Infrared divergences and the Bloch-Nordsieck cancellation in QED

Intuition Beginner

In the one-loop quantum-electrodynamic calculations of the previous four units, you met two kinds of infinity. The first kind is ultraviolet: the loop momentum runs to arbitrarily high energies and the integral diverges. That divergence is tamed by renormalisation — by absorbing the infinite shift into a redefinition of the bare mass and the bare charge, leaving a finite physical prediction.

The second kind is infrared: the loop momentum of a virtual photon runs to zero, the photon becomes arbitrarily soft, and the integral diverges again. This time the divergence is not in the high-energy regime where you might suspect a deeper theory takes over, but in the low-energy regime where ordinary classical electrodynamics is supposed to be most trustworthy. The infrared divergence is a structural feature of any theory with a massless gauge boson and unconfined charged particles.

The puzzle is that the divergence is real — it does not go away by being smarter about the calculation. Compute the one-loop vertex correction to an electron scattering off a nucleus, regulate the photon by giving it a tiny fictitious mass ${\mu }_{\gamma }$, and the answer contains a term proportional to $\log ({\mu }_{\gamma })$ that blows up as you remove the regulator. For the theory to be predictive, something has to cancel that log.

Bloch and Nordsieck found the cancellation in 1937. Their observation: any real detector has a finite energy resolution $\Delta E$. The "elastic" scattering process $e^{-}\to e^{-}$ that the one-loop vertex correction dresses is operationally indistinguishable from a process where the electron emits one or more soft photons too low in energy to be detected — photons with total energy below $\Delta E$.

The cross section for emitting such soft photons is itself infrared-divergent: integrating the bremsstrahlung emission rate down to arbitrarily low frequencies produces a $\log ({\mu }_{\gamma })$ of the same magnitude but opposite sign. Sum the elastic and the soft-photon-emission contributions and the photon-mass regulator drops out. The inclusive cross section, summed over all final-state photons with total energy below $\Delta E$, is finite and depends only on the physical scale $\Delta E/E$ rather than on the regulator.

This is more than a technical workaround. It tells you that in any quantum theory with a massless gauge boson, the elastic two-particle scattering process is not a well-defined observable. The asymptotic states of quantum electrodynamics are not bare electrons and photons; they are electrons surrounded by clouds of arbitrarily soft photons. The cross section you can compute is the inclusive one, summed over the soft-photon emissions the detector cannot resolve.

A generation later, Kinoshita, Lee, and Nauenberg extended the cancellation. When the final-state charged particles are massless (or so much more energetic than their masses that they behave that way), collinear singularities appear in addition to soft singularities — divergences from a photon emitted exactly along the direction of a charged particle. Summing over all degenerate final states (soft photons plus collinear photons) makes the answer finite.

The operational consequence at modern colliders: you don't measure cross sections for "an electron"; you measure cross sections for a jet — a collimated bundle of all the particles that emerge along a single direction with total energy in some specified window. The mathematical underpinning is the Kinoshita-Lee-Nauenberg theorem; the observable definition is the jet.

The infrared structure of gauge theories is one of the surprising places where classical and quantum field theory speak the same language: the classical Larmor formula for radiation from an accelerated charge contains exactly the same soft-photon factor as the quantum bremsstrahlung amplitude, and the Bloch-Nordsieck mechanism is the quantum-mechanical statement that any acceleration of a charged particle is accompanied by an infinite number of arbitrarily soft photons whose total energy and total momentum are finite. The infinity of photon number and the finiteness of physical observables coexist in the same theory.

Visual Beginner

INFRARED DIVERGENCES AND THE BLOCH-NORDSIECK CANCELLATION
==========================================================

   VIRTUAL one-loop                REAL soft emission
   ----------------                ------------------

     e-              e-              e-              e-
   ---*----.........----*---       ---*----------------*---
        \  virtual  /                       \
         \ photon  /                         \  soft gamma
          \  mu_g /                           \  omega < Delta E
           \     /                             \  detector cannot
            \   /                               \  resolve
             \./
              .

   F_1(q^2) ~ - (alpha/2pi) ·         sigma_brems ~ + (alpha/pi) ·
              log(m/mu_g)                            log(m/mu_g)
              log(-q^2/m^2)                          log(Delta E/m)

   [DIVERGES as mu_g to 0]          [DIVERGES as mu_g to 0,
                                     OPPOSITE sign]

   INCLUSIVE SUM (Bloch-Nordsieck 1937)
   ====================================

   sigma_incl(Delta E)  =  sigma_elastic  +  sigma_soft-gamma(Delta E)
                                          +  sigma_soft-gamma-gamma + ...

                       =  sigma_tree  *  (Delta E / E)^((2 alpha / pi) f(theta))

                                                    (Yennie-Frautschi-Suura
                                                     1961 exponentiation)

   FINITE.  mu_gamma drops out at every order in alpha.
   Depends only on the physical ratio Delta E / E.

HISTORY AT A GLANCE
====================
                                  year     contribution
                                  ----     -------------------------------
   Pauli-Fierz                    1938     long-wavelength radiation
                                           dressing of charged states
   Bloch-Nordsieck                1937     cancellation theorem in
                                           non-covariant form
   Sudakov                        1956     double-log resummation at
                                           high energies (form factor)
   Yennie-Frautschi-Suura         1961     all-orders covariant
                                           exponentiation
   Kinoshita                      1962     mass singularities; collinear
                                           and soft cancellation
   Lee-Nauenberg                  1964     degenerate-state generalisation
   Faddeev-Kulish                 1970     coherent-state asymptotic
                                           framework for quantum
                                           electrodynamics
   Sterman-Weinberg               1977     jet definition; first IR-safe
                                           QCD cross section
   Frenkel-Taylor                 1984     non-Abelian eikonal
                                           exponentiation
   Catani-Seymour                 1997     general NLO subtraction
                                           algorithm for IR-safe jets

Object	Symbol	Role
Photon mass regulator	${\mu }_{\gamma }$	fictitious IR cutoff; drops out of observables
Detector energy resolution	$\Delta E$	physical scale below which soft emissions are unresolved
One-loop electron form factor	$F_1(q^2)$	IR-divergent vertex correction; carries the Sudakov double log
Eikonal soft factor	$ϵ\cdot p/(k\cdot p)$	universal factorised emission of a soft photon $k$ from a charge with momentum $p$
Bremsstrahlung kernel	$d{\sigma }_{\text{soft}}=d{\sigma }_{\text{tree}}\cdot (\alpha /\pi )d\omega /\omega \cdot f(\theta )$	universal soft-emission rate; integrated up to $\Delta E$ gives the matching IR log
YFS exponent	$B(\theta )$	scattering-angle-dependent positive function in $(\Delta E/E{)}^{B(\theta )\alpha /\pi }$
Sudakov double log	$\alpha {\log }^2(s/m^2)$	dominant high-energy correction; resummed by the Sudakov form factor
Jet cone parameter	$R$	angular size around a final-state hard parton over which collinear emissions are summed (Sterman-Weinberg 1977)
Fine-structure constant	$\alpha =1/137.036$	the small parameter of the quantum-electrodynamics perturbative expansion

Worked example Beginner

Problem. A charged particle of energy $E$ scatters elastically off a heavy nucleus and is deflected through angle $\theta$. Order-of-magnitude estimate: how does the probability of no soft-photon emission above some detector threshold $\Delta E$ behave as $\Delta E$ is lowered toward zero? Use the leading-log Bloch-Nordsieck exponentiation $P_{\text{no emission}}(\Delta E)=(\Delta E/E{)}^B$ with $B=(2\alpha /\pi )f(\theta )$ and a representative value $f(\theta )\approx 1$ for hard scattering.

Solution.

Step 1. Plug numbers. Take $\alpha =1/137$, so $2\alpha /\pi =2/(137\pi )=4.64\times 10^{-3}$. With $f(\theta )\approx 1$, the exponent is $B\approx 4.64\times 10^{-3}$.

