The Cosmic Microwave Background

Summary

• The microwave sky is the single best-measured thing in cosmology, and most of the arguments made about it — on both sides — are about the wrong parts of it.
• The temperature was not predicted. The published hot-Big-Bang numbers ran 5 K, 5 K, 28 K, 3 K, 50 K, 7 K, 6 K between 1948 and 1956. The answer is 2.72548 K. This article concedes that point in full and does not need it.
• The "40 K prediction" of 1965 was not a prediction. Dicke, Peebles, Roll and Wilkinson wrote that they could not predict the temperature and gave 40 K as a rough upper limit. The creationist list that cites it as a failed prediction has misread the sentence.
• What was predicted, and what is hard, is the shape. A blackbody with one temperature at every frequency, filling all of space, scaling as T(z) = T₀(1 + z), carrying a harmonic series of acoustic peaks, polarized a quarter-cycle out of phase with its own temperature pattern, and gravitationally lensed by foreground mass. Every one of those has been measured.
• COBE/FIRAS is the sharpest measurement in cosmology. Deviations from a perfect Planck curve are under 50 parts per million of the peak, with |y| < 1.5 × 10⁻⁵ and |μ| < 9 × 10⁻⁵.
• A pile of ordinary sources cannot fake that. Adding blackbodies at different temperatures does not give a blackbody; it gives a y-distortion. And building a Planck spectrum requires creating photons, which only works in a dense, hot early universe.
• The dipole is a velocity, and it has been checked twice by different physics. Its amplitude is 3362.08 ± 0.99 µK, giving 369.82 ± 0.11 km/s. Planck independently recovered 384 ± 78 ± 115 km/s in the same direction from the aberration and modulation of the small-scale pattern.
• The acoustic peaks were computed before they were measured. The harmonic series was described in 1970, the first peak was placed near multipole 200 in 1994, the alternating odd/even peak heights were published in 1995-97, and the E-mode-on-temperature-trough phase relation was stated in the introduction of the paper that first detected polarization. Planck now measures 35 extrema, with the first peak at ℓ = 220.6 ± 0.6.
• The first test of the peak heights failed. BOOMERanG's 2000 data gave a baryon density far above the value from deuterium. It was published as a problem, and closed two years later by better pointing calibration — not by adjusting the theory. A skeptic should know this story; it is the best evidence that the fit was not rigged.
• Polarization is the strongest under-used check. E-mode peaks land on temperature troughs, to under one percent at high multipole. Primordial B-modes have never been detected; BICEP2's 2014 claim was retracted after a joint analysis with Planck showed the signal was Galactic dust. That retraction is told here in full, because a movement that hides its own errors deserves neither trust nor imitation.
• T(z) = T₀(1 + z) is measured out to redshift 6.34 — by carbon-monoxide excitation in quasar absorbers, by the Sunyaev-Zel'dovich effect in clusters, and by water absorption against the background in a starburst galaxy. The deviation index is 0.007 ± 0.027. A static bath of radiation cannot do this.
• The large-angle anomalies are real and small. The low quadrupole, the quadrupole-octopole alignment, the hemispherical asymmetry, the parity asymmetry and the Cold Spot all sit between 2σ and 3.5σ, all were found after looking, and a 2026 reanalysis using 99% of the sky cut two of them. The Cold Spot's supervoid explanation is now excluded by lensing. Neither dismissing them nor inflating them is honest.
• No young-earth or alternative cosmology explains the CMB power spectrum. Andrew Repp says so in the Creation Research Society Quarterly. Danny Faulkner's water-boundary proposal comes with his own admission that he "arbitrarily picked values." Russell Humphreys' 2014 Unruh-radiation proposal solves for its acceleration from the measured temperature and never mentions the anisotropies; his 1994 proposal borrowed the Big Bang's own mechanism, and he has since moved away from the expansion it needs. Jason Lisle's claim that "any universe with an average equilibrium temperature of 2.7 degrees will have a CMB" is false for a specific, checkable reason: an average temperature gives you a superposition, and a superposition is not a blackbody.
• The honest bottom line. The microwave background is the strongest single body of evidence for a hot, dense early universe, and the young-earth side currently has no answer to its power spectrum. Saying otherwise is not apologetics; it is wishful thinking.

This is the case file for the microwave background.  For how it fits into the wider argument, see the standard model overview and young-universe cosmologies.

What the thing actually is

Point a sensitive microwave receiver anywhere in the sky, subtract the Milky Way, and a faint hiss remains.  It comes from every direction.  Its spectrum is the spectrum of a body at 2.72548 ± 0.00057 kelvin.¹

That number is not one instrument's reading.  It is a weighted mean of fifteen independent measurements — cyanogen molecules excited in interstellar gas clouds, balloon-borne radiometers, rocket flights, instruments at the South Pole, the ARCADE balloons, and three separate analyses of COBE's spectrometer.¹  A thermometer made of molecules and a liquid-helium-cooled satellite agree to well within a percent.

In the standard picture the hiss has a simple history.  For the first few hundred thousand years the universe was hot enough that hydrogen stayed ionised.  Free electrons scatter light, so the mixture of photons and matter was opaque, like the inside of a cloud.  As it expanded it cooled.  At about 3,000 K the electrons combined with protons into neutral hydrogen, the fog cleared, and the photons have travelled freely ever since.  Expansion has stretched their wavelengths by a factor of roughly 1,100, cooling a 3,000 K glow to 2.7 K.

That is the claim.  The rest of this article is about how much of it is measurement and how much is story.

Who predicted it — and what the popular retelling gets wrong

Both sides tell this history badly.  Textbooks credit a clean prediction that was not clean.  Creationist lists cite a 1965 "prediction" that was explicitly not a prediction.  Here is the record.

The 1948 letter, and what it says

The first published estimate of a present-day radiation temperature is a two-paragraph letter by Ralph Alpher and Robert Herman in Nature, 13 November 1948, correcting an earlier paper of George Gamow's.  The temperature is one clause of one sentence:  "The temperature of the gas at the time of condensations was 600° K, and the temperature in the universe at the present time is found to be about 5° K."²

Jim Peebles — who was in the room in 1965 and won a Nobel Prize for this field — is blunt about what that number is worth:

the paper Alpher and Herman (1948a) presents the first estimate of the present radiation temperature.  Their numerical value is not significant because it depends on the present baryon density, which was very uncertain, and they compute element buildup along the wrong path.³

Peebles then names what was durable about it:  "Their lasting contribution is the first example of how the theory of what was happening when the expanding universe was a few minutes old may be tested by measurements some 14 Gyr later."³

The numbers wandered, badly

Helge Kragh's Cosmology and Controversy is the standard history, and it keeps the ledger.  Alpher and Herman published 5 K in 1948 and again in 1949.  In 1950 they published 28 K, because they adopted a new density estimate from Alfred Behr that was ten times the one they had used before.  Gamow published 3 K in 1950 with no derivation at all.  In The Creation of the Universe he published 50 K, obtained by plugging the present age of the universe into a formula that only holds in the radiation-dominated era.  Then 7 K in 1953 and 6 K in 1956.⁴

Seven published values spanning a factor of seventeen.  The answer is 2.72548 K.

Kragh names the deeper failure, and it is worth quoting against the standard triumphal account:

even if the prediction was understood, it may have seemed of theoretical interest only;  neither Gamow, Alpher, nor Herman suggested that it might actually be detected, and if so, how.⁴

A number nobody proposed to measure is not much of a prediction.

The 1965 "40 K" was an upper limit, and the creationist list gets this one flatly wrong

A standard creationist list of failed Big Bang temperature predictions ends with "40 K, Dicke 1965."  Here is the sentence, from page 415 of Dicke, Peebles, Roll and Wilkinson's paper in the Astrophysical Journal:

Without some knowledge of the density of matter in the primordial fireball we cannot predict the present radiation temperature.  However, a rough upper limit is provided by the observation that black-body radiation at a temperature of 40° K provides an energy density of 2 × 10⁻²⁹ gm cm⁻³, very roughly the maximum total energy density compatible with the observed Hubble constant and acceleration parameter.⁵

The sentence begins by saying they cannot predict the temperature.  The 40 K is the ceiling set by not over-closing the universe.  Listing it as a failed prediction of 40 K is a misreading of the plainest possible English.

One more detail from the same paper, checked in the primary:  its reference list cites Alpher, Bethe and Gamow 1948 and Alpher, Follin and Herman 1953, but no Alpher and Herman temperature paper.⁵  The Princeton group had rediscovered the idea independently and did not know the 1948 letter existed.  That is a real failure of the earlier work — not of memory, but of impact.

What Penzias and Wilson actually wrote

The discovery paper is one and a half pages long and mentions no cosmology whatsoever.  Its title is "A Measurement of Excess Antenna Temperature at 4080 Mc/s."  The opening:

Measurements of the effective zenith noise temperature of the 20-foot horn-reflector antenna … at 4080 Mc/s have yielded a value about 3.5° K higher than expected.  This excess temperature is, within the limits of our observations, isotropic, unpolarized, and free from seasonal variations (July, 1964-April, 1965).⁶

Their measured value:  "we compute the remaining unaccounted-for antenna temperature to be 3.5° ± 1.0° K at 4080 Mc/s."⁶  Their entire theoretical commitment is one sentence:  "A possible explanation for the observed excess noise temperature is the one given by Dicke, Peebles, Roll, and Wilkinson (1965) in a companion letter in this issue."⁶

That is the whole of it.  Two radio engineers reported an isotropic, unpolarised excess and pointed at the paper next door.

Eddington and McKellar, and why their numbers do not count

The creationist counter-history runs:  Arthur Eddington computed 3.18 K for the temperature of space in 1926, before anyone proposed a Big Bang;  Andrew McKellar measured 2.3 K from interstellar cyanogen in 1941.  Both are true, and both are conceded here.

