{os} S E = 2 I os , and S B = − ln P − 1 2 S E S_B=-\ln P-\frac{1}{2}S_E S B = − ln P − 2 1 S E , i.e. (14) , follows for all n n n . Together with the Onsager–Machlup representation of P P P with a sharp initial state given in [37 ] , this is the content of the proof of Proposition 1. The explicit checks in [48 ] were made for the three-point function. Further examples have since been checked [44 , 49 , 50 , 51 ] . For massive scalars, the Euclidean action of [51 ] has kernels that depend on the stochastic time.
b. Main contribution
We do not provide a new relation with Proposition 1. What we add is (i) its formulation as a statement about wavefunctions, P = Ψ U V Ψ I R / Z P=\Psi_{\mathrm{UV}}\Psi_{\mathrm{IR}}/Z P = Ψ UV Ψ IR / Z , which identifies the Langevin process as the ground-state (Doob) transform of the radial evolution (§ 1.3 C 2) — the classic correspondence between Euclidean field theory and ground-state diffusions [124 , 125 , 126 ] , which appears in stochastic quantization as e − S = ∣ Ψ 0 ∣ 2 e^{-S}=|\Psi_0|^2 e − S = ∣ Ψ 0 ∣ 2 [162 ] ; (ii) the consequences of this formulation: the initial distribution that corresponds to a general ultraviolet boundary condition (remark (b)), the fact that without radial homogeneity the same construction holds with a time-dependent drift (remark (c), as in [51 ] ), and the form of the loop-level statement (remark (d)); and (iii) its extension to gauge fields, where the gauge-fixed drift is not a gradient (§ IV).
Radial homogeneity gives the stationarity ∂ ϵ S I R = 0 \partial_\epsilon S_{\mathrm{IR}}=0 ∂ ϵ S IR = 0 of the radial Hamilton–Jacobi equation. Completing the square in the Onsager–Machlup action then reproduces the bulk action, and the Stokes term accounts for the shift − 1 2 G ( n ) -\frac{1}{2}G^{(n)} − 2 1 G ( n ) . This completion is the Bogomol’nyi, or “fake superpotential”, structure of holographic RG flows [101 , 118 , 119 , 120 ] . The proof also identifies the Langevin process as the ground-state (Doob) transform of the radial evolution (§ 1.3 C 2). This is the classic correspondence between Euclidean field theory and ground-state diffusions [124 , 125 , 126 ] , which appears in stochastic quantization as e − S = ∣ Ψ 0 ∣ 2 e^{-S}=|\Psi_0|^2 e − S = ∣ Ψ 0 ∣ 2 [162 ] . Two consequences follow. First, stationarity is not needed for the existence of the dictionary: without it, the same construction holds with a drift that depends on the stochastic time, as in [51 ] (remark (c)). What stationarity ensures is that S E S_E S E is a fixed action, so that the Langevin process is a stochastic quantization in the usual sense. Second, the construction fixes the initial data (remark (b)), and it extends to gauge fields whose gauge-fixed drift is not a gradient (§ IV).
