Vacuum polarization at one loop and the Uehling potential

Intuition Beginner

When you set up a positive charge in empty space and measure its electric field with a test charge, you find Coulomb's law: a $1/r$ potential. Quantum electrodynamics says this picture is incomplete. The vacuum is not really empty; it is a roiling sea of virtual electron-positron pairs that flicker into existence for time intervals too brief to detect directly. A positive charge polarises this sea — pulling virtual electrons slightly closer and pushing virtual positrons slightly farther — exactly the way a positive charge polarises a dielectric.

The net effect is that the bare charge in the Lagrangian is partly screened at long distances. What you measure from far away is a smaller effective charge than the bare one in the Lagrangian; what you would measure if you could probe at distances shorter than the electron Compton wavelength $\hslash /m_{e}c\approx 386$ fm is a larger effective charge, because at those distances you have penetrated inside the polarisation cloud.

This is vacuum polarisation, and the leading effect was computed by Edwin Uehling in 1935 — three years before Foley and Kusch measured the anomalous magnetic moment, twelve years before Bethe estimated the Lamb shift, thirteen years before the modern renormalisation programme. Uehling's calculation gave the explicit modification of the Coulomb potential to first order in $\alpha$. The Uehling correction is small in normal hydrogen (about 1.6% of the Lamb shift) but completely dominant in muonic hydrogen, where the muon orbits ~200 times closer to the proton than the electron does and consequently spends most of its time inside the electron-positron polarisation cloud.

The same calculation also produces the running coupling of quantum electrodynamics: $\alpha$ is not really a constant but a function of energy scale, slowly growing from $1/137$ at low energy to $1/128$ at the $Z$-boson mass. This running and the bound-state energy shifts are two faces of one object — the one-loop photon self-energy — and the structural fact that gauge invariance forces this self-energy to be purely transverse is what makes the whole renormalisation programme work.

Visual Beginner

ONE-LOOP QED VACUUM POLARISATION AND THE UEHLING POTENTIAL
===========================================================

                     (closed electron loop)
                     .--------------------.
                    /                      \
                   /                        \
                  /  internal e- carries k   \
                 /   internal e+ carries     \
   gamma (q)    |    k + q                    |   gamma (q)
   ~~~~~~~~~~--+                              +--~~~~~~~~~~
                |                              |
                 \                            /
                  \                          /
                   \                        /
                    \                      /
                     '--------------------'

   i Pi^mu-nu (q)  =  -(-ie)^2  Int  d^4 k / (2 pi)^4

                       tr [ gamma^mu (i (slash k + m) / (k^2 - m^2))
                             gamma^nu (i (slash(k+q) + m) / ((k+q)^2 - m^2)) ]

   WARD IDENTITY (gauge invariance):    q_mu Pi^mu-nu(q) = 0
       => Pi^mu-nu(q) = (q^2 g^mu-nu - q^mu q^nu) Pi(q^2)

   ON-SHELL RENORMALISATION:    Pi^R (q^2 = 0)  =  0
       (photon stays massless; renormalised charge = Coulomb-law charge)

   DRESSED PHOTON PROPAGATOR:    - i g^mu-nu / [ q^2 (1 - Pi^R(q^2)) ]

   UEHLING POTENTIAL (Uehling 1935):
   ==================================
       V_U(r)  =  - (Z alpha / r) [ 1 + (2 alpha / 3 pi) chi(2 m_e r) ]

   where    chi(x) = Int_1^infty d xi  e^{-x xi} (1 + 1/(2 xi^2)) sqrt(xi^2 - 1) / xi^2

   Short-r limit (r << 1/m_e):
       V_U(r) ~ -(Z alpha / r) [1 + (2 alpha / 3 pi) (log(1/(m_e r)) - gamma_E - 5/6)]
       (leading-log running of alpha)

   Long-r limit (r >> 1/m_e):
       V_U(r) ~ -(Z alpha / r) [1 + small e^{-2 m_e r} tail]
       (Coulomb law restored)

   HISTORY AT A GLANCE
   ====================
                                  year   contribution
                                  -----  --------------------------
   Dirac hole theory              1930   virtual e+e- pairs predicted
   Heisenberg-Euler-Weisskopf     1934   effective Lagrangian for QED vacuum
   Uehling                        1935   vacuum-polarisation potential
   Serber                         1935   independent (same year)
   Schwinger covariant            1949   renormalised computation
   Wichmann-Kroll                 1954   higher-loop bound-state corrections
   Gell-Mann-Low                  1954   running coupling, beta function
   Pohl muonic-H proton radius    2010   Uehling-dominated spectroscopy

Object	Symbol	Role
One-loop photon self-energy	${\Pi }^{\mu \nu }(q)$	sum of 1PI two-point diagrams at $O(\alpha )$
Scalar self-energy	$\Pi (q^2)$	transverse coefficient: ${\Pi }^{\mu \nu }=(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })\Pi (q^2)$
Renormalised self-energy	${\Pi }^R(q^2)$	$\Pi -\Pi (0)$; vanishes at $q^2=0$ by OS condition
Photon-field counterterm	${\delta }_3=Z_3-1$	absorbs UV pole into bare-charge rescaling
Uehling potential	$V_U(r)$	Fourier transform of dressed Coulomb interaction
Running coupling	$\alpha ({\mu }^2)$	$\alpha /[1-{\Pi }^R({\mu }^2)]$
Beta function (one loop)	$\beta (\alpha )=2{\alpha }^2/(3\pi )$	governs the energy dependence of $\alpha$
Compton wavelength	${\lambda }_C=1/m_e$	natural length scale of the Uehling screening

Worked example Beginner

Problem. Estimate the Uehling-potential contribution to the hydrogen $2S_{1/2}-2P_{1/2}$ Lamb-shift splitting, using the short-distance approximation $V_U(r)-V_C(r)\approx -(Z{\alpha }^2/15\pi )(4\pi /m_e^2){\delta }^3(\mathbf{r})$ for the leading $\delta$-function piece of the Uehling correction. Show numerically that the answer is roughly $-27$ MHz, partially cancelling against the much larger positive Lamb shift contributions from the electron self-energy and vertex correction.

Solution.

Step 1. Recognise that the ${\delta }^3(\mathbf{r})$ structure couples only to states with non-zero wavefunction at the origin. The hydrogen $2S_{1/2}$ state has $|{\psi }_{2S}(0){|}^2=1/(8\pi a_0^3)$ with Bohr radius $a_0=1/(\alpha m_e)$. The $2P_{1/2}$ state vanishes at the origin, so the Uehling shift acts entirely on $2S_{1/2}$.

Step 2. Compute the matrix element using first-order perturbation theory:

$$\Delta E_U(2S)=⟨2S|V_U-V_C|2S⟩=-\frac{Z{\alpha }^2}{15\pi }\cdot \frac{4\pi }{m_e^2}\cdot |{\psi }_{2S}(0){|}^2=-\frac{Z{\alpha }^2}{15\pi }\cdot \frac{4\pi }{m_e^2}\cdot \frac{1}{8\pi a_0^3}.$$

Step 3. Substitute $a_0=1/(\alpha m_e)$, giving $a_0^3=1/({\alpha }^3{m}_e^3)$:

$$\Delta E_U(2S)=-\frac{Z{\alpha }^2}{15\pi }\cdot \frac{4\pi }{m_e^2}\cdot \frac{{\alpha }^3{m}_e^3}{8\pi }=-\frac{Z{\alpha }^5{m}_e}{30\pi }.$$

Step 4. Plug in $Z=1$, $\alpha =1/137.036$, and $m_e{c}^2=0.511$ MeV converted to a frequency via $m_e{c}^2/h=1.236\times 10^{20}$ Hz:

$$\Delta E_U(2S)=-\frac{(1/137.036{)}^5\cdot (1.236\times 10^{20}\text{ Hz})}{30\pi }=-\frac{2.07\times 10^{-11}\cdot 1.236\times 10^{20}}{94.25}\text{ Hz}=-2.72\times 10^7\text{ Hz}.$$

So $\Delta E_U(2S)\approx -27.2$ MHz, matching the textbook Uehling contribution to the Lamb shift within the precision of the $\delta$-function approximation.

