betaFeedback

WorksPhysics and Astronomy

Fourier Transform of the Scalar Propagator with Quantum Light-Cone Fluctuations

Abstract

We derive the Lorentzian Fourier transform of the Gaussian-smeared light-cone contribution to the scalar propagator in Krein-space quantization in terms of a Meijer GG-function. We characterize its distributional structure, recover the massive propagator by adding the Bessel-function tail, and derive an exact expression for the massless one-loop bubble kernel using a position-space prescription.

Introduction

Krein-space quantization introduces an indefinite inner product and auxiliary negative-norm degrees of freedom. For the scalar two-point function considered here, the resulting time-ordered kernel is the real, principal-value part of the conventional Feynman propagator [1]. Metric fluctuations can broaden the classical light cone, motivating the replacement of its delta-function contribution by a Gaussian [2, 3]. We adopt a Gaussian of finite width, normalized as a distribution in the invariant interval.

The Fourier transform of this Gaussian is not an absolutely convergent four-dimensional integral, since the Gaussian remains constant along the noncompact surfaces of fixed Minkowski interval. Consequently, treating it as a Euclidean radial transform would alter the problem. We therefore compute the Lorentzian Fourier transform in four-dimensional Minkowski space in the sense of tempered distributions. An auxiliary Gaussian integral reduces the calculation to an Abel-summed sine integral, which we evaluate in terms of a Meijer GG-function.

We obtain explicit expressions for timelike and spacelike momenta and determine the distributional extension at the null cone. This analysis separates the principal-value massless pole from a continuous remainder. Adding the Bessel-function tail yields the massive propagator and demonstrates the cancellation of the massless pole when the light-cone smearing has unit normalization. We then apply this result to derive an exact expression for the massless one-loop bubble kernel using a position-space prescription.

Throughout this work, we use a constant positive smearing parameter μ\mu, which characterizes the Gaussian width and is related to the metric-fluctuation variance. This prescription defines a model for the light-cone contribution rather than a general metric-averaged propagator. In a more general treatment, the variance may depend on the two spacetime points and on the gravitational state; here, we approximate it by a constant to obtain an analytically tractable model.

Conventions and the smearing model

We work in four-dimensional Minkowski spacetime with

ηαβ=diag⁡(1,−1,−1,−1).(1)\eta_{\alpha\beta} = \operatorname{diag}(1,-1,-1,-1). \tag*{(1)}

We set

y=x−x′,q=p2,y2=(y0)2−∣y∣2,σ0=y22.(2)y = x-x', \qquad q = p^2, \qquad y^2 = (y^0)^2-\lvert\mathbf{y}\rvert^2, \qquad\sigma_0 = \frac{y^2}{2}. \tag*{(2)}

Our Fourier-transform conventions are

f~(p)=∫d4y eip⋅yf(y),f(y)=∫d4p(2π)4e−ip⋅yf~(p),(3)\widetilde{f}(p) = \int\mathrm{d}^4y\,e^{ip\cdot y}f(y), \qquad f(y) = \int\frac{\mathrm{d}^4p}{(2\pi)^4}e^{-ip\cdot y}\widetilde{f}(p), \tag*{(3)}

where p⋅y=p0y0−p⋅yp \cdot y = p^0y^0-\mathbf{p}\cdot\mathbf{y}.

The unsmeared scalar time-ordered kernel in Krein-space quantization is defined by [1, 3]

GT,m(0)(y)=−i⟨0∣Tϕ(x)ϕ(x′)∣0⟩,G~T,m(0)(p)=PV⁡1p2−m2.(4)G_{T,m}^{(0)}(y) = -i\langle0|T\phi(x)\phi(x')|0\rangle, \qquad\widetilde{G}_{T,m}^{(0)}(p) = \operatorname{PV}\frac{1}{p^2-m^2}. \tag*{(4)}

Its coordinate-space representation is

GT,m(0)(y)=−18πδ(σ0)+Bm(y)=−14πδ(y2)+Bm(y),(5)G_{T,m}^{(0)}(y) = -\frac{1}{8\pi}\delta(\sigma_0) + B_m(y) = -\frac{1}{4\pi}\delta(y^2) + B_m(y), \tag*{(5)}
Bm(y)=m28πθ(σ0)J1(2m2σ0)2m2σ0,m>0.(6)B_m(y) = \frac{m^2}{8\pi}\theta(\sigma_0)\frac{J_1\left(\sqrt{2m^2\sigma_0}\right)}{\sqrt{2m^2\sigma_0}}, \qquad m>0. \tag*{(6)}

The Bessel contribution is evaluated for σ0>0\sigma_0 > 0 and extended by zero for σ0<0\sigma_0 < 0. At the light cone, the ratio is defined by its continuous limit J1(z)/z→1/2J_1(z)/z \to1/2. The value assigned to θ(0)\theta(0) does not affect the distribution. For the massless field, we set B0=0B_0 = 0.