Eddington really did get 3.18 K, on page 371 of The Internal Constitution of the Stars.  But the next paragraph, on the same page, is the one that matters:

Radiation in interstellar space is about as far from thermodynamical equilibrium as it is possible to imagine, and although its density corresponds to 3°·18 it is much richer in high-frequency constituents than equilibrium radiation of that temperature.⁷

On the facing page he tabulates the temperature you would infer from starlight at each wavelength:  690 K at 6000 Å, 967 K at 4000 Å, 1750 K at 2000 Å, 4707 K at 600 Å.⁷  Eddington computed the equilibrium temperature of a black bulb sitting in starlight, and said in the same breath that starlight is nothing like a blackbody.  The microwave background is the opposite:  one temperature at every frequency, to fifty parts per million.  The two numbers agree by coincidence, and Eddington himself supplies the reason they are not the same thing.

McKellar really did get 2.3 K, and deserves more credit than he gets.  Cyanogen's first rotational level sits 3.79 wavenumbers above the ground state, and it is microwave background photons that populate it.  He was measuring the right quantity without knowing what it was:  "Thus from (3) we find, for the region of space where the CN absorption takes place, the 'rotational' temperature, T = 2°·3 K."⁸

But he hedged it three ways.  He said a thermal distribution "cannot be demonstrated but must be assumed."⁸  He doubted whether "the concept of such a temperature in a region with so low a density of both matter and radiation has any meaning."⁹  And he stated flatly that interstellar radiation "certainly does not correspond to a Planckian distribution for any one temperature."⁸  He made no cosmological claim of any kind.

Two footnotes on this history, both of which correct errors that circulate widely.

Fred Hoyle used McKellar's number against the Big Bang in 1950.  He argued that Gamow's model would give a present radiation temperature "much greater than McKellar's determination for some regions within the galaxy."³  The modern creationist use of McKellar is the exact reverse of the use his own contemporaries made of him.

And the sentence usually quoted as McKellar's caution is not McKellar's.  The line that the inferred temperature "has, of course, only a very restricted meaning" belongs to Gerhard Herzberg, who was commenting on McKellar's result in his 1950 textbook.¹⁰  It appears in neither of McKellar's 1941 papers.  Peebles's own 2014 history misattributes it, and so has this project.

The verdict, stated so it costs nothing

The temperature of the microwave background was not predicted.  Anyone who says "the Big Bang predicted 2.7 K" is overselling.

Two working cosmologists say it more sharply.  Luke Barnes and Geraint Lewis wrote a book to help critics of the standard model attack it properly, and they state it flatly:

But let's be very clear about one point.  The big bang theory does not predict the temperature of the CMB.  There is no calculation within the big bang theory that produces the number 2.725 K.¹⁰

Their reason is structural rather than apologetic.  In the model the temperature falls with time, so there is no such thing as "the" predicted value.  Three billion years ago it was 3.42 K.  Three billion years from now it will be 2.22 K.  What the theory does predict is that radiation coupled to matter in a hot early phase would decouple when atoms formed and survive as a uniform blackbody at a few kelvin today.

They also give the honest reason the omission has cost the model so little:  "no plausible alternative cosmology has predicted a perfect blackbody spectrum at 2.725 K."¹⁰

Hoyle put the objection in its sharpest form in 1994:

Had observation given 27 kelvins instead of 2.7 kelvins for the temperature, then 27 kelvins would have been entered in the catalogue.  Or 0.27 kelvin.  Or anything at all.¹¹

He is right about the temperature.  He is wrong about everything else in the measurement, and the rest of this article is why.

The blackbody spectrum

The measurement

COBE carried an instrument called FIRAS, the Far InfraRed Absolute Spectrophotometer.  It did not measure brightness at one frequency and compare it to a model.  It compared the sky, directly and continuously, against an on-board calibrator:  a cavity so nearly perfect that its emissivity, calculated and measured, was better than 0.99997, whose temperature could be dialled to match the sky.¹²  What FIRAS recorded was the difference between the sky and a known blackbody.

The difference was almost nothing.  From the full data set:

We find new tighter upper limits on general deviations from a blackbody spectrum.  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).¹²

Fifty parts per million of the peak.  Written out:  the measured spectrum and a mathematically perfect Planck curve agree to five thousandths of one percent, across the frequency range 2 to 21 inverse centimetres — 5,000 down to 480 micrometres — which spans the peak and both shoulders.¹²

Two of those numbers are worth naming, because they are the ones that constrain alternatives.

The Comptonization parameter, y.  It measures how much energy was dumped into the photons after the spectrum was fixed but by scattering that only shuffles photons in frequency without creating new ones.  Any such process moves photons from the low-frequency side to the high-frequency side, leaving a characteristic tilt.  FIRAS caps that tilt at fifteen parts per million.

The chemical potential, μ.  It measures a photon-number deficit or excess relative to a true blackbody.  Energy added early enough for scattering to reach equilibrium, but too late for photon-creating processes to top up the photon count, leaves exactly this signature.  FIRAS caps it at nine parts in a hundred thousand.

One housekeeping note, because it matters for citation hygiene.  The older, looser limits from the 1994 first-year analysis (|y| < 2.5 × 10⁻⁵, |μ| < 3.3 × 10⁻⁴) are frequently quoted alongside the 1996 numbers.¹³  They are the same instrument at an earlier stage.  Do not mix the two sets.

Why a blackbody is hard for any non-cosmological source to fake

This is where the argument lives, so it is worth being slow.

First:  adding blackbodies does not give a blackbody.  This is a theorem, not a rule of thumb.  Jens Chluba and Rashid Sunyaev worked it out in 2004:

These distortions are connected with the simple fact, that the superposition of blackbodies with different temperatures in general is not a blackbody.  We show, that in the limit of small temperature fluctuations, the superposition of blackbodies leads to a y-type spectral distortion.¹⁴

The reason is straightforward.  The Planck function is not linear in temperature.  Average two Planck curves at slightly different temperatures and the result is not the Planck curve at the average temperature — it is broader, with too much power in the tails.  That excess is precisely a y-distortion.

Now apply it.  Stars sit at a range of surface temperatures.  Dust grains sit at a range of temperatures.  Intergalactic gas, hydrogen clouds, galaxies, and anything else you might add up all sit at a range of temperatures.  Any proposal of the form "the microwave background is the summed emission of many sources" therefore predicts a y-distortion.  FIRAS caps y at 1.5 × 10⁻⁵.

Chluba and Sunyaev consider the effect so unavoidable that they propose using it:  the CMB dipole itself, seen as two blackbodies of slightly different temperature averaged over a beam, generates a known y ≈ 3.1 × 10⁻⁶ signal that can calibrate an instrument.¹⁴  The distortion from mixing temperatures is not a subtlety.  It is a tool.

Second:  you cannot build a blackbody without creating photons.  A Planck spectrum has a fixed relationship between energy density and photon number.  Compton scattering redistributes photons in frequency but conserves their number.  So scattering alone can never turn an arbitrary spectrum into a blackbody;  it can only reach the closest thing with the wrong photon count, which is a spectrum with a chemical potential.

To get all the way to μ = 0, photons have to be manufactured.  The processes that do it are double Compton emission and thermal bremsstrahlung, and both need high density and high temperature.  Chluba and Sunyaev again:

For very early energy release (z ≳ 2 × 10⁶), photon production at low frequencies by double Compton emission and the up-scattering of photons to high frequencies by Compton scattering are so efficient that the thermalization process is practically perfect for nearly arbitrary amounts of energy release … However, below this redshift the CMB spectrum becomes vulnerable, and spectral distortions that could still be observable today can form.¹⁵

Their two requirements, stated plainly:  "(i) the number of photons has to be readjusted and (ii) photons have to be redistributed over frequency to restore the shape of a blackbody."¹⁵

The window closes at a redshift of about two million — roughly one year after the beginning.  Anything that injects significant energy after that leaves a permanent scar.  FIRAS sees no scar.

Put those two facts together and the constraint is severe.  Whatever made this spectrum had to be a single medium in thermal equilibrium, optically thick, dense and hot enough to create photons, and it had to finish its work in the first year.  Today's universe cannot do it.  A screen of dust between here and the galaxies cannot do it, because it would be at a spread of temperatures and could not supply the photon-number budget.

Third, and honestly:  one alternative is constrained but not dead.  The best attempt to thermalize starlight in a static universe uses conducting "cosmic needles" — elongated iron whiskers filling intergalactic space, which absorb optical light and re-radiate in the far infrared.  Aigen Li worked through the optics in 2003:

We show here that the Rayleigh approximation is not valid since the Rayleigh criterion is not satisfied for highly conducting needles.  We also show that the available intergalactic iron dust, if modelled as infinite cylinders, is not sufficient to supply the required opacity at long wavelengths to obtain the observed isotropy and Planckian nature of the CMB.  If appealing to the antenna theory, conducting iron needles with exceedingly large elongations (> 10⁴) appear able to provide sufficient opacity to thermalize the CMB within the iron density limit.  But the applicability of the antenna theory to exceedingly thin needles of nanometer/micrometer in thickness needs to be justified.¹⁶

Two of the three calculational routes are killed.  The third survives, and requires iron needles more than ten thousand times longer than they are wide, in quantities that fill intergalactic space, with optical physics nobody has validated.

That is not a refutation and this article will not call it one.  It is also not a live research programme, and it does not escape the two harder walls.  A cloud of needles at a range of temperatures still runs into the superposition theorem.  And a static screen of needles still has to reproduce T(z) = T₀(1 + z), which is measured out to redshift 6.34 and is discussed below.

