Uranium-Series Dating of Corals and Speleothems

"Of course, a simpler explanation is that the radiometric dating is incorrect and that the speleothems grew only after the ice had melted." — Emil Silvestru, karst geologist, Creation Ministries International, 2012¹

"Maintaining the required gradient of thorium isotope ²³⁰Th over 3.6 meters for 1000 years, much less 10,000 years, seems exceedingly unlikely." — Tyler Coplen, US Geological Survey, in Science, arguing that a published set of uranium-series ages varies with sampling position by 2,500 years per metre²

"In searching the literature for growth rates, estimated growth rates that depend upon radiometric methods were avoided, since they always seem to greatly exaggerate the time." — Michael Oard, Journal of Creation, 2020³

Summary

Coral skeleton and cave calcite crystallise out of water that carries uranium and almost no thorium.  Thorium is insoluble.  It hydrolyses and sticks to particles within days.  So the daughter isotope starts at zero for a chemical reason, not because anyone assumed it.  Then ²³⁸U decays to ²³⁴U to ²³⁰Th, the thorium builds up, and the ratio gives an age.  A second chain in the same crystal, ²³⁵U to ²³¹Pa, gives an independent age from a different parent, a different daughter, a different element and a different half-life.

Five points, and the third is the one this article exists for.

1. The method's internal checks are real and they are tight. The two chains agree on the same fragment closely enough that agreement within 2σ is a routine acceptance criterion rather than a headline result.⁴ ²³⁰Th ages agree with tree rings at the young end,⁵ with Greenland ice-layer counts to within 800 years over 60,000,⁶ and with radiocarbon in a way that runs the wrong direction for accelerated ¹⁴C decay at all nineteen levels ever measured on the founding dataset.⁵
2. The method's failures are real too, and its own practitioners published them. Open-system behaviour in corals is "frequently observed."⁷ IntCal20 threw out every coral older than 25,000 calendar years.⁸ The strictest published screening protocol "would exclude nearly all of the (published) coral data presently available."⁴ And at Lake Mungo, uranium-series, ESR and OSL all agreed on 62,000 years and all were wrong by 55 per cent.^{9 10}
3. None of the internal checks refute accelerated decay, and one young-earth author reached the neighbouring conclusion before this article did. Thorium ingrowth goes as 1 − exp(−kλt). Multiply every decay constant by the same factor k and the apparent age is multiplied by exactly k. Secular equilibrium, the seawater ²³⁴U excess and Pa/Th concordance are all invariant under that operation. Anyone citing Pa/Th concordance as a refutation of accelerated decay has not done the algebra. Nathan Mogk, at the 2023 International Conference on Creationism, named the same four long-lived intermediates and concluded "that to first order, decay acceleration during the Flood must be pulse-like."¹¹ He also asked for exactly the measurement that settles the question, two years before it was made.¹¹ A uniform speed-up has to be refuted from outside the method, by counting clocks and by the heat budget, and that argument lives in other articles.
4. What does bite is a measurement, and a young-earth author called for it before it existed. Hu et al. 2025 measured δ²³⁴U on four carbonates from three continents that sit at secular equilibrium. Every value falls between −0.05 and +0.09 ‰.¹² Over 4,500 years only 1.26 per cent of any disequilibrium relaxes away, so a Flood-era offset would still be sitting there. That pins the ratio of the ²³⁸U and ²³⁴U decay constants to better than 0.015 per cent. A barrier-shift mechanism big enough to speed ²³⁸U by 10⁶ would drive them apart by a factor of 9.53, and would drive ²³⁰Th and ²³¹Pa apart by 2.85. The RATE project's own lever, alpha decay against beta decay, has nothing to grip on here, because every rate-limiting step in both chains is an alpha decay.¹³
5. The young-earth literature has barely engaged this method, and the size of the gap is now countable rather than asserted. A full-text sweep of all 457 International Conference on Creationism papers finds six that mention speleothems and five that mention ²³⁰Th or the uranium series, against controls of 37 papers for radiocarbon and 27 for zircon. Only one of the five, Mogk 2023, does anything with the chain. A 6,741-item harvest of creation.com, with Journal of Creation volumes 1 to 40 complete, and an enumeration of 1,403 Creation Research Society Quarterly articles, add a karst geologist's objection about recrystallisation, a three-part series about growth rates rather than dating, and one sentence calling the radiocarbon calibration circular. The boundaries of that search are stated below, and they matter: only the ICC corpus was read at body-text level, and ICR's own search endpoint has broken recall, so nothing about ICR can be established as a negative.

The verdict.  U-series dating of clean carbonate is one of the best-behaved radiometric methods there is.  Its self-consistency does not prove an old earth, and this article will not pretend it does.  What it proves is narrower and harder: whatever happened to nuclear decay, it happened to ²³⁸U, ²³⁴U, ²³⁰Th, ²³⁵U and ²³¹Pa in lockstep to a few parts in ten thousand.  No published young-earth mechanism can deliver that, because alpha decay is barrier tunnelling and barriers do not scale uniformly.

Every source this rests on is catalogued, with what it measured and how it was obtained, in the literature.

How the clock works, stated honestly

Uranium dissolves in seawater and in groundwater.  Thorium does not.  Thorium hydrolyses and adsorbs onto particles within days, so it never travels far in solution.  When coral aragonite or cave calcite crystallises, it takes uranium in at a few parts per million and takes essentially no thorium.  The daughter starts at zero.

That is the whole basis of the method, and it is worth pausing on, because it is different in kind from the assumption a potassium-argon age makes.  Nobody has to assume the initial daughter was zero.  Aqueous geochemistry puts it there.

Two chains run at once in the same crystal.

238U --a--> 234Th --b--> 234Pa --b--> 234U --a--> 230Th --a--> 226Ra --> ... --> 206Pb
     4.468 Gyr    24.1 d      1.17 min    245,620 yr   75,584 yr    1,600 yr

235U --a--> 231Th --b--> 231Pa --a--> 227Ac --> ... --> 207Pb
     704 Myr      25.5 h      32,760 yr

Note the structure now, because the last section of this article turns on it.  Every rate-limiting step in both chains is an alpha decay.  The beta decays are ²³⁴Th at 24.1 days, ²³⁴Pa at 1.17 minutes and ²³¹Th at 25.5 hours.  None of them controls anything.  Change the beta rate by a factor of a thousand and the chain does not notice.

The half-lives that set the clock:

Those two 2σ figures look loose against the ages people quote.  They are not, because the systematic part cancels when a ²³⁰Th age is computed from a measured ratio.  Cheng et al. give the effective dating uncertainties as ±70 years on the ²³⁴U half-life and ±30 years on the ²³⁰Th half-life, and the resulting age precision under ideal conditions as "±10 a at 10 ka, ±100 a at 130 ka, ±300 a at 200 ka, ±1 ka at 300 ka, ±2 ka at 400 ka, ±6 ka at 500 ka, and ±12 ka at 600 ka."¹⁵  A 2025 refit tightens the ²³⁴U half-life to 245,670 ± 260 years and the ages to ±3 kyr at 600 ka.¹²

There is a circularity worth naming before anyone else names it.  The half-lives were measured on calcite assumed to be at secular equilibrium.  Cheng et al. report the measured atomic ratios ²³⁴U/²³⁸U = 54.970 (±0.019) × 10⁻⁶ and ²³⁰Th/²³⁸U = 16.916 (±0.018) × 10⁻⁶ and note they are "identical within errors."¹⁵  That the two ratios come out equal is a consistency check, but the equilibrium assumption is still an assumption.  It gets tested later in this article, by an independent route, and it survives.

Hu et al. put the same worry more sharply than any critic has: "geological samples are frequently disturbed by environmental and climatic influences that prevent them from reaching and maintaining true secular equilibrium… Many laboratories now rely upon internally characterized standards, often assumed but not conclusively demonstrated to be in secular equilibrium.  This practice can inadvertently introduce systematic isotopic offsets, potentially biasing U-Th age determinations."¹²

The four things that can go wrong

One: the system is not closed.  Uranium or thorium moves in or out after crystallisation.  This happens often enough in corals that there is a published correction industry built around it.  Thompson and Goldstein, in Science in 2005, wrote that suborbital sea-level change "has been difficult to resolve, primarily because of problems with uranium/thorium coral dating," and offered a model that "corrects coral ages for the frequently observed open-system behavior of uranium-series nuclides."⁷  "Frequently observed" is their phrase, not a critic's.

