Radiocarbon Dating

"The initial vision that high precision AMS methods should make it possible to extend ¹⁴C dating of organic materials back as far as 90,000 years has not been realized.  The reason seems to be clear — few, if any, organic samples can be found containing so little ¹⁴C!" — John Baumgardner, 2005¹

"Of all of the possible explanations for the reports of the presence of ¹⁴C in these background samples, the least probable explanation has been advanced by those rejecting the validity of the ¹⁴C time scale." — Taylor, Southon & Santos, 2018²

Summary

Radiocarbon dating is not one of the "millions of years" methods.  It cannot reach past about 55,000 years even in principle, and it is the only radiometric method that has been checked, year by year, against clocks that were built without it.  That makes it a different kind of argument from uranium-lead or potassium-argon, and it deserves to be treated differently.

Seven conclusions, stated first.

1. The method itself is sound and unusually well tested. A radiocarbon age is a count of atoms, not an inference from an assumed starting condition. Where the answer is already known — Egyptian dynastic wood, dated tree rings, historically dated timbers — radiocarbon reproduces it, and its residual offsets are measured in decades, not millennia.
2. A raw radiocarbon "age" is not a date, and everybody in the field says so. The atmosphere's ¹⁴C level has moved, so raw ages must be converted through a calibration curve. The calibration curve is where the real argument lives.
3. Radiocarbon has genuine, documented failure modes, and they are worse than the public thinks: marine and hard-water reservoirs of several hundred years, volcanic carbon dilution, "old wood," systematic regional offsets in the eastern Mediterranean of one to three centuries, and roughly one blind inter-laboratory result in ten on old bone coming back a statistical outlier, in one case at 21,684 years where the consensus is 39,305.³ Every one of these is a percentage-level or century-level problem, or a replication problem. None of them is the ninety-percent compression a young-earth timeline needs.
4. The young-earth "carbon-14 in coal and diamonds" argument is real but much narrower than its popularisers claim, and its diamond half is refuted by the young-earth side's own follow-up experiment. In 2006 the RATE team put diamonds straight into the accelerator with no chemistry and measured 0.014 pMC, against 0.168 pMC for the twelve diamonds they had crushed and graphitised. Taking the chemistry away divided the signal by twelve.⁴ Their central statistic — that Phanerozoic biological samples average 0.292 pMC against 0.062 for Precambrian graphite — is answered by two controlled pairs sitting inside RATE's own Table 1, in which re-graphitising the same graphite added 0.25 and 0.14 pMC.⁵ The coal claim partly survives: RATE's coals average 0.324 pMC where a comparable laboratory gets 0.12 to 0.14 pMC on processed coal, a factor of two and a half that nobody has explained — and RATE's sharpest critic says so in print.⁵
5. The cleanest test in the whole dispute is a shape, not a number. If residual ¹⁴C were intrinsic to a sample, the measured ratio could not depend on how much sample you burn. Across 1,140 infinite-age background samples, it does: 0.2 pMC at 300 micrograms, rising to an apparent 20,000 years at 10 micrograms.² A single constant — 0.8 micrograms of added modern carbon — reproduces the whole curve.⁶ Intrinsic ¹⁴C cannot produce that trend at all.
6. The flood-dilution model that is supposed to convert those numbers into a young earth does not close its own books. Worked out for this article from published carbon-reservoir inventories, the model has to pick a pre-Flood carbon pool 500 times today's biosphere. Count only recoverable fossil fuels and you get three to five times, which predicts coal at 12 to 20 pMC — a hundred times more ¹⁴C than is measured. Count all the organic carbon in sedimentary rock, which is what flood geology says was buried, and you get 27,000 times, which predicts 0.002 pMC — a hundred times less than is measured.⁷ The number that works is picked because it works.
7. The hardest single fact for a young earth, in this whole subject, is a tree-ring count. IntCal20 rests on "securely dated tree-ring data" for "the last 12,310 years."⁸ That is roughly twice the age of the earth on a Masoretic biblical chronology, and it is not a radiometric inference — it is counted rings, cross-matched across thousands of trees. Anyone who wants a young earth has to answer dendrochronology, not radiocarbon. We do not think the young-earth literature has answered it.

What we therefore do not say.  We do not present ¹⁴C in coal, diamond, or dinosaur bone as a measurement of the age of the earth.  The parsimonious reading of nearly all of it is process contamination, with a genuine unexplained residue in coal.  The unexplained residue is worth stating exactly as what it is:  an open measurement problem, not a proof.

The one number to keep

Adding 1,975 parts per million of modern carbon — two parts in a thousand — to a sample containing genuinely no carbon-14 makes it read as 50,000 years old.²  We reproduced this from scratch and get 1,981 ppm.⁹  Everything both sides argue about in the deep-time range lives inside that two-parts-per-thousand window.

Part 1: What the clock actually is

Cosmic rays hit the upper atmosphere and knock neutrons loose.  A neutron striking an ordinary nitrogen-14 nucleus kicks out a proton and leaves carbon-14.  That ¹⁴C oxidises to CO₂, mixes through the atmosphere in a year or two, and gets taken up by plants, and by everything that eats plants.

While a thing is alive it keeps exchanging carbon with the air, so its ¹⁴C/C ratio tracks the atmosphere.  When it dies, exchange stops and the ¹⁴C decays back to ¹⁴N with a half-life of about 5,700 years.  Measure how much is left, and you have a clock.

Three details matter for everything that follows.

The half-life used to report ages is deliberately wrong.  Libby's original value was 5,568 years.  The better measurement is 5,730 years, and the current best is about 5,700.  The field kept reporting on the old figure so that decades of published ages would stay comparable, and fixed the discrepancy in calibration instead.  Minze Stuiver and Henry Polach codified the rules in 1977, and it is worth reading what a "conventional radiocarbon age" actually asserts:  "A. the use of the 5568 yr half-life…  B. the assumption of constancy of ¹⁴C atmospheric level during the past…  E. the year 1950 is automatically the base year."¹⁰  Point B is printed in the definition.  A conventional radiocarbon age is a defined quantity, not a claim about elapsed time, and turning it into a date is what calibration is for.  This is also why Taylor's table and ours agree only when both use 5,568.

Everything is normalised for fractionation.  Plants prefer light carbon.  A C₃ plant runs about 25 parts per thousand depleted in ¹³C relative to the marine-carbonate standard; a C₄ plant about 12.  Since ¹⁴C fractionates about twice as much as ¹³C, laboratories measure δ¹³C on every sample and correct the ¹⁴C to a common −25 ‰ baseline — "the postulated mean value of terrestrial wood."¹⁰  Without this, a corn-eater and a wheat-eater of the same age would date centuries apart.

Modern accelerator work counts atoms, not decays.  Beta counting waits for decay events — about 13.6 per minute per gram of modern carbon, dropping to under one per hour at 50,000 years, which is why the old limit was around 20,000 to 40,000 years.  Accelerator mass spectrometry converts the sample to graphite, ionises it, accelerates the ions to millions of electron volts, and separates them by mass.  It needs milligrams instead of grams and it is not fighting cosmic-ray background in a counter.  That change, in the early 1980s, is what created this entire debate:  it made it possible to see the last few parts in 10¹⁵.