The dipole, and what it measures

Subtract the mean temperature and the biggest thing left in the sky is a dipole:  one hemisphere hotter, the opposite hemisphere cooler, by about one part in a thousand.  It is a thousand times larger than everything else in the pattern.

Planck's measurement:

The direction points into the constellation Crater, near the border with Leo.¹⁷

The standard reading is that this is not a property of the background at all.  It is the Solar System's motion through it.  Move through an isotropic bath of radiation and the light ahead is blueshifted, the light behind redshifted, and the pattern on the sky is a dipole with amplitude v/c.  The measured amplitude gives v/c = (1.23357 ± 0.00036) × 10⁻³.¹⁷

Why that reading is not an assumption

A dipole could in principle be intrinsic — a genuine hot side and cold side of the universe.  The two readings are distinguishable, and the test has been run.

If the dipole is motion, then everything else in the sky is affected too, not just the mean brightness.  Motion aberrates the pattern:  the whole anisotropy field appears slightly squeezed toward the direction of travel and stretched away from it.  Motion also modulates it.  The fluctuations look slightly larger in amplitude ahead and smaller behind.  Both effects are of order v/c ≈ 10⁻³ applied to fluctuations that are already one part in 10⁵.  They are minuscule, and they depend on entirely different measurements from the dipole itself.

Planck measured them:

Here we report a first measurement of this velocity signature using the aberration and modulation effects on the CMB temperature anisotropies, finding a component in the known dipole direction, (l, b) = (264°, 48°), of 384 km s⁻¹ ± 78 km s⁻¹ (stat.) ± 115 km s⁻¹ (syst.).  This is a significant confirmation of the expected velocity.¹⁸

384 ± 78 ± 115 km/s, from the distortion of the small-scale pattern, against 369.82 ± 0.11 km/s from the dipole itself.  The precision is far worse.  The point is that it is an independent route to the same number, and it would have given zero if the dipole were an intrinsic hot-and-cold sky rather than a velocity.

The honest problem with the dipole

There is a real and unresolved tension here, and it does not get mentioned in most popular treatments of the CMB on either side.

If the dipole is the Solar System's motion, then distant sources should show the same dipole.  Counting quasars or radio galaxies across the sky, one should find slightly more in the direction of travel — an effect of aberration plus Doppler boosting of their fluxes across the survey's flux limit.  The predicted amplitude follows from the CMB velocity with no free parameters.

Nathan Secrest and colleagues measured it in two independent catalogues, one radio and one infrared, with no shared objects:

the null hypothesis is independently rejected by the radio galaxy and quasar samples with p-value of 8.9 × 10⁻³ and 1.2 × 10⁻⁵, respectively, corresponding to 2.6σ and 4.4σ significance.  The joint significance, using sample size-weighted Z-scores, is 5.1σ.¹⁹

The direction matches the CMB dipole.  The amplitude is roughly twice what the CMB velocity predicts.  A fully independent Bayesian reanalysis of the same quasar sample by Lawrence Dam, Geraint Lewis and Brendon Brewer confirmed it rather than dissolving it:

the amplitude of the dipole is found to be 2.7 times larger than that expected from the conventional kinematic explanation of the CMB dipole, with a statistical significance of 5.7σ.²⁰

That is a genuine anomaly, at higher significance than any of the large-angle CMB anomalies discussed later in this article, and it is not resolved.  It does not overturn the kinematic reading of the CMB dipole — the aberration measurement above is a direct check on that, and the quasar dipole points the same way.  What it says is that either the source populations have an intrinsic anisotropy in the CMB frame, or something is wrong with the standard picture of large-scale isotropy.  Dam and colleagues checked the obvious systematics.  The Galactic mask makes the discrepancy worse, by leaking dipole power into higher multipoles.  Source evolution improves it, but only mildly.

An honest presentation of the CMB has to include this.  It is the strongest live anomaly in the whole subject, and it is not the one creationist writers usually cite.

The dipole as a calibration source

One footnote with a nice symmetry.  Because the dipole means the sky is a superposition of blackbodies at slightly different temperatures within any finite beam, it generates its own tiny y-distortion.  Over the whole sky the induced parameter is y_d = 2.6 × 10⁻⁷;  taking the difference between the maximum and minimum of the dipole gives y_opt = 12 y_d = 3.1 × 10⁻⁶, which Chluba and Sunyaev note is only about five times below the FIRAS upper limit.¹⁴  The same theorem that forbids a superposition of sources from making the microwave background is used to calibrate the instruments that measure it.

The anisotropy power spectrum

Subtract the mean and the dipole and what remains is a mottled pattern at the level of about one part in a hundred thousand.  It took twenty-seven years after the discovery to detect it.

COBE, 1992

COBE's Differential Microwave Radiometers found the fluctuations in the first year of data.  Their abstract:

The first year of data from the Differential Microwave Radiometers (DMR) on the Cosmic Background Explorer (COBE) show statistically significant (>7 σ) structure that is well described as scale-invariant fluctuations with a Gaussian distribution. … The structure is consistent with a thermal spectrum at 31, 53, and 90 GHz as expected for cosmic microwave background anisotropy.²¹

Their numbers:  an rms sky variation of 30 ± 5 µK smoothed to 10 degrees, an rms cosmic quadrupole of 13 ± 4 µK, and a power-law index for the primordial fluctuations of n = 1.1 ± 0.5.²¹

The frequency check in that abstract is the part outsiders skip.  Galactic synchrotron emission, free-free emission and dust all have spectra that rise or fall steeply with frequency.  A signal that has the same amplitude in temperature units at 31, 53 and 90 GHz is not any of them.  That test has been repeated at every stage since, across WMAP's five bands and Planck's nine, and it is the reason the anisotropies are not attributed to the Galaxy.

Two consequences follow immediately, and one of them was a problem.

The good one:  fluctuations this small are exactly the seeds needed to grow today's galaxies and clusters by gravity in the available time.  Smoot's abstract says so:  "The low overall fluctuation amplitude is consistent with theoretical predictions of the minimal level gravitational potential variations that would give rise to the observed present day structure."²¹

The bad one:  Danny Faulkner's chapter "Problems with the Big Bang" states "As the first data from COBE were assembled in the early 1990s, we found that the CBR was perfectly smooth."²²  That is the opposite of what COBE found.  The detection of the anisotropies is what earned Smoot and Mather the 2006 Nobel Prize in Physics.  Answers in Genesis republished the chapter in 2013 and has not annotated it.

What makes the peaks

Before recombination, photons and baryons were a single fluid.  Gravity pulled the fluid into dark-matter potential wells;  radiation pressure pushed it back out.  The result is a standing sound wave in every Fourier mode, ringing at a frequency set by the mode's wavelength and the sound speed.

When the photons decoupled, the ringing stopped everywhere at once.  Modes that happened to be at maximum compression at that instant left a large temperature contrast.  Modes caught halfway left none.  So the temperature pattern carries a harmonic series:  peaks at angular scales corresponding to modes that had completed a half-integer number of oscillations by decoupling.

Three properties of that series carry information, and each maps onto a different physical quantity.

Where the first peak sits measures the geometry.  The physical size of the sound horizon at decoupling is calculable.  How large it appears on the sky depends on how light has travelled since.  In a spatially flat universe it subtends about one degree, which is multipole 200.

How high the odd peaks are relative to the even ones measures the baryon density.  Baryons add mass to the oscillating fluid.  Extra mass makes the compressions deeper and the rarefactions shallower, like hanging a weight on a spring.  Odd-numbered peaks are compressions, even-numbered peaks are rarefactions, so more baryons means a bigger odd/even contrast.

How high the third peak is measures the dark matter.  Gravitational potentials that decay while the wave is ringing feed energy into it, which erases the odd/even pattern.  Cold dark matter keeps the potentials from decaying.  Wayne Hu and Scott Dodelson put it plainly:  "By eliminating gravitational potentials, photon-baryon acoustic oscillations eliminate the alternating peak heights from baryon loading.  The observed high third peak (Halverson et al 2001) is a good indication that cold dark matter both exists and dominates the energy density at recombination."²³

What was predicted before it was measured

This is the part that a skeptic should check line by line, because "predicted in advance" is a claim people make loosely.

1970 — the existence of a harmonic series.  Two papers, three months apart, on opposite sides of the Iron Curtain.  Rashid Sunyaev and Yakov Zel'dovich, received 11 September 1969:

A detailed investigation of the spectrum of fluctuations may, in principle, lead to an understanding of the nature of initial density perturbations since a distinct periodic dependence of the spectral density of perturbations on wavelength (mass) is peculiar to adiabatic perturbations.²⁴

They add their own caveat in the next sentence:  "Practical observations are quite difficult due to the smallness of the effects and the presence of fluctuations connected with discrete sources of radio emission."²⁴

Jim Peebles and Jer Yu, in the same year:

We conclude from Figure 8 that the experimental search for small-scale irregularities in the microwave background provides a test for the first big peak in Figure 5 (if the radiation has not suffered further scattering).²⁵

Their honesty is worth reproducing in full, because it is the standard this article is trying to meet:

It is well to bear in mind that in this calculation the initial density fluctuations are invoked in an ad hoc manner because we do not have a believable theory of how they may have originated.  Also, it is entirely possible that we have left out some relevant force, possibly that provided by a primeval magnetic field.  Our calculation thus is at best exploratory.²⁵

Neither paper predicts "the first peak will be at multipole 220."  What they predict is the existence and physical origin of a harmonic series — twenty-two years before any anisotropy was detected at all.

