Lamb shift from one-loop QED

Intuition Beginner

When Dirac wrote his relativistic equation for the electron in 1928, he solved it in the Coulomb field of a proton and obtained the Sommerfeld fine-structure spectrum. That spectrum has a peculiar feature: the $2s_{1/2}$ and $2p_{1/2}$ levels of hydrogen come out exactly degenerate. Both have $n=2$, both have $j=1/2$, and the Dirac-Coulomb formula depends on $n$ and $j$ alone. For twenty years this was treated as a hard prediction of relativistic quantum mechanics.

In April 1947 Willis Lamb and Robert Retherford, working at Columbia, used a microwave-spectroscopy technique on a beam of metastable hydrogen atoms and measured the frequency required to drive transitions between the two supposedly-degenerate states. They found a splitting of about 1000 megahertz — not zero. The $2s_{1/2}$ state sits about a part in $10^7$ above the $2p_{1/2}$ state, by an energy of about $4.4\times 10^{-6}$ electronvolts. The experiment is precise and unambiguous; the Dirac equation is wrong about hydrogen at this level.

Hans Bethe heard the result at the Shelter Island Conference in June 1947 and within a week, on the train back from New York to Schenectady, he produced the first theoretical calculation. His idea was that the electron in hydrogen is constantly emitting and reabsorbing virtual photons — exactly as a free electron does, contributing to its self-energy and shifting its mass. The mass shift of a free electron is unobservable because it is absorbed into the definition of the physical mass. But the mass shift of a bound electron differs by a calculable amount from the free-electron shift, and that difference is observable. Bethe got a non-relativistic estimate of about 1040 megahertz, agreeing with Lamb-Retherford to two per cent.

The full relativistic calculation came in 1949 from three groups working independently: French and Weisskopf at MIT, Kroll and Lamb at Columbia, and Schwinger at Harvard. They split the calculation into three pieces. Most of the shift (about $+1052$ MHz) comes from the electron self-energy, refined to a properly relativistic treatment with the cutoff replaced by renormalisation. About $+68$ MHz comes from the anomalous magnetic moment that the same Schwinger had computed in 1948. And about $-27$ MHz — a small negative shift — comes from vacuum polarisation, the Uehling contribution of Edwin Uehling's 1935 paper.

The three pieces sum to $1052+68-27=1093$ MHz at the leading-order level; with all higher-order corrections included (recoil, two-loop, nuclear-size, two-photon-exchange) the theoretical total is $1057.845(2)$ MHz, matching the modern experimental value to roughly one part in $10^7$.

The Lamb shift is the most consequential precision test in twentieth-century physics. It was the empirical anchor that turned quantum electrodynamics from a theoretical curiosity into the most precisely tested theory ever constructed, and it forced the renormalisation programme that organises every quantum field theory built since.

Visual Beginner

HYDROGEN n = 2 LEVELS: DIRAC vs. EXPERIMENT
=============================================

   Dirac-Coulomb 1928               Lamb-Retherford 1947
   -------------------               -----------------------
                            
        2P_{3/2}   (j = 3/2)              2P_{3/2}
        ----------                        ----------
            |                                |
            | fine structure                 | fine structure
            | ~ 10 969 MHz                   | ~ 10 969 MHz
            v                                v
        2S_{1/2} = 2P_{1/2}               2S_{1/2}    <- shifted UP
        ------------------                ----------    by Lamb shift
        (both j = 1/2)                                  ~ 1057.845 MHz
        DEGENERATE                        2P_{1/2}
                                          ----------

ONE-LOOP QED DECOMPOSITION (n = 2, hydrogen)
=============================================
                                                contribution
   contribution        diagram                     to splitting
   ------------------  ---------------------     ----------------
   (a) electron        electron self-energy        + 1052 MHz
       self-energy     dressed by virtual photon
                       (Bethe 1947 NR estimate;
                        Kroll-Lamb / French-
                        Weisskopf 1949 relativistic)
   (b) anomalous       vertex correction;          + 68  MHz
       magnetic        F_2(0) = alpha/(2 pi)
       moment          (Schwinger 1948)
                       acts on 2P_{1/2}
   (c) vacuum          Uehling potential;          - 27  MHz
       polarisation    delta-function in 2S_{1/2}
                       (Uehling 1935; Serber 1935;
                        Schwinger 1949)
                                                ----------------
                                       TOTAL  ~ + 1093 MHz
                                       
   Higher-order (two-loop, recoil, finite-nucleus):  ~ - 35 MHz
                                                ----------------
                                       NET    ~ + 1057.85 MHz

HISTORY AT A GLANCE
====================
                              year     contribution
                              ----     -------------------------------
   Lamb-Retherford            1947     microwave 2S-2P splitting found
   Bethe                      1947     non-rel log-divergent estimate
   Welton                     1948     zitterbewegung picture
   French-Weisskopf           1949     full relativistic calculation
   Kroll-Lamb                 1949     independent relativistic calc.
   Schwinger                  1949     covariant counterterm calc.
   Bethe-Salpeter             1957     textbook synthesis
   Mohr-Plunien-Soff          1998     all-order bound-state QED review
   Eides-Grotch-Shelyuto      2001     light-hydrogenic-bound-state synth.
   Pohl et al. (CREMA)        2010     muonic-hydrogen Lamb-shift; r_p
   Beyer / Bezginov           2017-19  electronic-H revisions; r_p puzzle
                                       partially resolved

Object	Symbol	Role
Lamb-shift splitting	$\Delta E_L=E(2S_{1/2})-E(2P_{1/2})$	the observable, ~ 1057.845 MHz
Electron-self-energy contribution	$\Delta E^{\text{self}}$	dominant ~ $+1052$ MHz on $2S_{1/2}$
Vertex / $a_e$ contribution	$\Delta E^{a_e}$	~ $+68$ MHz on $2P_{1/2}$
Vacuum-polarisation contribution	$\Delta E^U$	~ $-27$ MHz on $2S_{1/2}$
Bethe logarithm	$\log k_0(n,\ell )$	numerical atomic matrix element; $\log k_0(2,0)=2.984128$
Proton charge radius	$r_p$	extracted from $2S-2P$ in hydrogen and muonic hydrogen; current value $\approx 0.841$ fm
Fine-structure constant	$\alpha =1/137.036$	the small parameter of the perturbative expansion
Compton wavelength	${\lambda }_C=1/m_e\approx 386$ fm	the scale of vacuum-polarisation screening

Worked example Beginner

Problem. Reproduce Bethe's 1947 non-relativistic estimate of the dominant electron-self-energy contribution to the hydrogen $2S-2P$ Lamb splitting. The Bethe formula, after cancellation of the free-electron mass shift, is

$$\Delta E_{\text{Bethe}}^{\text{self}}(n,\ell )=\frac{4{\alpha }^5{m}_e{c}^2}{3\pi n^3}\log \frac{K}{⟨E_n-E_m{⟩}_{n,\ell }}{\delta }_{\ell ,0},$$

with cutoff $K\sim m_e{c}^2$ and Bethe logarithm $\log k_0(2,0)=2.984128$ for the $2s$ state. Compute the numerical value and compare to the Lamb-Retherford measurement.

Solution.

Step 1. The ${\delta }_{\ell ,0}$ means only $s$-states are shifted at leading order in the non-relativistic estimate; in particular, $2P_{1/2}$ ($\ell =1$) is unshifted, so the entire splitting comes from $2S_{1/2}$. Substitute $n=2$, $\ell =0$:

$$\Delta E_{\text{Bethe}}^{\text{self}}(2,0)=\frac{4{\alpha }^5{m}_e{c}^2}{3\pi \cdot 8}\log \frac{K}{⟨E_2-E_m{⟩}_{2,0}}=\frac{{\alpha }^5{m}_e{c}^2}{6\pi }L,$$

with $L:=\log [K/⟨E_2-E_m{⟩}_{2,0}]$ the Bethe logarithm.

Step 2. Numerically, ${\alpha }^5=(1/137.036{)}^5=2.069\times 10^{-11}$, $m_e{c}^2=0.510999$ MeV, and $1/(6\pi )=0.05305$. So the prefactor is

$$\frac{{\alpha }^5{m}_e{c}^2}{6\pi }=2.069\times 10^{-11}\times 0.510999\times 0.05305\text{ MeV}=5.611\times 10^{-13}\text{ MeV}.$$

Converting to frequency via $1\text{ MeV}=2.418\times 10^{20}$ Hz: prefactor $=1.357\times 10^8$ Hz $=135.7$ MHz.