Colour decomposition in Yang–Mills
Let us consider a gauge theory with three-point and four-point interactions. For the antisymmetric structure constants f a b c f^{abc} f ab c , any n n n -point Witten diagram reads as
W α a 1 … a n = W α ∑ π ∈ S n − 2 b α , π f a 1 a π ( 2 ) c 1 f a π ( 3 ) c 1 c 2 × ⋯ f c n − 3 a π ( n − 1 ) a n (83) \mathcal{W}_{\alpha}^{a_1\ldots a_n}=W_\alpha\sum_{\pi\in S_{n-2}}b_{\alpha,\pi}f^{a_1a_{\pi(2)}c_1}f^{a_{\pi(3)}c_1c_2}\times\cdots f^{c_{n-3}a_{\pi(n-1)}a_n} \tag*{(83)} W α a 1 … a n = W α π ∈ S n − 2 ∑ b α , π f a 1 a π ( 2 ) c 1 f a π ( 3 ) c 1 c 2 × ⋯ f c n − 3 a π ( n − 1 ) a n (83)
where π \pi π is a permutation of { 2 , 3 , … , n − 1 } \{2,3,\ldots,n-1\} { 2 , 3 , … , n − 1 } and c i c_i c i are contracted dummy indices.62 Here, b α , π ∈ { 0 , ± 1 } b_{\alpha,\pi}\in\{0,\pm1\} b α , π ∈ { 0 , ± 1 } depends on the permutation π \pi π [163 ] and on the index α \alpha α , which enumerates the Witten diagrams for given n n n external legs with three-point vertices only.63
For given n n n external legs, we obtain the full amplitude by summing over all Witten diagrams:
A a 1 … a n = ∑ α W α a 1 … a n = ∑ π ∈ S n − 2 A ( 1 , π ( 2 ) , … , π ( n − 1 ) , n ) f a 1 a π ( 2 ) c 1 ⋯ f c n − 3 a π ( n − 1 ) a n (84) \begin{aligned}
\mathcal{A}^{a_1\ldots a_n}=\sum_\alpha\mathcal{W}^{a_1\ldots a_n}_\alpha\\
&=\sum_{\pi\in S_{n-2}}\mathcal{A}(1,\pi(2),\ldots,\pi(n-1),n)f^{a_1a_{\pi(2)}c_1}\cdots f^{c_{n-3}a_{\pi(n-1)}a_n} \tag*{(84)}
\end{aligned} A a 1 … a n = α ∑ W α a 1 … a n = π ∈ S n − 2 ∑ A ( 1 , π ( 2 ) , … , π ( n − 1 ) , n ) f a 1 a π ( 2 ) c 1 ⋯ f c n − 3 a π ( n − 1 ) a n (84)
where
A ( 1 , π ( 2 ) , … , π ( n − 1 ) , n ) ≡ ∑ α W α b α , π (85) \mathcal{A}(1,\pi(2),\ldots,\pi(n-1),n)\equiv\sum_\alpha W_\alpha b_{\alpha,\pi} \tag*{(85)} A ( 1 , π ( 2 ) , … , π ( n − 1 ) , n ) ≡ α ∑ W α b α , π (85)
is the colour-ordered amplitude in the Kleiss–Kuijf (KK) basis. (A2) is usually referred to as the Del Duca–Dixon–Maltoni (DDM) decomposition of the colour-dressed amplitude [163 , 164 , 165 , 166 ] .64
The DDM decomposition is convenient as it involves only ( n − 2 ) ! (n-2)! ( n − 2 )! terms. A more conventional basis is the trace basis,
A a 1 … a n = ∑ σ ∈ S n − 1 A ( 1 , σ ( 2 ) , … , σ ( n ) ) Tr ( T a 1 T a σ ( 2 ) ⋯ T a σ ( n ) ) (86) \mathcal{A}^{a_1\ldots a_n}=\sum_{\sigma\in S_{n-1}}\mathcal{A}(1,\sigma(2),\ldots,\sigma(n))\operatorname{Tr}\left(T^{a_1}T^{a_{\sigma(2)}}\cdots T^{a_{\sigma(n)}}\right) \tag*{(86)} A a 1 … a n = σ ∈ S n − 1 ∑ A ( 1 , σ ( 2 ) , … , σ ( n )) Tr ( T a 1 T a σ ( 2 ) ⋯ T a σ ( n ) ) (86)
