Tired Light

Summary

• Tired light is the proposal that a photon loses energy on a long journey, so redshift measures distance travelled rather than recession.  Fritz Zwicky published it in October 1929, seven months after Hubble's distance-velocity paper.¹  It explains Hubble's law with no expanding space, no beginning, and no dark sector.

• It is refuted — not unfashionable, but measured and excluded, by three independent classes of observation, over more than fifty years.  This article is a worked example of what a real falsification looks like, and the Berean Archive holds every other cosmology, including creationist ones, to the same standard.

• The first objection needs no telescope.  Any interaction that takes energy from a photon also changes its direction, and a billion small energy losses means a billion small deflections, which smears a point source into a blur.  Distant galaxies are not blurred.  Zwicky made this argument himself, in the same paper, and used it to kill the Compton-scattering version of his own idea:  a scattering explanation "will be in a hopeless position regarding the good definition of the images."¹  His own alternative — a "gravitational analogue of the Compton effect" — he could not derive properly, and he asked only that "a further elaboration of the theory seems to be worthwhile."¹¹  Later advocates were far more confident than the man who invented it.

• Decisive measurement one:  distant events run slow.  Energy loss drains each photon separately;  it cannot change how often photons arrive, so tired light predicts no stretching of event durations at all.  Supernova light curves show the (1 + z) stretching;² supernova spectra age at 1/(1 + z)^b with b = 0.97 ± 0.10, and the no-dilation model gives χ² = 150.3 for 35 degrees of freedom, a goodness of fit of 0.0 percent;³ and since 2023 the same stretching appears in the variability of 190 quasars monitored for over two decades.⁴

• Decisive measurement two:  surface brightness.  Tired light predicts dimming as (1 + z)¹.  Hubble Space Telescope photometry of three galaxy clusters near z = 0.85 gives an exponent near 3, and Lubin and Sandage report that the data "rule out the 'tired light' model at a significance level of better than 10 σ."⁵

• Why that test kills tired light but not a static Doppler model.  Etherington proved in 1933 that luminosity distance equals (1 + z)² times angular-diameter distance in any spacetime with Riemannian geometry, light on null geodesics, and conserved photon number.⁶  Holanda, Lima and Ribeiro state the scope exactly:  the relation is "valid for all cosmological models based on Riemannian geometry, being dependent neither on Einstein field equations nor the nature of matter-energy content."⁷  The (1 + z)⁻⁴ dimming is a theorem, and flat static Minkowski space obeys it as surely as an expanding one.  Tired light escapes only by destroying photon energy in transit — the one assumption the theorem forbids, and the reason it dies while a static Doppler model survives.

• Decisive measurement three:  the microwave background.  COBE/FIRAS measured it as a blackbody with RMS deviations "less than 50 parts per million of the peak."⁸  That needs a single thermalising surface at a single redshift, and a static, eternal universe has none, so its background would be a superposition of sources at every distance, each reddened differently — not a Planck spectrum.  Even Martín López-Corredoira, an astronomer sympathetic to static models, concludes that "it seems that this solution does not work."⁹  Carbon-monoxide thermometry adds a fourth blow:  T(z) = (2.725 ± 0.002) × (1 + z)^(1−β) with β = −0.007 ± 0.027, where a static radiation bath requires the exponent to be zero.¹⁰

• Modern variants deserve a fair hearing.  Nikita Lovyagin, Alexander Raikov, Vladimir Yershov and Yuri Lovyagin published a static-universe analysis of early JWST angular sizes in Galaxies in 2022, explicitly challenging the assumption that tired light is ruled out,¹¹ and their group's 2026 preprint reports a surface-brightness trend in 6,860 JWST galaxies departing from (1 + z)⁻⁴.¹²  These are real analyses and the anomaly may be real.  But the same authors write that tired light predicts no time dilation, that time dilation is observed, and that any static model still owes an account of the microwave background, the light elements and structure formation.¹¹¹¹

• Creationists should not reach for tired light, and the reason is not tactical.  It is refuted, and it does not deliver what a young-earth reader wants:  a static universe with photons decaying in flight is eternal or indefinitely old, and it does nothing for the distant-starlight problem.  Danny Faulkner — a young-earth creationist with a PhD in astronomy — warns that rejecting expansion "amounts to a rejection of general relativity" and that "if the universe is expanding, but we reject this, it would be impossible to construct a correct biblical cosmology."¹³¹³  Even Jason Lisle, who argues for a static universe, writes that the tired light model fails the Tolman test.¹⁴

What Zwicky actually proposed

Edwin Hubble's paper announcing that a galaxy's redshift rises with its distance appeared in the Proceedings of the National Academy of Sciences in March 1929.  Fritz Zwicky's reply appeared in the same journal seven months later, communicated on 26 August 1929.  It is seven pages long, and almost nobody who cites it has read it.

Zwicky opened not with his own idea but with a list of "some of the principal facts which any cosmological theory will have to account for," promising only at the end to suggest "a new effect of masses upon light… a sort of gravitational analogue of the Compton effect."¹  His list is still the right list, which is part of why the paper is worth reading a century later.  Four of his six observational facts matter here.