1994 — the position of the first peak.  Marc Kamionkowski, David Spergel and Naoshi Sugiyama, posted 3 January 1994:

In particular, the location of the primary Doppler peak depends primarily on [Ω] and is fairly insensitive to the other unknown parameters, such as [Ω_b], h, [Λ], and the shape of the power spectrum.  Therefore, measurements of CMB anisotropies on small scales may be used to determine [Ω].²⁶

and in their section 3:

If the Universe is flat but dominated by a cosmological constant, then the total mass density (non-relativistic matter plus vacuum-energy density) of the Universe is [Ω] = 1, and the Doppler peak is still located at l ≃ 200 (Bond et al. 1993).  The robustness of the prediction for the location of the Doppler peak as a function of [Ω] should come as no surprise:  This simply reflects the angular scale subtended by the horizon at the surface of last scattering.²⁶

Six years before the measurement.  And the prediction is not merely "there is a peak."  It is "its position tracks the curvature and almost nothing else," which is what turned the measurement into a test rather than a fit.

1995-97 — the alternating peak heights.  Wayne Hu, Naoshi Sugiyama and Joseph Silk, in a review posted to arXiv in April 1995 and published in Nature in March 1997:

Odd peaks thus represent the compression phase (temperature crests), whereas even peaks represent the rarefaction phase (temperature troughs), inside the potential wells.²⁷

and the falsifiable form:

Baryon drag can also increase the height of the first peak but its signal in the higher peaks is unambiguous as it predicts alternating peak heights.²⁷

The odd/even mechanism and its use as a baryon measurement are already in the April 1995 posting, before the first peak had been resolved at all:  "the difference in heights between odd and even peaks is a reasonably robust probe of the baryon-photon ratio."²⁸  The third peak was not measured until DASI and WMAP, six to eight years later.

The first test failed, and the failure was published

A skeptic reading the previous section will reasonably assume the parameters were tuned until the peaks matched.  The record shows the opposite, and this episode is the best evidence available.

BOOMERanG flew over Antarctica in 1998 and published in April 2000:

We compute the angular power spectrum of the microwave background, and find a peak at Legendre multipole ℓ_peak = (197 ± 6), with an amplitude ΔT₂₀₀ = (69 ± 8) µK.  This is consistent with that expected for cold dark matter models in a flat (euclidean) Universe, as favoured by standard inflationary scenarios.²⁹

MAXIMA-1 published weeks later with an independent balloon and independent detectors, and put the peak at ℓ ≈ 220.³⁰  Planck's answer is 220.6 ± 0.6.  MAXIMA was closer;  BOOMERanG's 197 is the number everyone remembers, and it was low.

Then the second peak came in, and it did not fit.  Fitting BOOMERanG's 2000 data gave a baryon density of Ω_b h² = 0.036 ± 0.005 under weak priors, and 0.031 ± 0.004 if flatness was assumed.³¹  Andrew Lange and colleagues said plainly what that meant:  their value was "both remarkably close to and significantly higher than that given by the observed light element abundances combined with BBN theory."³¹  That was a real, publicly discussed problem for the model, and it ran for two years.

The full BOOMERanG data set closed it, and the paper says exactly why:

These limits agree well with those found in Lange et al. (2001), with one exception.  There, the 1σ range for Ω_b h² (with the weak prior) is 0.036 ± 0.005.  Considering only the points in Figure 3 with ℓ ≤ 600 results in a 1σ limit of 0.027 ± 0.005.  This shift is presumably due to the improvement of the pointing solution and smaller error bars compared with B00.³²

Their new value was Ω_b h² = 0.022, and they say what it agrees with:  "The range for Ω_b h² is very consistent with determinations based on Big Bang Nucleosynthesis and measurements of light element abundances."³²

The disagreement was published, attributed to a pointing-reconstruction error in the telescope, and fixed by better instrument calibration.  Nothing in the theory was adjusted.  That is what a real test looks like when it initially fails.

Where the peaks are now

Planck's final release measures the extrema directly.  From Table 5 of the legacy overview paper:

Counting temperature, polarization and their cross-spectrum:  "Planck detects 18 peaks (with still only marginal detection of the eighth TT peak) and 17 troughs, for a total of 35 power spectra extrema."¹⁷

The first peak sits at ℓ = 220.6 ± 0.6.  That is a quarter of one percent.  The spacing after the first peak is near-constant at about 300, which is the harmonic series Sunyaev, Zel'dovich, Peebles and Yu described in 1970.

Six free parameters fit all of it.  Not six per peak — six for the whole spectrum, thirty-five extrema and thousands of individual bandpowers.

The cross-check that costs a skeptic the most

The peak heights give the baryon density.  So does the deuterium abundance measured in quasar absorption spectra, through big bang nucleosynthesis.  These are completely different measurements:  one is the relative heights of ripples in a microwave map, the other is the depth of an absorption line in the ultraviolet spectrum of a distant quasar.

Ryan Cooke, Max Pettini and Charles Steidel:

…we estimate that the cosmic abundance of baryons is 100 Ω_B,0 h² (BBN) = 2.166 ± 0.015 ± 0.011 … This value is based on the d(p, γ)³He reaction rate computed by Marcucci et al. (2016), and differs from the Planck CMB value by 2σ.  Alternatively, using an empirically determined d(p, γ)³He reaction rate, we estimate a baryon density of 100 Ω_B,0 h² (BBN) = 2.235 ± 0.016 ± 0.033, which is in better agreement with the CMB, albeit with larger errors.³³

Planck's value is 100 Ω_b h² = 2.237 ± 0.015.³⁴  The Atacama Cosmology Telescope, an entirely separate instrument, gets 0.0226 ± 0.0001.³⁵

So:  agreement to about three percent, from two methods that share no data, no instrument and no physics beyond nuclear reaction rates.  With a residual 1.8-2σ discrepancy that currently tracks a single laboratory cross section.³⁶

Do not write "perfect agreement."  Write what is true, which is more convincing:  a number read off the heights of ripples in a microwave map matches a number read off hydrogen and deuterium lines in quasar spectra to a few percent, and the leftover couple of sigma is blamed on a nuclear physics measurement, not on cosmology.

Polarization

Thomson scattering polarizes light when the radiation hitting the electron is not the same from every direction.  Specifically, it needs a quadrupole — hotter from two opposite sides, cooler from the other two.  At last scattering, that quadrupole is generated by the velocity field of the plasma.  So the polarization pattern is a map of how the fluid was moving when the photons left, while the temperature pattern is a map of how much it was compressed.

That distinction does all the work in this section.

E-modes and the phase relation

A polarization field on a sphere splits into two independent parts.  E-modes are the curl-free part:  the polarization directions form radial or tangential patterns around hot and cold spots.  B-modes are the divergence-free part:  the patterns swirl.  Density perturbations produce E-modes only.  B-modes require either gravitational waves or gravitational lensing.

Here is the prediction, and note where it appears.  The sentence below is in the introduction of the paper that first detected the polarization, citing three theory papers from 1997:

Since the level of the polarization depends on velocity, one expects that the peaks in the scalar E-mode power spectrum correspond to density modes that are at their highest velocities at decoupling and are therefore at minimum amplitude.  The location of the harmonic peaks in the scalar E-mode power spectrum are therefore expected to be out of phase with the peaks in the temperature spectrum.³⁷

In plain terms:  a standing wave is most compressed when it is moving slowest, and moving fastest when it is least compressed.  Compression makes temperature.  Velocity makes polarization.  So an E-mode peak must land where a temperature trough is.

Now the measurement.  Planck's Table 5 gives both series:

The polarization peaks sit on the temperature troughs, converging to under one percent at high multipole.¹⁷

This is the single most checkable advance prediction in the subject.  Anyone can put the two columns side by side.  There is no free parameter that could have been adjusted to make it come out;  the phase relation is forced by the physics of a standing wave, and it was stated in print before either spectrum was measured well enough to test it.

Planck says what it constrains:

One of the impressive consistency checks of the paradigm is that θ∗ determined from the temperature power spectrum matches to high precision those determined from the polarization power spectrum and from the cross-spectrum between temperature and polarization.  This limits the fraction of the perturbations that were not adiabatic in nature.¹⁷

The detection, and the amplitude nobody was allowed to tune

DASI, at the South Pole, reported the first detection in December 2002:

Specifically, E-mode polarization is detected at high confidence (4.9σ).  Assuming a shape for the power spectrum consistent with previous temperature measurements, the level found for the E-mode polarization is 0.80 (0.56 – 1.10), where the predicted level given previous temperature data is 0.9 – 1.1.  At 95% confidence, an upper limit of 0.59 is set to the level of B-mode polarization with the same shape and normalization as the E-mode spectrum.  The TE correlation of the temperature and E-mode polarization is detected at 95% confidence, and also found to be consistent with predictions.³⁷

Read the middle clause again.  The predicted amplitude, 0.9 to 1.1, was fixed entirely by the temperature data.  The polarization measurement then returned 0.80 with a range of 0.56 to 1.10.  Nothing was fitted to the polarization to make that happen.

The TE cross-correlation, and what it does and does not prove

Temperature and E-mode polarization are correlated, and the correlation changes sign as a function of angular scale.  Planck measures the first TE trough at ℓ = 146.0 ± 1.1 with amplitude −46.7 ± 1.4 µK², and the second at ℓ = 471.1 ± 0.7 with −74.1 ± 1.5 µK².¹⁷

The reason the first one matters is that it sits at a scale larger than the causal horizon at recombination.  A negative TE correlation there cannot be produced by anything that happened locally at last scattering, because nothing local could have coordinated the temperature and velocity fields across that separation.

Be careful about what follows.  This rules out a class of models — those in which structure is generated causally and locally at recombination, such as cosmic strings and topological defects.  It requires the perturbations to have been laid down before the modes entered the horizon.  It does not by itself select inflation.  The team that first reported the effect said so:  a pre-Big-Bang scenario and an ekpyrotic scenario also predict super-horizon fluctuations.³⁸

Two further cautions in the interest of not overclaiming.  The 33σ and 49σ figures that could be computed from the trough amplitudes above are the significance of those features, not of the statement "super-horizon fluctuations are detected."  No primary source located in this work gives a quantified significance for the latter, so no number is printed for it here.