What this tells us. The Uehling potential pulls the $2S_{1/2}$ level down by about 27 MHz, while the vertex correction and electron self-energy push it up by ~1085 MHz. The net Lamb splitting $\Delta E(2S_{1/2})-\Delta E(2P_{1/2})$ is +1058 MHz, the value Lamb and Retherford measured in 1947. Vacuum polarisation is the smallest of the three one-loop contributions in normal hydrogen but the largest in muonic hydrogen, where the muon's much smaller Bohr radius means its wavefunction sits squarely inside the electron-positron polarisation cloud and the Uehling potential becomes the dominant quantum-electrodynamic effect.

Check your understanding Beginner

Formal definition Intermediate+

Setup. Consider QED with the Lagrangian and Feynman rules established in 12.12.01. The full photon propagator $D^{\mu \nu }(q)$ is the resummation of the geometric series of one-particle-irreducible (1PI) two-point insertions, schematically

$$i{D}^{\mu \nu }(q)=i{D}_0^{\mu \nu }(q)+i{D}_0^{\mu \alpha }(q)\cdot i{\Pi }_{\alpha \beta }(q)\cdot i{D}_0^{\beta \nu }(q)+\cdots ,$$

where $D_0^{\mu \nu }(q)=-g^{\mu \nu }/(q^2+iϵ)$ is the bare Feynman-gauge propagator and ${\Pi }^{\mu \nu }(q)$ is the sum of all 1PI photon self-energy diagrams.

Ward-Takahashi transversality. Current conservation ${\partial }_{\mu }{J}^{\mu }=0$ at the operator level forces the photon self-energy to be transverse:

$$\boxed{q_{\mu }{\Pi }^{\mu \nu }(q)=0\text{for every }q.}$$

Lorentz covariance plus this transversality constraint reduce the most general two-tensor decomposition of ${\Pi }^{\mu \nu }$ to a single scalar function:

$$\boxed{{\Pi }^{\mu \nu }(q)=(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })\Pi (q^2).}$$

The longitudinal piece is absent. The transverse projector $(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })$ vanishes at $q^2=0$, so any candidate photon mass $m_{\gamma }^2\propto \Pi (0)$ is automatically forbidden by gauge symmetry rather than by a fine-tuning of counterterms. This is the QED analogue of the massless photon protection theorem that survives at all orders.

The one-loop diagram. The leading contribution to ${\Pi }^{\mu \nu }$ in perturbation theory is the single 1PI diagram in which the external photon converts into a virtual electron-positron pair (closed fermion loop) that propagates with loop momentum $k$ and reannihilates. Citing Berestetskii-Lifshitz-Pitaevskii §113-114 and Peskin-Schroeder §7.5 ^{[Berestetskii-Lifshitz-Pitaevskii §113-114; Peskin-Schroeder §7.5]}, in Feynman gauge the QED Feynman rules give

$$i{\Pi }^{\mu \nu }(q)=-(-ie{)}^2\int \frac{d^{d}k}{(2\pi {)}^d}\mathrm{tr}\left[{\gamma }^{\mu }\frac{i(\text{slashed}k+m)}{k^2-m^2+iϵ}{\gamma }^{\nu }\frac{i(\text{slashed}k+\text{slashed}q+m)}{(k+q{)}^2-m^2+iϵ}\right].$$

The leading overall minus sign is the universal closed-fermion-loop factor that comes from the anticommuting-Grassmann nature of fermion fields. The trace is over the four-dimensional Dirac-spinor index space. The dimension $d=4-2\epsilon$ is the dimensional-regularisation parameter ('t Hooft-Veltman 1972) ^{['t Hooft-Veltman 1972]}.

On-shell renormalisation. The renormalised photon self-energy is

$${\Pi }^R(q^2)=\Pi (q^2)-\Pi (0),$$

with the counterterm ${\delta }_3=-\Pi (0)$ absorbing the $1/\epsilon$ pole into the bare-field rescaling $A^{\mu }=Z_3^{-1/2}{A}_0^{\mu }$. The OS condition

$$\boxed{{\Pi }^R(q^2=0)=0}$$

ensures that the dressed propagator pole stays at $q^2=0$ (the photon remains massless) and that the residue at the pole is unity, so the renormalised electric charge $e_R$ is what Coulomb's law measures at long distance. The dressed propagator in Feynman gauge becomes

$$D^{\mu \nu }(q)=\frac{-i{g}^{\mu \nu }}{q^2[1-{\Pi }^R(q^2)]}+(\text{longitudinal gauge artefact, drops out of observables}).$$

The factor $[1-{\Pi }^R(q^2){]}^{-1}$ in the denominator is the running coupling: the effective fine-structure constant felt by a process at momentum scale $|q|$ is

$$\boxed{\alpha (q^2)=\frac{\alpha }{1-{\Pi }^R(q^2)}.}$$

Sign and metric conventions

Throughout this unit the metric is the mostly-plus particle-physics convention ${\eta }_{\mu \nu }=\mathrm{diag}(+1,-1,-1,-1)$ matching 12.16.01 and 12.16.02, the gamma matrices satisfy $\{{\gamma }^{\mu },{\gamma }^{\nu }\}=2{\eta }^{\mu \nu }1$ with $\mathrm{tr}({\gamma }^{\mu }{\gamma }^{\nu })=4{\eta }^{\mu \nu }$ and $\mathrm{tr}({\gamma }^{\mu }{\gamma }^{\nu }{\gamma }^{\rho }{\gamma }^{\sigma })=4({\eta }^{\mu \nu }{\eta }^{\rho \sigma }-{\eta }^{\mu \rho }{\eta }^{\nu \sigma }+{\eta }^{\mu \sigma }{\eta }^{\nu \rho })$, the QED vertex factor is $-ie{\gamma }^{\mu }$ with $e>0$ the proton charge, and the photon propagator in Feynman gauge is $-i{g}_{\mu \nu }/(q^2+iϵ)$. Berestetskii-Lifshitz-Pitaevskii use the opposite sign convention for $e$ and a Gaussian-Heaviside unit system; readers consulting BLP should flip $e\to -e$ in the vertex factor and convert $e^2=4\pi \alpha$ to $e^2=\alpha$ in Gaussian units (the numerical value of $\alpha$ is unchanged).

Key derivation Intermediate+

Theorem (Uehling-Serber-Schwinger, 1935-1949). The one-loop QED photon self-energy in dimensional regularisation with $d=4-2\epsilon$ has the transverse form ${\Pi }^{\mu \nu }(q)=(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })\Pi (q^2)$ with scalar coefficient

$$\Pi (q^2)=-\frac{\alpha }{3\pi }\left[\frac{1}{\epsilon }-{\gamma }_E+\log \frac{4\pi {\mu }^2}{m^2}\right]-\frac{2\alpha }{\pi }{\int }_0^{1}dxx(1-x)\log \left[1-x(1-x)\frac{q^2}{m^2}\right]+\mathcal{O}({\alpha }^2).$$

The OS renormalised self-energy is

$${\Pi }^R(q^2)=-\frac{2\alpha }{\pi }{\int }_0^{1}dxx(1-x)\log \left[1-x(1-x)\frac{q^2}{m^2}\right]+\mathcal{O}({\alpha }^2),$$

which vanishes at $q^2=0$ by inspection. The Fourier transform of the dressed Coulomb interaction sourced by a point charge $Ze$ is the Uehling potential

$$V_U(r)=-\frac{Z\alpha }{r}\left[1+\frac{2\alpha }{3\pi }\chi (2m_{e}r)\right]+\mathcal{O}({\alpha }^2),$$

with kernel $\chi (x)={\int }_1^{\infty }d\xi e^{-x\xi }(1+1/(2{\xi }^2))\sqrt{{\xi }^2-1}/{\xi }^2$.

