Renormalization That ClicksBonus ⑥ (the linking episode) / Three backbones, one spine

Three series written separately turn out to have been writing one and the same operation in three vocabularies ── the collection's linking episode

Three backbones, one spine "Cosmology That Clicks" and its c·t = constant.
"The Universe Is a Computer" and its F = 1/(Cn)D.
And the renormalization group ── all three were walking the same single axis, scale, in different words.

Tools you'll need: Episode 4's relevant/irrelevant, Episode 7's z↔1/E, Bonus ③'s fixed points and eigenvalues The heart of this one: D = a scaling dimension, c·t=const = a choice of μ

This is a linking episode. No new physics appears. Instead, three series written separately are laid on top of one another ── the backbone \(c\cdot t=\)constant of the sister series "Cosmology That Clicks," the backbone \(F=1/(Cn)^D\) of "The Universe Is a Computer," and this series' renormalization group. The conclusion first: the latter two are very nearly the same object, and the first is the same argument in a different field. "The Universe Is a Computer" says \(D\) runs ── that is not a metaphor; it is literally renormalization group flow. And "fix \(\alpha\), not \(c\)" has exactly the shape of renormalization's scheme independence. Better still, the three series had already passed through the same point three times without noticing ── the neutron, Landauer, and "a continuum limit exists only at a critical point."

01Lay the three backbones side by side

SeriesIts one-line backboneWhat it is saying
Cosmology That Clicks\(c\cdot t=\) constant
(fix \(\alpha\), not \(c\))
Dimensionful quantities are bookkeeping. Change the description and the observables don't move
The Universe Is a Computer\(F=1/(Cn)^D\)
(dimension is a read-off; \(D\) runs)
Change the scale and the read-off changes. How it runs is the phenomenon
Renormalization That Clicksthrow it away, same answer / relevant and irrelevantChange the scale and the theory flows. Only the fixed point's properties survive

All three are saying "when you move the scale, what changes and what doesn't." The only difference is where you stand to look ── cosmology from the bookkeeping side, the computer series from the ruler side, renormalization from the theory's own side.

02The \(D\) of \(F=1/(Cn)^D\) was a scaling dimension

Episode 10 of "The Universe Is a Computer" defines \(D\) like this:

Two quantities of the same shape
$$D=-\frac{d\ln F}{d\ln r}\qquad\Longleftrightarrow\qquad \beta(g)=\frac{dg}{d\ln \mu}$$

Both are derivatives with respect to the logarithm of scale. The first is the slope of how a force falls off, the second the running of a coupling ── the same kind of derivative along the same axis. So the title of Episode 10, "\(D\) runs," translates directly into the language of the renormalization group.

The Universe Is a ComputerRenormalization That Clicks
\(D\) is a read-off, not a number the world "has" (Ep. 1)the scaling dimension of an operator (Eps. 4, 6)
\(D\) runs (Ep. 10)renormalization group flow (Ep. 4)
weak force: \(D:2\to\infty\) (heavy W/Z give an exponential cutoff)the decoupling theorem ── heavy particles decouple from low-energy physics (Ep. 4)
strong force: \(D:2\to0\) (confinement; it stops falling with distance)a direction that grows toward the IR (relevant behaviour)
"the easiest dimension is per-phenomenon" = the upper critical dimension (Ep. 1)the boundary between a stable Gaussian fixed point and the Wilson–Fisher one (Ep. 4, Bonus ③)
2 in the UV as well; the cosmic web about 2 (Ep. 14)switching fixed points from scale to scale = a crossover

That Episode 1 already uses the phrase "upper critical dimension" settles it. That is renormalization-group vocabulary ── the measuring stick at the heart of "The Universe Is a Computer" was the renormalization group under another name.

One thing this side can add ── a footnote to "C and D are separate axes"

Episode 10 says "\(C\) (coupling) and \(D\) (geometry) are separate axes." Broadly true ── but from the renormalization group's point of view, the two axes mix once you leave a free theory. When the coupling runs, the dimension picks up a shift:

$$D_{\text{effective}}=D_{\text{classical}}+\gamma(g)$$

\(\gamma\) is called the anomalous dimension. Part of the reason \(D\) runs is that \(C\) runs. And this is exactly why, in Episode 4, the critical exponent \(\beta\approx0.326\) refused to be a clean number like \(1/2\) ── that awkward remainder is the departure from mean field (\(\gamma=0\)). The two axes are genuinely independent only very close to a Gaussian fixed point.

