betaFeedback

WorksPhysics and Astronomy

Weyl-induced ultrasoft renormalization of the static QCD potential

Abstract

We derive the leading curvature-dependent ultrasoft anomalous dimension of the static color-singlet potential in potential nonrelativistic QCD. In a Ricci-flat background the electric Weyl tensor introduces a t−2t^{-2} term in the short-time chromoelectric correlator. The resulting dipole self-energy is ultraviolet divergent at order αs2α_s^2 in the curvature sector, two orders earlier than the flat-space contribution. For W=CμανβrμuαrνuβW=C_{μανβ}r^μu^αr^νu^β and Vs(W)=cWW/rV_s^{(W)}=c_W W/r, we obtain ν ∂νcW=CFCAαs2/(3π)ν\,\partial_νc_W=C_F C_Aα_s^2/(3π). We calculate the compensating soft infrared pole and the finite one-loop matching coefficient in an optical dimensional subtraction scheme. A direct physical-metric fermion calculation reproduces the finite light-flavor term. We also exhibit the cancellation of the leading regulated infrared overlap and the compensating transformation of soft and ultrasoft constants under an evanescent curvature rescaling. In the perturbative adiabatic regime the anomalous dimension fixes the coefficient of a curvature-dependent αs2ln⁡αsα_s^2\lnα_s term in the static energy.

Introduction

The static QCD potential is a Wilson coefficient rather than an observable by itself. Its infrared sensitivity is canceled by the ultraviolet sensitivity of dynamical gluons at the ultrasoft scale. This separation, made explicit in potential nonrelativistic QCD (pNRQCD) [1, 2, 3], accounts for the first infrared divergence of the flat-space static potential at order αs4/r\alpha_s^4/r [4, 5, 6]. The ultrasoft dipole self-energy carries three powers of the singlet–octet energy gap. A background that changes the short-time chromoelectric correlator can therefore change the order at which the potential requires ultrasoft renormalization.

We consider this problem in a classical gravitational background. In vacuum the local tidal field is described by the Weyl tensor, and the propagation of gauge fields introduces a curvature dependence distinct from the gravitational force on the constituent masses. The separation of dynamical scales has a precedent in thermal pNRQCD [7], although the effect considered here is geometric and arises in the state-independent singularity of the field-strength correlator, rather than from thermal populations.

Gravitational corrections to electromagnetic two-body interactions have been calculated in weak-field geometries [8]. Curved-space gluodynamics has also been studied using low-energy gluon-dilaton models [9]. At the level of perturbative amplitudes, gravitational couplings of quarks and gluons are encoded in the stress-tensor–current–current correlator [10, 11]; the corresponding massless QED curvature effective action is known [12]. These calculations address different aspects of the gravitational response of gauge theories. Here we match a static source pair in an external Ricci-flat physical metric and determine the renormalization of its singlet potential by the chromoelectric dipole interaction. In particular, the soft Wilson-loop contributions and their ultrasoft subtraction are intrinsic to the static matching problem.

The static effective theory is formulated in Sec. II. Section III establishes the tree-level interaction, and Sec. IV computes the one-loop polarization and connected Wilson-line contributions. In Sec. V we derive the ultrasoft anomalous dimension, perform the infrared subtraction, and obtain the matched potential. Section VI gives an independent physical-metric calculation of the light-flavor term. We discuss the resulting perturbative hierarchy in Sec. VII.

Static effective theory in a tidal field

We set ℏ=c=1\hbar= c = 1, use signature (−,+,+,+)(-,+,+,+) and the curvature convention Rρσμν=∂μΓνσρ−∂νΓμσρ+⋯R^\rho{}_{\sigma\mu\nu} = \partial_\mu\Gamma^\rho_{\nu\sigma} - \partial_\nu\Gamma^\rho_{\mu\sigma} + \cdots. In a parallel tetrad along a central geodesic let Eij=Ri0j0E_{ij} = R_{i0j0}. For a Ricci-flat purely electric Weyl field, Eii=0E^i{}_i = 0 and R0ijk=0R_{0ijk} = 0. At quadratic Fermi order, neglecting curvature derivatives, the Lorentzian metric takes the frozen form

ds2=−N2dt2+γijdxidxj,(1)ds^2 = -N^2dt^2 + \gamma_{ij}dx^idx^j, \tag*{(1)}
N=1+12Eijxixj,γij=δij−13Rikjlxkxl.(2)N = 1 + \frac{1}{2}E_{ij}x^ix^j,\qquad\gamma_{ij} = \delta_{ij} - \frac{1}{3}R_{ikjl}x^kx^l. \tag*{(2)}

Two static sources are placed at x±=±r/2x_\pm= \pm r/2. The time coordinate tt equals proper time at the midpoint, not at either source. No thermal state or stationary Schwarzschild interior is assumed. We match at a fixed point of the central geodesic, treating the curvature as an external slowly varying field at the soft scale. The scale assumptions are

mQ≫r−1≫max⁡(∥E∥,ΛQCD),r∥∇E∥≪∥E∥,(3)m_Q \gg r^{-1} \gg\max(\sqrt{\lVert E\rVert},\Lambda_{\mathrm{QCD}}),\qquad r\lVert\nabla E\rVert\ll\lVert E\rVert, \tag*{(3)}

with weak coupling αs(r−1)≪1\alpha_s(r^{-1}) \ll1. In a nonstationary background the frozen soft coefficient also requires τΩE≪1\tau\Omega_E \ll1, where ΩE=∥DtE∥/∥E∥\Omega_E = \lVert D_tE\rVert/\lVert E\rVert measures the local rate of variation (away from zeros of the curvature; more generally the corresponding derivative expansion must be small). For a physical ultrasoft reference approximation, a distinct stronger hierarchy is specified in Sec. V. The mass mQm_Q enters the static limit only as a scale; nfn_f light flavors are massless. We set CF=(Nc2−1)/(2Nc)C_F = (N_c^2 - 1)/(2N_c), CA=NcC_A = N_c, TF=1/2T_F = 1/2, and g2=4παsg^2 = 4\pi\alpha_s. After Wick rotation of the frozen metric, the closed rectangular contour has temporal length TT and spatial connectors joining the sources at both ends. Denote its fundamental-representation Wilson loop by

W(T,r)=1Nctr⁡Pexp⁡(ig∮CT,rAμdxμ).(4)\mathcal{W}(T,r)=\frac{1}{N_c}\operatorname{tr}\mathcal{P}\exp\left(ig\oint_{C_{T,r}}A_\mu\mathrm{d}x^\mu\right). \tag*{(4)}

The coefficient of TT in −ln⁡⟨W⟩-\ln\langle\mathcal{W}\rangle, after subtraction of the single-source energies, defines the static energy in an auxiliary stationary completion. The pNRQCD matching coefficient is obtained by subtracting the ultrasoft contribution in the same convention [1, 3]. The auxiliary completion is an infrared regulator for the local static problem.

The relative displacement is a spacelike vector rμr^\mu in the midpoint tetrad, with u⋅r=0u\cdot r=0, r=(rμrμ)1/2r=(r^\mu r_\mu)^{1/2}, where uμu^\mu is the tangent to the central geodesic. The invariant

W=Cμανβrμuαrνuβ=Eijrirj(5)W=C_{\mu\alpha\nu\beta}r^\mu u^\alpha r^\nu u^\beta=E_{ij}r^i r^j \tag*{(5)}

is the trace-free electric Weyl contraction in vacuum. In the parity-even static sector it is the only scalar linear in curvature at the leading anisotropic short-distance order. With the midpoint worldline and separation fixed geometrically, this contraction is invariant under coordinate changes and rotations of the spatial tetrad. A change of the physical observer generally changes the electric Weyl tensor.

Let S=S11c/NcS=S_1\mathbf{1}_c/\sqrt{N_c} and O=OaTa/TFO=O^aT^a/\sqrt{T_F} denote the singlet and octet fields in the conventional color normalization. To leading multipole and static order the terms relevant for the matching are

LpNRQCD=tr⁡[S†(i∂t−Vs)S]+tr⁡[O†(iDt−Vo)O]+gVAtr⁡[O†r⋅ES+S†r⋅EO]+⋯ .(6)\mathcal{L}_{\mathrm{pNRQCD}}=\operatorname{tr}\left[S^\dagger(i\partial_t-V_s)S\right] +\operatorname{tr}\left[O^\dagger(iD_t-V_o)O\right] +gV_A\operatorname{tr}\left[O^\dagger\mathbf{r}\cdot\mathbf{E}S+S^\dagger\mathbf{r}\cdot\mathbf{E}O\right]+\cdots. \tag*{(6)}

where E\mathbf{E} is the chromoelectric field at the midpoint, VA=1V_A=1 at leading order, and integration over the relative coordinate and the center coordinate is understood. The curvature dependence of the leading static potential is represented by

δL=−cW(r;μ,ν)OW,OW=Wrtr⁡(S†S),(7)\delta\mathcal{L}=-c_W(r;\mu,\nu)\mathcal{O}_W,\qquad\mathcal{O}_W=\frac{W}{r}\operatorname{tr}(S^\dagger S), \tag*{(7)}

with μ\mu the coupling-renormalization scale and ν\nu the soft–ultrasoft factorization scale. The operator is a color singlet. Its leading ultrasoft counterterm is generated by two insertions of the dipole term in Eq. (6). At the accuracy needed for this counterterm, the singlet–octet gap is ΔV=Vo−Vs=CAαs/(2r)\Delta V=V_o-V_s=C_A\alpha_s/(2r). Curvature corrections to the dipole vertex or gap multiplied by the flat-space self-energy enter at higher perturbative order in the static sector.

Tree-level matching

Write the physical Euclidean metric as gμν=N2g~μνg_{\mu\nu}=N^2\widetilde{g}_{\mu\nu} with ds~2=dt2+γ~ijdxidxj\mathrm{d}\widetilde{s}^2=\mathrm{d}t^2+\widetilde{\gamma}_{ij}\mathrm{d}x^i\mathrm{d}x^j and γ~ij=N−2γij\widetilde{\gamma}_{ij}=N^{-2}\gamma_{ij}. In four dimensions the classical Yang–Mills and massless Dirac actions and the gauge connection in the loop are Weyl invariant (with the appropriate fermion rescaling). The renormalized Weyl transformation at one loop is treated in Sec. IV E.