Step 2. Evaluate at three representative resolution ratios. For $\Delta E/E=0.1$ (the detector resolves photons down to 10% of the beam energy), $P_{\text{no emission}}=0.1^{0.00464}=e^{0.00464\times \log 0.1}=e^{-0.01068}=0.989$. For $\Delta E/E=0.01$, $P_{\text{no emission}}=0.01^{0.00464}=e^{-0.02137}=0.979$. For $\Delta E/E=10^{-6}$, $P_{\text{no emission}}=10^{-6\times 0.00464}=10^{-2.78\times 10^{-2}}=0.938$.

Step 3. Now push to $\Delta E/E\to 0$. The probability $(\Delta E/E{)}^B$ tends to zero as $\Delta E\to 0$ for any positive $B$. The probability that a hard-scattering process produces no photon at all, with the detector resolving arbitrarily soft emissions, is zero.

Step 4. Compare to what the leading-log estimate would say if you tried to use plain perturbation theory. At one loop, the perturbative virtual correction to the elastic cross section is $\delta {\sigma }_{\text{virtual}}/{\sigma }_{\text{tree}}=-B\log (E/{\mu }_{\gamma })$, which blows up as the regulator ${\mu }_{\gamma }\to 0$. Adding the one-soft-photon bremsstrahlung gives $\delta {\sigma }_{\text{soft}}/{\sigma }_{\text{tree}}=+B\log (\Delta E/{\mu }_{\gamma })$, also divergent. Their sum is $-B\log (E/\Delta E)=B\log (\Delta E/E)$, which is finite but negative and large when $\Delta E\ll E$. The exponentiated form $(\Delta E/E{)}^B$ is the resummation of this leading log to all orders in $\alpha$; it never goes negative, and it correctly tends to zero (rather than to minus infinity) as $\Delta E\to 0$.

What this tells us. A charged particle scattering with no detectable photon emission is an increasingly improbable event the better your detector. At any finite detector resolution the probability is finite and computable. In the limit of a perfect detector that resolves arbitrarily soft photons, the probability of the "elastic" process drops to zero — every hard scattering process is accompanied, with probability one, by infinitely many arbitrarily soft photons whose total energy and momentum are nonetheless finite.

The exponentiated answer $(\Delta E/E{)}^{(2\alpha /\pi )f(\theta )}$ is the all-orders Yennie-Frautschi-Suura resummation of the leading soft logs. The fact that $2\alpha /\pi \approx 0.005$ is small means the exponent is small for ordinary detector resolutions, and the deviation from one (the no-radiation probability stays close to one) is correspondingly small. At collider experiments with $\Delta E/E$ ratios of $10^{-3}$ to $10^{-1}$, the effect is a few-percent correction — small but routinely measurable, and entirely absent if you ignore the inclusive prescription.

Check your understanding Beginner

Formal definition Intermediate+

Setup. Consider a QED scattering amplitude ${\mathcal{M}}_n(p_1,\dots ,p_n)$ for a process with $n$ external on-shell charged-particle and photon momenta, computed in covariant perturbation theory to some fixed order in $\alpha$. The amplitude is a Lorentz-covariant tensor / spinor expression whose modulus squared, integrated against the standard $n$-body Lorentz-invariant phase-space measure, defines the differential cross section $d{\sigma }_n$. Regulate the photon by a fictitious mass ${\mu }_{\gamma }$: the photon propagator becomes $-i{g}_{\mu \nu }/(k^2-{\mu }_{\gamma }^2+iϵ)$ and the on-shell photon dispersion is $k^0=\sqrt{{\mathbf{k}}^2+{\mu }_{\gamma }^2}$. All loop and phase-space integrals are then convergent at long wavelengths.

An infrared divergence is a $\log ({\mu }_{\gamma })$ or ${\log }^k({\mu }_{\gamma })$ singularity in ${\mathcal{M}}_n$ or in the phase-space integral $d{\sigma }_n$ as ${\mu }_{\gamma }\to 0$, originating from regions of loop or phase-space integration where one or more virtual or real photon momenta become arbitrarily soft.

The inclusive cross section with detector resolution $\Delta E$ is the sum over all final-state configurations that differ from a chosen hard configuration by emission of one or more soft photons of total energy at most $\Delta E$:

$$\boxed{{\sigma }_{\text{incl}}(\Delta E)=\sum _{k=0}^{\infty }{\int }_{\text{soft}(\Delta E)}d{\sigma }_{n+k}(p_1,\dots ,p_n;q_1,\dots ,q_k),}$$

where the sum runs over the number $k$ of unresolved soft photons and the integration domain $\text{soft}(\Delta E)$ restricts each soft-photon energy ${\omega }_i<\Delta E$ and the total soft energy ${\sum }_i{\omega }_i<\Delta E$.

The Bloch-Nordsieck theorem is the statement that ${\sigma }_{\text{incl}}(\Delta E)$ is infrared-finite to all orders in $\alpha$: the limit ${\mu }_{\gamma }\to 0$ exists and is finite for every $k$ in the sum and for the sum itself. Equivalently, the IR divergences of the virtual loop corrections at $O({\alpha }^n)$ cancel against the IR divergences of the real-soft-emission phase-space integrals at the same order, with the cancellation holding term by term.

The eikonal approximation for soft-photon emission is the universal factorisation, valid when the soft photon energy $\omega$ is much less than every charged-particle energy in the process:

$${\mathcal{M}}_{\text{soft emission}}(p_1,\dots ,p_n;k)={\mathcal{M}}_{\text{hard}}(p_1,\dots ,p_n)\cdot \sum _{i=1}^{n_{\text{charged}}}{\eta }_i{Q}_i\frac{{ϵ}^{\ast }(k)\cdot p_i}{k\cdot p_i}+\mathcal{O}({\omega }^0),$$

with $Q_i$ the charge of the $i$-th external charged particle in units of $e$, ${\eta }_i=+1$ for outgoing and $-1$ for incoming, and $ϵ(k)$ the polarisation vector of the emitted photon. The soft factor is universal: it depends only on the external momenta and charges of the hard process, not on its internal dynamics.

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 $-i{g}_{\mu \nu }/(k^2-{\mu }_{\gamma }^2+iϵ)$. Soft-photon energies are integrated up from ${\mu }_{\gamma }$ (the regulator) to $\Delta E$ (the detector resolution); the $\log ({\mu }_{\gamma })$ divergences cancel between virtual and real contributions and the surviving $\log (\Delta E)$ dependence is a physical, measurable feature of the cross section.

Key derivation Intermediate+

Theorem (Bloch-Nordsieck, 1937). In quantum electrodynamics with a photon mass regulator ${\mu }_{\gamma }$, the inclusive cross section for any QED scattering process, summed over final-state photons with total energy below a detector resolution $\Delta E$, is infrared-finite to all orders in $\alpha$. Concretely, for the elastic scattering $e^{-}+X\to e^{-}+X$ at tree level with one-loop QED corrections plus real soft-photon emission, the inclusive cross section through $O({\alpha }^3)$ takes the form

$${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}[1+\frac{2\alpha }{\pi }f(\theta )\log \frac{\Delta E}{E}+\mathcal{O}({\alpha }^2)],$$

with $f(\theta )>0$ depending on the scattering geometry, and the $\log ({\mu }_{\gamma })$ from the virtual and the real pieces canceling exactly. The all-orders resummation of the leading soft logs (Yennie-Frautschi-Suura, 1961 ^{[Yennie-Frautschi-Suura 1961]}) gives the exponentiated form

$$\boxed{{\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}\left(\frac{\Delta E}{E}{)}^{(2\alpha /\pi )f(\theta )}\right[1+\mathcal{O}(\alpha )],}$$

finite as ${\mu }_{\gamma }\to 0$ and as a power of the physical ratio $\Delta E/E$.

Proof. Work through the cancellation at $O({\alpha }^3)$ explicitly, then sketch the all-orders exponentiation.