1. Hubble's correlation is "roughly linear, corresponding to 500 km./sec. per 10⁶ parsecs."¹ That is about seven times the modern Hubble constant — an error in the distance scale that took decades to fix and that has no bearing on the argument.
2. "The relative shift of frequency… is apparently independent of the frequency."¹ Every wavelength stretches by the same factor.
3. "No appreciable absorption or scattering of light can be related to the above shift of spectral lines."¹
4. "The optical image of an extragalactic nebula seems to be as well defined as can be expected from the resolving power of the telescopes. The distance apparently is only geometrically involved and no additional blurring of the images occurs."¹

Points 2 through 4 are the constraints every tired-light model since has had to fight, and Zwicky wrote them down before proposing anything.

He then worked through the candidates.  De Sitter's universe he set aside because it reproduces Hubble's linear law only "by making some additional assumptions about the distribution of the nebulae."¹  Ordinary gravitational redshift he dismissed in a sentence:  the effect "has no relation to the distance of the observed galaxy from our own system and, therefore, cannot provide any explanation of the phenomenon discussed in this paper."¹

Compton scattering off free electrons is the version most people mean by "tired light," and Zwicky killed it himself, in this paper.  The arithmetic is simple.  Confine the deflection to less than the angular size of a nebula — under one degree — and a single Compton scattering changes the wavelength by at most 3 × 10⁻¹³ cm.  To produce the observed shift of 1 to 100 Å you therefore need between about 30,000 and 3 million scatterings, and every one of them sends light off in a new direction:

But then the light scattered in all directions would make the interstellar space intolerably opaque which disposes of the above explanation.¹

And, generalising:

Also, it is evident that any explanation based on a scattering process like the Compton effect or the Raman effect, etc., will be in a hopeless position regarding the good definition of the images as mentioned under B₄.¹

His own proposal was the gravitational drag of light.  Having ruled out scattering because it deflects, Zwicky needed a mechanism that drains energy without deflecting, and picked gravity:

According to the relativity theory, a light quantum hν has an inertial and a gravitational mass hν/c².  It should be expected, therefore, that a quantum hν passing a mass M will not only be deflected but it will also transfer momentum and energy to the mass M and make it recoil.  During this process, the light quantum will change its energy and, therefore, its frequency.¹

He is explicit that he cannot do the calculation properly:  "It is hardly possible to give a completely satisfactory theory of this gravitational analogue of the Compton effect, without making use of the general theory of relativity."¹  What follows is a Newtonian estimate, integrated over a uniform matter density ρ out to a cut-off distance D, giving Δν/ν ≈ 1.4π f ρ D L / c² for a path length L.

Then he compared it with the data, and that is the most honest paragraph in the paper.  Hubble's density estimate for intergalactic space ran from 10⁻²⁶ down to 10⁻³¹ grams per cubic centimetre — five orders of magnitude — and D was a guess, so the predicted redshift for L = 10⁶ parsecs came out anywhere between 3 × 10⁻² and 3 × 10⁻⁷.  The observed value is 1/600, or 1.7 × 10⁻³ — inside that range, but so is almost everything.  His verdict:  "In view of this agreement in order of magnitude, a further elaboration of the theory seems to be worthwhile."¹  His closing paragraph then asks for a general-relativistic derivation, a full treatment of the mutual gravitational interactions, and an account of the masses' proper motions — none of which the theory had.¹

That is a proposal offered tentatively, with the mechanism unfinished and the agreement admitted to be only an order of magnitude.  Almost every later use of "tired light" is more confident than Zwicky was.  The general-relativistic derivation he asked for was never supplied, and the reason is now understood:  a photon falling into a static gravitational well is blueshifted going in and redshifted climbing out, and the two shifts cancel exactly.  A static distribution of masses produces no net energy loss.  Getting a residue requires the potential to change while the photon crosses it — a real effect, the integrated Sachs-Wolfe effect, but second order, sign-indefinite, and not cumulative with distance.  Zwicky's own mechanism does not work in the theory he asked to have it rewritten in.

Danny Faulkner's summary, from inside the creationist community, is fair:  the hypothesis "proposed that light naturally loses energy as it propagated over large distance, with the amount of energy lost proportional to distance.  The mechanism whereby this happened was never identified."¹³

Why the idea is attractive

A model does not attract serious people for a century by being stupid.  Tired light has four real virtues, and a fair account states them before demolishing it.

It gets Hubble's law for free, and gets it exactly linear.  If a photon loses a fixed fraction of its energy per unit path length, redshift is exactly proportional to distance travelled, with no free function and no deceleration parameter.  In the standard model, linearity at low redshift has to be derived, and the curvature at high redshift is what forces dark energy into the fit.  In tired light, linearity is the mechanism.

It removes the beginning.  A static universe of indefinite age has no creation event and no cosmic clock running down.  Zwicky wrote before Lemaître's primeval atom was widely known, when a beginning in time struck many physicists as a theological intrusion.  That motive is worth naming honestly, because it is the mirror image of the motive a creationist brings.  Neither is evidence.

It removes the dark sector.  No expansion means no cosmological constant needed to explain why distant supernovae are dimmer than a coasting universe predicts.

It answers Olbers' paradox.  In an infinite, eternal, static universe every line of sight ends on a star and the night sky should blaze.  If light reddens with distance, remote starlight shifts out of the visible band and the sky is dark.