Step 3. Bethe estimated $L\approx 7.63$ (the logarithm of the cutoff $K=m_e{c}^2$ divided by an effective atomic-energy denominator $⟨E_2-E_m{⟩}_{2,0}$ of about 17.8 Rydberg, where 1 Rydberg $=13.6$ eV). Multiplying:

$$\Delta E_{\text{Bethe}}^{\text{self}}(2S)=135.7\text{ MHz}\times 7.63=1036\text{ MHz}.$$

Step 4. Compare to Lamb-Retherford 1947: measured splitting $1057.85$ MHz. Bethe's estimate gets 98% of the answer from a non-relativistic calculation with an ad-hoc cutoff. The remaining 22 MHz comes from the relativistic corrections that Schwinger, French-Weisskopf, and Kroll-Lamb computed in 1949: the vertex contribution from the anomalous magnetic moment ($+68$ MHz on $2P_{1/2}$), the vacuum-polarisation Uehling contribution ($-27$ MHz on $2S_{1/2}$), and the relativistic refinement of the self-energy calculation itself ($\sim -16$ MHz net).

What this tells us. Bethe's 1947 estimate is not magic; it is the non-relativistic limit of a properly renormalised quantum-electrodynamic calculation, with the cutoff $K$ standing in for the renormalisation scale and the Bethe logarithm $\log k_0$ standing in for the atomic-scale matrix element that bound-state perturbation theory produces.

The fact that the cutoff drops out of the final answer (the $K$-dependence of the bound-state self-energy is exactly cancelled by the $K$-dependence of the subtracted free-electron self-energy) is the renormalisation of bound-state quantum electrodynamics at the atomic scale, a year before Schwinger formalised the same idea in covariant perturbation theory. The remaining percent of the answer required all the technology of the 1949 papers, and confirmed the renormalisation programme as a predictive theory of radiative corrections.

Check your understanding Beginner

Formal definition Intermediate+

Setup. Consider the hydrogen atom with the proton treated as a static point charge $+e$ at the origin (Born-Oppenheimer / infinite-nuclear-mass limit). The bare Hamiltonian is the Dirac-Coulomb operator $H_0=\mathbf{\alpha }\cdot \mathbf{p}+\beta m_e-\alpha /r$ on the four-spinor Hilbert space $L^2({\mathbb{R}}^3;{\mathbb{C}}^4)$, with bound-state eigenfunctions ${\psi }_{n,j,\ell ,m}(\mathbf{r})$ and eigenvalues $E_{n,j}$ given by the Sommerfeld formula derived in 12.11.03. The eigenvalues depend only on $n$ and $j$ (the principal and total-angular-momentum quantum numbers), so for $j=1/2$ and $n=2$ the states $(2S_{1/2},2P_{1/2})$ are degenerate at the level of $H_0$.

The Lamb shift is the first-order shift in the $E_{n,j}$ eigenvalue produced by treating the radiative-correction Hamiltonian $H_{\text{rad}}$ as a perturbation of $H_0$:

$$\boxed{\Delta E_{n,j,\ell }^{\text{Lamb}}=⟨{\psi }_{n,j,\ell }|H_{\text{rad}}|{\psi }_{n,j,\ell }⟩+\mathcal{O}({\alpha }^2{H}_{\text{rad}}),}$$

with $H_{\text{rad}}$ the effective one-loop QED Hamiltonian. The Lamb-shift splitting is the difference

$$\Delta E^L(2S-2P)=\Delta E_{2,1/2,0}^{\text{Lamb}}-\Delta E_{2,1/2,1}^{\text{Lamb}},$$

the experimentally accessible quantity that Lamb-Retherford measured.

The three one-loop contributions. The radiative-correction Hamiltonian at one loop decomposes into three terms whose detailed structures are the subjects of 12.16.01, 12.16.02, and 12.16.03 respectively:

• $H_{\text{rad}}^{\text{self}}$ from the bound-state electron self-energy (the bound analogue of the on-shell mass renormalisation of 12.16.01),
• $H_{\text{rad}}^{a_e}$ from the anomalous magnetic moment $a_e=\alpha /(2\pi )$ (the on-shell limit of the vertex form factor $F_2(0)$ of 12.16.02),
• $H_{\text{rad}}^U$ from the Uehling vacuum-polarisation potential (the Fourier transform of the dressed Coulomb interaction of 12.16.03).

The total Lamb shift is

$$\boxed{\Delta E_{n,j,\ell }^{\text{Lamb}}=\Delta E_{n,j,\ell }^{\text{self}}+\Delta E_{n,j,\ell }^{a_e}+\Delta E_{n,j,\ell }^U+\mathcal{O}({\alpha }^2).}$$

Each of the three terms has a clean analytic expression at leading order in $Z\alpha$, exhibited below.

Cancellation of UV and IR scales. Each of the three contributions individually has the ultraviolet and infrared scale structure of its parent unit:

• $\Delta E^{\text{self}}$ inherits the bound-state self-energy UV divergence; the difference between the bound-state and free-electron self-energies is UV-finite (Bethe's cancellation), and the IR divergence in the wave-function renormalisation $Z_2$ from 12.16.01 is replaced in the bound-state context by the atomic-scale infrared cutoff $⟨E_n-E_m⟩$.
• $\Delta E^{a_e}$ inherits the IR-finiteness of $F_2(0)$ from 12.16.02; no UV or IR regulator is needed.
• $\Delta E^U$ inherits the UV-finiteness of ${\Pi }^R(q^2)$ after photon-side renormalisation from 12.16.03; the IR scale is set by the electron Compton wavelength $1/m_e$.

The Lamb-shift programme is therefore a textbook case of bound-state renormalisation: the perturbation $H_{\text{rad}}$ is well-defined after the usual on-shell renormalisation of $\delta m$, $Z_2$, $Z_3$, and the resulting bound-state energy shifts are finite, predictive, and testable.

Sign and metric conventions

This unit uses the same conventions as the QED prerequisites: mostly-plus metric ${\eta }_{\mu \nu }=\mathrm{diag}(+1,-1,-1,-1)$, gamma matrices $\{{\gamma }^{\mu },{\gamma }^{\nu }\}=2{\eta }^{\mu \nu }1$, QED vertex factor $-ie{\gamma }^{\mu }$ with $e>0$ the proton charge, Feynman-gauge photon propagator $-i{g}_{\mu \nu }/(q^2+iϵ)$. Energy shifts $\Delta E>0$ raise the level, $\Delta E<0$ lower it; the Lamb-shift convention is $\Delta E^L=E(2S_{1/2})-E(2P_{1/2})>0$ (the $2S$ sits above the $2P$). The hydrogen atom is treated in the non-recoil limit $m_p\to \infty$ with proton charge $+e$; recoil corrections appear at order $m_e/m_p\approx 1/1836$ and are systematically smaller than the leading Lamb shift.

Key derivation Intermediate+

Theorem (Bethe-Schwinger-French-Weisskopf-Kroll-Lamb, 1947-1949). The leading one-loop Lamb shift of a hydrogen-like atom of nuclear charge $Z$ at principal quantum number $n$, total angular momentum $j$, and orbital angular momentum $\ell$ is

$$\Delta E_{n,j,\ell }^{\text{Lamb}}=\frac{\alpha (Z\alpha {)}^4{m}_e{c}^2}{\pi n^3}\left[{\mathcal{L}}_{\text{self}}(n,\ell )+{\mathcal{L}}_{a_e}(j,\ell )+{\mathcal{L}}_U(\ell )\right]+\mathcal{O}(\alpha (Z\alpha {)}^5),$$

with the three coefficient functions

$${\mathcal{L}}_{\text{self}}(n,\ell )=\frac{1}{2}\left[\log \frac{1}{(Z\alpha {)}^2}-\log k_0(n,\ell )+\frac{11}{72}-\frac{1}{6}\right]{\delta }_{\ell ,0}+(\text{spin-orbit pieces for }\ell >0),$$

$${\mathcal{L}}_{a_e}(j,\ell )=\frac{1}{4(j+\frac{1}{2})(\ell +\frac{1}{2})}[{\delta }_{j,\ell +1/2}-{\delta }_{j,\ell -1/2}]\cdot (1-{\delta }_{\ell ,0})+\frac{1}{2}{\delta }_{\ell ,0},$$

$${\mathcal{L}}_U(\ell )=-\frac{1}{15}{\delta }_{\ell ,0}.$$

For hydrogen ($Z=1$) the prefactor $\alpha (Z\alpha {)}^4{m}_e{c}^2/(\pi n^3)$ at $n=2$ evaluates numerically to ${\alpha }^5{m}_e{c}^2/(8\pi )=135.6$ MHz $\cdot h$, and the Lamb-shift splitting $\Delta E^L(2S-2P)=\Delta E_{2,1/2,0}^{\text{Lamb}}-\Delta E_{2,1/2,1}^{\text{Lamb}}$ summing the three contributions gives $1052-27+68=1093$ MHz at the leading order. Higher-order corrections (two-loop QED, recoil, nuclear-finite-size) reduce this by $\sim 35$ MHz to the experimental value $1057.845$ MHz (Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}; Eides-Grotch-Shelyuto 2001 ^{[Eides-Grotch-Shelyuto 2001]}).

