Mass That ClicksBonus ③ / The series' final piece · the temptation of numerology

A match too beautiful, accurate to one part in 100,000 ── but it's not a "prediction"

Generations and the Koide Formula The mass ratio of the electron, muon, and tau somehow lands exactly on \(2/3\). It gives you chills.
Savoring the match, we diagnose why it is not a "prediction" (a circularity), with the same honesty as Bonus ①.

Tools you'll need: Episode 4's "running mass," Bonus ①'s method of diagnosis Beauty is the doorway to a hypothesis, not evidence

The series ends with the most too-beautiful temptation. Plug the masses of the three generations of charged leptons ── electron, muon, tau ── into a certain simple formula and it lands exactly on \(2/3\). And to an accuracy of one part in 100,000. It gives you chills. A conversational AI of the past praised this as "a prediction come true!" But ── this is not a prediction. We'll see why, with the same honesty as Bonus ① (the diagnosis of the rounding-error theory). This is a story about "how to tell numerology from physics," a fitting close for this series.

01The Koide formula ── yes, too beautiful

It's a formula found by Yoshio Koide in 1981. For the charged-lepton masses \(m_e,m_\mu,m_\tau\) ──

Let's try it ── plug in the measured values $$Q=\frac{m_e+m_\mu+m_\tau}{\big(\sqrt{m_e}+\sqrt{m_\mu}+\sqrt{m_\tau}\big)^2}$$

Measured values (MeV): \(m_e=0.511,\ m_\mu=105.658,\ m_\tau=1776.86\)

$$Q=\frac{1883.03}{(53.15)^2}=\frac{1883.03}{2824.6}\approx 0.66666\approx\frac{2}{3}$$

It agrees with the theoretical value \(2/3=0.66666\ldots\) to five decimal places. This is a genuine surprise ── far too clean to write off as "mere coincidence."

First, let's give it fair credit. Koide's discovery is a genuine surprise, still seriously researched, with ongoing attempts to derive it from grand unified theories or symmetry. Here it's the same as VSL and the rounding-error theory in Bonus ① ── at its core there really is something that draws people in.

The geometric face Koide's \(Q=2/3\) is actually equivalent to a geometric condition: "three vectors whose lengths are the \(\sqrt{m}\) fan out neatly at \(120^\circ\) to one another." That's what makes it feel all the more like "there's some structure here." This is the source of its beauty.

02But this is not a "prediction" ── the circularity trap

This is the heart of the diagnosis. A prediction is "hitting something you didn't put in." The Koide formula just plugs in three measured masses and confirms that one relation holds. None of the masses are derived from first principles, and it gives no reason "why \(2/3\)."

The difference between prediction and confirmation

Prediction = hitting a quantity that is not in the input.
Koide = confirming a relation from the input (the three measured masses). The origin of the value \(2/3\) is unexplained.

The AI of the past said, "plug in the measured values and you get \(0.6666\) ── a prediction come true!" But that's a circularity ── all it says is that a ratio built from the measured masses came out to \(2/3\); it didn't predict the masses. Let's confirm this "circularity" with our own eyes in the figure below.

Figure: fix \(m_e,m_\mu\) to their measured values, move only \(m_\tau\), and watch \(Q\). \(Q\) changes with \(m_\tau\) and crosses exactly \(2/3\) at the measured \(m_\tau=1776.86\) MeV. It's not that the formula determined the mass ── it's that the measured value happens to sit there
Move m_τ and Q moves.
Q(m_τ) 2/3 (the target value)

\(Q\) varies smoothly with \(m_\tau\), and \(2/3\) is just one point along the way. The measured tau happens to land there ── the formula did not "predict" the tau's mass. Since the other two also use measured values, this is not "hitting the target" but "it happened to agree."

03Numerology or physics ── how to tell

Whether a "beautiful match" is a law of physics can be told apart with a few questions. Let's run Koide through them.

✗
Is there a mechanism (an established theory that derives why that value)?There's still no accepted theory that derives \(2/3\) (Koide's model is unestablished).
✗
Does it hit a quantity not put in?No. It just plugs in three measured masses and confirms one relation.
✗
Is it robust to scale (mass runs ── Episode 4)?Mass runs with energy. Depending on which fineness of mass gives \(2/3\), the exactness shifts ── "exactly 2/3" is scheme/scale dependent.
✗
Does it generalize elsewhere?It fits charged leptons, but doesn't extend cleanly to quarks or neutrinos (requires tuning).

On top of that, there's the look-elsewhere caveat ── make enough dimensionless combinations and some will inevitably fall near a simple fraction like \(2/3\). So "being close to a simple value" is not, by itself, evidence of physics.

◇ ◇ ◇

04And why are there three generations at all?

Behind Koide lies a bigger unsolved problem: "why are there three copies (generations) that are identical except for mass?" Let's separate what's known from what isn't.

On "plugging in g=4 or 0" Playing by plugging \(4\) or \(0\) into the Koide formula or the generation index is fun, but the Koide formula is not a "law that forbids a fourth generation." A fourth generation is strongly constrained by experiment (\(Z\)-width and Higgs data), but that's an empirical constraint, not something derived from Koide's numerology ── here too, the honesty of "not loading the formula with too much meaning" is required.