Quantum metric fluctuations motivate a Gaussian smearing of the light-cone contribution [2, 3]. We introduce the normalized Gaussian

δμ(u)=12πμexp⁡(−u24μ2),∫Rδμ(u) du=1,μ>0.(7)\delta_\mu(u) = \frac{1}{2\sqrt{\pi}\mu}\exp\left(-\frac{u^2}{4\mu^2}\right), \qquad\int_{\mathbb{R}}\delta_\mu(u)\,\mathrm{d}u = 1, \qquad\mu>0. \tag*{(7)}

We smear the delta distribution in the Minkowski interval according to

δ(y2)⟶δμ(y2).(8)\delta(y^2) \longrightarrow\delta_\mu(y^2). \tag*{(8)}

Since σ0=y2/2\sigma_0 = y^2/2 and δμ/2(σ0)=2δμ(y2)\delta_{\mu/2}(\sigma_0) = 2\delta_\mu(y^2), this prescription is equivalent to replacing δ(σ0)\delta(\sigma_0) by δμ/2(σ0)\delta_{\mu/2}(\sigma_0) in (5). The resulting propagator model is

Gμ,m(y):=Kμ(y)+Bm(y),G_{\mu,m}(y) := K_\mu(y) + B_m(y),
Kμ(y)=−18ππμexp⁡[−(y2)24μ2].(9)K_\mu(y) = -\frac{1}{8\pi\sqrt{\pi}\mu}\exp\left[-\frac{(y^2)^2}{4\mu^2}\right]. \tag*{(9)}

Here (y2)2(y^2)^2 denotes the square of the Minkowski interval, not a Euclidean radial norm. The prescription smears only the light-cone contribution and leaves the Bessel-function tail unchanged.

For m>0m>0, the Fourier representation of the Bessel contribution is

Bm(y)=∫d4p(2π)4e−ip⋅yPV⁡m2p2(p2−m2).(10)B_m(y)=\int\frac{\mathrm{d}^4p}{(2\pi)^4}e^{-ip\cdot y}\operatorname{PV}\frac{m^2}{p^2(p^2-m^2)}. \tag*{(10)}

The principal-part prescription in this expression is defined by

PV⁡m2p2(p2−m2)≔PV⁡1p2−m2−PV⁡1p2.(11)\operatorname{PV}\frac{m^2}{p^2(p^2-m^2)}\coloneqq\operatorname{PV}\frac{1}{p^2-m^2}-\operatorname{PV}\frac{1}{p^2}. \tag*{(11)}

This specifies the distribution through its partial-fraction decomposition.

The Fourier transform of KμK_\mu is defined by

Kμ(y)=∫d4p(2π)4e−ip⋅yK~μ(p).(12)K_\mu(y)=\int\frac{\mathrm{d}^4p}{(2\pi)^4}e^{-ip\cdot y}\widetilde{K}_\mu(p). \tag*{(12)}

Consequently,

Gμ,m(y)=∫d4p(2π)4e−ip⋅yG~μ,m(p),(13)G_{\mu,m}(y)=\int\frac{\mathrm{d}^4p}{(2\pi)^4}e^{-ip\cdot y}\widetilde{G}_{\mu,m}(p), \tag*{(13)}
G~μ,m(p)=K~μ(p)+PV⁡m2p2(p2−m2),m>0.\widetilde{G}_{\mu,m}(p)=\widetilde{K}_\mu(p)+\operatorname{PV}\frac{m^2}{p^2(p^2-m^2)},\qquad m>0.

For m=0m=0, the Bessel contribution vanishes and G~μ,0=K~μ\widetilde{G}_{\mu,0}=\widetilde{K}_\mu. All Fourier transforms are understood in the sense of tempered distributions.

In units ℏ=c=1\hbar=c=1, the dimensions are

[μ]=mass−2,[Gμ,m]=mass2,[G~μ,m]=mass−2.(14)[\mu]=\mathrm{mass}^{-2},\qquad[G_{\mu,m}]=\mathrm{mass}^{2},\qquad[\widetilde{G}_{\mu,m}]=\mathrm{mass}^{-2}. \tag*{(14)}

If the first-order interval fluctuation σ1\sigma_1 is modeled as a centered Gaussian random variable with constant variance v=⟨σ12⟩>0v=\langle\sigma_1^2\rangle>0, then

⟨δ(σ0+σ1)⟩=12πvexp⁡(−σ022v).(15)\langle\delta(\sigma_0+\sigma_1)\rangle=\frac{1}{\sqrt{2\pi v}}\exp\left(-\frac{\sigma_0^2}{2v}\right). \tag*{(15)}

Under this Gaussian model, averaging broadens the light-cone contribution [2]. Comparison with the smearing in (9) gives μ2=2v\mu^2=2v. The variance depends on the gravitational state and its correlations; a quantitative relation to graviton density requires specifying that state. Here and below, single angle brackets ⟨⋅⟩\langle\cdot\rangle denote expectation values over metric fluctuations.