Proof. Treat each of the three contributions in turn.

(a) Electron self-energy (Bethe 1947 NR; French-Weisskopf / Kroll-Lamb 1949 relativistic). The bound-state electron self-energy is the diagonal matrix element

$$\Delta E_{n,\ell }^{\text{self}}=⟨{\psi }_{n,\ell }|{\Sigma }_{\text{bound}}(E_n,\mathbf{r},{\mathbf{r}}^{\prime })|{\psi }_{n,\ell }⟩,$$

where ${\Sigma }_{\text{bound}}$ is the bound-state self-energy operator: the analogue of the free-electron $\Sigma (p)$ of 12.16.01 but with the on-shell free-electron projector $\text{slashed}p-m$ replaced by the bound-state Coulomb-eigenstate projector. Bethe's 1947 reformulation evaluates ${\Sigma }_{\text{bound}}$ in the non-relativistic limit, where it reduces to the second-order Rayleigh-Schrödinger matrix element

$$\Delta E_{n,\ell }^{\text{self, NR}}=-\frac{2\alpha }{3\pi m_e^2}{\int }_0^{K}d\omega \sum _m\frac{|⟨m|\mathbf{p}|n⟩{|}^2(E_n-E_m)}{E_n-E_m-\omega }-\Delta E_n^{\text{self, free}}.$$

(Here $K$ is an ultraviolet cutoff on the photon momentum $\omega$, and the subtraction of the free-electron self-energy $\Delta E_n^{\text{self, free}}$ is what makes the difference finite.) The free-electron contribution has the same UV structure as the bound piece — both diverge linearly in $K$ — but with the bound-state-dependent matrix element replaced by its free-particle limit $|⟨p|\mathbf{p}|n⟩{|}^2\to p^2$. The leading-$K$ divergences cancel, and the remaining finite piece is

$$\Delta E_{n,\ell }^{\text{self, NR}}=\frac{4{\alpha }^5{m}_e{c}^2}{3\pi n^3}\log \frac{K}{⟨E_n-E_m{⟩}_{n,\ell }}{\delta }_{\ell ,0},$$

with $⟨E_n-E_m{⟩}_{n,\ell }$ the Bethe logarithm matrix element

$$\log k_0(n,\ell ):=\frac{⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\log |E_n-H_0|\mathbf{p}|{\psi }_{n,\ell }⟩}{⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\mathbf{p}|{\psi }_{n,\ell }⟩}.$$

(The ${\delta }_{\ell ,0}$ in the leading term reflects that only $s$-states have non-zero wavefunction at the origin in the non-relativistic limit; $p$-states contribute at sub-leading order in $Z\alpha$.) For hydrogen $2s$, numerical evaluation gives $\log k_0(2,0)=2.984128$ (Drake-Swainson 1990, Phys. Rev. A 41, 1243). Bethe estimated $K\sim m_e{c}^2$ on the grounds that the non-relativistic approximation breaks down above the electron rest energy; this gives the cutoff logarithm $\log (m_e{c}^2/17.8\text{Ry})=\log (0.511\text{MeV}/17.8\cdot 13.6\text{eV})=\log (2.11\times 10^3)=7.65$, and the numerical estimate $\Delta E_{2,0}^{\text{self, NR}}=({\alpha }^5{m}_e{c}^2/6\pi )\cdot 7.65\approx 1040$ MHz.

The relativistic reformulation by Kroll-Lamb 1949 ^{[Kroll-Lamb 1949]} and French-Weisskopf 1949 ^{[French-Weisskopf 1949]} replaces the ad-hoc cutoff $K$ with the proper on-shell renormalisation of the free-electron self-energy, and adds the spin-orbit and Darwin pieces that the non-relativistic calculation drops. The result for $\ell =0$ is

$$\Delta E_{n,0}^{\text{self}}=\frac{\alpha (Z\alpha {)}^4{m}_e{c}^2}{2\pi n^3}\left[\log \frac{1}{(Z\alpha {)}^2}-\log k_0(n,0)+\frac{11}{72}-\frac{1}{6}\right]+\mathcal{O}(\alpha (Z\alpha {)}^5).$$

The relativistic constants $11/72-1/6=11/72-12/72=-1/72$ replace Bethe's $K$-dependent log. The contribution is positive: $\Delta E_{2,0}^{\text{self}}\approx +1052$ MHz in hydrogen.

For $\ell \ne 0$ the leading self-energy contribution comes from a spin-orbit term that contributes $\sim -10$ MHz to $2P_{1/2}$ and $\sim -8$ MHz to $2P_{3/2}$, both much smaller than the $\ell =0$ contribution and not part of the leading Lamb-shift splitting at the precision considered.

(b) Anomalous magnetic moment (Schwinger 1948). The vertex-form-factor $F_2(0)=\alpha /(2\pi )$ derived in 12.16.02 generates an additional spin-orbit-like coupling in the non-relativistic Pauli reduction:

$$H_{\text{rad}}^{a_e}=\frac{\alpha }{2\pi }\cdot \frac{1}{2m_e^{2}r}\frac{dV}{dr}\mathbf{L}\cdot \mathbf{\sigma }=\frac{\alpha }{2\pi }\cdot \frac{(Z\alpha )}{2m_e^2{r}^3}\mathbf{L}\cdot \mathbf{\sigma },$$

with $V(r)=-Z\alpha /r$ the Coulomb potential. This is the same spin-orbit operator that appears in the Pauli reduction of the Dirac equation, but multiplied by the anomalous-moment factor $\alpha /(2\pi )$. Its expectation value on a hydrogenic eigenstate of total angular momentum $j$ and orbital angular momentum $\ell$ is (using $\mathbf{L}\cdot \mathbf{\sigma }=2\mathbf{L}\cdot \mathbf{S}$ and the standard $j(j+1)-\ell (\ell +1)-3/4$ algebra):

$$\Delta E_{n,j,\ell }^{a_e}=\frac{\alpha (Z\alpha {)}^4{m}_e{c}^2}{2\pi n^3}\cdot \frac{1}{2(\ell +\frac{1}{2})(j+\frac{1}{2})}[{\delta }_{j,\ell +1/2}-{\delta }_{j,\ell -1/2}](1-{\delta }_{\ell ,0})+(\text{Darwin term for }\ell =0).$$

For $\ell =0$ the spin-orbit operator vanishes (no orbital angular momentum), but the Foldy-Wouthuysen reduction of the anomalous-moment vertex generates a Darwin-like contact term that contributes to the $s$-state shift; in the standard organisation this Darwin contribution is included in the ${\mathcal{L}}_{a_e}$ coefficient $+\frac{1}{2}{\delta }_{\ell ,0}$ above (Bethe-Salpeter §21 ^{[Bethe-Salpeter 1957]}). For $\ell =1,j=1/2$, the contribution is

$$\Delta E_{2,1/2,1}^{a_e}=-\frac{\alpha (Z\alpha {)}^4{m}_e{c}^2}{2\pi \cdot 8}\cdot \frac{1}{2\cdot 1\cdot 1}⟹-\frac{{\alpha }^5{m}_e{c}^2}{32\pi }\approx -33.9\text{MHz}.$$

The minus sign comes from ${\delta }_{j,\ell -1/2}$ being the active piece for $j<\ell +1/2$. So $2P_{1/2}$ is shifted down by ~$34$ MHz, contributing $+34$ MHz to the splitting $\Delta E^L(2S-2P)=E(2S)-E(2P)$. The Darwin $\ell =0$ contribution to $2S_{1/2}$ contributes an additional $+34$ MHz to $2S$ in the same direction; the net is $+68$ MHz on the splitting (Bethe-Salpeter §21 ^{[Bethe-Salpeter 1957]}; Sakurai-Napolitano §5.4 ^{[Sakurai-Napolitano 2011]}).

