Do Stars Form By Themselves?

"The scary part is that if none of us knew in advance that stars exist, frontline research would offer plenty of convincing reasons for why stars could never form." — Neil deGrasse Tyson, Death by Black Hole, page 187¹

"It has been known for more than 30 years that star formation in giant molecular clouds (GMCs) is slow, in the sense that only ∼ 1% of the gas forms stars every free-fall time." — Mark Krumholz and Jonathan Tan, 2007²

"We haven't actually observed star formation in a telescope." — Danny Faulkner, Answers magazine, 2014³

Summary

Several young-earth writers argue that stars cannot form on their own.  Gas in space wants to expand, not contract.  Magnetic fields resist compression.  Spinning clouds fling themselves apart.  Nobody has ever watched a star switch on.  If all that is right, then every star in the sky was made directly, and the hot blue stars — which burn out fast — prove the universe is young.

That argument fails, and it fails on measurements rather than on theory.

The central physics claim is backwards.  Jason Lisle writes that "in a typical nebula, the gas pressure far exceeds the miniscule force of gravity."⁴ John Hartnett writes that as a cloud contracts "the required Jeans mass increases."⁵ Both statements are true only for a gas that cannot get rid of its heat.  Star-forming clouds get rid of their heat extremely well.  Their temperature has been measured across more than three orders of magnitude in density and it stays near 10 K, varying by about 20 %.^{6 6} Under that condition the Jeans mass falls as the cloud compresses, and gravity gains on pressure without limit.  A typical giant molecular cloud already holds hundreds to thousands of Jeans masses.

The real problem in the field is the exact opposite of the creationist claim.  If molecular gas simply fell together under gravity, the Milky Way would make roughly 300 solar masses of stars per year.  It makes 1.65 ± 0.19.⁷ Two entire research programmes — magnetic regulation and turbulent regulation — exist to explain why star formation is about a hundred times slower than free collapse.² The mainstream difficulty is that clouds hold themselves up too well, not that they cannot fall in.

The collapse itself has been detected.  In the globule B335, ALMA sees molecular lines in absorption against the central continuum source, redshifted relative to the cloud.  That geometry can only mean gas falling inward on the near side.  The authors call it "unambiguous evidence for infall," and derive an infall rate of 3 × 10⁻⁶ solar masses per year.⁸

The angular momentum objection has been answered by measurement.  Collapsing cores shed spin through jets and disk winds, and the rotation of those jets is now measured directly.^{9 10}

The magnetic objection has been answered by measurement.  Zeeman splitting gives the field strength in clouds directly.  Self-gravitating clouds turn out to be magnetically supercritical — the field is too weak to stop them.  Richard Crutcher and Athol Kemball put it bluntly: "The data strongly suggest that subcritical self-gravitating clouds are the exception and in fact none may exist."¹¹

The blue-star argument is weaker than advertised, and it is not independent.  Lisle writes that "the most massive blue stars cannot last even one million years."⁴ The standard Geneva stellar models give 2.67 million years of hydrogen burning for a 120-solar-mass star — the shortest life in the entire grid — and 3.53 million for a 60-solar-mass star.¹² More important, the argument only works if star formation never happens.  It is a corollary of the first claim, not a second line of evidence.

And there is a clock that needs no stellar theory at all.  Three massive stars — AE Aurigae, μ Columbae and ι Orionis — are flying apart from one another.  Run their measured positions and velocities backwards and they were in the same small volume about 2.5 million years ago, at the site of the Trapezium cluster in Orion.¹³ That is pure geometry.  It requires those stars to have existed 2.5 million years ago, and to have been ejected from a cluster that already existed then.

Where each argument lands

What this article does

It states the young-earth argument in the strongest form its own advocates give it, traces every number to a primary source, and shows the arithmetic in the open so it can be checked.  Where a figure is computed here rather than quoted, the computation is given.

It also reports the places where creationists are right.  Nobody has watched a star form from start to finish.  No Population III star has been confirmed.  Nobody can derive the stellar mass function from first principles.  Those are real gaps and they are stated plainly below, in the words of the people who work on them.

Planet formation is a separate problem with a very different answer, and it has its own article: Do Planets Form By Themselves?

Every source, with what it measured and which way it cuts, is catalogued in a companion page: the literature.

Part 1 — What creationists actually claim

There is no single young-earth position here.  There are three, and they disagree with each other.

Jason Lisle takes the hardest line.  He holds a PhD in astrophysics from the University of Colorado, was Director of Physical Sciences at the Institute for Creation Research when he wrote the article below, and now runs the Biblical Science Institute.  His 2012 article Blue Stars Confirm Recent Creation states the case in four sentences:

"Some astronomy textbooks even have pictures of nebulae labeled as 'star-forming regions' or 'stellar nurseries,' as if star formation were an observed fact.  But it is not.  Star formation has never been observed."⁴

He then gives three physical reasons: gas pressure, magnetic pressure, and rotation.