The conversion table

Because the whole argument is conducted in "percent modern carbon" and in "years BP," here is the translation, computed for this article on the conventional Libby half-life:¹¹

Read that column of "what sits here" carefully, because it is the entire fourth section of this article in one glance.  The mainstream numbers and the young-earth numbers are not far apart.  They are interleaved.  What separates them is whether the sample went through the laboratory's chemistry.

Part 2: Why a raw age is not a date

Libby's method assumed the atmospheric ¹⁴C/C ratio had been constant.  It has not been, and this was known by 1958.

• The Suess effect. Burning coal and oil dumps carbon with no ¹⁴C in it into the air, diluting the atmosphere. By the 1950s this had pushed the atmosphere several percent below its pre-industrial level.
• The bomb spike. Atmospheric nuclear testing did the opposite, roughly doubling atmospheric ¹⁴C by 1963–64 in the Northern Hemisphere.¹² It has been falling back ever since as the excess mixes into the ocean and biosphere. The falling limb is precise enough to date recent tissue to within a year or two.
• Solar and geomagnetic modulation. A stronger solar wind or a stronger geomagnetic field deflects cosmic rays and cuts ¹⁴C production. Over the Holocene this produces the familiar "wiggles" of a few percent, and over the glacial period a much larger drift.

None of this is a secret, and none of it is a concession dragged out of anyone.  It is the reason the calibration curve exists.

How the curve is built

IntCal20, the current standard, covers 0 to 55,000 calendar years and is assembled in three tiers.⁸

Tier one — counted tree rings, 0 to 12,310 years.  Wood is cut into single rings or small blocks, each ring dated by dendrochronology, and each dated ring measured for ¹⁴C.  This is the backbone.  IntCal20 says it plainly:  "the last 12,310 years for which securely dated tree-ring data are available."⁸  The German Hohenheim oak chronology alone rests on more than 7,000 individual trees, with an average of 108 crossdated trees covering each year.¹³

Tier two — floating tree-ring sequences, to about 13,910 years.  Beyond the absolutely dated chronologies, blocks of ring sequence are placed by wiggle-matching their ¹⁴C pattern against the curve.  Back to about 13,910 cal BP, IntCal20 is still "fully atmospheric" — every point comes from wood that breathed the air.⁸

Tier three — 13,910 to 55,000 years.  Past the trees the curve is built from Lake Suigetsu's varve-counted leaf macrofossils, uranium-thorium-dated speleothems, corals, and marine sediments.  IntCal20 calls Suigetsu "the only quasi-continuous, truly atmospheric record older than ca. 14,190 cal BP."⁸  This is the weakest tier, and its problems are treated at length in Do the Dating Methods Agree? — including a 5.4 % shortfall between Suigetsu's counted laminae and its uranium-thorium calendar scale at 50,000 years, which we computed ourselves and which is not advertised.

The single-year spikes, and why they are the strongest evidence here

In 2012 Fusa Miyake found an abrupt ¹⁴C step of about 15 parts per thousand — roughly 120 radiocarbon years — occurring in a single tree ring in Japanese cedar, at AD 774.¹⁴  The cause is now thought to be an extreme solar particle event.

The reason this matters is not the physics.  It is that the same event shows up in beryllium-10 and chlorine-36 in Greenland and Antarctic ice cores, archives that have nothing to do with radiocarbon and are counted independently.¹⁴  Four such spikes are now confirmed in more than one cosmogenic isotope:  AD 774, AD 993, 2,610 cal BP, and 9,125 cal BP.¹⁴  A fifth and larger one was published in 2023 at 14,300 cal BP, found in annually sampled subfossil Scots pines from the French Alps and matched by a ¹⁰Be anomaly.¹⁴

Take that seriously for a moment.  A one-year event, recorded in the rings of a tree, lines up with a spike in ice layers 9,000 years down a Greenland core.  Two archives, two different isotopes, two different counting methods, one year.  This is the kind of test that ought to settle arguments, and in this case it points against a young earth.  There is no young-earth treatment of the 9,125 and 14,300 cal BP spikes that we have been able to find.

The known-age tests

The other check is to date things whose age is already known from documents.

The Oxford Radiocarbon Accelerator Unit ran exactly this test on dynastic Egypt, sampling short-lived plant material from objects tied to specific reigns.  Their conclusion:  radiocarbon "does give a chronology which is in broad agreement with historical estimates when the right material (short-lived plant remains from good contexts) are used," with "a small regional offset… 19 radiocarbon years older," which they attribute to the Nile Valley's different growing season.¹⁵

Nineteen years.  On material three to four thousand years old.  That is a 0.5 % offset, in a region where the young-earth argument requires the method to be wrong by orders of magnitude.

Part 3: Where radiocarbon really does fail

An honest article has to print the failures at full strength, and there are more of them than a defender of the method usually admits.

Reservoir effects.  Carbon that has been out of contact with the atmosphere is already "old" when a living thing takes it in.  The surface ocean lags the atmosphere by roughly 400 radiocarbon years — Marine13 used a flat 405 years globally, and Marine20 replaced that with a modelled, spatially varying correction precisely because the flat number was too crude.¹⁶  A marine shell, a seal, a person who ate mostly fish:  all read centuries too old unless corrected.

Hard water and volcanic carbon.  Lakes fed through limestone dissolve ancient carbonate, and plants growing in volcanic CO₂ breathe carbon that has no ¹⁴C at all.  Both produce spuriously old ages for material that is genuinely modern.

"Old wood."  A charcoal fragment from the heartwood of a 400-year-old beam dates the tree ring, not the fire.  This is why the Oxford Egypt project insisted on short-lived plant remains.

Regional systematic offsets.  Douglas Keenan's 2002 paper in Radiocarbon documents that early-historical ¹⁴C dates downwind of the Mediterranean run consistently too early, and quotes practising archaeologists saying so bluntly.¹⁷  Manfred Bietak's work at Tell el-Dab'a in Egypt found an average offset of about 120 years, dates "too high," and identified it as "a wider eastern Mediterranean phenomenon."¹⁷

Inter-laboratory scatter.  The Fifth International Radiocarbon Intercomparison sent identical material to more than forty laboratories with the true value withheld.  On a mammoth bone with a consensus age of 39,305 BP, the 57 reported values ran from 21,684 to 42,500 BP, and about 10 % of all returned results were statistical outliers.³  This is the blind test that the charge of "the laboratories tune the answer" calls for, and it is worth being precise about what it shows.  The full range is driven by a handful of bad results, not by the typical laboratory: the interquartile range on that same sample is 35,320 to 40,400 BP.  The honest reading is that a single radiocarbon date near the limit carries a real chance — roughly one in ten — of being badly wrong, and that this is a reason to insist on replication, not a reason to discard the method.

And the scale of all of it.  Add these up honestly.  Reservoir effects:  a few hundred years.  Volcanic dilution:  up to a couple of thousand in the worst documented cases.  Regional offsets:  one to three centuries.  Laboratory scatter on hard, old bone:  tens of percent, on samples near the limit.  These are real and they matter for Bronze Age chronology, which is precisely where this site uses them — see Mesopotamian Absolute Chronology and Dating the Exodus.