where T a T^a T a are the generators of SU(N N N ) in the fundamental representation, normalized as Tr ( T a T b ) = δ a b \operatorname{Tr}(T^aT^b)=\delta^{ab} Tr ( T a T b ) = δ ab and [ T a , T b ] = i 2 f a b c T c [T^a,T^b]=i\sqrt{2}f^{abc}T^c [ T a , T b ] = i 2 f ab c T c , so that f a b c = − i 2 [ Tr ( T a T b T c ) − Tr ( T b T a T c ) ] f^{abc}=-\frac{i}{\sqrt{2}}[\operatorname{Tr}(T^aT^bT^c)-\operatorname{Tr}(T^bT^aT^c)] f ab c = − 2 i [ Tr ( T a T b T c ) − Tr ( T b T a T c )] . Using ∑ a ( T a ) i 1 j 1 ( T a ) i 2 j 2 = δ i 1 j 2 δ i 2 j 1 − δ i 1 j 1 δ i 2 j 2 / N \sum_a(T^a)_{i_1}^{j_1}(T^a)_{i_2}^{j_2}=\delta_{i_1}^{j_2}\delta_{i_2}^{j_1}-\delta_{i_1}^{j_1}\delta_{i_2}^{j_2}/N ∑ a ( T a ) i 1 j 1 ( T a ) i 2 j 2 = δ i 1 j 2 δ i 2 j 1 − δ i 1 j 1 δ i 2 j 2 / N , one can go back and forth between the DDM and trace bases. The partial amplitudes agree once f a b c f^{abc} f ab c in (A2) is replaced by f ~ a b c = i 2 f a b c = Tr ( [ T a , T b ] T c ) \tilde f^{abc}=i\sqrt{2}f^{abc}=\operatorname{Tr}([T^a,T^b]T^c) f ~ ab c = i 2 f ab c = Tr ([ T a , T b ] T c ) . With f a b c f^{abc} f ab c itself, the DDM coefficients are ( i 2 ) n − 2 (i\sqrt{2})^{n-2} ( i 2 ) n − 2 times the trace-basis partial amplitudes. The Kleiss–Kuijf relations map ( n − 1 ) ! (n-1)! ( n − 1 )! partial amplitudes of trace basis to the ( n − 2 ) ! (n-2)! ( n − 2 )! of DDM basis [163 , 166 ] .
Scalar seeds: slab integrals and Langevin diagrams
This appendix collects the scalar ingredients of § 2.2 : the slab propagators and the identities they obey, the closed forms of the exchange seed, and the Langevin diagrams whose sum gives the stochastic quadruple-trace weight. The seed theory is a massless scalar on the flat half-space with cubic coupling λ 3 ! ϕ 3 \frac{\lambda}{3!}\phi^3 3 ! λ ϕ 3 , i.e. the conformally coupled ϕ 3 \phi^3 ϕ 3 theory in AdS6 _6 6 after rescaling; its radial integrals do not depend on the boundary dimension. Its equilibrium action has G ( 3 ) = λ / ( 3 k 123 ) G^{(3)}=\lambda/(3k_{123}) G ( 3 ) = λ / ( 3 k 123 ) and, in the s s s -channel, G s ( 4 ) = − λ 2 / ( 12 k 1234 k 1212 k 3412 ) G_s^{(4)}=-\lambda^2/(12k_{1234}k_{1212}k_{3412}) G s ( 4 ) = − λ 2 / ( 12 k 1234 k 1212 k 3412 ) , so that the slab coupling is D s ( 4 ) = − λ 2 24 F ( ϵ ) D_s^{(4)}=-\frac{\lambda^2}{24}F(\epsilon) D s ( 4 ) = − 24 λ 2 F ( ϵ ) with F F F as in (62) .