05How to savor the temptation properly ── recovering the backbone

This series' backbone has been consistent ── physics is not the surface value but the invariant structure behind it (symmetry, running, mechanism). Koide's \(2/3\) is, unmistakably, a "beautiful surface value." It becomes a law when a mechanism is found and it hits a quantity not put in. Until then, it's a captivating puzzle.

This series' discipline (the same as Bonus ①)

Enjoy it. Get chills. But don't leap to "it predicted this" or "this is the source code."
Beauty is the doorway to a hypothesis, not evidence.

It's exactly the same shape as the rounding-error theory of Bonus ① ── the core is interesting, but don't lose your footing in the implementation or the claim. It's the mass-side close of what the sister series "Cosmology That Clicks" kept saying: "intuitiveness is a projection; only the ratio is real." When you meet numerology, neither kick it aside nor swallow it whole ── run it through these four questions (mechanism, prediction, scale, generalization). That is the most practical tool this series can hand you.

An honest line ── so as not to disparage numerology

Dismissing Koide as "nonsense" is also a mistake. Koide's match is sharp in a genuine way, and serious attempts to derive it from GUTs or extra dimensions continue today. History has examples that went from numerology to real physics ── but that was when they got a mechanism and a prediction. At present, no one can derive "why 2/3" for Koide = it's not a law but a puzzle, which is its accurate standing.

This time too, unrelated to the watchword \(c\cdot t=\text{constant}\). It was purely about "how to read a match."

Practice problems (solvable with today's material)
  1. In one line, why is the Koide formula "not a prediction"?
    See the answer
    It plugs in three measured masses and confirms that one relation (\(2/3\)) holds; it derives none of the masses from first principles and has no reason (mechanism) for \(2/3\). It hits no quantity not in the input ── so it's not a prediction but a puzzle.
  2. Name two questions for telling whether a "beautiful match" is a law of physics.
    See the answer
    (1) Is there a mechanism (an established theory that derives why that value)? (2) Does it hit a quantity not put in (a true prediction)? On top of these, robustness to scale (mass runs) and generalization elsewhere also count. Koide satisfies none / is uncertain on all.
  3. Regarding there being three generations, what is known and what isn't?
    See the answer
    Known: there are three species of light neutrino (from the \(Z\)-width measurement), and anomaly cancellation closes generation by generation. Not known: why exactly three generations. The Koide formula is not a law that forbids a fourth generation, and the constraint on a fourth generation is experimental.

Bonus ③ summaryBeauty is the doorway to a hypothesis

The Koide formula hits the charged-lepton mass ratio at \(2/3\) to an accuracy of \(10^{-5}\), a genuinely beautiful match. But it's confirmation, not prediction (plug in three measured masses and read off the relation ── circular). No mechanism, hits no quantity not put in, scale-dependent, hard to generalize ── it passes none of the four questions. Why there are three generations is also unsolved, and Koide doesn't explain it.

When you meet numerology, neither kick it aside nor swallow it whole ── run it through the four questions (mechanism, prediction, scale, generalization). Beauty is the doorway to a hypothesis, not evidence ── with the same discipline as the rounding-error theory of Bonus ①, this series comes to a close.

Mass That Clicks ── on the completion of the series From the single question "what is weight?", to the floor you can't erase (Episode 1), the symmetry that protects it (Episode 2), the two ways it wells up (Episode 3), the emergence of scale and the mass gap (Episode 4), the UV and IR where the smallest mass points to the largest (Episodes 5 & 6), and finally the meV where smallest and largest shake hands (the Finale) ── we've walked this far on a single backbone. Mass is "a floor of energy you can't erase"; zero is protected by symmetry, a finite value is born of "running," its minimum is set by the size of the universe, and the minimum always points to the maximum. The three bonus installments honestly diagnosed the temptations we deliberately kept out of play on that journey (rounding error, relativistic mass, numerology). Intuitiveness is a projection, \(c\cdot t=\text{constant}\) is a coordinate restatement, use things honestly only on the ground where they apply ── hold to this discipline and, from the single word "weight," you can walk on your own two feet all the way to the edge of the universe. The same single line that runs through the sister series "Cosmology That Clicks" ran through to the end here too. Thank you for the journey this far.
This document is Bonus ③ (the final piece) of the "Mass That Clicks" series, a piece for physics-loving high-schoolers and undergraduates. The following are all standard understanding: that Koide's formula \(Q=(m_e+m_\mu+m_\tau)/(\sqrt{m_e}+\sqrt{m_\mu}+\sqrt{m_\tau})^2\approx2/3\) (Yoshio Koide, 1981) holds to extremely high accuracy for the charged-lepton masses, but that it has no established mechanism and is a relation taking measured values as input, not a prediction from first principles; that its exactness depends on scale/scheme due to the running of mass under the renormalization group; that its generalization to quarks and neutrinos is not automatic; and that the number of generations being three (three light neutrino species from the \(Z\) decay width, anomaly cancellation per generation) is a fact whose reason is unexplained. This piece has no intent to disparage Koide's match; it acknowledges its appeal while diagnosing the "difference between a match and a prediction." The figure is a schematic of \(Q\) as \(m_\tau\) is varied with \(m_e,m_\mu\) fixed to measured values. ── 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 m_τ with the slider and you'll see Q land on 2/3 only at the measured value. Click "See the answer" to open each solution.