The advanced version of Episode 8, "The Ambiguity of i Gives Birth to Force" — why "not fixing" is the winning move
The bonus editions so far have built up one idea: "units, time, energy — what you fix is a choice, and physics lives only in the dimensionless." This time we lift that idea all the way up to the greatest weapon of 20th-century physics, the gauge principle. Yang–Mills theory worked, crudely put, because it chose a coordinate system that fixes nothing. That intuition is 70% right, and fixing the remaining 30% makes it deeper still. We'll rewrite Episode 8, "The moment it becomes local, force is born," this time from the side of "why not fixing is the winning move."
The idea of Yang–Mills is to take what we did in Episode 8 — "you're free to re-choose the ambiguous phase independently at every point (local gauge invariance)" — and extend it from the \(U(1)\) of electromagnetism to the non-abelian groups \(SU(2),SU(3)\). The paradox that arises here is the real identity of your intuition.
The more freedom of description you leave in (the less you fix), the more rigidly the dynamics get bound.
Demand a free re-choosing at every point, and a field (a force) that patches it up is forced into existence, and the terms you're allowed to write shrink to almost nothing. The Yang–Mills Lagrangian is essentially pinned down to the single term \(-\tfrac14 F^2\).
Almost no free parameters left = high predictive power = a short formula too. Free up the top (the description), and the bottom (the dynamics) becomes unique. This is the real identity of "not fixing makes it simple" — it isn't a matter of appearance, but a deep fact: the very form of the theory is bound and determined by the symmetry. "Not fixing" turned out to be synonymous with "binding it as tightly as possible with symmetry."
Here is the crux of precision. When you actually try to compute (quantize) Yang–Mills, without fixing, the integral diverges and you can't even define the propagator of the photon or gluon. Because you end up summing over infinitely many copies of the same physics (the gauge orbit). So you always need gauge fixing (the Faddeev–Popov procedure, 1967 — and the price is that an auxiliary field, the "ghost," springs up). "Fixing nothing" is powerful as a principle, but as an operation it simply doesn't hold up.
It's not "not fixing" that's admirable. What's admirable is to choose a fixing that keeps the symmetry visible, and avoid a fixing that erases it.
The winning path isn't to reduce the fixing, but to fix in a way that doesn't break the symmetry.
Even within "fixing," the two roads part into light and shadow. It's a trade of what you reveal and what you hide.
| Way of fixing | What gets revealed | What gets hidden / the price |
|---|---|---|
| Covariant gauge (Lorentz-type + ghosts) | Renormalizability is revealed. Lorentz symmetry is preserved too. | An "apparent field," the ghost, is added. It doesn't look like only physical particles. |
| Unitary gauge (erase every unneeded field = fix as much as possible) | Only the physical particles are visible, which feels nice (unitarity is revealed). | The formula runs wild at high energy, and it "appears" non-renormalizable. |
The decisive move that made Yang–Mills a "usable theory" was that 't Hooft and Veltman (1971) used the covariant gauge and proved it was renormalizable while keeping the symmetry visible. Had they only looked at the unitary gauge (fixing as much as possible), the theory might have stayed buried, looking like a "bad theory that breaks at high energy." Choose the fixing that reveals the most important property, and don't over-fix — that's the substance of the winning path.
Here we correct the intuition one notch. There are counterexamples — the Lorenz gauge (electromagnetic waves become a wave equation), the Coulomb gauge (electrostatics in one shot), the harmonic gauge (gravitational waves become a clean wave equation). A clever fixing can, on the contrary, dramatically simplify a specific problem. So "more fixing = more complex" doesn't hold in general.
The real failure mode isn't the "amount of fixing," but breaking, with your fixing, the invariant "core" that must not be fixed. And this is exactly the same structure as the VSL of Bonus 3.
VSL (Bonus 3): fix \(c,e,\hbar\) separately → accidentally move the invariant \(\alpha\) you were supposed to protect → dead end.
Unitary-gauge-style failure: erased the symmetry by fixing → renormalizability became invisible.
Neither is a case of "over-fixing"; both are cases of breaking, with your fixing, a core that must not be fixed (α / the symmetry).
Conversely, the successes (the covariant gauge / the α-invariant gauge of Bonus 4) made the core (the symmetry, α) a sanctuary and freely relativized only the remaining redundant directions. What we said in Bonus 4 — "the key to preserving equivalence isn't how you move \(c\) but whether you can protect \(\alpha\)" — becomes, word for word, the very "condition for a winning gauge fixing."
The thing you've been saying all along — "absolute values are bookkeeping; what must be protected is the invariant ratio / structure" — is, in the language of physics, the gauge principle itself. What Bonuses 2 through 7 showed through "how you move units, time, and energy," Yang–Mills does through "re-choosing the field itself at every point." One and the same principle wears the same face at the metrology level and at the field-theory level. The force we spoke of in Episode 8, "the ambiguity of i gives birth to force," was precisely the child born of "the freedom not to fix."
"Yang–Mills won because it doesn't fix" is correct as an intuition, but strictly it's "it won because it chose a fixing that keeps the symmetry visible." The ghosts of the covariant gauge, and the simultaneous holding of renormalizability and unitarity, are really quite technical matters; here we've drawn only the through-line. The position and height of the figure's "valley" are a schematic showing the trade-off of fixing, not a quantitative curve.
There's also a way of putting it that gauge symmetry, strictly speaking, is less a "symmetry" than a redundancy of description (when you count physical degrees of freedom, you discount the redundant part). Even in this view the conclusion is the same — a fixing that wipes out the redundancy also wipes out the hidden structure.
Yang–Mills is strong because when you demand local gauge invariance (= you can re-choose freely at every point, i.e. you don't fix), a force (the gauge field) is forced into existence and the form of the theory is almost uniquely determined. Freedom of description gives birth to the rigidity of the dynamics (few parameters, a short formula, high predictive power). But quantization requires fixing, and the winning path is to "choose a fixing that makes the symmetry visible, and never break the invariant core with your fixing."
"More fixing = more complex" isn't monotonic. A clever fixing simplifies a problem. The real failure is breaking, with your fixing, the core that must not be fixed (α / the symmetry) — the sin of VSL and the apparent breakdown of the unitary gauge were the same sin. Make the core a sanctuary, and relativize only the redundant directions. The backbone running since Bonus 2 was, in the language of physics, the gauge principle itself.
Print / save as PDF: ⌘+P (Ctrl+P on Windows). On screen, use the slider to see the valley between over-fixing and under-fixing. "See the answer" opens each solution.