"In space, there is no container.  So gas expands indefinitely.  If the gas could be forced into a sphere that is very small (in comparison to a nebula) such as the sun, then the gas would be held together by its own gravity.  However, in a typical nebula, the gas pressure far exceeds the miniscule force of gravity."⁴

In a 2025 interview he goes further:

"I'm pretty well convinced that star formation either doesn't happen or it's so rare that it cannot account for the stars that we have in the universe."¹⁴

He also reports a calculation done with the physicist Russell Humphreys: that compressing a typical nebula would take "about four supernovae happening about the same time in the same area."¹⁴ The point is a chicken-and-egg problem.  If it takes four exploding stars to make one new star, the first stars cannot be explained at all.

John Hartnett presses the same case in more technical detail.  A physicist who worked on precision frequency standards, he wrote three long articles for Creation Ministries International between 2015 and 2016.  His argument turns on the Jeans criterion:

"From Eq. (1) it is obvious that as the density of the cloud is increased with collapse the temperature of the cloud is increased (without some rapid cooling method; i.e. radiated away) and the required Jeans mass increases, as does the required Jeans length.  This runs counter to a collapse through the Jeans limit."⁵

He adds a second charge: that computer simulations of star formation cheat.

"In computer simulations of star formation the computer program is usually started with an over-density such that the Jeans mass is already achieved, hence the limit is not a problem because the simulation is started past that point."¹⁵

And a third: that the whole enterprise depends on dark matter, which he calls a "god of the gaps."

"One must invent unknown stuff—dark matter—with the right properties—the unknown 'god of the gaps'—to get stars to form naturalistically.  Without it, it just can't happen!"¹⁵

Danny Faulkner does not agree with either of them.  Faulkner was professor of physics and astronomy at the University of South Carolina Lancaster for 26½ years and is now the astronomer at Answers in Genesis.  His 2014 article Are Stars Still Forming Today? runs through the same physics, agrees that gas clouds are "stable against collapse," but then declines to draw the conclusion:

"Creationists may not have biblical reasons to deny that some physical processes might form stars today, but we reject the idea that present-day processes could explain the ultimate creation of stars."³

And he ends:

"As an astronomer and biblical creationist, do I believe that stars form today?  I'm not sure.  I understand both sides of the biblical arguments.  I don't see that the Bible absolutely precludes star formation today, nor do I see that the Bible demands it.  So I suggest a continuing discussion of this question."³

Faulkner's honesty is worth noting, because the rest of this article shows that his caution was justified and Lisle's confidence was not.

Part 2 — The physics, done carefully

What the Jeans criterion actually says

James Jeans asked in 1902 how big a cloud has to be before its own gravity beats its own pressure.  The answer is the Jeans mass:

M_J ∝ T^(3/2) × ρ^(−1/2)

where T is temperature and ρ is density.  Both Lisle and Hartnett quote this correctly.  The disagreement is about what happens to T while ρ rises.

Two cases:

If the gas cannot get rid of its heat (adiabatic), compression heats it.  For a monatomic gas, T ∝ ρ^2/3, so T^3/2 ∝ ρ, and the Jeans mass goes as ρ^+1/2 — it rises as the cloud shrinks.  Collapse stalls.  This is Hartnett's case, and in this case he is right.

If the gas can radiate its heat away as fast as compression makes it (isothermal), T stays fixed, and the Jeans mass goes as ρ^−1/2 — it falls as the cloud shrinks.  Every further contraction makes the cloud more unstable than it already was.  Collapse runs away.

So the whole question reduces to a single empirical one: do star-forming clouds stay cold while they compress?

They do, and it has been measured

This is not a modelling assumption.  It is a measurement, made with two independent thermometers.

Carbon monoxide lines thermalise at densities around a few thousand molecules per cubic centimetre and give gas temperatures of 10–15 K in regions with no star formation.  Ammonia inversion lines trace much denser gas — above 10,000 per cubic centimetre — and give about 10 K.  Edwin Bergin and Mario Tafalla, reviewing the field in Annual Review of Astronomy and Astrophysics, report that in two Taurus cores mapped in ammonia "the temperature remains constant over the central 0.1 pc with a value of about 10 K."⁶ Across the whole set of cores studied, the temperature differences they find are "relatively small (20 %)."⁶

That is the answer.  Density climbs by a factor of ten thousand or more; temperature moves by twenty per cent.  In some cores it goes down, not up, because dust radiates the heat away faster than the gas can hold it.⁶

Hartnett calls the isothermal treatment "more assumptions and more storytelling."⁵ It is neither.  It is what the thermometers read.

