Cosmology That ClicksBonus 5 / Standing on the "moving" side (Trilogy, Part 1)

Bonuses 3 and 4 were about "protecting α." This time we flip it: once you decide to move α, we count what else you can still hold fixed.

If You Move α,
What Can You Hold Fixed? "c = constant" and "c·t = constant" were the same physics — they just fixed different things.
So one level down: once you decide to move α itself, how much freedom is left?

Tools you'll need: "only dimensionless numbers are physics" from Bonus 2, and division Of the 4 constants, at most 3 can be fixed

The bonus issues so far have been written from the side that says "α is something to protect." In Bonus 3, VSL failed because it couldn't protect α; in Bonus 4, we saw that α being invariant is the very core of the equivalence condition. This time we turn the viewpoint inside out: if we really decide to move α, how much "freedom to hold things fixed" do we have left? The answer can be counted with surprising precision — and it turns out to be a reprise, one level down, of this series' watchword: "c = constant and c·t = constant are the same thing; they just fix different quantities." This is part one of a trilogy.

01First, count — the 4 constants that build α are worth only "3 people"

Let's write the fine-structure constant in the form from Episode 2.

The 4 dimensionful constants that build α
$$\alpha = \frac{e^2}{4\pi\varepsilon_0\,\hbar\,c}\approx\frac{1}{137}$$

The ingredients are the four quantities \(c,\ e,\ \varepsilon_0,\ \hbar\). And yet, out of these four, there is only one "number with the units gone" you can build: \(\alpha\). This is no accident — it has a deep meaning. Against the four kinds of units (length, time, mass, charge), these four constants carry only three units' worth of independent information. So the leftover one unit's worth spills over as a "relation with the units gone" — that is \(\alpha\).

In other words — why "3 people's worth"? If the four constants were truly independent (four kinds' worth), no dimensionless number could be built and \(\alpha\) wouldn't exist. \(\alpha\) exists precisely because there is one constraint among the four constants (a combination in which the dimensions cancel out exactly). One constraint means you're free to choose at most 4 − 1 = 3 of them.

From here comes the iron rule of fixing.

The heart of this issue (the iron rule of fixing)

Of \(c,e,\varepsilon_0,\hbar\), the number you can freely fix by convention (units, gauge) is at most 3. The remaining one is always dragged along by \(\alpha\) and moves.
So once you've decided to "move \(\alpha\)," this "one that moves" can never be erased.

02c is already fixed — so what's left is "at most 2"

From the very beginning, this series has spoken in a coordinate frame where \(c=\text{constant}\). In other words, we've already spent one fixing-slot on c. By the iron rule, then, we can additionally fix at most 2 of \(\{e,\varepsilon_0,\hbar\}\). The third always moves, and that "one that moves" is exactly the bookkeeping face of "α changing." Depending on whom you assign to be the mover, the same physics can be written three ways.

What you fix (= the yardstick)The moverWhat it means
c, ε₀, ħe (∝√α)A theory where "electric charge changes over time." Bekenstein-type varying-e is this one.
c, e, ħε₀ (∝1/α)A picture where "the vacuum's response (how hard it is for electricity to get through) changes."
c, e, ε₀ħ (∝1/α)A picture where "the quantum step size changes."

Whichever you choose, the colors of atoms and the rhythms of clocks — that is, the observations are completely identical. The only difference is the bookkeeping choice of "whom to make the mover." In the figure below, drag to watch how, when \(\alpha\) grows a little, each of the three movers shifts (and yet \(\alpha\) stays the same).

Figure: As α grows, e increases slightly (∝√α) while ħ and ε₀ decrease (∝1/α). The three ledgers look different, but reassembled they always give the same α.
α (what we move — the real physics) e = as mover, ∝√α ħ = as mover, ∝1/α ε₀ = as mover, ∝1/α
To be honest — ε₀ is a "bookkeeping role," and the real candidate is e \(\varepsilon_0\) (the permittivity of the vacuum) is really an SI-unit bookkeeping quantity: in Gaussian units it becomes \(\alpha=e^2/\hbar c\) and drops out of the formula. So the "candidates you'd actually move" are \(e\) or \(\hbar\). The historical varying-α theory (Bekenstein) took the most straightforward form: moving the charge \(e\). The picture where you move ε₀ is just the same thing as moving e, said in a different ledger.

03This is a one-level-down reprise of "c = constant vs. c·t = constant"

What I want you to notice here is that what just happened is a strikingly close, scaled-down copy of the structure we saw in Bonuses 3 and 4. Back then it was "fix \(c\), or fix \(c\cdot t\) — same physics, just different things fixed." This time we go one level down: "fix \(e\), or fix \(\hbar\), or fix \(\varepsilon_0\) — same physics, just different things fixed." "Whom you make the mover is gauge (a rebalancing of the ledger); what shows up in observation is only the dimensionless \(\alpha\)" — the same single principle keeps showing its face over and over, at one hierarchy level after another.