(c) Vacuum polarisation (Uehling 1935). The Uehling potential derived in 12.16.03 modifies the Coulomb potential at short distances. For the leading bound-state shift on $s$-states, the ${\delta }^3(\mathbf{r})$ approximation of 12.16.03 Lemma 7 gives

$$\Delta E_{n,0}^U=-\frac{Z{\alpha }^2}{15\pi }\cdot \frac{4\pi }{m_e^2}|{\psi }_{n,0}(0){|}^2=-\frac{4Z{\alpha }^2}{15m_e^2{n}^3}\cdot \frac{(Z\alpha m_e{)}^3}{\pi }=-\frac{4(Z\alpha {)}^5{m}_e{c}^2}{15\pi n^3},$$

after substituting $|{\psi }_{n,0}(0){|}^2=(Z\alpha m_e{)}^3/(\pi n^3)$ for the hydrogenic $s$-state. Writing this in the canonical $\alpha (Z\alpha {)}^4{m}_e{c}^2/(\pi n^3)$ units extracts the coefficient ${\mathcal{L}}_U(\ell =0)=-1/15$, and the $\ell \ne 0$ contribution is zero at leading order because $|{\psi }_{n,\ell }(0){|}^2=0$ for $\ell \ge 1$. Numerically for hydrogen $2s$:

$$\Delta E_{2,0}^U=-\frac{4{\alpha }^5{m}_e{c}^2}{15\pi \cdot 8}=-\frac{{\alpha }^5{m}_e{c}^2}{30\pi }\approx -27.1\text{MHz}.$$

The Uehling contribution is negative and lowers the $2S$ level. This was computed by Uehling in 1935 (Phys. Rev. 48, 55) ^{[Uehling 1935 via 12.16.03]}, twelve years before Lamb-Retherford measured the shift; the calculation was an isolated prediction in the pre-renormalisation era and was confirmed to be a real contribution to the splitting only after Bethe's 1947 calculation revealed that the bigger self-energy piece was also a real, calculable effect.

Summing the three contributions. For hydrogen $n=2$, $j=1/2$:

Contribution	$2S_{1/2}$ shift	$2P_{1/2}$ shift	Net contribution to $\Delta E^L=E(2S)-E(2P)$
(a) Self-energy	$+1052$ MHz	$\sim 0$	$+1052$ MHz
(b) Anomalous moment	$+34$ MHz	$-34$ MHz	$+68$ MHz
(c) Uehling	$-27$ MHz	$0$	$-27$ MHz
Total (leading)	$+1059$ MHz	$-34$ MHz	$+1093$ MHz

Higher-order corrections — two-loop QED (Källén-Sabry, Appelquist-Brodsky), recoil ($m_e/m_p$), nuclear finite size ($r_p^2$), and the Wichmann-Kroll bound-state vacuum-polarisation contribution beyond Uehling — reduce the prediction to $1057.845(2)$ MHz, matching the experimental value to one part in $10^7$ (Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}; Eides-Grotch-Shelyuto 2001 ^{[Eides-Grotch-Shelyuto 2001]}). $\square$

Counterexamples to common slips

• The Lamb shift is not a hyperfine effect. Hyperfine structure (proton-electron-spin coupling) splits each of $2S_{1/2}$ and $2P_{1/2}$ into singlet and triplet sub-levels separated by $\sim 178$ MHz (for $2S_{1/2}$) and $\sim 59$ MHz (for $2P_{1/2}$); the Lamb shift is the splitting between the centroids of the two levels and is roughly an order of magnitude larger. Lamb-Retherford 1947 resolved both the Lamb shift and the hyperfine structure in the same experiment.

• The Uehling potential's sign is negative: it lowers the $2S$ level. The self-energy and anomalous-moment contributions are both positive on $2S_{1/2}$. The cancellations and the sum-to-positive happen at the level of the three terms; the net Lamb shift $E(2S)-E(2P)>0$ is the result of competing effects, not of a single dominant mechanism.

• Bethe's 1947 estimate uses a cutoff $K\sim m_e{c}^2$, but the cutoff drops out of the final answer after subtracting the free-electron self-energy. The relativistic calculation by French-Weisskopf and Kroll-Lamb replaces the cutoff by the proper on-shell renormalisation scheme; the underlying physics — bound-state minus free-state self-energy is finite — is the same.

• The leading three-term sum overshoots experiment by $\sim 35$ MHz, not undershoots. This is a feature of the perturbative expansion, not an error: the next-order terms (two-loop, recoil) are individually larger than the residue and combine to bring the prediction down to $1057.845$ MHz. Truncating at one loop gives 1093 MHz; truncating at two loops gives $\sim 1058$ MHz; truncating at all known orders gives $1057.845(2)$ MHz.

• The Lamb shift in hydrogen is dominated by the electron self-energy ($\sim 95\%$), with only $\sim 6\%$ from the vertex and $-2.5\%$ from vacuum polarisation. In muonic hydrogen, the ratios are inverted: the Uehling contribution dominates at $\sim 99\%$ of the Lamb shift because the muon's much smaller Bohr radius ($a_0^{\mu }\approx 250$ fm, comparable to $1/m_e\approx 386$ fm) puts the muon wavefunction inside the electron-positron polarisation cloud. The same calculation, applied to different atoms, makes radically different physical predictions.

Exercises Intermediate+

Advanced results Master

The Lamb-shift programme since 1949 has extended in three directions: ever-more-precise theoretical calculations of bound-state QED, complementary high-precision experiments in electronic and muonic atoms, and applications of the same calculational technology to heavy hydrogenic ions and to the muon $g-2$ anomaly.

Two-loop QED bound-state corrections. Källén and Sabry 1955 (Dan. Mat. Fys. Medd. 29, 17) computed the two-loop photon self-energy that contributes the Källén-Sabry potential, the two-loop analogue of the Uehling potential. The two-loop electron self-energy was computed by Appelquist and Brodsky 1970 (Phys. Rev. A 2, 2293), Sapirstein-Yennie 1990 (in Quantum Electrodynamics, ed. Kinoshita, World Scientific 1990), and the comprehensive two-loop calculation was completed by Pachucki 1993-94 (Phys. Rev. A 48, 2609; Phys. Rev. A 49, 4639), reaching coefficients of order ${\alpha }^2(Z\alpha {)}^4{m}_e{c}^2$ at the few-tens-of-kHz level for hydrogen. Yerokhin-Indelicato-Shabaev 2003-2008 (Phys. Rev. Lett. 91, 073001; Phys. Rev. A 77, 062510) computed all ${\alpha }^2(Z\alpha {)}^6$ contributions, bringing the theoretical Lamb-shift precision to $\sim 2$ kHz on the 1058 MHz total — an agreement of $\sim 10^{-7}$ with the experimental value, the most precise comparison between theory and experiment in atomic physics.

Three-loop and the running coupling. Three-loop QED contributions to the bound-state energy were computed by Eides-Shelyuto 1994 (Phys. Rev. A 50, 4521) and Pachucki 1999 (Phys. Rev. A 60, 1006), with current theoretical precision limited by the ~10 ppb uncertainty in the proton charge radius $r_p$ — itself extracted from the same Lamb-shift measurements at the muonic-hydrogen level. The leading effect of the running QED coupling between the electron Compton wavelength and the Bohr scale is captured by the substitution $\alpha \to \alpha (\mu =m_e)$ in the leading-order formula, but the running between $m_e$ and the hydrogen ionisation scale ($\sim 13.6$ eV $=27\times 10^{-6}{m}_e{c}^2$) is a 0.01% effect and well below current experimental precision.

Bethe-Salpeter equation and recoil corrections. The leading recoil correction at order $(m_e/m_p)(Z\alpha {)}^4{m}_e{c}^2$ was computed by Salpeter 1952 (Phys. Rev. 87, 328) using the two-body Bethe-Salpeter equation; the result is a $\sim +24$ MHz shift on the hydrogen $2S$ level. Higher-order recoil corrections ($m_e/m_p\cdot \alpha (Z\alpha {)}^4,(m_e/m_p{)}^2(Z\alpha {)}^4$, etc.) have been computed by Pachucki-Karshenboim 1995 (J. Phys. B 28, L221) and Czarnecki-Melnikov-Yelkhovsky 1999-2002 (Phys. Rev. Lett. 82, 311; Phys. Rev. A 65, 062106), with the comprehensive review in Eides-Grotch-Shelyuto 2001 ^{[Eides-Grotch-Shelyuto 2001]}. The Bethe-Salpeter equation is the rigorous two-body framework that generalises the Schrödinger equation for bound systems of two relativistic particles; its connection to the perturbative renormalisation programme was the subject of an extensive 1950s-1960s development by Salpeter, Mandelstam, and others, ultimately providing the systematic framework for all bound-state-QED calculations in light hydrogenic systems.