The gas does eventually stop being able to radiate.  That happens when the collapsing core becomes opaque to its own infrared emission, at around 10⁻¹³ grams per cubic centimetre — roughly 10¹⁰ molecules per cubic centimetre.  At that point the collapse does halt, exactly as Hartnett describes, and a "first hydrostatic core" forms.¹⁶ But by then the object is already a protostar in everything but name.  The adiabatic barrier arrives at the end of the story, not the beginning.

The pressure-versus-gravity argument is a unit error

Lisle's version of the objection — and the version in many popular treatments, including a note in this site's own research files — compares the wrong two quantities.

The argument runs: shrink the cloud tenfold, and the volume drops a thousandfold, so the pressure rises a thousandfold; but gravity only rises a hundredfold; so pressure wins.

Pressure is force per unit area.  Gravity is a force.  They cannot be compared directly.  Convert pressure to a force by multiplying by the surface area:

• Pressure at fixed mass and temperature: P ∝ 1/R³
• Area: A ∝ R²
• Pressure force: F_P = P × A ∝ 1/R
• Gravitational force: F_g ∝ GM²/R² ∝ 1/R²

Shrink the cloud tenfold and the pressure force rises ten times while gravity rises a hundred times.  The ratio F_g / F_P goes as 1/R.  Gravity gains without limit as the cloud shrinks.  That is the Jeans instability, stated in forces instead of masses, and it is why the runaway happens.

This correction matters because the erroneous version is what makes the claim feel obvious.  Once the units are fixed, the intuition reverses.

A molecular cloud is already far past the threshold

The Jeans mass can be computed for real, measured cloud conditions.  Using the standard form M_J = (5kT/Gμm_H)^{3/2} (3/4πρ)^{1/2} with a mean molecular weight of 2.33:

A giant molecular cloud has a mass of 10⁴ to 10⁶ solar masses at densities of order 100 per cubic centimetre.  Divide: a 100,000-solar-mass cloud contains about two thousand Jeans masses.  It is not marginally unstable.  It is wildly unstable, by three orders of magnitude in mass.

Hartnett's complaint that a textbook "nowhere explains how such a core of a GMC got that way, i.e. how it exceeded the Jeans mass" has a short answer.⁵ The cloud never had to get that way.  At the temperature and density at which molecular gas exists at all, essentially every giant molecular cloud in the Galaxy is already thousands of times over the limit.  The textbook does not explain how the threshold was crossed because for these objects it was never not crossed.

The mainstream problem is that clouds collapse too slowly

This is the point that reverses the whole argument, and it is over fifty years old.

In 1974 Ben Zuckerman and Neal Evans pointed out that if the Galaxy's molecular clouds were simply falling together, the Milky Way's star formation rate would be enormously higher than what is seen.  The calculation is easy to redo.  The Galaxy holds (1.0 ± 0.3) × 10⁹ solar masses of molecular hydrogen.¹⁷ At a density of 100 molecules per cubic centimetre the free-fall time is 3.1 million years.  Divide one by the other and the Milky Way should be producing about 325 solar masses of new stars per year.

The measured rate is 1.65 ± 0.19 solar masses per year, from a hierarchical Bayesian combination of every published Milky Way measurement.⁷

The observed rate is about two hundred times too low on this set of inputs.  The figure depends on which mean density is used for the free-fall time, and the standard published statement is put in a form that does not: about 1 % of a cloud's mass turns into stars per free-fall time, a hundredfold shortfall.  Krumholz and Tan state the standing result: "It has been known for more than 30 years that star formation in giant molecular clouds (GMCs) is slow, in the sense that only ∼ 1 % of the gas forms stars every free-fall time.  This result is entirely independent of any particular model of molecular cloud lifetime or evolution."²

Read that against Lisle's claim that gas pressure "far exceeds the miniscule force of gravity."  If that were true the observed star formation rate would be zero, and there would be no puzzle to write papers about.  The actual puzzle is that clouds are strongly self-gravitating and still only convert about one per cent of themselves per free-fall time.  The competing explanations — magnetic support, supersonic turbulence, stellar feedback — are all attempts to find something that holds the clouds back.

Neal Evans and colleagues found the same thing inside individual clouds with Spitzer: "Nonetheless, star formation is still slow compared to that expected in a free fall time, even in the dense cores."¹⁸

Magnetic fields: measured, and too weak

Lisle's second objection is magnetic.  He cites Lee Hartmann's textbook for the claim that "the magnetic pressure would halt a shrinking cloud and drive it to re-expand."⁴

The strength of the field is not a matter of opinion.  The Zeeman effect splits spectral lines in proportion to the line-of-sight magnetic field, and it has been measured in hydrogen, hydroxyl and cyanide lines across densities from about 10 to 10¹⁰ particles per cubic centimetre.  The quantity that decides the outcome is the mass-to-flux ratio.  Above a critical value, the field cannot stop gravity.