The model (9) is an approximation to the metric-averaged propagator, rather than a general expression for that average. The normalized smearing ensures the distributional zero-width limit

Gμ,m⟶GT,m(0)in S′(R4),μ↓0.(16)G_{\mu,m}\longrightarrow G_{T,m}^{(0)}\quad\text{in }\mathcal{S}'(\mathbb{R}^4),\qquad\mu\downarrow0. \tag*{(16)}

Here S′(R4)\mathcal{S}'(\mathbb{R}^4) denotes the space of tempered distributions, the continuous dual of the Schwartz space S(R4)\mathcal{S}(\mathbb{R}^4) of smooth functions whose derivatives of every order decrease faster than any inverse power at infinity. The convergence means that

lim⁡μ↓0Gμ,m[φ]=GT,m(0)[φ]for every φ∈S(R4),(17)\lim_{\mu\downarrow0}G_{\mu,m}[\varphi]=G_{T,m}^{(0)}[\varphi]\quad\text{for every }\varphi\in\mathcal{S}(\mathbb{R}^4), \tag*{(17)}

where G[φ]G[\varphi] denotes the action of the distribution GG on the test function φ\varphi. For the bounded function Gμ,mG_{\mu,m}, this action is given by

Gμ,m[φ]=∫R4d4y Gμ,m(y)φ(y).(18)G_{\mu,m}[\varphi]=\int_{\mathbb{R}^4}\mathrm{d}^4y\,G_{\mu,m}(y)\varphi(y). \tag*{(18)}

The alternative Gaussian coefficient −18ππ2μ2-\frac{1}{8\pi}\sqrt{\frac{\pi}{2\mu^2}} multiplies KμK_\mu by π/2\pi/\sqrt{2}. It therefore gives a different zero-width normalization when the Bessel tail is unchanged. We use (9) consistently below; Appendix § 9 records the conversion to the alternative coefficient.

Distributional Lorentzian Fourier transform

Since KμK_\mu is bounded, it defines a tempered distribution on Minkowski spacetime R1,3\mathbb{R}^{1,3}. Its Fourier transform can be obtained by Gaussian regularization:

K~μ(p)=lim⁡ε↓0∫R1,3d4y eip⋅ye−ε[(y0)2+∣y∣2]Kμ(y),(19)\widetilde{K}_\mu(p)=\lim_{\varepsilon\downarrow0}\int_{\mathbb{R}^{1,3}} \mathrm{d}^{4}y\,e^{i p\cdot y}e^{-\varepsilon[(y^{0})^{2}+|\mathbf{y}|^{2}]}K_\mu(y), \tag*{(19)}

where the limit is taken in S′(R4)\mathcal{S}'(\mathbb{R}^{4}). The integration variables remain real. The positive-definite Gaussian regulator ensures absolute convergence for ε>0\varepsilon>0 without a Wick rotation. Although the regulator breaks Lorentz invariance at finite ε\varepsilon, the limiting Fourier transform is Lorentz invariant because KμK_\mu depends only on y2y^{2}.

We use the auxiliary identity

e−(y2)2/(4μ2)=1π∫−∞∞ds e−s2eisy2/μ.(20)e^{-(y^{2})^{2}/(4\mu^{2})}=\frac{1}{\sqrt{\pi}}\int_{-\infty}^{\infty}\mathrm{d}s\,e^{-s^{2}}e^{i s y^{2}/\mu}. \tag*{(20)}

With the regulator retained, the coordinate integrations are Gaussian. For α∈R∖{0}\alpha\in\mathbb{R}\setminus\{0\}, the corresponding distributional Fresnel limit is

∫d4y eip⋅y+iαy2=−iπ2sgn⁡(α)α2e−ip2/(4α).(21)\int\mathrm{d}^{4}y\,e^{i p\cdot y+i\alpha y^{2}}=-i\pi^{2}\frac{\operatorname{sgn}(\alpha)}{\alpha^{2}}e^{-i p^{2}/(4\alpha)}. \tag*{(21)}

The phase factor follows from one time integration and three spatial integrations. Evaluating the coordinate integrals at finite ε\varepsilon and then taking the distributional limit gives

K~μ(p)=iμ8∫−∞∞ds sgn⁡(s)s2e−s2−iμq/(4s).(22)\widetilde{K}_\mu(p)=\frac{i\mu}{8}\int_{-\infty}^{\infty}\mathrm{d}s\,\frac{\operatorname{sgn}(s)}{s^{2}}e^{-s^{2}-i\mu q/(4s)}. \tag*{(22)}

The integral at s=0s=0 inherits its prescription from the original regulated Fourier transform; it is not an ordinary improper integral.

Pairing the contributions from positive and negative ss, and then setting t=1/st=1/s on the positive half-line, yields

K~μ(p)=μ4I(μq4),I(a)=Abel⁡∫0∞dt e−1/t2sin⁡(at).(23)\widetilde{K}_\mu(p)=\frac{\mu}{4}I\left(\frac{\mu q}{4}\right),\qquad I(a)=\operatorname{Abel}\int_{0}^{\infty}\mathrm{d}t\,e^{-1/t^{2}}\sin(at). \tag*{(23)}

Here Abel summation means inserting e−ηte^{-\eta t} and taking η↓0\eta\downarrow0 in the sense of tempered distributions. The unregulated sine integral is not an ordinary improper integral because its amplitude tends to one at infinity.