The local optical curvature at the midpoint is R~ij=2Eij\widetilde{R}_{ij}=2E_{ij}, R~=0\widetilde{R}=0. For example, the spatial physical Ricci tensor at the midpoint equals EijE_{ij} by four-dimensional Ricci flatness, and the conformal transformation of the three-metric contributes another EijE_{ij}. The optical geodesic distance between symmetric endpoints is

ℓ=∫−r/2r/2(1−12Eijninjλ2)dλ=r(1−W/24),(8)\ell=\int_{-r/2}^{r/2}\left(1-\frac{1}{2}E_{ij}n^i n^j\lambda^2\right)\mathrm{d}\lambda=r(1-W/24), \tag*{(8)}
Rz=R~ijzizj=2W+O(E2r4),(9)\mathcal{R}_z=\widetilde{R}_{ij}z^i z^j=2W+O(E^2r^4), \tag*{(9)}

where ni=ri/rn^i=r^i/r and ziz^i is the optical separation vector in a parallel frame. The spatial Riemann correction to the length along this radial segment vanishes by antisymmetry. The Van Vleck factor is U=Δ1/2=1+Rz/12U=\Delta^{1/2}=1+\mathcal{R}_z/12 to this order.

We continue the optical spatial dimension to d=3−2ϵd=3-2\epsilon, set b=(d−1)/2=1−ϵb=(d-1)/2=1-\epsilon, hold Rz\mathcal{R}_z fixed, and retain R~\widetilde{R} throughout. The bare coupling is gB2=g2(μ)μ2ϵeEγ/(4π)ϵZg2g_B^2=g^2(\mu)\mu^{2\epsilon}e^\gamma_E/(4\pi)^\epsilon Z_g^2. The residual soft pole is subtracted in a tree-normalized dimensionally continued coefficient, with its opposite assigned to the dipole self-energy. In particular the dimensionally continued tree factor and gap must be used consistently before taking d→3d\to3. Finite evanescent changes in the continuation or subtraction may shift the Wilson coefficient and ultrasoft reference term oppositely.

Static Coulomb Green function

The physical static A0A_0 operator is the density-weighted optical scalar Laplacian, −∂i(γ N−1γij∂j)-\partial_i(\sqrt{\gamma}\,N^{-1}\gamma^{ij}\partial_j). With L=−∇~2L=-\widetilde{\nabla}^2 its Green function away from coincidence is, in three spatial dimensions,

G1=U4πℓ=14πr(1+524W).(10)G_1=\frac{U}{4\pi\ell}=\frac{1}{4\pi r}\left(1+\frac{5}{24}W\right). \tag*{(10)}

Multiplication by −g2CF-g^2C_F gives the leading static interaction,

Vs(0)(r)=−CFαsr(1+524W).(11)V_s^{(0)}(r)=-\frac{C_F\alpha_s}{r}\left(1+\frac{5}{24}W\right). \tag*{(11)}

Smooth boundary-dependent terms are distinct from the local nonanalytic term proportional to W/rW/r.

An independent physical-metric check uses the static Maxwell operator P=−∇i(N−1∇i)P=-\nabla_i(N^{-1}\nabla^i) with the invariant spatial delta function, where the derivatives now use γ\gamma. With f=ln⁡Nf=\ln N,

P=N−1/2[−∇γ2−12∇γ2f+14(∇γf)2]N−1/2.(12)P=N^{-1/2}\left[-\nabla_{\gamma}^{2}-\frac{1}{2}\nabla_{\gamma}^{2}f+\frac{1}{4}(\nabla_{\gamma}f)^{2}\right]N^{-1/2}. \tag*{(12)}

Since ∇γ2f=0\nabla_{\gamma}^{2}f=0 at linear order, the local Green function is N+1/2N−1/2N_{+}^{1/2}N_{-}^{1/2} times the spatial scalar Green function. The physical spatial Ricci tensor is EijE_{ij} and the radial spatial length is rr, so

Gphys=(1+W/8)(1+W/12)4πr=1+5W/244πr+O(E2).(13)G_{\mathrm{phys}}=\frac{(1+W/8)(1+W/12)}{4\pi r}=\frac{1+5W/24}{4\pi r}+O(E^{2}). \tag*{(13)}

This agrees also with the static spin-independent term of Caicedo and Urrutia [8]: in their Eqs. (9), (18), (19), (33) and (36), symmetric endpoints give Aab=rarb/12A^{ab}=r^{a}r^{b}/12, a00=−W/12a_{00}=-W/12, ξh=0\xi_{h}=0, and h00(1)=h00(2)=−W/4h_{00}^{(1)}=h_{00}^{(2)}=-W/4. Thus a~00=5W/12\tilde{a}_{00}=5W/12 and their symmetrized Coulomb factor is 1+5W/241+5W/24.

One-loop soft matching

The order-g4g^{4} static interaction is determined by the vacuum polarization and the connected two-exchange contribution to the logarithm of the Wilson loop. We first compute these diagrams in the optical metric and then restore the physical metric. The coupling is renormalized in MS; the standard flat-space coefficients and coordinate-space logarithm are

a1=319CA−209TFnf,a_{1}=\frac{31}{9}C_{A}-\frac{20}{9}T_{F}n_{f},
β0=113CA−43TFnf,\beta_{0}=\frac{11}{3}C_{A}-\frac{4}{3}T_{F}n_{f},
Lλ=ln⁡(λr)+γE,Lℓ=ln⁡(μℓ)+γE.(14)L_{\lambda}=\ln(\lambda r)+\gamma_{E},\qquad L_{\ell}=\ln(\mu\ell)+\gamma_{E}. \tag*{(14)}

Here γE\gamma_{E} is Euler’s constant.

Local propagators and curvature convolution

Let Z=t2+ℓ2Z=t^{2}+\ell^{2}, cb=Γ(b)/(4πb+1)c_{b}=\Gamma(b)/(4\pi^{b+1}) and F=cbZ−bF=c_{b}Z^{-b}. The free Euclidean Feynman-gauge operators are the Hodge Laplacian on one-forms and the scalar Laplacian on ghosts. We employ the local heat-kernel expansion of [13, 14, 15, 17]. The optical heat-kernel coefficients give

H(t,z)=UF,(15)H(t,z)=UF, \tag*{(15)}
Dij′(t,z)=gij′H−cb4(b−1)R~ij′Z1−b,(16)D_{ij'}(t,z)=g_{ij'}H-\frac{c_{b}}{4(b-1)}\widetilde{R}_{ij'}Z^{1-b}, \tag*{(16)}
D00′=H,D0i′=0,Dgh=H.(17)D_{00'}=H,\qquad D_{0i'}=0,\qquad D_{\mathrm{gh}}=H. \tag*{(17)}

Here gij′g_{ij'} is the parallel propagator. The vector correction follows from the first heat-kernel coefficient (a1)ij=R~g~ij/6−R~ij(a_{1})_{ij}=\widetilde{R}\widetilde{g}_{ij}/6-\widetilde{R}_{ij}, whose proper-time integral supplies Γ(b−1)/Γ(b)\Gamma(b-1)/\Gamma(b). These are local parametrices with the dimensionally continued infrared prescription, not full Green functions. Smooth additions are considered in Sec. V C.

The scalar fractional Green kernels are particularly useful for attaching the Coulomb legs:

Ks(x,y)=⟨x∣L−s∣y⟩=ksℓ2s−dU,(18)K_{s}(x,y)=\langle x|L^{-s}|y\rangle=k_{s}\ell^{2s-d}U, \tag*{(18)}
ks=Γ(d/2−s)4sπd/2Γ(s).(19)k_{s}=\frac{\Gamma(d/2-s)}{4^{s}\pi^{d/2}\Gamma(s)}. \tag*{(19)}

The formulas hold at linear curvature and neglect derivatives of curvature; powers outside the convergent domain are interpreted as homogeneous distributions. The flat static polarization kernel scales as ℓ1−2d\ell^{1-2d}. Its covariant scalar component corresponds to LbL^{b} and becomes L−(2−b)L^{-(2-b)} after the two Coulomb legs are attached. Any additional first-order anisotropic kernel can instead be convolved with the flat Coulomb legs.

Since Rz\mathcal{R}_{z} is a trace-free quadratic polynomial, ∇2Rz=0\nabla^{2}\mathcal{R}_{z}=0, and

−∇2(Rzℓp)=−p(p+d+2)Rzℓp−2.(20)-\nabla^{2}\left(\mathcal{R}_{z}\ell^{p}\right)=-p(p+d+2)\mathcal{R}_{z}\ell^{p-2}. \tag*{(20)}

For the isotropic power the last factor is p+d−2p+d-2. Starting from Rzℓ1−2d\mathcal{R}_{z}\ell^{1-2d} and inverting twice with the exponents 1−4b1-4b and 3−4b3-4b gives the ratio to the scalar convolution,

K(b)=(−2b)(2−2b)(4−2b)(6−2b)=b(b−1)(2−b)(3−b).(21)\mathcal{K}(b)=\frac{(-2b)(2-2b)}{(4-2b)(6-2b)}=\frac{b(b-1)}{(2-b)(3-b)}. \tag*{(21)}

In particular K(1−ϵ)=−ϵ/2+O(ϵ2)\mathcal{K}(1-\epsilon)=-\epsilon/2+O(\epsilon^{2}). This zero cannot be taken before multiplying the one-loop poles.

Light-fermion polarization

The spinor parametrix is obtained by acting with −γμ∇~μ-\gamma^{\mu}\widetilde{\nabla}_{\mu} on the spin-parallel scalar Green kernel. The spin connection contributes through the contracted derivative of the spin parallel transporter, −γiR~ijzj/4-\gamma^{i}\widetilde{R}_{ij}z^{j}/4; the derivative of UU gives +γiR~ijzj/6+\gamma^{i}\widetilde{R}_{ij}z^{j}/6. The scalar curvature endomorphism in the squared Dirac operator vanishes here. In a parallel spin frame,

S(t,z)=2cbγμzμZb+1U+cb12γiR~ijzjZb.(22)S(t,z)=\frac{2c_{b}\gamma^{\mu}z_{\mu}}{Z^{b+1}}U+\frac{c_{b}}{12}\frac{\gamma^{i}\widetilde{R}_{ij}z^{j}}{Z^{b}}. \tag*{(22)}

with z0=tz^{0}=t and four-component spinor trace. Contract the static bubble −gB2TFnftr⁡[γ0S(x,y)γ0S(y,x)]-g_{B}^{2}T_{F}n_{f}\operatorname{tr}[\gamma^{0}S(x,y)\gamma^{0}S(y,x)]. The reversed separation sign can be factored out of the numerator. The flat numerator is then 4(t2−ℓ2)(2bcb)2/Z2b+24(t^{2}-\ell^{2})(2bc_{b})^{2}/Z^{2b+2}, and the term linear in the second part of Eq. (22) is −8(2bcb)(cb/12)Rz/Z2b+1-8(2bc_{b})(c_{b}/12)\mathcal{R}_{z}/Z^{2b+1}. Using