They are not remotely enough to turn 50,000 radiocarbon years into 4,500 calendar years.  Anyone using Keenan and Bietak for that purpose is using them against their own arguments;  both propose naturalistic corrections of a few centuries that leave the method intact.

Part 4: Carbon-14 where there should be none

Now the argument that actually matters.

The claim

Samples that ought to be completely ¹⁴C-dead — coal from the Carboniferous, fossil wood from the Jurassic, marble, natural gas, even diamond — routinely come back with measurable ¹⁴C, typically 0.1 to 0.5 pMC, which on the standard conversion is 40,000 to 55,000 "years."  This is not disputed, and it is not a creationist discovery.  Paul Giem catalogued about seventy such measurements from the mainstream radiocarbon literature between 1984 and 1998.¹⁸  Baumgardner extended the table for the RATE project.¹  Radiocarbon laboratories have been publishing these numbers, and arguing about them, for forty years.

The young-earth reading is that this residual ¹⁴C is real and original — that the organisms really did live within the last few tens of thousands of years.  Baumgardner's central statistic is a comparison of two groups inside his own Table 1:  samples of Precambrian, non-biological affinity average 0.062 pMC, while biological samples from Phanerozoic rock average 0.292 pMC, five times higher.¹  On the standard timescale both groups should be equally dead.  The difference, he argues, is intrinsic ¹⁴C from organisms that were alive recently.

The mainstream reading is that the difference is the chemistry.

What a "background" actually is

Everything turns on a distinction that the popular young-earth presentations blur.

• Machine background is what the accelerator reads with no sample and no chemistry — a bare or silver-filled aluminium target holder. Vogel, Nelson and Southon measured 0.00128 ± 0.00056 pMC on a graphite rod, 0.00142 ± 0.00023 on their best graphitised anthracite, and 0.00125 ± 0.00060 on graphitised calcite.¹⁹ This is the number quoted when someone says "AMS can see 100,000 years."
• Process blank is what the accelerator reads on material that is certainly dead after it has been through the identical chemistry a real sample goes through: combustion or acidification, CO₂ transfer, graphitisation over an iron catalyst, packing into the target. In the same 1987 paper, the total system background on large anthracite samples was 0.48 ± 0.16 pMC, equivalent to 43,000 years — three hundred and seventy times the machine floor, on the same instrument, in the same study.¹⁹

Taylor, Southon and Santos list sixteen distinct places contamination can enter, twelve of them in the sample handling rather than the instrument:  adsorbed CO₂ on the oxidiser, adsorbed CO₂ on tube walls, atmospheric leakage into the combustion tube, carbon in the graphitisation catalyst, vacuum-line memory, the sample holder itself, and physical transfer.²

The load-bearing error in the popular young-earth case is treating the machine number as the achievable floor.  Baumgardner writes that the AMS detection limit is "about 0.0005 pMC, corresponding to about 100,000 ¹⁴C years," and describes the blank subtraction as a device "to keep this dilemma largely under wraps."⁴  A bare holder cannot rebut a chemistry-introduced floor, because it has been through no chemistry.

Gregg Davidson and Ken Wolgemuth put the demonstration in one sentence:  ancient graphite "analyzed directly on a modern accelerator mass spectrometer yield values around 0.003 pMC," and "the same graphite samples run after taking them through the whole chemical treatment processes yield values around 0.3 pMC."²⁰  Same material.  Factor of a hundred.

The controls are printed inside RATE's own table

The strongest single answer to the 0.062-versus-0.292 comparison does not require any source RATE did not already cite.  It is in RATE's own Table 1.

Kirk Bertsche — who took his physics PhD under Richard Muller, the inventor of radiocarbon AMS, and then worked in the AMS laboratory at Lawrence Livermore — pointed out two paired entries.⁵

The chemistry added 0.25 pMC in the first pair and 0.14 pMC in the second.  Those are direct, controlled, published measurements of exactly the quantity in dispute — and the gap Baumgardner's argument rests on is 0.292 − 0.062 = 0.23 pMC.

Nearly all the samples in Baumgardner's low-mean "Precambrian" group are geological graphite, which needs only a mechanical surface clean and no chemical processing.  Nearly all the samples in his high-mean "Phanerozoic biological" group were combusted and graphitised.  Baumgardner's own footnote to the figure concedes half of this:  "In most cases, the graphite results do not include such 'background' corrections since they are usually intended themselves to serve as procedural blanks."¹  His two groups are not old carbon versus young carbon.  They are unprocessed carbon versus processed carbon.

The sample-mass test, on 1,140 samples

There is a cleaner discriminator still, and it is decisive.

If the ¹⁴C is intrinsic to the sample, the measured ratio must be independent of how much sample you burn.  A ratio is a ratio.  If instead a roughly constant mass of modern carbon is added by the chemistry, the measured ratio must rise as the sample gets smaller — because the same contaminant is being divided by less material.

Taylor and colleagues plot 1,140 infinite-age background samples of calcite, wood, and coal from the UC Irvine laboratory against sample size.  Above about 0.3 mg the values hold flat at about 0.2 pMC, an apparent age near 50,000 years.  Below 0.3 mg they climb:  a 50 µg sample reads about 30,000 years, and a 10 µg sample about 20,000.²

We fitted that curve for this article with a one-parameter model — a constant added mass m of modern carbon:⁶

One number, 0.8 micrograms of modern carbon, reproduces the whole curve.  And the same figure is independently measured:  Vogel and colleagues determined by direct experiment in 1987 that "total contamination equivalent to 2.2 ± 1.1 µg of modern material was added during processing," a value modern laboratories have since driven down.¹⁹

Intrinsic ¹⁴C cannot produce a mass dependence at all.  This is not an interpretive dispute;  it is a shape in the data that one hypothesis predicts and the other forbids.  We are not aware of any young-earth treatment of it.

The diamonds, refuted by RATE's own follow-up

Diamond is the best possible test material.  It is chemically inert, it has no pore space, it hosts no bacteria, and it can be measured either as a raw crushed chip loaded straight into the target or as fully processed graphite.

The RATE project crushed twelve African diamonds, sent them to a commercial accelerator laboratory, and got 0.10 to 0.39 pMC — mean 0.168.¹  Baumgardner's conclusion is that "diamond, the material that ought to be the ultimate 'standard background' because of its extreme resistance to external contamination, displayed a ¹⁴C/C ratio above the laboratory's standard background for each of the twelve samples analyzed."

Two things dissolve this.

First, the blank was measured on the wrong material.  RATE's laboratory determined its blank by measuring purified natural gas, and got 0.077 pMC.  Baumgardner records this himself:  the subtraction "is justified by the laboratory under the assumption that the ¹⁴C in the natural gas analyses must somehow represent contamination."¹  Gas goes straight into a combustion line;  a diamond has to be crushed, burned, and graphitised.  Comparing the two is comparing a processed measurement against an unprocessed blank.  Against the right blank — a mainstream laboratory's processed diamond, at 0.15 pMC — RATE's first six stones (five from kimberlite pipes, one alluvial from Guinea) average 0.122 pMC, which is 81 % of background, not above it.²¹  The young-earth side has since set that bar higher still:  Cupps and Thomas, reading the same mainstream table, derive a 99.7 % detection limit of 0.18 pMC for processed diamond.²²  By their own threshold, all six of RATE's first diamonds and three of the six Namibian stones are non-detections.