A useful distributional decomposition is

I(a)=PV⁡1a+∫0∞dt (e−1/t2−1)sin⁡(at).(24)I(a)=\operatorname{PV}\frac{1}{a}+\int_{0}^{\infty}\mathrm{d}t\,(e^{-1/t^{2}}-1)\sin(at). \tag*{(24)}

The first term follows from

∫0∞e−ηtsin⁡(at) dt=aa2+η2⟶PV⁡1ain S′(R),η↓0.(25)\int_{0}^{\infty}e^{-\eta t}\sin(at)\,\mathrm{d}t=\frac{a}{a^{2}+\eta^{2}}\longrightarrow\operatorname{PV}\frac{1}{a}\quad\text{in }\mathcal{S}'(\mathbb{R}),\qquad\eta\downarrow0. \tag*{(25)}

The amplitude in the second integral is integrable:

∫0∞∣e−1/t2−1∣dt<∞.(26)\int_{0}^{\infty}\left|e^{-1/t^{2}}-1\right|\mathrm{d}t<\infty. \tag*{(26)}

Consequently, the second term in (24) defines a continuous odd function of aa.

To justify the corresponding decomposition in four-momentum space, subtract the unsmeared massless light-cone kernel before removing the regulator in (19). In the auxiliary representation, this subtraction replaces e−s2e^{-s^2} by e−s2−1e^{-s^2}-1. After pairing with a Schwartz function, the regulated coordinate integrals are bounded uniformly by a constant times ∣s∣−2|s|^{-2} for s≠0s\ne0: the modulus of the regulated Gaussian prefactor is at most π2μ2/s2\pi^2\mu^2/s^2, and the remaining exponential has modulus at most one. Thus the subtracted integrand has an integrable majorant proportional to

∣e−s2−1∣s2,∫R∣e−s2−1∣s2 ds<∞.(27)\frac{\lvert e^{-s^2}-1\rvert}{s^2}, \qquad\int_{\mathbb{R}}\frac{\lvert e^{-s^2}-1\rvert}{s^2}\,\mathrm{d}s<\infty. \tag*{(27)}

Dominated convergence therefore applies to the subtracted integral after pairing with Schwartz test functions. The unsmeared massless light-cone term is fixed by (5) and transforms to PV⁡(1/p2)\operatorname{PV}(1/p^2). Thus (23) is understood explicitly as

K~μ(p)=PV⁡1p2+μ4∫0∞dt (e−1/t2−1)sin⁡(μp2t4).(28)\widetilde{K}_{\mu}(p)=\operatorname{PV}\frac{1}{p^2}+\frac{\mu}{4}\int_{0}^{\infty}\mathrm{d}t\,\left(e^{-1/t^2}-1\right)\sin\left(\frac{\mu p^2t}{4}\right). \tag*{(28)}

Here

PV⁡1p2:=lim⁡ρ→0p2(p2)2+ρ2in S′(R4).(29)\operatorname{PV}\frac{1}{p^2}:=\lim_{\rho\to0}\frac{p^2}{(p^2)^2+\rho^2}\quad\text{in }S'(\mathbb{R}^4). \tag*{(29)}

This separates the principal-value massless pole from a continuous remainder, which vanishes at p2=0p^2=0, and fixes the transform without additional terms supported on the null cone. All exchanges of integrations are understood with the stated regulators or after pairing with Schwartz test functions.

Meijer GG representation

For a>0a>0, integration by parts in the Abel sense gives an absolutely convergent cosine integral:

I(a)=1a∫0∞f′(t)cos⁡(at) dt,f(t)=e−1/t2,f′(t)=2t−3e−1/t2.(30)I(a)=\frac{1}{a}\int_{0}^{\infty}f'(t)\cos(at)\,\mathrm{d}t,\qquad f(t)=e^{-1/t^2},\quad f'(t)=2t^{-3}e^{-1/t^2}. \tag*{(30)}

Define M[I](s)=∫0∞as−1I(a) da\mathcal{M}[I](s)=\int_{0}^{\infty}a^{s-1}I(a)\,\mathrm{d}a. Using a regulator for the oscillatory interchange gives

M[I](s)=Γ(s−1)cos⁡(π(s−1)2)∫0∞f′(t)t1−s dt=12Γ(s)Γ(s−12)sin⁡(πs2),1<ℜs<2.(31)\begin{aligned} \mathcal{M}[I](s)&=\Gamma(s-1)\cos\left(\frac{\pi(s-1)}{2}\right)\int_{0}^{\infty}f'(t)t^{1-s}\,\mathrm{d}t\\ &=\frac{1}{2}\Gamma(s)\Gamma\left(\frac{s-1}{2}\right)\sin\left(\frac{\pi s}{2}\right),\qquad1<\Re s<2. \tag*{(31)} \end{aligned}

This route avoids using the elementary Mellin transform of the sine outside its initial strip of validity. With s=2us=2u, the duplication and reflection identities give

I(a)=π212πi∫c−i∞c+i∞Γ(u−12)Γ(u+12)Γ(1−u)(a24)−u du,12<c<1.(32)I(a)=\frac{\sqrt{\pi}}{2}\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}\frac{\Gamma\left(u-\frac{1}{2}\right)\Gamma\left(u+\frac{1}{2}\right)}{\Gamma(1-u)}\left(\frac{a^2}{4}\right)^{-u}\,\mathrm{d}u,\qquad\frac{1}{2}<c<1. \tag*{(32)}