Slab propagators and bridge identities. The bulk-to-cutoff and bulk-to-bulk propagators of the slab 0 < z < ϵ 0<z<\epsilon 0 < z < ϵ are S k S_k S k and G k U V G_k^{\mathrm{UV}} G k UV of (54) . Since G k U V G_k^{\mathrm{UV}} G k UV is the Green’s function of − ∂ z 2 + k 2 -\partial_z^2+k^2 − ∂ z 2 + k 2 with Dirichlet conditions at both ends,
( ∂ z + k ) ( ∂ w + k ) G k U V ( z , w ) = δ ( z − w ) − k e k ( z + w − ϵ ) sinh k ϵ , (87) (\partial_z+k)(\partial_w+k)G_k^{\mathrm{UV}}(z,w)=\delta(z-w)-\frac{k e^{k(z+w-\epsilon)}}{\sinh k\epsilon}, \tag*{(87)} ( ∂ z + k ) ( ∂ w + k ) G k UV ( z , w ) = δ ( z − w ) − sinh k ϵ k e k ( z + w − ϵ ) , (87)
as follows from ( ∂ z + k ) sinh k z = k e k z (\partial_z+k)\sinh kz=k e^{kz} ( ∂ z + k ) sinh k z = k e k z and ( ∂ w + k ) sinh k ( ϵ − w ) = − k e − k ( ϵ − w ) (\partial_w+k)\sinh k(\epsilon-w)=-k e^{-k(\epsilon-w)} ( ∂ w + k ) sinh k ( ϵ − w ) = − k e − k ( ϵ − w ) together with the jump − 1 -1 − 1 of ∂ z G k U V \partial_zG_k^{\mathrm{UV}} ∂ z G k UV across z = w z=w z = w . Its k → 0 k\to0 k → 0 limit is ∂ z ∂ w G 0 U V = δ ( z − w ) − 1 / ϵ \partial_z\partial_wG_0^{\mathrm{UV}}=\delta(z-w)-1/\epsilon ∂ z ∂ w G 0 UV = δ ( z − w ) − 1/ ϵ , the covariance identity of the Brownian bridge used in (76) ; at finite k k k , (A5) is used in (77) with k → γ k\to\gamma k → γ .
Closed form of the exchange seed. Expanding the products of sinh k r z \sinh k_rz sinh k r z into hyperbolic cosines reduces F ( ϵ ) F(\epsilon) F ( ϵ ) to the function
J s ( a , b ) : = ∫ 0 ϵ ∫ 0 ϵ d z d w cosh ( a z ) G s U V ( z , w ) cosh ( b w ) , (88) \mathcal{J}_s(a,b):=\int_0^\epsilon\int_0^\epsilon dz\,dw\,\cosh(az)G_s^{\mathrm{UV}}(z,w)\cosh(bw), \tag*{(88)} J s ( a , b ) := ∫ 0 ϵ ∫ 0 ϵ d z d w cosh ( a z ) G s UV ( z , w ) cosh ( b w ) , (88)
as in (66) . The integrals are elementary and give
J s ( a , b ) = X s ( a , b ) ( s 2 − a 2 ) ( s 2 − b 2 ) , (89) \mathcal{J}_s(a,b)=\frac{X_s(a,b)}{(s^2-a^2)(s^2-b^2)}, \tag*{(89)} J s ( a , b ) = ( s 2 − a 2 ) ( s 2 − b 2 ) X s ( a , b ) , (89)
X s ( a , b ) = s ( ch a + ch b − ch s − ch a ch b ch s ) sh s + ( s 2 − b 2 ) a sh a ch b − ( s 2 − a 2 ) b ch a sh b a 2 − b 2 , (90) X_s(a,b)=\frac{s(\operatorname{ch}_a+\operatorname{ch}_b-\operatorname{ch}_s-\operatorname{ch}_a\operatorname{ch}_b\operatorname{ch}_s)}{\operatorname{sh}_s}