03A complex dimension is a renormalization eigenvalue gone complex

Episode 8 of "The Universe Is a Computer" opens \(D\) up to the complex plane ── \(D=a+i\beta\), with the imaginary part producing log-periodic oscillation and a preferred zoom factor \(\lambda=e^{2\pi/\beta}\), described there as "the mark of continuous scale symmetry broken down to a discrete one."

That is exactly the complex version of the eigenvalues around a fixed point from Bonus ③.

What happens when an eigenvalue goes complex

One step of coarse-graining multiplies a deviation from the fixed point by an eigenvalue \(\Lambda\) (Bonus ③'s \(N^{1-k/2}\) was a concrete case). If \(\Lambda\) is real, the deviation simply grows or shrinks monotonically. But if \(\Lambda\) is complex ──

$$\Lambda^n = |\Lambda|^n e^{\,i n\theta}\quad\Longrightarrow\quad \underbrace{|\Lambda|^n}_{\text{ordinary growth}}\times\underbrace{\cos(n\theta+\varphi)}_{\text{log-periodic oscillation}}$$

Because the coarse-graining count \(n\) is the logarithm of scale, the oscillation is periodic in the logarithm of scale. The same view returns only at particular magnifications ── continuous scale invariance has fallen to discrete scale invariance. Episode 8's \(\lambda=e^{2\pi/\beta}\) is nothing but a restatement of this \(\theta\).

Episode 4 said "at a critical point it looks the same at every magnification." The world of complex dimensions is one notch weaker ── not at every magnification, but at every step of a fixed ladder. Self-similar sets like the Cantor set, and log-periodic precursor phenomena, live here.

04\(c\cdot t=\)constant was a choice of \(\mu\)

Now the other backbone. "Cosmology That Clicks" draws its honest line as: "fix \(\alpha\), not \(c\)"; "you can fix at most three"; "only dimensionless quantities are invariant."

The renormalization group has a demand of exactly the same logical shape. When you make a theory finite, the energy \(\mu\) at which you cut is a bookkeeping mark chosen by a human. And physical quantities must not depend on it. As an equation:

The Callan–Symanzik equation ── the demand "do not depend on the bookkeeping," as an equation
$$\mu\frac{d}{d\mu}\,\bigl(\text{observable}\bigr)=0$$

Astonishingly, this is what the renormalization group equation is. Turn the single sentence "the answer must not depend on how you chose to describe it" into a differential equation and the running of the coupling (the beta function) is forced. The very freedom to pick a gauge constrains the physics.

It has the same shape as \(c\cdot t=\)constant saying "how you measure time is a gauge choice, and if \(\alpha\) is invariant it leaves no trace on observation" ── you may choose the ledger; you may not choose the answer.

Cosmology's Bonus ⑥ "two axes" was already this distinction Bonus ⑥ of "Cosmology That Clicks" split the ways \(\alpha\) can move into two ──
・"moving in time" (gauge freedom present / disputed) = the ledger axis
・"running with energy" (no freedom / established fact) = the renormalization group axis
That split is this series' subject exactly. 1/137 → 1/128 is the latter, where there is no freedom of choice ── which is why it is an "established fact." Conversely \(c\cdot t=\)constant is the former, which is why it "leaves no trace on observation." That bonus was already a renormalization episode.

And Bonus ⑦, "the resolution knob has two ends (a floor at 1/137, a ceiling at the Planck edge)" ── that knob becomes a direction in space in Episode 7 of this series. In AdS/CFT the radial coordinate satisfies \(z\leftrightarrow 1/E\), so turning the knob is diving deeper. The two ends are the deep interior of the bulk (IR = the 1/137 floor) and the very edge of the boundary (UV = the Planck edge). Cosmology's Bonus ⑦ becomes geometry in Renormalization's Episode 7.

05Try it ── read one axis in three languages

The figure below deliberately has only one horizontal axis ── distance \(r\), and its flip side, energy \(E\sim\hbar c/r\). The top panel is the figure from Episode 10 of "The Universe Is a Computer" (the running \(D\) of the four forces); the bottom is the figure from Episode 6 of "Cosmology That Clicks" (\(\alpha\) running from 1/137 to 1/128). Two figures from two different series, put on one shared axis.

Move the cursor with the slider and the readout gives three readings at once. What the three series say about the same single point ── that is this whole episode.

Figure: the horizontal axis is distance r (log) and, behind it, energy E=ħc/r. Top = the four forces' D=−dlnF/dlnr (electromagnetism and gravity stay at 2, weak goes 2→∞, strong goes 2→0). Bottom = the running of the electromagnetic coupling α (1/137 → 1/128). Read the cursor position in three series' languages at once
electromagnetism & gravity (D=2) weak force (2→∞) strong force (2→0) running of 1/α

06Three times already, through the same point

The surprise in writing this linking episode was here. The three series had passed through the same three points without meaning to. These are not analogies; they are overlaps of content.