Heavy-ion bound-state QED. For hydrogen-like ions with high $Z$ — uranium U${}^{91+}$ has $Z\alpha =92/137.036\approx 0.671$ — the perturbative expansion in $Z\alpha$ ceases to converge and bound-state QED must be done non-perturbatively in $Z\alpha$ from the start. The radiative corrections are computed by numerically solving the Dirac equation in the strong external Coulomb field and constructing the self-energy and vacuum-polarisation contributions in the resulting bound-state basis. Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]} is the canonical review of this non-perturbative bound-state QED programme; current measurements of the U${}^{91+}$ ground-state Lamb shift (Gumberidze et al. 2005, Phys. Rev. Lett. 94, 223001) test bound-state QED at $Z\alpha \sim 0.7$ to 1% precision and constitute one of the few precision tests of QED in the strong-field regime where the small parameter $Z\alpha$ ceases to be small.

Connection to the muon $g-2$ anomaly. The same calculational technology that produces the Lamb shift in hydrogen produces the QED contribution to the muon anomalous magnetic moment $a_{\mu }$, the subject of the Brookhaven 2006 (Bennett et al., Phys. Rev. D 73, 072003) and Fermilab 2021-2023 (Abi et al., Phys. Rev. Lett. 126, 141801; Aguillard et al., Phys. Rev. Lett. 131, 161802) measurements. The current experimental world average is $a_{\mu }^{\text{exp}}=0.00116592059(22)$, in tension with the Standard Model prediction at the 4-5$\sigma$ level depending on whether the leading hadronic vacuum-polarisation contribution is taken from $e^{+}{e}^{-}\to$ hadrons dispersion-relation evaluations (Aoyama et al. 2020, Phys. Rep. 887, 1) or from the BMW 2020 lattice-QCD calculation (Borsanyi et al., Nature 593, 51). The Lamb-shift QED is the same QED that enters the muon $g-2$; the precision tests are complementary, and any new physics that affects one must affect the other in calculable ways. The current $4-5\sigma$ tension in the muon $g-2$ is the most striking open precision-test anomaly in particle physics, and its resolution would either confirm new physics at the $\sim 2$ TeV scale or refine the hadronic contribution beyond current uncertainties.

Muonic-atom spectroscopy and the proton-radius puzzle. The CREMA collaboration's 2010 measurement (Pohl et al., Nature 466, 213 ^{[Pohl 2010]}) of the 2S-2P transition in muonic hydrogen with 50 ppm precision extracted a proton charge radius $r_p^{\mu }=0.84087(39)$ fm, smaller by 7$\sigma$ than the then-standard electronic-hydrogen and elastic-electron-scattering values $r_p^e=0.879(8)$ fm. The discrepancy — the proton-radius puzzle — drove a decade of theoretical and experimental scrutiny. The 2017 1S-3S measurement by Beyer et al. (Science 358, 79 ^{[Beyer-Bezginov 2017]}) and the 2019 2S-2P measurement by Bezginov et al. (Science 365, 1007) revised the electronic-hydrogen value downward to $r_p^e\approx 0.833(10)$ fm, partially resolving the puzzle by bringing the two measurements into agreement at the $\sim 1\sigma$ level. The remaining tension with the elastic-scattering value $r_p^{\text{ep}}=0.879(8)$ fm awaits further high-precision $ep$ experiments at MAMI/A1 and at MUSE (PSI). The lesson is that high-precision Lamb-shift spectroscopy is sensitive at the percent level to the proton's internal structure, not just to the QED radiative corrections; the proton-radius extraction is now the limiting precision uncertainty on the theoretical Lamb shift.

Alternate hydrogenic systems. The Lamb-shift calculations apply directly to positronium (electron-positron bound state; Stroscio 1975, Phys. Rep. 22, 215), muonium (${\mu }^{+}{e}^{-}$; Hughes et al. 2001 review in Atomic and Molecular Beams), and antihydrogen (CERN ALPHA collaboration, Ahmadi et al. 2017 Nature 541, 506 — first 1S-2S antihydrogen spectroscopy). Each system provides a different sensitivity to different terms in the QED expansion: positronium has no nuclear-finite-size correction but a large annihilation contribution; muonium tests the QED of a purely leptonic two-body system without strong-interaction contamination; antihydrogen tests CPT symmetry at the bound-state level. The calculations in all four systems use the same technology developed in the 1947-1957 Lamb-shift programme, with minor modifications for the different mass ratios and gyromagnetic ratios.

Connection to the renormalisation group. The leading-log term $\log (1/(Z\alpha {)}^2)$ in the self-energy contribution is the bound-state analogue of the renormalisation-group running of the coupling: it captures the leading contribution of all orders in $\alpha$ that scales as powers of $\alpha \log (1/\alpha )$. The structure $[\log (1/\alpha ){]}^n\cdot {\alpha }^{n+1}$ at $n$-loop is the leading log approximation, and its resummation via the Gell-Mann-Low equation (Gell-Mann-Low 1954, Phys. Rev. 95, 1300) is equivalent to using the running coupling $\alpha (\mu =m_e{c}^2)$ in place of $\alpha$ in the leading-order formula. The fact that this reorganisation gives the same answer is the bound-state-QED analogue of the renormalisation-group invariance that connects the running-coupling and explicit-log languages in flat-space QED.

Full proof set Master

The Key derivation supplies the leading-order Lamb-shift formula and the three individual contributions. The following auxiliary results are stated with proof outlines verifiable against the cited literature.

Lemma 1 (Bethe-logarithm definition and numerical value). The Bethe logarithm $\log k_0(n,\ell )=⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\log |E_n-H_0|\mathbf{p}|{\psi }_{n,\ell }⟩/⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\mathbf{p}|{\psi }_{n,\ell }⟩$ is well-defined for any hydrogenic eigenstate with $\ell \ge 0$, and for $n=2$, $\ell =0$ takes the value $2.984128556$ (Drake-Swainson 1990, Phys. Rev. A 41, 1243). Proof outline. The denominator is the standard closure-relation evaluation: $⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\mathbf{p}|{\psi }_{n,\ell }⟩=⟨{\psi }_{n,\ell }|\mathbf{p}(E_n-H_0)\mathbf{p}|{\psi }_{n,\ell }⟩=⟨{\psi }_{n,\ell }|\mathbf{p}\cdot [\mathbf{p},V]|{\psi }_{n,\ell }⟩$ since $H_0={\mathbf{p}}^2/(2m)+V$; for the Coulomb $V=-Z\alpha /r$ this gives $2\pi Z\alpha |\psi (0){|}^2/m$ — a finite, computable matrix element. The numerator $⟨\mathbf{p}(E_n-H_0)\log |E_n-H_0|\mathbf{p}⟩$ requires summing over the full hydrogen spectrum (discrete and continuum) and is computed numerically by spectral methods. Drake-Swainson 1990 give the value to 10 decimal places using Slater-orbital basis expansions. (Detailed proof: Bethe-Salpeter §19 ^{[Bethe-Salpeter 1957]}; Drake-Swainson 1990; Pachucki 1993 Phys. Rev. A 48, 2609.) $\square$

Lemma 2 (UV cancellation in the bound-state self-energy). The Bethe non-relativistic self-energy of a bound electron, after subtracting the free-electron self-energy, is finite as the UV cutoff $K\to \infty$ for any fixed hydrogenic eigenstate. Proof outline. The bound-state and free-state self-energies share the same leading high-frequency behaviour: at $\omega \gg E_{\text{atomic}}$ the matrix element $|⟨m|\mathbf{p}|n⟩{|}^2/(E_n-E_m-\omega )\to -|⟨m|\mathbf{p}|n⟩{|}^2/\omega$ for both bound and free intermediate states (by completeness, ${\sum }_m|⟨m|\mathbf{p}|n⟩{|}^2\to ⟨n|{\mathbf{p}}^2|n⟩$ in both cases). The difference, evaluated at fixed $\omega$, scales as $|⟨m|\mathbf{p}|n⟩{|}^2\cdot (E_n-E_m)/{\omega }^2$, which is UV-finite under the $\omega$-integral up to $\omega =K\to \infty$. The leading log $\log (K/E_{\text{atomic}})$ is the cutoff-dependent piece; the constant part as $K\to \infty$ is the Bethe logarithm. (Detailed proof: Bethe 1947 ^{[Bethe 1947]}; Sakurai-Napolitano §5.4 ^{[Sakurai-Napolitano 2011]}.) $\square$