Two: the initial ²³⁴U/²³⁸U is not what it is assumed to be.  Seawater is not at equilibrium.  It runs about 14.6 per cent rich in ²³⁴U, written δ²³⁴U ≈ +146 ‰, because alpha recoil knocks ²³⁴U preferentially off continental mineral surfaces.  A coral inherits that value, and the method assumes it has not changed.  But this is not merely an assumption.  It is a number that gets measured on every sample, and it is the method's own most sensitive alarm.  The Barbados group rejects anything outside a window: "the δ²³⁴U_initial of coral samples must be between 138 and 150 per mil."⁴  Bard et al. found that for every deglacial Barbados sample the initial ratios "fall in the range of modern sea water (1.14-1.15)," and concluded that "their U-Th systematics have not been strongly altered after deposition."⁵

Three: there was initial ²³⁰Th after all.  Detrital clay carries thorium into a growing crystal.  The monitor is ²³²Th, which has no source in either uranium chain on these timescales, so its abundance measures dirt.  A high ²³⁰Th/²³²Th means the thorium is homegrown; a low one means contamination.  Bard et al.: "the ²³²Th/²³⁰Th ratios are generally lower than 30 and therefore no correction for detrital ²³⁰Th is required because the typical crustal ²³²Th/²³⁰Th values are of the order of 10⁵."⁵  Four orders of magnitude of headroom.

Four: the mineral recrystallised.  Aragonite converting to calcite, or dissolving and reprecipitating in fresh water, moves uranium around without leaving the sample.  The screen is X-ray diffraction: samples must show "no detectable calcite based on a documented 0.2% calcite or better detection limit."⁴

Each of these has failed on real samples.  The next three sections are about what happens when you test whether they failed on a given one.

The internal checks, with the numbers

Two decay chains in one crystal

This is the strongest structural argument the method has, and popular treatments on both sides skip it.

The ²³⁰Th clock and the ²³¹Pa clock share nothing except the sample.

Different parent.  Different daughter.  Different element, which means a different chemical separation on a different column with a different spike.  Half-lives differing by a factor of 2.3.  If a coral had gained or lost uranium, or if the decay constants had ever moved by different amounts, there is no reason for the two ages to land on the same number.

They land on the same number.  Edwards et al. 1997 in Science: "Barbados corals that have identical ²³¹Pa and thorium-230 (²³⁰Th) ages indicate that the timing of sea level change over parts of the last glacial cycle is consistent with the predictions of the Astronomical Theory.  Two Devils Hole calcite subsamples record identical ²³¹Pa and ²³⁰Th ages, suggesting that the chronology of this climate record is accurate."¹⁶

A limitation, stated up front.  The sample-by-sample agreement in years and per cent is in that paper's tables, and the full text could not be obtained for this article.  Only the publisher abstract was.  So "identical" is the authors' word, and the residuals behind it are not in hand.

What is documented from bytes is the tolerance the Barbados group actually applies to hundreds of samples, which is arguably better evidence than one paper's tables.  Concordance is not a headline there.  It is an entry requirement.  Their published ranking puts a coral in Category III if it has "been exposed to vadose (rainwater percolating zone) fresh water and [has] ²³¹Pa/²³⁵U and ²³⁰Th/²³⁴U/²³⁸U ages that agree within 2σ analytical error."  A coral that reached the fresh water table and is "not validated by concordant ²³⁰Th/²³⁴U/²³⁸U and ²³¹Pa/²³⁵U dates" is Category VI and is thrown away.⁴  The stated purpose is blunt: corals older than 30,000 years that were exposed to fresh water get a ²³¹Pa age "in order to validate the ²³⁰Th/²³⁴U/²³⁸U age determinations."¹⁷

How tight is a 2σ agreement?  On the thorium side, tight.  Fairbanks et al. state ²³⁰Th ages are measured "with a precision of better than ±3.0 per mil (one sigma) over the entire calibration," and give an average age uncertainty "of about 0.5% (2 RSD)."^{17 17}  At 30,000 years that is roughly ±150 years.  The ²³¹Pa error is larger and is not stated in the papers obtained, so the honest summary is that routine Pa/Th concordance constrains the two chains against each other at somewhere around the one per cent level, not tighter.  That is still a hard test to pass by accident, and it is passed routinely.

Radiocarbon on the same slab of coral

Bard, Hamelin, Fairbanks and Zindler published the founding comparison in Nature in 1990.  Nineteen Barbados corals, each cut into a 4 mm slab, with the uranium-series sample and the radiocarbon sample drilled as two concentric cores out of the same few millimetres of skeleton.⁴  Same coral, same growth increment, two clocks.

Table 1 of that paper was recovered in full for this article by OCR of a scanned copy, and every row was recomputed.  The script is at research/chain-equilibrium.ts and its output at research/chain-equilibrium.out.txt.  Nine of the nineteen rows:

Δ¹⁴C is the amount by which the atmosphere's radiocarbon content sat above the modern reference level.  It is computed here as [exp(−T₁₄C ÷ 8033) ÷ exp(−T_cal ÷ 8267) − 1] × 1000, using the Libby half-life for the conventional age and the true half-life for the calendar age.

Three things come out of this table.

The recomputation reproduces the authors' own number, so nothing was mis-transcribed.  Bard et al. wrote that the offset "is equivalent to a Δ¹⁴C of ~400-500‰ above the present-day reference level for ages between 18 and 22 kyr BP."⁵  The rows in that window give 445 to 513 ‰.  The table and the prose agree.

The largest gap is 3,910 years, at a ²³⁰Th age of 20,610 years.  The authors round it to "a maximum difference of ~3.5 kyr at ~20 kyr BP."⁵

The sign is wrong for accelerated radiocarbon decay.  Faster ¹⁴C decay leaves less ¹⁴C in an old sample, which makes the radiocarbon age look older.  Every one of the nineteen Barbados corals has a radiocarbon age younger than its uranium-series age.  Nineteen out of nineteen, in the direction that accelerated decay does not predict.

And the anchor at the young end matters more than any of that.  The three youngest samples fall inside the tree-ring calibration, and Bard et al. report the dendrochronologically calibrated ¹⁴C dates as 7,440-7,140, 8,970-8,380 and 9,420-8,970 yr BP against ²³⁰Th ages of 7,460, 8,450 and 9,265.⁵  Tree rings involve no decay at all.  A ring is a year because somebody counted it.  The ²³⁰Th clock agrees with a counting clock, to within a few hundred years at nine thousand.

One more observation from the same paper, about the raw data rather than the interpretation.  Nine duplicate uranium-series analyses all agree within 2σ, and the ages come out in depth order without exception.  Bard et al.: "It is remarkable that there is not even a single stratigraphic inversion in the U-Th data."⁵  Nineteen ages from three cores, in the right order, with no help from any model.

Counting clocks

These numbers are already verified in this project's memory file agents/memory/decay-free-clocks.md and were not re-derived here.  The provenance matters, so it is stated.

• Greenland ice layers. Svensson et al. built GICC05 by counting annual layers in Greenland ice back 60,000 years. Compared against the Hulu Cave speleothem chronology, which is ²³⁰Th-dated: "GICC05 agrees with the Hulu Cave time scale within 800 yrs throughout the 60 ka period."⁶ The Austrian Kleegruben Cave record matches to under 300 years. One cave does not fit, Moomi Cave on Socotra, which is off by more than a thousand years on adjacent interstadials. That exception is part of the record and belongs in it.
• Wang et al. 2001, the paper the Hulu chronology comes from: "Between 11,000 and 30,000 yr. B.P., the timing of changes in the monsoon, as established with ²³⁰Th dates, generally agrees with the timing of temperature changes from the Greenland Ice Sheet Project Two (GISP2) core, which supports GISP2's chronology in this interval."¹⁸
• Lake Suigetsu varves. Counted layers in a Japanese lake against the ²³⁰Th-anchored calendar scale, computed in house over 593 dated levels: residual −32 yr at 10-15 ka, +1,409 at 20-25 ka, +2,655 at 30-35 ka, +2,744 at the deepest level. The varve count runs short by 5.4 to 7.5 per cent. This is worth dwelling on because it runs the wrong way for a young-earth reading. If the ²³⁰Th clock were fast, a true annual count would come out shorter than the radiometric age, which is what happens here, but the shortfall is 7 per cent and a young-earth chronology needs something like 700 per cent. And the honest caveat is that Suigetsu is half model: roughly half its years are interpolated by algorithm because the seasonal layers did not form or did not survive, and its authors concede the record "on its own would not be useful without the age constraints provided by the speleothem and coral data."
• Radiocarbon offset between two continents. Japanese lake leaves and Chinese cave calcite, 1,600 km apart, both carrying ¹⁴C and both dated independently, differ by 12 ± 17 years, "consistent with a zero systematic offset."

Devils Hole, where the disagreement is worth more than the agreement

Devils Hole is not a stalagmite.  It is vein calcite that grew under water, lining a fault-controlled fissure below the water table in the Nevada desert.  It is dense, slow-growing and clean.  Core DH-11 is 36 centimetres long and covers half a million years.