The reionization bump

There is one more E-mode feature, and it is the odd one out.  Planck's EE spectrum has a peak at ℓ = 145 ± 3 with amplitude 1.11 ± 0.04 µK².¹⁷  That is not part of the acoustic series;  it is much too large in angular scale.

It comes from a second scattering.  When the first stars and quasars reionized the intergalactic hydrogen, free electrons reappeared, and a small fraction of microwave background photons scattered off them.  That rescattering re-polarized them, at the angular scale of the horizon at the time of reionization rather than at recombination.

The amplitude of that bump measures the total scattering probability — the optical depth τ.  Planck's value is τ = 0.0544 ± 0.0073, meaning about five percent of background photons were rescattered.³⁴  That in turn dates the bulk of reionization, and the date it gives is consistent with the independent evidence from quasar absorption spectra.

B-modes, and the retraction that should be taught

Primordial gravitational waves from inflation would leave a B-mode pattern that no density perturbation can make.  Detecting it would be extraordinary.  It has not been detected.

Lensing B-modes have been detected, and are a different thing.  Gravitational lensing by intervening matter converts a small fraction of E-modes into B-modes.  This is a prediction with no free parameters given the measured E-modes and the measured matter distribution, and it is now measured at high significance:  the BICEP2/Keck-Planck joint analysis detected lensing B-modes at 7.0σ, and Planck detects the reduction in B-mode power after delensing at 9σ.³⁹⁴⁰

Primordial B-modes have not.  In March 2014 the BICEP2 collaboration announced that they had found them:

We find an excess of B-mode power over the base lensed-ΛCDM expectation in the range 30 < ℓ < 150, inconsistent with the null hypothesis at a significance of > 5σ. … The observed B-mode power spectrum is well fit by a lensed-ΛCDM + tensor theoretical model with tensor-to-scalar ratio r = 0.20 (+0.07/−0.05), with r = 0 disfavored at 7.0σ.⁴¹

It was reported worldwide as the direct detection of inflation.  It was wrong, and the reason is instructive:  the excess was polarized emission from Galactic dust in the observed patch.  BICEP2's own abstract flagged the vulnerability honestly — "these models are not sufficiently constrained by external public data to exclude the possibility of dust emission bright enough to explain the entire excess signal"⁴¹ — but the headline claim was made anyway.

Eleven months later, a joint analysis with Planck's 353 GHz data, which measures dust polarization directly, settled it:

We find strong evidence for dust and no statistically significant evidence for tensor modes. … The final result is expressed as a likelihood curve for r, and yields an upper limit r₀.₀₅ < 0.12 at 95% confidence.³⁹

The 5σ detection became an upper limit.  The current best constraint, from the same collaboration seven years on:

The likelihood analysis yields the constraint r₀.₀₅ < 0.036 at 95% confidence.  Running maximum likelihood search on simulations we obtain unbiased results and find that σ(r) = 0.009.  These are the strongest constraints to date on primordial gravitational waves.⁴²

Three things follow, and a young-earth reader should hold all three.

First, this is a failure for inflation, not a success.  A large class of inflationary models predicted r above the current limit and are now excluded.  The BICEP/Keck paper names two casualties of its own result:  "The φ^2/3 model now lies outside the 95% contour as does the band of natural inflation models."⁴²  The measurement is a null result, and the theory has been shrinking to accommodate it.

Second, the field corrected itself in public, fast, at the cost of a Nobel-scale announcement.  Eleven months from claim to retraction, and the retraction was co-authored by the people who made the claim.  That is what integrity looks like, and it is exactly the standard young-earth work is entitled to expect of itself.  John Hartnett met it in 2018 when he withdrew his own CMB-shadows argument in the Journal of Creation.⁴³  Danny Faulkner met it in 2017 when he told creationists to stop using the missing-supernova-remnant argument.⁴⁴  Those are the models.

Third, "they announced it before it was checked" is not an argument against the CMB.  It is an argument about a specific claim that was checked and rejected.  The blackbody spectrum, the acoustic peaks and the polarization phase relation were all checked too, and they survived.

The temperature-redshift relation

This is the sharpest test in the whole subject and almost nobody uses it.

Why it is sharp

If the microwave background is relic radiation in an expanding universe, then a blackbody stays a blackbody as space expands, and its temperature falls exactly as

T(z) = T₀ × (1 + z)

with no adjustable freedom.  The reason is that expansion stretches every wavelength by the same factor, and a Planck spectrum stretched uniformly is another Planck spectrum at a lower temperature.  This is a theorem about photons on null geodesics with conserved photon number, not a model-fitting result.

Now compare the alternatives.

A static bath of radiation — the CMB as a local or ambient glow filling a non-expanding space — predicts the same temperature everywhere.  T(z) = T₀, flat.

Tired light, in which photons lose energy en route, destroys photon number as it goes.  It cannot preserve a blackbody at all, which is a separate problem, but its temperature scaling is also different.

Thermalized starlight in a static universe — the cosmic-needle proposal above — predicts a temperature set by the local starlight energy density at each epoch, which is not the (1 + z) law and would have to be arranged by hand at every redshift.

So the relation is a discriminator, and it can be measured directly, at distances where the difference is enormous.  At redshift 2.7 the prediction is 10 K against 2.7 K for a static bath.  At redshift 6.34 it is 20 K.

The molecular thermometers

The trick is to find a cloud of gas at high redshift whose molecular energy levels are close enough together that microwave background photons populate them, and whose level populations are not disturbed by anything else.  Then the excitation temperature of the molecule is the background temperature at that redshift.  This is exactly what McKellar did with cyanogen at redshift zero in 1941 — the same physics, run at cosmological distance.

Carbon monoxide works well because its rotational levels are closely spaced.  Pasquier Noterdaeme and colleagues measured five CO absorption systems in quasar spectra with the VLT:

Their combined fit, allowing a deviation index β where T(z) = T₀(1 + z)^1−β:

Combining our precise measurements with previous constraints, we obtain T_CMB(z) = (2.725 ± 0.002) × (1 + z)^1−β K with β = −0.007 ± 0.027, a more than two-fold improvement in precision.⁴⁵

β = 0 is the standard prediction.  The measurement is −0.007 ± 0.027.

The cluster thermometer

A completely different method uses the thermal Sunyaev-Zel'dovich effect.  When background photons cross the hot gas in a galaxy cluster, some scatter off fast electrons.  The resulting spectral distortion has a shape that depends on the background temperature at the cluster's redshift, so measuring the distortion in several frequency bands recovers T(z) without any molecular physics at all.

Guillaume Hurier and colleagues, using Planck:

Focusing on the redshift dependance of T_CMB, we obtain T_CMB(z) = (2.726 ± 0.001) × (1 + z)^1−β K with β = 0.009 ± 0.017, which improves on previous constraints.  Combined with measurements of molecular species absorptions, we derive β = 0.006 ± 0.013.⁴⁶

Ivan de Martino and colleagues, using 481 X-ray-selected clusters and a different analysis:

If the CMB black-body temperature scales with redshift as T(z) = T₀(1 + z)^1−α, we constrain deviations of adiabatic evolution to be α = −0.007 ± 0.013, consistent with the temperature-redshift relation of the standard cosmological model.⁴⁷

Three groups, three data sets, two entirely different physical mechanisms, all landing on zero deviation with errors of one to three percent.

The record holder

In 2022 Dominik Riechers and colleagues pushed it much further, using water absorption against the background in a dusty starburst galaxy:

The strength of this effect implies a microwave background temperature of 16.4-30.2 K (1σ range) at z = 6.34, which is consistent with a background temperature increase with redshift as expected from the standard ΛCDM cosmology.⁴⁸

The prediction at that redshift is 20.0 K.⁴⁸  The measurement is loose.  The 1σ range is nearly a factor of two wide.  But it is at redshift 6.34, a lookback time of 12.8 billion years in the standard chronology, and it is 6 to 11 times the local value.

What this rules out

Any model in which the microwave background is a local, ambient or contemporary phenomenon predicts a flat T(z).  The measured deviation index is 0.007 ± 0.027.  A flat relation would require β = 1, which is thirty-seven standard deviations away.

The dilemma this poses for a static cosmology is worth stating precisely, because it is easy to state loosely and get it wrong.

If the radiation bath is genuinely static — the same everywhere, at rest in one frame — then a cloud at redshift 2.4 is moving through it at a large fraction of the speed of light.  Such a cloud does not see a hotter blackbody.  It sees a strongly aberrated field, ranging from T₀/(1+z) behind to T₀(1+z) ahead, whose mean energy density corresponds to a far lower temperature than T₀(1+z).  The observed excitation temperatures do not match that picture.

If instead the radiation bath is comoving with the clouds — cooling as they recede, so that each cloud sits at rest in a locally isotropic bath at T₀(1+z) — then the bath is doing exactly what expansion does.  Calling it something other than expansion does not change the physics.

So the honest statement is:  a static bath of radiation is excluded, and a comoving bath that reproduces the measurement is the expanding-universe answer under a different name.  That is a much stronger constraint than most alternatives realise they are under.

Is it actually behind everything?

Every argument above assumes the microwave background comes from farther away than the galaxies and clusters.  That assumption is testable, and it has been tested two ways.

It is gravitationally lensed

Mass between here and the last-scattering surface deflects the photons, slightly remapping the pattern.  The effect is small, deflections of a few arcminutes, but it has a specific statistical signature that can be extracted.