Crutcher and Kemball's 2019 review of every Zeeman detection reports the result:

"Clouds with N_H ≳ 10²² cm⁻² have a mean M/Φ that is supercritical by a factor of 2–3.  The data strongly suggest that subcritical self-gravitating clouds are the exception and in fact none may exist."¹¹

They also report how the field scales.  Above about 300 particles per cubic centimetre — the density at which clouds become self-gravitating — the field grows as roughly the two-thirds power of density, "which suggest that magnetic fields do not provide significant support at high densities."¹¹

A two-thirds power law is exactly what happens if the field is dragged along by gas that is collapsing in all three dimensions.  It is the signature of a field being carried, not a field pushing back.

There is a genuine caveat, and Crutcher states it: fields vary a lot from cloud to cloud at any given density, so magnetic support may matter in some regions and not others.¹¹ What the data do not support is the flat claim that magnetism prevents collapse.

Angular momentum: the objection was real, and it was answered

This is the strongest of the three classical objections, and creationists cite a real textbook admission for it.  Carroll and Ostlie write: "Conservation of angular momentum arguments lead us to expect that all main-sequence stars ought to be rotating very rapidly, at rates close to breakup.  However, observations show that this is not generally the case."⁵

Hartnett's response is that invoking magnetic fields to fix this is special pleading: "Exclude magnetic fields from the discussion of the initial collapse of the giant molecular cloud to form the star when they are a problem, but include them when you want to solve a different problem."⁵

The charge would stick if the magnetic angular momentum transport were only a theoretical device.  It is not.  It is observed.

Bipolar jets shoot out of forming stars along their rotation axes.  If those jets carry away angular momentum magnetically, they should be rotating — and the rotation should have the same sense as the disk.  Francesca Bacciotti and colleagues used the Hubble Space Telescope's imaging spectrograph on the outflow from DG Tauri and found exactly that: systematic velocity differences of 6–15 km/s across the jet, consistent with rotation.⁹ The measured angular momentum flux was enough to matter for the disk's evolution.

The same signature has been mapped in the HH 212 system with ALMA, where the flattened infalling envelope is seen "spiraling into the central source," the inner region has become a rotationally supported disk, and a jet emerges along the axis.¹⁰ The whole chain — envelope, pseudodisk, Keplerian disk, jet — is resolved in one object.

This is a case where creationist writers should update.  Even a creationist source has noticed:

"Herbig-Haro objects are bipolar jets of material ejected by stars in the early stages of their development.  Astronomers have used Doppler measurements to determine that these jets are rotating.  This rotation transfers angular momentum away from these 'protostars,' and calculations show that this process can effectively transfer enough angular momentum within the timeframe of that particular stage of stellar development.  This provides a plausible solution to one aspect of the angular momentum problem in stellar evolution.  It is important for creationists to stay informed about the work being done in evolutionary astronomy so that we don't mischaracterize opposing views."¹⁹

That is the right attitude, and it is the correct call on the evidence.

Part 3 — What is actually observed

The sequence, and how long each stage lasts

Forming stars are sorted by the shape of their infrared spectrum.  Class 0 objects are buried in an envelope more massive than the star.  Class I have shed most of the envelope.  Class II are visible T Tauri stars with disks.  Class III have lost the disk.

If this were an arbitrary filing scheme, as Hartnett suggests when he calls it "stamp-collecting,"⁵ the classes would have no relation to each other.  They do, and the relation is quantitative.  Counting objects in each class across five nearby molecular clouds and assuming a steady flow through the sequence gives the time spent in each stage.  Evans and colleagues obtained, from the Spitzer Cores to Disks survey:

^{18 18}

Three things follow.

First, this is why nobody has watched a star form.  The shortest stage lasts about a hundred thousand years.  Human astronomy is about four hundred years old.  Expecting to see one star complete the process is like expecting to watch one redwood grow from seed during a lunch break.  Faulkner's observation that "we haven't actually observed star formation in a telescope" is true and is not an argument.³

Second, the class fractions are a consistency test that could have failed.  If the classes were unrelated kinds of object rather than stages, there would be no reason for their relative numbers to reproduce a sensible time ordering, or for the youngest classes to be the ones still buried in dense gas.  Evans found that the youngest objects "are very strongly associated with dense cores."¹⁸ Class 0 sources sit inside dense gas; Class III sources do not.  That is what a sequence looks like.

The count has since been repeated on a second, independent sample.  The Herschel Orion Protostar Survey finds 93 Class 0 and 222 Class I sources; the Spitzer clouds give 63 and 132.  The two datasets yield Class 0 fractions of 0.30 and 0.32, and Class 0 lifetimes of 0.15 and 0.16 million years.²⁰ Two different telescopes, two different regions of sky, the same answer to within seven per cent.

Third, the ages are far longer than 6,000 years but far shorter than the age of the universe.  A Class I protostar half a million years old is a problem for a 6,000-year timescale on its own, quite apart from how it got there.