Jd=∫−∞∞dtZd=π Γ(d−1/2)Γ(d)ℓ1−2d,(23)J_{d}=\int_{-\infty}^{\infty}\frac{dt}{Z^{d}}=\frac{\sqrt{\pi}\,\Gamma(d-1/2)}{\Gamma(d)}\ell^{1-2d}, \tag*{(23)}
∫−∞∞t2−ℓ2Zd+1 dt=−d−1dJd,(24)\int_{-\infty}^{\infty}\frac{t^{2}-\ell^{2}}{Z^{d+1}}\,dt=-\frac{d-1}{d}J_{d}, \tag*{(24)}

we obtain the integrated bubble curvature factor

1+Rz(16+d24b2).(25)1+\mathcal{R}_{z}\left(\frac{1}{6}+\frac{d}{24b^{2}}\right). \tag*{(25)}

Relative to the scalar-covariantized contribution, which already contains UU, the extra factor after Coulomb convolution is

ρf=(112+d24b2)K(b)=548ϵ+O(ϵ2).(26)\rho_f=\left(\frac{1}{12}+\frac{d}{24b^2}\right)\mathcal{K}(b)=\frac{5}{48}\epsilon+O(\epsilon^2). \tag*{(26)}

With the conventional coordinate-space normalization (per TFnfαs/(4π)T_F n_f\alpha_s/(4\pi) and per dimensionally continued tree exchange) the flat bubble is

ff=−43ϵ−209−83Lℓ+O(ϵ),(27)f_f=-\frac{4}{3\epsilon}-\frac{20}{9}-\frac{8}{3}L_\ell+O(\epsilon), \tag*{(27)}

Thus the additional finite optical fermion term is

lim⁡ϵ→0ffρfRz=536Rz.(28)\lim_{\epsilon\to0}f_f\rho_f\mathcal{R}_z=\frac{5}{36}\mathcal{R}_z. \tag*{(28)}

Gluon–ghost polarization

Static current and Wick contractions

Split the temporal gauge field as A0a=Ba+a0aA^a_0=B^a+a^a_0, with external BaB^a static, and set Aia=aiaA^a_i=a^a_i. In ordinary (not background-field) covariant Feynman gauge, the cubic term linear in BB follows from F0i2/2F^2_{0i}/2. After integrating the derivative on BB by parts it is gB∫J0agB\int J^a_0, with the measure understood and

J0a=fabc[−a˙ibaic+2(∇~ia0b)aic+a0b∇~ia˙ic].(29)J^a_0=f^{abc}\left[-\dot{a}^b_i a^c_i+2(\widetilde{\nabla}_i a^b_0)a^c_i+a^b_0\widetilde{\nabla}_i\dot{a}^c_i\right]. \tag*{(29)}

The factor of two in the mixed derivative current comes from differentiating the external and the quantum temporal fields. The ghost vertex linear in BB has one temporal derivative. The contraction identities fabcfa′bc=CAδaa′f^{abc}f^{a'bc}=C_A\delta^{aa'} and fabcfa′cb=−CAδaa′f^{abc}f^{a'cb}=-C_A\delta^{aa'} fix the relative signs. The Hodge identity Δ1d=dΔ0\Delta_1d=d\Delta_0 gives ∇~xiDij′=−∇~j′H\widetilde{\nabla}^i_xD_{ij'}=-\widetilde{\nabla}_{j'}H for compatible Green functions, away from contact. The three types of current in Eq. (29) are denoted JT,JM,JDJ_T,J_M,J_D.

The first entry is the sum of a product of first time derivatives and a term with a second time derivative, integrated by parts. The divergence–divergence contraction is H∂t2HH\partial_t^2H off contact. The ghost loop has the usual

Grassmann sign. Combining Table I, the gluon–ghost current kernel divided by its color factor is

ContractionIntegrand
JTJTJ_TJ_T2D˙ij′D˙ij′2\dot{D}_{ij'}\dot{D}^{ij'}
JMJMJ_MJ_M4Dij′H;ij′4D^{ij'}H_{;ij'}
JMJDJ_MJ_D−2∣∇~xH∣2-2|\widetilde{\nabla}_xH|^2
JDJMJ_DJ_M−2∣∇~yH∣2-2|\widetilde{\nabla}_yH|^2
JDJDJ_DJ_D−H˙2-\dot{H}^2
Ghost loop−H˙2-\dot{H}^2

Table I. TABLE I. Contractions contributing to the static gluon–ghost kernel, after temporal integration by parts. The common color factor CAC_A and time integral are suppressed.

P=∫−∞∞dt[2D˙ij′D˙ij′+4Dij′H;ij′−2∣∇~xH∣2−2∣∇~yH∣2−2H˙2].(30)\mathcal{P}=\int_{-\infty}^{\infty}dt\left[2\dot{D}_{ij'}\dot{D}^{ij'}+4D^{ij'}H_{;ij'}-2|\widetilde{\nabla}_xH|^2-2|\widetilde{\nabla}_yH|^2-2\dot{H}^2\right]. \tag*{(30)}

The inverse-propagator insertion is minus the current–current kernel. Ordinary local contact terms are handled with the flat kinetic counterterm and covariant dimensional extension. The trace-free local part of the four-gluon tadpole vanishes since it is proportional to tr⁡a1\operatorname{tr}a_1; smooth infrared terms have to be treated consistently with the overlap. A cubic vertex with three temporal legs is zero because it cannot arise from the antisymmetric field strength. At this order no other bulk diagrams with only temporal source attachments occur. Spatial attachments are discussed in § 2.3 C.

Curvature numerator

Using the bitensor identities collected in App. A1, and with primes indicating ZZ derivatives, one obtains

gij′H;ij′=U(−2dF′−4ℓ2F′′)−RzF′,(31)g^{ij'}H_{;ij'}=U(-2dF'-4\ell^2F'')-\mathcal{R}_zF', \tag*{(31)}
∣∇~xH∣2=4U2ℓ2(F′)2+23RzFF′.(32)|\widetilde{\nabla}_xH|^2=4U^2\ell^2(F')^2+\frac{2}{3}\mathcal{R}_zFF'. \tag*{(32)}

The additional spatial-vector correction contracts as R~ij′F;ij′=−4RzF′′\widetilde{R}_{ij'}F^{;ij'}=-4\mathcal{R}_zF''. The term in Dij′D˙ij′D_{ij'}\dot{D}^{ij'} linear in the vector endomorphism is trace zero. The curvature-dependent pieces in Eq. (30), beyond the common U2U^2 factor, are therefore

δPint=−4RzFF′−83RzFF′+4cbb−1RzZ1−bF′′=−cb2RzZ−d[20b3+4b(b+1)b−1].(33)\begin{aligned} \delta\mathcal{P}_{\mathrm{int}}=-4\mathcal{R}_zFF'-\frac{8}{3}\mathcal{R}_zFF'+\frac{4c_b}{b-1}\mathcal{R}_zZ^{1-b}F'' \\ &=-c_b^2\mathcal{R}_zZ^{-d}\left[\frac{20b}{3}+\frac{4b(b+1)}{b-1}\right]. \tag*{(33)} \end{aligned}

This isolates the derivative geometry from the vector endomorphism; treating each vector component as a scalar would miss the second, dimensionally singular term. The flat integrated combination reduces to (2d−6)F2−4∣∇F∣2(2d-6)F^2-4|\nabla F|^2. Inserting ∫t2Z−d−1dt=Jd/(2d)\int t^2Z^{-d-1}dt=J_d/(2d) gives the full reduced kernel

P=−cb2Jd[Ab(1+Rz/6)−BbRz],(34)\mathcal{P}=-c_b^2J_d\left[A_b(1+\mathcal{R}_z/6)-B_b\mathcal{R}_z\right], \tag*{(34)}
Ab=4b2(3+1/d),Bb=20b3+4b(b+1)b−1.(35)A_b=4b^2(3+1/d),\qquad B_b=\frac{20b}{3}+\frac{4b(b+1)}{b-1}. \tag*{(35)}

The same BbB_b follows by performing the derivatives in Eq. (30) before time integration. The sign convention in Eq. (34) is fixed by the momentum-space check in App. A2.

Flat normalization and anisotropic contribution

Attaching the Coulomb legs and normalizing to tree exchange gives (per CAαs/(4π)C_A\alpha_s/(4\pi)):

fse=π(3ϵ−5)eγEϵ(μℓ)2ϵ4Γ(1/2−ϵ)Γ(5/2−ϵ)×Γ(−ϵ)Γ(1/2−2ϵ)=53ϵ+319+103Lℓ+O(ϵ).(36)f_{\mathrm{se}}=\frac{\sqrt{\pi}(3\epsilon-5)e^{\gamma_E\epsilon}(\mu\ell)^{2\epsilon}}{4\Gamma(1/2-\epsilon)\Gamma(5/2-\epsilon)}\times\Gamma(-\epsilon)\Gamma(1/2-2\epsilon) =\frac{5}{3\epsilon}+\frac{31}{9}+\frac{10}{3}L_\ell+O(\epsilon). \tag*{(36)}

The light-fermion factor obeys ff/fse=−8b/(3d+1)f_f/f_{\mathrm{se}}=-8b/(3d+1), another check on the flat normalization.

Subtract the curvature already contained in UU and use Eq. (21) to attach the Coulomb legs. The normalized extra anisotropic factor is

ρg=(112−BbAb)K(b)(37)\rho_g=\left(\frac{1}{12}-\frac{B_b}{A_b}\right)\mathcal{K}(b) \tag*{(37)}
=3ϵ3+24ϵ2−67ϵ+3612(ϵ+1)(ϵ+2)(3ϵ−5)=−310+497600ϵ+O(ϵ2).(38)=\frac{3\epsilon^3+24\epsilon^2-67\epsilon+36}{12(\epsilon+1)(\epsilon+2)(3\epsilon-5)} =-\frac{3}{10}+\frac{497}{600}\epsilon+O(\epsilon^2). \tag*{(38)}

The product with Eq. (36) is

fseρg=−12ϵ+2572−Lℓ+O(ϵ).(39)f_{\mathrm{se}}\rho_g=-\frac{1}{2\epsilon}+\frac{25}{72}-L_\ell+O(\epsilon). \tag*{(39)}

The finite constant 25/72=(5/3)(497/600)−(3/10)(31/9)25/72=(5/3)(497/600)-(3/10)(31/9) depends on the terms evanescent in the curvature convolution. In particular a direct calculation with d=3d=3 before performing the integrals would not determine this number.

An additional local ultraviolet curvature counterterm for two temporal fields at this engineering dimension would be proportional to R~\widetilde{R} or R~00\widetilde{R}_{00}; both vanish in the continuation. Covariant kinetic terms and their ordinary renormalization are retained.