Proof. Start from the dimensionally-regularised loop integral

$$i{\Pi }^{\mu \nu }(q)=-e^2\int \frac{d^{d}k}{(2\pi {)}^d}\frac{\mathrm{tr}[{\gamma }^{\mu }(\text{slashed}k+m){\gamma }^{\nu }(\text{slashed}k+\text{slashed}q+m)]}{[k^2-m^2+iϵ][(k+q{)}^2-m^2+iϵ]}.$$

The trace identity $\mathrm{tr}[{\gamma }^{\mu }\text{slashed}a{\gamma }^{\nu }\text{slashed}b]=4(a^{\mu }{b}^{\nu }+a^{\nu }{b}^{\mu }-g^{\mu \nu }a\cdot b)$ together with $\mathrm{tr}[{\gamma }^{\mu }{\gamma }^{\nu }]=4g^{\mu \nu }$ evaluates the numerator. With $a=k$ and $b=k+q$, after cross-terms in $m$ (which contribute $4m^2{g}^{\mu \nu }$ via $\mathrm{tr}[{\gamma }^{\mu }{\gamma }^{\nu }]m^2$),

$$N^{\mu \nu }=\mathrm{tr}[{\gamma }^{\mu }(\text{slashed}k+m){\gamma }^{\nu }(\text{slashed}k+\text{slashed}q+m)]=4\left[k^{\mu }(k+q{)}^{\nu }+k^{\nu }(k+q{)}^{\mu }-g^{\mu \nu }(k\cdot (k+q)-m^2)\right].$$

Combine the two denominators with the Feynman-parameter identity $1/(AB)={\int }_0^{1}dx/[xA+(1-x)B{]}^2$ — taking $A=(k+q{)}^2-m^2$, $B=k^2-m^2$ — and shift the loop momentum to $\ell =k+xq$. The combined denominator becomes $[{\ell }^2-\Delta +iϵ{]}^2$ with

$$\Delta =m^2-x(1-x)q^2.$$

The numerator, expressed in $\ell$, picks up cross-terms ${\ell }^{\mu }(\ell +(1-x)q{)}^{\nu }+{\ell }^{\nu }(\ell +(1-x)q{)}^{\mu }-g^{\mu \nu }(\ell \cdot (\ell +(1-x)q)-x(1-x)q^2-m^2)$, plus linear-in-$\ell$ terms that integrate to zero by parity, plus terms with no $\ell$. The ${\ell }^{\mu }{\ell }^{\nu }$ contribution is symmetrised to ${\ell }^2{g}^{\mu \nu }/d$ inside the rotationally-invariant Euclidean integral. After this algebra, dropping the parity-odd zero contributions, the integrand contains the structures

• $4[{\ell }^{\mu }{\ell }^{\nu }+(\text{loops with one }\ell )]$ symmetrised to $(8{\ell }^2/d)g^{\mu \nu }$,
• $-4g^{\mu \nu }{\ell }^2$ from the $-g^{\mu \nu }\ell \cdot \ell$ piece,
• $-2(2x-1{)}^2(q^{\mu }{q}^{\nu }-g^{\mu \nu }{q}^2/2)$ from the $-g^{\mu \nu }x(1-x)q^2$ recombination,
• $-2x(1-x)(q^{\mu }{q}^{\nu }-g^{\mu \nu }{q}^2)$ — a transverse piece.

The cleanest organising principle is that after the Wick rotation and the standard $d$-dimensional master integrals

$$\int \frac{d^d{\ell }_E}{(2\pi {)}^d}\frac{1}{[{\ell }_E^2+\Delta {]}^2}=\frac{\Gamma (\epsilon )}{(4\pi {)}^{d/2}}{\Delta }^{-\epsilon },\int \frac{d^d{\ell }_E}{(2\pi {)}^d}\frac{{\ell }_E^2}{[{\ell }_E^2+\Delta {]}^2}=\frac{d}{2}\frac{\Gamma (\epsilon -1)}{(4\pi {)}^{d/2}}{\Delta }^{1-\epsilon },$$

all of the non-transverse $g^{\mu \nu }$ pieces from the ${\ell }^2$ contractions cancel against the $-g^{\mu \nu }{m}^2$ piece — a tedious but mechanical cancellation that ultimately enforces the Ward identity term-by-term. The surviving transverse combination is

$${\Pi }^{\mu \nu }(q)=(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })\Pi (q^2),\Pi (q^2)=-\frac{8e^2}{(4\pi {)}^{d/2}}\Gamma (\epsilon ){\int }_0^{1}dxx(1-x){\Delta }^{-\epsilon },$$

with $\Delta =m^2-x(1-x)q^2$. The Wick rotation supplied an overall factor of $i$ that cancelled the $i$ on the left-hand side of $i{\Pi }^{\mu \nu }$.

Substitute $d=4-2\epsilon$ and $e^2=4\pi \alpha$, and use $1/(4\pi {)}^{d/2}=(1/(4\pi {)}^2)(4\pi {)}^{\epsilon }$:

$$\Pi (q^2)=-\frac{2\alpha }{\pi }\Gamma (\epsilon )(4\pi {)}^{\epsilon }{\mu }^{-2\epsilon }{\int }_0^{1}dxx(1-x){\Delta }^{-\epsilon },$$

where the dimensional-regularisation mass scale $\mu$ was reinstated by the standard replacement $e\to e{\mu }^{\epsilon }$ to keep $e$ dimensionless. Expand $\Gamma (\epsilon )=1/\epsilon -{\gamma }_E+\mathcal{O}(\epsilon )$, $(4\pi {)}^{\epsilon }=1+\epsilon \log (4\pi )+\mathcal{O}({\epsilon }^2)$, and ${\Delta }^{-\epsilon }=1-\epsilon \log (\Delta /{\mu }^2)+\mathcal{O}({\epsilon }^2)$:

$$\Pi (q^2)=-\frac{2\alpha }{\pi }\left[\frac{1}{\epsilon }-{\gamma }_E+\log (4\pi )\right]{\int }_0^{1}dxx(1-x)+\frac{2\alpha }{\pi }{\int }_0^{1}dxx(1-x)\log \frac{\Delta }{{\mu }^2}+\mathcal{O}(\epsilon ).$$

The first integral is ${\int }_0^{1}dxx(1-x)=1/6$, so the divergent prefactor becomes

$$\Pi (q^2)=-\frac{\alpha }{3\pi }\left[\frac{1}{\epsilon }-{\gamma }_E+\log (4\pi )-\log \frac{m^2}{{\mu }^2}\right]-\frac{2\alpha }{\pi }{\int }_0^{1}dxx(1-x)\log \left[1-x(1-x)\frac{q^2}{m^2}\right]+\mathcal{O}({\alpha }^2),$$

after replacing $\log (\Delta /{\mu }^2)=\log (m^2/{\mu }^2)+\log (1-x(1-x)q^2/m^2)$ and absorbing $\log (m^2/{\mu }^2)$ into the divergent bracket. This is the announced structure.

The OS counterterm ${\delta }_3=-\Pi (0)$ subtracts the $1/\epsilon$ pole and the accompanying constants, leaving the renormalised self-energy

$${\Pi }^R(q^2)=\Pi (q^2)-\Pi (0)=-\frac{2\alpha }{\pi }{\int }_0^{1}dxx(1-x)\log \left[1-x(1-x)\frac{q^2}{m^2}\right]+\mathcal{O}({\alpha }^2),$$

which vanishes at $q^2=0$ by inspection of the logarithm.

