70
Part Two · Quantum fields and particles

Renormalization

The electron’s charge is not a fixed number. Measured from far away, the fine-structure constant is 1/137.036. Measured at the energy of the Z boson, at LEP, it is about 1/128. The vacuum around an electron is full of fleeting electron–positron pairs that line up and partly screen it, like a charge in water; probe more closely and you see more of the bare charge. Renormalization is the theory of how such couplings change with the scale you look at, and it turns the infinities of Chapter 69 into predictions.

On the plate: the strength of electromagnetism against the energy at which you measure it.

Bare and measured

Write the QED Lagrangian with “bare” mass m0m_0 and charge e0e_0. Split each into the measured value plus a counterterm: e0=e+δee_0 = e + \delta e, m0=m+δmm_0 = m + \delta m, and so on. Choose the counterterms to cancel the divergences of Chapter 69 exactly, by demanding renormalization conditions: for example, that the charge measured in very-low-energy scattering is 1/137.036. Every other prediction is then finite and expressed in terms of measured quantities. The bare values are infinite, but they are never observed. Established

The running coupling

The finite part of the photon loop depends on momentum. Absorbing it into an effective coupling at scale QQ gives, at leading order,

1α(Q)=1α(0)−23π∑fNcQf2 ln⁡Qmf\frac{1}{\alpha(Q)} = \frac{1}{\alpha(0)} - \frac{2}{3\pi}\sum_f N_cQ_f^2\,\ln\frac{Q}{m_f}

summed over every charged particle lighter than QQ, with Nc=3N_c = 3 colours for quarks. Each new particle that becomes light enough adds to the screening. Probe up to 1.96 (log₁₀ GeV). This crude estimate reaches about 1/127.8 at the Z mass; the precise value, using measured hadron production for the quark loops, is 1/127.95. The plate marks both. Established

The Landau pole

Follow the QED formula upward and 1/α1/\alpha eventually hits zero: the coupling becomes infinite at a scale Lev Landau identified in the 1950s. With the electron alone it is around 1028010^{280} MeV, absurdly far beyond the Planck scale, where gravity takes over anyway. So the pole is not a practical problem, but it tells you QED cannot be the complete theory at arbitrarily short distances. The strong force runs the other way (Chapter 79).

What “renormalizable” means today

Old textbooks divided theories into renormalizable ones, like QED, where a finite number of counterterms suffice, and the rest, which were considered sick. Kenneth Wilson’s view, now standard, is different. Every theory is an effective field theory, valid below some scale MM. Terms in the Lagrangian with dimension above four come with powers of 1/M1/M and so are tiny at low energy. Renormalizable theories are simply the ones that survive when MM is far away. Fermi’s theory of the weak force (Chapter 65) is non-renormalizable and works perfectly well below the W mass, where it should. General relativity is the same kind of theory, with MM the Planck mass. Established

What you now have

Bare and renormalized parameters, counterterms and renormalization conditions, the running of α\alpha from 1/137 to 1/128, the Landau pole, and the effective-field-theory meaning of renormalizability.

· · ·

Three questions to carry with you.

Why is α larger at higher energy?tap
Vacuum pairs screen the charge; at short distances you see through more of the screening.
Are bare parameters measurable?tap
No. Only renormalized quantities, fixed by measurement, appear in predictions.
Is a non-renormalizable theory useless?tap
No; it is an effective theory, accurate below its scale M, with corrections suppressed by powers of E/M.

And one to work out. By how much does 1/α1/\alpha change from zero energy to the Z mass, using 137.036 and 127.95?

Δ(1/α) =

Sources: Particle Data Group, Review of Particle Physics, electroweak review (α(m_Z)⁻¹ = 127.95 in the MS-bar scheme). L. D. Landau, in Niels Bohr and the Development of Physics (1955). K. G. Wilson and J. Kogut, Phys. Rep. 12, 75 (1974). M. D. Schwartz, ch. 22–23. Checked by code: leading-log running with lepton and approximate quark thresholds gives 1/α(m_Z) ≈ 127.8; Landau pole for the electron alone at 10²⁸⁰ MeV.

Next · Chapter 71
The renormalization group

Boiling water and the early universe obey the same mathematics.

Turn the page