Planck measures it at 40σ.⁴⁰  The Atacama Cosmology Telescope measures it independently at 43σ:

We determine the amplitude of the CMB lensing power spectrum at 2.3% precision (43σ significance). … This signal-to-noise ratio independently matches the 42σ achieved in the latest Planck lensing analysis and is competitive with the precision achieved in any galaxy weak lensing analysis.⁴⁹

Lensing also smears the acoustic peaks slightly.  Planck can remove the smearing using its own reconstructed lensing map and detects the resulting sharpening at 11σ from Planck alone and 15σ when combined with the cosmic infrared background.⁴⁰

Individual clusters have been caught doing it, in temperature and in polarization separately, and the masses recovered agree with masses from optical weak lensing and from the X-ray gas.⁵⁰

A local emission source cannot be lensed by a cluster at redshift 0.5.  That is the whole argument, and it is a 43σ argument.

This matters because a widely cited creationist claim runs the other way.  Richard Lieu and Jonathan Mittaz published a 2005 paper titled "On the Absence of Gravitational Lensing of the Cosmic Microwave Background," and John Hartnett's flagship anti-Big-Bang article for Creation magazine rests on Lieu.  Three things are wrong with that use of it, and the full treatment is in the standard model overview.  In short:  the paper did not claim a null lensing detection but a smearing of peak positions;⁵¹  it was rebutted by name in the standard review of the subject within six months;⁵²  and the effect it said was missing is now measured well enough to be subtracted.  In 2005 CMB lensing genuinely had not been detected, and theirs was a legitimate challenge that failed.  Citing a 2005 paper in 2026 as though the twenty years since had not happened is not legitimate.

Ten thousand clusters have been found by their shadow in it

When background photons cross the hot gas in a cluster, some gain energy by scattering off fast electrons.  The result is a distortion with a fixed spectral shape, a deficit below 218 GHz and an excess above it, whose brightness does not fade with distance.  It is a fractional change in a background that is the same everywhere.

The Atacama Cosmology Telescope's 2026 catalogue reports 10,040 clusters found this way, including 1,180 beyond redshift 1 and 124 beyond redshift 1.5.⁵³  Planck's all-sky catalogue lists 1,653 detections with 1,203 confirmed.⁵⁴  The South Pole Telescope confirmed 516 of 677 candidates over 2,500 square degrees.⁵⁴

These clusters were found by looking for the shadow, then confirmed by optical and X-ray telescopes.  An artifact does not work as a finding chart that keeps turning up real objects, and a background in front of the clusters cannot be scattered by their gas.

One caution against overclaiming.  The relativistic correction to this effect — a percent-level refinement — sits at only about 2σ, and the paper that claimed a 5σ detection was withdrawn by its own authors after an error was found in review.⁵⁵  Do not cite that number.

The large-angle anomalies

At the largest angular scales the microwave background does things ΛCDM does not predict.  These are real, and creationist writers cite them constantly.  Both the dismissal and the celebration are wrong, and this section tries to give each one its actual number.

The best statement of the situation is Planck's own, from the isotropy paper:

The existence of these features is uncontested, but, given the modest significances at which they deviate from the standard ΛCDM cosmological model, and the a posteriori nature of their detection, the extent to which they provide evidence for a violation of isotropy in the CMB remains unclear.  It is plausible that they are indeed simply statistical fluctuations.⁵⁶

The five features

The low quadrupole and the missing large-angle correlation.  The ℓ = 2 multipole is smaller than expected, and more generally the two-point angular correlation function is nearly flat beyond about 60 degrees, where ΛCDM predicts a modest negative tail.  The statistic usually used is S(1/2), the integrated squared correlation over that range.

The quadrupole-octopole alignment, popularly the "axis of evil."  The quadrupole and octopole are each unusually planar, and their preferred axes point in nearly the same direction — a direction that is uncomfortably close to the ecliptic pole and the dipole direction.

The hemispherical power asymmetry.  Fluctuation power is slightly larger in one half of the sky than the other.  Planck fits it as a dipole modulation of amplitude 0.070 (+0.032/−0.015) toward Galactic (l, b) = (221°, −22°) ± 31°, over the multipole range 2 to 220 — the range chosen because it gives the lowest p-value.⁵⁶  Planck adds a caution that is easy to miss:  "Removing angular scales ℓ < 100 eliminates the significance of the modulation entirely."⁵⁶

The parity asymmetry.  Odd multipoles carry more power than even ones at low ℓ.

The Cold Spot.  An unusually cold region about ten degrees across in the southern sky, with a central decrement of roughly −150 µK.

The significances, and the reason to be careful

Every one of these was found by looking at the map and noticing something, then constructing a statistic to quantify what had been noticed.  That is the "look-elsewhere effect," and it is not a technicality.  Planck quantified it for the hemispherical asymmetry:

In PCIS15 we applied a look-elsewhere correction to the corresponding results and demonstrated that if we allow the range of ℓ_max to vary (as we have done here) then small p-values (of order 0.1-1 %) occur of order 10 % of the time.⁵⁶

Read that carefully.  A p-value of 0.1 to 1 percent, nominally 2.6 to 3.3σ, arises by chance one time in ten once you allow yourself to choose the range you test.

Charles Bennett and the WMAP team made the point with an example that should be memorised:

one of the authors (L.P.) noted that the initials of Stephen Hawking appear in the temperature map… We pose the question, what is the probability of this occurrence?  It is certainly infinitesimal; in fact, much less likely than several claimed cosmological anomalies.  Yet, we do not take this anomaly seriously because it is silly.⁵⁷

What changed in 2026

The anomaly significances depend heavily on how much of the sky is thrown away to avoid the Milky Way.  Planck's standard analysis masks 26 percent.  Laura Herold, Graeme Addison, Charles Bennett and colleagues re-derived foreground-cleaned maps that need only a 1 percent mask, and recomputed all five:

Their summary:

We find a notable impact of the choice of galactic mask on the significance of two anomalies: the significance of the low real-space correlation function and of the local-variance asymmetry reduces from ∼ 3σ for the Planck common mask with 26% masked fraction to ∼ 2σ for the 1% mask.⁵⁸

and their conclusion:

Overall, for an alternative physical model to be convincingly favored over ΛCDM with statistically isotropic Gaussian fluctuations, it would need to explain multiple CMB anomalies, or better describe some other type of measurement in addition to a CMB anomaly.⁵⁸

Note that the alignment went the other way:  using more sky made it stronger, not weaker.  This is not a paper that made the anomalies go away.  It is a paper that showed the numbers move when the mask moves, which is itself a reason not to build a cosmology on them.

They also flag the correction nobody has managed to apply:  "in order to assess an accurate estimate of the significance, a more thorough study of the look-elsewhere effect is necessary, which would lower the significance of the individual anomaly tests."⁵⁸

Two of the five weakened.  None reaches 5σ.  The low quadrupole itself, measured on the same maps in the companion paper, is 2.2σ.⁵⁸

The Cold Spot's best explanation is dead

For a decade the leading physical explanation of the Cold Spot was a supervoid on the line of sight, cooling photons through the integrated Sachs-Wolfe and Rees-Sciama effects.  A large void really was found in that direction — the Eridanus supervoid, detected in the Dark Energy Survey — but it accounts for only 10 to 20 percent of the decrement.

Then the lensing test was run.  A void deep enough to make the Cold Spot would also leave a lensing signature, and the signature is absent:

Here we analyze the lensing signal in the Planck CMB data and rule out the hypothesis that the Cold Spot could be due to a large void located anywhere between us and the surface of last scattering.  In particular, computing the evidence ratio we find that a model with a large void is disfavored compared to ΛCDM, with odds 1 : 13 (1 : 20) for SMICA (NILC) maps, compared to the original odds 56 : 1 (21 : 1) using temperature data alone.⁵⁹

Odds of 56 to 1 for a void became 13 to 1 against.  The authors state it plainly:  "The absence of a clear lensing signature now puts the supervoid interpretation to rest."⁵⁹

So the Cold Spot is a real cold spot, is now unexplained by any structure, and is most likely a fluctuation.  Anyone using it as evidence against the standard model should notice that its removal removes the only physical explanation anyone had.

What the anomaly advocates actually argue

The strongest case for taking them seriously is not any single number.  Dominik Schwarz, Craig Copi, Dragan Huterer and Glenn Starkman argue that three of them are statistically independent "atoms," so the joint probability is what matters:  "it is quite unlikely that two of them, or even all three of them are statistical flukes (the corresponding p-values would be at most ~10⁻⁵)."⁶⁰

They are also honest about the weak points.  On the key statistic:  "Undoubtedly, S₁/₂ is an ad hoc and a posteriori statistic."⁶⁰  On whether more data can help:  "Since all three missions observed the same sky, we could be unlucky and live in a very atypical realization of the Universe."⁶¹  And they explicitly rule out the explanation a geocentric reading would need:  "we ignore speculations on instrumental effects, as it seems to us that the consistency of WMAP and Planck 2015 results makes such an explanation quite unlikely."⁶⁰

That last point is the one that matters for creationist use of the alignment.  If the alignment with the ecliptic were caused by something in the Solar System — dust, zodiacal light, an instrumental artifact tied to the scan pattern — it would be a foreground, and foregrounds have frequency spectra.  The alignment is present in maps built from different frequency bands by different instruments a decade apart.  A real solar-system signal would not be.