Winograd and colleagues published its oxygen-isotope record in Science in 1992, and the chronology was built on "21 replicated mass-spectrometric uranium-series dates."  Then they used it to attack the reigning theory of the ice ages:

"Between the middle and latest Pleistocene, the duration of the last four glacial cycles recorded in the calcite increased from 80,000 to 130,000 years; this variation suggests that major climate changes were aperiodic.  The timing of specific climatic events indicates that orbitally controlled variations in solar insolation were not a major factor in triggering deglaciations.  Interglacial climates lasted about 20,000 years.  Collectively, these observations are inconsistent with the Milankovitch hypothesis for the origin of the Pleistocene glacial cycles but they are consistent with the thesis that these cycles originated from internal nonlinear feedbacks within the atmosphere-ice sheet-ocean system."¹⁹

This belongs in a young-earth-friendly article, and not for the obvious reason.  For twenty-four years a set of ²³⁰Th dates was published, defended, and cited as a refutation of the standard theory of ice-age timing.  The field's response was to doubt Milankovitch, not to doubt the dates.  A dating method that gets pointed at the orbital theory and used as a weapon against it is not a method being quietly tuned to produce expected answers.  That charge is the commonest one levelled at radiometric dating, and Devils Hole is a documented counterexample.

The resolution came in 2016, from a second core.  Moseley et al.:

"previous versions of the iconic Devils Hole (Nevada) subaqueous calcite record exhibit shifts to interglacial values ~10,000 years before orbitally forced ice age terminations, and interglacial durations ~10,000 years longer than other estimates.  Our measurements from Devils Hole 2 replicate virtually all aspects of the past 204,000 years of earlier records, except for the timing during terminations, and they lower the age of the record near Termination II by ~8000 years, removing both ~10,000-year anomalies."²⁰

Read that carefully, because it cuts both ways.  The isotope record replicated.  The ages near the terminations moved by about 8,000 years out of about 130,000, which is 6 per cent.  That is the honest size of an error that survived twenty-four years of scrutiny at the best-behaved uranium-series site in the world, and it is roughly eighty times the ideal analytical uncertainty at that age.  Anybody who tells you U-series ages are good to a tenth of a per cent in the field is quoting a laboratory number.

And it is still disputed.  Tyler Coplen, a co-author of the 1992 paper and a working uranium-series geochemist at the USGS, refused to concede, in Science:

"Moseley et al's… preferred-Termination-II age is subjective, as evidenced by variation in their Termination-II ages of 2500 years per meter.  Termination-II-age bias decreases to zero at ~1.5 meters below the present-day water table, if one assumes linear variation with core-sample height.  Maintaining the required gradient of thorium isotope ²³⁰Th over 3.6 meters for 1000 years, much less 10,000 years, seems exceedingly unlikely."²

A published set of ²³⁰Th ages varying systematically with where in the core you sampled, by 2,500 years per metre, said out loud in Science by someone who does this for a living.  That is a real problem in a real dataset and there is no reason to soften it.

But state the scale of the fight as well as its existence.  The two sides are arguing about 8,000 years inside 130,000.  Neither is arguing about whether the record is 4,500 years old.  A young-earth reader who takes Coplen's sentence as evidence that the method is worthless has borrowed a geochemist's complaint about the third significant figure and pretended it was about the first.

One correction to the research this article is built on.  An earlier working note put a "Devils Hole record extended to 570,000 years, Winograd et al. 2006" in a table.  That citation does not support that number.  Winograd et al. 2006, Quaternary Research 66:202-212, is titled "Devils Hole, Nevada, δ¹⁸O record extended to the mid-Holocene," and its abstract says the record "has been extended from 60,000 to 4500 yr ago."²¹  It extends the record at the young end, not the old one.  The 570 kyr figure has been dropped from this article and moved to quarantine until somebody pins it to a paper.

The correction this article owes the other side

Here is the algebra, and it is not complicated.

The amount of ²³⁰Th in a closed carbonate grows toward equilibrium as

[230Th / 238U] = 1 − exp(−λ t)

with λ the ²³⁰Th decay constant.  Now suppose every decay constant in nature was multiplied by some factor k during some episode.  The ingrowth becomes

[230Th / 238U] = 1 − exp(−k λ t)

A laboratory today measures that ratio and inverts it using the present-day λ.  The age it reports is therefore

apparent age = k × t

exactly.  Not approximately.  The decay constant and the elapsed time appear in the physics only as the product λt, so no measurement made inside the chain can tell a large λ from a long t.

Every internal check in this article is invariant under that operation, and this needs saying plainly.

• Secular equilibrium is invariant. Equilibrium is the state where every step in the chain has the same activity. Scale all the decay constants together and every activity scales together, so the equilibrium condition is untouched. It just arrives sooner.
• The seawater ²³⁴U excess is invariant. The excess exists because alpha recoil supplies ²³⁴U to the ocean faster than ²³⁴U decays. The supply term goes as λ₂₃₈N₂₃₈ and the sink term as λ₂₃₄N₂₃₄. Both scale by k. The steady-state ratio does not move.
• Pa/Th concordance is invariant. Both clocks read k × t. They still agree with each other.

So anyone who cites Pa/Th concordance as a refutation of accelerated decay has not done the algebra.  This is a common move in counter-apologetic writing and it is simply wrong.  Two clocks that were both sped up by the same factor still agree, and their agreement carries no information about whether they were sped up.

One young-earth author has done the neighbouring piece of this algebra, and he published it before this article was written.  It is not in Silvestru, not in Oard and not in Hebert, and an earlier draft of this article said flatly that the ground was empty.  That was wrong.  Nathan Mogk got to the same structure from the other side in 2023, and he is dealt with two subsections below, on the merits and at length.

Two things follow, and they set up the rest of the article.

First, uniform acceleration has to be refuted from outside the method.  The refutations that work are the decay-free counting clocks in the section above, and the heat budget, which asks where 4.5 billion years of decay energy went in a few thousand years.  Both live in other articles.  This one cannot settle them and will not pretend to.

Second, the k = 10⁶ row has a consequence that only one young-earth author has stated, and he stated it for a different purpose.  Mogk 2023 notes that most of the uranium chain has "sufficiently short half-lives to have established secular equilibrium in the time since the Flood."¹¹  The same is true of the whole chain during the acceleration itself.  At k = 10⁶ a single year drives both chains to full secular equilibrium.  A carbonate present during such an episode would come out undatable, not young: the ratio pins at 1.000000 and there is no age to read.  Every coral and speleothem that yields a finite ²³⁰Th age must therefore have crystallised after the episode ended.  Which raises the question the next section is about.

What a young-earth chronology actually needs

Take 4,500 years as a generous post-Flood interval and ask what factor K each published record requires.

This is where the young-earth position is weakest, and there is no way to dress it up.  The RATE project's chapter on accelerated decay proposes "three episodes: one in the early part of the Creation week, another between the Fall and the Flood, and the third during the year of the Genesis Flood."²³  Corals and speleothems that grew after the Flood have no acceleration available to them at all on that model, so their ages are simply unexplained.  Rescuing them requires a post-Flood acceleration of K around 110 to 145 averaged over the whole 4,500 years, and larger still if it ran for only part of that time.  That is a different claim from RATE's and it multiplies the heat problem.

One young-earth author has put a post-Flood acceleration in print, and he applies it to this very method.  Michael Oard, in the Creation Research Society Quarterly in 2021, places accelerated decay "during the Flood and Ice Age."²⁴  He infers it from aluminium-26/beryllium-10 burial ratios and then writes:  "Thus, other Quaternary dating methods can potentially be telescoped into post-Flood time because of accelerated radiometric decay.  For instance, U-series dating on speleothems, which gives dates as old as 500 kyr., can be explained."²⁴  That is the whole of his treatment.  He gives no factor and no history of how the rate changed, and in the same paragraph he names the cost himself:  accelerated decay "presents a problem of too much radiation during the Ice Age."²⁴  Two sentences inside a paper on cosmogenic nuclides are a claim, not an engagement with the method.  They sit in a paper whose title and abstract never mention the uranium series, and the sweep of the Quarterly set out below did not find them.  The table above is what his proposal has to deliver:  factors from 7 to 851 on samples that all formed within the same 4,500 years.  An acceleration strongest early and fading through the Ice Age could in principle give older samples more of it, but nobody has written down that curve.  Whatever form it takes, the δ²³⁴U bound below shows it must hold ²³⁸U and ²³⁴U in lockstep to 0.015 per cent, and the counting clocks above must still agree with the results.

No other young-earth source found for this article both puts the acceleration after the Flood and applies it to this method.  The most technical young-earth paper on uranium-series disequilibrium puts it in the Flood year, and it is the subject of the next subsection.

Nathan Mogk, who arrived at the same structure from the other side

Nathan Mogk read a paper to the Ninth International Conference on Creationism in 2023 called "Tapping the Hourglass: Disequilibrium Relaxation Following Accelerated Nuclear Decay."¹¹  It is the most technically serious young-earth engagement with uranium-series disequilibrium that exists, and this article missed it in an earlier draft.  Three things in it bear directly on everything above.