Set against that, the standard model needs an expansion, a beginning, inflation, dark matter and dark energy, and its two best measurements of the expansion rate currently disagree at around the 5σ level, a genuine unsolved problem treated in its own article on the Hubble tension.  A reader who wonders whether the establishment is defending its turf is asking a reasonable question.  This article answers it by showing the measurements, with their numbers and their weaknesses.

The objection that needs no telescope

Before any measurement, tired light has a mechanism problem, and it is the one Zwicky identified on page 775.

Energy and momentum come in the same package.  A photon carries momentum p = E/c.  Take energy out of it and you have taken momentum out of it.  Momentum is a vector, so whatever absorbed it recoiled in some direction, and the photon's direction changed to match.  There is no known interaction that removes a photon's energy while leaving its direction untouched, and that is not a gap in our knowledge:  "energy loss with no direction change" would mean the recoiling particle absorbed energy without absorbing momentum, which no dynamics allows.

Now count.  The redshift accumulates over billions of light years, so if each interaction takes a tiny bite there must be an enormous number of them, and deflections add up like a random walk.  A point source becomes a fuzzy disc and a galaxy's image becomes a smear.

Distant galaxies are not smeared.  Their images are as sharp as the telescope allows — what Zwicky reported in 1929 and what a century of better instruments has confirmed.  High-redshift quasars are point sources.  The angular-size measurements that static-model advocates themselves use, with medians near 0.2 arcseconds for JWST galaxies at z > 5, presuppose sharp images, because a blurred image has no measurable size.

López-Corredoira — who is not a defender of the standard model, and who publishes static-universe analyses himself — states the two standing problems in a 2017 review:

But there are two problems:  1) the φ-bath smears out the coherence of the radiation from the source, and so all images of distant objects would look blurred if intergalactic space produced scattering, something that is incompatible with present-day observations;  2) the scattering effect and the consequent loss of energy would be frequency dependent, which is again incompatible with what we observe in galaxies.⁹

The second problem is as bad as the first.  Every real photon-matter interaction has a wavelength-dependent cross-section.  Rayleigh scattering goes as λ⁻⁴;  Compton scattering shifts wavelength by a fixed amount, not a fixed fraction.  The observed redshift is a fixed fraction:  a one-nanometre wave becomes a two-nanometre wave and a one-metre radio wave becomes a two-metre radio wave, so a galaxy's redshift from its 21 cm hydrogen line agrees with its redshift from optical lines — wavelengths differing by a factor of about 400,000.¹⁵  Luke Barnes and Geraint Lewis put the whole difficulty in one sentence:  "There just isn't a way to make a uniform light stretcher out of matter."¹⁵

The escape hatches, and what they cost

Advocates know all this, and the modern literature is mostly attempts to build an interaction that drains energy without deflecting.  Jean-Pierre Vigier proposed a stochastic covariant superfluid aether whose excitations dissipate photon energy;  Paul Marmet and Grote Reber proposed a "plasma redshift" in which a photon gives up energy to an electron without changing direction;  others have proposed photon-Raman interactions with atomic hydrogen, non-linear optics with an anharmonic correction, or multiple scattering in a turbulent medium.⁹

Two things should be said, and they pull in opposite directions.

In fairness:  they aim at exactly the right target.  Each is a serious attempt to satisfy the no-blur and no-colour-dependence constraints, and their authors state those constraints as clearly as any critic would.

Against them:  every one is built backwards.  The desired output is specified first and a mechanism is reverse-engineered to produce it.  None has been seen in a laboratory;  each requires a new field, a new aether, or a departure from quantum electrodynamics.  And there is a cost usually left unsaid:  the plasma-redshift route needs an intergalactic density of about 10⁴ atoms per cubic metre, roughly 10⁵ times the accepted value of order 10⁻¹ atoms per cubic metre.⁹  That is not an adjustment to a poorly known number.  It is a different universe.

None of this is decisive on its own.  A mechanism nobody has seen might still exist.  But it means tired light enters the observational tests already owing an explanation, and the tests are where it dies.

Decisive measurement one: distant events run slow

This is the cleanest of the three.

The logic.  A moving source, or a stretching geometry, does not only lengthen the waves.  It lengthens the interval between anything the source does — wave crests, flashes, explosions, flickers, all spread out by the same factor.  An energy-loss mechanism cannot do that.  Tired light drains each photon separately as it flies;  it has no way to change how often photons leave the source or arrive here.  Barnes and Lewis state the prediction in one sentence:

In a tired light universe, if a supernova intrinsically brightens and darkens over a period of 30 days, it will be observed to take 30 days to brighten and darken through Earthly telescopes, albeit with the light redshifted to longer wavelengths.¹⁵

So there is a hard split, with no free parameter anywhere in it:

The universe supplies the clocks.  Type Ia supernovae rise and fade on a schedule set by the radioactive decay of nickel-56 and cobalt-56.  Quasars flicker.  All that is needed is to watch a distant one and time it.