Second, and decisively, RATE ran the follow-up themselves and it went against them.  In 2006, after the RATE book had gone to press, the team put diamonds directly into the AMS sample holder with no chemistry at all.  Baumgardner reports the result:  "Our ten diamond samples displayed ¹⁴C values between 0.008 and 0.022 pMC, with a mean value of 0.014 pMC."⁴  He does not say whether these were the same stones, or which laboratory ran them, so this is a follow-up rather than a matched control.  It is still RATE's own material, RATE's own technique, and RATE's own number.

Taking the chemistry away divided the signal by twelve.²³  That is RATE's own experiment, on RATE's own material, published on a young-earth website, and it is the answer to RATE's own diamond claim.  Baumgardner reports it fairly and adds, "caution is obviously advisable in their interpretation" — though in his next sentence he still concludes that "the most plausible explanation, astonishing as it may be to some, is that natural diamond contains measurable and reproducible levels of intrinsic ¹⁴C."⁴

One detail deserves to be kept, because it is to Baumgardner's credit.  The six Namibian alluvial stones scatter far more widely than the deep-mine ones — 0.12 to 0.39 pMC — and he noticed this and changed his hypothesis in print, moving from "primordial ¹⁴C formed during the creation of the earth itself" to in-situ production from neutrons.¹  Alluvial diamonds sat near the surface in water-saturated gravels;  deep-mine stones did not.  That the surface-residence stones are the dirty ones and the deep-mine stones the clean ones is what a contamination model predicts, and it is mostly what was found.  The one exception should be stated.  The Guinea stone in the first set was alluvial too, and it read 0.11 pMC, indistinguishable from the kimberlite stones.¹  Whatever made the Namibian stones dirty was not alluvial residence alone.

He does still argue that unprocessed readings of this size are intrinsic, because Taylor and Southon's 0.015 pMC fragments exceed a bare aluminium holder, "with no sample or silver powder present," by a factor of twenty-five.⁴  But there is no single machine background.  Taylor and colleagues' own 2018 report lists silver powder packed into an aluminium target, with no sample in it, at inferred ages of 59.7 and 60.9 ka — 0.051 to 0.059 pMC, four times higher than the unprocessed diamonds themselves.^{2 24}  Baumgardner himself quotes the 2007 paper as saying that "¹⁴C count rates from silver powder cathodes were comparable with those from diamonds."⁴  A diamond reading below a target with no sample in it is not evidence of carbon in the diamond.  And even taken at face value, 0.014 pMC is an inferred age of 71,000 years, which is no help to a young earth in any case.

Conclusion on diamonds:  this argument should be dropped.  Not because a critic said so, but because RATE's own follow-up measurement says so.

The Taylor and Southon exchange

For completeness, because the young-earth literature cites it heavily:  Taylor and Southon's 2007 paper on natural diamonds prompted a published exchange.  Bertsche argued that their ¹⁴C values correlated inversely with ion-source current, which is the signature of instrument memory rather than intrinsic carbon.  Baumgardner replied in 2014 that the correlation appears only in part B of their Table 2 and is absent from part A, where six fragments of one diamond gave stable values across a fifty-percent range of beam current, and he quotes the authors as writing that "¹⁴C from the actual sample is probably the dominant component of the 'routine' background."⁴

Only the abstract of Taylor and Southon 2007 could be obtained.  It confirms the sentence Bertsche quotes, verbatim:  "The oldest ¹⁴C age equivalents were measured on natural diamonds which exhibited the highest current yields."  It also gives the numbers:  fourteen diamond samples, ion currents of 110 to 250 µA of ¹²C⁻ where graphite typically yields about 150, and apparent ages from 64.9 to 80.0 ka.²⁵  The section-1 sentence Baumgardner quotes, and the part A / part B structure of Table 2, are in the body of the paper and remain unchecked, so we will not lean on them either way.  What we can check is what the same authors wrote eleven years later, in a paper written specifically to address this argument:  that of all the explanations for ¹⁴C in background samples, intrinsic ¹⁴C "would be, in our view, the least likely scientific explanation."²  Whatever the 2007 sentence meant, its authors do not read it the way Baumgardner does.

What the young-earth side has said since

The unprocessed-diamond result has been on the Answers in Genesis website since 2007.  Every young-earth treatment of diamonds written after it either does not mention it or reads it as confirmation.

Snelling, in 2007, presented Taylor and Southon as "confirmation that there is in situ carbon-14 in diamonds," and in the same article conceded the mechanism this section rests on.  The Irvine diamonds read lower than RATE's, he wrote, because they "were mounted directly in the beam," whereas RATE's were combusted and graphitised, and "that process may have introduced some more carbon-14 to the analyses."²⁶  His remaining question — why the same instrument read more ¹⁴C in Precambrian graphite than in diamond — has a published answer.  Ion-source memory depends on the surface of the target, and on the Irvine instrument graphite gives 0.020 to 0.035 pMC where diamond gives 0.005 to 0.02.⁵

Sarfati's 2006 Creation article, still the most-read version of the argument and updated as recently as November 2024, compares the 0.12 pMC diamond mean with the 0.08 pMC natural-gas blank and answers four objections:  "background radiation in the detector," radium cluster decay, neutron capture, and the 55,700-year figure.  None of the four is the objection that matters.  The article never mentions what the blank was measured on, and its 2024 update adds only a pointer to Cupps and Thomas.²⁷  Joel Tay's 2016 reply for Creation Ministries International quotes Bertsche's summary paragraph in full — including the sentence "the unprocessed diamond samples probably reflect instrument background" — and asks "where is the evidence of such contamination?" without mentioning the 0.014 pMC measurement that sentence refers to.²⁸

The most technical response is Cupps and Thomas in the Creation Research Society Quarterly for 2019, answering Taylor, Southon and Santos directly.  They re-plot Taylor and Southon's fourteen diamond measurements against a three-sigma detection limit and conclude that "from six to ten of their ten diamonds contained intrinsic radiocarbon."²²  Their central argument, as Thomas puts it in his 2020 summary, is that "background samples already account for lab contamination," so a sample above the blank cannot be blamed on the laboratory.²⁹  That is right, and it is exactly why the material of the blank matters.  RATE's blank was gas, its samples were solids, and a gas blank does not account for the chemistry a solid goes through.  Cupps and Thomas also derive their own detection limits from the Irvine table — 0.18 pMC for processed diamond — without applying them to RATE's diamonds, nine of which fall below that limit.  And the paper does not mention RATE's 2006 unprocessed result at all.²²

The coal, which is the part that does not go away

Here the young-earth side has something, and — importantly — its sharpest critic grants it.