Infall, detected

The decisive observation is simple in principle.  If gas is falling onto a central object, then gas on the near side is moving away from Earth as it falls inward.  Molecular lines formed in that gas, seen in absorption against the bright central source, must therefore be redshifted relative to the cloud.  Nothing else produces that combination.

ALMA observed the isolated globule B335 and found it.  The paper's own words:

"Observations of the isolated globule B335 with ALMA have yielded absorption features against the continuum that are redshifted from the systemic velocity in both HCN and HCO+ lines.  These features provide unambiguous evidence for infall toward a central luminosity source."⁸

Fitting the line shapes gives an infall radius of 0.012 parsecs, a mass infall rate of 3 × 10⁻⁶ solar masses per year, and an age of about 50,000 years.⁸

Lisle writes that "observations confirm that gas clouds expand; they do not appear to collapse into stars."⁴ The B335 measurement is the direct contradiction of that sentence, and it was published three years after his article.  The claim needs withdrawing.

Disks, at every age

The disk is the fossil of the collapse.  Angular momentum guarantees that infalling gas cannot land on the star directly; it must flatten into a disk first.  So the theory makes a hard prediction: young stars must have disks, and old ones must not.

The prediction holds and the timing is measured.  Karl Haisch, Elizabeth Lada and Charles Lada surveyed six young clusters in the infrared and found the fraction of stars with disks dropping steadily with cluster age, from over 80 % in the youngest to essentially zero by about 6 million years, with a half-life near 3 million.²¹

Two individual objects make the point sharper.  HL Tauri, a Class I–II source in Taurus no more than 1–2 million years old, was imaged by ALMA in 2014 and resolved into a disk with a series of bright and dark rings.²² IRS 63, a Class I protostar in Ophiuchus, shows four annular structures in a disk around a star less than 500,000 years old.²³

A 500,000-year-old star with a structured disk is not compatible with a 6,000-year-old universe under any reading.  And the age does not come from stellar evolution theory; it comes from the object still being wrapped in the envelope it is eating.

Massive stars, and how long they can live

Lisle's blue-star argument depends on a number: how long the most luminous stars can burn.  He states it as under a million years.⁴

The standard reference grid is Ekström and colleagues' Geneva models, computed at solar composition for masses from 0.8 to 120 solar masses, with and without rotation.  Their Table 2 gives the hydrogen-burning lifetime directly:

¹²

The shortest life anywhere in that grid is 2.67 million years, for the most massive star modelled.  Rotation lengthens it.  Lisle's "cannot last even one million years" is short by a factor of about three, and his conclusion that spiral galaxies "cannot be even one million years old" understates the number his own argument produces.⁴

That matters less than it might, because 3 million years is still vastly less than 13.8 billion.  The blue-star argument survives the correction as an argument that something must replenish massive stars.  What it cannot do is establish 6,000 years.  Even taken entirely at face value it gives a few million — about five hundred times too long.

The argument's real weakness is different.  It is not independent evidence.  It is a conditional: if stars never form, then blue stars imply a young universe.  Lisle says as much: "You have to then challenge them and say, well, wait a minute, is star formation something that's likely to happen in this universe?"¹⁴ Everything therefore rests on the first claim, which is the one the measurements contradict.

A clock with no stellar physics in it

Here is a test that uses no model of stellar interiors, no nuclear physics, and no assumption about how stars form.  It uses only positions and velocities.

About 10–30 % of O stars are "runaways" — moving fast, far from any cluster.  If a runaway was flung out of a cluster, its path traced backwards should intersect that cluster at a definite past time.

Ronny Hoogerwerf, Jos de Bruijne and Tim de Zeeuw used Hipparcos astrometry to retrace 56 runaway stars.  Three of them — AE Aurigae, μ Columbae and ι Orionis — converge:

"…the orbits of the runaways AE Aurigae and μ Columbae and of the eccentric binary ι Orionis intersected each other in a very small volume ∼ 2.5 Myr ago, and show that they were ejected from the nascent Trapezium cluster."¹³

Two O-type stars and a massive binary were at the same place 2.5 million years ago.  That place is the Orion Nebula Cluster, which is still there, still full of gas, and still making stars.

The young-earth position can only answer this by saying the stars were created already in flight, on trajectories that falsely converge.  That answer is available.  But it costs the whole apparent-age line of argument: if God can create stars in flight along fake trajectories, He can equally create a universe that looks old in every other respect, and no observation can then count for a young universe either.  A position cannot use apparent youth as evidence while dismissing apparent age as illusion.

Part 4 — The dark matter claim

Hartnett's most striking assertion is that star formation cannot work without dark matter:

"Nowadays, dark matter is added as an essential ingredient to all simulations on star formation…"¹⁵

and

"…so much is needed, such that 85 % of all matter in the solar system needs to be dark matter if the solar system and all other stars formed naturally and were not directly created by the Creator."⁵

Both statements are wrong, and the second is wrong by about ten orders of magnitude.