Light curves: Goldhaber 2001

Type Ia supernovae are not identical — a brighter one fades more slowly — so the test has to be statistical.  Gerson Goldhaber and the Supernova Cosmology Project fitted a width to nearly 1,300 normalised intensity measurements from high-redshift supernovae and nearly 300 B-band points from 18 nearby Calán/Tololo supernovae at z < 0.11.  With a time-axis width factor w = s(1 + z), the intrinsic stretch s came out independent of redshift while the (1 + z) factor was required:  "This argues strongly against alternative explanations, such as tired light, for the redshift of distant objects."²  Their §4 names the models it kills, Zwicky's included:  the tired light theories "would not yield this slowing of the light curves, and thus do not fit this dataset."²

The honest weakness of this version:  it assumes the intrinsic width distribution does not itself change with redshift.  Goldhaber's own paper flags it — "this analysis does not account for sample selection effects — such as possible preferred discovery of high stretch SNe near the flux limit of each SNe search."²  That is exactly what the next paper set out to remove.

Spectra: Blondin 2008, and why it is the stronger test

A supernova's spectrum changes as it ages.  Different elements dominate on different days;  the absorption features shift as the ejecta thin out.  That progression is a clock that does not depend on how bright the supernova is, and it cannot be counterfeited by an intrinsically wide light curve.

Stéphane Blondin and 31 co-authors took 35 spectra of 13 high-redshift Type Ia supernovae between z = 0.28 and z = 0.62, matched each against a library of nearby supernova spectra to read off its age in its own rest frame, and compared that with the elapsed time an Earth clock recorded between observations.  Fitting an ageing rate of 1/(1 + z)^b gives b = 0.97 ± 0.10 for the whole sample and 0.95 ± 0.10 for the high-redshift objects alone.  Relativity predicts exactly 1.  Tired light predicts 0.

Those are Blondin's Table 4 numbers for the full sample.³  A goodness of fit of 0.0 percent is not "the data mildly prefer the other model."  It means that if tired light were true, a data set this bad would essentially never occur.  Blondin's own verdict:  "This hypothesis is favored beyond doubt over models that predict no time dilation."³

Quasars: the loophole that stayed open until 2023

A critic had a real reply available until recently.  Supernovae are one class of object, and quasar variability had failed to show the expected stretching in several published analyses.  López-Corredoira's 2017 review listed that failure among the standard model's problems, and he was right to.

It closed in 2023.  Geraint Lewis and Brendon Brewer analysed 190 quasars monitored for over two decades in multiple wavebands, using Bayesian model comparison rather than a structure-function fit, and detected the (1 + z) stretching:  "This detection counters previous claims that observed quasar variability lacked the expected redshift-dependent time dilation."⁴

A 2025 follow-up by the same authors with Yuan Li did the sceptic's work on their own result.  Quasar timescales depend on luminosity and on rest-frame wavelength, and the population evolves;  all three could counterfeit a redshift dependence.  Allowing for an intrinsic scatter of 0.288 ± 0.021 dex that the earlier work had not modelled, they refitted (1 + z)ⁿ and obtained n = 1.14 ± 0.34 — consistent with the predicted 1, and 3.4σ from the tired-light prediction of 0.  The telling detail is what happened in the model comparison:  "The only exception was a model that assumes n = 1, which outperformed the main model."¹⁶  The free parameter did not earn its keep.  The prediction fit better than the fit.

The one place this test is still weak: gamma-ray bursts

Honesty requires flagging the item on López-Corredoira's list that has not closed.  Amitesh Singh and Shantanu Desai fitted duration ∝ (1 + z)^B to 247 Swift bursts, where B should be 1.  Their opening sentence is the fair summary of the field:  "the cosmological time dilation signature for GRBs has not been unequivocally demonstrated."  Their own result is mixed — binned analyses consistent with B = 1 within 1σ, unbinned analyses showing intrinsic scatter above 100 percent and "no definitive conclusion… can be inferred."¹⁷  That is a genuine loose end.  It is also weaker than it looks, because burst durations vary intrinsically by three orders of magnitude while supernova light curves vary by tens of percent.  A test whose scatter swamps the signal is uninformative, not contrary.

Can tired light be saved here?

Only by making distant supernovae intrinsically different from nearby ones, in exactly the way (1 + z) time dilation would look.  Barnes and Lewis put the objection precisely:

To account for the observed data on the duration of supernova fading, the actual evolution must be very specific;  it must be the form expected from the big bang theory.  To make tired light and supernova evolution work, we need to fine-tune the evolution, dialling in a specific but unmotivated relationship.¹⁵

The comparison with the standard model's own fudges is the point.  Stellar ageing and K-corrections are unavoidable in any cosmology and independently testable against nearby stars.  A conspiracy that slows ancient supernovae by precisely (1 + z), and their spectra by precisely (1 + z), and quasar variability by precisely (1 + z), is testable nowhere else.  That is what "ad hoc" means.

Decisive measurement two: the Tolman surface-brightness test

This test is misused by nearly everyone who cites it, in both directions.  Getting it right takes one theorem, and it changes what the test is worth — including what it is worth against tired light.

What Tolman proposed, and what he was aiming at

Richard Tolman noticed in 1930 that surface brightness — light received per unit area of a galaxy's image on the sky — behaves very differently in an expanding universe than in a static one.  Allan Sandage, who spent thirty-five years on the measurement, states Tolman's result exactly:

Tolman (1930, 1934) discovered the effect that the surface brightness of a "standard" radiating object that is receding with redshift z will be fainter than a similar stationary "standard" object at rest by (1 + z)⁴.  However, if the manifold is stationary but nevertheless has a redshift due to an "unknown law of nature", the factor is only (1 + z).¹⁸

Read that carefully.  Sandage's contrast is between an object receding and an object at rest whose redshift comes from an unknown law of nature.  It is not a contrast between expanding space and motion through static space.  Tolman's test was built to catch a non-kinematic redshift.  That is what it catches, and tired light is the thing it was built to catch.