+\frac{(s^2-b^2)a\operatorname{sh}_a\operatorname{ch}_b-(s^2-a^2)b\operatorname{ch}_a\operatorname{sh}_b}{a^2-b^2}, \tag*{(90)} X s ( a , b ) = sh s s ( ch a + ch b − ch s − ch a ch b ch s ) + a 2 − b 2 ( s 2 − b 2 ) a sh a ch b − ( s 2 − a 2 ) b ch a sh b , (90)
with ch x ≡ cosh x ϵ \operatorname{ch}_x\equiv\cosh x\epsilon ch x ≡ cosh x ϵ and sh x ≡ sinh x ϵ \operatorname{sh}_x\equiv\sinh x\epsilon sh x ≡ sinh x ϵ . J s \mathcal{J}_s J s is symmetric in a ↔ b a\leftrightarrow b a ↔ b and even in each argument, and its apparent poles at a 2 = b 2 a^2=b^2 a 2 = b 2 and s 2 ∈ { a 2 , b 2 } s^2\in\{a^2,b^2\} s 2 ∈ { a 2 , b 2 } are removable. The longitudinal seed F ( 0 ) = F ∣ s → 0 F^{(0)}=F|_{s\to0} F ( 0 ) = F ∣ s → 0 follows from
J 0 ( a , b ) = 1 a 2 b 2 [ ch a + ch b − 1 − ch a ch b ϵ + a b ( a ch a sh b − b sh a ch b ) a 2 − b 2 ] . (91) \mathcal{J}_0(a,b)=\frac{1}{a^2b^2}\left[\frac{\operatorname{ch}_a+\operatorname{ch}_b-1-\operatorname{ch}_a\operatorname{ch}_b}{\epsilon}
+\frac{ab(a\operatorname{ch}_a\operatorname{sh}_b-b\operatorname{sh}_a\operatorname{ch}_b)}{a^2-b^2}\right]. \tag*{(91)} J 0 ( a , b ) = a 2 b 2 1 [ ϵ ch a + ch b − 1 − ch a ch b + a 2 − b 2 ab ( a ch a sh b − b sh a ch b ) ] . (91)
Langevin diagrams for the exchange identity. On the Langevin side, R k ( τ ) = e − k τ θ ( τ ) R_k(\tau)=e^{-k\tau}\theta(\tau) R k ( τ ) = e − k τ θ ( τ ) and G k ( t , t ′ ) G_k(t,t') G k ( t , t ′ ) are as in (23) , and g k ( t ) = G k ( t , t ) g_k(t)=G_k(t,t) g k ( t ) = G k ( t , t ) . At order λ 2 \lambda^2 λ 2 , the s s s -channel part of the connected stochastic four-point function is a sum of three terms. With r ∈ { 1 , 2 } r\in\{1,2\} r ∈ { 1 , 2 } , r ′ ∈ { 3 , 4 } r'\in\{3,4\} r ′ ∈ { 3 , 4 } , r ˉ \bar r r ˉ the partner of r r r in the channel, κ ≡ k 12 \kappa\equiv k_{12} κ ≡ k 12 , and G 3 L ≡ G ( 3 ) ( k 1 , k 2 , κ ) G_3^L\equiv G^{(3)}(k_1,k_2,\kappa) G 3 L ≡ G ( 3 ) ( k 1 , k 2 , κ ) , G 3 R ≡ G ( 3 ) ( κ , k 3 , k 4 ) G_3^R\equiv G^{(3)}(\kappa,k_3,k_4) G 3 R ≡ G ( 3 ) ( κ , k 3 , k 4 ) :
