Continuous and discrete are two representations of the same physics ── the observables must agree in either one
In Episode 6 we found that mass has a lower wall ── \(m_{\min}\sim\hbar H/c^2\sim10^{-33}\) eV. And this floor, as the universe grows and \(H\) drops, drops as \(1/t\). Just as you said ── because the universe expands, the "smallest visible mass" is set over time, and drops. This time, next to this moving floor, we place a completely different "smallest mass" ── Yang–Mills' mass gap (the \(\Lambda\) born of confinement, which appeared in Episode 4). The two come out of the same equation, yet one moves and one does not. The neutron testifies to the difference. ── But this view of "two floors" is a story that draws the universe in one representation, the continuous one. Finally, we switch to another representation, the discrete, and check ── and we see that nothing that observation can catch changes at all. If it looks different, that is a word that dissolves into the representation.
A particle of mass \(m\) has a measure of its quantum-mechanical spread, the Compton wavelength \(\lambda_C=\hbar/(mc)\). The lighter it is, the longer the wavelength. Into this single equation, put two lengths.
① \(\lambda_C=\) the length of confinement (\(\approx1\) fm, proton size) → \(m\approx\) 0.2 GeV (the mass gap)
② \(\lambda_C=\) the cosmic horizon (\(c\,t\approx13.8\) billion light-years) → \(m\approx\) \(10^{-33}\) eV (the IR floor)
Let's check with numbers. Using the handy conversion \(\hbar c=197\) MeV·fm, the mass corresponding to \(\lambda_C=1\) fm is \(mc^2=\hbar c/\lambda_C=197\ \text{MeV·fm}/1\ \text{fm}\approx200\) MeV. That is exactly \(\Lambda_{\rm QCD}\), the scale of the strong force that confines quarks. On the other hand, putting in \(\lambda_C=c\,t\) returns Episode 6's \(m_{\min}=\hbar/(c^2t)=\hbar H/c^2\approx1.4\times10^{-33}\) eV.
These two are 41 orders of magnitude apart. It is the same single equation, "Compton wavelength = a length," yet whether the length you put in is inside (1 fm) or outside (the edge of the universe) makes this much difference. The structure of "the smallest mass always points to a largest," seen in Episode 6 §05, works here too ── the gap points to the inner confinement length, the IR floor to the outer horizon.
Here we recall the sister series' rule (Cosmology That Clicks, Bonus ④) ── "'the speed of light slows down / \(c\cdot t=\)constant' is a rephrasing of coordinates, and the physics is read in dimensionless quantities." So "visible vs. frozen" must also be judged not by the dimensionful \(m\) or \(t\), but by a dimensionless ratio. That ratio is this.
\(N>1\): it can oscillate before the universe changes = visible, distinguishable from zero.
\(N<1\): pinned by Hubble friction, it freezes = indistinguishable from zero.
The boundary \(N=1\) is exactly \(m_{\min}=\hbar H/c^2\).
This is Episode 6's equation of motion \(\ddot\phi+3H\dot\phi+(mc^2/\hbar)^2\phi=0\) itself ── whether the restoring-force rate \(mc^2/\hbar\) can beat the Hubble friction \(H\). And because \(N\) is dimensionless, whether you read it in a gauge where \(c\) varies or as space stretching, its value does not change. Episode 6 §06's "trap of units" can be avoided from first principles. Use \(t\) naively as a clock and you fall into the trap. The right thing is to read it with the ratio to \(H\), namely \(N\).
The line \(N=1\), that is \(m_{\min}=\hbar H/c^2\), drops as \(H\) drops. The younger the universe, the larger \(H\) was (\(H\approx1/t\)), and the higher the floor. Let's line up the numbers.
Put \(m_{\min}=\hbar H/c^2\), \(H\approx1/t\) into each era
$$\text{BBN}\ (t\approx100\,\text{s}):\quad m_{\min}\approx\hbar/t\approx6.6\times10^{-18}\ \text{eV}$$ $$\text{now}\ (t\approx4.4\times10^{17}\,\text{s}):\quad m_{\min}\approx1.5\times10^{-33}\ \text{eV}$$At the time of BBN the floor was 15 orders of magnitude higher than now. As the universe grew, the floor dropped and dropped, and ever-lighter masses became "visible" ── expansion is an ally of low mass (Episode 6 §06). This is the moving floor.
Then, the other floor ── how does the mass gap move over time? The answer is it does not move. The reason lies in the dimensional transmutation we saw in Episode 4. The strong force's coupling runs with distance, and from a theory with classically no scale (scale invariant), a single scale \(\Lambda\) wells up through quantum effects.