Surface brightness is a good discriminator because distance cancels out of it.  A galaxy twice as far away delivers a quarter of the light, but covers a quarter of the sky.  In a static, Euclidean universe with no redshift at all, surface brightness does not depend on distance.  Everything that survives that cancellation is telling you about geometry and about what happens to photons in flight.

Why (1 + z)⁻⁴ is a theorem, not a prediction

Four factors of (1 + z) appear.  Each photon arrives with its energy reduced by (1 + z).  Photons arrive less often, by another (1 + z) — the time dilation of the previous section.  The remaining two come from the relation between the two ways of measuring distance:  the luminosity distance D_L, from how faint a known brightness looks, and the angular-diameter distance D_A, from how small a known size looks.  Surface brightness is flux divided by solid angle, which is (D_A/D_L)² times the intrinsic value.

The formal statement is Etherington's reciprocity theorem, proved in 1933:

D_L = (1 + z)² D_A

Holanda, Lima and Ribeiro state its scope as plainly as anyone has:

This equation is completely general, valid for all cosmological models based on Riemannian geometry, being dependent neither on Einstein field equations nor the nature of matter-energy content.  It only requires that source and observer are connected by null geodesics in a Riemannian spacetime and that the number of photons are conserved.⁷

Renzi, Hogg and Giarè, who tested the relation against data, give the same conditions in the negative.  Deviations "would appear if photons did not propagate on null geodesics (i.e. were not massless), if photon number was not conserved… or if the spacetime was not described by a pseudo-Riemannian manifold."¹⁹  Their joint fit to late-time data gives a deviation parameter ε = 0.001 ± 0.025 — consistent with zero at the couple-of-percent level.¹⁹

Why tired light fails a test that a static universe passes

Here is the sentence this article is built around.

Flat, static, non-expanding Minkowski space satisfies all three conditions of the theorem.  Light travels on null geodesics, photons are conserved, and the geometry is Riemannian.  So a universe in which galaxies simply fly apart through fixed space predicts (1 + z)⁻⁴ dimming, exactly like an expanding one.  So does every other metric theory anybody has written down.  The (1 + z)⁻⁴ law is a theorem, not a prediction of expansion, and the Tolman test cannot tell expansion from motion through static space.  It never could.

Tired light is the one candidate that escapes the theorem, and it escapes by violating a premise.  It removes energy from photons in transit, so the redshift is not a property of the geometry, photon energy is not conserved along the ray, and Etherington does not apply.  With a Euclidean static geometry and no time dilation, only one factor of (1 + z) survives:  the predicted exponent is 1 instead of 4.

That is why tired light dies here and a static Doppler model does not.  Escaping the theorem is the only way to make a different prediction, and the escape route is exactly the assumption the data then rule out.

Lubin and Sandage found the price of that escape when they tried to compute a tired-light prediction at all:

In contrast with the standard model with the Mattig equations, there is no metric theory of how distances and magnitudes are measured in a "tired light" model.  Therefore, we must guess at a reasonable equation for distance.⁵

That sentence is worth more to a sceptical reader than any of the numbers.  Tired light is not a rival cosmology.  It is a proposal about photons with no geometry attached, which is why every quantity has to be hand-fitted, and why it fails so badly when it is.

The measurement, and the honest problem with it

Lubin and Sandage measured surface brightness for early-type galaxies in three clusters near z = 0.85, imaged with the Hubble Space Telescope, against a local comparison sample:  "The HST data rule out the 'tired light' model at a significance level of better than 10 σ."⁵

A reader who hears only "10 σ" is being sold something.  The exponent has to be extracted from real galaxies whose intrinsic surface brightness is unknown and changes as their stars age, and the raw result was n = 2.59 ± 0.17 in the R band and 3.37 ± 0.13 in the I band.⁵  Neither is 4.  The gap is attributed to luminosity evolution, in the amount independent stellar-population models predict.⁵¹⁸  The band dependence is the tell:  a geometric exponent cannot depend on which colour filter you use, so 2.59 in one band and 3.37 in another announces at high significance that something band-dependent is present.  That something is stellar evolution.  The full argument is in the article on the expanding universe.  So the standard model does not sail through this test either;  Barnes and Lewis, who accept it, say the big bang "survives the surface brightness test.  It doesn't fail… But it doesn't pass with flying colours either."¹⁵

Does that rescue tired light?  No, and the reason is a direction, not a magnitude.  Stellar evolution makes distant galaxies brighter than their descendants, which means less observed dimming.  It moves every prediction the same way:  toward smaller measured exponents.  The standard model needs that help to get from 4 down to the observed 3.  Tired light predicts 1 and needs the correction to push the other way.  Barnes and Lewis make exactly this point:  the corrections "move the prediction upwards, to give us less dimming than we might expect.  This would also be the case in a tired light universe, meaning that its prediction is even worse than is shown in Figure 4.3."¹⁵

An observed exponent near 3 sits one unit below the standard model's 4 and two units above tired light's 1, and the only known astrophysical correction moves both predictions away from tired light.  Sandage's closing verdict — "The Tolman prediction is verified.  The expansion would seem to be real"¹⁸ — is a correct refutation of tired light and an overstatement about expansion, because a static Doppler model predicts the same (1 + z)⁻⁴ and would pass the same test.