What would settle it

Polarization.  The temperature maps are already cosmic-variance limited:  there is only one sky and Planck has measured all of it that is measurable.  Planck says so:  determining whether the anomalies are flukes "cannot be determined by further investigation of the temperature fluctuations on the angular scales of interest, since those data are already cosmic-variance limited."⁵⁶

Polarization is a partly independent realization of the same underlying field.  If the anomalies are physical they should leave a polarization signature;  if they are flukes they should not.  So far the polarization evidence is close to null.  An analysis of the hemispherical asymmetry in Planck's PR4 polarization maps finds "a slightly lower significance (2.8% for the Sevem method) for the amplitude of the E-mode local variance dipole as well as a significant variability with the considered mask."⁶²  A p-value of 2.8 percent is not a detection.  The Simons Observatory and LiteBIRD will do better.

The honest position is narrow and worth stating in full.  The anomalies are real features that ΛCDM does not predict.  No rival model retrodicts them either.  None reaches 5σ.  Two of five weakened in 2026 when more sky was used.  The Cold Spot's physical explanation is dead.  And the alignment does not imply geocentrism, because a real geocentric signal would be a foreground, and the frequency-independence rules that out.

What young-earth and alternative writers say about the CMB

This section is the reason the article exists.  The microwave background is the piece of evidence most often waved away in creationist writing, and the waving away does not hold.

There is no accepted creationist explanation, and a creationist said so

Andrew Repp surveyed fifty-seven years of creation astronomy for the Creation Research Society Quarterly in 2022.  On the CMB:

However, there is still no generally accepted explanation for the CMB in a creationary framework:  for instance, Humphreys (1994b) explains it as relict light from the first day, whereas Faulkner (2016) explains it as blackbody radiation from a water boundary ('the waters above').⁶³

And on the power spectrum:

The situation worsens when we consider the power spectrum of the CMB—that is, the patterns exhibited by its minute temperature variations (anisotropies) … The yet more detailed CMB mapping by Planck … seems to have laid to rest the idea that these anisotropies are artificial.  However, there has been little if any attempt to explain them within a creation framework.⁶³

That is the state of play, stated from inside the movement, in the movement's own journal, four years ago.  This review found no later creationist treatment of the power spectrum.

Humphreys' two proposals, in his own words

Repp's one-line summary of Russell Humphreys is fair for 1994 and out of date since 2014.  Humphreys has offered two different origins for the CMB, and the second replaces the first.

1994:  the Big Bang's mechanism inside a bounded universe.  In Starlight and Time, the universe begins as a ball of water that collapses, heats, and is then stretched out.  When space is about a thousand times smaller than today, the plasma in the "expanse" between the waters below and the waters above cools to about 3,000 K and turns transparent.  Then:

At this point, thermal radiation in the expanse would be very uniform and have a blackbody spectrum, having been surrounded by optically thick walls above and below during the previous stage of expansion. … At the end of the expansion, the radiation temperature would have dropped to the 2.74 Kelvin we see today.  This explanation of the cosmic microwave background is not too different from that provided by the big bang theories, except for the boundedness and the optically thick "walls" around the expanse.⁶⁴

This is the strongest creationist position on the CMB, and the reason is plain:  it keeps the physics that makes the blackbody, the T(z) scaling and the acoustic peaks, a hot opaque phase followed by expansion, and changes only the boundary and the clock.  The cost is that it grants most of what this article argues.  It also does not predict the temperature:  a factor of about a thousand is what the observed 2.74 K requires (3,000 ÷ 2.74 ≈ 1,100), and the passage gives no independent reason for it.  The book never works out the anisotropies;  it records only that the radiation varies with direction by "no greater than one part in 100,000."⁶⁴

2014:  Unruh radiation from an accelerating "fabric of space."  Humphreys has since moved away from the expansion that the 1994 mechanism needs.  He writes that in Starlight and Time "I interpreted scriptures mentioning a stretching-out of the heavens as supporting a large expansion of the universe.  But in 2011 John Hartnett pointed out a scriptural reason why that interpretation is probably incorrect," and he now looks for redshifts caused by stretching "without much expansion."⁶⁵  His new source for the CMB is the Unruh effect, William Unruh's 1976 result that an accelerating observer finds itself in a thermal bath.  If space itself is being accelerated in a fourth spatial dimension, the particles bound in it would glow:  "If the cosmic acceleration is large enough, we should be able to detect such radiation, and it should have the spectrum of a black body."⁶⁵

Credit where it is due:  this is the only creationist proposal that gets a Planck-shaped spectrum from a named physical law rather than from a surface assumed to be at the right temperature.  Three problems remain.

1. The temperature is an input, not a prediction. Humphreys uses the measured 2.725 K to find the acceleration, not the other way round: "If indeed the CMB is caused by cosmic acceleration, then we can calculate the amount of acceleration by solving eq. (11) for g," which gives 6.721 × 10²⁰ m/s².⁶⁵ Like Faulkner's water, it can be fitted to any temperature.
2. The emitter is transparent, and the radiation is made close by. Humphreys states that "The fabric of space is exceedingly transparent," so the radiation would arrive "from a wide range of distances."⁶⁵ Two of this article's checks bite here. A medium that barely absorbs also barely emits (the Kirchhoff point raised against Lisle below), so it would fall short of the full blackbody intensity FIRAS measured. And radiation produced everywhere, including between us and the clusters, could not be lensed or shadowed by those clusters as a whole background. A defender could reply that every detector riding in the accelerating membrane feels the Unruh bath directly, with no distant emitters needed; the paper does not make that argument, and a bath felt locally would face the lensing objection even more sharply. These are this article's objections; no published exchange on them was found.
3. It is silent on everything else. The paper never mentions the anisotropies, the CMB's polarization, lensing, or the temperature-redshift relation, so whether the model can reproduce T₀(1 + z) is untested. Its abstract promises later papers on galaxy redshifts and the Pioneer anomaly, not on the power spectrum.

Faulkner's water boundary, in his own words

Danny Faulkner's 2016 proposal is that the CMB is thermal emission from a shell of water at the edge of the universe — the "waters above" of Genesis 1:7 — redshifted down to microwave frequencies.  It is a serious attempt, and Repp names it beside Humphreys'.

His own worked example assumes water at 300 K at redshift 10, and gets "a blackbody spectrum of 30K" against the observed 2.7255 K — off by a factor of eleven.⁶⁶  He then says why that does not worry him:

Here I arbitrarily picked values for the temperature and redshift for the water … At this time I have no model expectations of exact temperature of the water or its redshift, but it is easy to pick pairs of their values that would result in a match with the CMB.⁶⁶

Give him credit for the candour;  most proposals do not carry their own disclaimer.  But a model that can be tuned to any temperature predicts none.  And the temperature is the easy part.  A shell of water at a single temperature would produce a blackbody, so the spectrum is not the difficulty.  The difficulty is everything else:  the acoustic peaks, the polarization phase relation, the lensing, and T(z).  The proposal addresses none of them and does not claim to.

Lisle's claim, and why it is checkably wrong

Jason Lisle's argument is the most commonly repeated version, so it is worth quoting at length:

So, this cosmic microwave background (CMB) wasn't really "predicted" by the big bang, so much as it was offered as evidence for the big bang after its discovery.  But it is at least consistent with the big bang, right?  Well, yes.  And it is also consistent with any universe that has an average temperature.  The physics of electrodynamics predicts that anything that has a positive temperature must radiate electromagnetic energy.  And the wavelength of that energy will be a function of the temperature of the source.  So any universe with an average equilibrium temperature of 2.7 degrees will have a CMB.  The expectation of a CMB is neither unique nor robust.⁶⁷

The first sentence is correct and this article agrees with it.  The rest is wrong, and it is wrong for a reason that can be checked without trusting anybody.

"Any universe with an average equilibrium temperature of 2.7 degrees will have a CMB" is false.  A universe with an average temperature of 2.7 K is a universe containing things at a range of temperatures whose average is 2.7 K.  The sum of blackbodies at different temperatures is not a blackbody — it is a Planck curve plus a y-distortion, by the Chluba-Sunyaev theorem.¹⁴  FIRAS caps that distortion at fifteen parts per million.  So an average temperature does not give you the observed spectrum;  it gives you a spectrum that FIRAS would have rejected.

And "anything with a positive temperature must radiate" does not give you a Planck spectrum either.  Kirchhoff's law says a body radiates as a blackbody only if it also absorbs as one — that is, only if it is optically thick at every wavelength being measured.  The universe today is transparent at microwave frequencies.  That is precisely why we can see through it to redshift 1,100.  A transparent medium cannot be in radiative equilibrium with the radiation passing through it.

And the photon-number problem stands.  Reaching a true Planck spectrum requires processes that create photons.  Compton scattering conserves photon number.  A local equilibrium at 2.7 K in a transparent universe has no mechanism to supply the photon budget.¹⁵

Three independent reasons, each checkable, each in a mainstream physics paper.  The claim that the CMB is generic is the single weakest argument in the creationist repertoire on this subject, and it should be retired.

The Lieu-Mittaz argument, and Hartnett's own precedent

John Hartnett's Creation magazine treatment leans on Richard Lieu's 2005 lensing claim.⁶⁸  That argument is dead;  the reasons are in the lensing section above.

The more useful thing about Hartnett is that he has already done the right thing once.  In 2018 he withdrew his own "big bang casts no shadows" argument in the Journal of Creation:

Without anything to contradict this new result, and the analysis seems strong, one must entertain the possibility that the anomaly first found by Lieu et al in 2006 has been adequately explained… At best this no-shadow argument is now equivocal and hence I suggest that it should no longer be used as an argument against the big bang hypothesis.⁴³

Creation Ministries International hosts the retraction under the slug cmb-shadows-argument-do-not-use.  That is the standard.  The Lieu lensing argument earned the same treatment in 2006 and has not received it.