A = 9 G 3 L G 3 R [ ∑ r ∫ 0 t d s R k r ( t − s ) G k r ˉ ( s , t ) ∫ 0 s d s ′ R κ ( s − s ′ ) G k 3 ( s ′ , t ) G k 4 ( s ′ , t ) + ( 12 ↔ 34 ) ] , (92) A=9G_3^LG_3^R\left[\sum_r\int_0^t ds\,R_{k_r}(t-s)G_{k_{\bar r}}(s,t)\int_0^s ds'\,R_\kappa(s-s')G_{k_3}(s',t)G_{k_4}(s',t)+(12\leftrightarrow34)\right], \tag*{(92)} A = 9 G 3 L G 3 R [ r ∑ ∫ 0 t d s R k r ( t − s ) G k r ˉ ( s , t ) ∫ 0 s d s ′ R κ ( s − s ′ ) G k 3 ( s ′ , t ) G k 4 ( s ′ , t ) + ( 12 ↔ 34 ) ] , (92)
B = 9 G 3 L G 3 R ∑ r , r ′ ∫ 0 t d s ∫ 0 t d s ′ R k r ( t − s ) R k r ′ ( t − s ′ ) G k r ˉ ( s , t ) G κ ( s , s ′ ) G k r ˉ ′ ( s ′ , t ) , (93) B=9G_3^LG_3^R\sum_{r,r'}\int_0^t ds\int_0^t ds'\,R_{k_r}(t-s)R_{k_{r'}}(t-s')G_{k_{\bar r}}(s,t)G_\kappa(s,s')G_{k_{\bar r'}}(s',t), \tag*{(93)} B = 9 G 3 L G 3 R r , r ′ ∑ ∫ 0 t d s ∫ 0 t d s ′ R k r ( t − s ) R k r ′ ( t − s ′ ) G k r ˉ ( s , t ) G κ ( s , s ′ ) G k r ˉ ′ ( s ′ , t ) , (93)
C = − 12 G s ( 4 ) ∑ r = 1 4 ∫ 0 t d s R k r ( t − s ) ∏ r ′ ′ ≠ r G k r ′ ′ ( s , t ) . (94) C=-12G_s^{(4)}\sum_{r=1}^{4}\int_0^t ds\,R_{k_r}(t-s)\prod_{r''\ne r}G_{k_{r''}}(s,t). \tag*{(94)} C = − 12 G s ( 4 ) r = 1 ∑ 4 ∫ 0 t d s R k r ( t − s ) r ′′ = r ∏ G k r ′′ ( s , t ) . (94)
A A A has a response function as internal line and arises from iterating the cubic drift; B B B has a correlator as internal line; C C C comes from the quartic term of S E S_E S E . Inserting ⟨ ϕ 4 ⟩ c = A + B + C \langle\phi^4\rangle_c=A+B+C ⟨ ϕ 4 ⟩ c = A + B + C into (63) and comparing with D s ( 4 ) = − λ 2 24 F ( t ) D_s^{(4)}=-\frac{\lambda^2}{24}F(t) D s ( 4 ) = − 24 λ 2 F ( t ) gives D s ( 4 ) = P s ( 4 ) − 1 2 G s ( 4 ) D_s^{(4)}=P_s^{(4)}-\frac{1}{2}G_s^{(4)} D s ( 4 ) = P s ( 4 ) − 2 1 G s ( 4 ) , the exchange identity behind (64) . Proposition 1 guarantees it; we have also verified it directly for generic norms and in the limits κ → 0 \kappa\to0 κ → 0 and k 2 → 0 k_2\to0 k 2 → 0 (see the ancillary file). A direct analytic evaluation of A + B + C A+B+C A + B + C , inserting (65) and (A7) and (A8) together with the elementary time integrals of the response functions and correlators, would turn this into an algebraic identity; since Proposition 1 already establishes the result, we leave it to future work.