First, he identifies the same four nuclides, for the same reason.  On page 339:

"The vast majority of radioisotopes in the U/Pb system have sufficiently short half-lives to have established secular equilibrium in the time since the Flood, with four notable exceptions.  In the radium series, ²³⁴U has a half-life of 245,500 years, ²³⁰Th has a half-life of 75,400 years, and ²²⁶Ra has a half-life of 1,600 years.  In the actinium series, ²³¹Pa has a half-life of 32,760 years… It is possible that following the AND epoch, that samples would have an excess of these daughter products, causing the radiometric system to tick at an anomalously high rate until these isotopes were depleted."¹¹

That is the same observation this article's δ²³⁴U section is built on.  A short half-life relaxes away and carries no memory; a long one does not.  Mogk gets there by asking what accelerated decay would leave behind, and this article gets there by asking what a measurement could still see.  The two routes meet on the same four isotopes.

Second, he concludes on the same page "that to first order, decay acceleration during the Flood must be pulse-like."¹¹  His reasoning is an averaging argument, not an isotope argument.  He takes the consensus young-earth correlation in which the Cambrian through the Cretaceous is Flood rock, which is 475 million radiometric years, and divides accumulated radiometric age by Ussher's duration for each period of history.  A shorter interval carrying comparable radiometric age needs a much larger average acceleration, so the Flood year dominates and the acceleration looks like a spike rather than a plateau.

Third, he asked for the measurement that decides it.  In the same paragraph: "Standard radiometric analyses do not include measurements of these intermediate isotopes, and so the measurements are not common.  These additional measurements should be prioritized to better constrain the nature of the underlying mechanism for AND and the amount of decay acceleration experienced."¹¹

That measurement has since been made, and it is Hu et al. 2025.  Nine determinations of δ²³⁴U on four carbonates from three continents, all landing between −0.05 and +0.09 ‰.¹²  On a young-earth chronology 4,500 years erases 1.26 per cent of any ²³⁴U/²³⁸U offset, so whatever a pulse left in those rocks would still be there.  Nothing is there.  Mogk proposed the test in 2023 and the test came back null in 2025.  He deserves the credit for calling for it, and the result deserves to be reported as the answer to his question rather than as an argument against him.

Now the merits.  A pulse confined to the Flood year has two problems on carbonates, and one number in the paper needs reading for what it is.

A pulse in the Flood year makes post-Flood carbonates harder to explain, not easier.  Coral aragonite and cave calcite that crystallised after the Flood were never inside the pulse.  Their ²³⁰Th ingrowth ran at ordinary rates for at most about 4,500 years, so they should read at most about 4,500 years.  Bard's Barbados corals span 30,470.  Devils Hole DH-11 spans 500,000.  Confining the acceleration to the Flood year removes the only escape the previous table left open, which was a sustained post-Flood K of 110 to 145.

Mogk's own alternative, inherited excess daughter, cannot work in a carbonate.  His mechanism for anomalously old post-Flood dates is reservoir relaxation: a magma chamber holds an excess of intermediate daughters left over from the acceleration, and rocks drawn from it inherit that excess and date old.  A coral does not draw from a magma chamber.  It draws from seawater, and a speleothem draws from cave drip water, and neither fluid can hold a thorium or protactinium excess for any length of time.  Both elements hydrolyse and adsorb onto particles within days.  That is the same chemistry that puts the daughter at zero and makes the method work at all, and it means the reservoir this mechanism needs does not exist for carbonates.  It also has an independent check: inherited ²³⁰Th arrives with detrital ²³²Th, and Bard's corals sit below ²³²Th/²³⁰Th = 30 against a crustal value near 10⁵.⁵

Mogk's figure for the post-Flood period is an average, not a proposal, and the carbonates rule it out as a rate.  He allots "66 million years for the modern period" and divides it by Ussher's duration for that period, as one of three time-averaged accelerations.¹¹  Ussher puts the Flood at 2349 BC, so the modern period is about 4,370 years and the average implied by his two inputs is about 1.5 × 10⁴.  Mogk does not argue that decay ran that fast after the Flood.  He uses the averages to conclude that the acceleration was a pulse in the Flood,¹¹ and for old ages from after the Flood he offers relaxation, not continued acceleration:  his mechanism "obviates the need for a gradual ramp-down of AND" (accelerated nuclear decay) while still reproducing "the anomalously old measured ages from the immediate post-Flood era."  He leaves post-Flood acceleration open only by citing Oard:  "There may still be evidence of post-Flood AND in the ratios of cosmogenic radionuclides which are commonly used for Quaternary dating (Oard, 2021)."¹¹  So the 1.5 × 10⁴ is a number no model claims, and it is used here only for scale.  Spread evenly over the time since the Flood, it would give a coral that grew 4,000 years ago 6 × 10⁷ years of ²³⁰Th ingrowth.  That coral would be saturated at [²³⁰Th/²³⁸U] = 1.000000 and would have no readable age at all.  Real corals and speleothems are not saturated.  They yield finite ages, in stratigraphic order, with no inversions in nineteen Barbados samples from three cores.⁵  So whatever accounts for the 66 million radiometric years Mogk assigns to the modern period, it cannot be an acceleration spread evenly through it, which agrees with his own choice of a Flood pulse.

What survives of his paper, and it is not nothing.  The pulse structure is a real prediction and it is testable, which is more than most of this literature offers.  His reservoir-relaxation mechanism is aimed at magmatic systems, where an excess-daughter reservoir is physically plausible, and this article does not touch that case.  He never applies any of it to Holocene corals or speleothems, so the gap the table above identifies is one he has not closed and did not set out to close.

What does bite: a measurement, not an argument

Uniform acceleration is invisible from inside the chain.  Non-uniform acceleration is not, and this is where the argument turns.

Why acceleration cannot be uniform

Alpha decay is quantum tunnelling through a Coulomb barrier.  The nucleus is not falling apart on a schedule; an alpha particle is rattling against a wall it does not have the energy to climb, and occasionally it gets through.  The probability of getting through depends exponentially on the height and width of that wall, which is why half-lives inside a single decay chain span twenty-four orders of magnitude while the decay energy Q spans only a factor of two.

That exponential sensitivity is the whole problem for a differential-acceleration model.  Any mechanism that changes the barrier changes every nuclide by a different amount, because every nuclide sits at a different Q.

The standard parameterisation is the Viola-Seaborg formula,

log10 T(seconds) = (a Z + b) / sqrt(Q) + (c Z + d)

with Sobiczewski's constants a = 1.66175, b = −8.5166, c = −0.20228, d = −33.9069.  Check it against measurement first, because a formula nobody has tested is worth nothing:

The four even-even emitters are reproduced to about 0.2 in the logarithm across five orders of magnitude of half-life.  ²³⁵U is odd-N and ²³¹Pa is odd-Z, and both carry the usual hindrance offsets.  Hindrance shifts the intercept, not the sensitivity to Q, so it cancels out of the ratios below.

Now take RATE's own headline figure, a speed-up of ²³⁸U by a factor of 10⁶, treat any barrier-changing mechanism as an effective shift ΔQ, solve for the shift that produces it, and apply that same shift to the other emitters.  The answer is ΔQ = +0.840 MeV, and it gives:

Two ratios matter.

• ²³⁰Th against ²³¹Pa: a factor of 2.85. Because apparent age is linear in the acceleration factor, the two clocks would read ages differing by that factor. A coral whose ²³⁰Th age is 20,000 years would give a ²³¹Pa age near 7,000. Edwards et al. report Barbados corals and Devils Hole calcite with identical ²³¹Pa and ²³⁰Th ages, and the Barbados group makes 2σ agreement a routine acceptance criterion applied to hundreds of samples.^{16 4} A factor of 2.85 is not a 2σ disagreement. It is a factor of 2.85.
• ²³⁸U against ²³⁴U: a factor of 9.53. This one has a sharper test, and it is the best thing in this article.

The δ²³⁴U bound

Suppose λ₂₃₈ was scaled by k₂₃₈ and λ₂₃₄ by k₂₃₄ during some episode.  While the episode ran, the parent-daughter steady state shifted, so immediately afterwards the measured activity ratio would be

[234U / 238U] = k238 / k234

which in delta notation is δ²³⁴U₀ = 1000 × (k₂₃₈/k₂₃₄ − 1).  Once decay rates returned to normal, that offset would relax back toward zero as exp(−λ₂₃₄ t), governed by the ²³⁴U half-life alone.

And here is the part that makes the bound tight, which is that it uses the young-earth timescale as an input.  The e-folding time for that relaxation is 1/λ₂₃₄ = 354,000 years.  Over 4,500 years, exp(−λ₂₃₄ t) = 0.98738.  Only 1.26 per cent of any disequilibrium would have decayed away.  Whatever offset a Flood-era differential acceleration created is still sitting in the rock, essentially undiminished.  A long-age model has 4.5 billion years to erase the evidence.  A young-earth model has 4,500 years, and 4,500 years erases nothing.

So: go and measure δ²³⁴U on carbonates that should be at secular equilibrium.