Decisive measurement three: the microwave background

The cosmic microwave background is the most perfect blackbody ever measured, anywhere, by anyone.  COBE's Far InfraRed Absolute Spectrophotometer compared it against a precision blackbody carried on board:

The RMS deviations are less than 50 parts per million of the peak of the CMBR.  For the Comptonization and chemical potential we find |y| < 15 × 10⁻⁶ and |µ| < 9 × 10⁻⁵ (95% CL).⁸

Fifty parts per million, at a temperature of 2.728 ± 0.004 K.  That is where tired light runs into a wall — but the argument is subtler than the popular version, and the popular version is wrong.

The version that does not work

It is often said that tired light "distorts a blackbody."  Taken literally that is false, and a critic who says it can be embarrassed.  Take a Planck spectrum at temperature T and apply a strictly proportional energy loss:  every photon's frequency divided by (1 + z), no photons created or destroyed.  Work through the bookkeeping and the result is a Planck shape at temperature T/(1 + z).  A uniform redshift does not by itself deform a blackbody — which is exactly why an expanding universe keeps the background thermal as it cools.

So the naive objection fails.  Three real ones remain.

Problem one: the amplitude comes out too high

In an expanding universe the four powers of (1 + z) dim a blackbody source exactly as needed to keep it on the Planck curve.  In a static Euclidean universe with tired light, only one power survives, so the spectrum has the Planck shape at T/(1 + z) but an intensity (1 + z)³ too high for that temperature.  Convert that to a brightness temperature — in the long-wavelength part of the spectrum, intensity is proportional to temperature — and it comes out at (1 + z)² × T, hotter than the source that emitted it.  Focus that light and you could heat an object above the temperature of the furnace it came from.  That is a second-law violation, and it is the deep reason surface brightness must dim as (1 + z)⁻⁴ in any theory where photons are just travelling.

Problem two: a static universe has no last-scattering surface

The standard model produces a blackbody because it has one:  a single moment when the universe became transparent everywhere at once.  Every microwave photon we see left that surface at essentially the same redshift.  One temperature in, one temperature out.

A static, indefinitely old universe has no such moment.  Whatever produces the background is spread through space, so the photons arriving now left their sources at every distance and have been reddened by every amount.  The observed spectrum is a superposition of Planck functions at a continuum of temperatures.  A sum of blackbodies at different temperatures is not a blackbody.  The broadening is not a 50-parts-per-million effect;  it is order unity.

Problem three: thermalising it needs an opacity the sky does not have

The obvious repair is a medium that scattered the background into equilibrium.  But thermalisation needs a large optical depth, and a large optical depth means we could not see through it.  López-Corredoira states the problem and concedes it:

But in a tired light model in a static universe the photons suffer a redshift that is proportional to the distance travelled, and in the absence of absorption or emission the photon number density remains constant, we would not see a blackbody background.  The universe cannot have an optical depth large enough to preserve a thermal background spectrum in a tired light model because we could not observe radio galaxies at z ∼ 3 with the necessary optical depth.  Therefore, it seems that this solution does not work.⁹

That is a professional astronomer who publishes static-universe analyses, writing in a review devoted to the standard model's failures, concluding that the tired-light account of the microwave background does not work.

The fourth blow: the background is hotter in the past

In an expanding universe the background cools as the universe grows, so its temperature at redshift z must be T₀(1 + z).  In a static universe filled with a uniform radiation bath the temperature is the same everywhere — the exponent is zero.

Carbon-monoxide absorption in quasar spectra gives T_CMB(z) = (2.725 ± 0.002) × (1 + z)^(1−β) with β = −0.007 ± 0.027.¹⁰  The exponent is 1.007 ± 0.027;  zero is about 37 standard deviations away.  Water absorption in a starburst galaxy independently puts the background at 16.4 to 30.2 K at z = 6.34 against 20.0 K predicted,²⁰ and Sunyaev-Zel'dovich scattering in hundreds of galaxy clusters agrees.

A tired-light advocate has one move, and it is the one López-Corredoira's own summary table records:  put the source of the background at a definite great distance, so a cloud at redshift z sees radiation tired-lighted less than the radiation reaching Earth.⁹  That can be made to work numerically.  But look at the cost.  It reintroduces a special emitting shell at a definite distance — a preferred place, and in an eternal universe effectively a preferred time.  The whole appeal of the static model was that it had neither.  And a shell at one distance is precisely what problem two requires, so the two repairs are the same repair:  reinventing the surface of last scattering with none of the physics that produced it.

The scorecard

Two rows are not losses, and this article prints them because leaving them out would be dishonest.  The gamma-ray burst test is uninformative, not favourable.  The angular-size test is the one place where a static model genuinely fits better than the standard model does without help, and it is why the modern static-universe literature exists.