Dark matter is negligible inside a molecular cloud.  The local dark matter density is about 0.008–0.013 solar masses per cubic parsec, or roughly 0.3 GeV per cubic centimetre.²⁴ Take 0.01 solar masses per cubic parsec, which is 6.8 × 10⁻²⁵ grams per cubic centimetre.  A molecular cloud at 100 hydrogen molecules per cubic centimetre has a density of 4.7 × 10⁻²² grams per cubic centimetre — about 700 times higher.  A dense core at 10,000 per cubic centimetre is 70,000 times higher; a protostellar core at a million is seven million times higher.  Dark matter is not merely unnecessary at these scales.  It is a rounding error.

It also cannot clump on these scales.  Dark matter is collisionless.  It cannot radiate, so it cannot cool, so it cannot lose the kinetic energy needed to settle into a structure a tenth of a parsec across.  With a velocity dispersion of order 200 km/s it flies straight through a molecular core.  This is exactly the property Hartnett quotes from Scientific American — "because the dark-matter particles would not emit radiation or lose energy, they would remain scattered in the primordial cloud"¹⁵ — and it cuts against his conclusion, not for it.  The passage he quotes is describing dark matter staying behind while the gas collapses without it.

The solar system figure is measurable, and it has been measured.  Elena Pitjeva and Nikolay Pitjev used the EPM2011 planetary ephemeris — decades of radar ranging, spacecraft tracking and optical astrometry — to look for the extra perihelion drift that a distributed mass would produce.  Nothing was found.  Their limit from Saturn is a dark matter density below 1.1 × 10⁻²⁰ grams per cubic centimetre, giving a dark matter mass inside Saturn's orbit below 7.1 × 10⁻¹¹ solar masses.²⁵ They note that this is about half the uncertainty on the mass of the asteroid belt.

Hartnett's 85 % would require several solar masses of dark matter in the solar system.  The measured ceiling is under a ten-billionth of one solar mass.

Where the dark matter claim comes from.  Hartnett's sources are correct about their own subject.  The BBC article he quotes is about a cosmological simulation of galaxy formation, where dark matter genuinely does dominate.  The Scientific American article is about Population III star formation in the early universe, where dark-matter minihalos genuinely are the sites where the first gas gathers.  Neither is about present-day star formation in the Milky Way, where dark matter plays no role at all.  Simulations of molecular cloud collapse — the work of Bate, Federrath, Klessen, Padoan, Krumholz and others — do not include a dark matter component, because at those densities there is nothing to include.

This is a specific and recurring error pattern: a statement about one length scale imported into an argument about a completely different one.

Part 5 — Where the creationists are right

An honest audit has to say what survives.  Five things do.

1. No star has been watched forming

True, and unavoidable.  The shortest phase in the sequence lasts about 100,000 years.  Every claim about the process is therefore an inference from a population of objects caught at different stages, plus physics, plus simulation.  That is a weaker kind of evidence than watching one event, and saying so is fair.

What blunts the objection is that the same limitation applies to almost everything astronomical, including things nobody disputes — nobody has watched a star reach the end of its main sequence either.  And the inference here is unusually well constrained, because the stages are not merely ranked but timed, and because infall, rotation, disks and jets are each detected separately.

2. Population III stars have never been confirmed

This is the strongest surviving point, and the mainstream states it in the same words.  Writing in October 2025, Eli Visbal, Ryan Hazlett and Greg Bryan open with:

"Up to this point, there have been no unambiguous direct detections of Population III (Pop III) stars, defined by their extremely low metallicities."²⁶

The first generation of stars had to form from hydrogen and helium alone — no dust, no carbon monoxide, none of the coolants that keep present-day clouds at 10 K.  Molecular hydrogen is a poor coolant, and it is fragile.  Lisle's chicken-and-egg objection lands here with full force, and it is the version of the argument that a specialist would take seriously.

The current best candidate is LAP1-B, a gravitationally lensed system at redshift 6.6 reported in 2025.  Visbal and colleagues argue it matches three theoretical predictions at once — a low-metallicity halo of the right virial temperature, a top-heavy mass function, and a small cluster containing a few thousand solar masses of massive stars.²⁶ That is a genuine result.  It is also a candidate, at redshift 6.6 rather than the redshift 20 or more where most Population III stars are supposed to form, and it has not been confirmed by a second method.

The honest statement: the first stars are predicted, modelled in detail, and unobserved.  Anyone arguing that the whole account of star formation is secure should not include this part of it.

3. Nobody can derive the initial mass function

Stars are not born at random masses.  The distribution is steep, roughly universal, and peaks near a third of a solar mass.  No first-principles derivation of it exists.  Krumholz devotes a third of his 2014 review to the problem and does not solve it.