Lemma 3 (Relativistic relation between Bethe cutoff and on-shell renormalisation). The Bethe cutoff $K\sim m_e{c}^2$ in the non-relativistic formula and the on-shell mass-counterterm $\delta m$ of the free-electron self-energy (12.16.01) are connected: replacing the cutoff with the on-shell-renormalised free-electron self-energy gives the same result at leading order. Proof outline. The free-electron self-energy at on-shell momentum has the dimensional-regularisation structure ${\Sigma }_2(\text{slashed}p=m)=(3m\alpha /4\pi )[1/\epsilon -{\gamma }_E+\log (4\pi {\mu }^2/m^2)+5/3]+\mathcal{O}({\alpha }^2)$; the $1/\epsilon$ pole is absorbed by $\delta m$. The bound-state self-energy in the same regularisation scheme contains both the free-particle pieces and the bound-state-specific Bethe-logarithm pieces. Subtracting $\delta m$ from the bound-state self-energy reproduces the Bethe-cutoff structure, with the cutoff $K$ replaced by the renormalisation scale $\mu$. The $\mu$-dependence cancels between the constants $-{\gamma }_E+\log (4\pi {\mu }^2/m^2)$ and the explicit $-\log ⟨E_n-E_m{⟩}_{n,\ell }$ in the bound-state Bethe-log structure, with the final answer dependent only on $\log k_0(n,\ell )$ and the relativistic constants $11/72-1/6$. (Detailed proof: French-Weisskopf 1949 ^{[French-Weisskopf 1949]}; Kroll-Lamb 1949 ^{[Kroll-Lamb 1949]}; Bethe-Salpeter §21 ^{[Bethe-Salpeter 1957]}.) $\square$

Lemma 4 (Anomalous-moment spin-orbit operator). The one-loop vertex contribution $F_2(0)=\alpha /(2\pi )$ produces an additional spin-orbit term in the Pauli-reduced bound-state Hamiltonian, $H_{\text{rad}}^{a_e}=(\alpha /2\pi )\cdot (1/2m_e^{2}r)(dV/dr)\mathbf{L}\cdot \mathbf{\sigma }$. Proof outline. Start from the QED vertex $-ie[{\gamma }^{\mu }{F}_1(q^2)+i{\sigma }^{\mu \nu }{q}_{\nu }{F}_2(q^2)/(2m)]$ of (12.16.02), evaluated at on-shell $q^2\to 0$. The Pauli term $i{\sigma }^{\mu \nu }{q}_{\nu }/(2m)$ couples to the electromagnetic field as $-(e/2m)F_2(0)\bar{\psi }{\sigma }^{\mu \nu }{F}_{\mu \nu }\psi$. Apply the Foldy-Wouthuysen reduction to the non-relativistic limit: $\bar{\psi }{\sigma }^{\mu \nu }{F}_{\mu \nu }\psi \to -(2/m)(\mathbf{\sigma }\cdot \mathbf{E}\times \mathbf{p}+\mathbf{\sigma }\cdot \mathbf{B})-(1/m^2)(\nabla \cdot \mathbf{E})\mathbf{\sigma }+\cdots$. The first term is the spin-orbit coupling $(\mathbf{\sigma }\cdot \mathbf{L})(1/r)dV/dr$ for a central Coulomb potential; the second term is the Darwin-like contact term that contributes to $s$-states. Multiplying by $F_2(0)=\alpha /(2\pi )$ and identifying the contribution as the additional perturbation beyond the Dirac equation gives $H_{\text{rad}}^{a_e}$. (Detailed proof: Bethe-Salpeter §21 ^{[Bethe-Salpeter 1957]}; Itzykson-Zuber §7-1-2.) $\square$

Lemma 5 (Uehling potential $\delta$-function approximation in bound states). For bound-state $s$-states with Bohr radius $a_0\gg 1/m_e$ (electronic hydrogen), the leading contribution of the Uehling potential to the bound-state energy is $\Delta E_{n,0}^U=-(Z{\alpha }^2/15\pi )(4\pi /m_e^2)|{\psi }_{n,0}(0){|}^2=-4(Z\alpha {)}^5{m}_e{c}^2/(15\pi n^3)$. Proof outline. The Uehling correction $\delta V_U(r)$ from (12.16.03) has the momentum-space form ${\Pi }^R(-|\mathbf{q}{|}^2)/|\mathbf{q}{|}^2$, which at low $|\mathbf{q}|\ll m_e$ is $\approx -\alpha /(15\pi m_e^2)$, a constant. The Fourier transform of a constant is a ${\delta }^3(\mathbf{r})$. The first-order matrix element of $\delta V_U(\mathbf{r})=-(Z{\alpha }^2/15\pi )(4\pi /m_e^2){\delta }^3(\mathbf{r})$ on an $s$-state of wavefunction-squared $|\psi (0){|}^2=(Z\alpha m_e{)}^3/(\pi n^3)$ gives the announced result. The approximation is valid for electronic hydrogen because $a_0^e=1/(\alpha m_e)\approx 5.3\times 10^4$ fm dominates the Compton scale $1/m_e\approx 386$ fm. (Detailed proof: see (12.16.03) Lemma 7; Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}; Bethe-Salpeter §21 ^{[Bethe-Salpeter 1957]}.) $\square$

Lemma 6 (Recoil correction at order $m_e/m_p$). The leading proton-recoil correction to the hydrogen $2S$ Lamb shift is $\Delta E^{\text{recoil}}(2S)=(m_e/m_p)(Z\alpha {)}^4{m}_e{c}^2/n^3\cdot [\text{coefficient }\sim 1]\approx +24$ MHz. Proof outline. Start from the two-body Bethe-Salpeter equation for the bound state $(e^{-}p)$, which generalises the one-body Dirac-Coulomb equation by replacing the static proton charge with a dynamical proton field. The leading recoil correction is the first-order expansion in $m_e/m_p\approx 1/1836$ of the bound-state energy. Salpeter 1952 (Phys. Rev. 87, 328) computed the result explicitly using the two-body QED Hamiltonian; the contribution adds $\sim +24$ MHz to the hydrogen $2S$ Lamb shift and is one of the dominant higher-order corrections beyond the one-loop QED programme. (Detailed proof: Salpeter 1952; Bethe-Salpeter §22-23 ^{[Bethe-Salpeter 1957]}; Eides-Grotch-Shelyuto 2001 ^{[Eides-Grotch-Shelyuto 2001]}.) $\square$

Lemma 7 (Two-loop QED bound-state corrections of order ${\alpha }^2(Z\alpha {)}^4{m}_e{c}^2$). The two-loop QED contribution to the hydrogen $2S$ Lamb shift is approximately $-10$ MHz, comprising contributions from two-loop self-energy ($-7.5$ MHz), two-loop vacuum polarisation Källén-Sabry ($-0.7$ MHz), two-loop vertex (~$-1.5$ MHz), and crossed photon-exchange diagrams. Proof outline. Each of the four contributions is computed by the same dimensional-regularisation programme applied at two loops, with the bound-state evaluation handled by the same matrix-element technology as the one-loop case. The full calculation was completed by Pachucki 1993-94 (Phys. Rev. A 48, 2609; A 49, 4639) building on partial earlier work by Appelquist-Brodsky 1970 (Phys. Rev. A 2, 2293) and Sapirstein-Yennie 1990 (in Quantum Electrodynamics, ed. Kinoshita, World Scientific 1990). The total two-loop correction brings the predicted Lamb shift from the leading-order 1093 MHz down to 1058 MHz; the residual $\sim 0.04$ MHz discrepancy between theory and the experimental 1057.845 MHz is consumed by recoil, three-loop, and nuclear-finite-size corrections. (Detailed proof: Pachucki 1993-94; Eides-Grotch-Shelyuto 2001 ^{[Eides-Grotch-Shelyuto 2001]}; Mohr-Plunien-Soff 1998 ^{[Mohr-Plunien-Soff 1998]}.) $\square$

Connections Master

• 12.16.01 Electron self-energy and mass renormalisation at one loop supplies the free-electron self-energy ${\Sigma }_2(\text{slashed}p)$ that the bound-state Lamb-shift calculation subtracts from the bound-state self-energy to produce the UV-finite difference. The on-shell renormalisation conditions $\Sigma (\text{slashed}p=m)=0$ and $d\Sigma /d\text{slashed}p{|}_{\text{slashed}p=m}=0$ are the same conditions that pin the Bethe-cutoff dependence in the bound-state calculation.