The anomalies as evidence for geocentrism

Some creationist writing takes the quadrupole-octopole alignment with the ecliptic as evidence that Earth occupies a special place.  Hartnett has written that "this is the evidence we would expect for a galactocentric universe, with certain special features even attributed to the solar system," while elsewhere explicitly disclaiming geocentrism:  "I do not believe in nor promote geocentrism."⁶⁹⁷⁰

The argument does not work, and the reason is simple.  If the alignment were caused by anything in the Solar System — zodiacal dust, an instrument artifact tied to the scan pattern, anything local — it would be a foreground.  Foregrounds have frequency spectra.  This one does not:  it is present in maps made by two different satellites, a decade apart, from different frequency bands, and the cleaned maps agree.  Schwarz, Copi, Huterer and Starkman rule out the instrumental reading for exactly that reason:  "we ignore speculations on instrumental effects, as it seems to us that the consistency of WMAP and Planck 2015 results makes such an explanation quite unlikely."⁶⁰

A geocentric or heliocentric signal would be a foreground.  The alignment is not a foreground.  Therefore the alignment is not a geocentric signal.

One more caution.  Danny Faulkner's 2018 Answers in Genesis article on the axis of evil and the Cold Spot contains at least five factual errors, including describing the axis of evil as "a long stretch of warmer than average temperature" (it is an alignment of multipole axes, not a hot band) and saying "COBE found that the CMB was perfectly smooth" (COBE detected the anisotropies).  It should not be cited.

Verschuur's correlation claim

Gerrit Verschuur has argued since 2007 that features in the WMAP maps correlate with Galactic neutral hydrogen structures, which would make some of the "CMB" a local foreground.  His numbers are morphological pair counts — 41 close pairs against 23 expected in one region, 22 against 13 in another — with no p-value or significance stated anywhere in the paper.⁷¹

The one quantitative test of the claim was run by Kate Land and Anže Slosar, who "find no statistically significant correlation."⁷²  Their methodological warning is the same one Bennett later used on the anomalies:  "the human brain is particularly susceptible to discerning structure in random fields.  After all, one can easily be convinced that Stephen Hawking's initials are imprinted in the WMAP ILC map."⁷²

Verschuur's later papers have almost no citations, and his 2010 reply concedes that the statistical test "casts little additional light."  The fair verdict is that the claim is dormant rather than formally refuted.

Half of it is real, and should be granted.  Susan Clark and colleagues showed in 2015 that Galactic hydrogen filaments genuinely do trace the orientation of Planck's 353 GHz polarized dust emission — a correlation now used to build B-mode foreground templates.⁷³  Interstellar hydrogen does leave a fingerprint in microwave maps.  It just is not the primary anisotropies.

And there is a decisive argument neither side ran.  Galactic free-free emission a hundred parsecs away cannot be gravitationally lensed by clusters at redshift 0.5 to 5.  The microwave background is, at 43σ.

What would falsify this

A body of evidence that cannot be broken is not evidence.  Here is what would break this one.

A spectral distortion.  Find |y| or |μ| above the FIRAS limits and the thermalization story fails.  This is a live experimental target;  a proposed mission called PIXIE would improve the limits by three orders of magnitude, and the standard model itself predicts a small y-distortion from reionization and structure formation that is just below current sensitivity.  If the measured distortion came in with the wrong sign or the wrong magnitude, the model would be in serious trouble.

A failure of T(z).  Measure the background temperature at high redshift and get something other than T₀(1 + z).  The current constraint on the deviation index is ±0.027 out to redshift 3, and much looser at 6.34.  A well-measured deviation of a few percent would rule out adiabatic expansion of a conserved photon field.

Acoustic peaks in the wrong place.  The peak positions and the polarization phase relation are fixed by physics with no freedom once the six parameters are set.  A measured EE peak landing on a TT peak rather than a trough would falsify the adiabatic acoustic picture outright.

A cluster that is not lensing the background.  CMB-cluster lensing is measured object by object.  Systematically null results for well-characterized clusters would break the claim that the background is behind them.

A large-angle anomaly confirmed in polarization at high significance.  If the hemispherical asymmetry shows up in polarization at 5σ with a consistent axis, it is physical, and ΛCDM does not have a mechanism for it.  This is the test the field is waiting for.

A cluster too massive too early.  The SZ catalogues are also a test:  the ACT team explicitly checked and reported "we find no cluster with a mass and redshift combination high enough to falsify the standard ΛCDM cosmology with Gaussian initial perturbations."⁵³  A single sufficiently extreme object would do it.

None of these has happened.  That is the honest state of the evidence, and pretending otherwise helps nobody.

Objections and responses

"The Big Bang predicted 2.7 K."  No, it did not, and this article says so at length.  The published predictions ran from 3 K to 50 K over eight years.  The theory predicts that a hot early phase leaves a blackbody at a few kelvin whose temperature falls with time;  it does not produce the number 2.725.  Two working cosmologists writing a book for critics say exactly this.¹⁰

"Eddington predicted 3.18 K in 1926, before the Big Bang."  He calculated the temperature a black bulb would reach in starlight, and on the same page said starlight "is about as far from thermodynamical equilibrium as it is possible to imagine."⁷  His own table gives equivalent temperatures from 690 K to 4707 K depending on wavelength.⁷  A quantity that takes different values at different wavelengths is the opposite of what the CMB is.

"McKellar measured it in 1941 and it had nothing to do with cosmology."  True, and he deserves credit.  He also said a thermal distribution "must be assumed," doubted the concept had meaning, and stated that interstellar radiation "certainly does not correspond to a Planckian distribution for any one temperature."⁸  Fred Hoyle used McKellar's number against the Big Bang in 1950.³

"Any universe with an average temperature of 2.7 K would have a CMB."  No.  A universe with an average temperature has sources at a range of temperatures, and a sum of blackbodies at different temperatures is not a blackbody.¹⁴  Reaching a true Planck spectrum also requires creating photons, which needs a dense, hot, optically thick medium.¹⁵  Today's universe is transparent at microwave frequencies, which is why we can see through it.

"The CMB is starlight thermalized by intergalactic dust."  The only worked version uses conducting iron needles, and Aigen Li showed two of its three calculations are invalid.¹⁶  The surviving version needs needles with aspect ratios above ten thousand and optical physics nobody has validated.  It also has to beat the superposition theorem and reproduce T(z) = T₀(1 + z) out to redshift 6.34.

"The CMB is not gravitationally lensed, so it is not behind the clusters."  It is lensed, at 40σ by Planck and 43σ by ACT.⁴⁰⁴⁹  The 2005 paper this claim rests on was rebutted by name in the standard review within six months.⁵²

"The power spectrum is just a six-parameter fit — of course it matches."  Six parameters fit thirty-five measured extrema and thousands of bandpowers across temperature, polarization and their cross-correlation.¹⁷  More to the point, the peak position was published in 1994 before the measurement, the alternating peak heights in 1995, and the polarization phase relation before either spectrum was resolved.  And the first fit failed:  BOOMERanG's 2000 baryon density was "significantly higher" than the light-element value, and its own authors said so.³¹  A tuned fit does not do that.

"The Big Bang has to invent dark matter and dark energy to make it work."  Barnes and Lewis answer this one better than an apologist could:  the darkness of the night sky, Hubble's law, supernova time dilation, the CMB blackbody and its temperature-redshift relation, the Lyman-alpha forest and nucleosynthesis "do not depend on the amount of dark matter and dark energy."¹⁰  Everything in this article except the third-peak height survives dark matter's disappearance.

"The axis of evil and the Cold Spot show the standard model is failing."  They show it is incomplete on the largest scales.  Every one sits between 2σ and 3.5σ, all were found after looking, and two of five weakened in 2026 when more sky was used.⁵⁸  The Cold Spot's only physical explanation — a supervoid — is now excluded by lensing at odds of 13 to 1 against.⁵⁹  Nobody has a competing model that retrodicts them either.

"The CMB proves the universe is billions of years old, so a young universe is refuted."  The CMB does not date anything by itself.  What it does is constrain the physics of a hot dense phase, and it is the constraint no young-universe model has answered.  Every young-earth cosmology on offer reproduces standard observational cosmology by construction, which means each of them inherits the CMB power spectrum as an unexplained fact rather than a prediction.  That is a real problem and this article does not soften it.

"You are just repeating what secular astronomers say."  Check who says what.  The statement that there is no accepted creationist explanation for the CMB is in the Creation Research Society Quarterly.⁶³  The admission that the water-boundary temperature was "arbitrarily picked" is Faulkner's own.⁶⁶  The withdrawal of the no-shadow argument is Hartnett's own.⁴³  The claim that COBE found a smooth CMB is Faulkner's, and it is false.²²  Nearly every negative verdict here was reached first by someone inside the movement.

Where this leaves things

The microwave background is not a proof of the Big Bang, and it is not a triumph of prediction.  Its temperature was never predicted, and the people who came closest to predicting it never suggested looking for it.

What it is, is a very large, very precise, internally cross-checked body of measurement with a small number of unexplained features at the largest angular scales.  The spectrum is a blackbody to fifty parts per million.  The temperature scales as (1 + z) out to redshift 6.34.  The ripples form a harmonic series whose positions and polarization phases were computed in advance.  The whole thing is lensed by foreground mass at 43σ and has been used to discover ten thousand galaxy clusters.

Against that, the young-earth side has three proposals from two authors.  Humphreys' 1994 version borrows the Big Bang's own mechanism, and he has since moved away from the expansion it needs.  His 2014 Unruh version and Faulkner's water boundary both take the temperature as an input, and Faulkner says in as many words that his was picked by hand.  None of the three addresses the power spectrum.  Andrew Repp asked for a creationist treatment of the power spectrum in 2022, and this review found none since.

Saying the evidence is weak would be false, and a skeptical reader would catch it in ten minutes.  The honest position is that the microwave background is the strongest single argument for a hot dense early universe, that several arguments made against it are dead and should be retired, and that the work of answering it has not been done.

Sources