WKB hierarchy for the radial ground state
For completeness we recall how the loop corrections to a ground-state wavefunction are determined by its tree-level part, and then record the form this takes for the kernels of S E S_E S E . Write Ψ I R = e − S E / 2 \Psi_{\mathrm{IR}}=e^{-S_E/2} Ψ IR = e − S E /2 with S E = S E ( 0 ) + S E ( 1 ) + ⋯ S_E=S_E^{(0)}+S_E^{(1)}+\cdots S E = S E ( 0 ) + S E ( 1 ) + ⋯ ordered by loops, and H R G = ∫ d d x [ − 1 2 δ 2 δ ϕ 2 + V ] H_{RG}=\int d^dx[-\frac{1}{2}\frac{\delta^2}{\delta\phi^2}+\mathcal{V}] H R G = ∫ d d x [ − 2 1 δ ϕ 2 δ 2 + V ] . Since − 1 2 ∇ 2 e − S E / 2 = [ 1 8 ( ∇ S E ) 2 − 1 4 ∇ 2 S E ] e − S E / 2 -\frac{1}{2}\nabla^2e^{-S_E/2}=[\frac{1}{8}(\nabla S_E)^2-\frac{1}{4}\nabla^2S_E]e^{-S_E/2} − 2 1 ∇ 2 e − S E /2 = [ 8 1 ( ∇ S E ) 2 − 4 1 ∇ 2 S E ] e − S E /2 , the eigenvalue equation H R G Ψ I R = E 0 Ψ I R H_{RG}\Psi_{\mathrm{IR}}=E_0\Psi_{\mathrm{IR}} H R G Ψ IR = E 0 Ψ IR reads
V = 1 8 ( ∇ S E ) 2 − 1 4 ∇ 2 S E + E 0 , (95) \mathcal{V}=\frac{1}{8}(\nabla S_E)^2-\frac{1}{4}\nabla^2S_E+E_0, \tag*{(95)} V = 8 1 ( ∇ S E ) 2 − 4 1 ∇ 2 S E + E 0 , (95)
which is H F P = H R G − E 0 H_{FP}=H_{RG}-E_0 H F P = H R G − E 0 with H F P H_{FP} H F P as in (7) . At lowest order this is the Hamilton–Jacobi equation (30) , 1 8 ( ∇ S E ( 0 ) ) 2 = V \frac{1}{8}(\nabla S_E^{(0)})^2=\mathcal{V} 8 1 ( ∇ S E ( 0 ) ) 2 = V ; at the next order it is the transport equation (44) . These are the first two equations of the WKB hierarchy of the Euclidean Schrödinger equation [168 , 169 ] , here for the functional Schrödinger equation [170 , 171 , 172 ] of the radial evolution [112 ] .
Equation (44) is linear and first order in S E ( 1 ) S_{\mathrm{E}}^{(1)} S E ( 1 ) , so it is solved by the method of characteristics. The characteristics are the integral curves of ∇ S E ( 0 ) \nabla S_{\mathrm{E}}^{(0)} ∇ S E ( 0 ) , i.e. the noiseless Langevin flow ∂ z Φ = − 1 2 ∇ S E ( 0 ) \partial_z\Phi=-\frac{1}{2}\nabla S_{\mathrm{E}}^{(0)} ∂ z Φ = − 2 1 ∇ S E ( 0 ) , which are the regular classical solutions of § III B 1. Along such a curve
∂ z S E ( 1 ) [ Φ ( z ) ] = − 1 2 ∇ S E ( 0 ) ⋅ ∇ S E ( 1 ) = − 1 2 ( ∇ 2 S E ( 0 ) − 4 E 0 ) , \partial_z S_{\mathrm{E}}^{(1)}[\Phi(z)]=-\frac{1}{2}\nabla S_{\mathrm{E}}^{(0)}\cdot\nabla S_{\mathrm{E}}^{(1)}=-\frac{1}{2}(\nabla^2 S_{\mathrm{E}}^{(0)}-4E_0), ∂ z S E ( 1 ) [ Φ ( z )] = − 2 1 ∇ S E ( 0 ) ⋅ ∇ S E ( 1 ) = − 2 1 ( ∇ 2 S E ( 0 ) − 4 E 0 ) ,
hence