• 12.16.02 One-loop QED vertex function and the anomalous magnetic moment supplies the Schwinger $a_e=\alpha /(2\pi )$ that, via the spin-orbit Foldy-Wouthuysen reduction, contributes the $\sim +68$ MHz piece of the Lamb splitting. The Ward-Takahashi identity $Z_1=Z_2$ that the vertex unit emphasises is the structural reason the renormalised charge is independent of the vertex form-factor normalisation.

• 12.16.03 Vacuum polarization at one loop and the Uehling potential supplies the $-27$ MHz Uehling contribution, the smallest of the three one-loop pieces in electronic hydrogen but completely dominant in muonic hydrogen. The same Uehling potential also produces the running of the QED coupling that ultimately reorganises the Lamb-shift formula at the leading-log level via the Gell-Mann-Low equation.

• 12.11.03 Dirac equation in a Coulomb field supplies the Dirac-Coulomb spectrum $E_{n,j}$ that the Lamb shift perturbs. The Sommerfeld degeneracy of $2S_{1/2}$ and $2P_{1/2}$ at the level of the Dirac equation is the prediction that the Lamb shift falsifies; the lift of the degeneracy by ~$1058$ MHz is the first observational signature of QED corrections beyond single-particle relativistic quantum mechanics.

• [12.16.05 — pending] Bloch-Nordsieck and the cancellation of infrared divergences concerns the IR cancellations that make physical cross-sections finite; the same mechanism is at work in the bound-state Lamb shift, where the atomic-scale matrix elements provide a natural IR cutoff via the Bethe logarithm.

• [12.16.06 — pending] Two-loop QED radiative corrections extends the one-loop programme of this unit to second order in $\alpha$ and produces the Källén-Sabry potential, the Appelquist-Brodsky two-loop self-energy, and the Pachucki comprehensive two-loop bound-state result, bringing the theoretical Lamb-shift precision to the few-kHz level.

• 12.11.01 Dirac equation and relativistic spin supplies the spinor framework — gamma matrices, Foldy-Wouthuysen reduction, on-shell projectors — that the Lamb-shift perturbation calculation uses to expand the radiative-correction operators in powers of $1/m_e$.

• 12.06.01 Schrödinger-Coulomb hydrogen is the non-relativistic limit underlying Bethe's 1947 calculation: the matrix elements $⟨m|\mathbf{p}|n⟩$ that enter the second-order Rayleigh-Schrödinger expression for the self-energy are computed on the unperturbed Schrödinger eigenfunctions.

• [12.07.05 — pending] Time-independent bound-state perturbation theory is the formal framework in which the Lamb shift is computed: the radiative Hamiltonian $H_{\text{rad}}$ is treated as a first-order perturbation of the Dirac-Coulomb Hamiltonian $H_0$, and the energy-level shifts are extracted from the diagonal matrix elements $⟨{\psi }_{n,j,\ell }|H_{\text{rad}}|{\psi }_{n,j,\ell }⟩$.

• [14.04.01 — pending] Relativistic effects in heavy atoms (chemistry) uses the same Sommerfeld $(Z\alpha {)}^4$ scaling and the same vacuum-polarisation corrections that produce the Lamb shift; for elements with $Z\alpha \to 1$ (gold, mercury, uranium), the corrections become non-perturbative and account for the colour of gold, the liquidity of mercury, and the chemistry of the transactinides.

• [10.04.05 — pending] Casimir effect shares the conceptual origin with the Lamb shift: both are vacuum-fluctuation effects that produce measurable shifts in bound-state or boundary-condition-modified vacuum energies. The historical connection between the Casimir 1948 calculation and the Bethe 1947 Lamb-shift calculation is direct: both were responses to the same wave of late-1940s vacuum-energy thinking that culminated in the systematic renormalisation programme.

Historical & philosophical context Master

The Lamb shift is the most consequential precision-physics measurement of the twentieth century, and the calculation that explained it is the founding moment of modern quantum field theory. The historical record is unusually well-documented: Lamb's Nobel Prize lecture (1955), Bethe's autobiographical account (1986), Schwinger's reminiscences in his collected papers (1979), and the Shelter Island Conference proceedings (1947) preserve a remarkable view into how the renormalisation programme emerged in the eighteen months between Lamb-Retherford 1947 and the French-Weisskopf / Kroll-Lamb / Schwinger papers of 1949.

The experimental story begins with William Houston's 1937 spectroscopic measurement (Phys. Rev. 51, 446) of an unexplained discrepancy in the hydrogen Balmer-series fine structure, which suggested the Dirac equation might be missing something at the part-per-thousand level. The work was inconclusive — the optical-spectroscopy precision of the era could not resolve the splittings cleanly. The breakthrough was the development of microwave-spectroscopy techniques during World War II for radar applications, which gave 1940s atomic physicists a tool with far better frequency resolution than optical spectroscopy. Willis Lamb, who had worked on radar at Columbia's Wave Propagation Group during the war, applied microwave spectroscopy to a beam of metastable hydrogen and on 26 April 1947 measured the 1058 MHz $2S-2P$ splitting (announced in Lamb-Retherford 1947, Phys. Rev. 72, 241 ^{[Lamb-Retherford 1947]}). The measurement was definitive: the Dirac equation's prediction of $2S_{1/2}$–$2P_{1/2}$ degeneracy was wrong by an unambiguous, reproducible, atomic-scale energy splitting.

The Shelter Island Conference (1-3 June 1947) brought together a small group of the leading American theoretical physicists — Bethe, Oppenheimer, Schwinger, Feynman, Wheeler, Weisskopf, Marshak, Rabi, Pais, Serber, Uhlenbeck, Kramers, van Vleck, and Lamb himself — to discuss the implications of Lamb's measurement and of the contemporaneous Foley-Kusch 1947 measurement of the electron's anomalous magnetic moment (Phys. Rev. 73, 412). The conference concluded that the Dirac equation needed radiative corrections from quantum electrodynamics, but no one knew how to compute them in a way that gave finite results. Hans Bethe, on the train back to Schenectady on 4 June 1947, performed the first calculation: a non-relativistic estimate of the bound-state electron self-energy with an ad-hoc cutoff at $K\sim m_e{c}^2$. His Phys. Rev. 72, 339 (1947) ^{[Bethe 1947]} gave $\sim 1040$ MHz for the $2S-2P$ splitting, agreeing with Lamb-Retherford to 2%. The calculation was published in August 1947, just three months after the measurement.

Bethe's 1947 paper was the first quantitative success of perturbative QED. It demonstrated that the radiative corrections could be calculated and that the divergences could be tamed by subtracting the free-electron self-energy from the bound-state self-energy — an early version of the mass-renormalisation that would soon be formalised by Schwinger, Tomonaga, Feynman, and Dyson. The subsequent relativistic calculations by Schwinger 1949 (Phys. Rev. 76, 790 ^{[Schwinger 1949]}), French-Weisskopf 1949 (Phys. Rev. 75, 1240 ^{[French-Weisskopf 1949]}), and Kroll-Lamb 1949 (Phys. Rev. 75, 388 ^{[Kroll-Lamb 1949]}) replaced Bethe's ad-hoc cutoff with the proper renormalisation programme and added the vertex and vacuum-polarisation contributions. The three groups reached the same answer by different routes, providing the first substantive cross-check on the consistency of the new renormalisation framework. Theodore Welton's 1948 zitterbewegung interpretation (Welton 1948, Phys. Rev. 74, 1157 ^{[Welton 1948]}) provided a physical picture for the calculation that helped diffuse the technical apparatus into the broader physics community: the electron position is jiggled by vacuum-field fluctuations, and the resulting smearing of the Coulomb potential produces the observed shift.

Feynman 1949 (Phys. Rev. 76, 769 ^{[Feynman 1949]}) recast the Schwinger-Tomonaga calculation in terms of the diagrammatic perturbation theory that became the standard pedagogical and computational tool for the rest of the twentieth century. Dyson 1949 (Phys. Rev. 75, 486) proved that the Schwinger-Tomonaga, Feynman, and Bethe approaches were three formulations of the same underlying theory, and that the renormalisation programme could be made systematic to all orders. The mid-1950s and 1960s consolidated these insights into the textbook treatments of Bethe-Salpeter 1957 ^{[Bethe-Salpeter 1957]} and Bjorken-Drell 1964-65, completing the absorption of bound-state QED into the standard physics curriculum.

