Draw the diagram. Watch the physics follow.
A complete QED course, one real calculation at a time — from Feynman diagrams you draw yourself, through tree level and into renormalization, to numbers an experiment can check.
every lesson rides the same pipeline
- Diagrams
- Amplitude
- Squaring
- Spin sum
- Phase space
- Cross section
QED Part 1 — Foundations
We start where a first QFT course hands you the keys: the fields are already quantized and the Lagrangian is on the table. These lessons turn it into a working rulebook.
- 1.1 The QED Lagrangian Where the course begins: the electron field, the photon field, and how demanding local gauge invariance conjures the one interaction term that powers everything that follows. the Dirac & photon fields · local U(1) gauge invariance · the interaction term start →
- 1.2 Propagators How disturbances in the fields travel: derive the photon and fermion propagators from the free theory — and meet the gauge-fixing trick that makes the photon’s possible at all. Green functions · gauge fixing & the photon propagator · the fermion propagator start →
- 1.3 The Feynman rules & available interactions Assemble the complete rulebook — external legs, propagators, the vertex — then take it to the canvas and discover what QED lets you draw, and what it forbids. the QED vertex · external legs · the feasibility frontier start →
QED Part 2 — Tree level
Every calculation rides the same pipeline; every lesson adds the minimum new physics.
- 2.1 e⁺e⁻ → μ⁺μ⁻ at leading order Draw the contributing diagram, watch the amplitude assemble itself, then follow it through squaring, spin sums, and phase space to σ(√s) on a collider plot. Feynman rules · the amplitude · spin sums · phase space → σ(√s) start →
- 2.2 Crossing & Mandelstam e⁻μ⁻ → e⁻μ⁻ The same amplitude, read sideways: rotate annihilation into electron–muon scattering, meet s, t, u — and watch q² go spacelike. crossing symmetry · Mandelstam variables · spacelike q² start →
- 2.3 Interference e⁺e⁻ → e⁺e⁻ Two diagrams contribute — so amplitudes add before they square. Interference, the relative minus sign from exchanging fermions, and your first two-diagram checkpoint. amplitude superposition · interference · the relative minus sign start →
- 2.4 Identical particles e⁻e⁻ → e⁻e⁻ Indistinguishable electrons: the u-channel partner diagram, antisymmetrization, and why the cross section carries a factor of ½. identical fermions · u-channel · symmetry factor ½ start →
- 2.5 Propagating fermions & photon polarization e⁻γ → e⁻γ A photon scatters off an electron: the first internal fermion propagator, photon polarization sums, and the Klein–Nishina formula. fermion propagators · polarization sums · Klein–Nishina start →
- 2.6 Annihilation — putting it together e⁺e⁻ → γγ Cross Compton into pair annihilation: identical photons in the final state, and a checkpoint that uses everything Part 1 taught. crossing in practice · identical bosons · the spin-statistics sign · a collinear logarithm start →
- 2.7 From amplitude to number σ → N Every lesson so far ended at a curve. No experiment measures a curve — it counts events. Flux and why σ is an area, luminosity, and why a detector’s acceptance defines the observable rather than approximating it. Then point it at LEP, and watch tree-level QED miss by a factor of 191.5. the flux factor · luminosity and N = σ∫L dt · acceptance as definition · when a discrepancy is a discovery start →
QED Part 3 — Loops & renormalization
Where the infinities are. Every integral in this part has one or two propagators and an exact closed form — enough to meet a divergence, decide what to do with it, and watch the fine structure constant stop being constant.