The Uehling potential. A static point charge $Ze$ at the origin sources a classical four-current $J^{\mu }(x)=Ze{\delta }^{\mu 0}{\delta }^3(\mathbf{x})$ whose Fourier transform is $J^{\mu }(q)=2\pi Ze{\delta }^{\mu 0}\delta (q^0)$. The dressed Coulomb potential is the dressed photon propagator contracted with this current. In the static limit $q^0=0$, the four-momentum is spacelike with $q^2=-|\mathbf{q}{|}^2$, and the dressed scalar potential is

$${\tilde{V}}_U(\mathbf{q})=\frac{-Ze}{|\mathbf{q}{|}^2[1-{\Pi }^R(-|\mathbf{q}{|}^2)]}\approx \frac{-Ze}{|\mathbf{q}{|}^2}[1+{\Pi }^R(-|\mathbf{q}{|}^2)]+\mathcal{O}({\alpha }^2),$$

valid to first order in $\alpha$. The static-position potential is the inverse Fourier transform $V_U(\mathbf{r})=\int d^{3}q/(2\pi {)}^3{\tilde{V}}_U(\mathbf{q})e^{i\mathbf{q}\cdot \mathbf{r}}$.

The zeroth-order piece reproduces the classical Coulomb potential $V_C(r)=-Z\alpha /r$. The first-order correction is

$$\delta V_U(\mathbf{r})=-Ze\int \frac{d^{3}q}{(2\pi {)}^3}\frac{e^{i\mathbf{q}\cdot \mathbf{r}}}{|\mathbf{q}{|}^2}{\Pi }^R(-|\mathbf{q}{|}^2).$$

Substitute the integrand expression for ${\Pi }^R$ and switch the order of integration. The ${\int }_0^{1}dxx(1-x)$ Feynman-parameter measure is unchanged; the inner $\mathbf{q}$-integral becomes the three-dimensional Fourier transform of $-\log [1+x(1-x)|\mathbf{q}{|}^2/m^2]/|\mathbf{q}{|}^2$. Use the standard identity (Peskin-Schroeder §7.5 ^{[Peskin-Schroeder §7.5]})

$$\int \frac{d^{3}q}{(2\pi {)}^3}\frac{e^{i\mathbf{q}\cdot \mathbf{r}}}{|\mathbf{q}{|}^2}\log \left[1+\frac{|\mathbf{q}{|}^2}{4m^2{t}^2}\right]=\frac{1}{4\pi r}\left(1-e^{-2mr/t}\right)\cdot \text{ (multiplied by }2t\text{)},$$

valid for $t\in (0,1)$, after substituting the Feynman parameter $x=(1-t)/2$ (so $x(1-x)=(1-t^2)/4$). Re-collecting the $x$-integral as a $t$-integral over $t\in (0,1)$ with measure $dx=-dt/2$ and $4x(1-x)=1-t^2$, one obtains

$$\delta V_U(r)=-\frac{Z\alpha }{r}\cdot \frac{2\alpha }{3\pi }\chi (2m_{e}r),\chi (x)={\int }_1^{\infty }d\xi e^{-x\xi }\left(1+\frac{1}{2{\xi }^2}\right)\frac{\sqrt{{\xi }^2-1}}{{\xi }^2},$$

after substituting $\xi =1/t$. The full Uehling potential is then

$$V_U(r)=V_C(r)+\delta V_U(r)=-\frac{Z\alpha }{r}\left[1+\frac{2\alpha }{3\pi }\chi (2m_{e}r)\right]+\mathcal{O}({\alpha }^2).\square$$

Counterexamples to common slips

• The leading factor $-2/3$ in $\Pi (q^2)=-(\alpha /3\pi )[1/\epsilon +\dots ]$ is one-loop QED specific. For a Dirac fermion in the loop the coefficient is $-2/3$; for a complex scalar it is $-1/6$; for a Majorana fermion it is $-1/3$; for an $SU(N)$-coloured Dirac fermion it picks up an extra factor of $N$. The beta function $\beta (\alpha )=2{\alpha }^2/(3\pi )$ is the specific one-Dirac-flavour result; the Standard Model running involves species-dependent thresholds at every charged-particle mass.

• The Uehling potential's leading short-distance behaviour is logarithmic, not power-law. Naive dimensional analysis suggests a $1/r^2$ tail; the actual result is a $1/r$ tail multiplied by $\log (1/m_{e}r)$. The $\log$ is the leading-log running of the coupling at short distance, and it is what makes the Uehling correction grow without bound as $r\to 0$ — formally a Landau-pole signature, in practice cut off by higher-loop physics or by the deeper Standard Model.

• The transversality $q_{\mu }{\Pi }^{\mu \nu }=0$ holds only after counterterm subtraction if one uses a non-gauge-invariant regulator (momentum cutoff $\Lambda$). With dimensional regularisation it holds term-by-term in the loop integral. This is the practical reason dim reg is the modern default: it preserves gauge invariance manifestly at every intermediate step, removing the need to verify cancellations by hand.

• The renormalised photon self-energy ${\Pi }^R(q^2)$ does not vanish for $q^2\ne 0$; it is the running coupling. The vanishing condition ${\Pi }^R(0)=0$ is what makes $\alpha =\alpha (0)$ the Thomson-limit coupling and what protects the photon mass.

• The Uehling potential is not the Yukawa potential. Yukawa screening produces $e^{-mr}/r$ — pure exponential decay starting at $r=0$. Uehling screening produces $e^{-2m_{e}r}/r$ as a tail added to an undamped Coulomb $1/r$ piece, so the long-distance behaviour is Coulomb-like and the short-distance behaviour is enhanced rather than suppressed.

Exercises Intermediate+

Advanced results Master

Vacuum polarisation at one loop is the entry point to a network of higher-order calculations and modern precision tests that connect the Uehling-Serber-Schwinger result of 1935-1949 to current frontier measurements in muonic-atom spectroscopy, electroweak precision physics, and the muon $g-2$ programme.

Wichmann-Kroll higher-order bound-state corrections. Wichmann and Kroll (Phys. Rev. 96, 232, 1954) ^{[Wichmann-Kroll 1954]} computed the next-order vacuum-polarisation contribution to bound-state energies in strong external Coulomb fields: diagrams with three or more photon attachments to the electron loop, in which the loop interacts non-perturbatively with the binding Coulomb field of a nucleus. For low-$Z$ atoms the Wichmann-Kroll contribution is suppressed by $(Z\alpha {)}^2$ relative to the Uehling term and contributes at the few-kHz level in hydrogen. For high-$Z$ hydrogenic ions (uranium U${}^{91+}$ with $Z\alpha \approx 0.67$), the Wichmann-Kroll contribution is comparable in magnitude to the Uehling contribution and the perturbative expansion in $Z\alpha$ ceases to converge — bound-state QED in heavy ions must be done non-perturbatively in $Z\alpha$ from the start. The Mohr-Plunien-Soff 1998 review (Phys. Rep. 293, 227) ^{[Mohr-Plunien-Soff 1998]} is the canonical reference for the full Wichmann-Kroll treatment and its extensions to two-loop bound-state QED.

Two-loop vacuum polarisation: Källén-Sabry potential. Källén and Sabry (Dan. Mat. Fys. Medd. 29, 17, 1955) computed the two-loop QED photon self-energy, obtaining a finite analytic expression for the renormalised ${\Pi }^R(q^2)$ at $O({\alpha }^2)$. The Fourier transform of the two-loop dressed Coulomb potential is the Källén-Sabry potential, which adds a few-percent correction to the leading Uehling contribution in heavy muonic atoms. The Källén-Sabry analysis was the first to confirm that the renormalisation programme extends consistently to the photon-side beyond one loop, and the analytical form of the two-loop ${\Pi }^R$ involves Clausen functions and dilogarithms of the form ${\mathrm{Li}}_2({\sin }^2\theta )$ — a foreshadowing of the multi-polylogarithm structure that pervades all of multi-loop QED.