Hu et al. did exactly that in 2025, for an entirely unrelated reason, and got the answer to a precision nobody had before.  They measured δ²³⁴U on four carbonates from three continents: three Nullarbor Plain stalagmites from Western Australia, U-Pb dated to 3.83 ± 0.11, 3.45 ± 0.18 and 3.76 ± 0.13 Ma, and a hydrothermal calcite vein from the Bohemian Massif in the Czech Republic, U-Pb dated to 266.20 ± 0.62 Ma.¹²  Nine measurements in all.  Every one falls between −0.05 and +0.09 ‰, with 2σ uncertainties of 0.09 to 0.12 ‰.¹²  Their conclusion: "all four samples yielded ²³⁴U/²³⁸U activity ratios indistinguishable from unity (i.e., δ²³⁴U = 0)."¹²

Take |δ²³⁴U| < 0.15 ‰ at 2σ, which covers every measurement with room to spare.  Divide by the relaxation factor:

|k238 / k234 − 1|  <  0.00015 / 0.98738  =  1.52e-4

The ²³⁸U and ²³⁴U decay constants must have been scaled identically to better than 0.015 per cent.

The Viola-Seaborg calculation above puts them a factor of 9.53 apart.  That is not a near miss.  It is off by more than four orders of magnitude.

The arithmetic is in research/chain-equilibrium.ts and the captured output in research/chain-equilibrium.out.txt, both in this article's folder.  Inputs are the published half-lives and Q values; nothing is fitted.  Anyone who thinks the Sobiczewski constants are the wrong ones, or that ΔQ is the wrong way to parameterise a barrier change, can change the numbers and re-run it.  The conclusion does not hinge on the exact constants.  It hinges on the fact that alpha decay is exponentially sensitive to the barrier, which no mechanism can avoid.

The strongest counter to that bound, which comes from inside the young-earth literature

Somebody will say that ²³⁴U moves.  They will be right, and the person who documented it best for a young-earth readership is Andrew Snelling.

All three parts of his Answers Research Journal series on the U-Pb method open with a list of assumptions the method needs, and that list includes "Deviations of the ²³⁴U/²³⁸U ratio from secular equilibrium."²⁵  Part 2 then develops it.  Working through Davis and Krogh's 2000 leaching experiments on Archean zircon, Snelling reports that residues "showed evidence of preferential leaching of radiogenic Pb isotopes and ²³⁴U from damaged crystal lattice sites," and that this finding was corroborated by "the observation of significant ²³⁴U anomalies in groundwater and the oceans."²⁵

Stated at full strength, the objection runs like this.  The bound above assumes that a measured δ²³⁴U of zero means the two decay constants were scaled identically.  But ²³⁴U sits in a lattice site wrecked by the alpha recoil that made it, so it leaches preferentially.  Groundwater and seawater carry measurable ²³⁴U anomalies for exactly that reason, which nobody disputes.  So a δ²³⁴U of zero could be an artefact of chemistry, and a real differential in the decay constants could have been erased, or faked, by uranium moving in and out of the rock.

The objection is real about uranium and irrelevant to decay constants, and the difference is the whole point.

Uranium mobility changes how many atoms of each isotope are in the rock.  It cannot change the ratio of two decay constants.  Those are properties of two nuclei, and no amount of leaching touches them.  What mobility can do is destroy the record that would otherwise show a differential, and that is precisely why the four Hu samples are the right test rather than the wrong one.

Consider what open-system behaviour would have to do to produce this result.  Secular equilibrium is not a value that mobility drifts toward.  It is a single point, [²³⁴U/²³⁸U] = 1.000000, and alpha-recoil leaching is directional: it preferentially removes ²³⁴U, which drives δ²³⁴U negative in the rock and positive in the water leaving it.  That is the mechanism Snelling describes and it is why groundwater is ²³⁴U-rich.  A rock that has been open to it does not land on zero; it lands somewhere else, and the further it is from zero the more it has moved.  Every one of the nine Hu measurements lands on zero to within 0.12 ‰.¹²  Four samples, on three continents, in two entirely different settings — arid-zone stalagmites under the Nullarbor Plain and a hydrothermal calcite vein in the Bohemian Massif — do not all arrive at the same single point by nine independent accidents of leaching.

The equilibrium is also predicted before it is measured, from a different daughter.  The three Nullarbor stalagmites are U-Pb dated at 3.83, 3.76 and 3.45 Ma and the Czech vein at 266.20 Ma.¹²  The Bateman solution says a carbonate older than about 1.5 Myr should sit within a few hundredths of a per mil of equilibrium.  It does.  A sample that had lost or gained ²³⁴U would fail that prediction, and these do not.

What the objection does establish, and it is worth conceding, is the reason the screening exists.  Snelling's ²³⁴U mobility is not an exotic claim; it is the standard justification for the δ²³⁴U_initial window of 138 to 150 ‰ that the Barbados group applies to every coral,⁴ and for the fact that this article's own section on where the method fails is as long as it is.  The correct summary is that the young-earth literature has published the mechanism that makes U-series practitioners screen their samples, and has not published anything that turns that mechanism against a decay-constant ratio, because it does not reach one.

Why RATE's own lever does not reach

RATE's stated evidence for differential acceleration is a pattern across dating methods.  From the introduction to volume 1, page 11:

"Although the discordant isochron 'ages' are the normal pattern, the discordance seems to differ in a predictable fashion.  Alpha-decay methods generally seem to give older 'ages' than β-decay methods."¹³

That is a claim about alpha versus beta.  It is a real observation and it is the strongest empirical thing RATE has.

It cannot be applied here, because both of these chains are all-alpha in their rate-limiting steps.  Look back at the chain diagram.  The beta decays are ²³⁴Th at 24.1 days, ²³⁴Pa at 1.17 minutes and ²³¹Th at 25.5 hours.  Not one of them is anywhere near controlling the ingrowth of ²³⁰Th or ²³¹Pa.  You could turn the beta rates up or down by orders of magnitude and neither clock would move.

So on this system, the alpha-beta lever has nothing to grip on.  Any differential acceleration must be within alpha decay, which is the harder claim, and it is the claim the δ²³⁴U bound excludes at 1.5 × 10⁻⁴ and routine Pa/Th concordance excludes at roughly the per-cent level.

Where the method genuinely fails

None of the above means uranium-series ages are always right.  They are not, and the best documentation of that comes from the people who make them.

Open-system behaviour in corals is common and admitted.  Thompson and Goldstein's Science paper exists because coral ages were scattered enough to hide a real sea-level signal.  Their words: "primarily because of problems with uranium/thorium coral dating."⁷  Their correction model is a three-parameter fit that assumes a particular mechanism of uranium gain and loss,²⁶ which means a corrected age is a model output, not a measurement, and inherits every assumption in the model.  That distinction should be kept sharp in both directions.

IntCal20 threw out every coral older than 25,000 calendar years.  This is the most damaging admission about coral uranium-series work in the mainstream literature, and it is the calibration community's own.  Reimer et al. 2020, page 739:

"some of the coral data used in IntCal13 exhibit wide scatter in Δ¹⁴C (~500 ‰), especially around 37 kyr BP (Figure 2).  It is likely that diagenesis due to exposure to freshwater occurred when these corals were above sea level during the lowstand of the Last Glacial Maximum (LGM; 21 ± 2 cal kBP).  This effect appears to have been particularly prevalent in corals from areas with rapid uplift such as Vanuatu for which coral Δ¹⁴C appears very elevated (>1100‰ around 27-29 cal kBP; Cutler et al. 2004)… It was therefore decided to remove all coral data older than 25 cal kBP for construction of the IntCal20 curves to avoid the potential inclusion of erroneous data.  Further work is needed to develop a way of identifying which corals are affected by diagenesis."⁸

Both halves of that have to be reported, and a young-earth writer who quotes only the first will be caught inside a day.  What it concedes is that pre-25,000-year coral data are unreliable enough to discard wholesale.  What it does not concede is that the ²³⁰Th ages are wrong.  The stated cause is freshwater diagenesis altering the carbon, and IntCal20 replaces the discarded corals with speleothems whose ²³⁰Th ages it trusts and uses.  Reporting the first sentence without the second is a misrepresentation, and this article will not make it.

The screening protocol excludes most published data, and the mainstream said so in an argument with itself.  Reimer et al. 2006, quoted verbatim by Fairbanks et al. in their reply:

"Fairbanks et al. (2005) and Chiu et al. (2005) insist that a protocol of 'no detectable calcite' be adopted as the new standard.  This may well be good advice, but doing so would exclude nearly all of the (published) coral data presently available."⁴

Fairbanks answers by applying his criteria to a rival calibration dataset and showing that nearly all of it fails.  That exchange is worth more to a skeptical reader than any critic's summary, because both parties are inside the field and both are on the record.