(1) The virtual IR divergence at one loop. The one-loop vertex correction to the elastic amplitude is, from the analysis of 12.16.02, proportional to the Dirac form factor $F_1(q^2)$. Splitting $F_1(q^2)=1+\delta F_1(q^2)$ where $\delta F_1$ contains the one-loop correction, and isolating the IR-divergent piece by the Feynman-parameter method at ${\mu }_{\gamma }\to 0$, gives (Peskin-Schroeder §6.4 ^{[Peskin-Schroeder 1995]}):

$$\delta F_1(q^2){|}_{\text{IR}}=-\frac{\alpha }{2\pi }\left[\log \frac{-q^2}{m^2}\log \frac{m^2}{{\mu }_{\gamma }^2}+(\text{single-log and finite pieces})\right].$$

The leading IR divergence is the Sudakov double log $\log (-q^2/m^2)\log (m^2/{\mu }_{\gamma }^2)$ — one $\log$ from the soft region of the loop momentum, one $\log$ from the collinear region along the external charged-particle directions. The wave-function renormalisation $Z_2$ from 12.16.01 contributes an additional single $\log (m^2/{\mu }_{\gamma }^2)$ piece that combines with the vertex divergence to give the net one-loop infrared structure. The correction to the elastic cross section is

$$\frac{{\sigma }_{\text{elastic, 1-loop}}}{{\sigma }_{\text{tree}}}-1=-\frac{\alpha }{\pi }\left[B(\theta )\log \frac{m^2}{{\mu }_{\gamma }^2}+(\text{finite})\right],$$

with $B(\theta )$ a positive scattering-geometry-dependent function — the Bloch-Nordsieck $B$-coefficient. The divergent piece is the issue: as ${\mu }_{\gamma }\to 0$ the one-loop correction diverges and the perturbative expansion is ill-defined.

(2) The real soft-emission piece. The same elastic process, with an additional final-state photon of momentum $k$, polarisation $ϵ$, and energy $\omega =|\mathbf{k}|$ small compared to all hard scales, factorises in the eikonal limit (Low 1958; Peskin-Schroeder §6.1 ^{[Peskin-Schroeder 1995]}):

$${\mathcal{M}}_{\text{brems}}(p_1\to p_1^{\prime }+k)={\mathcal{M}}_{\text{hard}}(p_1\to p_1^{\prime })[e(ϵ\cdot p_1^{\prime })/(k\cdot p_1^{\prime })-e(ϵ\cdot p_1)/(k\cdot p_1)]+\mathcal{O}({\omega }^0),$$

with the eikonal soft factor identical to the classical Lienard-Wiechert radiation pattern from an instantaneous deflection $p_1\to p_1^{\prime }$. The bremsstrahlung cross section in the soft region, summed over photon polarisations and integrated over photon directions, is

$$d{\sigma }_{\text{brems, soft}}=d{\sigma }_{\text{tree}}\frac{\alpha }{2{\pi }^2}\frac{d^{3}k}{2\omega }\sum _{\text{pol}}{\left|\frac{ϵ\cdot p_1^{\prime }}{k\cdot p_1^{\prime }}-\frac{ϵ\cdot p_1}{k\cdot p_1}\right|}^2.$$

Carrying out the polarisation sum and the angular integral, then integrating $\omega$ from ${\mu }_{\gamma }$ (regulator) to $\Delta E$ (detector cut), gives

$${\sigma }_{\text{brems}}(\omega <\Delta E)={\sigma }_{\text{tree}}\frac{\alpha }{\pi }B(\theta )\log \frac{\Delta E^2}{{\mu }_{\gamma }^2}+\mathcal{O}({\sigma }_{\text{tree}}(\Delta E/E{)}^2),$$

with the same coefficient $B(\theta )$ as in the virtual one-loop correction. (This is the structural content of the Bloch-Nordsieck cancellation: the virtual and real soft-emission contributions have the same coefficient times the divergent log, with opposite signs.)

(3) The cancellation. Add the elastic-with-virtual-correction and the soft-bremsstrahlung pieces:

$${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}+\delta {\sigma }_{\text{elastic, 1-loop}}+{\sigma }_{\text{brems}}(\omega <\Delta E)={\sigma }_{\text{tree}}\left[1-\frac{\alpha }{\pi }B(\theta )\log \frac{m^2}{{\mu }_{\gamma }^2}+\frac{\alpha }{\pi }B(\theta )\log \frac{\Delta E^2}{{\mu }_{\gamma }^2}+\cdots \right].$$

The two $\log ({\mu }_{\gamma }^2)$ pieces cancel:

$$-\log \frac{m^2}{{\mu }_{\gamma }^2}+\log \frac{\Delta E^2}{{\mu }_{\gamma }^2}=\log \frac{\Delta E^2}{m^2}=2\log \frac{\Delta E}{m}.$$

After substituting and using $\log (\Delta E/m)=\log (\Delta E/E)+\log (E/m)$ (where $E$ is the characteristic hard scale of the process), the residual ${\mu }_{\gamma }$-independent term gives

$${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}\left[1+\frac{2\alpha }{\pi }B(\theta )\log \frac{\Delta E}{E}+(\text{finite, geometry-dependent terms})\right],$$

with ${\mu }_{\gamma }$ absent at this order. This is the $O(\alpha )$ form of the Bloch-Nordsieck theorem.

(4) Higher orders and exponentiation. The same cancellation happens at every order in $\alpha$. At $O({\alpha }^2)$, the two-loop virtual correction (containing ${\log }^2({\mu }_{\gamma })$ Sudakov double logs) cancels against the two-soft-photon bremsstrahlung (also ${\log }^2({\mu }_{\gamma })$). At $O({\alpha }^n)$, the $n$-loop virtual correction cancels against the $n$-soft-photon bremsstrahlung, with the leading ${\log }^n({\mu }_{\gamma })$ pieces matching exactly. The combinatorial structure of the cancellation (Yennie-Frautschi-Suura 1961 ^{[Yennie-Frautschi-Suura 1961]}) exponentiates the leading-log dependence:

$${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}\exp \left[\frac{2\alpha }{\pi }B(\theta )\log \frac{\Delta E}{E}\right][1+\mathcal{O}(\alpha )]={\sigma }_{\text{tree}}{\left(\frac{\Delta E}{E}\right)}^{(2\alpha /\pi )B(\theta )}[1+\mathcal{O}(\alpha )],$$

the all-orders form announced. The $\mathcal{O}(\alpha )$ correction inside the brackets contains the non-leading-log finite pieces of the inclusive cross section at one loop and is computed by the standard QED perturbative machinery. $\square$

Counterexamples to common slips

• The Bloch-Nordsieck cancellation works for inclusive observables. The elastic cross section by itself — ${\sigma }_{\text{elastic}}({\mu }_{\gamma })$ with all real photon emission forbidden — diverges as ${\mu }_{\gamma }\to 0$ and is not a physical observable. The operational redefinition of the cross section to include unresolved soft emission is mandatory, not optional.

• The cancellation is term-by-term in $\alpha$ but not within a single Feynman diagram. The $n$-loop virtual amplitude is IR-divergent on its own; the $n$-real-soft-photon cross section is IR-divergent on its own; their sum is finite. There is no diagrammatic local cancellation; the cancellation is between distinct diagrams of distinct topology and distinct final-state particle content.

• The "photon mass" ${\mu }_{\gamma }$ is a regulator, not a physical mass. Equivalent IR regulators (dimensional regularisation with $d=4+2{\epsilon }_{\text{IR}}$, off-shell external legs, analytic regularisation) all give the same physical inclusive cross section; only intermediate expressions look different. The photon-mass regulator is conceptually clean for the Bloch-Nordsieck cancellation but breaks gauge invariance at intermediate stages; dimensional regularisation preserves gauge invariance and is the modern standard for explicit calculations.

• For massless final-state charged particles (or for processes at energies far above the charged-particle mass), additional collinear singularities appear from photons emitted exactly along a charged-particle direction. The Bloch-Nordsieck cancellation as originally stated does not handle these; the Kinoshita-Lee-Nauenberg theorem (Kinoshita 1962 ^{[Kinoshita 1962]}; Lee-Nauenberg 1964 ^{[Lee-Nauenberg 1964]}) extends the cancellation by also summing over collinear final-state configurations, which corresponds operationally to the definition of a jet.