Spectral representation and the dispersion-relation extraction of $\alpha (M_Z)$. Analyticity of ${\Pi }^R(q^2)$ in the complex $q^2$-plane, together with the unitarity-derived spectral representation, gives

$${\Pi }^R(q^2)=\frac{q^2}{\pi }{\int }_{4m_e^2}^{\infty }\frac{ds}{s(s-q^2)}\mathrm{Im}{\Pi }^R(s),$$

with the imaginary part fixed by the optical theorem to $\mathrm{Im}{\Pi }^R(s)=-{\sigma }_{e^{+}{e}^{-}\to \text{hadrons}}(s)\cdot s/(4\pi \alpha )$ for the hadronic-contribution channel. The dispersion-relation extraction of the hadronic contribution to ${\Pi }^R$ from $e^{+}{e}^{-}\to$ hadrons cross-section data is the standard way to compute the leading hadronic contribution to the muon $g-2$ (the "HVP" piece). The discrepancy between the dispersion-relation extraction and the BMW 2020 lattice-QCD calculation (Borsanyi et al., Nature 593, 51, 2021) is currently the central open question in the muon $g-2$ programme; resolving it determines whether the experimental Fermilab Muon g-2 result (Aguillard et al., Phys. Rev. Lett. 131, 161802, 2023) is in $4.2\sigma$ tension with the Standard Model or only $1.5\sigma$.

Running coupling and the Landau pole. The QED beta function $\beta (\alpha )=2{\alpha }^2/(3\pi )+\mathcal{O}({\alpha }^3)$ integrates to the running coupling $\alpha ({\mu }^2)=\alpha ({\mu }_0^2)/[1-(\alpha ({\mu }_0^2)/3\pi )\log ({\mu }^2/{\mu }_0^2)]$. The denominator vanishes — the Landau pole — at ${\mu }_L^2={\mu }_0^2\exp (3\pi /\alpha ({\mu }_0^2))=m_e^2\exp (3\pi \cdot 137.036)\sim 10^{286}{\text{GeV}}^2$. This is vastly beyond the Planck scale $M_P\approx 10^{19}$ GeV, so the Landau pole is not a physical pathology of QED but a signal that perturbation theory in $\alpha$ ceases to organise the dynamics long before the formal pole is reached. The Standard Model embedding of QED in the $SU(2{)}_L\times U(1{)}_Y$ electroweak theory and ultimately in whatever ultraviolet completion replaces it (grand unification, string theory, asymptotic safety) means the QED Landau pole is averted dynamically far below the formal value.

Vacuum polarisation in the muon $g-2$. The leading $O({\alpha }^2)$ contribution to the muon anomalous magnetic moment from vacuum polarisation is the Petermann-Sommerfield electron-loop insertion in the one-loop muon vertex: a virtual photon spans the two muon legs and the photon is dressed by an electron loop, giving an additional contribution $a_{\mu }^{\text{vp,e}}=(\alpha /\pi {)}^2\cdot (\log (m_{\mu }/m_e)/3+\text{finite})$. The logarithm $\log (m_{\mu }/m_e)\approx 5.33$ is the leading-log enhancement of the muon $g-2$ over the electron $g-2$ from the same diagram, and is the origin of the $(m_{\mu }/m_e{)}^2\approx 4\times 10^4$ relative sensitivity of the muon to new physics: a heavy virtual particle of mass $M$ in the muon vertex contributes $\sim (m_{\mu }/M{)}^2(\alpha /\pi )$, enhanced over the electron case by exactly this factor. By the five-loop calculation, vacuum-polarisation insertions of all charged Standard-Model species (leptons, quarks, $W$, $Z$) contribute and are tracked individually in the Theory Initiative white paper (Aoyama et al., Phys. Rep. 887, 1, 2020).

Schwinger pair production: vacuum polarisation in strong external fields. Schwinger 1951 (Phys. Rev. 82, 664) extended the one-loop vacuum-polarisation calculation to constant external electric fields, obtaining the imaginary part of the effective Lagrangian as a pair-production rate per unit volume $\Gamma =(eE{)}^2/(4{\pi }^3)\cdot \exp (-\pi m^2/(eE))$. The exponential suppression for $eE\ll m^2$ explains why Schwinger pair creation has never been observed: laboratory fields are $\sim 10^{10}$ V/m, the critical Schwinger field is $E_c=m_e^2/e\approx 1.3\times 10^{18}$ V/m, and the exponential factor at $E/E_c=10^{-8}$ is $\exp (-\pi \cdot 10^8)$. ELI and other ultraintense-laser programmes aim at $E/E_c\sim 10^{-4}$, where the rate is $\exp (-\pi \cdot 10^4)$ — still vanishingly small, but possibly observable with very long integration times in the most extreme focal spots. The vacuum-polarisation function ${\Pi }^R(q^2)$ in this regime is the analytic-continuation companion of the Uehling potential to the Lorentzian strong-field setting.

Modern precision tests. Muonic hydrogen Lamb-shift spectroscopy by the CREMA collaboration (Pohl et al., Nature 466, 213, 2010 ^{[Pohl 2010]}; Antognini et al., Science 339, 417, 2013) measured the 2S-2P transition to 50 ppm precision and extracted a proton charge radius $r_p=0.84087(39)$ fm, smaller than the electronic-hydrogen and elastic-scattering values of $\sim 0.879(8)$ fm at the level of $5\sigma$. The discrepancy — the "proton-radius puzzle" — was partially resolved by subsequent electronic-hydrogen measurements (Beyer et al., Science 358, 79, 2017; Bezginov et al., Science 365, 1007, 2019) that found smaller values consistent with the muonic-hydrogen extraction. The Uehling potential is the dominant QED contribution to the muonic 2S-2P splitting (205 GHz of the ~206 GHz total) and the proton-finite-size contribution that the spectroscopy extracts is the next-largest piece (10 GHz, scaling as $r_p^2$). Any unaccounted-for QED contribution to vacuum polarisation at the few-percent level would shift the extracted radius substantially.

Comparison of regularisation schemes. As with the electron self-energy of 12.16.01, the photon self-energy can be regulated by Pauli-Villars, by a four-dimensional momentum cutoff, or by dimensional regularisation. Pauli-Villars introduces fictitious heavy fermion fields with masses $M_i$ and signs $C_i$ tuned to give a UV-finite loop integral; momentum cutoff breaks gauge invariance and produces a spurious quadratically-divergent photon mass that must be hand-cancelled; dimensional regularisation preserves gauge invariance manifestly and is the modern default. The Ward identity $q_{\mu }{\Pi }^{\mu \nu }=0$ holds term-by-term in dim reg but requires explicit cancellation in cutoff schemes — a practical reason for the universal adoption of dim reg in modern QFT.

Full proof set Master

The Key derivation supplies the dimensional-regularisation evaluation of $\Pi (q^2)$, its renormalisation, and the Fourier transform to the Uehling potential. The following auxiliary results are stated with proof outlines verifiable against the cited literature.