Speleothem radiocarbon starts out old.  Cave drip water dissolves limestone with no ¹⁴C left in it, so the calcite is born apparently ancient.  IntCal20's posterior dead-carbon fractions: 472, 470 and 481 ± 50 ¹⁴C years for three Hulu Cave stalagmites, which is three stalagmites in one cave agreeing to eleven radiocarbon years, but 1,523 ± 208 and 2,891 ± 323 for a Bahamas speleothem, which is why that record was downweighted.⁸  This affects the ¹⁴C read off a speleothem.  It does not affect the ²³⁰Th age of the same calcite, and conflating the two is a mistake both sides make.

The worst documented failure of uranium-series concordance is Lake Mungo 3, and a skeptic should be handed it before he finds it.  Thorne et al. 1999 dated the Mungo 3 skeleton with combined ESR and uranium-series at 62,000 ± 6,000 years, "in close agreement with OSL age estimates on the sediment into which the skeleton was buried of 61,000 ± 2,000 years."⁹  Bowler et al. 2003 then got 40 ± 2 kyr from 25 optical ages.¹⁰  Four apparently independent methods agreed to within 2 per cent and were wrong by 55 per cent.

Why did concordance fail there and hold on corals?  Because ESR, Th/U and Pa/U were all reading the same uranium in the same bone, and bone is the textbook open system.  All three share one uranium-uptake assumption, so they are not independent at all.  They agreed because they were wrong together.  That is exactly the reason the coral screening criteria in the section above exist, and it is why an unscreened concordance means much less than a screened one.  It is also a warning against the general move of treating agreement as proof, which is a move this article has already declined to make about accelerated decay.

The honest error bars.  Ideal laboratory conditions on modern instruments give ±10 yr at 10 ka and ±100 yr at 130 ka.¹⁵  Real-world dispersion is very much larger.  The Devils Hole Termination II revision was 8,000 years at 130,000, about eighty times the ideal figure.

And one number nobody publishes.  What fraction of analysed corals and speleothems fail the δ²³⁴U window, the X-ray diffraction screen and the Pa/Th check?  No source found for this article reports it.  That number would sharpen this debate in whichever direction it fell, and its absence is a genuine weakness in the literature.  A published concordance rate among samples pre-selected for concordance is not the same thing as a concordance rate.

The young-earth literature, and how little of it there is

One young-earth author has engaged uranium-series disequilibrium as a method, and he is Nathan Mogk.  Nobody else found in these searches has.  That is a strong claim about an absence, so the corpora and the queries are set out here, run between 25 and 30 August 2026, and a reader can check them rather than take them.

State the boundaries of that, because they are the difference between a countable finding and a hopeful one.  Only the ICC corpus was read at body-text level;  the creation.com, ARJ and CRSQ sweeps were run over titles, abstracts and site indexes, with full text pulled only for hits.  ICR is excluded from any negative at all.  And two books were never obtained:  if Snelling's Earth's Catastrophic Past or Woodmorappe's The Mythology of Modern Dating Methods contains a real critique of this method, this section is wrong and should be rewritten.

Within those limits, nothing found outside Mogk 2023 engages the ²³¹Pa/²³⁰Th cross-check, the detrital ²³²Th correction, or the Cheng et al. half-life determination.  One source engages the δ²³⁴U_initial screening criterion sideways:  Snelling's ARJ work on ²³⁴U mobility, dealt with above.

What does exist divides into five items.  Only the first is about the clock's chemistry.  The last is not an argument at all.

Recrystallisation, which is the best objection anyone on that side has made

Emil Silvestru holds a doctorate in karst sedimentology.  In a 2011 creation.com feedback reply, answering a correspondent who had cited U-Th ages, stable isotopes and laminae counts as mutually corroborating, he named the failure mode that actually matters:

"The ²³⁴U–²³⁰Th method applied to speleothems has yielded so many erroneous ages that we have lost count.  Of course, the accepted dates are those that are in accordance with the standard long-age, evolutionary karstogenesis model.  However, there are many uncontrollable variables, from contamination (never properly countered regardless the method), to recrystallization (which has long been ignored on a microscopic scale, most assessments being made on megascopic grounds)."²⁷

The mechanism is real and the parenthesis is the sharp part.  Recrystallisation moves uranium and thorium around inside a sample without either element leaving it, so a bulk chemical analysis can look clean while the age is wrong.  It is the same failure the coral literature calls diagenesis, and it is why this article's list of things that can go wrong has it as item four.  Silvestru's specific complaint is that screening is done at the scale of a hand specimen or a cut slab, while the resetting happens at the scale of a crystal boundary.  That is a fair description of how a lot of speleothem work is actually screened.

What the field does about it, stated so a reader can weigh the answer against the objection.  Four things, and they are not equally strong.

• X-ray diffraction catches aragonite that has inverted to calcite, which is the commonest and most destructive route. The Barbados protocol requires "no detectable calcite based on a documented 0.2% calcite or better detection limit."⁴ This is a strong screen, and it is why corals are better behaved than speleothems.
• The initial δ²³⁴U value is a direct open-system detector on every sample. Uranium that moved drags the number off the seawater or drip-water value, so a disturbed sample announces itself.⁴
• Petrographic fabric screening is what speleothem workers use where XRD cannot help, because most cave calcite was calcite to begin with and recrystallising calcite to calcite does not change the mineral. Columnar fabric with continuous crystal boundaries is taken as primary; mosaic or blocky fabric with truncated boundaries is taken as altered. This is a judgement made down a microscope, not a measurement, and Silvestru's objection lands hardest here.
• Stratigraphic ordering and replicate dating are the empirical backstop. A recrystallised horizon does not know where it sits in a stalagmite, so disturbance shows up as age inversions with depth and as disagreement between coeval samples. Bard's Barbados corals give nineteen ages from three cores in the right order with no inversions,⁵ and IntCal20's three Hulu Cave stalagmites agree with each other to eleven radiocarbon years on their dead-carbon fractions.⁸

The residual, which this article will not paper over.  Calcite-to-calcite recrystallisation is not caught by XRD, fabric screening is a petrographer's call rather than a number, and nobody publishes what fraction of speleothems are rejected on fabric grounds.  Silvestru is right that this is thinner than the coral screening, and he is right that most assessments are megascopic.  What he does not supply is a case, with a sample and an age, where a screened speleothem was shown to be recrystallised and wrong.  His own paragraph offers the phrase "we have lost count" and no count.

And the same reply contains the conflation the next subsection is about.  Two paragraphs later Silvestru offers, as corroboration for the inaccuracy of radiometric dating, "the case of numerous speleothems grown on show cave facilities, from concrete pavements to hand rails and even cabling.  In other words we see speleothems forming quickly."²⁷  That is an observation about growth rate offered as an argument about dating.  A decade later his own organisation told its readers not to make it that way.

Speleothems can grow fast, which is true and is not an argument about dating

Michael Oard's three-part series in the Journal of Creation is the fullest young-earth treatment of speleothems in print, and it is careful, well-referenced work.  It identifies five controlling variables for growth rate, adds evaporation, and concludes that present-day rates are "around 0.1 to 2 mm/yr, but can vary from zero to more than 5 mm/yr," and that in "the early-to-mid post-Flood Ice Age" growth "could easily be over a hundred times that of today in some caves."³

That claim is probably right, and it is not an argument against the ²³⁰Th clock.  The two get confused constantly, so it is worth being explicit about why they are different.

Growth rate is millimetres per year of calcite.  The ²³⁰Th clock measures an isotope ratio inside a given millimetre of calcite.  A stalagmite that grew a metre in a century contains one century of ²³⁰Th ingrowth, spread thinly over a lot of rock.  The clock reads a hundred years and the stalagmite is a metre tall.  Nothing is in tension.  Fast growth and correct dating are fully compatible, and the argument only becomes a dating argument if the ages are also wrong, for some separate reason.

Oard does not argue that the ages are wrong.  He assumes it, in print, in part 3:

"In searching the literature for growth rates, estimated growth rates that depend upon radiometric methods were avoided, since they always seem to greatly exaggerate the time.  Instead, only observed growth rates were used."³

Quote that against the argument rather than waiting for a critic to find it.  Using the conclusion as a filter on the evidence is the exact failure young-earth writers accuse mainstream geologists of committing, and the credibility lost by hiding it exceeds anything the argument gains.

And a creationist made this same distinction against his own side, on chemistry rather than on logic.  The popular version of the growth-rate argument is the stalactite hanging under a concrete bridge:  it grew in decades, so speleothems grow fast.  Gavin Cox, writing for Creation Ministries International in December 2021, told his readers to stop using it flat.  He opens by warning that "Biblical creationists do need to be careful about how they use evidence, and as we shall see, stalactites growing under modern concrete bridges are not directly analogous to how speleothems grow in natural caverns."²⁸  The reason is a chemical one, and it is worth having exactly:

"Nelson's data corroborate the C&KS conclusion that the calthemites growing from concrete absorbed carbon dioxide from the air, rather than losing it (degassing) as in the case of cave speleothems.  Nelson therefore recognizes that in modern concrete buildings the chemistry is different from that of cave systems for the formation of speleothems.  Therefore, drawing comparisons with modern concrete stalactites to that of cave formations should only be done in a qualified sense."²⁸

A calthemite grows from hyperalkaline concrete leachate at pH above 9, taking CO₂ in from the air.  A cave speleothem grows from near-neutral calcium bicarbonate solution, giving CO₂ off.  Opposite reactions, so the rate of one says nothing about the rate of the other.  Cox does not abandon the comparison — he keeps it as an "object lesson" and he corrects a sceptical site on a point of fact along the way, since the man-made deposits are calcite rather than gypsum — but he qualifies it, in print, in the movement's largest apologetics outlet.