Comparison with the Mellin–Barnes definition [4] gives

I(a)=π2G0,32,0(a24 | −−12,12,0),a>0.(33)I(a)=\frac{\sqrt{\pi}}{2}G_{0,3}^{2,0}\left(\frac{a^2}{4}\,\middle|\,\begin{matrix}-\\-\frac{1}{2},\frac{1}{2},0\end{matrix}\right),\qquad a>0. \tag*{(33)}

Oddness supplies the result for negative aa. Consequently, away from q=0q=0,

K~μ(p)=sgn⁡(q)πμ8G0,32,0(μ2q264 | −−12,12,0).(34)\widetilde{K}_{\mu}(p)=\operatorname{sgn}(q)\frac{\sqrt{\pi}\mu}{8}G_{0,3}^{2,0}\left(\frac{\mu^2q^2}{64}\,\middle|\,\begin{matrix}-\\-\frac{1}{2},\frac{1}{2},0\end{matrix}\right). \tag*{(34)}

The factor sgn⁡(q)\operatorname{sgn}(q) is essential. Dependence on q2q^2 alone would give the wrong spacelike sign.

Define the dimensionless function

H(z)=G0,32,0(z|−0,1,12),z>0.(35)H(z)=G^{2,0}_{0,3}\left(z\middle|\begin{matrix}-\\0,1,\frac{1}{2}\end{matrix}\right),\qquad z>0. \tag*{(35)}

The shift identity for the Meijer GG-function gives the equivalent pointwise formula

K~μ(p)=πqH(μ2q264),q≠0.(36)\widetilde{K}_{\mu}(p)=\frac{\sqrt{\pi}}{q}H\left(\frac{\mu^2q^2}{64}\right),\qquad q\ne0. \tag*{(36)}

At the null cone, the Fourier transform determined by (19) has the explicit decomposition

K~μ(p)=PV⁡1p2+Rμ(p2),Rμ(q)=πH(μ2q2/64)−1q,Rμ(0)=0.(37)\widetilde{K}_{\mu}(p)=\operatorname{PV}\frac{1}{p^2}+R_{\mu}(p^2),\qquad R_{\mu}(q)=\frac{\sqrt{\pi}H\left(\mu^2q^2/64\right)-1}{q},\qquad R_{\mu}(0)=0. \tag*{(37)}

Here PV⁡(p2)−1\operatorname{PV}(p^2)^{-1} is the real part of the distributional boundary value of (p2+i0)−1(p^2+i0)^{-1}. Equation (37), equivalently (24), specifies the meaning of the special-function expression at p2=0p^2=0.

Subtracting the massless version of (5) from the massive one fixes the tail transform without an additional integration:

B~m(p)=PV⁡1p2−m2−PV⁡1p2.(38)\widetilde{B}_{m}(p)=\operatorname{PV}\frac{1}{p^2-m^2}-\operatorname{PV}\frac{1}{p^2}. \tag*{(38)}

The frequently used notation PV⁡{m2/[p2(p2−m2)]}\operatorname{PV}\{m^2/[p^2(p^2-m^2)]\} means precisely this difference, not an unspecified product of singular distributions. Combining (37) and (38), we obtain

G~μ,m(p)=PV⁡1p2−m2+Rμ(p2),m≥0.(39)\widetilde{G}_{\mu,m}(p)=\operatorname{PV}\frac{1}{p^2-m^2}+R_{\mu}(p^2),\qquad m\ge0. \tag*{(39)}

For m>0m>0, the massless pole cancels exactly. For m=0m=0, the massless pole remains. The massive physical pole at p2=m2p^2=m^2 is also retained. A principal-value pole is a well-defined distribution, but it is not a bounded function. Both limiting operations are consistent:

lim⁡μ↓0G~μ,m=PV⁡1p2−m2,lim⁡m↓0G~μ,m=K~μ,(40)\lim_{\mu\downarrow0}\widetilde{G}_{\mu,m}=\operatorname{PV}\frac{1}{p^2-m^2},\qquad\lim_{m\downarrow0}\widetilde{G}_{\mu,m}=\widetilde{K}_{\mu}, \tag*{(40)}

in S′(R4)\mathcal{S}'(\mathbb{R}^4). The second limit is not obtained by taking a fixed-mm infrared expansion and then setting m=0m=0.

Infrared and large-invariant-momentum behavior

Expansion near the null cone

The coincident poles in the Mellin–Barnes representation of HH generate logarithms. Its residue expansion is

H(z)=1π+∑n=1∞zn{ln⁡z−ψ(n+1)−ψ(n)−ψ(n+12)}n!(n−1)!Γ(n+12),z>0,(41)H(z)=\frac{1}{\sqrt{\pi}}+\sum_{n=1}^{\infty}\frac{z^n\{\ln z-\psi(n+1)-\psi(n)-\psi(n+\frac{1}{2})\}}{n!(n-1)!\Gamma(n+\frac{1}{2})},\qquad z>0, \tag*{(41)}

where ψ\psi is the digamma function. In particular, as z↓0z\downarrow0,

πH(z)=1+2z[ln⁡z−3+3γ+2ln⁡2]+O(z2∣ln⁡z∣).(42)\sqrt{\pi}H(z)=1+2z[\ln z-3+3\gamma+2\ln2]+O(z^2|\ln z|). \tag*{(42)}