Lemma 1 (Ward-Takahashi transversality to all orders). The photon self-energy in QED satisfies $q_{\mu }{\Pi }^{\mu \nu }(q)=0$ to all orders in perturbation theory. Proof outline. Current conservation ${\partial }_{\mu }{J}^{\mu }(x)=0$ is an operator identity in QED — it follows from the global $U(1)$ symmetry of the Lagrangian via Noether's theorem and survives quantisation. Apply ${\partial }_{\mu }^{(x)}$ to the connected time-ordered two-point function $⟨0|T{J}^{\mu }(x)J^{\nu }(y)|0⟩$. The derivative acts on the operator (giving zero by conservation) and on the time-ordering $\theta$-function (giving equal-time canonical commutators). The equal-time canonical commutators $[J^0(x),J^{\nu }(y)]$ at $x^0=y^0$ are local Schwinger terms that contribute at most to gauge-artefact pieces of ${\Pi }^{\mu \nu }$ and drop out of the physical transverse projection. In momentum space the relation reads $q_{\mu }{\Pi }^{\mu \nu }(q)=0$. The argument is non-perturbative: it survives at every order in $\alpha$ and at any loop count. (Detailed proof: Weinberg Vol. I §10.4; Peskin-Schroeder §7.4 for the operator derivation.) $\square$

Lemma 2 (UV degree of divergence by power counting). The one-loop photon self-energy in four dimensions has superficial degree of divergence $D=2$ (quadratic), but the actual divergence is logarithmic ($D=0$) by the Ward identity. Proof outline. Power-count the loop integral: $D=(\text{measure})-2\cdot (\text{fermion props})+(\text{numerator powers})=4-4+2=2$, suggesting a quadratic divergence with coefficient proportional to a candidate photon mass $m_{\gamma }^2$. However, the transversality ${\Pi }^{\mu \nu }=(q^2{g}^{\mu \nu }-q^{\mu }{q}^{\nu })\Pi (q^2)$ forces the leading $q$-independent piece to vanish identically; the scalar $\Pi (q^2)$ must extract two powers of $q$ from the integrand to match the transverse projector. This reduces the effective degree of divergence to $D=0$ — logarithmic — and the divergence appears as a $1/\epsilon$ pole in $\Pi (q^2)$ at $q^2\ne 0$. The cancellation between the naive quadratic divergence and the Ward-identity-enforced transversality is automatic in dimensional regularisation but requires explicit subtraction in cutoff schemes. (Detailed proof: Peskin-Schroeder §7.5; Weinberg Vol. I §11.2.) $\square$

Lemma 3 (Feynman-parameter combination and Wick rotation). The two-denominator combination $1/[(k+q{)}^2-m^2][k^2-m^2]$ admits the Feynman-parameter form ${\int }_0^{1}dx/[{\ell }^2-\Delta +iϵ{]}^2$ with $\ell =k+xq$ and $\Delta =m^2-x(1-x)q^2$, and Wick rotation ${\ell }^0=i{\ell }_E^0$ gives the Euclidean measure $i{d}^d{\ell }_E$. Proof outline. The Feynman-parameter identity $1/(AB)={\int }_0^{1}dx/[xA+(1-x)B{]}^2$ follows from direct computation as in Exercise 4 of 12.16.01: substitute $u=B+x(A-B)$, integrate $du/[(A-B)u^2]$ from $u=B$ to $u=A$, and recognise the result as $1/(AB)$. Substituting $A=(k+q{)}^2-m^2$, $B=k^2-m^2$ and completing the square in $k$ produces ${\ell }^2-\Delta +iϵ$ with $\Delta =m^2-x(1-x)q^2$ (positive for $q^2<4m^2$, ensuring the Wick rotation is unobstructed). The Wick rotation of the loop time component ${\ell }^0\to i{\ell }_E^0$ is licit because the Feynman $+iϵ$ prescription keeps the poles of $1/[{\ell }^2-\Delta +iϵ{]}^2$ away from the imaginary ${\ell }^0$-axis, and the integrand at large $|{\ell }^0|$ decays as $|{\ell }^0{|}^{-4}$ faster than the arcs grow. (Detailed proof: Peskin-Schroeder §A.4; Itzykson-Zuber §6-2-3.) $\square$

Lemma 4 (Fourier transform to position space). The Fourier transform of $-1/|\mathbf{q}{|}^2\cdot \log [1+|\mathbf{q}{|}^2/(4m_e^2{t}^2)]$ in three spatial dimensions, integrated against the Feynman parameter $x=(1-t)/2$, produces the kernel $\chi (2m_{e}r)/r$ of the Uehling potential. Proof outline. Use the integral representation $\log [1+z]=z{\int }_0^{1}d\lambda /(1+\lambda z)$ to rewrite the logarithm as a single integral. The inner three-dimensional Fourier transform of $1/(|\mathbf{q}{|}^2+4m_e^2{t}^2/\lambda )$ is the Yukawa kernel $e^{-(2m_{e}t/\sqrt{\lambda })r}/(4\pi r)$ via the standard identity $\int d^{3}q{e}^{i\mathbf{q}\cdot \mathbf{r}}/(|\mathbf{q}{|}^2+{\mu }^2)=2{\pi }^2{e}^{-\mu r}/r$. Reorganise the $\lambda$-integral as a $\xi =1/\sqrt{1-\lambda (1-t^2)}$ substitution and recombine with the $t$-integral via $t\to \xi$. The result is $\chi (2m_{e}r)={\int }_1^{\infty }d\xi e^{-2m_{e}r\xi }(1+1/(2{\xi }^2))\sqrt{{\xi }^2-1}/{\xi }^2$. (Detailed proof: Peskin-Schroeder §7.5; BLP §114; Itzykson-Zuber §7-1-1.) $\square$

Lemma 5 (small-$r$ asymptotic of $\chi$). For $r\ll 1/m_e$, the kernel satisfies $\chi (2m_{e}r)\to \log (1/(m_{e}r))-{\gamma }_E-5/6+\mathcal{O}(m_{e}r)$. Proof outline. Split the $\xi$-integral at $\xi =1/(m_{e}r)$. For $\xi \ll 1/(m_{e}r)$ the exponential $e^{-2m_{e}r\xi }$ is close to one and $\chi \approx {\int }_1^{1/(m_{e}r)}d\xi (1+1/(2{\xi }^2))\sqrt{{\xi }^2-1}/{\xi }^2$. The integrand at large $\xi$ behaves as $1/\xi +1/(2{\xi }^3)+\mathcal{O}(1/{\xi }^4)$, integrating to $\log (\xi )+\text{const}$. Evaluating at the upper limit $1/(m_{e}r)$ gives the dominant $\log (1/(m_{e}r))$. The constants combine to $-{\gamma }_E-5/6$ after careful tracking of the finite pieces. (Detailed proof: Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}, Eq. 14-15.) $\square$

Lemma 6 (large-$r$ asymptotic of $\chi$). For $r\gg 1/m_e$, the kernel satisfies $\chi (2m_{e}r)\to \sqrt{\pi /(8m_{e}r)}{e}^{-2m_{e}r}(1+\mathcal{O}(1/(m_{e}r)))$. Proof outline. The integrand $e^{-2m_{e}r\xi }\cdot (1+1/(2{\xi }^2))\sqrt{{\xi }^2-1}/{\xi }^2$ for large $m_{e}r$ is dominated by the saddle at $\xi =1$. Expand $\sqrt{{\xi }^2-1}=\sqrt{2(\xi -1)}(1+\mathcal{O}(\xi -1))$ near $\xi =1$, factor out $e^{-2m_{e}r}$, and evaluate the resulting Gaussian-like integral ${\int }_0^{\infty }d\eta \sqrt{2\eta }{e}^{-2m_{e}r\eta }=\sqrt{\pi /(8m_{e}r)}{e}^0$ via standard Laplace-method asymptotics. (Detailed proof: BLP §114; Greiner-Reinhardt QED, 4e, §5.5.) $\square$

Lemma 7 (Uehling shift via $\delta$-function approximation). For $r\ll a_0$ the Uehling correction $\delta V_U(r)$ is well-approximated by a Dirac $\delta$-function in coordinate space: $\delta V_U(r)\approx -(Z{\alpha }^2/15\pi )(4\pi /m_e^2){\delta }^3(\mathbf{r})$. Proof outline. The momentum-space form of the Uehling correction at low $|\mathbf{q}|\ll m_e$ is ${\Pi }^R(-|\mathbf{q}{|}^2)/|\mathbf{q}{|}^2\approx -(\alpha /15\pi )(|\mathbf{q}{|}^2/m_e^2)/|\mathbf{q}{|}^2=-\alpha /(15\pi m_e^2)$, a $\mathbf{q}$-independent constant. The Fourier transform of a constant is a $\delta$-function. The factor $1/15$ comes from expanding ${\Pi }^R$ to second order in $q^2/m^2$: ${\Pi }^R(q^2)\approx -(2\alpha /\pi ){\int }_0^{1}dxx(1-x)\cdot (-x(1-x)q^2/m^2)=(2\alpha /\pi )(q^2/m^2){\int }_0^{1}dx[x(1-x){]}^2=(\alpha /15\pi )(q^2/m^2)$ using ${\int }_0^{1}dx[x(1-x){]}^2=1/30$. The validity of the $\delta$-function approximation requires that the wavefunction be smooth on the scale $1/m_e$, which holds for electronic hydrogen (where $a_0\gg 1/m_e$) and fails for muonic hydrogen. (Detailed proof: Peskin-Schroeder §7.5; Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}.) $\square$

Connections Master

• 12.11.01 supplies the Dirac equation and the bare fermion propagator $i(\text{slashed}k+m)/(k^2-m^2+iϵ)$ that the closed loop in this unit traces; the gamma-matrix algebra is also the source of the trace identity used in the Key derivation.