• The Sudakov double log $\alpha {\log }^2(s/m^2)$ that dominates at high energies is not part of the Bloch-Nordsieck IR-divergent structure; it is a finite, large logarithm that survives the cancellation and contributes a calculable enhancement to the inclusive cross section. The Bloch-Nordsieck theorem says the divergent pieces cancel; the surviving large logs are physical and must be resummed (Sudakov 1956 ^{[Sudakov 1956]}) for quantitative predictions at very high energies.

Exercises Intermediate+

Advanced results Master

The Bloch-Nordsieck-Yennie-Frautschi-Suura programme is the entry point to a structurally rich body of results that has shaped quantum field theory from 1937 to the present. The post-1960s developments fall into four streams: the all-orders soft-photon exponentiation theorems, the high-energy Sudakov resummation, the non-Abelian extension required for QCD, and the modern asymptotic-symmetry reinterpretation.

Faddeev-Kulish asymptotic states. Faddeev and Kulish 1970 (Theor. Math. Phys. 4, 745) showed that the IR divergences of QED can be moved entirely out of the inclusive cross section and into a redefinition of the asymptotic states. Replace the standard Fock-space asymptotic electron $|e^{-}(p)⟩$ by a dressed state $|e^{-}(p){⟩}_{\text{coh}}=D(p)|e^{-}(p)⟩$ with $D(p)$ a coherent-state displacement operator carrying an infinite number of arbitrarily soft photons with momenta correlated with $p$. The dressed S-matrix between dressed asymptotic states is IR-finite without any cutoff, and the Bloch-Nordsieck cancellation in the inclusive cross section is interpreted as the unitarity of the dressed S-matrix. The coherent-state dressing carries finite energy and momentum but infinite photon number, which is exactly what the physical asymptotic states of QED look like. Kibble 1968-1969 (J. Math. Phys. 9, 315; Phys. Rev. 173, 1527) had independently developed the same formalism in the framework of coherent states. The Faddeev-Kulish formulation is the structurally cleanest statement of the IR structure of QED but is computationally cumbersome compared to the inclusive-cross-section approach, and the bulk of practical IR calculations still use the Bloch-Nordsieck-YFS framework.

Non-Abelian eikonal exponentiation. The Yennie-Frautschi-Suura exponentiation generalises to non-Abelian gauge theories with modifications. In a non-Abelian theory the soft-gluon emissions from different external legs no longer commute (the colour-charge generators $T^a$ at different legs do not commute), and the exponentiation involves colour-coherence effects. Frenkel-Taylor 1984 (Nucl. Phys. B246, 231 ^{[Frenkel-Taylor 1984]}) established the all-orders non-Abelian eikonal exponentiation: the leading soft-gluon emissions from a fixed external-line configuration exponentiate, but the exponent is not the simple Yennie-Frautschi-Suura sum over individual-line contributions; it involves additional terms from colour-correlated emissions. Catani-Marchesini-Webber 1991-1992 (Nucl. Phys. B349, 635; B370, 310) developed the colour-coherent parton shower as the algorithmic implementation of the non-Abelian exponentiation, the structural foundation of all modern parton-shower Monte Carlo generators (PYTHIA, HERWIG, SHERPA).

Sudakov resummation and the QCD parton shower. The Sudakov form factor of Exercise 6 above generalises to QCD, where the leading double log is ${\alpha }_s{\log }^2(Q^2/{\mu }_F^2)$ for a hard-scattering process at scale $Q^2$ with a factorisation scale ${\mu }_F$. Collins-Soper-Sterman 1985 (Nucl. Phys. B250, 199) developed the resummation framework for transverse-momentum distributions of Drell-Yan dileptons; Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) 1972-1977 (Sov. J. Nucl. Phys. 15, 438; Sov. Phys. JETP 46, 641; Nucl. Phys. B126, 298) developed the parton-distribution-evolution equations that resum the leading-log collinear emissions from initial-state partons. The interplay of Sudakov resummation and DGLAP evolution gives the modern picture of high-energy QCD as a hierarchy of factorisation scales, with hard-scattering cross sections, parton distribution functions, jet shapes, and fragmentation functions all controlled by the resummed leading and subleading logs of the relevant scale ratios. The convergence of the Sudakov-resummed cross section to the exact result at NLO and NNLO is one of the most extensively tested predictions of QCD at the LHC.

Modern asymptotic symmetries. Strominger 2014-2018 (in Lectures on the Infrared Structure of Gravity and Gauge Theory ^{[Strominger 2018]}) reinterpreted the soft-photon and soft-graviton theorems as Ward identities for asymptotic symmetries of QED and gravity at null infinity. The Low 1958 / Weinberg 1965 soft-photon theorem, ${\mathcal{M}}_{n+1}^{\text{soft}}(k)={\mathcal{M}}_n\cdot S^{(0)}(k)+S^{(1)}(k){\mathcal{M}}_n+\mathcal{O}(k)$, is the Ward identity for an infinite-dimensional symmetry group of large gauge transformations — gauge transformations that do not vanish at null infinity. The Bloch-Nordsieck cancellation in this language is the statement that physical asymptotic states must be invariant (or covariantly transform) under the large-gauge-transformation group; the Faddeev-Kulish dressing is the explicit construction of those invariant states. The same framework generalises to gravity, where the Bondi-Metzner-Sachs 1962 supertranslation group plays the role of the QED large-gauge group, the soft-graviton theorem (Weinberg 1965, Phys. Rev. 140, B516) is the supertranslation Ward identity, and the gravitational memory effect (Zel'dovich-Polnarev 1974, Sov. Astron. 18, 17; Strominger-Zhiboedov 2014, JHEP 01, 086) is the classical manifestation of the soft-graviton emission. This modern asymptotic-symmetry reading of Bloch-Nordsieck is conceptually deep but has not yet led to practical computational improvements over the inclusive-cross-section approach for terrestrial-physics observables.

Jet algorithms and IR safety at colliders. The Sterman-Weinberg 1977 ^{[Sterman-Weinberg 1977]} paper introduced the first IR-safe definition of a jet — a collimated cone of energy in a fixed angular size $\delta$ containing all but a fraction $ϵ$ of the total event energy in two back-to-back hemispheres. The Sterman-Weinberg cross section is IR-finite by construction. Modern jet algorithms — the $k_t$ algorithm (Catani-Dokshitzer-Seymour-Webber 1991, Nucl. Phys. B377, 295), the anti-$k_t$ algorithm (Cacciari-Salam-Soyez 2008, JHEP 04, 063), the SISCone algorithm (Salam-Soyez 2007, JHEP 05, 086) — generalise Sterman-Weinberg to sequential clustering procedures that satisfy IR safety (cross sections finite in the perturbative expansion to all orders) and infrared and collinear safety (cross sections insensitive to the addition of soft particles or to collinear splittings of existing particles). The anti-$k_t$ algorithm is the LHC standard and is used in every ATLAS and CMS jet measurement. The IR-and-collinear-safety property is the operational expression of the Kinoshita-Lee-Nauenberg theorem, and the modern jet-algorithm zoology is the practical implementation of the KLN cancellation in the high-multiplicity hadron-collider environment.

Bloch-Nordsieck violation in non-Abelian theories. A subtle feature of the non-Abelian case: the standard Bloch-Nordsieck theorem, summing only over final-state soft-gluon configurations at fixed initial state, fails in QCD when there are coloured initial-state particles whose colour configurations can mix under soft-gluon exchange. Doria-Frenkel-Taylor 1980 (Nucl. Phys. B168, 93) constructed an explicit two-loop counterexample to the QCD-naive Bloch-Nordsieck theorem at the inclusive Drell-Yan cross section, with an uncanceled IR divergence at $O({\alpha }_s^2)$ from initial-state colour interference. The resolution requires initial-state colour averaging (which is automatic in the hadron-collider context where the initial colour is averaged by the colour-singlet nature of the hadrons) and a more refined statement of the Bloch-Nordsieck theorem that explicitly accounts for initial-state colour structure. The Doria-Frenkel-Taylor counterexample was the conceptual entry point to the modern understanding of factorisation-breaking effects in QCD and is a subtle technical issue that affects only a small class of observables; the Catani-Seymour 1997 ^{[Catani-Seymour 1997]} dipole subtraction algorithm and its successors handle initial-state colour averaging automatically and produce IR-finite NLO QCD cross sections for all practical hadron-collider observables.