Here γ\gamma denotes the Euler–Mascheroni constant. Thus, for fixed μ\mu and q→0q\to0,

Rμ(q)=μ2q32[ln⁡(μ2q264)−3+3γ+2ln⁡2]+O(μ4q3ln⁡(μ2q2)).(43)R_{\mu}(q)=\frac{\mu^2q}{32}\left[\ln\left(\frac{\mu^2q^2}{64}\right)-3+3\gamma+2\ln2\right]+O\left(\mu^4q^3\ln(\mu^2q^2)\right). \tag*{(43)}

This proves that RμR_\mu is continuous at q=0q=0 but is not differentiable there. For m>0m>0 and ∣q∣≪m2|q|\ll m^2,

G~μ,m(p)=−1m2−qm4+O(q2/m6)+Rμ(q).(44)\widetilde{G}_{\mu,m}(p)=-\frac{1}{m^2}-\frac{q}{m^4}+O(q^2/m^6)+R_\mu(q). \tag*{(44)}

For m=0m=0, the leading term is instead PV⁡(1/q)\operatorname{PV}(1/q). Accordingly, smearing does not remove the massless propagator pole. Whether a particular amplitude has an infrared divergence depends on its full integrand and external kinematics.

Large invariant momentum

The function f′(t)f'(t) in (30) is flat at t=0t=0, and all its derivatives are integrable and vanish at both integration endpoints. Repeated integration by parts therefore shows that, for every integer N>0N>0,

I(a)=O(∣a∣−N),∣a∣→∞.(45)I(a)=O(|a|^{-N}),\qquad|a|\to\infty. \tag*{(45)}

Consequently K~μ\widetilde{K}_\mu decreases faster than every inverse power of ∣q∣|q| for fixed μ>0\mu>0. No particular exponential rate is needed for this conclusion. The massive tail then determines the algebraic asymptotics:

G~μ,m(p)=m2q2+m4q3+O(q−4),∣q∣→∞,m>0.(46)\widetilde{G}_{\mu,m}(p)=\frac{m^2}{q^2}+\frac{m^4}{q^3}+O(q^{-4}),\qquad|q|\to\infty,\quad m>0. \tag*{(46)}

This is an expansion away from the pole surfaces. Large ∣p2∣|p^2| is not the same as large Euclidean norm of the four-momentum: arbitrarily large momenta can remain close to the light cone.

Coordinate-space regularity and a massless bubble kernel

At fixed μ>0\mu>0, KμK_\mu is smooth and bounded. For m>0m>0, the Bessel-function tail Bm(y)B_m(y) is bounded but has a finite jump across the light cone. As y2↓0y^2\downarrow0,

Bm(y)=m216π−m4y2128π+O(m6(y2)2),y2>0,(47)B_m(y)=\frac{m^2}{16\pi}-\frac{m^4y^2}{128\pi}+O\left(m^6(y^2)^2\right),\qquad y^2>0, \tag*{(47)}

whereas Bm(y)=0B_m(y)=0 for y2<0y^2<0. For fixed m>0m>0 and σ0→+∞\sigma_0\to+\infty,

Bm(y)=2−1/4m1/28ππσ0−3/4[cos⁡(m2σ0−3π4)+O((m2σ0)−1)].(48)B_m(y)=\frac{2^{-1/4}m^{1/2}}{8\pi\sqrt{\pi}}\sigma_0^{-3/4}\left[\cos\left(m\sqrt{2\sigma_0}-\frac{3\pi}{4}\right)+O\left((m\sqrt{2\sigma_0})^{-1}\right)\right]. \tag*{(48)}

Thus the full kernel has finite one-sided limits at the light cone, although it is not continuous there for m>0m>0. Its value on the light cone, including at coincidence, depends on the convention chosen for the jump. This choice does not affect the associated distribution.

For a concrete application, let PP denote the external four-momentum flowing through the bubble, and set Q=P2Q=P^2. We define the massless bubble kernel as the Fourier transform of an ordinary bounded product:

Bμ(P):=F[Kμ2](P),(49)\mathcal{B}_\mu(P):=\mathcal{F}[K_\mu^2](P), \tag*{(49)}

where the Fourier transform is understood in S′(R4)\mathcal{S}'(\mathbb{R}^4). Direct multiplication gives

Kμ(y)2=164π3μ2e−(y2)/(2μ2)=−182 π3/2μK~μ/2(y).(50)K_\mu(y)^2=\frac{1}{64\pi^3\mu^2}e^{-(y^2)/(2\mu^2)}=-\frac{1}{8\sqrt{2}\,\pi^{3/2}\mu}\widetilde{K}_{\mu/\sqrt{2}}(y). \tag*{(50)}

Hence the exact result in this prescription is

Bμ(P)=−182 π3/2μK~μ/2(P).(51)\mathcal{B}_\mu(P)=-\frac{1}{8\sqrt{2}\,\pi^{3/2}\mu}\widetilde{K}_{\mu/\sqrt{2}}(P). \tag*{(51)}

For Q≠0Q \ne0, this becomes

Bμ(P)=−sgn⁡(Q)128πG0,32,0(μ2Q2/128|−−12,12,0).(52)\mathcal{B}_{\mu}(P)=-\frac{\operatorname{sgn}(Q)}{128\pi}G^{2,0}_{0,3}\left(\mu^{2}Q^{2}/128\middle|\begin{matrix}-\\-\frac{1}{2},\frac{1}{2},0\end{matrix}\right). \tag*{(52)}