RATE took ten coals from the U.S. Department of Energy Coal Sample Bank at Penn State — collected in argon-purged drums from active mines, ground to 20 mesh, stored refrigerated — spanning Eocene to Pennsylvanian.  The measured values, before the laboratory's 0.077 pMC subtraction, average 0.324 pMC.²¹  The same style of laboratory measures 0.12 to 0.14 pMC on processed coal.²

Bertsche's own words:  "Unlike the literature values, Baumgardner's coal samples do show significant radiocarbon above background, inviting explanation."⁵  He also concedes the general point:  "some materials, e.g., coals and carbonates, often do show radiocarbon contamination that cannot be fully accounted for."⁵

So a factor of two and a half stands unexplained.  Four candidate explanations, none decisive:

1. Different laboratories have different process blanks, sometimes by a factor of several. RATE's laboratory characterised its own blank on gas and so may never have measured its own solid-sample floor.
2. Grinding to 0.85 mm creates enormous surface area, and coal is a strong physical adsorber of CO₂ — the property behind proposals to sequester CO₂ in coal seams. Finely divided coal exposed to modern air is close to a worst case for adsorbed modern carbon.
3. Biology. David Lowe's 1989 note in Radiocarbon proposed that the finite ages routinely found in coal come from fungal and microbial action in the substrate, since wood-rotting fungi such as Polyporus versicolor fix atmospheric CO₂ into the coal, and autoclaving kills the organisms but leaves their ¹⁴C-bearing hyphae and waste behind.³⁰ RATE's argon-purged, refrigerated sourcing is a partial answer to this.
4. Uranium. Coal often sits in or near uranium-bearing rock, and the sample handler's own view, relayed by Bertsche, was that the RATE coals were contaminated before they arrived.⁵

Note also that RATE's coals show no trend with geological age — Eocene and Pennsylvanian coals give the same range.  Both sides claim this, and it discriminates nothing:  a constant process blank predicts a flat distribution just as a single burial event does.

What is diagnostic is internal scatter.  Andrew Snelling sent splits of one body of fossil wood from Crinum, Queensland, to two laboratories and got conventional ages of 29,544, 37,800, 44,700, and >35,620 BP.³¹  Converted to pMC that is roughly 2.5 down to 0.38 — a factor of six and a half between splits of the same trees.  Snelling calls the differences "incongruous," which they are.  A genuine, uniform, pre-Flood ¹⁴C level does not do that.  Variable contamination does.

In-situ production

Could the ¹⁴C be made in place, underground, after burial?

The mechanism is real and nobody disputes it.  Uranium and thorium in rock throw off alpha particles.  Those alphas strike light nuclei — aluminium, sodium, silicon, magnesium, oxygen — and knock neutrons loose;  uranium-238 also fissions on its own.  A freed neutron that finds a nitrogen-14 nucleus makes carbon-14, by exactly the reaction that fills the atmosphere with radiocarbon.  Two weaker reactions do the same job:  ¹³C(n,γ)¹⁴C on the 1.07 % of carbon that is ¹³C, and ¹⁷O(n,α)¹⁴C on the 0.038 % of oxygen that is ¹⁷O.

So the question is never whether.  It is how much, and that is arithmetic.

Start with the structural point, because it decides more than the numbers do.  ¹⁴C made underground also decays underground, and production and decay settle into a steady state after about 38,000 years.  A sample buried for 70 million years and one buried for 300,000 years, in the same rock, should therefore hold the same amount of in-situ ¹⁴C.  Burial time drops out of the answer entirely.  What is left is a prediction with teeth:  if in-situ production is what laboratories are measuring, the measurements should track the sample's nitrogen and its uranium, and should be flat against its age.

Giem saw this and said so, and his version is sharper because he chose the right variable:  since nitrogen-14 captures thermal neutrons about 110,000 times more readily than carbon-13, if neutron capture mattered then radiocarbon ages "should vary wildly with the nitrogen content of the sample," and nobody has found that.¹⁸  We recomputed that factor from cross sections and isotopic abundances alone and got 107,000.  Giem is right.³²

Working the numbers

The clearest published calculation is Baumgardner's own.  Using measured thermal-neutron production in deep-mine granite, the ¹⁴N(n,p)¹⁴C cross-section of 1.8 barns, and a generous 1,000 ppm nitrogen in diamond, he obtains a steady-state ratio of 9 × 10⁻⁶ pMC — "more than 13,000 times smaller than the mean value of 0.12 ± 0.01 pMC we measured."¹  His arithmetic is correct;  we reproduced it in his own units and got his number to three digits.³²

But his calculation is narrow.  It covers diamond, at one flux, by one reaction.  So we built the general case as a tool you can run yourself, and put it on this site:  the in-situ radiocarbon calculator.  It computes neutron production from the rock's actual composition using the yields in Šrámek et al. 2017, works out what fraction of those neutrons nitrogen wins against everything else competing to absorb them, runs all three ¹⁴C-producing reactions, takes the steady state, and reports the result as pMC, as an uncalibrated radiocarbon age, and as a calibrated calendar age.

Every default in it is set to favour in-situ production:  the largest measured neutron production rather than the typical one, no trace neutron poisons unless you add them, no leakage out of the sample, and the sample placed inside a host rock that outproduces it.  It is graded against Šrámek's published totals, Baumgardner's arithmetic, Giem's cross-section ratio, and Fabryka-Martin's independent thesis calculation, and it is required to return exactly zero when given no uranium.

Set it to a coal seam in average continental crust and it returns 5.1 × 10⁻⁵ pMC, an apparent age of 116,000 years.  RATE's coals measured 0.324 pMC.  The gap is a factor of about 6,300.

The concession that has to be made first

There is a real weakness in the standard dismissal of this argument, and it should be stated before the rebuttal.

Fossil bone is not ordinary rock, and it really can out-uranium a granite.  Buried bone scavenges uranium from groundwater — that is the basis of uranium-series dating of bone, and on the Colorado Plateau prospectors used fossil bone as an ore guide.  Measured values:  37.8 ppm uranium in the Tyrannosaurus rex femur MOR 1125, by laser-ablation ICP-MS;³³  283 ppm in a proboscidean tooth from Chad;³⁴  and "in excess of 800 ppm U and 200 ppm Th" in some bone cortices from Olduvai Gorge and Kanam.³⁵  Millard and Hedges put the ceiling at "as much as 0.1 % by weight" — 1,000 ppm.³⁶  Stripa granite, chosen for Swedish nuclear-waste research because it is unusually radioactive, holds 44 ppm.³⁷  On uranium alone, fossil bone can be six to eighteen times hotter than the hottest well-studied granite.

Bone collagen is also about 16 % nitrogen by weight, which is more nitrogen than any other material anyone submits for radiocarbon dating.

So bone is the best case in the world for this argument, and any treatment that waves it away with "uranium concentrations are negligible" has not done the work.

What the numbers do to Brian Thomas's catalogue

Brian Thomas's 2018 University of Liverpool PhD thesis is the most careful young-earth compilation of radiocarbon measurements on dinosaur bone.  His Table 7.1 lists 21 Mesozoic determinations across nine dinosaurs, from three geological provinces, in three bone fractions, at five laboratories.³⁸  They run from 0.57 pMC (an unidentified dinosaur, apatite) to 6.61 pMC (a stegosaurid, apatite).