Modern variants, and who still argues for them

Tired light did not die quietly in 2001.  A small, serious literature continues, and it deserves to be described accurately rather than waved at.

David Crawford has argued for a static universe with photon energy loss over three decades;  his most sustained statement is a 2011 paper whose title is its thesis, "Observational evidence favors a static universe."²¹  Paul LaViolette published a 2021 paper arguing the same.²²  Paul Marmet and Grote Reber — the latter described by López-Corredoira as "a co-initiator of radio astronomy" — proposed the plasma-redshift mechanism above.⁹  Naming them is better than saying "some researchers."

The most active group is at Saint Petersburg State University:  Nikita Lovyagin, Alexander Raikov, Vladimir Yershov and Yuri Lovyagin, who published "Cosmological Model Tests with JWST" in Galaxies in December 2022.  Their thesis is stated without hedging:

We compare this result with predictions of the standard ΛCDM cosmological model and some static cosmological models, including Zwicky's "tired-light" model.  The latter is currently assumed to be ruled out by observations.  We challenge this assumption and show that a static model can provide a natural and straightforward way of solving the puzzle of the well-evolved galaxies.¹¹

Their argument is the angular-size relation.  In an expanding universe angular diameter has a minimum and then rises again at high redshift, because very distant objects were physically close when they emitted.  In a static tired-light universe it falls monotonically toward a floor.  Early JWST data showed high-redshift galaxies smaller and more evolved-looking than ΛCDM comfortably predicts, and the static curve fitted better.  In April 2026 the same group posted a preprint extending the work to surface brightness, analysing 6,860 galaxies with spectroscopic redshifts from the ASTRODEEP-JWST catalogue and reporting "a significant departure from the prediction of the standard cosmological model."¹²

These are real analyses on real data, and the anomaly may well be real.  Galaxy sizes and brightnesses at high redshift are a live problem in mainstream cosmology, discussed at length in the companion articles on the expanding universe and the Doppler model.  Nothing here requires the JWST size measurements to be wrong.

But the same authors concede the case this article is making, in the same paper:

The static cosmological model with photon-energy dissipation predicts no time dilation…  The time-dilation effect is actually observed in the light-curves of the type-Ia supernovae, which strongly supports the expanding-Universe models or the static-Universe models of the Einstein-de-Sitter type.¹¹

And their summary of the remaining bill:

Although our analysis of JWST observations favours a static (TL) model of the Universe rather than the standard (expanding-Universe) cosmology, the latter is strongly supported by other observational evidence.  The main challenges which any static cosmological model must face are the cosmic microwave background (CMB), the abundance of chemical elements in the Universe and the formation of cosmic structure.¹¹

They even note that their own Figure 6 shows the universe evolving relative to the tired-light model, so that "the real picture is more complicated than a pure static cosmological solution."¹¹

That is the honest state of play.  The best modern advocates agree that tired light predicts no time dilation, agree that time dilation is observed, agree that it owes an account of the microwave background and the light elements, and use tired light as an example of a static model rather than the model they are defending.  What survives their work is the possibility of some static or slowly-expanding cosmology, not Zwicky's mechanism.

What a static model would still have to explain.  A mechanism that drains photon energy without deflecting and without wavelength dependence, and that has been seen somewhere;  the (1 + z) ageing of supernova light curves, supernova spectra and quasar variability;  a 50-parts-per-million blackbody with no last-scattering surface to make it;  a background whose temperature scales as (1 + z)^1.007 ± 0.027^ in a universe with no cooling;  helium at 24 percent by mass and the observed deuterium and lithium, without a hot early phase;  and the growth of structure.  The standard model answers all of these with a single parameter set.  A rival that answers one of them better is interesting.  It is not yet a cosmology.

Why creationists should not reach for tired light

Tired light appears regularly in popular creationist writing as a way of dismissing the big bang without argument.  It should be dropped, for two reasons, and the second matters more.

First, it is refuted.  Everything above.  Citing a model that lost at 10σ, at 0.0 percent goodness of fit, and at 37σ on three separate measurements does not undermine the big bang.  It damages the credibility of whoever cites it, and it should, because the pattern on display is a willingness to use any stick.

Second, it does not deliver a young universe.  This is the part that seems to go unnoticed.  Tired light describes a universe that is static — no expansion, no beginning, no cosmic clock.  Photons take billions of years to cross it, and its stars have to be old enough to have made the heavy elements we see.  It is an eternal or indefinitely old universe, a worse fit to Genesis than the big bang, not a better one.  It does nothing for the distant-starlight problem, and Jason Lisle — the creationist most associated with static-universe cosmology — says as much about his own static model:  it "is not needed to explain distant starlight in a young universe."²³  Lisle also states flatly that the tired light model fails the Tolman test, in the same sentence in which he claims his own model passes it.¹⁴

Danny Faulkner has been making this argument inside the creationist community for years.  He describes tired light accurately, including its fatal gap — "the mechanism whereby this happened was never identified"¹³ — and notes that rejecting expansion "was tantamount to rejection of general relativity, upon which the interpretation of the Hubble relation as expansion is based."¹³  His warning to his own community is blunt:

Rejection of universal expansion seems to be motivated by opposition to the big bang model.  If the universe is not expanding, then the big bang model is not viable.  However, the big bang is just one possible expanding universe cosmology.  This short-cut way of undermining the big bang model could be a huge mistake, for if the universe is expanding, but we reject this, it would be impossible to construct a correct biblical cosmology.¹³

That is the right position and this article endorses it.  The expansion is a measurement.  Whether it started 13.8 billion years ago is a separate question with separate evidence, treated in the big bang article.  Attacking the measurement to avoid the conclusion is bad method, and it is bad method whoever does it.