A connecting voice — Duff's "there are zero fundamental constants" One physicist pushed this counting to its limit. Michael Duff argued that dimensionful constants are essentially none of them "fundamental" (= they are all conventions of units), and that only dimensionless ratios are physics (see also the reading list at the end of the bonus issues). This issue's "at most 3 can be fixed, the one that moves is a choice of ledger" is exactly a small, hands-on instance of that stance.
◇ ◇ ◇

04But this time it isn't just "gauge" — α moving is real

Here is the decisive difference from Bonuses 3 and 4, and the point you absolutely must not miss. In Bonus 4's \(c\cdot t=\text{constant}\), \(\alpha\) was invariant. So that was pure gauge (freedom of viewpoint), with zero observational difference — it was "the same physics." But this time, \(\alpha\) itself moves. The moving of \(\alpha\) is a genuine physical change that no choice of units or gauge can erase. It leaves traces in observation: atomic clocks tell us "α has moved by less than \(10^{-17}\) per year," and \(\dot\alpha/\alpha\) is still being hunted in the absorption lines of distant quasars.

A correction of wording

"Consider a coordinate frame in which \(\alpha\) changes" should, more precisely, be —
"Add one new piece of physics (= assume α moves), then choose which constant to blame it on."
The coordinate = gauge freedom that remains is only in "which constant to blame it on"; the fact itself that "α moves" cannot be erased in any coordinate frame.

The honest line — don't fix too much / it isn't necessarily only α

First, you cannot fix all four of \(\{c,e,\varepsilon_0,\hbar\}\). Freezing all of them would force \(\alpha=\text{constant}\), which contradicts the premise "move α." You can fix at most three.

Also, \(\alpha\) is only one of the dimensionless numbers. There's also the proton-to-electron mass ratio \(m_p/m_e\), the strong coupling, gravity's \(\alpha_G=Gm^2/\hbar c\), and more — there are several "numbers with the units gone." This time we assumed "only \(\alpha\) moves, all other ratios fixed," but every extra dimensionless number you let move adds one independent new piece of physics, each carrying its own separate observational constraints. You don't get it for free.

Practice problems (solvable with this issue's formula alone)
  1. Having fixed \(c\), you also fix \(e\) and \(\hbar\). How does the remaining \(\varepsilon_0\) move in relation to \(\alpha\)?
    See the answer
    In \(\alpha=e^2/(4\pi\varepsilon_0\hbar c)\), with \(e,\hbar,c\) fixed, \(\alpha\propto 1/\varepsilon_0\). That is, \(\varepsilon_0\propto 1/\alpha\). If α increases, ε₀ decreases.
  2. Why is "fixing all four of \(c,e,\varepsilon_0,\hbar\)" no good? In one sentence.
    See the answer
    Fixing all four automatically pins \(\alpha\) to a constant too, contradicting this issue's premise of "moving α." You can fix at most three (since there is one constraint).
  3. Bonus 4's \(c\cdot t=\text{constant}\) versus this issue's "move α" — what's the difference in whether it shows up in observation?
    See the answer
    In Bonus 4, α is invariant, so it's pure gauge = zero observational difference (a restatement of the same physics). This time α itself moves, so it's a genuine physical change = it leaves traces in observation (atomic clocks, quasars). The gauge freedom remains only in "which constant to blame it on."

SummaryYou can fix three; the one that moves is a choice of ledger

The four constants \(c,e,\varepsilon_0,\hbar\) that build \(\alpha\) are, dimensionally, only "3 people's worth" independent (one constraint = that's α). So by convention you can fix at most three; the remaining one always moves along with α. In this series, where c is already fixed, you can additionally fix at most two of \(\{e,\varepsilon_0,\hbar\}\). Whom you make the mover is gauge (ledger), and varying-e, varying-ε₀, and varying-ħ are all restatements of the same physics — this was a one-level-down reprise of "c = constant vs. c·t = constant."

The decisive difference from Bonus 4, however, is that \(\alpha\) itself moves. That is real physics that gauge can't erase, and it shows up in observation. So the right phrasing is not "a coordinate frame where α moves" but "choose which constant to blame the physics of moving α on." The freedom you can fix is three, but the fact that α moves can't be erased in any coordinate frame — this is the starting point of the trilogy.

This document is Bonus 5 of the "Cosmology That Clicks" series, a reading piece for physics-loving high-schoolers and undergraduates. The four dimensionful constants that build \(\alpha=e^2/4\pi\varepsilon_0\hbar c\) span three independent dimensions, and only one dimensionless combination — \(\alpha\) — arises; therefore at most three can be fixed by the freedom of units and gauge. This is a consequence of dimensional analysis. \(\varepsilon_0\) is a quantity specific to SI units and does not appear in Gaussian units. The stance that the values of dimensionful constants depend on unit conventions and that the physical content lies in dimensionless quantities is laid out in Duff, Okun, and Veneziano, "Trialogue on the number of fundamental constants" (2002). The time variation of \(\alpha\) is constrained by atomic clocks and the like to about \(|\dot\alpha/\alpha|\lesssim 10^{-17}\)/year, and claims of variation in distant quasars remain under debate. — To print, use your browser's "Print" and "Save as PDF" (in the print version, the slider and answers are static and hidden).

Print / PDF: ⌘+P (on Windows, Ctrl+P). On screen, the slider lets you see that the same α can be written in three ledgers. Click "See the answer" to open a solution.