• 12.12.01 supplies the Feynman rules of QED — vertex factor $-ie{\gamma }^{\mu }$, photon propagator $-i{g}_{\mu \nu }/(q^2+iϵ)$, closed-fermion-loop minus sign — that this unit composes into the photon self-energy integral.

• 12.16.01 Electron self-energy and mass renormalisation at one loop is the fermion-side companion of this unit: where this unit computes the photon's wave-function counterterm $Z_3$, the self-energy unit computes the electron's mass shift $\delta m$ and wave-function counterterm $Z_2$. The Ward identity $Z_1=Z_2$ from the vertex unit (12.16.02) makes the renormalised electric charge $e_R=e_0\sqrt{Z_3}$ depend only on the photon-side renormalisation computed here.

• 12.16.02 One-loop QED vertex function and the anomalous magnetic moment shares the gauge-invariance machinery of this unit: the Ward-Takahashi identity that forces ${\Pi }^{\mu \nu }$ to be transverse is the same identity that forces $F_1(0)=1$ in the vertex calculation. The two units together complete the photon-side and vertex-side one-loop dressing of QED.

• [12.16.04 — pending] Lamb shift in the Bethe-Welton-Feynman style uses the Uehling potential of this unit as one of three contributions to the hydrogen 2S-2P splitting, alongside the self-energy contribution from 12.16.01 and the vertex contribution from 12.16.02. The Uehling piece is the smallest in normal hydrogen (27 MHz of 1058 MHz) but completely dominant in muonic hydrogen (205 GHz of 206 GHz).

• [12.16.05 — pending] Bloch-Nordsieck and the cancellation of infrared divergences concerns the IR side of the renormalisation programme; vacuum polarisation is the UV side of the photon line. The two cancellation mechanisms operate on different divergences and at different sides of the diagrams.

• [12.16.06 — pending] Two-loop vacuum polarisation and the Källén-Sabry potential extends this unit's one-loop computation to second order in $\alpha$ and produces the next-order correction to the bound-state Lamb shift, important in heavy-ion atomic physics.

• 12.05.05 Free Dirac quantum field gives the canonical-quantisation framework — CAR, mode expansion, Feynman propagator — that the bare fermion propagator inside the loop comes from.

• 03.09.02 Clifford algebra underlies the gamma-matrix trace identities $\mathrm{tr}({\gamma }^{\mu }{\gamma }^{\nu })=4{\eta }^{\mu \nu }$ and $\mathrm{tr}({\gamma }^{\mu }{\gamma }^{\nu }{\gamma }^{\rho }{\gamma }^{\sigma })=4({\eta }^{\mu \nu }{\eta }^{\rho \sigma }-{\eta }^{\mu \rho }{\eta }^{\nu \sigma }+{\eta }^{\mu \sigma }{\eta }^{\nu \rho })$ on which every step of the photon self-energy computation depends.

• 12.06.04 Crossing symmetry and the CPT theorem lies one step downstream: the same dressed photon propagator survives under crossing $s\leftrightarrow t$ in $2\to 2$ scattering and under CPT in the comparison of particle and antiparticle one-photon exchange.

Historical & philosophical context Master

Vacuum polarisation in quantum electrodynamics was the first manifestation of a genuinely new physical effect predicted by Dirac's 1928-1930 reformulation of the relativistic electron equation. Dirac's hole theory (Proc. Roy. Soc. A 126, 360, 1930) had identified the negative-energy solutions of the Dirac equation as occupied states of the vacuum, with holes in the negative-energy sea interpreted as positrons. Anderson's 1932 discovery of the positron (Phys. Rev. 43, 491, 1933) confirmed the hole-theory picture and made concrete the idea that the QED vacuum could be polarised by external electromagnetic fields: an external charge could pull virtual electrons out of the negative-energy sea and reorganise the vacuum charge distribution at distances comparable to the Compton wavelength $1/m_e$.

Edwin Uehling, then a graduate student at the University of Michigan working with Robert Oppenheimer, computed the leading correction in 1935 (Phys. Rev. 48, 55, 1935) ^{[Uehling 1935]}. Uehling's calculation was done in the pre-renormalisation framework of Dirac hole theory and used a momentum cutoff $\Lambda$ to render the integrals finite; the cutoff dropped out of the leading correction to the Coulomb potential, leaving the famous expression $V_U(r)=-(Z\alpha /r)[1+(2\alpha /3\pi )\chi (2m_{e}r)]$. Robert Serber independently obtained the same result in the same volume of Physical Review (Phys. Rev. 48, 49, 1935) ^{[Serber 1935]}, using a slightly different organisation of the calculation. Both papers were technical achievements of the pre-Feynman-diagram era — derivations of a finite correction to a classical potential from the divergent perturbation theory of Dirac hole theory, carried out by direct manipulation of the negative-energy projector.

Heisenberg, Euler, and Weisskopf had earlier (Z. Phys. 90, 209, 1934; Z. Phys. 98, 714, 1936) computed the effective Lagrangian of QED in slowly-varying external electromagnetic fields, obtaining the Euler-Heisenberg Lagrangian that contains the same one-loop vacuum-polarisation information in the constant-field limit. Schwinger's 1949 covariant treatment (Phys. Rev. 75, 651, 1949) ^{[Schwinger 1949]} reorganised the Uehling-Serber calculation in the language of the renormalisation programme, replacing the momentum cutoff with the modern counterterm structure ${\delta }_3=Z_3-1$ and clarifying that the Uehling potential is the Fourier transform of the dressed Coulomb interaction sourced by the renormalised photon propagator. This was the first instance in which the renormalisation programme was applied to a photon-side observable, and the success of the calculation was a key piece of evidence that the renormalisation framework was systematic rather than ad hoc.

The bound-state extension to higher orders in $Z\alpha$ — needed for heavy-atom and high-precision spectroscopy — was carried out by Wichmann and Kroll in 1954 (Phys. Rev. 96, 232, 1954) ^{[Wichmann-Kroll 1954]}. Their calculation showed that the next-order vacuum-polarisation contribution to bound-state energies comes from diagrams with three or more photon attachments to the electron loop, in which the loop interacts non-perturbatively with the strong external Coulomb field of a heavy nucleus. Mohr, Plunien, and Soff's 1998 review (Phys. Rep. 293, 227, 1998) ^{[Mohr-Plunien-Soff 1998]} is the comprehensive modern reference for bound-state QED including vacuum polarisation to two loops and all orders in $Z\alpha$.

Gell-Mann and Low (Phys. Rev. 95, 1300, 1954) ^{[Gell-Mann-Low 1954]} extracted the running coupling and the QED beta function from the renormalised vacuum-polarisation function. Their paper introduced the renormalisation-group equation $\mu d\alpha /d\mu =\beta (\alpha )$ as a tool for resumming leading-log contributions to all orders, and the resulting $\beta (\alpha )=2{\alpha }^2/(3\pi )$ at one loop became the prototype for every subsequent calculation of running couplings in non-Abelian gauge theories. The positive sign of the QED beta function — $\alpha$ grows with energy — is what distinguishes QED from QCD's asymptotic freedom (Gross-Wilczek-Politzer 1973) and is what underlies the QED Landau pole.