Connected temporal Wilson-line web

The logarithm of the Wilson loop removes independent repetitions of tree exchange. It does not remove the connected non-Abelian color correlation between two exchanges. For t>0t>0 write

A(t)=cbt−2b,λ=1+Rz/12,A(t)=c_b t^{-2b},\qquad\lambda=1+R_z/12,
B(t)=cbλ(t2+ℓ2)−b.(40)B(t)=c_b\lambda(t^2+\ell^2)^{-b}. \tag*{(40)}

The time propagators for the symmetric and antisymmetric source couplings are D±=(A±B)/2D_\pm=(A\pm B)/2. Define Qa=T1a+T2aQ^a=T_1^a+T_2^a, Pa=T1a−T2aP^a=T_1^a-T_2^a, with the antiquark generator in the antifundamental. The singlet satisfies Qa∣s⟩=0Q^a|s\rangle=0, and P2=4CF−Q2P^2=4C_F-Q^2. Acting once with PaP^a produces an octet. Both crossed and nested four-PP color orderings give 4CF(4CF−CA)4C_F(4C_F-C_A); subtracting the disconnected factor (4CF)2(4C_F)^2 leaves −4CFCA-4C_FC_A. The nested PQQPPQQP ordering contributes +4CFCA+4C_FC_A.

With u,v,w≥0u,v,w\ge0 the three successive time gaps between ordered vertices, the extensive connected correction is VW=−4gB4CFCAIV_W=-4g_B^4C_FC_AI with

I=∫0∞du dv dw[D−(u+v+w)B(v)−D−(u+v)D−(v+w)].(41)I=\int_0^\infty du\,dv\,dw\left[D_-(u+v+w)B(v)-D_-(u+v)D_-(v+w)\right]. \tag*{(41)}

The purely same-source part is removed by the single-source subtraction. The mixed ABAB terms factorize exactly: changing variables in the first term and the two mixed crossed terms gives

IAB=12∫0∞ds dt A(s)B(t)[(s−t)++min⁡(s,t)]=12(∫0∞sA(s) ds)(∫0∞B(t) dt).(42)\begin{aligned} I_{AB}=\frac{1}{2}\int_0^\infty ds\,dt\,A(s)B(t)\left[(s-t)_++\min(s,t)\right] \\ &=\frac{1}{2}\left(\int_0^\infty sA(s)\,ds\right)\left(\int_0^\infty B(t)\,dt\right). \tag*{(42)} \end{aligned}

The first factor is scaleless in the all-hard dimensional expansion. With a separate infrared regulator its ultraviolet–infrared decomposition must not be discarded; see § 2.3 C.

The remaining opposite-source term integrates to

IBB=−12∫0∞dv B(v)∫v∞dt tB(t)=−cb2λ2ℓ3−4b4(b−1)πΓ(2b−3/2)2Γ(2b−1).(43)\begin{aligned} I_{BB}=-\frac{1}{2}\int_0^\infty dv\,B(v)\int_v^\infty dt\,tB(t) \\ &=-\frac{c_b^2\lambda^2\ell^{3-4b}}{4(b-1)}\frac{\sqrt{\pi}\Gamma(2b-3/2)}{2\Gamma(2b-1)}. \tag*{(43)} \end{aligned}

The nested part contributes a factor t−vt-v under this integral; the crossed part supplies vv. Normalizing by the tree exchange gives (per CAαs/(4π)C_A\alpha_s/(4\pi))

fW=−2πeγEϵ(μℓ)2ϵΓ(−ϵ)Γ(1/2−2ϵ)Γ(1/2−ϵ)2=2ϵ+4Lℓ+O(ϵ).(44)f_W=-\frac{2\sqrt{\pi}e^{\gamma_E\epsilon}(\mu\ell)^{2\epsilon}\Gamma(-\epsilon)\Gamma(1/2-2\epsilon)}{\Gamma(1/2-\epsilon)^2} =\frac{2}{\epsilon}+4L_\ell+O(\epsilon). \tag*{(44)}

Since the web carries λ2\lambda^2 and the scalar tree carries λ\lambda, its additional curvature factor is fWRz/12f_W R_z/12. In combination with Eq. (39), the additional optical gluon result is

CARz[−13ϵ+2572−23Lℓ].(45)C_AR_z\left[-\frac{1}{3\epsilon}+\frac{25}{72}-\frac{2}{3}L_\ell\right]. \tag*{(45)}

The flat combination CA(fse+fW)+TFnfffC_A(f_{\mathrm{se}}+f_W)+T_Fn_ff_f reproduces β0/ϵ+a1+2β0Lℓ\beta_0/\epsilon+a_1+2\beta_0L_\ell. Subtracting the coupling pole gives the known flat potential. This check constrains the color and normalization conventions, although it does not uniquely determine the curvature calculation.

Physical-metric restoration

Classical conformal invariance alone is insufficient to relate the renormalized physical and optical quantum theories. We use the one-loop Weyl Ward insertion [16] at fixed infrared completion. With f=ln⁡Nf=\ln N and with the coupling absorbed into the gauge connection A=gA\mathcal{A}=gA, the gauge-field part of the renormalized action variation is

δfS=−β032π2∫d4xg~ f FμνFμν.(46)\delta_f S=-\frac{\beta_0}{32\pi^2}\int d^4x\sqrt{\tilde{g}}\,f\,\mathcal{F}^{\mu\nu}\mathcal{F}_{\mu\nu}. \tag*{(46)}

where the sign is fixed by the shift ln⁡(μℓ)→ln⁡(μℓ)+f\ln(\mu\ell)\to\ln(\mu\ell)+f for constant ff. Purely gravitational terms cancel in vacuum-normalized loop expectation values. The loop is renormalized with Weyl-related infrared conditions.

The local normalization of this insertion can be checked before taking the dimensional limit. In D=4−2ϵD=4-2\epsilon, the physical bare action with the coupling absorbed into the connection is

SB=14gB2∫dDxg~ e−2ϵfF2.(47)S_B=\frac{1}{4g_B^2}\int d^Dx\sqrt{\tilde{g}}\,e^{-2\epsilon f}\mathcal{F}^2. \tag*{(47)}

Using Zg2=1−β0g2/(16π2ϵ)+O(g4)Z_g^2=1-\beta_0g^2/(16\pi^2\epsilon)+O(g^4), the product of −2ϵf-2\epsilon f and the inverse-coupling counterterm gives −β0fF2/(32π2)-\beta_0 f\mathcal{F}^2/(32\pi^2), with the same normalization as Eq. (46). The fermion sector admits the direct check of Sec. VI.

The static part is governed by Lf=−∇~i(f∇~i)L_f=-\tilde{\nabla}_i(f\tilde{\nabla}^i), for which

Lf=12{f,L}+12∇~2f.(48)L_f=\frac{1}{2}\{f,L\}+\frac{1}{2}\tilde{\nabla}^2f. \tag*{(48)}

Locally ∇~2f=0\tilde{\nabla}^2f=0 at the retained order. Consequently L−1LfL−1=12(fL−1+L−1f)L^{-1}L_fL^{-1}=\frac{1}{2}(fL^{-1}+L^{-1}f) up to terms from the infrared completion. At both symmetric sources f=W/8f=W/8, so Eq. (46) produces +β0W/4+\beta_0W/4 in the normalized one-loop bracket. (8) gives the optical distance correction −β0W/12-\beta_0W/12 from the logarithm. The total restoration is +β0W/6+\beta_0W/6.

Since Rz=2W\mathcal{R}_z=2W, the finite extra terms in Eqs. (28) and (45) become

cA=22572+1118=4736,(49)c_A=2\frac{25}{72}+\frac{11}{18}=\frac{47}{36}, \tag*{(49)}
cf=2536−29=118.(50)c_f=2\frac{5}{36}-\frac{2}{9}=\frac{1}{18}. \tag*{(50)}

The remaining optical soft infrared pole, in physical normalization, is

Vs,IR(E)=+CFCAαs2W6πrϵ.(51)V^{(E)}_{s,\mathrm{IR}}=+\frac{C_FC_A\alpha_s^2W}{6\pi r\epsilon}. \tag*{(51)}

This step identifies the coefficient before ultrasoft subtraction; consistency requires removing this pole with the dipole operator of pNRQCD, not a new ultraviolet counterterm in fundamental QCD.

Ultrasoft renormalization and the matched potential

The soft pole in Eq. (51) is the first indication of curvature-dependent ultrasoft running. We now determine the ultraviolet counterterm of the singlet operator directly from the dipole self-energy in pNRQCD.

Chromoelectric correlator

The free chromoelectric correlator in the midpoint frame can be read from Eqs. (15) and (16). The parallel correlator contains an adjoint temporal Wilson line; the latter is unity at the order needed for the free correlator. With the Euclidean–Lorentzian electric-field sign, define Gij(t)=∂t2Dij′(t,0)−H;ij′(t,0)G_{ij}(t)=\partial_t^2D_{ij}'(t,0)-H_{;ij}'(t,0) for t>0t>0. Explicitly,

H;ij′(t,0)=2bcbδijt−2b−2−cbR~ijt−2b6,(52)H_{;ij}'(t,0)=2b c_b\delta_{ij}t^{-2b-2}-\frac{c_b\widetilde{R}_{ij}t^{-2b}}{6}, \tag*{(52)}
−∂t2Dij′(t,0)=−2b(2b+1)cbδijt−2b−2+2b−12cbR~ijt−2b.(53)-\partial_t^2D_{ij}'(t,0)=-2b(2b+1)c_b\delta_{ij}t^{-2b-2}+\frac{2b-1}{2}c_b\widetilde{R}_{ij}t^{-2b}. \tag*{(53)}

Thus its local Laplace spectral reference is

Gij(t)=∫0∞dω e−ωtJij(ω),(54)G_{ij}(t)=\int_0^\infty d\omega\,e^{-\omega t}\mathcal{J}_{ij}(\omega), \tag*{(54)}
Jij(ω)=4b2cbΓ(2b+2)δijω2b+1−cb(b−2/3)Γ(2b)R~ijω2b−1.(55)\mathcal{J}_{ij}(\omega)=\frac{4b^2c_b}{\Gamma(2b+2)}\delta_{ij}\omega^{2b+1}-\frac{c_b(b-2/3)}{\Gamma(2b)}\widetilde{R}_{ij}\omega^{2b-1}. \tag*{(55)}

In four spacetime dimensions this becomes (δijω3−Eijω)/(6π2)(\delta_{ij}\omega^3-E_{ij}\omega)/(6\pi^2). This is a local ultraviolet asymptotic reference, not a globally positive spectral density. In particular the linear curvature term cannot be extrapolated into the low-frequency physical spectrum without additional information.

