Case Study 2: Feynman Diagrams — Pictures That Changed Physics

Overview

In 1948, Richard Feynman introduced a pictorial method for computing scattering amplitudes in quantum electrodynamics. These "Feynman diagrams" transformed theoretical physics. What had been pages of opaque algebraic manipulations became simple pictures that even a graduate student could draw and compute. The diagrams encode deep physics — causality, locality, Lorentz invariance, unitarity — in a visual language that simultaneously aids calculation and physical intuition.

This case study examines how Feynman diagrams work, why they were revolutionary, and what they reveal about the structure of quantum field theory.

Part 1: Before Feynman — The Problem

The Computational Nightmare

Before Feynman diagrams, calculating scattering amplitudes in QED was extraordinarily tedious. Julian Schwinger, whose operator approach was mathematically equivalent to Feynman's, produced correct results — but his calculations were monuments of algebraic endurance that few physicists could follow.

Consider the simplest non-trivial QED process: electron-electron scattering at tree level (one virtual photon exchange). In Schwinger's formalism, this required:

1. Writing the interaction Hamiltonian density
2. Computing the second-order term in the Dyson series
3. Evaluating time-ordered products of field operators
4. Contracting creation and annihilation operators (Wick's theorem)
5. Performing the momentum-space integrals
6. Taking traces over Dirac matrices
7. Summing over polarizations

The result is a single formula — but arriving at it is a multi-page calculation even for the simplest case. At higher orders (loop corrections), the number of terms explodes.

Feynman's Insight

Feynman realized that each step in the calculation could be associated with a visual element:

• External particles = lines entering or leaving the diagram
• Virtual particles = internal lines
• Interactions = vertices where lines meet
• Conservation laws = momentum conservation at each vertex
• Integration = summing over all possible internal momenta

Instead of grinding through the algebra, you could draw the picture and read off the mathematical expression directly, using a small set of rules.

🔵 Historical Note: Feynman first presented his diagrams at the Pocono Conference in 1948. The audience — which included Bohr, Dirac, and most of the leading physicists of the era — was deeply skeptical. Bohr reportedly asked whether Feynman had forgotten the uncertainty principle. It was Freeman Dyson who showed that Feynman's approach was mathematically equivalent to Schwinger's and Tomonaga's, and who systematized the "Feynman rules" that physicists use today.

Part 2: How Feynman Diagrams Work

The Basic Elements (QED)

Every QED Feynman diagram is built from three ingredients:

1. Electron lines (solid lines with arrows): - Arrow indicates the direction of the fermion number flow - An electron moving forward in time = a positron moving backward in time - Each external electron contributes a spinor: $u(p)$ (incoming electron), $\bar{u}(p)$ (outgoing electron), $v(p)$ (incoming positron), $\bar{v}(p)$ (outgoing positron) - Each internal electron line contributes a propagator: $\frac{i(\not{k} + m)}{k^2 - m^2 + i\epsilon}$

2. Photon lines (wavy lines): - No arrow (photons are their own antiparticles) - Each external photon contributes a polarization vector: $\epsilon^\mu(k)$ - Each internal photon line contributes a propagator: $\frac{-ig^{\mu\nu}}{k^2 + i\epsilon}$ (in Feynman gauge)

3. Vertices (where an electron line meets a photon line): - Always one photon and two electron lines (one incoming, one outgoing) - Each vertex contributes: $-ie\gamma^\mu$ (where $\gamma^\mu$ is a Dirac gamma matrix) - Momentum is conserved at each vertex

Reading the Diagram

Given a diagram, the scattering amplitude $\mathcal{M}$ is obtained by:

1. Assigning momenta to all lines (external momenta are fixed; internal momenta are determined by conservation or integrated over).
2. For each element, writing the corresponding mathematical factor.
3. Multiplying all factors together.
4. Integrating over undetermined (loop) momenta: $\int \frac{d^4k}{(2\pi)^4}$.
5. Including appropriate signs for fermion loops and symmetry factors.

The cross-section is then $\sigma \propto |\mathcal{M}|^2$, summed/averaged over spins and colors as appropriate.

Example: Coulomb Scattering (Tree Level)

Two electrons scatter by exchanging one virtual photon. The diagram has: - Two incoming electron lines ($p_1$, $p_2$) - Two outgoing electron lines ($p_3$, $p_4$) - One internal photon line carrying momentum $q = p_1 - p_3$ - Two vertices

The amplitude is:

$$\mathcal{M} = \left[\bar{u}(p_3)(-ie\gamma^\mu)u(p_1)\right] \frac{-ig_{\mu\nu}}{q^2} \left[\bar{u}(p_4)(-ie\gamma^\nu)u(p_2)\right]$$

Simplifying:

$$\mathcal{M} = \frac{ie^2}{q^2}\left[\bar{u}(p_3)\gamma^\mu u(p_1)\right]\left[\bar{u}(p_4)\gamma_\mu u(p_2)\right]$$

In the non-relativistic limit, $q^2 \approx -|\mathbf{q}|^2$, and this reduces to the Fourier transform of the Coulomb potential $V(r) = e^2/(4\pi r)$. Feynman diagrams reproduce Coulomb's law as a special case.

💡 Key Insight: The Coulomb force between two electrons is, in QFT language, the exchange of a virtual photon. The $1/r$ potential arises from the $1/q^2$ factor in the photon propagator (its Fourier transform is $1/r$). The concept of "force" in classical physics is replaced by "particle exchange" in QFT. This is a profound reinterpretation: there is no "action at a distance" — forces are mediated by particles.