Related: the star formation efficiency — that stubborn 1 % per free-fall time — is measured, not predicted.  Competing theories reproduce it qualitatively.  None derives it.

4. Protostars are ten times too faint

This one is not usually raised by creationists, and it should be, because it is a straightforward failed prediction.

A protostar's brightness should come mostly from material crashing onto its surface.  Feed a quarter-solar-mass protostar at the rate that gravitational collapse implies, and it should shine at about 25 times the Sun's luminosity.  The observed median is 1.3 solar luminosities in the Spitzer sample and 1.2 in the Herschel sample.

Michael Dunham and colleagues state it plainly in the Protostars and Planets VI review: "Over the past two decades various theories have attempted and failed to explain observed protostellar luminosites, which are found to be generally ∼10 times less luminous than expected."²⁰ Evans's Spitzer survey said its data "confirm and aggravate" the problem.¹⁸

The leading answer is that accretion is episodic — long quiet stretches punctuated by short violent bursts, so that most protostars are caught between meals.  Evans estimated that half a star's mass is accreted during only 7 % of the Class I lifetime.¹⁸ Bursts of the required kind are seen — the FU Orionis outbursts — but whether they occur often enough to close the gap is not settled.

This is a real, unresolved, quantitative discrepancy in the middle of the standard picture.  It does not point toward a young universe, and nobody claims it does.  But an audit that lists the field's problems should list it.

5. Simulations do prescribe their starting conditions

Hartnett's charge is half right, and the half that is right is worth stating.  A simulation that resolves an individual protostar cannot also resolve a galaxy.  Such runs do start from a prescribed turbulent, self-gravitating cloud, and the results depend on that choice.

What makes the charge only half right is that the choice is not free.  The prescribed initial conditions are the observed ones: the measured density, temperature, line width and magnetic field of real clouds.  And galaxy-scale simulations, which do not prescribe clouds at all, form molecular clouds as an output and then form stars inside them.  The scales are stitched together, imperfectly, rather than assumed.

There is also a genuine unsolved problem in this area that creationists have not picked up: the magnetic braking catastrophe.  In ideal magnetohydrodynamics, the field is so efficient at removing angular momentum that no rotationally supported disk forms at all — the opposite of what is seen.  Resolving it requires non-ideal effects, misalignment between field and rotation axis, or turbulence, and the resolution is still argued about.²⁷ Ironically, this is a case where the field is too good at solving the angular momentum problem creationists raise.

Part 6 — The quotations, checked

Popular creationist articles on this subject rest heavily on quotations from mainstream astronomers.  Several of them reverse meaning when the surrounding text is restored.

Neil deGrasse Tyson

The most-quoted sentence in this whole debate is on page 187 of Death by Black Hole:

"The scary part is that if none of us knew in advance that stars exist, frontline research would offer plenty of convincing reasons for why stars could never form."¹

Tyson is writing a chapter on how star formation works.  The paragraph is a rhetorical setup: he lists rotation, turbulence, gas pressure and magnetic fields as the obstacles a cloud must overcome, and two paragraphs later, still on the same page, he begins explaining how it overcomes them:

"If a cloud cools to a low enough temperature (less than about 100 degrees above absolute zero), its constituent atoms will bump and stick rather than careen off one another, as they do at higher temperatures."¹

Quoting the obstacle list while omitting the resolution inverts the author's point.  The sentence is real, it is quoted accurately, and it means the opposite of what it is used to mean.

Mark Krumholz

Also widely quoted:

"Star formation is one of the least understood phenomena in cosmic evolution.  It is difficult to formulate a general theory for star formation in part because of the wide range of physical processes involved… As a result of these complexities, there is nothing like a generally agreed-upon theory of star formation as there is for stellar structure."²⁸

Every word is accurate.  What is missing is what Krumholz means by "a theory."  Two paragraphs later he says the goal is to make star formation "like the theory of stellar evolution: a field that continues to hold unsolved problems, but one for which enough basic results have been established that researchers in other areas of astrophysics can make use of them with some confidence."²⁸ The three questions he wants a theory to answer are named in his title: the star formation rate, stellar clustering, and the initial mass function.  Not whether stars form.  The same review's opening sentence describes star formation as a process that generates radiant energy, synthesises elements, forms planets and leads to life.²⁸

The paper Lisle cites

Lisle's footnote 3, supporting "star formation is problematic at best," points to Wiebe and colleagues, Problems of Star Formation Theory and Prospects of Submillimeter Observations.⁴ The title is promising.  The paper's second sentence is not:

"By the end of twentieth century the overall picture of star formation process in dense interstellar clouds became fully formed.  Contraction (collapse) of protostellar clumps is initiated by gravitation and external pressure; thermal pressure, rotation of the clumps and magnetic field counteract contraction.  Thus, general scenario of star formation is defined by a complex interaction of these factors.  However, many fundamental problems of star formation remain unsolved."²⁹

It is a case for building a space-based submillimetre interferometer.  The "problems" in the title are problems of angular resolution and sensitivity.