Connection to renormalisation: the running coupling at low scales. The same IR structure that produces the Bloch-Nordsieck cancellation also controls the running of the QED coupling at low scales. The leading-log running $\alpha (Q^2)=\alpha (0)/[1-(\alpha (0)/3\pi )\log (Q^2/m_e^2)]$ from vacuum polarisation (12.16.03) breaks down at very low $Q^2\ll m_e^2$, where the infrared renormalisation-group flow becomes important. The conceptual structure of the low-$Q^2$ running of QED — the Landau pole in the infrared rather than the better-known Landau pole in the ultraviolet — has been the subject of theoretical reanalyses by Wilson 1971 (Phys. Rev. B4, 3174) and by Polchinski 1983 (Nucl. Phys. B231, 269). The IR structure is intimately related to the asymptotic states and the Bloch-Nordsieck cancellation, since the asymptotic charged states of QED are precisely those that survive the low-$Q^2$ renormalisation-group flow.

Full proof set Master

The Key derivation supplies the explicit Bloch-Nordsieck cancellation at $O(\alpha )$ and sketches the all-orders exponentiation. The following auxiliary results are stated with proof outlines verifiable against the cited literature.

Lemma 1 (Eikonal factorisation of soft-photon emission). For any QED scattering amplitude ${\mathcal{M}}_n$ with $n$ external charged-particle and photon momenta, the amplitude ${\mathcal{M}}_{n+1}(p_1,\dots ,p_n;k)$ for the same process with an additional final-state photon of momentum $k$ and polarisation $ϵ$ factorises in the soft limit $\omega =|k^0|\to 0$ as ${\mathcal{M}}_{n+1}^{\text{soft}}(k)={\mathcal{M}}_n\cdot e{\sum }_i{\eta }_i{Q}_i(ϵ\cdot p_i)/(k\cdot p_i)+\mathcal{O}({\omega }^0)$, where the sum runs over external charged legs with charges $Q_i$ and incoming/outgoing signs ${\eta }_i$. Proof outline. The soft photon attaches to one of the external charged legs, sending the leg propagator $i(\text{slashed}{p}_i+m)/(p_i^2-m^2+iϵ)$ off-shell to $i(\text{slashed}{p}_i\mp \text{slashed}k+m)/((p_i\mp k{)}^2-m^2+iϵ)$, where the sign depends on incoming/outgoing. Expanding in small $k$: the propagator denominator becomes $\mp 2k\cdot p_i+\mathcal{O}(k^2)$ and the numerator (after on-shell projector application) is dominated by $2mϵ\cdot p_i/(\mp 2k\cdot p_i)=\mp ϵ\cdot p_i/(k\cdot p_i)$ in the soft limit, with the overall $-ie{\gamma }^{\mu }{ϵ}_{\mu }$ vertex factor giving the eikonal soft factor. Internal photon attachments are subleading in $\omega$. This is the Low 1958 soft-photon theorem; the proof is given in Peskin-Schroeder §6.1 ^{[Peskin-Schroeder 1995]} and in Weinberg Vol I §13 ^{[Weinberg 1995]}. $\square$

Lemma 2 (Soft-photon phase-space measure). The Lorentz-invariant phase-space measure for a single emitted soft photon of energy $\omega$, integrated over the photon direction $\hat{k}$, gives $d^{3}k/[(2\pi {)}^{3}2\omega ]=\omega d\omega d{\Omega }_k/(16{\pi }^3)$. The combined eikonal-squared-amplitude and phase-space integral over the soft region ${\mu }_{\gamma }<\omega <\Delta E$ produces the universal IR-log structure ${\sigma }_{\text{soft brems}}={\sigma }_{\text{tree}}\cdot (\alpha /\pi )\mathcal{B}(\theta )\log (\Delta E^2/{\mu }_{\gamma }^2)$. Proof outline. The polarisation sum ${\sum }_{\text{pol}}{ϵ}_{\mu }^{\ast }{ϵ}_{\nu }=-g_{\mu \nu }$ (Feynman gauge with photon mass regulator) gives a $-1/(k\cdot p_i)(k\cdot p_j)$ structure in the squared eikonal amplitude. The energy integral $\int d\omega \cdot {\omega }^2/{\omega }^2=\int d\omega /\omega$ produces the $\log (\Delta E/{\mu }_{\gamma })$ piece, while the angular integral $\int d{\Omega }_k/[4\pi (k\cdot p_i)(k\cdot p_j)/{\omega }^2]$ produces the angular geometric factor ${\mathcal{B}}_{ij}(\theta )$ that depends on the angle between $p_i$ and $p_j$. The total $\mathcal{B}(\theta )={\sum }_{ij}{Q}_i{Q}_j{\mathcal{B}}_{ij}(\theta )$ is positive (by general arguments about radiated energy being positive) and matches exactly the $B(\theta )$ coefficient appearing in the virtual one-loop $\log (m^2/{\mu }_{\gamma }^2)$ piece. Detailed angular evaluation: Peskin-Schroeder §6.4 ^{[Peskin-Schroeder 1995]} eqs. 6.93-6.95. $\square$

Lemma 3 (Term-by-term cancellation of $\log ({\mu }_{\gamma })$ at one loop). The one-loop virtual correction to the elastic cross section contains a piece $-{\sigma }_{\text{tree}}\cdot (\alpha /\pi )B(\theta )\log (m^2/{\mu }_{\gamma }^2)$ from the IR-divergent vertex and wave-function renormalisation integrals. The one-soft-photon bremsstrahlung cross section integrated to $\omega <\Delta E$ contains a piece $+{\sigma }_{\text{tree}}\cdot (\alpha /\pi )B(\theta )\log (\Delta E^2/{\mu }_{\gamma }^2)$ from Lemma 2. The sum is ${\sigma }_{\text{tree}}\cdot (\alpha /\pi )B(\theta )\log (\Delta E^2/m^2)={\sigma }_{\text{tree}}\cdot (2\alpha /\pi )B(\theta )\log (\Delta E/m)$, finite as ${\mu }_{\gamma }\to 0$. Proof outline. The cancellation requires the same coefficient $B(\theta )$ on the virtual and real sides; this matching is the structural content of the Bloch-Nordsieck theorem. The proof is by direct computation of both sides using Lemmas 1 and 2 and the explicit IR-divergent piece of the one-loop vertex form factor of 12.16.02; the matching coefficient is automatic, given the universality of the eikonal soft factor. (Detailed proof: Bloch-Nordsieck 1937 ^{[Bloch-Nordsieck 1937]} for the non-covariant original; Peskin-Schroeder §6.4-§6.5 ^{[Peskin-Schroeder 1995]} for the modern covariant version; Weinberg Vol I §13.2 ^{[Weinberg 1995]} for the abstract-S-matrix-axiomatic version.) $\square$