The bubble kernel is dimensionless, as required. Its distributional extension at the null cone is

Bμ(P)=−182π3/2μ[PV⁡1P2+Rμ2(P2)].(53)\mathcal{B}_{\mu}(P)=-\frac{1}{8\sqrt{2}\pi^{3/2}\mu}\left[\operatorname{PV}\frac{1}{P^{2}}+\frac{R_{\mu}}{\sqrt{2}(P^{2})}\right]. \tag*{(53)}

In particular, it retains a principal-value singularity at P2=0P^{2}=0. One may write the symbolic momentum-space convolution

Bμ(P)≡∫d4k(2π)4K~μ(k)K~μ(P−k),(54)\mathcal{B}_{\mu}(P)\equiv\int\frac{\mathrm{d}^{4}k}{(2\pi)^{4}}\widetilde{K}_{\mu}(k)\widetilde{K}_{\mu}(P-k), \tag*{(54)}

provided the right-hand side is understood through (49), or a compatible limiting prescription. Here kk denotes the loop momentum, so the two internal lines carry momenta kk and P−kP-k. A convolution of arbitrary tempered distributions is not automatically defined by an ordinary integral.

For the interaction Lint=−λϕ4/4!\mathcal{L}_{\mathrm{int}}=-\lambda\phi^{4}/4!, we use vertex factor −iλ-i\lambda and internal-line factor iK~μi\widetilde{K}_{\mu}. We adopt the convention S=1+iTS=1+iT and define the invariant scattering amplitude Mfi\mathcal{M}_{fi} by

⟨f∣iT∣i⟩=i(2π)4δ(4)(Pf−Pi)Mfi,(55)\langle f|iT|i\rangle=i(2\pi)^{4}\delta^{(4)}(P_{f}-P_{i})\mathcal{M}_{fi}, \tag*{(55)}

where PiP_i and PfP_f are the total initial and final four-momenta. With these conventions, the formal amputated ss-channel contribution is

iMs(1)(P)=(−iλ)2i22Bμ(P)=λ22Bμ(P),(56)i\mathcal{M}_{s}^{(1)}(P)=\frac{(-i\lambda)^{2}i^{2}}{2}\mathcal{B}_{\mu}(P)=\frac{\lambda^{2}}{2}\mathcal{B}_{\mu}(P), \tag*{(56)}

where 1/21/2 is the diagram’s symmetry factor. Interpreting this expression as a physical scattering amplitude additionally requires specifying the interacting Krein-space rules and the physical-state prescription.

For the massive case, the same position-space prescription gives

F[Gμ,m2](P)=Bμ(P)+2F[KμBm](P)+F[Bm2](P).(57)\mathcal{F}[G_{\mu,m}^{2}](P)=\mathcal{B}_{\mu}(P)+2\mathcal{F}[K_{\mu}B_{m}](P)+\mathcal{F}[B_{m}^{2}](P). \tag*{(57)}

All three coordinate-space products are bounded and define tempered distributions. Only the first term has been reduced here to the displayed Meijer GG-function. Moreover, multiplying averaged propagators is a specified effective prescription: in general,

⟨G(y)2⟩≠⟨G(y)⟩2(58)\langle G(y)^{2}\rangle\ne\langle G(y)\rangle^{2} \tag*{(58)}

when the same fluctuating metric enters both factors.

Conclusion

The Gaussian light-cone model admits an exact Lorentzian Fourier transform. Its Meijer GG-function representation must include the factor sgn⁡(p2)\operatorname{sgn}(p^{2}) (the timelike-spacelike sign) and a specified extension at the null cone. With unit-normalized light-cone smearing, the massive result is the principal-value massive pole plus a continuous correction; the massless result retains its principal-value massless pole. For fixed μ>0\mu>0, the Gaussian contribution decays faster than any inverse power of ∣p2∣\lvert p^{2}\rvert as ∣p2∣→∞\lvert p^{2}\rvert\to\infty, while the massive tail has the algebraic behavior m2/(p2)2m^{2}/(p^{2})^{2}. At fixed positive smearing width, the bounded coordinate-space kernels allow ordinary products for the nonderivative examples considered here. The exact bubble illustrates both the usefulness and the limits of that statement: its Fourier transform is well defined as a tempered distribution, although it retains a principal-value singularity at P2=0P^2=0. To establish ultraviolet and infrared finiteness and determine whether renormalization conditions are required, further work should specify the metric correlations, the averaging of products, and the interacting physical-state prescription before extending the calculation to general amplitudes.

Acknowledgements

The author acknowledges ChatGPT for its assistance in discussions of the Meijer GG-function.