Modern experimental tests of vacuum polarisation have moved from electronic hydrogen Lamb-shift spectroscopy to muonic hydrogen and to the muon $g-2$ programme. The CREMA collaboration's 2010 measurement (Pohl et al., Nature 466, 213, 2010) ^{[Pohl 2010]} of the muonic-hydrogen 2S-2P Lamb shift to 50 ppm precision was the first measurement in which vacuum-polarisation contributions dominated by orders of magnitude over the other QED corrections, and the resulting proton-radius extraction precipitated the "proton-radius puzzle" that has driven much of low-energy precision QED for the past fifteen years. The Fermilab Muon g-2 experiment's results (Abi et al., Phys. Rev. Lett. 126, 141801, 2021; Aguillard et al., Phys. Rev. Lett. 131, 161802, 2023) currently quote $a_{\mu }$ to 0.20 ppm, with a tension of $4.2\sigma$ relative to the dispersion-relation Standard-Model prediction or $1.5\sigma$ relative to the BMW 2020 lattice-QCD prediction (Borsanyi et al., Nature 593, 51, 2021). Resolving this discrepancy — which is a discrepancy between two ways of computing the hadronic vacuum-polarisation contribution to ${\Pi }^R$ — is the central open problem in the field.

Bibliography Master

Primary literature:

• Dirac, P. A. M. A theory of electrons and protons. Proc. Roy. Soc. A 126, 360 (1930). The hole-theory framework that predicted vacuum polarisation as a physical effect.

• Uehling, E. A. Polarization effects in the positron theory. Phys. Rev. 48, 55 (1935). The original derivation of the vacuum-polarisation correction to the Coulomb potential.

• Serber, R. Linear modifications in the Maxwell field equations. Phys. Rev. 48, 49 (1935). Independent, contemporaneous derivation of the same result by a slightly different method.

• Heisenberg, W. & Euler, H. Consequences of Dirac's theory of the positron. Z. Phys. 98, 714 (1936). The effective Lagrangian for QED in constant external electromagnetic fields, equivalent to the vacuum-polarisation calculation in the slow-field limit.

• Schwinger, J. Quantum electrodynamics. II. Vacuum polarization and self-energy. Phys. Rev. 75, 651 (1949). The covariant renormalisation-programme treatment of vacuum polarisation, replacing the Uehling-Serber cutoff with modern counterterm subtraction.

• Pauli, W. & Villars, F. On the invariant regularization in relativistic quantum theory. Rev. Mod. Phys. 21, 434 (1949). The first Lorentz-invariant UV regulator, used in the contemporary verification of Schwinger's vacuum-polarisation result.

• Källén, G. & Sabry, A. Fourth order vacuum polarization. Dan. Mat. Fys. Medd. 29, 17 (1955). The two-loop QED photon self-energy and the resulting Källén-Sabry potential.

• Wichmann, E. H. & Kroll, N. M. Vacuum polarization in a strong Coulomb field. Phys. Rev. 96, 232 (1954). The higher-order bound-state QED corrections beyond the Uehling potential; extended to two loops in subsequent papers.

• Gell-Mann, M. & Low, F. E. Quantum electrodynamics at small distances. Phys. Rev. 95, 1300 (1954). The renormalisation-group derivation of the running coupling $\alpha ({\mu }^2)$ from the leading-log piece of vacuum polarisation.

• Ward, J. C. An identity in quantum electrodynamics. Phys. Rev. 78, 182 (1950); Takahashi, Y. On the generalized Ward identity. Nuovo Cim. 6, 371 (1957). The Ward-Takahashi identity that enforces the transversality of ${\Pi }^{\mu \nu }$.

• 't Hooft, G. & Veltman, M. Regularization and renormalization of gauge fields. Nucl. Phys. B44, 189 (1972). The dimensional-regularisation framework that makes the transversality of ${\Pi }^{\mu \nu }$ manifest at every intermediate step.

• Lamb, W. E. & Retherford, R. C. Fine structure of the hydrogen atom by a microwave method. Phys. Rev. 72, 241 (1947); Phys. Rev. 79, 549 (1950). The experimental measurement of the 2S-2P splitting against which the Uehling and other one-loop contributions are calibrated.

• Bethe, H. A. The electromagnetic shift of energy levels. Phys. Rev. 72, 339 (1947). The first renormalised QED prediction of the Lamb shift, including a placeholder for the vacuum-polarisation contribution later given exactly by the Uehling result.

• Mohr, P. J., Plunien, G. & Soff, G. QED corrections in heavy atoms. Phys. Rep. 293, 227 (1998). The comprehensive review of bound-state QED including Uehling, Wichmann-Kroll, and two-loop vacuum-polarisation contributions, with explicit numerical values for hydrogenic ions throughout the periodic table.

• Pohl, R. et al. (CREMA Collaboration). The size of the proton. Nature 466, 213 (2010); Antognini, A. et al. Proton structure from the measurement of 2S-2P transition frequencies of muonic hydrogen. Science 339, 417 (2013). The muonic-hydrogen Lamb-shift measurements in which the Uehling potential dominates the QED contribution and the proton-radius puzzle was discovered.

• Aoyama, T. et al. The anomalous magnetic moment of the muon in the Standard Model. Phys. Rep. 887, 1 (2020). The Theory Initiative white paper, with detailed treatment of the hadronic vacuum-polarisation contribution to the muon $g-2$.

• Borsanyi, S. et al. (BMW Collaboration). Leading hadronic contribution to the muon magnetic moment from lattice QCD. Nature 593, 51 (2021). The lattice-QCD evaluation of the leading hadronic vacuum-polarisation contribution, currently in tension with dispersion-relation evaluations.

• Aguillard, D. P. et al. (Fermilab Muon g-2 Collaboration). Measurement of the positive muon anomalous magnetic moment to 0.20 ppm. Phys. Rev. Lett. 131, 161802 (2023). The current most-precise experimental measurement of $a_{\mu }$, in tension with the dispersion-relation Standard-Model prediction.

Textbook treatments:

• Berestetskii, V. B., Lifshitz, E. M. & Pitaevskii, L. P. Quantum Electrodynamics. Vol. 4 of the Landau-Lifshitz Course of Theoretical Physics, 2e. Pergamon / Butterworth-Heinemann, 1982. §113-114 (vacuum polarisation and the Uehling potential), §117 (radiative corrections to bound-state energies including Lamb shift). The canonical physicist-process-driven treatment.

• Peskin, M. E. & Schroeder, D. V. An Introduction to Quantum Field Theory. Westview Press, 1995. §7.5 (vacuum polarisation, dressed photon propagator, Uehling potential, running coupling, beta function), with explicit Feynman-parameter manipulations and Fourier-transform identities.

• Weinberg, S. The Quantum Theory of Fields, Vol. I: Foundations. Cambridge University Press, 1995. §11.2 (photon self-energy and the renormalisation of the electric charge), §13.2 (the Ward identity in full generality).

• Schwartz, M. D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014. §16.2 (vacuum polarisation in detail), §19 (renormalisation-group running of $\alpha$).

• Itzykson, C. & Zuber, J.-B. Quantum Field Theory. McGraw-Hill, 1980. §7-1-1 (vacuum polarisation with explicit gamma-matrix trace algebra), §7-3-2 (the Uehling correction to bound-state energies).

• Bjorken, J. D. & Drell, S. D. Relativistic Quantum Fields. McGraw-Hill, 1965. §19.10 (vacuum polarisation and the Uehling potential in cutoff-regularised form).

• Srednicki, M. Quantum Field Theory. Cambridge University Press, 2007. Chs. 62-64 (vacuum polarisation, Uehling potential, beta function in dim reg).

• Greiner, W. & Reinhardt, J. Quantum Electrodynamics, 4e. Springer, 2009. §5.5 (vacuum polarisation with worked numerical evaluation of the $\chi$ kernel and explicit bound-state Lamb-shift evaluation).

• Jentschura, U. & Adkins, G. Quantum Electrodynamics: Atoms, Lasers and Gravity. World Scientific, 2022. Modern monograph treatment of bound-state QED including vacuum polarisation in muonic hydrogen and Wichmann-Kroll corrections to the proton-radius puzzle.