Independent physical-metric Hadamard check

The electric singularity can be checked without the optical endomorphism entering Eq. (16). In the physical Euclidean Ricci-flat metric, the local free vector Green function is Pμν′/(8π2σ)P_{\mu\nu}'/(8\pi^2\sigma), where PP is parallel transport and σ\sigma is the world function. The Van Vleck and first logarithmic coefficients vanish at linear curvature after neglecting curvature derivatives. Set c=1/(4π2)c=1/(4\pi^2) and let the endpoints have Euclidean time gap t>0t>0 and small spatial offsets x,y\mathbf{x},\mathbf{y} from the central worldline. To the required order in the offsets,

2σ=t2+∣x−y∣2+t23Ekl(xkxl+ykyl+xkyl)+⋯ ,(56)\begin{aligned} 2\sigma=t^2+|\mathbf{x}-\mathbf{y}|^2 \\ &+\frac{t^2}{3}E_{kl}(x^kx^l+y^ky^l+x^ky^l)+\cdots, \tag*{(56)} \end{aligned}
P00′=1+f(x)+f(y)+⋯ ,(57)P_{00}'=1+f(x)+f(y)+\cdots, \tag*{(57)}
P0j′=−t2Ejk(x+y)k+⋯ ,(58)P_{0j}'=-\frac{t}{2}E_{jk}(x+y)^k+\cdots, \tag*{(58)}
Pi0′=+t2Eik(x+y)k+⋯ .(59)P_{i0}'=+\frac{t}{2}E_{ik}(x+y)^k+\cdots. \tag*{(59)}

The world-function term is obtained by integrating the physical g00g_{00} perturbation on the unperturbed segment; the mixed parallel components follow by integrating the physical connection, Γ0i0=−Eijxj\Gamma^0_{0i}=-E_{ij}x^j and Γ00i=−Eijxj\Gamma^i_{00}=-E^i{}_jx^j. Spatial-metric corrections to σ\sigma are quiet in the spatial offsets and do not enter this check. On the central geodesic Pij′=δijP_{ij}'=\delta_{ij}.

Write Dphys=cP/(2σ)D^{\mathrm{phys}}=cP/(2\sigma). Taking the four derivatives in the field-strength two-point function, then setting x=y=0x=y=0, gives

∂i∂jD00′phys=2cδij/t4−cEij/(3t2),(60)\partial_i\partial_jD^{\mathrm{phys}}_{00'}=2c\delta_{ij}/t^4-cE_{ij}/(3t^2), \tag*{(60)}
−∂i∂0D0j′phys=cEij/(2t2),(61)-\partial_i\partial_0D^{\mathrm{phys}}_{0j'}=cE_{ij}/(2t^2), \tag*{(61)}
−∂0∂jDi0′phys=cEij/(2t2),(62)-\partial_0\partial_jD^{\mathrm{phys}}_{i0'}=cE_{ij}/(2t^2), \tag*{(62)}
∂0∂0Dij′phys=−6cδij/t4.(63)\partial_0\partial_0D^{\mathrm{phys}}_{ij'}=-6c\delta_{ij}/t^4. \tag*{(63)}

The Euclidean field-strength correlator is the sum of these terms. Including the same electric-field Wick-rotation sign as in the optical calculation gives the local positive-frequency reference

Gijphys(t)=δijπ2t4−Eij6π2t2+⋯ .(64)\mathcal{G}^{\mathrm{phys}}_{ij}(t)=\frac{\delta_{ij}}{\pi^2t^4}-\frac{E_{ij}}{6\pi^2t^2}+\cdots. \tag*{(64)}

Thus the optical result in Eq. (55) is reproduced independently in four physical dimensions, including its sign and tensor normalization. This is a check of the residue, not of its dimensionally evanescent terms.

Dipole anomalous dimension and leading logarithm

Using the curvature term in Eq. (55), the dipole self-energy is

ΣUS=−gB2CFrirj∫0∞dt e−tΔVGij(t).(65)\Sigma_{\mathrm{US}}=-g_B^2C_Fr^ir^j\int_0^\infty dt\,e^{-t\Delta V}\mathcal{G}_{ij}(t). \tag*{(65)}

In D=4−2ϵD=4-2\epsilon dimensions the curvature term has time dependence t−2+2ϵt^{-2+2\epsilon}. Its logarithmic divergence follows from the linear term in the gap expansion, or directly from

∫0∞ω1−2ϵdωω+ΔV=(ΔV)1−2ϵΓ(2−2ϵ)Γ(−1+2ϵ).(66)\int_0^\infty\frac{\omega^{1-2\epsilon}d\omega}{\omega+\Delta V}=(\Delta V)^{1-2\epsilon}\Gamma(2-2\epsilon)\Gamma(-1+2\epsilon). \tag*{(66)}

Since the gamma-function product begins with −1/(2ϵ)-1/(2\epsilon), the pole is

ΣUS,UV(W)=−CFCAαs2W6πrϵ.(67)\Sigma_{\mathrm{US,UV}}^{(W)}=-\frac{C_FC_A\alpha_s^2W}{6\pi r\epsilon}. \tag*{(67)}

The counterterm in the singlet potential has the opposite sign. The infrared pole in Eq. (51) therefore cancels Eq. (67). Renormalizing the potential and the dipole self-energy at the same factorization scale yields

ν∂cW∂ν=CFCA3παs2+O(αs3).(68)\nu\frac{\partial c_W}{\partial\nu}=\frac{C_FC_A}{3\pi}\alpha_s^2+O(\alpha_s^3). \tag*{(68)}

Equation (68) is the leading ultrasoft mixing of the time-ordered dipole product into the local singlet operator (7). In the same normalization, the flat-space correlator gives a running proportional to αsr2(ΔV)3\alpha_sr^2(\Delta V)^3 and hence O(αs/r)O(\alpha_s/r). The Weyl term instead gives a running proportional to αsWΔV\alpha_sW\Delta V and hence O(αs2W/r)O(\alpha_s^2W/r). A finite, scale-independent evanescent subtraction at this order cannot change Eq. (68).

For r−1≫ΔV≫max⁡(∥E∥,ΩE,ΛQCD)r^{-1}\gg\Delta V\gg\max(\sqrt{\lVert E\rVert},\Omega_E,\Lambda_{\mathrm{QCD}}) and curvature gradients small on 1/ΔV1/\Delta V, the local short-time approximation can be used for the leading ultrasoft logarithm in a state with no additional competing scales. The renormalized dipole term supplies +CFCAαs2Wln⁡(ΔV/ν)/(3πr)+C_FC_A\alpha_s^2W\ln(\Delta V/\nu)/(3\pi r); matching the soft coefficient at the inverse separation therefore gives

δEW,log⁡=CFCAαs2W3πrln⁡(CAαs2).(69)\delta\mathcal{E}_{W,\log}=\frac{C_FC_A\alpha_s^2W}{3\pi r}\ln\left(\frac{C_A\alpha_s}{2}\right). \tag*{(69)}

This is the curvature analogue of the leading ultrasoft logarithm O(αs4ln⁡αs/r)O(\alpha_s^4\ln\alpha_s/r) in the flat static energy. The stronger hierarchy is needed for the physical logarithm with the gap as the infrared scale, not for the ultraviolet residue in Eq. (68).

Regulated overlap and spatial connectors

A local parametrrix by itself does not fix the smooth infrared part of a propagator. This is especially relevant for the vector endomorphism in Eq. (16), whose proper-time integral is infrared divergent at b=1b=1. We isolate the leading sensitivity to such smooth terms rather than assigning a physical meaning to the unregulated propagators at large separation.

Static-source contour at fixed regulator

Consider a static completion which agrees with the local geometry in a region of scale κ−1\kappa^{-1} and has compatible gauge boundary conditions, no perturbative zero modes relevant to the correlator, and a gap of order κ\kappa. We assume that a perturbative completion can be chosen with r−1≫κ≫max⁡(ΔV,∥E∥,ΛQCD)r^{-1}\gg\kappa\gg\max(\Delta V,\sqrt{\lVert E\rVert},\Lambda_{\mathrm{QCD}}). These conditions also allow expansions in κr\kappa r and ΔV/κ\Delta V/\kappa. The source paths, subtraction and curvature expansion are held fixed in this comparison.

At fixed such regulator, the spatial connectors prepare and annihilate a gauge-invariant static-pair state. If it has nonzero overlap with the lowest singlet state, the spectral representation takes the form ⟨W(T,r)⟩=∑nZn(r)e−TEn(r)\langle\mathcal{W}(T,r)\rangle=\sum_n Z_n(r)e^{-T\mathcal{E}_n(r)}. Hence

−ln⁡⟨W(T,r)⟩=TE0−ln⁡Z0+O(e−ωκT),(70)-\ln\langle\mathcal{W}(T,r)\rangle=T\mathcal{E}_0-\ln Z_0+O(e^{-\omega_\kappa T}), \tag*{(70)}

where ωκ>0\omega_\kappa>0. Endpoint factors and cusp renormalizations are nonextensive. Equivalently, connected endpoint diagrams cluster into regions of finite time extent at fixed gap, while repetitions of disconnected bulk factors cancel in the logarithm. This argument justifies removing connector overlap effects from the extensive energy under these assumptions, without taking the gap to zero before the long-time limit.

Leading smooth contributions

Decompose the temporal and spatial propagators into their local parametrices and smooth infrared remainders, Hκ=H+δHH_{\kappa}=H+\delta H, Dκ=D+δDD_{\kappa}=D+\delta D. For a symmetric perturbative completion, retain their trace-free curvature-linear pieces at the midpoint:

Cij=δDij(0,0),Mij=∫0∞dt t δH;ij′(t,0).(71)C_{ij}=\delta D_{ij}(0,0),\qquad M_{ij}=\int_{0}^{\infty}\mathrm{d}t\,t\,\delta H_{;ij}'(t,0). \tag*{(71)}

One may compare two gapped completions to obtain decaying remainders, or use analytic continuation of these moments for a difference from the local parametrix. Terms at infinity in integrations by parts must vanish in the convergence domain before continuation. In the following calculation the leading mixed hard–infrared, nonanalytic W/rW/r structure is retained; further terms suppressed by κr\kappa r and analytic all-infrared terms are outside this coefficient.