Part 3: Beyond Tree Level — Loops, Infinities, and Renormalization

Why Loops Matter

Tree-level diagrams (no closed loops) give the leading-order approximation. Higher precision requires loop diagrams, which involve integrals over internal momenta.

The one-loop correction to electron-electron scattering includes diagrams where the exchanged virtual photon splits into a virtual electron-positron pair and recombines. This "vacuum polarization" diagram modifies the effective charge of the electron at short distances.

The Ultraviolet Catastrophe of QFT

Loop integrals typically diverge at large momenta (the "ultraviolet" region):

$$\int \frac{d^4k}{(2\pi)^4} \frac{1}{k^2(k-p)^2} \sim \int \frac{k^3 dk}{k^4} = \int \frac{dk}{k} \sim \ln\Lambda \to \infty$$

This logarithmic divergence (and worse, quadratic divergences in some diagrams) initially seemed to doom QFT.

Renormalization: The Solution

The infinities are absorbed into redefinitions of the physical parameters. The "bare" electron mass $m_0$ and charge $e_0$ in the Lagrangian are infinite, but the "renormalized" (physical) mass $m$ and charge $e$ are finite:

$$m = m_0 + \delta m \quad (\text{both } m_0 \text{ and } \delta m \text{ infinite; their sum finite})$$ $$e = e_0 + \delta e \quad (\text{same})$$

This is not a trick but a reflection of physical reality: the properties of a particle depend on the scale at which you observe it. An electron surrounded by a cloud of virtual pairs has a different effective charge at close range (where you penetrate the cloud) than at long range (where you see the screened charge).

The criterion for a "renormalizable" theory: only a finite number of parameters need to be redefined (the masses and coupling constants already in the Lagrangian). QED, the electroweak theory, and QCD are all renormalizable. Quantum gravity is not — which is why it remains unsolved.

📊 By the Numbers: The electron $g-2$ calculation at each order involves:

Part 4: What Feynman Diagrams Teach Us About Physics

Particles Mediate Forces

In classical physics, forces act at a distance (or through fields). In QFT, forces arise from particle exchange:

• Electromagnetism: photon exchange
• Strong force: gluon exchange
• Weak force: $W^\pm$ and $Z^0$ exchange
• Gravity (in the quantum gravity program): graviton exchange

The range of a force is related to the mass of the exchanged particle: $R \sim \hbar/(mc)$. Massless photons give infinite-range electromagnetism. Massive $W/Z$ bosons give short-range ($\sim 10^{-18}$ m) weak interactions.

Everything Interacts with Everything (Almost)

Feynman diagrams make visible the intricate web of interactions. A Higgs boson can decay into two photons — even though the Higgs does not couple directly to photons. The process goes through a loop of virtual top quarks (which couple to both the Higgs and the photon). The discovery channel $H \to \gamma\gamma$ at the LHC relies on this loop-mediated process.

The Sum Over Histories

Each Feynman diagram represents one "way" the process can happen. The total amplitude is the sum over all diagrams:

$$\mathcal{M}_{\text{total}} = \mathcal{M}_{\text{tree}} + \mathcal{M}_{\text{1-loop}} + \mathcal{M}_{\text{2-loop}} + \ldots$$

This is the QFT version of Feynman's "sum over histories" (Chapter 31): every possible intermediate configuration contributes, with a weight given by $e^{iS/\hbar}$. The diagrams organize this infinite sum into a systematic perturbation series.

What Diagrams Cannot Do

Feynman diagrams are a perturbative tool. They work when the coupling constant is small ($\alpha \ll 1$ for QED). When the coupling is large — as in QCD at low energies ($\alpha_s \sim 1$) — the perturbation series does not converge, and diagrams fail. Non-perturbative methods (lattice QCD, instantons, effective field theory) are needed.

Diagrams also do not capture topological or global effects in QFT (instantons, vacuum tunneling, confinement). The modern understanding of QFT goes well beyond Feynman diagrams, but diagrams remain the primary computational tool for weak and electromagnetic processes.

Discussion Questions

1. Feynman diagrams translate abstract mathematics into pictures. Discuss the role of visualization in physics. Can you think of other examples where a pictorial representation transformed a field? (Consider circuit diagrams, free-body diagrams, spacetime diagrams.)

2. Virtual particles are "real enough" to produce measurable effects (Casimir force, Lamb shift) but cannot be directly detected. How should we think about the ontological status of virtual particles? Are they "real"?

3. The number of Feynman diagrams grows factorially with the order of perturbation theory. At high enough order, the perturbation series must diverge. What does this mean for the status of QED as a fundamental theory? (Research "asymptotic series" and the Dyson argument for the divergence of perturbation theory.)

4. Feynman introduced his diagrams for QED in 1948. Today they are used in QCD, electroweak theory, condensed matter physics, and even string theory. Why were they so broadly applicable? What feature of Feynman diagrams makes them useful across such different domains?

Further Investigation

• Read Feynman's own account of developing his approach in his Nobel Lecture (1965), reprinted in QED: The Strange Theory of Light and Matter.

• Research the history of the Shelter Island (1947) and Pocono (1948) conferences, where the post-war generation of physicists (including Feynman, Schwinger, Dyson, Bethe, and Oppenheimer) laid the foundations of modern QFT.

• Look up the "Feynman diagram" for your favorite particle physics process on the Particle Data Group website (pdg.lbl.gov). Can you identify the vertices, propagators, and external lines?

• Research amplitude methods (spinor-helicity formalism, BCFW recursion, amplituhedron) — modern techniques that go beyond Feynman diagrams. How do they achieve simpler expressions for scattering amplitudes?