This matters more than the point it settles.  Ten years earlier Silvestru had used the unqualified version, hand rails and cabling and all.²⁷  Cox corrected it in 2021.  The distinction between growth rate and age that this article is pressing is therefore not a distinction invented by a critic and imposed on the young-earth literature.  It is one that literature has already drawn against itself.

Speleothems dated during the Ice Age, which is a real observation with no mechanism

Emil Silvestru holds a doctorate in karst sedimentology, which makes his 2012 Creation article the most technically credentialled young-earth source on caves.  He makes two arguments.

The first is geomorphological, and it is the better one:

"Well, I knew—and all karstologists know—that the surface of limestone terrains above caves changes dramatically in short periods of time.  And any change at the surface also changes the location of the water droplets inside the cave.  However, the stalagmites do not indicate any changes.  So the conclusion is simple: they cannot be that old."¹

That is a real observation and it deserves a real answer.  It is not an argument about ²³⁰Th, and the answer comes from the cave rather than from geochronology.  Cave Formations gives it:  the drop falls from the tip of a stalactite welded to the ceiling, so when water returns along the same fracture after a pause it leaves from the same tip and lands on the same spot.  That answer is reasoning from the stalactite, not a study;  no paper yet follows drip points across long pauses with dated stalagmites before and after.

The second is Vancouver Island.  Speleothems there "have yielded radiometric ages of between 12 and 18 thousand years," but the geological evidence says the island lay under up to two kilometres of ice during the Fraser Glaciation, so they should not have grown at all.  Silvestru:

"But rather than question the radio-isotopic dates (and hence the methodology involved), some scientists have proposed that the 2-km—(6,500-ft—) thick ice cover melted and grew back in a few thousand years, even though there is no evidence for this melting and they cannot explain how it could have happened!  Of course, a simpler explanation is that the radiometric dating is incorrect and that the speleothems grew only after the ice had melted."¹

The anomaly is real.  The proposed resolution is not worked out.  Silvestru concludes that the speleothems "have recorded the post-glacial climate variations over the last 4,000 years,"¹ which requires the ²³⁰Th clock to run 3 to 4.5 times fast over exactly the interval from 4,000 years ago to now.  He proposes no mechanism, and he does not do the arithmetic on his own claim.

Somebody should do that arithmetic, because it is testable and it fails.  If the ²³⁰Th clock ran three times fast over the last few thousand years, Bard's three youngest Barbados corals, which read 7,460, 8,450 and 9,265 ²³⁰Th years, would really be about 2,500, 2,800 and 3,100 years old.  Their radiocarbon ages, calibrated against tree rings, come out at 7,440-7,140, 8,970-8,380 and 9,420-8,970 years.⁵  A ring is a year because somebody counted it.  So Silvestru's required speed-up puts the ²³⁰Th clock at odds with a counting clock by a factor of three, on samples that were measured thirty-five years ago and are on the public record.

And Silvestru and Oard contradict each other, which nobody at Creation Ministries International appears to have noticed.  Silvestru's Ice Age argument requires that speleothems did not grow during the Ice Age.  Oard's argument is that post-Flood Ice Age growth was "over a hundred times that of today in some caves."  These are the two leading young-earth cave specialists at the same organisation, and their models point in opposite directions.

The calibration is circular, which is false and is easy to answer

Jake Hebert, in Answers Research Journal, gives the only direct young-earth engagement with Bard et al. 1990 that these searches found, and it is one sentence:

"it is fairly common for uniformitarian scientists to use one radioisotope dating method to calibrate another radioisotope dating method.  For instance, uranium-thorium ages for corals have been used to calibrate the carbon-14 timescale (Bard et al. 1990).  Of course, the fact that radioisotope dates need to be 'calibrated' or 'synchronized' … is a clear indication that such dates are not absolute, despite popular perception."²⁹

The premise is true.  The inference does not follow, for two reasons.

Radiocarbon needs calibration for a reason that has nothing to do with decay.  Its production rate in the upper atmosphere varies with the geomagnetic field and with the carbon cycle, so the amount of ¹⁴C a living thing starts with is not constant.  That is a source-term problem.  Uranium-series has no source term: a coral's ²³⁰Th starts at zero by chemistry, not at a variable atmospheric level.  The two methods do not share an assumption, so using one to fix the other's initial condition is not circular.

And the calibration is anchored at both ends by clocks that involve no decay at all.  Tree rings at the young end, ice layers and varve counts through the middle.  If the ²³⁰Th clock and the ¹⁴C clock had been calibrated into a mutually consistent fiction, both would have to disagree with the counting clocks, and they do not: 800 years over 60,000 against Greenland ice.⁶

The fifth item, which is a misdescription rather than an argument

The most-read young-earth description of this method is Andrew Snelling's, in Answers magazine in September 2018, and it describes a different method:

"Another method occasionally used is called uranium-series dating.  As uranium decays to lead, it passes through a series of other elements.  Measuring the quantities of those elements enables us to estimate an age based on how far along the process has gone.  For this method to work with human artifacts or bones, however, they need to be entombed in limestone containing uranium, such as that in cave deposits."³⁰

Two things are wrong in three sentences, and neither is a matter of interpretation.

Progress toward lead is not what is measured.  A ²³⁰Th age is read from the disequilibrium among the intermediate daughters, ²³⁴U and ²³⁰Th, on a timescale where essentially no ²⁰⁶Pb has accumulated at all.  Reading progress along the chain toward its stable terminus is U-Pb dating, a different measurement over a different range, and Snelling has written a three-part, 140-page series on it elsewhere.

And the geochemistry is backwards.  The method works because thorium is insoluble and uranium is soluble, so the growing calcite takes in uranium and excludes thorium and the daughter starts at zero.  What matters is not that the limestone "contains uranium" but that the water does and carries no thorium.  Get that the wrong way round and the whole basis of the method disappears.

This is not quoted to score a point off a busy writer.  It is quoted because it is the clearest available measure of how thin the engagement is:  the movement's most prominent geologist, writing in its most-read outlet, in the article that most young-earth readers will meet this method through.

Objections

"The detrital thorium correction is a fudge factor.  You tune it until the answer looks right."

It would be, if the correction were a free parameter.  It is not, because the thing being corrected for is measured on the same aliquot.  Detrital contamination brings ²³²Th along with ²³⁰Th, and ²³²Th has no source in either uranium chain on these timescales, so the measured ²³⁰Th/²³²Th ratio tells you directly how much of the thorium is dirt.  Crustal material sits near ²³²Th/²³⁰Th ≈ 10⁵.  Bard's Barbados corals sit below 30.⁵  That is four orders of magnitude of headroom, which means the correction is not applied at all on those samples, because there is nothing to correct.

Where the correction is applied, on dirty speleothems, the objection has some force, and the honest answer is that the size of the correction should be reported and rarely is in popular summaries.  But note what would happen if the correction were being tuned: the ²³¹Pa age would break, because protactinium and thorium have different chemistry and a tuned thorium correction does not fix a protactinium age.  Concordance is not proof against tuning, but it makes tuning expensive.

"You admit corals are open systems.  So why trust any of them?"

Because "open system" is a property of a particular sample, not of corals in general, and it can be detected on that sample.  The δ²³⁴U_initial window of 138 to 150 ‰ is the detector,⁴ and it is a good one: uranium moving in or out of a coral drags that number off the seawater value, so a sample that gained or lost uranium announces itself.  X-ray diffraction catches recrystallisation at the 0.2 per cent level.⁴  Pa/Th concordance catches what the first two miss.

The fair version of this objection is different and harder, and it is dealt with two objections down.

"The radiocarbon calibration is circular, because it is calibrated against uranium-thorium."

Answered in full above.  Radiocarbon needs calibration because its production rate varies; uranium-series has no production rate to vary.  The two methods share no assumption, and both are anchored against clocks that count rather than decay.

There is a narrower version worth conceding: for the interval beyond 25,000 years, IntCal20 rests substantially on speleothem ²³⁰Th ages, so a systematic error in the ²³⁰Th clock in that window would propagate into the calibration curve.  That is a real dependency and it should be stated.  It is bounded, though, by the Greenland ice-layer comparison, which runs to 60,000 years and involves no decay.⁶

"Concordance between two decay chains proves nothing if both were accelerated together."