The only constant trace-free spatial-propagator addition in Eq. (30) that survives at this order multiplies 4H;ij′4H_{;ij}'. Inserting the two hard Coulomb legs gives a correction to the Coulomb Green function

δGse=4gB2CACij∫ddq(2π)dqiqjeiq⋅r(q2)3.(72)\delta G_{\mathrm{se}}=4g_{B}^{2}C_{A}C_{ij}\int\frac{\mathrm{d}^{d}q}{(2\pi)^{d}}\frac{q_iq_j e^{i\mathbf{q}\cdot\mathbf{r}}}{(q^{2})^{3}}. \tag*{(72)}

Using the exact flat identity

Cij∫ddq(2π)dqiqjeiq⋅r(q2)3=−Cij∂i∂j(k3r6−d)=−18G1,d(r)Cijrirj,(73)C_{ij}\int\frac{\mathrm{d}^{d}q}{(2\pi)^{d}}\frac{q_iq_j e^{i\mathbf{q}\cdot\mathbf{r}}}{(q^{2})^{3}} =-C_{ij}\partial_i\partial_j\left(k_{3}r^{6-d}\right) =-\frac{1}{8}G_{1,d}(r)C_{ij}r^ir^j, \tag*{(73)}

where G1,d=k1r2−dG_{1,d}=k_{1}r^{2-d}, the corresponding singlet term is

δVse=gB2CFΔVdrirjCij,ΔVd=CAgB22G1,d.(74)\delta V_{\mathrm{se}}=g_{B}^{2}C_{F}\Delta V_{d}r^{i}r^{j}C_{ij},\qquad\Delta V_{d}=\frac{C_{A}g_{B}^{2}}{2}G_{1,d}. \tag*{(74)}

This identity is dimensionally exact and so also controls the finite part multiplying an evanescent coefficient.

For the time-line web, the same- versus opposite-source propagators for the smooth remainder differ by δA(t)−δB(t)=rirjδH;ij′(t,0)/2+O(r4)\delta A(t)-\delta B(t)=r^{i}r^{j}\delta H_{;ij}'(t,0)/2+O(r^{4}). In the leading mixed region of Eq. (41) one exchange is hard and the other has time separation of order κ−1\kappa^{-1}. Expanding the ordered time factors accordingly gives

δImixed=12∫0∞Bhard(v)dv∫0∞t[δA(t)−δB(t)]dt=18G1,drirjMij.(75)\begin{aligned} \delta I_{\mathrm{mixed}}&=\frac{1}{2}\int_{0}^{\infty}B_{\mathrm{hard}}(v)\mathrm{d}v\int_{0}^{\infty}t[\delta A(t)-\delta B(t)]\mathrm{d}t\\ &=\frac{1}{8}G_{1,d}r^{i}r^{j}M_{ij}. \tag*{(75)} \end{aligned}

The separated hard–infrared term of the crossed contribution has additional powers of κr\kappa r; the all-infrared terms do not generate a leading nonanalytic 1/r1/r Coulomb factor. Therefore

δVW=−gB2CFΔVdrirjMij.(76)\delta V_{W}=-g_{B}^{2}C_{F}\Delta V_{d}r^{i}r^{j}M_{ij}. \tag*{(76)}

Combining Eqs. (74) and (76) yields the leading sensitivity gB2CFΔVdrirj(Cij−Mij)g_{B}^{2}C_{F}\Delta V_{d}r^{i}r^{j}(C_{ij}-M_{ij}).

In the ultrasoft EFT the corresponding electric correlator variation is δGij=∂t2δDij−δH;ij′\delta G_{ij}=\partial_{t}^{2}\delta D_{ij}-\delta H_{;ij}'. Taking the first time moment,

∫0∞t δGij(t)dt=Cij−Mij,(77)\int_{0}^{\infty}t\,\delta G_{ij}(t)\mathrm{d}t=C_{ij}-M_{ij}, \tag*{(77)}

by integration by parts with the stated boundary prescription. Expanding Eq. (65) through first order in ΔVd/κ\Delta V_{d}/\kappa leads to the identity

δVse+δVW=δΣUS(g4).(78)\boxed{\delta V_{\mathrm{se}}+\delta V_{W}=\delta\Sigma_{\mathrm{US}}^{(g^{4})}}. \tag*{(78)}

The cancellation occurs prior to removal of the dimensional regulator and includes the finite terms in the leading smooth overlap. The explicit proper-time separation check in App. B.1 verifies both the pole and the finite cutoff-dependent part of this cancellation.

Finite matching and subtraction

With the pole and the infrared overlap identified, we can specify the finite subtraction and assemble the matched coefficient. For illustration of the finite convention, extending the local spectral reference in Eq. (55) over all frequencies and subtracting minimally in the tree-normalized convention at scale ν\nu gives

Σref(E)=g2CFWΔV6π2[ln⁡ΔVν+12+ln⁡2].(79)\Sigma_{\mathrm{ref}}^{(E)}=\frac{g^{2}C_{F}W\Delta V}{6\pi^{2}}\left[\ln\frac{\Delta V}{\nu}+\frac{1}{2}+\ln2\right]. \tag*{(79)}

One may derive the constant directly in Euclidean time using ∫0∞t−2be−tΔVdt=Γ(1−2b)(ΔV)2b−1\int_{0}^{\infty}t^{-2b}e^{-t\Delta V}\mathrm{d}t=\Gamma(1-2b)(\Delta V)^{2b-1}. In this reference expression the gap is held fixed in the tree-normalized bracket; evanescent factors from the tree normalization must be retained consistently when converting the bracket to a dimensionful subtraction. A physical correlator contains a finite remainder relative to Eq. (55); Eq. (79) is not a calculation of that remainder. The physical hierarchy for use of this reference was specified above. The auxiliary matching completion in Sec. VC instead has gap κ≫ΔV\kappa\gg\Delta V and is used in the matching expansion before regulator removal.

The matched soft logarithm is −4CALV/3-4C_{A}L_{V}/3 in Eq. (82). The two factorization derivatives cancel:

ν∂νVs(E)=−ν∂νΣref(E)=CFCAαs2W3πr.(80)\nu\partial_{\nu}V_{s}^{(E)}=-\nu\partial_{\nu}\Sigma_{\mathrm{ref}}^{(E)}=\frac{C_{F}C_{A}\alpha_{s}^{2}W}{3\pi r}. \tag*{(80)}

In addition, μdαs/dμ=−β0αs2/(2π)+⋯\mu\mathrm{d}\alpha_{s}/\mathrm{d}\mu=-\beta_{0}\alpha_{s}^{2}/(2\pi)+\cdots cancels the explicit LμL_{\mu} dependence through the displayed order. Collecting the finite optical polarization and web terms, physical-metric restoration, and the ultrasoft subtraction, we obtain

Vs(E)(r;μ,ν)=−CFαs(μ)rW[524+αs(μ)4πAE(μ,ν)],(81)V_s^{(E)}(r;\mu,\nu)=-\frac{C_F\alpha_s(\mu)}{r}W\left[\frac{5}{24}+\frac{\alpha_s(\mu)}{4\pi}A_E(\mu,\nu)\right], \tag*{(81)}
AE(μ,ν)=524(a1+2β0Lμ)+CA(4736−43Lν)+TFnf18,(82)A_E(\mu,\nu)=\frac{5}{24}(a_1+2\beta_0L_\mu)+C_A\left(\frac{47}{36}-\frac{4}{3}L_\nu\right)+\frac{T_Fn_f}{18}, \tag*{(82)}

The flat static potential is −CFαs[1+αs(a1+2β0Lμ)/(4π)]/r-C_F\alpha_s[1+\alpha_s(a_1+2\beta_0L_\mu)/(4\pi)]/r. The first line of Eq. (82) arises from the scalar-covariantized flat correction; the remaining terms require the anisotropic polarization and the connected Wilson-line web. The ν\nu-logarithm agrees with Eq. (68), and the explicit μ\mu-dependence cancels the running of the tree coupling.

OriginCAC_ATFnfT_Fn_f
Local polarization, Rz=2WR_z=2W25/3625/365/185/18
Connected web (log and pole only)00
Optical distance and Weyl restoration11/1811/18−2/9-2/9
Extra finite coefficient47/3647/361/181/18

Table II. Additional finite terms in the coefficient multiplying Wαs/(4π)W\alpha_s/(4\pi), excluding 5(a1+2β0Lμ)/245(a_1+2\beta_0L_\mu)/24. Color factors are not included in the individual numerical entries. The remaining −4CALν/3-4C_AL_\nu/3 is separately fixed by ultrasoft subtraction.

Equivalently,

AE=CA(437216+5536Lμ−43Lν)−TFnf(1127+59Lμ).(83)A_E=C_A\left(\frac{437}{216}+\frac{55}{36}L_\mu-\frac{4}{3}L_\nu\right)-T_Fn_f\left(\frac{11}{27}+\frac{5}{9}L_\mu\right). \tag*{(83)}

For SU(3)SU(3) and μ=ν=e−γE/r\mu=\nu=e^{-\gamma_E}/r this reduces to 437/72−11nf/54437/72-11n_f/54.

To demonstrate cancellation of the factorization scale in a finite expression, combine Eq. (81) with the reference self-energy Eq. (79). In the normalization of AEA_E, their sum has the reference bracket

AE,refcomb=512β0Lμ−1127TFnf+CA[293216−43(ln⁡(CAαs)+γE)].(84)A_{E,\mathrm{ref}}^{\mathrm{comb}}=\frac{5}{12}\beta_0L_\mu-\frac{11}{27}T_Fn_f+C_A\left[\frac{293}{216}-\frac{4}{3}\left(\ln(C_A\alpha_s)+\gamma_E\right)\right]. \tag*{(84)}

The ln⁡αs\ln\alpha_s term reproduces Eq. (69). The displayed constant is the sum with the local spectral reference, rather than with the finite state-dependent remainder of the physical correlator.

Finite continuation transformation

Consider a family of trace-free optical continuations with the same four-dimensional endpoint geometry and with R~ij=2(1+ηϵ)Eij\widetilde{R}_{ij}=2(1+\eta\epsilon)E_{ij} at fixed WW. The optical distance is held fixed. The four-dimensional tree coefficient and logarithms are unchanged. This family is a useful test because the optical Ricci tensor of a spatial conformal transformation of a d+1d+1 dimensional Ricci-flat static geometry would instead be (d−1)Eij(d-1)E_{ij}: the spatial Ricci tensor is EijE_{ij}, and the transformation contributes (d−2)Eij(d-2)E_{ij}. Its optical curvature normalization corresponds to η=−1\eta=-1. This does not account for the additional nonconformal bare vertices in a fully physical d+1d+1 dimensional theory.