S E ( 1 ) [ Φ ] = 1 2 ∫ 0 ∞ d z [ ∫ d d x δ 2 S E ( 0 ) δ Φ ( x ) δ Φ ( x ) ∣ Φ c l ( z ) − 4 E 0 ] + S E ( 1 ) [ 0 ] , (96) S_{\mathrm{E}}^{(1)}[\Phi]=\frac{1}{2}\int_0^\infty dz\left[\int d^d x\,\frac{\delta^2 S_{\mathrm{E}}^{(0)}}{\delta\Phi(x)\delta\Phi(x)}\bigg|_{\Phi_{\mathrm{cl}}(z)}-4E_0\right]+S_{\mathrm{E}}^{(1)}[0], \tag*{(96)} S E ( 1 ) [ Φ ] = 2 1 ∫ 0 ∞ d z [ ∫ d d x δ Φ ( x ) δ Φ ( x ) δ 2 S E ( 0 ) Φ cl ( z ) − 4 E 0 ] + S E ( 1 ) [ 0 ] , (96)
where Φ c l ( z ) \Phi_{\mathrm{cl}}(z) Φ cl ( z ) is the regular solution with Φ c l ( 0 ) = Φ \Phi_{\mathrm{cl}}(0)=\Phi Φ cl ( 0 ) = Φ . Convergence at the fixed point Φ c l ( ∞ ) = 0 \Phi_{\mathrm{cl}}(\infty)=0 Φ cl ( ∞ ) = 0 fixes E 0 = 1 4 ∇ 2 S E ( 0 ) [ 0 ] = 1 2 Tr Ω E_0=\frac{1}{4}\nabla^2 S_{\mathrm{E}}^{(0)}[0]=\frac{1}{2}\operatorname{Tr}\Omega E 0 = 4 1 ∇ 2 S E ( 0 ) [ 0 ] = 2 1 Tr Ω , the zero-point energy; for free fields S E ( 1 ) S_{\mathrm{E}}^{(1)} S E ( 1 ) is a constant. Equation (A11) is the one-loop fluctuation determinant around the regular solution, written as a radial integral in the manner of Gel’fand–Yaglom [173 ] .
Expanded in the kernels of (14) , with S E ( L ) = ∑ n G ( n , L ) Φ n S_{\mathrm{E}}^{(L)}=\sum_n G^{(n,L)}\Phi^n S E ( L ) = ∑ n G ( n , L ) Φ n , the term of order Φ n \Phi^n Φ n of (44) reads
2 k 1 ⋯ n G ( n , 1 ) + ⋯ = ( n + 1 ) ( n + 2 ) ∫ p G ( n + 2 , 0 ) ( k 1 , … , k n , p , − p ) (97) 2k_{1\cdots n}G^{(n,1)}+\cdots=(n+1)(n+2)\int_p G^{(n+2,0)}(k_1,\ldots,k_n,p,-p) \tag*{(97)} 2 k 1 ⋯ n G ( n , 1 ) + ⋯ = ( n + 1 ) ( n + 2 ) ∫ p G ( n + 2 , 0 ) ( k 1 , … , k n , p , − p ) (97)
for transverse legs, where the dots are products of tree kernels with lower-point one-loop kernels: the one-loop n n n -point kernel is the tree ( n + 2 ) (n+2) ( n + 2 ) -point kernel with two legs sewn, divided by the total energy of the external legs. For a quartic contact kernel, (A12) reproduces the one-loop tadpole ∫ p ∫ 0 ∞ d z e − 2 k z G p ( z , z ) = ∫ p 1 2 k ( 2 k + 2 p ) \int_p\int_0^\infty dz\,e^{-2kz}G_p(z,z)=\int_p\frac{1}{2k(2k+2p)} ∫ p ∫ 0 ∞ d z e − 2 k z G p ( z , z ) = ∫ p 2 k ( 2 k + 2 p ) 1 of the double-trace coupling, the factor 1 / 2 p 1/2p 1/2 p of the propagator cancelling against its 1 − e − 2 p z 1-e^{-2pz} 1 − e − 2 p z . Equation (A12), applied to the closed-form quartic kernel (32) of Yang–Mills theory with the regularization of the sewn legs that point (ii) of § VII calls for, is the starting point of the one-loop computation in progress [129 ] ; we do not pursue it here.
References