Lemma 4 (Yennie-Frautschi-Suura all-orders exponentiation). The Bloch-Nordsieck cancellation at $O({\alpha }^n)$ between $n$-loop virtual corrections and $n$-soft-photon bremsstrahlung produces, after summing the leading ${\log }^n(\Delta E/E)$ pieces over all $n$, the exponentiated cross section ${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}\cdot (\Delta E/E{)}^{(2\alpha /\pi )B(\theta )}\cdot [1+\mathcal{O}(\alpha )]$. Proof outline. The combinatorial structure of the cancellation has the cluster-decomposition property: a single coherent soft-photon emission cloud factorises from the hard process, and the multi-soft-photon emission is the exponential of the single-soft-photon emission. The mathematical statement is that the connected $n$-soft-photon emission diagrams contribute $(1/n!)[B(\theta )\log (\Delta E/{\mu }_{\gamma }){]}^n$ to the cross-section ratio at $O({\alpha }^n)$, and the sum over $n$ exponentiates. The matching virtual $n$-loop calculation gives $(-1{)}^n/n![B(\theta )\log (m^2/{\mu }_{\gamma }^2){]}^n$, and the combined sum is $\exp [(2\alpha /\pi )B(\theta )\log (\Delta E/m)]=(\Delta E/m{)}^{(2\alpha /\pi )B(\theta )}$. The ${\mu }_{\gamma }$ dependence cancels at every order. Detailed proof: Yennie-Frautschi-Suura 1961 ^{[Yennie-Frautschi-Suura 1961]} for the original construction; Weinberg Vol I §13.4 ^{[Weinberg 1995]} for the modern abstract version. $\square$

Lemma 5 (Sudakov double-log resummation for the form factor). The electron electromagnetic form factor at large spacelike momentum transfer $Q^2=-q^2\gg m^2$ satisfies the Sudakov exponential $F_1(-Q^2{)}^{\text{Sudakov LL}}=\exp [-(\alpha /4\pi ){\log }^2(Q^2/m^2)]$. Proof outline. As in Exercise 6 above, the $n$-loop leading-double-log piece of the form factor is $(-1{)}^n(\alpha /4\pi {)}^n(1/n!){\log }^{2n}(Q^2/m^2)$, with the $1/n!$ from the time-ordering combinatorics of multiple virtual soft photons inserted between the external legs. Summing ${\sum }_n(-{)}^n/n![(\alpha /4\pi ){\log }^2(Q^2/m^2){]}^n=\exp [-(\alpha /4\pi ){\log }^2(Q^2/m^2)]$. The result was first derived by Sudakov 1956 ^{[Sudakov 1956]} from the eikonal approximation to the multi-loop integrals; modern proofs use the Collins-Soper-Sterman renormalisation-group formalism or the SCET (soft-collinear effective theory) framework. (Detailed proof: Collins-Soper-Sterman 1985 Nucl. Phys. B250, 199; Becher-Neubert 2009 Phys. Rev. Lett. 102, 162001 for the modern SCET treatment.) $\square$

Lemma 6 (Kinoshita-Lee-Nauenberg cancellation for massless charged particles). For a process with massless final-state charged particles, the cross section summed over all degenerate final states (soft + collinear) is IR-finite to all orders in the gauge coupling. The corresponding IR-safe observable is a jet — a collimated bundle of energy in a fixed angular cone, defined operationally by the Sterman-Weinberg ^{[Sterman-Weinberg 1977]} criterion or its modern descendants (anti-$k_t$, $k_t$, SISCone). Proof outline. The Bloch-Nordsieck cancellation handles the soft singularity from $\omega \to 0$; the additional collinear singularity from $\hat{k}\parallel \hat{p}$ requires summing over photon (gluon) configurations that are collinear with a charged-particle direction. The KLN theorem states: the cross section summed over all soft + collinear degenerate final states is finite. The Kinoshita 1962 ^{[Kinoshita 1962]} proof uses dimensional regularisation and the structure of the propagators in the collinear region; the Lee-Nauenberg 1964 ^{[Lee-Nauenberg 1964]} proof is more general and applies to any unitary theory with degenerate states. Modern applications: Catani-Seymour 1997 ^{[Catani-Seymour 1997]} dipole subtraction; NLO Monte Carlo generators (MCFM, MadGraph_aMC@NLO, POWHEG). $\square$

Lemma 7 (Sterman-Weinberg jet cross section as IR-safe observable). The Sterman-Weinberg 1977 ^{[Sterman-Weinberg 1977]} cross section ${\sigma }_{2\text{-jet}}(ϵ,\delta )$ for two-jet events in $e^{+}{e}^{-}\to$ hadrons, with $ϵ$ the fraction of total energy permitted outside the two jet cones of half-angle $\delta$, is IR-finite for any $ϵ,\delta >0$ and has the leading-log behaviour ${\sigma }_{2\text{-jet}}(ϵ,\delta )/{\sigma }_{\text{tot}}=1-(4{\alpha }_s/3\pi )\log (1/ϵ)\log (1/{\delta }^2)+\mathcal{O}({\alpha }_s^2)$. Proof outline. The KLN sum over degenerate states for the $e^{+}{e}^{-}\to q\bar{q}$ at NLO contains soft, collinear-to-quark, and collinear-to-antiquark emissions. The Sterman-Weinberg cone definition restricts the non-jet energy to less than $ϵE$ and the out-of-cone angular range to outside $\delta$. The IR-divergent pieces from soft and collinear emissions inside the cones are absorbed into the jet definition; the residual finite piece is the leading-log $\log (1/ϵ)\log (1/{\delta }^2)$ coefficient. The structural similarity to the Sudakov double log is not accidental: both arise from the soft + collinear region of the bremsstrahlung phase-space integral. (Detailed proof: Sterman-Weinberg 1977 ^{[Sterman-Weinberg 1977]}; Ellis-Stirling-Webber, QCD and Collider Physics, Cambridge 1996, Ch. 3.) $\square$

Connections Master

• 12.16.01 Electron self-energy and mass renormalisation at one loop supplies the IR-divergent wave-function renormalisation $Z_2$ that is one half of the Bloch-Nordsieck cancellation. The same $\log (m/{\mu }_{\gamma })$ structure that appears in $Z_2$ appears in the bremsstrahlung integral, with opposite sign, and the cancellation in the inclusive cross section closes the loop.

• 12.16.02 One-loop QED vertex function and the anomalous magnetic moment supplies the IR-divergent $F_1(q^2)$ form factor whose Sudakov double log $\log (m/{\mu }_{\gamma })\log (-q^2/m^2)$ is the canonical example of the IR structure cancelled by Bloch-Nordsieck. The finite anomalous moment $F_2(0)=\alpha /(2\pi )$ is IR-finite on its own and is not affected by the cancellation; only $F_1$ requires the inclusive prescription to be physically meaningful.

• 12.12.01 Canonical quantum field theory supplies the Feynman rules for QED that produce the loop integrals analysed here, and the asymptotic-state framework whose modification by soft-photon dressing (Faddeev-Kulish 1970) is the structurally clean alternative to the Bloch-Nordsieck inclusive-cross-section prescription.

• 12.16.04 Lamb shift from one-loop QED uses a different mechanism — the atomic-scale matrix-element structure of the bound state, rather than the soft-photon detector cut — to render the bound-state self-energy IR-finite. The Bethe logarithm $\log k_0(n,\ell )$ plays the role of the detector resolution $\Delta E$ in the inclusive-cross-section framework; the same IR-divergent free-electron self-energy that drives the Lamb-shift calculation is the IR-divergent piece that Bloch-Nordsieck cancels in scattering processes.

• [12.16.06 — pending] Two-loop QED radiative corrections extends the one-loop Bloch-Nordsieck cancellation to second order in $\alpha$. The two-loop virtual amplitudes have ${\log }^2({\mu }_{\gamma })$ Sudakov double logs that cancel against the two-soft-photon bremsstrahlung; the combinatorial structure of the cancellation is the entry point to the all-orders Yennie-Frautschi-Suura exponentiation.

• [12.12.06 — pending] Bethe-Heitler bremsstrahlung and pair production supplies the high-energy bremsstrahlung cross section that interpolates between the soft-photon limit (where Bloch-Nordsieck applies) and the hard-photon limit (where the Bethe-Heitler formula gives the explicit $\omega$-dependent emission spectrum). The radiation-length parameter $X_0$ that characterises bremsstrahlung in matter is the integrated bremsstrahlung cross section over all photon energies, weighted by the eikonal soft factor in the soft region.