CrossingCosmology That ClicksThe Universe Is a ComputerRenormalization That Clicks
the neutronEp. 7: BBN passes judgment on \(c\cdot t=\)const (\(\Gamma\propto Q^5\))──Ep. 2: 880 s = the time for the phase to become untrackable; bottle vs beam
LandauerEp. 10: decay is the erasure of informationEp. 6: only erasing has a costBonus ②: \(k_BT\ln2\); the observer as a coarse-graining device
the continuum limit──Ep. 9: a continuum limit exists only at a critical pointEp. 4: diverging correlation length and scale invariance
The third one lands hardest

The caveat in Episode 9 of "The Universe Is a Computer" ── "the continuum limit stands only at a critical point, and the area law is an assumption" ── writes down the central proposition of the renormalization group independently, from the computational side. You can send a lattice spacing to zero only when the correlation length diverges (otherwise the mesh survives in the answer). In other words, a continuum field theory can exist only on top of a fixed point.
So Episode 9's open question "is the lattice scaffolding or foundation?" and renormalization's open question "does that fixed point really exist (asymptotic safety)?" are the same hole, looked into from two sides.

And the finale of "Cosmology That Clicks" ended right here The finale of "Cosmology That Clicks" closes: "gravity climbs as \(E^2\) and catches up at the Planck scale ── and there it can no longer be renormalized."
Which means this series is also the continuation of that finale. The name of the door left open there turned out to be the title of this series ── I did not notice while writing it.

07The whole collection on one sheet

Move the scale and…Cosmology That ClicksThe Universe Is a ComputerRenormalization That Clicks
what changes\(\alpha\) runs with energythe read-off \(D\) runscouplings flow (the beta function)
what doesn'tdimensionless observablesinvariants (unmoved by changing representation)fixed-point properties, critical exponents
what you may choose (the ledger)which constants to fix (at most three)which representation to read in (Ep. 3's zoo)the renormalization scheme, \(\mu\)
what may be discardeddimensionful numbersredundancyirrelevant operators
the limits (both ends)the 1/137 floor and the Planck wall (Bonus ⑦)\(10^{90}\) bits, \(10^{120}\) operationswhere coarse-graining breaks (Ep. 5)
where it arrives"only dimensionless quantities are invariant""I drew the map; I planted no flag""the universe isn't slacking off; we are the ones with no choice"
◇ ◇ ◇
The honest line ── identical, isomorphic, and merely similar

Established: the beta function and anomalous dimension \(\gamma(g)\); the Callan–Symanzik equation (that \(\mu\)-independence of observables yields the renormalization group equation); the decoupling theorem; the running of the electromagnetic coupling (\(1/\alpha\approx137\), about 128 at \(M_Z\)); the upper critical dimension; the correspondence between complex renormalization eigenvalues and discrete scale invariance with log-periodic oscillation; that a continuum limit requires a critical point (fixed point); and Landauer's principle. All standard physics.

Please keep these apart (the most important caveat in this episode):
"\(D\) = a scaling dimension" is not an identity but an isomorphism. The \(D\) of "The Universe Is a Computer" is a general name for read-offs applied to many objects, and in practice it covers at least three different things: the exponent of a force's fall-off, the fractal dimension of the cosmic web, and the spectral dimension of quantum gravity. These share the form "logarithmic derivative with respect to scale" but are technically distinct quantities with no guarantee of agreeing (which is precisely what Episode 14 means by "the numbers do not converge to one"). The tables above are correspondences of form, not identifications of quantity.
\(c\cdot t=\)constant and \(\mu\)-independence merely share a logical structure. One is a choice of coordinates and units in cosmology, the other a choice of regularisation scheme in field theory; the contents of the ledgers differ. Only the form "you may choose the ledger, not the answer" is shared.
③ By contrast, the three crossings in section 06 are overlaps of content, not of form (the same neutron, the same Landauer principle, the same "a continuum limit stands only at a critical point"). Those are not metaphors.
④ This episode is a linking episode and makes no new claim. It translates three series into one vocabulary and lays them side by side; that is all.