All additional polarization and web terms linear in RzR_z acquire the factor (1+ηϵ)(1+\eta\epsilon); the curvature-dependent pole in the normalized soft bracket is −2CA/(3ϵ)-2C_A/(3\epsilon). In the same tree-normalized scheme the flat ultraviolet pole is canceled before taking the limit. The finite fermionic extra has no remaining pole and is unchanged under this particular deformation. Thus minimal subtraction in the deformed continuation gives

AE(η)=AE(0)−2η3CA.(85)A_E^{(\eta)}=A_E^{(0)}-\frac{2\eta}{3}C_A. \tag*{(85)}

The normalized reference self-energy has the opposite pole +2CA/(3ϵ)+2C_A/(3\epsilon) and its electric tensor acquires precisely the same (1+ηϵ)(1+\eta\epsilon) factor. Define AUSA_{\mathrm{US}} by Σref(E)=−CFαs2WAUS/(4πr)\Sigma_{\mathrm{ref}}^{(E)}=-C_F\alpha_s^2W A_{\mathrm{US}}/(4\pi r). The corresponding finite change is

AUS(η)=AUS(0)+2η3CA.(86)A_{\mathrm{US}}^{(\eta)}=A_{\mathrm{US}}^{(0)}+\frac{2\eta}{3}C_A. \tag*{(86)}

Consequently AE(η)+AUS(η)=AE,refcombA_E^{(\eta)}+A_{\mathrm{US}}^{(\eta)}=A_{E,\mathrm{ref}}^{\mathrm{comb}} for every η\eta. For η=−1\eta=-1, the extra soft gluonic constant is 71/3671/36 instead of 47/3647/36, with a compensating −2CA/3-2C_A/3 shift in the normalized reference bracket. This illustrates explicitly why the individual soft constant cannot be called continuation-independent. It tests a common curvature normalization change, not arbitrary changes to the local propagator algebra, ultraviolet scheme, source operators or evanescent vertices.

The finite non-Abelian constant is quoted in ordinary Feynman gauge with the optical continuation defined in Sec. III; the closed Wilson loop and the adjoint-linked electric correlator are gauge invariant in a compatible regulator. The continuation test concerns the common trace-free curvature rescaling, while the physical-metric light-flavor calculation below tests a different set of vertices.

Physical-metric check of the light-flavor term

A separate check of the fermion bubble follows from the physical Dirac parametrix. We continue the physical vacuum condition Rμν[g]=0R_{\mu\nu}[g] = 0 in D=d+1D = d + 1 dimensions. The corresponding optical relation is R~ij=(d−1)Eij\widetilde{R}_{ij} = (d - 1)E_{ij} rather than 2Eij2E_{ij} at nonzero ϵ\epsilon. The following calculation retains the physical evanescent Maxwell kinetic operator.

The Lichnerowicz formula, Ricci flatness, and the algebraic Bianchi identity imply, to linear curvature without curvature derivatives, that the contracted derivative of the spin parallel propagator has no curvature contribution: γμRμνabγab\gamma^\mu R_{\mu\nu ab}\gamma^{ab} reduces to a Ricci contraction. The spinor Green function thus has the form

Sg(x,y)=2bcbγμvμ(sg2)b+1Pspin(x,y),vμ=∇xμσg,(87)S_g(x,y)=\frac{2b c_b\gamma_\mu v^\mu}{(s_g^2)^{b+1}}P_{\mathrm{spin}}(x,y),\qquad v^\mu=\nabla_x^\mu\sigma_g, \tag*{(87)}

where sg2=2σgs_g^2 = 2\sigma_g. The Van Vleck factor is unity at this order in physical Ricci-flat geometry. For the unit static time directions ux,uyu_x,u_y and their parallel transport, the spinor trace (after removing the reversed separation sign) is

14tr⁡(u ⁣ ⁣ ⁣/xϕ ⁣ ⁣ ⁣/xu ⁣ ⁣ ⁣/yϕ ⁣ ⁣ ⁣/y)=2(ux⋅v)(uyP⋅v)−(ux⋅uyP)v2=2pt2NxNy−sg2,(88)\frac{1}{4}\operatorname{tr}(u\!\!\!/_{x}\phi\!\!\!/_{x}u\!\!\!/_{y}\phi\!\!\!/_{y})=2(u_x\cdot v)(u_y^P\cdot v)-(u_x\cdot u_y^P)v^2 = \frac{2p_t^2}{N_xN_y}-s_g^2, \tag*{(88)}

where ptp_t is the conserved geodesic momentum associated with the frozen time translation and ux⋅uyP=1+O(E2)u_x\cdot u_y^P=1+O(E^2).

Write f=ln⁡Nf=\ln N and let fˉ\bar f be its average along the zeroth-order optical segment, with Z=t2+ℓ2Z=t^2+\ell^2. Varying the geodesic action gives sg2=Z(1+2fˉ)s_g^2=Z(1+2\bar f) and pt=t(1+fˉ)p_t=t(1+\bar f). For the quadratic lapse,

Qf=fx+fy−2fˉ=Eijzizj6.(89)Q_f=f_x+f_y-2\bar f=\frac{E_{ij}z^iz^j}{6}. \tag*{(89)}

The two color-current vertices and volume factors contribute NxdNydN_x^dN_y^d relative to the optical measures. Expanding the trace times the Green-function denominators gives the integrand (up to its flat normalization)

2t2−Z+Qf[d(2t2−Z)−2t2]Zd+1.(90)\frac{2t^2-Z+Q_f[d(2t^2-Z)-2t^2]}{Z^{d+1}}. \tag*{(90)}

Using the integrals in Sec. IV B, the relative static polarization factor becomes

1+16(d+1d−1)Eijzizj.(91)1+\frac{1}{6}\left(d+\frac{1}{d-1}\right)E_{ij}z^iz^j. \tag*{(91)}

The curvature factor from the optical spinor kernel equals (91) after substituting R~ij=(d−1)Eij\widetilde{R}_{ij}=(d-1)E_{ij}. This checks the curvature numerator at continued dd, not only at d=3d=3.

The extra anisotropic factor relative to the scalar Van Vleck term, in units of EijzizjE_{ij}z^iz^j, is

Pf(d)=16(d+1d−1)−d−112=512−ϵ12+O(ϵ2).(92)P_f(d)=\frac{1}{6}\left(d+\frac{1}{d-1}\right)-\frac{d-1}{12} =\frac{5}{12}-\frac{\epsilon}{12}+O(\epsilon^2). \tag*{(92)}

Multiplying by K(b)=−ϵ/2+O(ϵ2)K(b)=-\epsilon/2+O(\epsilon^2) and the flat fermionic pole −4/(3ϵ)-4/(3\epsilon) gives the additional finite term 5TFnf/185T_Fn_f/18 in units of WW.

The physical dimensionally continued Maxwell action contains the static density gg00gij=N−2ϵγ~γ~ij\sqrt{g}g^{00}g^{ij}=N^{-2\epsilon}\sqrt{\widetilde{\gamma}}\widetilde{\gamma}^{ij}. Its operator is P=L−2ϵLfP=L-2\epsilon L_f, Lf=−∇~i(f∇~i)L_f=-\widetilde{\nabla}_i(f\widetilde{\nabla}^i), to the required order. With G=L−1G=L^{-1} and ∇~2f=0\widetilde{\nabla}^2f=0 locally,

GP=G+ϵ(fG+Gf)+O(ϵ2,E2).(93)G_P=G+\epsilon(fG+Gf)+O(\epsilon^2,E^2). \tag*{(93)}

Because the divergent part of the massless Dirac bubble is proportional to LL, the two external legs supply GPLGPG_PL G_P, whereas the coupling counterterm supplies GPG_P. Their difference times β0,f/ϵ\beta_{0,f}/\epsilon leaves

β0,fϵ(GPLGP−GP)=β0,f(fG+Gf),(94)\frac{\beta_{0,f}}{\epsilon}(G_PL G_P-G_P)=\beta_{0,f}(fG+Gf), \tag*{(94)}

where β0,f=−4TFnf/3\beta_{0,f}=-4T_Fn_f/3. At the symmetric endpoints f++f−=W/4f_++f_-=W/4, so the leg correction is −TFnf/3-T_Fn_f/3 in units of WW. The distance conversion supplies −β0,f/12=TFnf/9-\beta_{0,f}/12=T_Fn_f/9. Thus the net additional term is

TFnf(518−13+19)=TFnf18.(95)T_Fn_f\left(\frac{5}{18}-\frac{1}{3}+\frac{1}{9}\right)=\frac{T_Fn_f}{18}. \tag*{(95)}

Including the scalar-covariantized flat-space term, the full light-flavor bracket is −TFnf(11/27+5Lμ/9)-T_Fn_f(11/27+5L_\mu/9), in agreement with (82). This calculation uses the physical Ricci-flat spinor propagator and evanescent Coulomb legs, with the same local matching and coupling-renormalization convention.

Discussion

A tidal field modifies the renormalization structure of pNRQCD. The short-time electric Weyl contribution reduces the gap dependence of the dipole divergence from (ΔV)3(\Delta V)^3 to ΔV\Delta V, generating anomalous running of the operator OWO_W at order αs2\alpha_s^2 rather than αs4\alpha_s^4. The soft Wilson-loop calculation reproduces the infrared residue required by this running. The finite hard matching coefficient supplements the anomalous dimension with the boundary condition in the specified subtraction convention.

The result applies locally to vacuum tidal fields. For example, a radial freely falling tetrad in Schwarzschild spacetime has Eij=(M/R3)diag⁡(−2,1,1)E_{ij}=(M/R^3)\operatorname{diag}(-2,1,1), with RR the areal radius. In a Coulombic bound state, direct curvature corrections to the static interaction are parametrically of the same order as relativistic kinetic tidal terms; a complete spectroscopic calculation requires both. The static EFT result does not depend on that bound-state calculation or on extrapolation into a large-curvature regime. Its principal implication is the early onset of ultrasoft renormalization and the corresponding logarithm in (69).

Conclusion

The electric Weyl tensor generates a new local ultraviolet term in the chromoelectric dipole self-energy. In the static singlet sector its anomalous dimension, Eq. (68), is of order αs2\alpha_s^2 and fixes the leading curvature logarithm in the weak-coupling static energy. The one-loop Wilson-loop matching yields the corresponding compensating infrared pole and the finite coefficient in Eq. (82). The light-flavor term is reproduced by a separate calculation in the physical Ricci-flat continuation. These results determine the leading curvature-dependent running and its one-loop boundary condition for pNRQCD in a vacuum tidal field.

Data Availability

No data were created or analyzed in this study.

