Five Sigma of Daylight
Two rigorous rulers measure how fast the universe expands — the early-universe CMB and the local distance ladder — and their error bars don't touch. Our own-code stats put the gap at five sigma, far too big to be chance. But the cause — an uncaught systematic, or new physics — is genuinely unresolved. We report the gap at its true width and leave the question open.
Ask the universe how fast it is expanding and you get two answers. One comes from the infant cosmos— the microwave afterglow of the Big Bang, read through our best model of physics — and says 67.4. The other comes from the nearby, present-day universe— a ladder of stars and exploding suns whose distances we measure almost directly — and says 73.0. Each comes with a small error bar. They do not overlap. Both camps have checked their work for a decade, and the gap has only gotten harder to explain away.
Where we land:held open. The five-sigma gap is real and settled; its cause — an uncaught systematic or new physics — is not. Reported at its true width.
1 · Two rulers, two numbers
The Hubble constant H₀ is the expansion rate — kilometres per second of recession for every megaparsec of distance. The early-universe value (Planck / CMB) and the late-universe value (SH0ES / Cepheids+SNe) are measured in completely different ways, with completely different possible mistakes. That’s what makes their disagreement so serious: there’s no shared error to blame. Plotted as probability curves, they sit in different places on the axis, with clear air between them:
2 · It is not a coincidence
Could two honest measurements land this far apart by pure chance? Combine their error bars and the separation is 5.0 standard deviations. The probability of a fluctuation that large is about 5.5×10⁻⁷ — roughly 1 in 1,825,729. In physics, 5σ is the threshold to claim a discovery. By that same standard, the disagreementis a discovery: something real is going on. The two numbers are not both allowed to be right about a single universe with a single expansion rate — unless a hidden ingredient is missing.
When two careful rulers disagree by five sigma, one of them is wrong — or the thing you assumed they were both measuring isn’t what you thought it was.
3 · Two kinds of answer, neither in hand
The resolution is one of two shapes, and after years of work neither has arrived. Both are live; nothing here picks between them.
Systematic error
A subtle bias nobody has caught — in the Cepheid distance ladder, or in how the CMB is modelled. If so, the crisis dissolves into a mistake, and the standard model survives untouched.
New physics
Both measurements are right, and the bridge between the infant universe and today — the standard cosmological model — is missing something: early dark energy, a new relativistic particle, evolving dark energy, a tweak to gravity.
Tellingly, the ladder has survived its hardest test: JWST re-checked the Cepheid distances and confirmed them(Riess et al.), closing that escape hatch on the Cepheid side. The tension didn’t break — though independent JWST programs (Freedman’s CCHP, via the TRGB) land nearer Planck, so JWST has not settled it either way.
Verdict — held open
Both measurements are rigorous and reproducible; the 5.0σ gap between them is real, not chance. What it means— an uncaught systematic, or new physics beyond the standard model — is unresolved, and actively fought over in the current literature. Nothing here is hidden: the data is public, the methods are open, and the answer still isn’t in. Dark Math reports the gap at its true width — five sigma of daylight between two views of the same sky — and marks the cause as one of the biggest open questions in physics today.
Why our math sees more
A single headline number (“the universe expands at ~70”) hides the whole story; the story is the disagreement. Structure-first analysis treats each measurement as an invariant with an honest error bar, quantifies the tension between them (5σ, not a fluke), and refuses to average the two into a false peace or to crown one without cause. The gap is the finding — and naming it precisely, rather than smoothing it over, is exactly the discipline this whole series is built on.
Sources & method
measurements — Planck 2018 (Planck Collab., A&A 641, 2020): H₀ = 67.4 ± 0.5. SH0ES: Riess et al., ApJL934 (2022): 73.0 ± 1.0; JWST confirmation: Riess et al. (2023–24). TRGB: Freedman et al. (CCHP, 2019–21) ~69.8 ± 1.7.
overview — Hubble tension (Wikipedia) (third-party overview — linked, not re-hosted)
computed here — tension = ΔH₀ / √(σ₁² + σ₂²) = 5.6 / 1.12 = 5.01σ; P(chance) = 2(1 − Φ(5.01)) = 5.5×10⁻⁷ via math.erf; curves = normal densities. Own-code, stdlib math only.
method own-code tension in sigma · chance-probability via erf · non-overlapping error bars · systematic vs new-physics fork
ethos the gap is the finding · don’t average two rulers into a false peace · a 5σ disagreement is itself a discovery · earned vs reaching
Dark Math The Observer’s Index — dark = the consistent, light = the medium of observation. Release 012 · held open — the gap is the finding.