Give in-situ production every advantage at once — permineralised bone at 100 ppm uranium in its own right, sitting inside a 44 ppm uranium granite, and collagen at the full 16 % nitrogen for the collagen fractions — and the calculator returns a ceiling of 9.5 × 10⁻³ pMC for a collagen fraction and 1.0 × 10⁻³ pMC for an apatite fraction.³²

Against Thomas's own measurements that is a shortfall of 228-fold at best and 6,500-fold at worst.  Not one of his 21 values is within three orders of magnitude of what neutron capture can supply.

Push it harder.  Take the hottest fossil bone ever published, 800 ppm uranium and 200 ppm thorium, and let it make its own neutrons:  in-situ production reaches 0.018 pMC, still 32 times short of his lowest value.  To reach 0.57 pMC the bone would have to be roughly 4 % uranium by weight;  to reach 6.61 pMC, about 65 % uranium, which is to say the specimen would have to be pitchblende rather than bone.³²

Thomas rejects in-situ production too, and this is where it gets awkward for both sides.  His Table 7.4 lists it as a possible contaminant and records "None" detected, on the ground that "concentrations from Hell Creek Fm., Montana were far too low, at 0.020 mg/kg for U and 0.078 mg/kg for T[h]."³⁸

That number is almost certainly wrong, and wrong in the direction against his own case.  0.020 ppm uranium is a hundred and thirty times below average continental crust and about 1,900 times below the measured uranium in a T. rex femur.  No permineralised bone has ever been published at that level.  The thesis cites Appendix 7.1 for it, but Appendix 7.1 reproduces only the cover sheet of a commercial analytical report — addressed to Hugh Miller and dated 2012 — and not the data page, so the figure cannot be checked from the thesis at all.³⁸

He also names the wrong mechanism.  His table calls it "cluster decay of U & Th generate ¹⁴C from N in collagen," and his note argues that "for Uranium's beta decay to have supplied the detected levels of ¹⁴C… the majority of each fossil and its surroundings would have to have been composed of U."  Cluster decay, beta decay and neutron capture are three different processes, and the one that actually matters here — a neutron from an (α,n) reaction hitting ¹⁴N — is not any of the three he addresses.

So his conclusion is right and his reasoning is not.  The honest way to put it:  in-situ production fails for dinosaur bone, but it fails by a much larger margin than Thomas's own figures would give, and it fails for a reason he never states.

Why it fails, in one sentence

Dilution.

In-situ production makes ¹⁴C atoms.  A radiocarbon measurement reports a ratio, ¹⁴C to total carbon.  Any material with a normal amount of ordinary carbon in it swamps the handful of atoms neutron capture supplies.  This is not our finding — it is the conclusion of the standard mainstream treatment of subsurface radionuclide production, Fabryka-Martin's 1988 Arizona thesis, whose abstract states flatly that "deep subsurface production of ³H, ¹⁴C, ⁸⁵Kr and ⁹⁹Tc is generally below detection."³⁷

She then rigs a worst case on purpose:  put all the uranium and boron in a granite into 1 % of the rock volume, giving effective concentrations of 900 ppm boron and 300 ppm uranium, and ¹⁴C production jumps fiftyfold.  Her verdict:  "Even in this case, however, the ¹⁴C content of the bulk rock would only be 0.005 pmC, still well below detection, because of dilution by the high C content of these rocks."³⁷

And permineralised bone is a worse case than her granite, on both variables at once.  Its carbonate-fluorapatite carries carbon in the mineral lattice at the percent level against her granite's 300 ppm, which enlarges the denominator;  and the collagen that held the nitrogen is precisely what fossilisation removes, which shrinks the numerator.  A permineralised bone has lost the target nucleus the reaction needs while gaining ordinary carbon to hide the product in.

Where in-situ ¹⁴C is real, and it is worth saying so

Two places, both documented, and neither of them helps the young-earth argument.

In uranium ore.  Barker and Torkelson measured ¹⁴C by AMS in uranium-bearing minerals and found it "proportional to the concentration of uranium in that mineral," averaging (2.2 ± 0.5) × 10⁻¹⁵ ¹⁴C atoms per uranium atom.³⁹  This is the only place in-situ radiocarbon has actually been detected rather than calculated.  It works there because ore is nearly carbon-free, so a small number of ¹⁴C atoms makes a large ratio.  Applied to Thomas's bones it would give still less than the neutron-capture model does — the calculator reports both, and our modelled channels sit above their measurement, not below it, which is the direction an honest bound should sit.

In groundwater running through an orebody — but here the field data are worth reading carefully, because they are usually quoted only halfway.  Fabryka-Martin calculated that water flowing through the Koongarra uranium deposit could in principle pick up 5.5 to 400 pMC from in-situ production, and noted that the 51 pMC measured in borehole PH 49, which intersects the primary ore, "could conceivably be due entirely to in-situ nuclear processes."³⁷

That is a statement about an upper bound, not a detection, and her own Table 6-5 two chapters later shows why the distinction matters.  The ore-intersecting borehole reads 51 pMC.  The upgradient control outside the deposit's influence reads 64.  The four downgradient wells read 60, 92, 94 and 101.³⁷  The well inside the orebody is the lowest number in the set.  Nothing there looks like uranium adding radiocarbon to water.

And the same waters are not short of in-situ nuclides:  their iodine-129 is enriched one to three orders of magnitude over the equilibrium ratio of the source minerals.³⁷  So the method detects in-situ production when it is there.  At Koongarra it finds no ¹⁴C anomaly at all.

Where the effect is real, it makes water look younger, not older, which is the opposite of what a young-earth argument needs.

The one direct measurement in ancient organic carbon

There is a measurement that settles the size of this effect in exactly the material the argument is about — ancient buried organic carbon — and it comes from particle physics, not from either side of this dispute.

The Borexino collaboration needed the most radiocarbon-free liquid scintillator obtainable, so they made it from petroleum pumped from deep underground and then measured what ¹⁴C was left.  Their prototype detector found ¹⁴C/¹²C = (1.94 ± 0.09) × 10⁻¹⁸.  A later batch, from a different oilfield, came in at 11 × 10⁻¹⁸.⁴⁰

Convert to the radiocarbon convention:  1.94 × 10⁻¹⁸ is 1.7 × 10⁻⁶ pMC.  That is real, ancient, buried organic carbon, sitting in real rock, with its residual ¹⁴C measured to nine per cent precision — and it is about five orders of magnitude below the 0.1 to 0.5 pMC that young-earth radiocarbon work reports from coal, diamond and bone.

Two further details make it more useful still.  Schönert and Resconi identified ¹⁴N(n,p)¹⁴C as the main source of that residual, which is the same reaction modelled here.  And Bonvicini and colleagues derive the dependence as r ∝ U·M·N/B — uranium, the light nuclei that turn alphas into neutrons, and nitrogen on top;  boron on the bottom.⁴⁰  That is the structure of the calculation above, arrived at independently, including the sign of the boron term.