• 10.07.01 Larmor formula (electromagnetism) is the classical limit of the soft-photon emission rate from an accelerated charge. The Bloch-Nordsieck $B(\theta )$ coefficient is the angular integral of the Larmor radiation pattern, weighted by the quantum density of states. The classical-quantum correspondence between Larmor radiation and bremsstrahlung in the soft limit is the structural content of the soft-photon theorem (Low 1958; Weinberg 1965).

• 12.16.03 Vacuum polarisation at one loop and the Uehling potential supplies the running of the QED coupling that controls the high-energy structure of the Sudakov form factor. The leading-log running of $\alpha$ between ${\mu }_{\gamma }$ and $\Delta E$ contributes a subleading correction to the YFS exponential at the $O({\alpha }^2)$ level.

• 12.11.01 Dirac equation and relativistic spin supplies the on-shell projector $\text{slashed}p-m$ that appears in the eikonal factorisation of the soft-photon emission amplitude; the projector structure is the technical reason the soft factor is universal.

• [13.06.05 — pending] Bondi-Metzner-Sachs supertranslations and the soft graviton theorem (general relativity) is the gravitational analogue of the QED IR structure analysed here. The soft-graviton theorem of Weinberg 1965 is the BMS supertranslation Ward identity, and the gravitational analogue of the Bloch-Nordsieck cancellation operates in graviton-emission cross sections in gravitational scattering. The modern asymptotic-symmetry reading of both QED and gravity IR structure (Strominger 2017 ^{[Strominger 2018]}) unifies them under a single conceptual framework.

Historical & philosophical context Master

The Bloch-Nordsieck cancellation is the oldest of the IR-divergence cancellation theorems in quantum field theory, predating the Schwinger-Tomonaga-Feynman-Dyson UV renormalisation programme of 1947-1949 by a decade. Felix Bloch (1905-1983) and Arnold Nordsieck (1911-1971), then both at Stanford, were studying the long-known infrared catastrophe of quantum electrodynamics: the observation, originally made in heuristic computations by Heisenberg and Pauli 1929-1930 (Zeits. f. Physik 56, 1; 59, 168) and made explicit by Mott 1931 (Proc. Roy. Soc. A 135, 429) and by Heitler 1937 (Proc. Camb. Phil. Soc. 33, 277), that the cross section for electron scattering with no photon emission diverges logarithmically when the photon-emission threshold is taken to zero. The "catastrophe" was conceptually puzzling: classical electrodynamics also has soft-photon emission (the Lienard-Wiechert radiation pattern), but the integrated radiated energy is finite for any finite acceleration; the divergence appears only in the quantum theory, in the probability of finding the scattered electron with no photon companion.

Bloch and Nordsieck's 1937 paper (Phys. Rev. 52, 54 ^{[Bloch-Nordsieck 1937]}) is a brief — eight pages — but conceptually decisive contribution. They observed that the relevant physical observable is not the probability of "no photon emission" (which is in fact zero in QED for any nonzero acceleration) but the probability of emission of photons with total energy below a detector-resolution threshold $\Delta E$. They computed this inclusive probability in non-covariant perturbation theory and found that it is finite and depends only on the physical ratio $\Delta E/E$. The cancellation works because the same soft-photon coupling that makes the photon-emission rate diverge in the soft limit also makes the photon-non-emission probability vanish — and the two divergences are precisely of the form to cancel. The Bloch-Nordsieck paper is one of the earliest examples in physics of a reorganisation of observables to extract a finite prediction from an apparently divergent perturbative expansion, a strategy that later became the conceptual backbone of the renormalisation programme.

The 1937 Bloch-Nordsieck calculation was non-covariant — the modern Dirac-equation-and-covariant-perturbation-theory framework was not yet established — and the all-orders extension was unclear. Pauli and Fierz 1938 (Nuovo Cim. 15, 167 ^{[Pauli-Fierz 1938]}) generalised the argument to the long-wavelength radiation field of a charged particle and showed that the asymptotic states of QED carry a coherent soft-photon dressing, a result that anticipated the Faddeev-Kulish 1970 formulation by three decades but was not pursued at the time. The next major step came after the 1947-1949 covariant-renormalisation programme: Schwinger 1949 (Phys. Rev. 76, 790) noticed that the IR divergences of the one-loop QED vertex correction match in coefficient with the soft bremsstrahlung in covariant form, confirming the Bloch-Nordsieck cancellation in the language of covariant perturbation theory. Berestetskii-Lifshitz-Pitaevskii 1953 / 1980 (LL Vol 4 §§118-120 ^{[Berestetskii-Lifshitz-Pitaevskii 1982]}) gave the explicit process-by-process matching for elastic electron-nucleus scattering and for $e^{+}{e}^{-}\to$ exclusive hadronic processes.

Yennie, Frautschi, and Suura 1961 (Ann. Phys. 13, 379 ^{[Yennie-Frautschi-Suura 1961]}) completed the all-orders covariant exponentiation. Their paper showed that the leading soft-log dependence at every order in $\alpha$ exponentiates to give a $(\Delta E/E{)}^{B(\theta )\cdot 2\alpha /\pi }$ correction to the tree-level cross section, and that this exponentiation is universal — independent of the specific scattering process and depending only on the external-charge configuration. The YFS exponentiation is the bound between the Bloch-Nordsieck theorem (an existence statement: the cancellation holds order by order) and the practical computation: with the exponentiated form in hand, one can compute the inclusive cross section to arbitrary precision in $\alpha$ without ever encountering the IR regulator.

The 1962-1964 Kinoshita-Lee-Nauenberg theorem (Kinoshita 1962, J. Math. Phys. 3, 650 ^{[Kinoshita 1962]}; Lee-Nauenberg 1964, Phys. Rev. 133, B1549 ^{[Lee-Nauenberg 1964]}) extended the Bloch-Nordsieck cancellation to massless charged particles, where collinear singularities appear in addition to soft singularities. The KLN theorem was conceptually decisive for the development of quantum chromodynamics in the 1970s: in QCD, the asymptotic states are confined hadrons rather than free coloured partons, but the perturbative QCD calculations work at the parton level, and the IR-and-collinear-safe-observable framework supplied by KLN is what makes parton-level perturbative QCD compatible with the hadronic-observable phenomenology. The Sterman-Weinberg 1977 (Phys. Rev. Lett. 39, 1436 ^{[Sterman-Weinberg 1977]}) jet definition and its descendants (anti-$k_t$, $k_t$, SISCone) are the algorithmic implementation of the KLN cancellation for hadron-collider observables.

The Faddeev-Kulish 1970 (Theor. Math. Phys. 4, 745) coherent-state formulation reinterprets the Bloch-Nordsieck cancellation as a redefinition of the asymptotic states of QED, with the physical electron carrying an infinite coherent cloud of arbitrarily soft photons. Their formulation makes the IR structure manifest at the level of the asymptotic states rather than at the level of the inclusive cross section, and is the structurally cleanest statement of the IR finiteness of QED. The recent reinterpretation of QED soft-photon theorems as Ward identities for asymptotic symmetries — the large gauge transformations of QED at null infinity, in the language of Strominger 2014-2018 ^{[Strominger 2018]} — places the Bloch-Nordsieck cancellation in a broader framework that connects it to the analogous structure of soft-graviton theorems in general relativity and to the Bondi-Metzner-Sachs 1962 supertranslation symmetry of asymptotically flat spacetimes.

The Bloch-Nordsieck cancellation is the founding example of infrared-safe observables in quantum field theory: physical quantities that are finite in perturbation theory without requiring any IR regulator and that depend only on physical scales. The conceptual content of the cancellation — that elastic scattering is not a well-defined observable in any massless-gauge-boson theory, and that only inclusive sums over arbitrarily soft and collinear emissions give finite predictions — is one of the deep features of gauge theories that the renormalisation programme of the 1940s did not anticipate. The IR structure of QED, QCD, and gravity continues to be an active area of research seventy years after Bloch-Nordsieck, with the modern asymptotic-symmetry reading providing a unifying conceptual framework that places these theorems in a broader mathematical context.