- 3.1 Why loops diverge A closed cycle leaves one momentum undetermined, and you integrate over it. Count the powers of that momentum and you know in advance which diagrams diverge — a list that turns out to be short, and to contain every loop the canvas will let you draw. loop momenta · superficial degree of divergence · the divergent list · choosing a regulator start →
- 3.2 Poles, contours, and continuation Two promises come due at once. The $+i\varepsilon$ you were told to carry silently turns out to be what licenses swinging a contour, and the analytic continuation dimensional regularization runs on is a theorem nobody has shown you yet. Residues, arcs that die, and a function that outlives the integral defining it. residues and contours · Jordan's lemma · the iε prescription · analytic continuation start →
- 3.3 Evaluating a loop integral The recipe, once, in the abstract: combine the denominators with Feynman parameters, shift the loop momentum, rotate the contour into Euclidean space, and do the angular integral in d dimensions. What comes back is a Γ function with a pole sitting exactly at d = 4. Feynman parameters · Wick rotation · the master formula · expanding in ε start →
- 3.4 Vacuum polarization The first complete calculation. A photon spends part of its life as an electron–positron pair; the Ward identity fixes the shape of the answer before you compute it, and above 4m² the result goes complex — because up there the pair can be real. the photon self-energy · transversality from the Ward identity · the branch cut at 4m² · the optical theorem start →
- 3.5 The one-loop masters You have just computed a bubble at equal masses. Here is the same object with any masses, in every kinematic region, beside the tadpole that accompanies it — the two integrals every one-loop calculation in this part is assembled from. the tadpole A₀ · the bubble B₀ · thresholds & branch cuts · a basis of masters start →
- 3.6 Renormalization What to do with a 1/ε. Counterterms reorganize the Lagrangian rather than patch its output, Z factors connect the parameters you wrote down to the ones you can measure, and a scale μ appears that no observable is allowed to remember. counterterms · Z₃ and the photon field · on-shell vs MS-bar · the scale μ start →
- 3.7 Mass renormalization The electron dresses itself in its own field. The self-energy moves the pole of the propagator, so the mass in the Lagrangian is not the mass anybody measures — and there is more than one honest answer to which one deserves the name. the self-energy −iΣ(p) · Z₂ and field renormalization · pole mass vs running mass evaluates locked
- 3.8 The running coupling Sum the polarization bubbles and the coupling starts to move: vacuum screening makes the charge you measure depend on how closely you look. The prediction is checkable against real experiments — and where it falls short is where the rest of the Standard Model is hiding. resummation · screening · α(q²) · the Landau pole locked
- 3.9 The renormalization group Turn “no observable may remember μ” into a differential equation and it starts predicting. The β function runs any coupling, not just this one — so run all three of the Standard Model’s, and watch them very nearly meet at an energy nobody will ever build a machine to reach. the RG equation · the β function · running any coupling · the near-miss at 10¹⁶ GeV locked
QED Part 4 — Precision QED
Where the integrals stop being solvable one at a time. Triangles and boxes reduce onto a basis of masters — and out of the vertex fall the electron’s anomalous magnetic moment and the cancellation that makes a cross section finite.
- 4.1 The vertex & tensor reduction Three propagators, and a numerator carrying loop momenta upstairs. Passarino–Veltman reduction decomposes it onto scalar integrals, and what survives is two form factors — everything a photon is able to learn about an electron’s structure. tensor reduction · the triangle C₀ · form factors F₁, F₂ · Ward identity Z₁ = Z₂ evaluates locked
- 4.2 g − 2 F₂(0) = α/2π. One line of algebra at the end of a long calculation predicts that the electron is not quite as magnetic as Dirac said — and the measurement agrees to more decimal places than almost anything else in physics. the anomalous magnetic moment · Schwinger’s α/2π · prediction vs measurement locked
- 4.3 The box Two photons exchanged at once: four propagators around a loop, and the last master integral. Reducing it closes the one-loop toolkit — past four points, nothing new is ever needed. the box D₀ · two-photon exchange · reduction to lower points locked
- 4.4 Soft photons & the infrared A photon of arbitrarily low energy costs arbitrarily little to emit, and the integral says so. Where the ultraviolet divergence came from short distances, this one comes from long ones — the same regulator, doing the opposite job. the eikonal approximation · the soft factor · UV ε vs IR ε · collinear divergences locked
- 4.5 Real emission & the cancellation e⁺e⁻ → μ⁺μ⁻(γ) Ask for two muons and the answer is infinite. Ask for two muons plus any photon too soft to see, and it is a number. The virtual and real divergences cancel against each other — and what makes them cancel is the detector’s own resolution. Bloch–Nordsieck · virtual + real cancellation · resolution as definition · finite σ_NLO locked
- 4.6 One-loop curiosities — Furry’s theorem & light-by-light Two amplitudes that power counting says should diverge, and do not. Charge conjugation kills every odd-photon loop before you compute it, and gauge invariance rescues light-by-light — symmetry doing the work of an integral. Furry’s theorem · light-by-light · symmetry as computation evaluates locked