The older quotations

Several quotations circulating in creationist literature date from 1981 to 1990 — "if stars did not exist, it would be easy to prove that this is what we expect"; "the origin of stars represents one of the most fundamental unsolved problems of contemporary astrophysics."  These are real and were fair when written.  They predate the Zeeman survey results, the Spitzer and Herschel protostar censuses, the ALMA infall detections, the resolved disks, and the jet rotation measurements.  Quoting a 1990 assessment of the observational state of a field that has since been transformed by three new observatories is not evidence about the field today.

Common objections

"You still cannot show me a star forming."  Correct, and neither can anyone else, for the reason given above: the shortest stage takes about 100,000 years.  What can be shown is the collapse in progress (B335), the disk that the collapse produces (HL Tau, IRS 63), the jet that carries off its spin (DG Tau, HH 212), the population statistics that time the sequence (Evans 2009), and the galaxy-wide rate that results (Licquia & Newman 2015).  Demanding a single continuous observation is a standard that would also rule out plate tectonics, the erosion of the Grand Canyon, and the life cycle of a bristlecone pine.

"Every trigger for collapse requires a pre-existing star, so the first stars are unexplained."  For the first stars this is a real and unresolved problem, and this article says so.  For present-day star formation it is not true: the dominant picture is that supersonic turbulence in the cloud creates the overdensities itself, without any external trigger, and turbulence is generated by the galaxy's own shear, by supernovae, and by the clouds' own gravity.  Triggering by an external shock is one channel, not the required channel.

"Lisle and Humphreys calculated that it takes four supernovae."  The calculation has not been published, so it cannot be checked.  What can be checked is whether real clouds need external compression at all, and the Jeans arithmetic above says they do not: a giant molecular cloud is already a thousand times over the threshold at its observed temperature and density.  A calculation that concludes otherwise is presumably assuming an adiabatic cloud.

"Star formation efficiency is only a few per cent, so most gas never makes stars."  True, and this is the mainstream's own headline problem.  It does not help the young-earth case.  A process that turns 1 % of a cloud into stars per free-fall time still turns the whole Galaxy over many times in 13 billion years, and it produces the 1.65 solar masses a year that are observed.

"None of this proves the universe is old."  It does not by itself, and no single line of evidence does.  What it does is remove the young-earth position's main astronomical argument.  If stars form, then hot blue stars carry no implication about the age of anything, because they are being replaced.  And a Class I protostar 500,000 years old, or three runaway stars that were together 2.5 million years ago, are each independently incompatible with 6,000 years.

"Astronomers are driven by naturalistic assumptions."  Faulkner makes this point and it is fair as far as it goes.³ But the specific claims audited here — cloud temperatures, magnetic field strengths, redshifted absorption lines, jet rotation, stellar lifetimes, runaway star trajectories — are measurements.  Their values do not change with the observer's philosophy.  The argument has to be won or lost on them.

Verdict

Against the argument.

The claim that stars cannot form naturally is not a live scientific position.  Its three physical pillars have each been tested directly and each has failed: clouds stay cold as they compress, their magnetic fields are too weak to hold them, and their spin is carried off by jets that are observed to rotate.  The collapse itself has been detected.  The stages have been counted and timed.  The Galaxy's output has been measured.  And the field's actual open problem is that star formation runs about a hundred times slower than gravity alone would drive it — a problem that only exists because clouds do collapse.

Two of the arguments contain outright errors that should be corrected regardless of one's position.  The pressure-versus-gravity comparison confuses pressure with force.  The claim that dark matter is needed to make stars mistakes a statement about galaxy-scale and early-universe structure for a statement about molecular clouds, and the accompanying figure for the solar system is off by ten orders of magnitude against a published measurement.

The blue-star argument, which is the practical payoff of the whole case, is weaker than its advocates state — the shortest stellar lifetime is about 2.7 million years, not under one million — and it is entirely parasitic on the star formation claim.  When that claim goes, it goes.

Three things survive and deserve to be said clearly.  No one has watched a star form, and no one will.  No Population III star has been confirmed, and the specialists say so in print.  And nobody can derive the stellar mass function or the star formation efficiency from first principles.  A young-earth case built on those three points would be honest.  The case built on gas pressure, magnetism and rotation is not.

Danny Faulkner declined to commit on this question and asked for continuing discussion.  On the evidence gathered here, his caution was the right call, and the answer to his question is yes.

Related articles

• Do Planets Form By Themselves? — the companion piece, with a very different verdict
• The Age of the Solar System — comets, the Moon's recession, Saturn's rings, and Humphreys' magnetic fields

Sources