What would revive tired light

A model is only worth taking seriously if you can say what would change your mind.  The list is short.

• A laboratory demonstration of an interaction that removes energy from photons in proportion to path length, with no deflection and no wavelength dependence. Nothing else on this list matters without it.
• A metric formulation. Lubin and Sandage had to "guess at a reasonable equation for distance" because tired light supplies none. A version with an actual spacetime attached could be tested properly instead of hand-fitted.
• A demonstration that supernova spectral ageing is an artefact — not the light curves, the spectra, which is the harder result.
• An account of the 50-parts-per-million blackbody that does not require a scattering surface at a single redshift.

If the first item ever arrives, the rest becomes worth arguing about.  Until then, tired light is a historical episode:  a good scientist's speculative idea, tested honestly for seventy years, and found false.  That is not a failure of the process.  It is the process working, and it is the standard this site applies to every other model, including the ones it would prefer to be true.

Objections and responses

"Tired light was rejected because it threatened the big bang, not because of the data."

Check the dates and the authors.  The first person to publish a fatal objection to a tired-light mechanism was Zwicky, in the paper that proposed one, seven months after Hubble and years before the big bang was a serious contender.  The supernova time-dilation result came from the Supernova Cosmology Project, whose famous result — dark energy — was itself an unwelcome surprise to theorists.  Lubin and Sandage reported honestly that their raw exponent was 2.59 in one band and 3.37 in another rather than the 4 their own model wanted.  And the sharpest concessions quoted in this article come from López-Corredoira and from Lovyagin and colleagues, both of whom publish static-universe analyses.  That is not a consensus closing ranks.

"Nobody has ever seen space expand.  At least photons and matter are things we know exist."

Fair instinct.  But nobody has seen a photon lose energy in flight either, and the mechanism has been sought for ninety-seven years without success.  More to the point, the argument here does not depend on expansion being real.  Motion through static space produces (1 + z) time dilation and (1 + z)⁻⁴ dimming just as expansion does.  The measurements rule out energy loss;  they do not rule in expanding space.  That question is argued in the Doppler-model article.

"The Tolman measurement gives an exponent near 3.  The standard model predicts 4.  So both models are wrong."

The first two sentences are correct and this article says so.  But look at the distances.  The standard model is one unit off and has an independently calibrated correction — stellar ageing — of the right size and sign to cover it.  Tired light is two units off in the other direction, and the same correction makes its fit worse.  A model that needs the known systematic to run backwards is not in the same position as one that needs it to run forwards.

"Maybe the right mechanism just hasn't been discovered yet."

That may be true, and it would answer the blurring and wavelength objections.  It would not help with the decisive one.  Time dilation is not about the mechanism.  Any process that takes energy from photons individually, by whatever means, leaves the rate at which photons depart the source untouched, so it cannot stretch the duration of an event.  That is a statement about what "losing energy" means, not about which particle does the taking.  A perfect, undiscovered, non-deflecting, colour-blind energy sink would still predict that a distant supernova rises and fades on its own local schedule.  Distant supernovae do not.

"Eddington and Nernst predicted a few-kelvin background from static-universe reasoning before Gamow did."

They did, and Lovyagin and colleagues are right to point it out:  Eddington got about 3 K in 1926 from starlight, Nernst 2.8 K in 1937.¹¹  The coincidence is real and it is a point against triumphalist accounts of Gamow's prediction.  But guessing a temperature is a much smaller achievement than accounting for a spectrum that is Planckian to 50 parts per million, with an angular power spectrum whose acoustic peaks were predicted before they were measured.  The static-universe literature has no account of either.

"So the big bang is proved?"

No.  Killing tired light leaves two survivors:  an expanding geometry, and galaxies moving through static space.  Both predict everything in this article identically.  Whether the expansion began in a hot dense state 13.8 billion years ago is a third question with its own evidence.  This article establishes exactly one thing:  the redshift is not photons getting tired.

Sources

Every source worked up for this article is listed in literature-review.md beside it, marked used or not-used with a reason.

The primary document is Zwicky's own 1929 paper, read from the scanned PNAS original and quoted from 200 dpi page images rather than OCR, because the OCR of that scan corrupts exponents and superscripts.  The three decisive measurements are cited to the papers that made them:  Goldhaber et al. 2001, Blondin et al. 2008, Lewis and Brewer 2023 with Brewer, Lewis and Li 2025, Lubin and Sandage 2001 with Sandage 2010, and Fixsen et al. 1996.  The reciprocity theorem's scope is quoted from Holanda, Lima and Ribeiro 2010, not from Renzi et al. 2022, which does not contain that sentence and to which it has been misattributed elsewhere on this site.  For the case for tired light the article uses its advocates directly:  López-Corredoira 2017, and Lovyagin, Raikov, Yershov and Lovyagin 2022 with the 2026 Tolman-test preprint from the same group.