Into this equation, neither the age of the universe \(t\) nor the expansion rate \(H\) enters at all. The gap is an intrinsic property of the theory itself, in an infinitely wide, flat spacetime (the arena the mathematicians ask about). So even if the universe expands, \(\Lambda\) does not move. Seen through the dimensionless \(N\), this is obvious at a glance.
\(N_{\rm gap}=(\Delta/\hbar)/H\), \(\Delta=0.2\) GeV → \(\Delta/\hbar=3.0\times10^{23}\) s\(^{-1}\)
$$\text{now}:\ N_{\rm gap}=\frac{3.0\times10^{23}}{2.2\times10^{-18}}\approx1.4\times10^{41}$$Today's \(N_{\rm gap}\approx10^{41}\) is the same number as the "41-order ratio" from \(\S01\) (naturally, since \(N_{\rm gap}=\Delta/m_{\min}\)). Even at the time of BBN, \(N_{\rm gap}\approx10^{25}\). In every era of the universe, the gap has been far above the floor ── it has never once come close to the floor.
The gap touches the floor (\(N_{\rm gap}=1\)) when \(H=\Delta/\hbar\), that is, at \(t=\hbar/\Delta\approx3\times10^{-24}\) seconds. This is when the whole universe was a mere 1 fm ── even earlier than the QCD phase transition (\(\sim10^{-5}\) seconds), far before BBN (100 seconds), completely outside the region we can observe. So throughout the entire observable history of the universe, the gap has remained an unmoving floor.
"The gap does not move" ── can this really be confirmed by observation? It can. And the sister series "Cosmology That Clicks," in its Episode 7, has already done it. A handshake between the same episode numbers.
The mass-gap scale \(\Lambda_{\rm QCD}\) sets the masses of the proton and neutron, and the \(Q\)-value of neutron decay (the mass-energy difference between the neutron and proton + electron). If the gap had been dropping with the universe as \(1/t\), then both \(\Lambda_{\rm QCD}\) and \(Q\) would have differed by orders of magnitude between the time of BBN and now. And yet ──
The amount of helium produced 13.8 billion years ago requires that the \(Q\) at the BBN era agreed with today's to within \(0.34\%\) (with the fifth-power amplification of the decay rate, the window is this narrow).
In other words, \(\Lambda_{\rm QCD}\) (= the mass-gap scale) has not moved even 0.34% over the 13.8 billion years since BBN. Had it been dropping as \(1/t\), it would have changed by orders of magnitude and the universe's helium would not have remained at today's amount ── the neutron and helium reject that. Neutron decay is a testimony, from real data, that the mass gap is an intrinsic floor that does not ride on cosmic time.
So far we have drawn the universe in one representation, the continuous one. Here we recall the foundation of this series (Cosmology That Clicks, Bonus ④) ── just as "space stretches" and "the speed of light slows" were equivalent, continuous and discrete too are nothing but two representations of the same physics. If so, what observation catches (the observables) must agree exactly whichever representation you compute in. If the conclusion changes in only one of them, that is not physics but a word that dissolves into representation.
The most certain example is the mass gap itself. Compute Yang–Mills' gap \(\Delta\) by carving space into a lattice (discrete) and sending the lattice spacing to zero (the continuum limit), and ── you get the same value as the \(\Delta\) computed in the continuous representation. The discrete is a regularization = a scaffold for the computation (a representation), and \(\Delta\) is an invariant independent of representation. This is exactly why lattice QCD is trusted (Wilson's lattice gauge theory, seen in this series' Bonus ④ "The four forces in one discrete equation"). Discrete or continuous does not change the answer.
The continuous and discrete representations are rephrasings of the same physics. The observables agree in both representations. If it looks different, that is a word that dissolves into representation ── not physics.
With this single line, let's sort the items that appeared in \(\S01\)–\(\S05\). First, the things that are necessarily the same in continuous and discrete ── these are physics.
| Quantities independent of representation (= physics) | Why they are representation-independent |
|---|---|
| The threshold \(\hbar H/c^2(t)\): masses below it do not affect observation, since \(N=(mc^2/\hbar)/H<1\) | The dimensionless \(N\) (whether it can oscillate once in the age of the universe) is representation-invariant. The same in continuous or discrete. |
| The YM mass gap \(\Delta\sim\Lambda\) | Lattice (discrete) → continuum limit gives the same value as the continuous representation. That's why lattice QCD works. |
| \(\Delta\) remains as \(L\to\infty\), while the mode spacing \(\hbar c/L\) vanishes | A statement about how the limit is taken; it is representation-independent (= the true nature of the difference between the moving and unmoving floors) |
| The \(1/t\) dependence of the threshold | The contents of \(N\) behave that way. "Drops with expansion" is the same in both representations. |
On the other hand, the things that differ only in wording between continuous and discrete ── these are not physics, and dissolve into representation.
| Questions that dissolve | Continuous representation's wording | Discrete representation's wording |
|---|---|---|
| What is below the threshold? | continuous but frozen and inactive | no rung |
| What is the IR floor's status? | an operational limit | a really existing gap |
| Spacetime itself? | a continuum | a lattice |
These three cannot be distinguished by any observation ── because the threshold is that very resolution. Whether you say there is a "frozen continuous spectrum" below the threshold or say "there is no rung," the observed result is not one whit different. So the phrasing "make it discrete and the floor turns into a 'really existing gap'" is superfluous ── all that changed was the words, and the physics (= what is below the threshold does not affect observation) is the same in both representations. Swapping a difference of representation for a difference of physics is exactly the trap this series' Bonus ① (diagnosing tantalizing-but-wrong theories) warned against.
So the correct meaning of the "single ladder" is this ── a single map that lines up representation-independent landmarks on a log scale. Whether you draw it continuous or discrete, the landmarks stand in the same places.
And on this map, what \(t\) moves is only the lower end (the threshold); the YM gap \(\Delta\) is immovable (the neutron + BBN testify to \(0.34\%\)) ── this conclusion is exactly identical in the continuous and discrete representations. The difference between the two floors (moving / unmoving) comes not from "continuous vs. discrete" but from a representation-invariant distinction: "a lower end that vanishes as \(L\to\infty\), or a middle rung that remains." Not confusing this point is the heart of the whole sorting.
The core of this episode is that "the conclusion is the same whether continuous or discrete" ── so you may safely tell it in the easier-to-grasp representation (the "freedom of viewpoint" that Cosmology That Clicks, Bonus ④, guarantees). But what the map unifies is the "place" of the landmarks, not their "origin." The lower end (threshold, IR, vanishes in the counting limit) and the middle rung (gap, UV, remains by dynamics) are different physics, even if they line up on the map. Lumping them together as "both are gaps born of the discrete/finite" is a confusion to be avoided.
A remaining caveat ── the mathematical proof of existence of the Yang–Mills mass gap is unsolved, as a Clay Millennium Prize Problem (as of 2025). It is not "a gap is visible on the lattice" but a proof that "the invariant \(\Delta>0\) remains in the continuum limit," which is exactly the remaining hard part. And whether "a finite integer number of states" (Technical Appendix ⑮) is really true is an unproven wager, and even if it is true, what it changes is other observables like recurrence and unitarity, not the values of the threshold or \(\Delta\). We keep the closing chapter's rule "if a chat says it's 'solved,' be suspicious" here too.
The same single equation, "Compton wavelength = a certain length," gives birth to two "smallest masses." Put in the inner 1 fm and you get the mass gap \(\approx0.2\) GeV; put in the outer horizon \(c\,t\) and you get the IR floor \(\approx10^{-33}\) eV ── two floors 41 orders of magnitude apart. The IR floor is a moving floor that drops as \(1/t\) together with \(H\); the mass gap is an unmoving floor set by dimensional transmutation \(\Lambda\), and the neutron + BBN testify to its immobility at \(0.34\%\).
And the crucial point ── this conclusion is exactly the same whether you draw the universe continuous or discrete. Continuous and discrete are two representations of the same physics, and the observables must agree (Cosmology That Clicks, Bonus ④). \(\Delta\) is invariant in the lattice → continuum limit, and the threshold, \(\Delta\), meV, and Planck are landmarks independent of representation. "Make it discrete and the floor turns into a really existing gap" was nothing but a word that dissolves into representation. What the map unifies is the place of the landmarks, not their origin, and the lower end (IR, counting) and the middle rung (dynamics) are different things ── we won't overclaim. Clay's continuum limit is left open, and "a finite integer number" is a wager that affects other observables.
Print / save as PDF: Ctrl+P (⌘+P on Mac). On screen, moving the "now" age of the universe with the slider shows only the indigo floor (the IR floor) dropping, while the rust-red floor (the mass gap) stays put. "See the answer" opens each solution.