Alternative Gaussian normalizations

For μ>0\mu> 0, we consider the alternative Gaussian term

Kμalt(y)=−18ππ2μ2e−y2/(4μ2)=CKμ(y),C=π2.(59)K^{\mathrm{alt}}_{\mu}(y)=-\frac{1}{8\pi}\sqrt{\frac{\pi}{2\mu^2}}e^{-y^2/(4\mu^2)}=CK_{\mu}(y),\qquad C=\frac{\pi}{\sqrt{2}}. \tag*{(59)}

Keeping this coefficient is mathematically possible, but its zero-width limit is −Cδ(y2)/(4π)-C\delta(y^2)/(4\pi), rather than the light-cone term in (5). Its exact transform is

K~μalt(p)=sgn⁡(q)π3/282G0,32,0(μ2q264 | −12,12,0),q≠0.(60)\widetilde{K}^{\mathrm{alt}}_{\mu}(p)=\operatorname{sgn}(q)\frac{\pi^{3/2}}{8\sqrt{2}}G_{0,3}^{2,0}\left(\frac{\mu^2q^2}{64}\,\middle|\,-\frac{1}{2},\frac{1}{2},0\right),\qquad q\ne0. \tag*{(60)}

If the Bessel tail is left unchanged, the full transform with the alternative normalization is

G~μ,malt(p)=PV⁡1p2−m2+(C−1)PV⁡1p2+CRμ(p2).(61)\widetilde{G}^{\mathrm{alt}}_{\mu,m}(p)=\operatorname{PV}\frac{1}{p^2-m^2}+(C-1)\operatorname{PV}\frac{1}{p^2}+CR_{\mu}(p^2). \tag*{(61)}

For m>0m > 0, the massless pole therefore cancels only when C=1C=1. Equation (9) adopts unit normalization to recover the stated unsmeared propagator. This changes the model normalization; it is not an algebraic rearrangement of the alternative coefficient.

The coefficient in Eq. (2.5) of Ref. [3] must be compared using the variance appearing there. Writing v=⟨σ12⟩v=\langle\sigma_1^2\rangle and μ2=2v\mu^2=2v, its light-cone term is

−18ππ2ve−σ02/(2v)=πKμ(y).(62)-\frac{1}{8\pi}\sqrt{\frac{\pi}{2v}}e^{-\sigma_0^2/(2v)}=\pi K_{\mu}(y). \tag*{(62)}

Thus that printed coefficient differs by a factor π\pi from the unit-normalized Gaussian average adopted here. The factor C=π/2C=\pi/\sqrt{2} above instead concerns the alternative expression written directly in terms of μ\mu with exponent −y2/(4μ2)-y^2/(4\mu^2). For the coefficient printed in Ref. [3], the preceding conversion formulas apply with C=πC=\pi after matching the widths.

Notation and computational checks

The Meijer GG-function H(z)H(z) is represented in grouped computational notation as

H(z)=G0,32,0(z | 0,1,12)≡MeijerG⁡(((),()),((0,1),(12)),z),z>0.(63)H(z)=G_{0,3}^{2,0}\left(z\,\middle|\,0,1,\frac{1}{2}\right)\equiv\operatorname{MeijerG}(((),()),((0,1),(\tfrac{1}{2})),z),\qquad z>0. \tag*{(63)}

There are no upper parameters. The lower parameters are divided into two groups: 00 and 11 contribute the numerator factors Γ(u)Γ(u+1)\Gamma(u)\Gamma(u+1), whereas 12\frac{1}{2} contributes the denominator factor Γ(12−u)\Gamma(\frac{1}{2}-u). Explicitly,

H(z)=12πi∫cH−i∞cH+i∞Γ(u)Γ(u+1)Γ(12−u)z−u du,cH>0,z>0.(64)H(z)=\frac{1}{2\pi i}\int_{c_H-i\infty}^{c_H+i\infty}\frac{\Gamma(u)\Gamma(u+1)}{\Gamma(\frac{1}{2}-u)}z^{-u}\,\mathrm{d}u,\qquad c_H>0,\quad z>0. \tag*{(64)}

Here cHc_H is a positive real constant specifying the vertical contour Re⁡u=cH\operatorname{Re} u = c_H in the complex uu-plane. The contour lies to the right of all poles of Γ(u)\Gamma(u) and Γ(u+1)\Gamma(u + 1).

For a numerical check of (34), one may compare the Meijer GG-function expression for I(a)I(a) with the convergent cosine integral (30), rather than truncating the nonconvergent, unregularized sine integral. Near q=0q = 0, the expansion (43) can be used to avoid loss of numerical precision caused by subtractive cancellation.

References

  1. [1]M. V. Takook, Negative norm states in de Sitter space and QFT without renormalization procedure, Int. J. Mod. Phys. E 11 (2002), 509–518. doi:10.1142/S0218301302001071; arXiv:gr-qc/0006019.DOI
  2. [2]L. H. Ford, Gravitons and light cone fluctuations, Phys. Rev. D 51 (1995), 1692–1700. doi:10.1103/PhysRevD.51.1692.DOI
  3. [3]M. V. Takook, “Krein” regularization method, Mod. Phys. Lett. A 37 (2022), 2250059. doi:10.1142/S0217732322500596; arXiv:2112.05390.DOI
  4. [4]NIST Digital Library of Mathematical Functions, Chapter 16, Sec. 16.17, Definition of the Meijer G-function, https://dlmf.nist.gov/16.17.

Paper details

vlnc:W_y2ej9j23
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

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

  • Physics and Astronomy

ChatGPT

View statement
Looking for related papers…
Contents