Exercises
  1. How are \(D=-d\ln F/d\ln r\) and \(\beta(g)=dg/d\ln\mu\) the same, and how do they differ?
    See the answer
    Same: both are derivatives with respect to the logarithm of scale, measuring how a read-off moves when you move the scale. Different: what is being differentiated ── the size of a force in one case, a coupling constant in the other. So they are isomorphic, not identical.
  2. What footnote does the renormalization group add to "C (coupling) and D (geometry) are separate axes"?
    See the answer
    Away from a free theory the anomalous dimension \(\gamma(g)\) enters and \(D_{\text{eff}}=D_{\text{classical}}+\gamma(g)\), mixing the two axes. Part of why \(D\) runs is that \(C\) runs. The same \(\gamma\) is why the critical exponent \(\beta\approx0.326\) isn't a clean number.
  3. What does the complex dimension \(D=a+i\beta\) correspond to in the renormalization group?
    See the answer
    A complex renormalization eigenvalue around a fixed point. The phase of \(\Lambda^n=|\Lambda|^ne^{in\theta}\) produces an oscillation periodic in the coarse-graining count (= the logarithm of scale) ── the mark of continuous scale invariance falling to discrete scale invariance, with \(\lambda=e^{2\pi/\beta}\) as the rung spacing.
  4. State in one sentence the "form" shared by \(c\cdot t=\)constant and the Callan–Symanzik equation.
    See the answer
    You may choose the ledger (the value of \(c\) / where to cut, \(\mu\)) but you may not choose the answer (the dimensionless observables) ── and that demand itself becomes an equation that constrains the physics.

Bonus ⑥ summary / the linking episodeAll three were walking the same axis: scale

The \(D=-d\ln F/d\ln r\) of "The Universe Is a Computer" and renormalization's \(\beta(g)=dg/d\ln\mu\) are the same "logarithmic derivative with respect to scale." So Episode 10's "\(D\) runs" translates directly into renormalization group flow: the weak force's \(2\to\infty\) is the decoupling theorem, Episode 1's upper critical dimension is the boundary between the Gaussian and Wilson–Fisher fixed points, and Episode 8's complex dimension is a complex renormalization eigenvalue = discrete scale invariance. The one footnote this side can add: because of the anomalous dimension \(\gamma(g)\), \(C\) and \(D\) mix once you leave a free theory.

"Cosmology That Clicks" and its \(c\cdot t=\)constant share their argument with the Callan–Symanzik equation \(\mu\,d/d\mu(\text{observable})=0\) ── choose the ledger, not the answer. Its Bonus ⑥ split of "the time axis (gauge) / the energy axis (RG)" is this series' subject, and its Bonus ⑦ "two ends of the resolution knob" becomes, in Episode 7, the depth of the bulk.

And the three series had already crossed at the same three points ── the neutron (Cosmology Ep. 7 / Renormalization Ep. 2), Landauer (all three), and "a continuum limit exists only at a critical point" (Computer Ep. 9 / Renormalization Ep. 4). That last one writes down the central proposition of the renormalization group independently, from the computational side. And I have only now noticed that the finale of "Cosmology That Clicks" ended on "there it can no longer be renormalized" ── this series was the continuation of that door.

This document is Bonus Episode ⑥ (the linking episode) of the "Renormalization That Clicks" series, a reading piece for physics-loving high-schoolers and undergraduates. The beta function and anomalous dimension; the Callan–Symanzik equation (that renormalization-point independence of observables gives the renormalization group equation); the Appelquist–Carazzone decoupling theorem; the running of the electromagnetic coupling (\(1/\alpha\simeq137\), about 128 at \(M_Z\)); the upper critical dimension; the correspondence between complex renormalization eigenvalues and discrete scale invariance with log-periodic oscillation; that a continuum limit requires a fixed point; and Landauer's principle ── all established physics. That the correspondence between the \(D\) of "The Universe Is a Computer" and renormalization's scaling dimension is an isomorphism and not an identity (the exponent of a force's fall-off, a fractal dimension and a spectral dimension are distinct quantities that merely share the form "logarithmic derivative with respect to scale"); that \(c\cdot t=\)constant and \(\mu\)-independence agree in logical structure rather than in content; and that this episode is a linking episode making no new claim ── all spelled out in the body's "honest line." The figure draws \(D(r)\) for the four forces and the running of the electromagnetic coupling using models simplified for teaching (a Yukawa-type exponential cutoff for the weak force, an added string tension for the strong force, and a logarithmic interpolation for \(\alpha\) calibrated to give 1/128 at \(M_Z\)); it is not a precision calculation. ── To print, use your browser's "Print" and "Save as PDF" (in the print version the slider and answers are frozen and hidden). Sister series: Cosmology That Clicks / The Universe Is a Computer / Black Holes That Click. Related: Episode 4 / Episode 7 / Bonus ③ / Contents.

Print / make a PDF: ⌘+P (Ctrl+P on Windows). On screen, moving the cursor with the slider reads the same single point in three series' languages at once. "See the answer" opens each solution.