Bitensor identities and flat-space normalization

Local derivative identities

The bitensor derivatives used in Eq. (30) are, in the common parallel frame,

gij′σ;ij′=−d−Rz6,σ=ℓ22,(96)g^{ij'}\sigma_{;ij'}=-d-\frac{R_z}{6},\qquad\sigma=\frac{\ell^2}{2}, \tag*{(96)}
∇~xU=R~⋅z/6,∇~yU=−R~⋅z/6,(97)\widetilde{\nabla}_x U=\widetilde{R}\cdot z/6,\qquad\widetilde{\nabla}_y U=-\widetilde{R}\cdot z/6, \tag*{(97)}
gij′U;ij′=−R~/6=0.(98)g^{ij'}U_{;ij'}=-\widetilde{R}/6=0. \tag*{(98)}

Flat normalization from momentum space

An independent check is available from the flat ordinary-gauge three-gluon vertex and Faddeev–Popov loop. At static momentum q=(0,q)q=(0,\mathbf{q}), the contracted numerator including the gluon symmetry factor and ghost subtraction is

(2d−2)k02+12∣k−q∣2+12∣k+2q∣2.(99)(2d-2)k_0^2+\frac{1}{2}\lvert\mathbf{k}-\mathbf{q}\rvert^2+\frac{1}{2}\lvert\mathbf{k}+2\mathbf{q}\rvert^2. \tag*{(99)}

The gluon-only numerator has (2d−1)k02(2d-1)k_0^2; the ghost loop subtracts k02k_0^2. In d+1d+1 spacetime dimensions the bubble integral is

I0(q)=∫dd+1k(2π)d+11k2(k+q)2(100)I_0(q)=\int\frac{\mathrm{d}^{d+1}k}{(2\pi)^{d+1}}\frac{1}{k^2(k+q)^2} \tag*{(100)}
=(q2)−ϵ(4π)2−ϵΓ(ϵ)Γ(1−ϵ)2Γ(2−2ϵ).(101)=\frac{(q^2)^{-\epsilon}}{(4\pi)^{2-\epsilon}}\frac{\Gamma(\epsilon)\Gamma(1-\epsilon)^2}{\Gamma(2-2\epsilon)}. \tag*{(101)}

Using ∫k⋅q/[k2(k+q)2]=−q2I0/2\int k\cdot q/[k^2(k+q)^2]=-q^2I_0/2 and ∫k02/[k2(k+q)2]=−q2I0/(4d)\int k_0^2/[k^2(k+q)^2]=-q^2I_0/(4d), with the same integration measure, the contracted bubble is (3d+1)q2I0/(2d)(3d+1)q^2I_0/(2d). Fourier transformation gives Eq. (36).

Auxiliary regulator integrals and dimensional expansions

Proper-time separation

One can explicitly test the moment identity by splitting the local heat-kernel integral at s=S=κ−2s=S=\kappa^{-2}. This is an auxiliary separation of the parametrix, not by itself a gauge-invariant regulator of the complete theory. Removing the tail s>Ss>S from the vector endomorphism changes the coincident propagator by

Cij=R~ij(4π)b+1∫S∞s−bds=R~ij(4π)b+1S1−bb−1,(102)C_{ij}=\frac{\widetilde{R}_{ij}}{(4\pi)^{b+1}}\int_S^\infty s^{-b}\mathrm{d}s=\frac{\widetilde{R}_{ij}}{(4\pi)^{b+1}}\frac{S^{1-b}}{b-1}, \tag*{(102)}
Mij=R~ij3(4π)b+1S1−bb−1.(103)M_{ij}=\frac{\widetilde{R}_{ij}}{3(4\pi)^{b+1}}\frac{S^{1-b}}{b-1}. \tag*{(103)}

Here MijM_{ij} follows from the mixed derivative of UU and ∫0∞te−t2/(4s)dt=2s\int_0^\infty te^{-t^2/(4s)}\mathrm{d}t=2s. The tails are evaluated for Re⁡b>1\operatorname{Re} b>1 and continued afterwards. Thus

Cij−Mij=−4Eij3(4π)2−ϵκ−2ϵϵ.(104)C_{ij}-M_{ij}=-\frac{4E_{ij}}{3(4\pi)^{2-\epsilon}}\frac{\kappa^{-2\epsilon}}{\epsilon}. \tag*{(104)}

Normalized to the dimensionally continued tree exchange, the additional one-loop bracket is

Aov=2CA3ϵeγEϵ(μκ)2ϵ=CA[23ϵ+43ln⁡μκ+23γE]+O(ϵ).(105)\begin{aligned} A_{\mathrm{ov}}=\frac{2C_A}{3\epsilon}e^{\gamma_E\epsilon}\left(\frac{\mu}{\kappa}\right)^{2\epsilon} \\ &=C_A\left[\frac{2}{3\epsilon}+\frac{4}{3}\ln\frac{\mu}{\kappa}+\frac{2}{3}\gamma_E\right]+O(\epsilon). \tag*{(105)} \end{aligned}

This explicitly reproduces the opposite of the soft infrared pole in the bracket, together with its finite separation constant. Subtracting the EFT overlap renormalized at ν\nu removes both ln⁡κ\ln\kappa and the auxiliary 2γE/32\gamma_E/3 constant. This gives Eq. (82) in the prescription used here.

Laurent-series identities

We collect the Laurent-series inputs for the finite coefficient (each loop normalized by the dimensionally continued tree Green function and its own color factor):

fse=53ϵ+319+103Lϵ,(106)f_{\mathrm{se}}=\frac{5}{3\epsilon}+\frac{31}{9}+\frac{10}{3}L_\epsilon, \tag*{(106)}
fW=2ϵ+4Lϵ,(107)f_{\mathrm{W}}=\frac{2}{\epsilon}+4L_\epsilon, \tag*{(107)}
ff=−43ϵ−209−83Lϵ,(108)f_{\mathrm{f}}=-\frac{4}{3\epsilon}-\frac{20}{9}-\frac{8}{3}L_\epsilon, \tag*{(108)}
ρg=−310+(497600)ϵ,(109)\rho_g=-\frac{3}{10}+\left(\frac{497}{600}\right)\epsilon, \tag*{(109)}
ρf=−548ϵ.(110)\rho_f=-\frac{5}{48}\epsilon. \tag*{(110)}

Only the displayed orders are required in the products. The gluon residue is −1/2+2/12=−1/3-1/2+2/12=-1/3 per RzR_z and the finite part is 25/7225/72. The fermion residue vanishes because ρf=O(ϵ)\rho_f=O(\epsilon) while the product has finite part 5/365/36. The isotropic pole is 5CA/3+2CA−4TFnf/3=β05C_A/3+2C_A-4T_Fn_f/3=\beta_0. The ordinary coupling renormalization removes that scalar-covariantized pole, including the tree factor UU. The residual infrared pole is treated separately, using the same dimensionally continued tree normalization in the soft and ultrasoft sectors.

The short-distance curvature calculation is performed at O(E)O(E), retaining the nonanalytic ErEr behavior. Curvature derivatives, higher multipoles and powers of curvature are not included. The numerical constants quoted here are exact rational numbers within this expansion.

References

References

  1. [1]N. Brambilla, A. Pineda, J. Soto, and A. Vairo, “Potential NRQCD: An effective theory for heavy quarkonium,” Nucl. Phys. B 566, 275 (2000), arXiv:hep-ph/9907240.DOI
  2. [2]N. Brambilla, A. Pineda, J. Soto, and A. Vairo, “Effective-field theories for heavy quarkonium,” Rev. Mod. Phys. 77, 1423 (2005), doi:10.1103/RevModPhys.77.1423.DOI
  3. [3]A. Pineda, “Review of heavy quarkonium at weak coupling,” Prog. Part. Nucl. Phys. 67, 735 (2012), arXiv:1111.0165.DOI
  4. [4]T. Appelquist, M. Dine, and I. J. Muzinich, “Static limit of quantum chromodynamics,” Phys. Rev. D 17, 2074 (1978), doi:10.1103/PhysRevD.17.2074.DOI
  5. [5]N. Brambilla, A. Pineda, J. Soto, and A. Vairo, “Infrared behavior of the static potential in perturbative QCD,” Phys. Rev. D 60, 091502(R) (1999), doi:10.1103/PhysRevD.60.091502.DOI
  6. [6]A. Pineda and J. Soto, “The renormalization group improvement of the QCD static potentials,” Phys. Lett. B 495, 323 (2000), arXiv:hep-ph/0007197.DOI
  7. [7]N. Brambilla, J. Ghiglieri, A. Vairo, and P. Petreczky, “Static quark-antiquark pairs at finite temperature,” Phys. Rev. D 78, 014017 (2008), arXiv:0804.0993.DOI
  8. [8]J. A. Caicedo and L. F. Urrutia, “Relativistic two-body Coulomb–Breit Hamiltonian in an external weak gravitational field,” Phys. Lett. B 705, 143 (2011), arXiv:1110.0109.DOI
  9. [9]P. Gaete and E. Spallucci, “Confinement from gloudynamics in curved space-time,” Phys. Rev. D 77, 027702 (2008), doi:10.1103/PhysRevD.77.027702.DOI
  10. [10]R. Armillis, C. Corianò, and L. Delle Rose, “Trace anomaly, massless scalars, and the gravitational coupling of QCD,” Phys. Rev. D 82, 064023 (2010), arXiv:1005.4173.DOI
  11. [11]C. Corianò, S. Lionetti, D. Melle, and R. Tommasi, “The gravitational form factors of hadrons from CFT in momentum space and the dilaton in perturbative QCD,” Eur. Phys. J. C 85, 498 (2025), arXiv:2409.05609.arxiv.org/abs/2409.05609
  12. [12]J. F. Donoghue and B. K. El-Menoufi, “Covariant non-local action for massless QED and the curvature expansion,” J. High Energy Phys. 10, 044 (2015), arXiv:1507.06321.
  13. [13]T. S. Bunch and L. Parker, “Feynman propagator in curved spacetime: A momentum-space representation,” Phys. Rev. D 20, 2499 (1979), doi:10.1103/PhysRevD.20.2499.DOI
  14. [14]D. V. Vassilevich, “Heat kernel expansion: User’s manual,” Phys. Rept. 388, 279 (2003), arXiv:hep-th/0306138.DOI
  15. [15]D. J. Toms, “Local momentum space and the vector field,” Phys. Rev. D 90, 044072 (2014), arXiv:1408.0636.DOI
  16. [16]J. C. Collins, A. Duncan, and S. D. Joglekar, “Trace and dilatation anomalies in gauge theories,” Phys. Rev. D 16, 438 (1977), doi:10.1103/PhysRevD.16.438.DOI
  17. [17]M. B. Fröb and M. Taslimi Tehrani, “Green’s functions and Hadamard parametrices for vector and tensor fields in general linear covariant gauges,” Phys. Rev. D 97, 025022 (2018), arXiv:1708.00444.DOI

Paper details

vlnc:W_uqwjwbsb
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents