Mass That ClicksBonus ① / An honest diagnosis of a near-miss theory (revised)

Throughout the main series we kept saying "this is not a rounding error." This time we diagnose that near-miss theory head-on

Is the Mass Gap
a "Rounding Error"? "The universe is a finite computer, and rounding error becomes the mass gap" ── a near-miss theory.
Its motivation was sharp. Where it stumbled was mixing up "the resolution of which scale."

Tools you'll need: the dimensional transmutation of Episode 4, the IR floor of Episode 6, the CKN bound of the Finale What to implement is not rounding error but "symmetry and dimensionless quantities"

In the main series (especially Episodes 4 and 6), we repeatedly drew a line: "this is not a rounding error." This time we take that near-miss theory and, rather than hunting for a culprit, diagnose it precisely with the tools we've built up in this series. It's the same structure as when Bonus ③ of the sister series diagnosed that "variable-speed-of-light theory (VSL) had the right motivation but lost its footing in the implementation." To state the conclusion up front ── the core idea (finite resources = an upper limit on information) is genuine, but it mixed up UV and IR "resolution." This installment carefully untangles that mix-up.

01What the theory says ── and where it's sharp

The near-miss theory claims: "The universe is a finite-bit computer. When you compute the non-Abelian Yang–Mills field with finite bits, the rounding error (machine epsilon \(\epsilon\)) is amplified through self-loops and can't converge to zero, leaving a minimum footprint ── that is the mass gap \(\Delta>0\)." This deserves fair credit, because it has a sharp core.

02First, a correction ── this theory does NOT "fix" the lattice

The lazy criticism of the rounding-error theory goes: "Fixing a lattice breaks Lorentz symmetry." But this theory doesn't fix the lattice. The resolution (the lattice's fineness) goes as \(c\cdot t\) ── that is, it moves with the age of the universe. So that criticism, as stated, isn't fair. Let's plainly grant that.

However ── even a time-varying lattice does not automatically preserve local Lorentz invariance. This calls for a careful distinction.

Don't confuse two "frames"

The universe already has a preferred frame ── the CMB rest frame (comoving frame). This one was picked out by the matter distribution and does not violate the Lorentz invariance of the laws (harmless).
The problem is a different one. Tie a regular spatial lattice to cosmic time \(t\), and under a "local boost" the lattice looks anisotropic. Time-variation is a rebuttal to a "static absolute lattice," but not a rebuttal to a "local boost."

The interesting thing is that a Lorentz-invariant discretization really is possible. But it isn't a regular lattice ── it's a random scattering (causal set theory). Being random, on average it picks no direction and no frame, and stays invariant. But at that very moment, the computer picture ── "compute the field at each grid point and rounding error appears" ── no longer holds (because it's not a regular lattice). This is the surviving version of your intuition ── finite and discrete can be saved, but it stops being "rounding error."

03The real stumble ── mixing up UV and IR

This is the heart of the revised version. Even after granting the time-variation, one problem remains ── the scale being called "resolution" is actually two different scales.

One lattice can't do both jobs

The Lorentz-safe resolution that moves as \(c\cdot t\) = the IR floor (\(10^{-33}\) eV).
The lattice that produces the mass gap (gluons) = UV short-distance (GeV to Planck).
The two are more than 40 orders of magnitude apart ── "it's Lorentz-OK because it moves as \(c\cdot t\)" uses the IR, while "rounding error = gap" needs the UV. It's swapping in a different scale.

Figure: the axis of mass/energy (logarithmic). Move the slider on the scale being called "resolution," and you can only ever reach one side: the IR side (Lorentz-safe, moves as c·t = the floor from Episode 6) or the UV side (able to chop up gluons = the mass gap, but breaking local Lorentz). Across those 40 orders of magnitude, the two are incompatible
IR: Lorentz-safe (the floor from Episode 6) UV: chops up the gap but breaks local Lorentz

04So, gap ≠ neutrino still stands

The consequence of this mix-up is a confusion about what \(\Delta\) really is. The near-miss theory set "mass gap = the lightest particle = the neutrino mass." Three distinct quantities are crammed together here.

•
The Yang–Mills mass gap ~ 1.5 GeVThe pure-gauge glueball (strong-force sector, the \(\Lambda\) of Episode 4). The \(\Delta\) of the Clay problem.
•
The neutrino ~ 0.05 eVThe lightest actual "particle." Lepton sector; its origin is the seesaw of Episode 5 (\(v^2/M\)).
•
The IR floor ~ 10⁻³³ eVThe "smallest measurable" (Episode 6). 30 orders of magnitude below the neutrino.

"The lightest observed particle is the neutrino" ── that's correct. But it is not the IR floor (30 orders above) and it is not the Yang–Mills gap either (10 orders below, and in a different sector). Furthermore, if you compute the floor from the \(c\cdot t\) resolution you get \(10^{-33}\) eV, which doesn't reach the neutrino's \(0.05\) eV. The source of that \(0.05\) eV isn't resolution ── it's the seesaw of Episode 5. The three are different fields, different origins, different scales.

◇ ◇ ◇

05The dilemma, and a third path

The same unavoidable either-or appears as in VSL Bonus ③.

Path A: keep the discreteness as a "computational convenience"
Take the continuum limit \(a\to0\) properly. Lorentz stays intact; consistent with observation.
But ── that's ordinary lattice gauge theory itself, not a "new discovery."
→ Harmless, but not new
Path B: claim the regular-lattice discreteness is "real"
Don't take the continuum limit; insist "the universe really is discrete, and rounding error is mass." You can say you "solved it."
But ── the UV regular spatial lattice breaks local Lorentz and collides with experiment.
→ Bold, but harmful

The near-miss theory in effect chose Path B (or was indifferent to the A/B distinction) and paid Path B's penalty. But there was a third path in the main series ── the same "finite," depending on how you implement it, can survive.

Two ways to use "finite" without breaking anything

① CKN / holography (Finale): use finiteness as a "black hole = upper limit on information" → break neither Lorentz nor symmetry, and link UV and IR at the meV scale.
② Causal sets: make the discreteness a "random scattering" → preserve Lorentz invariance (but since it's not a regular lattice, the "rounding error" picture disappears).

The intuition you first grasped ── "the universe can only hold finite information" ── was correct. It's just that if you implement it as regular-lattice rounding error you lose your footing, whereas if you implement it as an upper limit on information (holography) or random discreteness (causal sets) it survives. What you must protect is not the discreteness of the implementation, but symmetry (Lorentz) and the running of dimensionless quantities ── this is the mass-side version of Bonus ③'s lesson, "protect \(\alpha\), not \(c\)."

✓
Core idea: finite resources / an upper limit on informationGenuine. The foundation of the Finale's CKN bound. Lives on as holography / causal sets.
✓
The resolution moves as c·t (time-variation)Correct. That's the IR floor of Episode 6, \(\hbar H/c^2\). Lorentz-safe.
✗
Chopping up UV gluons (the gap) with that IR resolutionA UV/IR mix-up. Treating two things 40 orders apart as one lattice.
✗
gap = neutrino mass = rounding errorConfusing three distinct quantities. The mechanism is not an implementation-dependent \(\epsilon\) but the dimensional transmutation of Episode 4 (fixed uniquely by the running of a dimensionless quantity).
An honest line ── so as not to gloat

This diagnosis isn't meant to knock down the rounding-error theory. Just as VSL was genuine research, this theory too shares the same core as the modern frontier of physics that captures the universe with finite information (holography, the swampland, causal sets). The lesson we could learn ── protect symmetry and dimensionless quantities, not the implementation, and don't confuse the UV and IR "resolutions" ── is what's valuable.

And as we saw in the Finale, even the "correct implementation" ── CKN / holographic dark energy ── still has unpaid homework (the equation-of-state \(w\) problem). It's not that one side is perfect. Every line of thought is carrying unpaid homework somewhere.

Practice problems (solvable with today's material)
  1. Is "a time-varying lattice preserves local Lorentz" correct?
    See the answer
    Not correct. The universe's CMB rest frame is a harmless preferred frame, but tying a regular spatial lattice to cosmic time \(t\) still makes the lattice look anisotropic under a local boost, breaking local Lorentz. Time-variation is a rebuttal to a "static absolute lattice," not to a "local boost." A Lorentz-invariant discretization is possible if it's a random scattering (causal sets), but that isn't a regular lattice and the "rounding error" picture no longer holds.
  2. What is "mixing up UV and IR"?
    See the answer
    The resolution that moves as \(c\cdot t\) and is Lorentz-safe is the IR floor \(\hbar H/c^2\sim10^{-33}\) eV (Episode 6). Meanwhile, the lattice needed to compute gluons = the mass gap via rounding error is UV short-distance (GeV to Planck). The two are more than 40 orders of magnitude apart, and one lattice can't do both jobs. "It's Lorentz-OK because of c·t" uses the IR, while "rounding error = gap" needs the UV ── a bait-and-switch.
  3. Name the ways to use the finite-resources idea "without breaking anything."
    See the answer
    (1) The Finale's CKN / holography (use finiteness as a "black hole = upper limit on information" → preserve Lorentz). (2) Causal sets (make the discreteness a random scattering to preserve Lorentz invariance). Neither is "regular lattice + rounding error."

Bonus ① summaryWhat you must protect is not the implementation but "symmetry and dimensionless quantities"

The core of the rounding-error theory (finite resources = an upper limit on information) is genuine, the same as the frontier. Moving the lattice as \(c\cdot t\) is correct too ── that's the IR floor of Episode 6. Where it stumbled was trying to chop up UV gluons (GeV to Planck) with that IR resolution (\(10^{-33}\) eV, Lorentz-safe) (a UV/IR mix-up). So push it with a regular lattice and you break local Lorentz (harmful); keep it at the continuum limit and it's just ordinary lattice gauge theory (harmless but not new). And it was confusing three distinct quantities into one (the gap \(\sim\)1.5 GeV / the neutrino \(\sim\)0.05 eV / the IR floor \(\sim10^{-33}\) eV).

The genuine mass gap is the dimensional transmutation of Episode 4 (fixed uniquely by the running of a dimensionless quantity); the correct use of finite information is the CKN / holography of the Finale, or causal sets (which don't break symmetry). Don't cling to implementation-dependent values; protect the symmetry and dimensionless quantities you must protect, and don't confuse UV and IR ── VSL's lesson holds, unchanged, for the mass gap.

This document is Bonus ① (revised) of the "Mass That Clicks" series, a piece for physics-loving high-schoolers and undergraduates. The following are all established physics / standard understanding: that the Yang–Mills mass gap (the Clay Millennium Prize problem) concerns the glueball mass of pure gauge theory (\(\sim1.5\) GeV scale, strong-force sector) and is distinct from the neutrino mass (\(\sim0.05\) eV, lepton sector, seesaw origin) and the IR floor (\(\sim10^{-33}\) eV, cosmological lower bound); that a regular spatial lattice breaks local Lorentz invariance (even when made time-varying) whereas causal sets based on random scattering are statistically Lorentz-invariant; that the cosmological comoving (CMB) rest frame is a preferred frame but does not break the Lorentz invariance of the fundamental laws; that dimensional transmutation gives a scale uniquely from the running of a dimensionless coupling whereas the rounding error of a discretization is implementation-dependent; and that the CKN / holographic handling of finite information is a framework that preserves symmetry. This piece has no intent to disparage any particular individual's theory; it appreciates the core idea (finite information) while diagnosing the implementation-level flaw of mixing up the UV/IR "resolution." ── 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: Ctrl+P (Cmd+P on Mac). On screen, move the slider on "resolution" and you'll see that the IR (Lorentz-safe) and the UV (chopping up the gap) can't coexist. Click "See the answer" to open each solution.