Bibliography Master

Primary literature:

• Bloch, F. & Nordsieck, A. Note on the radiation field of the electron. Phys. Rev. 52, 54 (1937). The originating cancellation theorem: the inclusive cross section for electron scattering, summed over soft-photon emission below a detector resolution $\Delta E$, is infrared-finite to all orders in $\alpha$.

• Pauli, W. & Fierz, M. Zur Theorie der Emission langwelliger Lichtquanten. Nuovo Cim. 15, 167 (1938). The long-wavelength radiation field of a charged particle as a coherent soft-photon dressing of the asymptotic states; conceptual precursor to the Faddeev-Kulish 1970 formulation.

• Mott, N. F. The polarisation of electrons by double scattering. Proc. Roy. Soc. A 135, 429 (1931). Early heuristic identification of the IR catastrophe in non-relativistic Coulomb scattering of electrons.

• Heitler, W. The influence of radiation damping on the scattering of electrons. Proc. Camb. Phil. Soc. 33, 277 (1937). Pre-Bloch-Nordsieck attempt to handle the IR catastrophe by introducing radiation-damping back-reaction; superseded by the inclusive-cross-section formulation.

• Schwinger, J. Quantum electrodynamics. III. The electromagnetic properties of the electron — radiative corrections to scattering. Phys. Rev. 76, 790 (1949). The covariant matching of the IR-divergent piece of the one-loop vertex correction to the soft bremsstrahlung cross section; confirms the Bloch-Nordsieck cancellation in covariant perturbation theory.

• Low, F. E. Bremsstrahlung of very low-energy quanta in elementary particle collisions. Phys. Rev. 110, 974 (1958). The soft-photon theorem in its modern covariant form: the eikonal factorisation ${\mathcal{M}}_{n+1}^{\text{soft}}={\mathcal{M}}_n\cdot S^{(0)}(k)+S^{(1)}(k){\mathcal{M}}_n+\mathcal{O}(k)$ with universal soft factors.

• Weinberg, S. Infrared photons and gravitons. Phys. Rev. 140, B516 (1965). The soft-photon and soft-graviton theorems in QED and gravity, including the generalisation of the Bloch-Nordsieck cancellation to gravity and the IR structure of low-energy graviton amplitudes.

• Sudakov, V. V. Vertex parts at very high energies in quantum electrodynamics. Sov. Phys. JETP 3, 65 (1956). The Sudakov double logarithm in the QED form factor at large spacelike $Q^2$; the all-orders resummation $\exp [-(\alpha /4\pi ){\log }^2(Q^2/m^2)]$.

• Yennie, D. R., Frautschi, S. C. & Suura, H. The infrared divergence phenomena and high-energy processes. Ann. Phys. (N.Y.) 13, 379 (1961). The all-orders covariant exponentiation of the Bloch-Nordsieck cancellation: ${\sigma }_{\text{incl}}(\Delta E)={\sigma }_{\text{tree}}(\Delta E/E{)}^{B(\theta )2\alpha /\pi }$.

• Kinoshita, T. Mass singularities of Feynman amplitudes. J. Math. Phys. 3, 650 (1962). The first generalisation of Bloch-Nordsieck to massless charged particles, with collinear singularities cancelled by summing over collinear-degenerate final states.

• Lee, T. D. & Nauenberg, M. Degenerate systems and mass singularities. Phys. Rev. 133, B1549 (1964). The full general statement of the Kinoshita-Lee-Nauenberg theorem: the cross section summed over all initial- and final-state degenerate configurations is IR-finite to all orders.

• Faddeev, L. D. & Kulish, P. P. Asymptotic conditions and infrared divergences in quantum electrodynamics. Theor. Math. Phys. 4, 745 (1970). The coherent-state asymptotic-state formulation of QED that makes the IR cancellation manifest at the level of the dressed asymptotic states.

• Sterman, G. & Weinberg, S. Jets from quantum chromodynamics. Phys. Rev. Lett. 39, 1436 (1977). The first IR-safe definition of a jet in QCD; the algorithmic implementation of the KLN cancellation for hadron-collider observables.

• Frenkel, J. & Taylor, J. C. Non-Abelian eikonal exponentiation. Nucl. Phys. B246, 231 (1984). The all-orders extension of the Yennie-Frautschi-Suura exponentiation to non-Abelian gauge theories with colour-coherence corrections.

• Catani, S. & Seymour, M. H. A general algorithm for calculating jet cross sections in NLO QCD. Nucl. Phys. B485, 291 (1997). The dipole-subtraction algorithm for IR-finite NLO QCD calculations; the modern implementation of the KLN cancellation in automated NLO Monte Carlo generators.

• Strominger, A. Lectures on the infrared structure of gravity and gauge theory. Princeton University Press (2018); arXiv:1703.05448. The modern asymptotic-symmetry reading of the soft-photon and soft-graviton theorems as Ward identities for large gauge transformations of QED and gravity at null infinity.

• Collins, J. C., Soper, D. E. & Sterman, G. Back-to-back jets in QCD. Nucl. Phys. B250, 199 (1985). The resummation framework for transverse-momentum distributions of Drell-Yan dileptons; one of the foundational results of modern Sudakov resummation in QCD.

• Doria, R., Frenkel, J. & Taylor, J. C. Counter-example to non-Abelian Bloch-Nordsieck theorem. Nucl. Phys. B168, 93 (1980). The two-loop counterexample to the naive Bloch-Nordsieck theorem in QCD; the entry point to the modern understanding of factorisation-breaking effects in QCD.

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. §98 (the infrared catastrophe) and §§118-120 (explicit Bloch-Nordsieck matching for elastic electron-nucleus scattering and for exclusive hadronic processes); the canonical physicist-process-driven treatment.

• Peskin, M. E. & Schroeder, D. V. An Introduction to Quantum Field Theory. Westview Press, 1995. §6.4 (infrared divergences and soft bremsstrahlung) and §6.5 (Sudakov double logarithm and the leading-log resummation); the standard pedagogical reference for the modern covariant Bloch-Nordsieck cancellation.

• Weinberg, S. The Quantum Theory of Fields, Vol. I: Foundations. Cambridge University Press, 1995. Ch. 13 (infrared effects), with the Bloch-Nordsieck theorem in its abstract S-matrix-axiomatic form, the YFS exponentiation, and the soft-photon-and-soft-graviton theorems.

• Schwartz, M. D. Quantum Field Theory and the Standard Model. Cambridge University Press, 2014. §20 (infrared divergences in QED and the cancellation theorem); modern pedagogical treatment with explicit dimensional-regularisation calculations.

• Itzykson, C. & Zuber, J.-B. Quantum Field Theory. McGraw-Hill, 1980. §8-3 (the Bloch-Nordsieck cancellation) and §8-4 (Sudakov form factor); the Franco-Russian peer to BLP with slightly more functional-integral emphasis.

• Sterman, G. An Introduction to Quantum Field Theory. Cambridge University Press, 1993. Ch. 13 (IR divergences and the KLN theorem); explicit treatment of QCD jet observables and parton showers.

• Collins, J. C. Foundations of Perturbative QCD. Cambridge University Press, 2011. Ch. 5-7 (IR divergences, factorisation, parton distribution functions); the modern reference for IR structure of QCD and Sudakov resummation in the SCET framework.

• Ellis, R. K., Stirling, W. J. & Webber, B. R. QCD and Collider Physics. Cambridge University Press, 1996. Ch. 3 (jet cross sections, Sterman-Weinberg definition, anti-$k_t$ and related modern algorithms); the standard hadron-collider phenomenology reference.

• Greiner, W. & Reinhardt, J. Quantum Electrodynamics, 4e. Springer, 2009. §6.4 (Bloch-Nordsieck cancellation with explicit one-loop computation for $e^{-}+Z\to e^{-}+Z$); the teaching companion that follows BLP's process-driven pedagogy.

• Weinberg, S. The Quantum Theory of Fields, Vol. II: Modern Applications. Cambridge University Press, 1996. §19 (asymptotic states and IR structure in non-Abelian gauge theories); the canonical reference for the QCD extension of Bloch-Nordsieck.