The calculator's answer for a nitrogen-poor hydrocarbon in ordinary crust falls inside the range Borexino measured.  Two ways of getting at the same quantity — one a calculation from cross sections, one a low-background physics experiment with no stake in the argument — agree.

A wrinkle that cuts slightly the other way

Honesty requires one correction to the picture drawn so far, and it is one this article got wrong at first.

Nitrogen is not the only route, and boron is not only a neutron poison.  ¹¹B(α,p)¹⁴C makes radiocarbon from an alpha particle directly, with no neutron in the chain at all, and Fabryka-Martin classes it as a major deep-lithosphere mechanism — one of four that can each supply more than 5 % of the total — alongside ¹⁴N(n,p)¹⁴C and ¹⁷O(n,α)¹⁴C.³⁷  In clays and shales it dominates, because those rocks are boron-rich.

So a model that carries boron only as a sink has the sign of the boron term wrong.  The calculator now runs this fourth channel, using her published α-capture coefficients, and reproduces her worked example for high-Ca granite — 9 ppm boron, 3 ppm uranium, 8.5 ppm thorium, 13.8 ¹⁴C atoms per gram.³²

It does not rescue the argument.  Run boron up through realistic concentrations and the poisoning still wins:  in a coal seam, 50 ppm of boron lowers in-situ ¹⁴C by about 13 %, and 500 ppm lowers it by half.  Only above roughly half a per cent boron by weight does α-capture overtake the poisoning, and that is borate ore, not rock.  The point of saying so is that the honest version of the argument is not "nitrogen is everything";  it is that all four channels together, at every realistic composition, come to three or four orders of magnitude less than what is measured.

The verdict

In-situ production at present-day decay rates is ruled out for coal, for diamond, and for dinosaur bone — by the young-earth side's own arithmetic in Baumgardner's case, and by three or more orders of magnitude in every case we can compute.  It survives only if you assume a past episode of accelerated nuclear decay, which is a much larger claim, treated on this site under radiometric concordance.  To reach RATE's coal value that way needs neutron production about 42,000 times today's.  Thomas's bones start from a higher baseline, so they need less — about 560 times for his lowest value and 6,500 times for his highest — but that is still a demand for an episode of nuclear decay three to four orders of magnitude above anything measured, imposed on top of the two free parameters the flood-dilution model already needs.³²

And the falsifiable prediction still sits there unclaimed.  In-situ production says the signal should track nitrogen and uranium and ignore age.  Thomas's own catalogue is the natural test — 21 measurements, nine animals, three provinces — and it reports no uranium data for the specimens measured, and no nitrogen data either.  That experiment has not been run.  It should be.

Part 5: Can a flood model actually carry these numbers?

Suppose the residual ¹⁴C in coal were entirely real.  What would it take to fit it into a young earth?

The standard answer is dilution.  Before the Flood, the argument runs, there was far more carbon circulating in the biosphere — all the carbon now locked in coal, oil, gas, and some fraction of limestone.  Cosmic rays make ¹⁴C at a rate set by the cosmic-ray flux, not by how much ordinary carbon is around, so a much bigger carbon pool holding the same number of ¹⁴C atoms gives a much smaller ratio.  Baumgardner's estimate:  the pre-Flood biosphere held "conservatively, 300–700 times the total C relative to our present world," giving a pre-Flood ratio of about 0.2 pMC, decaying to about 0.12 pMC today.¹

Which is, indeed, about what is measured.  So the model reproduces the data.  The question is whether it was predicted or fitted.

The carbon budget does not close

We worked the inventory out from published reservoir sizes.⁷  Living biomass today is about 550 gigatonnes of carbon.⁴¹  The published fossil-fuel reserves are 1,002 to 1,940 GtC — 383–1,135 for gas, 173–264 for oil, 446–541 for coal.⁴²  The dispersed organic carbon in sedimentary rock, which is where most buried organic carbon actually is, is about 1.25 × 10²¹ moles, or 15 million GtC.⁴³

Look at where the required row falls.  It is not near either end.  Count what is actually recoverable as fossil fuel and the dilution is three to five times, which predicts coal at 13 to 21 pMC — a hundred times more ¹⁴C than anyone measures.  Count everything that flood geology says was buried by the Flood, which must include the dispersed organic carbon in every Phanerozoic shale, and the dilution is 27,000 times, which predicts 0.002 pMC — a hundred times less than anyone measures, and below the machine floor.

To land on 500× you have to select about 1.8 % of the sedimentary organic carbon reservoir and no more, with no independent reason for that fraction.  Giem is admirably open about this;  his derivation is an explicit chain of guesses — "1/200 × 1/4 × 1/5 × 60%" — and he says outright that "much of the spread in the upper limit is due to differences in assumptions regarding the pre-Flood magnetic field, assumptions that are presently untestable."¹⁸  Baumgardner says his own figure is "very soft."¹

That is not dishonest.  It is just not evidence.  A model with a free parameter that is set by the answer cannot then be offered as confirmation of the answer.

The refill problem

There is a second difficulty, and it is arguably worse.  If the atmospheric ratio was 0.2 pMC at the Flood and is 100 pMC now, it rose by a factor of 500 in about 4,500 years.  Baumgardner acknowledges this directly:  "it is necessary for the ¹⁴C/C ratio to have increased dramatically and rapidly, by a factor on the order of 500, to reach its present value after the cataclysm.  The means by which this could have occurred has long been a difficulty."¹

The difficulty is not just mechanistic.  It is that we can see what the ratio did over that interval, ring by ring, in wood.  Across the tree-ring-anchored portion of IntCal20 the atmospheric ratio wanders by a few percent, not by a factor of 500.  A 500-fold rise over the same window would leave a monstrous, unmistakable signature:  organic material from 4,400 years ago would date tens of thousands of years old, material from 4,000 years ago somewhat less, and the apparent ages of Egyptian and Mesopotamian objects would be wildly, systematically wrong in a way no archaeologist could miss.  Instead the Oxford project finds a 19-year offset on dynastic Egypt.¹⁵

Accelerated decay cuts the wrong way

The RATE project's central proposal elsewhere is that nuclear decay ran far faster in the past.  Applied to ¹⁴C, that proposal destroys the very evidence this chapter depends on:  any acceleration large enough to age uranium by hundreds of millions of years would annihilate every ¹⁴C atom in a pre-Flood organism many times over.  Baumgardner sees this and proposes that the acceleration was selective — smaller for short-half-life isotopes — deriving "only about 2000 years worth of accelerated ¹⁴C decay" by extrapolating discordances in a Grand Canyon diabase sill, and calling the scaling "speculative as it may be."¹

He is right that it is speculative.  It is also an extra free parameter, tuned in the direction the argument needs.  Two tuned parameters — the dilution factor and the decay-acceleration exemption — is a great deal of adjustment for one measured number.

Part 6: The fact that has to be answered

Strip away everything above and one thing remains.