This objection is correct, and this article concedes it without qualification.  The algebra is worked in the open above.  Apparent age is exactly k × t for a uniform k, so ²³⁰Th and ²³¹Pa clocks that were both sped up by the same factor still agree with each other, and their agreement says nothing about whether they were sped up.  The same goes for secular equilibrium and for the seawater ²³⁴U excess.

What concordance does constrain is differential acceleration, and the constraint is sharp, because alpha decay is barrier tunnelling and a barrier shift cannot scale nuclides of different Q by the same factor.  The refutation of a uniform speed-up has to come from somewhere else, and this article does not pretend to supply it.

"²³⁴U is mobile.  It is preferentially leached out of recoil-damaged lattice sites, and there are significant ²³⁴U anomalies in groundwater and the oceans.  Your δ²³⁴U bound is measuring chemistry, not decay constants."

The best objection in this list, it comes from Andrew Snelling's own Answers Research Journal work rather than from a critic, and the first two sentences of it are simply true.²⁵  It is answered at length above.  In short:  leaching changes how many atoms are in the rock, not the ratio of two decay constants, and alpha-recoil leaching is directional, so it drives δ²³⁴U away from zero rather than onto it.  Nine measurements on four samples from three continents and two geological settings all land on zero to within 0.12 ‰,¹² at the value their independently determined U-Pb ages predict.¹²  Mobility is why the method screens on initial δ²³⁴U in the first place;  the objection is the screening criterion.

"How do you know those four Hu carbonates were at equilibrium?  You assumed it, and then used the assumption to bound your own parameter."

The sharpest objection in the list, and the answer is that equilibrium was predicted independently before it was measured.

The Bateman solution for a closed carbonate says how long the approach to equilibrium takes.  The rate is set by the slowest daughter, ²³⁴U, whose e-folding time is 354,000 years:

The Nullarbor stalagmites were dated by uranium-lead at 3.83, 3.76 and 3.45 Ma.  That is the same decay chain read at the opposite end — U-Pb counts ²⁰⁶Pb, the stable terminus that ²³⁸U reaches after fourteen steps, while U-series reads the disequilibrium among the intermediates on the way there.  Different daughter, different half-life, different laboratory measurement, and no shared assumption beyond the identity of the parent.¹²  The table says a 3.8 Myr carbonate should sit at 0.999978, or δ²³⁴U within about 0.02 ‰ of zero.  The measurement says +0.06 ± 0.09, +0.06 ± 0.10 and −0.03 ± 0.12 ‰.¹²  The prediction was made from a U-Pb age, and the U-series measurement confirmed it, on four samples from three continents.

Somebody could still argue that the U-Pb ages are themselves wrong.  But they would then have to explain why four independently U-Pb dated carbonates, spanning 3.45 Ma to 266 Ma, all land at exactly the δ²³⁴U their U-Pb ages predict, and none of them lands anywhere else.

"You only publish concordant samples.  Of course concordance is high."

Partly fair, and the fair part is real.  A coral is accepted into the Barbados calibration set partly because its two chains agree.⁴  So the published concordance rate is a rate among samples pre-selected for concordance, and nobody reports the rejection rate.  That number should exist and does not.

The part that is not fair is the implied conclusion.  A discordant sample is not silently discarded; it is put in Category IV, V or VI, and the categories are published with their criteria.  And the screening decision is made on the δ²³⁴U and XRD evidence, which is independent of the age, before concordance is checked.  Still, until somebody publishes what fraction of samples fail, this objection cannot be fully answered, and this article will not pretend otherwise.

"Lake Mungo shows that four concordant methods can all be wrong by 55 per cent."

Correct, and the case is discussed above rather than buried, because it is the strongest thing a skeptic of the method has.  The lesson it teaches is specific: concordance between methods that share an assumption is worth nothing.  ESR, Th/U and Pa/U all read the same uranium in the same bone and all assume the same uranium-uptake history, so they were never independent.

That lesson cuts against naive concordance arguments everywhere, including some made by defenders of radiometric dating.  It does not transfer to screened coral aragonite, where the shared assumption is exactly what the δ²³⁴U window and the XRD screen are designed to test.  The way to break the analogy is to find screened corals whose Pa and Th ages diverge, and this article did not find any in the literature it obtained.

"Devils Hole moved by 8,000 years.  Why should anybody trust a number that can move that much?"

Because 8,000 out of 130,000 is 6 per cent, and the argument this article is grading needs 99.7 per cent.  A method whose worst documented systematic error at a well-studied site is 6 per cent is not a method that can be off by a factor of 29.

The objection does succeed against a different claim, which nobody in this article is making: that field uranium-series ages are good to the laboratory precision.  They are not.  Quote the laboratory number and a hostile reader will produce Devils Hole, and be right to.

"Speleothems can grow a hundred times faster than assumed, so the timescale is wrong."

Growth rate and age are different quantities and the confusion between them is the single commonest error in popular treatments.  Oard's growth-rate work is probably right and does not touch the ²³⁰Th clock, for the reason set out above: a stalagmite that grew fast has the same isotope ratio in it that a stalagmite that grew slowly would have, if the two are the same age.

Where the young-earth case is strongest

Four places, stated without softening.

1. The self-consistency arguments do not touch uniform accelerated decay, and that is a genuine and important concession. Every internal check in this article, equilibrium, Pa/Th concordance, seawater δ²³⁴U, is invariant under a uniform scaling of all decay constants, because every observable depends only on the product λt. Counter-apologetic writing that presents Pa/Th concordance as a refutation of accelerated decay is wrong, and the error is arithmetical rather than interpretive.
2. Open-system behaviour is admitted, common, and the screening is partly circular in a way that deserves a name. A sample is accepted partly because its two chains agree. That is a good test, but it means the published concordance rate is a rate among pre-screened samples, and no source found reports the rejection rate. That number would be genuinely informative and nobody publishes it.
3. Silvestru's recrystallisation objection is right about the screening even though he supplies no case. Calcite-to-calcite recrystallisation in a speleothem is invisible to the X-ray diffraction screen that catches aragonite inversion in a coral, and what replaces it is petrographic fabric judgement rather than a measured threshold.²⁷ Nobody publishes a rejection rate on fabric grounds. Speleothem screening is genuinely thinner than coral screening and this article says so.
4. Silvestru's geomorphological objection has an answer, but not yet a test. If limestone surfaces change on short timescales, and drip points migrate with them, a stalagmite that grew under one drip for 300,000 years is odd. Cave Formations answers that the stalactite fixes the drip point. No study has yet followed drip points across long pauses to test that answer directly.

Where the young-earth case fails

Four places, and none of them can be softened either.

1. RATE's actual claim, differential acceleration by decay mode, is refuted quantitatively by these two chains. Both are all-alpha in their rate-limiting steps, so the alpha-versus-beta lever does not exist here. And the within-alpha differential that any barrier-changing mechanism must produce is excluded at the 1.5 × 10⁻⁴ level by measured δ²³⁴U in equilibrium carbonates, and at roughly the per-cent level by routine Pa/Th concordance. The Viola-Seaborg estimate puts ²³⁸U and ²³⁴U a factor of 9.53 apart.
2. On RATE's own placement of the acceleration episodes, post-Flood coral and speleothem ages have no explanation at all. Devils Hole needs K ≈ 111 and the Sanbao/Hulu composite K ≈ 143, both after the Flood. Mogk 2023 keeps the acceleration in the Flood, as a pulse. The one young-earth author who has put a post-Flood acceleration in print, Oard 2021, says it can explain speleothem ages, but he gives no factor and concedes it brings "too much radiation during the Ice Age."²⁴ Nobody has proposed a number that works.
3. The radiocarbon sign test runs the wrong way. Nineteen out of nineteen Barbados corals have ¹⁴C ages younger than their ²³⁰Th ages, by up to 3,910 years. Accelerated ¹⁴C decay predicts older.
4. The two leading young-earth cave specialists contradict each other on whether speleothems grew explosively during the Ice Age or should not have grown during it at all, and neither has engaged the other.

What this article does not claim

It does not claim that uranium-series concordance proves an old earth.  It does not, and the algebra above shows exactly why.

What the concordance proves is narrower and harder to accommodate.  Whatever happened to nuclear decay rates, it happened to ²³⁸U, ²³⁴U, ²³⁰Th, ²³⁵U and ²³¹Pa in lockstep, to a few parts in ten thousand for the uranium pair and to roughly a per cent for the thorium-protactinium pair.  Alpha decay is exponentially sensitive to the Coulomb barrier, so a barrier-changing mechanism cannot deliver that.  Any young-earth model that wants these ages compressed needs a mechanism that scales five alpha emitters of different Q by identical factors, and no such mechanism has been proposed.

That is a much narrower target than the RATE literature has ever acknowledged, and hitting it is the work that has not been done.

Every source behind this, with what it measured, how it was obtained, and which way it cuts, is in the literature.

Sources