The Lamb shift is also philosophically consequential. It was the first direct measurement of a vacuum effect — a measurable energy shift produced by the quantum-fluctuations of an unobserved field. The conceptual claim that empty space carries physical properties detectable through its effect on a probe particle was not new (the Casimir effect of 1948 makes the same claim from a complementary direction), but the Lamb-shift measurement was the first quantitative confirmation that the vacuum-energy concept could be extracted from quantum field theory and put to predictive use. Within the next two decades the same conceptual move underlay the QCD instanton, the Higgs mechanism, and the cosmological constant — each a vacuum-energy effect with measurable consequences, each justified ultimately by the precedent that the QED vacuum had been measured and the measurement matched theory to a part in $10^4$.

A modern open question is whether the proton-radius puzzle indicates a breakdown of QED at the muonic level or merely a systematic effect in electronic-hydrogen spectroscopy. The 2017-2019 revisions of the electronic-hydrogen value (Beyer 2017 Science 358, 79; Bezginov 2019 Science 365, 1007 ^{[Beyer-Bezginov 2017]}) brought the two measurements into agreement at the $1\sigma$ level, but the elastic-scattering value $r_p\approx 0.879(8)$ fm remains in tension. Whether this tension is a real anomaly or a systematic effect in electron-proton elastic scattering is the subject of ongoing experimental programmes at MAMI/A1 and MUSE/PSI. The Lamb-shift QED is correct at the part-per-$10^{-7}$ level when the proton-radius input is properly accounted for; the residual uncertainties are in atomic-physics inputs (Rydberg constant, proton magnetic moment) and in the proton structure itself, not in the QED radiative corrections.

Bibliography Master

Primary literature:

• Lamb, W. E. & Retherford, R. C. Fine structure of the hydrogen atom by a microwave method. Phys. Rev. 72, 241 (1947). The discovery of the $2S_{1/2}-2P_{1/2}$ splitting, the empirical anchor for the entire post-war QED renormalisation programme.

• Bethe, H. A. The electromagnetic shift of energy levels. Phys. Rev. 72, 339 (1947). The first non-relativistic estimate of the Lamb shift; established that the bound-state-minus-free-state self-energy is finite and calculable.

• Schwinger, J. Quantum electrodynamics. III. The electromagnetic properties of the electron — radiative corrections to scattering. Phys. Rev. 76, 790 (1949). The Schwinger relativistic-renormalised calculation of the Lamb shift, including the vertex and vacuum-polarisation contributions.

• French, J. B. & Weisskopf, V. F. The electromagnetic shift of energy levels. Phys. Rev. 75, 1240 (1949). Independent relativistic calculation of the Lamb shift using the renormalisation programme; one of the three 1949 papers that established the modern QED treatment.

• Kroll, N. M. & Lamb, W. E. On the self-energy of a bound electron. Phys. Rev. 75, 388 (1949). The third 1949 relativistic calculation, by Lamb himself together with Kroll. Confirmed the agreement of three independent calculational schemes.

• Welton, T. A. Some observable effects of the quantum-mechanical fluctuations of the electromagnetic field. Phys. Rev. 74, 1157 (1948). The zitterbewegung-fluctuation interpretation of the Lamb shift that provided a heuristic physical picture for the radiative-correction calculation.

• Feynman, R. P. Space-time approach to quantum electrodynamics. Phys. Rev. 76, 769 (1949). The diagrammatic-perturbation-theory formulation of QED, including a Lamb-shift calculation as one of the worked examples.

• Dyson, F. J. The radiation theories of Tomonaga, Schwinger, and Feynman. Phys. Rev. 75, 486 (1949). The proof that the three competing 1947-1948 formulations of QED are equivalent and that the renormalisation programme is systematic to all orders.

• Uehling, E. A. Polarization effects in the positron theory. Phys. Rev. 48, 55 (1935). The Uehling potential, the vacuum-polarisation contribution to the Coulomb potential that eventually contributes $-27$ MHz to the hydrogen Lamb shift; computed twelve years before Lamb-Retherford measured the splitting.

• Salpeter, E. E. Mass corrections to the fine structure of hydrogen-like atoms. Phys. Rev. 87, 328 (1952). The first calculation of the proton-recoil correction to the Lamb shift, using the two-body Bethe-Salpeter equation.

• Appelquist, T. & Brodsky, S. J. Two-loop electrodynamic corrections to the Lamb shift. Phys. Rev. A 2, 2293 (1970). The two-loop electron self-energy contribution to the Lamb shift, one of the dominant higher-order corrections.

• Pachucki, K. Complete two-loop binding correction to the Lamb shift. Phys. Rev. A 48, 2609 (1993); Higher-order binding corrections to the Lamb shift. Phys. Rev. A 49, 4639 (1994). The full two-loop bound-state QED calculation, bringing the theoretical precision to the few-kHz level.

• Yerokhin, V. A., Indelicato, P. & Shabaev, V. M. Two-loop self-energy contribution to energy levels of light hydrogenlike atoms. Phys. Rev. Lett. 91, 073001 (2003); Phys. Rev. A 77, 062510 (2008). The all-order two-loop bound-state QED self-energy calculation.

• 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 measurement that precipitated the proton-radius puzzle.

• Beyer, A. et al. The Rydberg constant and proton size from atomic hydrogen. Science 358, 79 (2017); Bezginov, N. et al. A measurement of the atomic hydrogen Lamb shift and the proton charge radius. Science 365, 1007 (2019). The 2017-2019 electronic-hydrogen revisions that partially resolved the proton-radius puzzle.

• 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 all-order radiative corrections.

• Eides, M. I., Grotch, H. & Shelyuto, V. A. Theory of light hydrogenic bound states. Phys. Rep. 342, 63 (2001); expanded as Springer Tracts in Modern Physics 222, 2007. The current standard reference for the comprehensive Lamb-shift calculation in light hydrogenic systems.

• Bennett, G. W. et al. (Muon g-2 Collaboration, BNL). Final report of the E821 muon anomalous magnetic moment measurement at BNL. Phys. Rev. D 73, 072003 (2006); Abi, B. et al. (Fermilab Muon g-2 Collaboration). Measurement of the positive muon anomalous magnetic moment to 0.46 ppm. Phys. Rev. Lett. 126, 141801 (2021); Aguillard, D. P. et al. Measurement of the positive muon anomalous magnetic moment to 0.20 ppm. Phys. Rev. Lett. 131, 161802 (2023). The muon $g-2$ measurements that use the same QED technology as the Lamb shift; the current $4-5\sigma$ tension with the Standard Model is the leading precision-physics anomaly.

• Drake, G. W. F. & Swainson, R. A. Bethe logarithms for hydrogen up to $n=20$, and approximations for two-electron atoms. Phys. Rev. A 41, 1243 (1990). The high-precision numerical evaluation of the Bethe logarithm $\log k_0(n,\ell )$.

Textbook treatments:

• Bethe, H. A. & Salpeter, E. E. Quantum Mechanics of One- and Two-Electron Atoms. Springer, 1957. The canonical textbook treatment of the Lamb shift and all related bound-state QED; written by Bethe himself with Salpeter as co-author, with the original Lamb-shift calculation and its 1949 relativistic refinement as worked examples.

• 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. §123 (Lamb shift) with the explicit decomposition into self-energy, vertex, and vacuum-polarisation contributions; the canonical physicist-process-driven treatment.

• Peskin, M. E. & Schroeder, D. V. An Introduction to Quantum Field Theory. Westview Press, 1995. §7.5 (Uehling potential and Lamb-shift contribution) and §6.3 (vertex contribution).

• Sakurai, J. J. & Napolitano, J. Modern Quantum Mechanics, 2e. Pearson, 2011. §5.4 (Lamb shift in the Welton zitterbewegung picture) — the pedagogically accessible heuristic derivation.

• Itzykson, C. & Zuber, J.-B. Quantum Field Theory. McGraw-Hill, 1980. §7-3-2 (Lamb shift via bound-state QED, including the Bethe-logarithm derivation).

• Greiner, W. & Reinhardt, J. Quantum Electrodynamics, 4e. Springer, 2009. §5.5 (Lamb shift with explicit numerical evaluation of all three contributions and the Bethe logarithm).

• Jentschura, U. & Adkins, G. Quantum Electrodynamics: Atoms, Lasers and Gravity. World Scientific, 2022. Modern monograph treatment of the Lamb shift in muonic hydrogen and the proton-radius puzzle.

• Weinberg, S. The Quantum Theory of Fields, Vol. I: Foundations. Cambridge University Press, 1995. Ch. 14 (bound-state methods); the discussion of how the bound-state perturbation theory of this unit fits into the broader Wigner-Bargmann axiomatic framework.

• Schwartz, M. D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014. §16.3 (Lamb shift with explicit derivation of the three one-loop contributions).