Radiocarbon calibration back to about 12,310 years rests on counted tree rings.  Not on radiometric assumptions.  Not on decay constants.  On rings, cross-matched between thousands of overlapping trees.  The German Hohenheim oak chronology draws on more than 7,000 individual oaks, averages 108 crossdated trees for each year of its length, and was independently confirmed against the Irish and Göttingen chronologies back to 9,741 years.¹³  When the floating Preboreal Pine Chronology was finally matched to it dendrochronologically, the ring-count position and the radiocarbon wiggle-match position differed by 8 years at about 11,000 — 0.07 %.¹³  A separate chronology, a different species on a different continent in a different climate, runs continuously from bristlecone pine in the White Mountains of California back to 6700 BC — 8,681 years, as of 1983.⁴⁴

Twelve thousand three hundred and ten calendar years before 1950 is 10,360 BC.  That is about twice the age of the earth on a Masoretic biblical chronology, and about eight thousand years earlier than the usual young-earth date for the Flood.  A young-earth position must therefore hold that the tree-ring chronologies are wrong by a factor of two or more — that thousands of trees have been systematically mis-crossmatched in a way that inflates the count without producing detectable inconsistency, and that the German, Irish, and Californian chronologies happen to be wrong by about the same amount.

The young-earth literature has tried.  The usual proposal is that bristlecone pines and other stressed trees lay down more than one ring in a year, so the counts are inflated.  Multiple rings do occur in stressed conifers.  Three things weaken the argument.  Cross-dating exists precisely to catch them:  a false ring in one tree does not appear in its neighbours, so the pattern-match fails and the anomaly is found.  The argument is aimed at the wrong trees:  IntCal20's backbone is European oak, growing in a temperate climate with a single unambiguous growing season, not high-altitude bristlecone.  And the required error is enormous — not a few percent but a factor of two, in the same direction, in every chronology at once.

We are not aware of any young-earth work that has demonstrated a factor-of-two error in the Hohenheim oak chronology, or explained the ¹⁴C spike at 9,125 cal BP that is matched by ¹⁰Be in polar ice.

This article says plainly:  that is an unanswered problem, and it is a serious one.  We would rather print it than pretend it away.  The rest of this site holds that the earth is young;  this particular line of evidence runs against that, and a reader is entitled to know it from us rather than from a critic.

Objections and responses

"Carbon dating has been proven wrong — living snails dated 27,000 years old." The number is real and it is not exaggerated:  Melanoides tuberculatus taken alive from artesian springs in southern Nevada gave "¹⁴C contents as low as 3.3 ± 0.2 percent modern (apparent age, 27,000 years)."⁴⁵  But read the next clause of the same sentence.  Riggs attributes it "to fixation of dissolved HCO₃⁻ with which the shells are in carbon isotope equilibrium" — the hard-water reservoir effect — and says the point of publishing is "so that erroneous ages are not attributed to freshwater biogenic carbonates."  A known, named, mechanistically explained, correctable artifact in freshwater shell is not evidence that dates on terrestrial plants are wrong.  Quoting the first half of that sentence without the second is the kind of argument that loses a careful reader on the spot.

"They subtract 0.8 pMC from every sample to hide the carbon-14." The subtraction is real;  the characterisation is wrong.  A process blank is a measured quantity — the ¹⁴C that certifiably dead carbon picks up going through the identical chemistry — and it varies by laboratory and by material, which is why it is measured rather than assumed.  Not subtracting it would be the error.  The legitimate complaint is narrower and better:  that a laboratory should measure its blank on the same kind of material as the sample.  RATE's laboratory did not, and that is where their diamond result comes from.

"AMS can detect one atom in 10¹⁵, so it should see 100,000 years." The instrument can.  The instrument plus the chemistry cannot.  The floor that matters is the process blank, around 0.1 to 0.5 pMC, which corresponds to 47,000 to 57,000 years, and that is exactly the practical limit laboratories quote.

"Carbon-14 in diamonds proves a young earth." It does not, and the reason is RATE's own measurement, not a critic's.  When the RATE team put diamonds directly into the accelerator with no chemistry, they read 0.014 pMC instead of 0.168.⁴  No young-earth treatment of diamonds since 2007 has engaged that measurement;  the ones that mention it at all call it intrinsic anyway.  Drop this argument.

"But nobody has explained the coal." Correct, and we said so.  RATE's coals sit about two and a half times above a mainstream processed-coal blank.  That is an open measurement problem worth pursuing, and the honest way to pursue it is the way Giem proposed:  run coal from different parts of one seam, from different depths, plus coke, coal tar, petroleum, and wood, through identical chemistry at a laboratory that has characterised its solid-sample blank, and see whether the values track the material or track the chemistry.¹⁸  Until someone does that, "unexplained" is the right word, and "proves a young earth" is not.

"Radiocarbon assumes a constant atmosphere, and that assumption is false." The assumption was abandoned in the 1950s.  Everything since then is calibration, which is the opposite of assuming constancy — it is measuring the variation directly against independent clocks.

"The calibration curve is circular:  they use radiocarbon to date the tree rings." No.  Tree rings are dated by counting and cross-matching, with no radiocarbon input.  The ¹⁴C measurement is then read off the independently dated ring.  Circularity would require the ring dating to depend on ¹⁴C, and for the first 12,310 years it does not.  The charge has more force in the third tier of the curve, past 14,000 years, where floating sequences and marine records are placed partly by wiggle-matching — and there it is a fair objection, treated at length in Do the Dating Methods Agree?

"Archaeologists throw out radiocarbon dates that disagree with their expectations." Some do, and named archaeologists have said so in print;  Keenan collects the quotations.¹⁷  But the frequently circulated line — "among archaeologists, carbon dating is a big joke… they never date their finds by carbon-14" — has no traceable source that we can find, and we do not use it.  The documented version of this complaint is real and is worth citing;  the punchy version appears to be apocryphal.

What we conclude

Radiocarbon dating works, it is tested against clocks it did not build, and its failures are the kind you find in a mature method:  named, mechanistic, and measured in percent.

The young-earth argument from residual ¹⁴C in ancient carbon has one strong form and several weak ones.  The weak ones — diamonds, dinosaur bone, "they hide it by subtracting," the 0.062-versus-0.292 comparison — should be abandoned, and this article abandons them.  Three of the four are answered by measurements the young-earth side itself published or tabulated:  RATE's own unprocessed diamonds, the two re-graphitised graphite controls inside RATE's own Table 1, and Baumgardner's own neutron-flux calculation.  The strong form is narrow:  RATE's coal measurements sit meaningfully above the process blank of comparable mainstream laboratories, and nobody has published an explanation.  Bertsche, who wrote the sharpest critique of RATE in print, grants exactly this and no more.  It deserves to be pressed, by the experiment Giem outlined, at a laboratory that characterises its blank on solid samples of the same material.

But even in its strong form, that residue is not an age for the earth.  Converting it into one requires a dilution model whose central parameter has no independent support and whose consistent versions overshoot in both directions, plus a selective decay-acceleration exemption invented for the purpose.  And it has to get past the tree rings, which it does not.

The honest summary is uncomfortable and it is the one we will print:  on this particular question, the evidence as it currently stands runs against a young earth, and the strongest thing the young-earth side has is an unexplained factor of two and a half in a coal blank.

Sources