The panels above show
block-spin renormalization (Kadanoff 1966): start from a 2D Ising configuration at temperature T, then repeatedly replace each 2×2 block by its majority spin. Each panel is one more "zoom out."
Watch what the flow does:
•
T < T_c: blocks agree more and more — the system flows to the
ordered fixed point (all aligned). Ferromagnet.
•
T > T_c: blocks decorrelate — flow to the
disordered fixed point (pure noise). Paramagnet.
•
T = T_c: the configuration is statistically
self-similar — fractal clusters within clusters at every scale. Coarse-graining changes nothing. This scale-invariant state is an
unstable fixed point of the RG flow.
Universality: near a fixed point, almost every microscopic detail is an
irrelevant perturbation — it shrinks under the flow. Only a few
relevant directions (temperature, field) survive. That's why a magnet, a liquid-gas transition, and a binary alloy — completely different microscopically — share the
same critical exponents: they flow to the same fixed point. Critical behavior is a property of the flow, not the material.
Wilson's revolution (Nobel 1982): Kenneth Wilson made this quantitative — RG flow in the space of all possible Hamiltonians, with critical exponents as eigenvalues of the flow linearized at the fixed point. This solved the Kondo problem, tamed the infinities of quantum field theory (renormalization isn't a dirty trick; it's the statement that physics at low energy
shouldn't care about unknown physics at high energy), and gave us
effective field theory — the modern framing in which the Standard Model itself is just the long-distance limit of something deeper.
Everywhere: RG fixed points organize phase transitions, the flow of coupling constants with energy (asymptotic freedom of QCD — quarks become free at short distance), deep-learning theory (training as RG flow), and even
MERA tensor networks, where the holographic direction
is RG scale. Asking "what survives zooming out?" turns out to be the most powerful question in science.