Lithium has a strange role in astrophysics: it is the heaviest element that had time to form in the first minutes after the Big Bang, and the only one watched by two sciences from opposite ends at once—the nuclear physics of the first 3 minutes and the physics of stellar interiors. In a popular-science piece about the composition of the Solar System, I came across a line tossed in as scenery: the Sun is “poorer in lithium than it should be for its age.” Nothing unusual, a star has processed it away. But that line has a cosmological tail stretching back 40 years, called the lithium problem: the standard Big Bang model calls for roughly 3 times more lithium than we find in the Galaxy’s oldest stars.
Worse still, all the neighboring predictions of that same model—for helium and deuterium—agree with observations almost perfectly. So the error is not in the model. The problem is local, neat, and very stubborn: one figure does not match, and you have to handle it very carefully, because the matching ones are right beside it.
Let’s start with what everything rests on. The Standard Big Bang model (in English, Big Bang nucleosynthesis, BBN) has exactly one free parameter—the ratio of the number of baryons to the number of photons. Today, that ratio is measured not from stars but from the microwave background—from the map of relic radiation made by Planck and WMAP. The baryon density derived from it is known to within fractions of a percent, and it is fed into the nuclear equations of the first minutes. The result is a prediction with no fitting: how much hydrogen, helium-4, deuterium, and lithium-7 the Universe should have baked.
The test is brilliant for 3 of the 4 elements. Deuterium, the most fragile of them, agrees with observations within the errors. Helium does especially well: a fresh review of emission lines gave Y_p = 0.2458 ± 0.0013, meaning that for the first time the observational error has been cut in half and brought level with the theoretical systematic uncertainty.
But lithium-7 from those same equations comes out at around A(Li) ≈ 2.7—in the logarithmic scale where hydrogen is set to 12. In the Galaxy’s oldest stars, it is found roughly 0.4–0.5 dex lower.
That is a factor of 2.5–3.
The fate of attempts to blame nuclear physics is telling, too. Last spring, a detailed BBN “sensitivity atlas” came out: the authors ran their computational scheme over 77 parameters—14 fundamental constants and cosmological quantities plus 63 thermonuclear reactions—and examined how each one affected the output. The conclusion was merciless: shifting lithium-7 by the required amount through nuclear rates would take several coordinated adjustments to key reactions, and existing experiments forbid such adjustments. Within one sigma, the lithium problem has no nuclear solution. The good news from the same work: helium is fine, while deuterium can be brought into line with modest, physically plausible shifts in reaction rates.
Now for where lithium is measured at all. There is a class of stars everyone looks at: warm dwarfs in the Galactic halo—old, low-mass, slowly evolving. It was in them that Monique and François Spite found something astonishing in 1982: lithium abundance barely depends on either stellar temperature or metallicity. A solid horizontal line, which is why it was named the Spite plateau.
The plateau is not merely a pretty regularity. It means that all these stars have retained exactly as much lithium as was present in the cloud from which they formed. The Spites suggested that this was the primordial abundance, inherited from the Big Bang. And since lithium does not survive inside the star—it burns at temperatures above 2.5 million kelvin—the surface abundance can persist only if mixing does not drag matter down into the hot interior. Warm low-mass dwarfs have shallow convective zones, and so their lithium survives.
Just how even this plateau is becomes clear from the numbers. In one careful study from the mid-2000s—41 stars in the 6000–6800 kelvin range—the scatter was only 0.06 dex. 20 years later, in an updated selection, the mean was A(Li) = 2.322 ± 0.013 with a scatter of 0.078 dex. It is one of the cleanest correlations in all of stellar spectroscopy, and that is precisely why it is so hard to explain away as error: it is simply too even.
But even, not where it needs to be. Until 2003, the observed value was considered almost identical to theory, then it emerged that explaining the deuterium ratio required a higher baryon density—and the lithium prediction immediately jumped by half a dex, pulling away from the plateau. Numbers that had agreed came apart because of a refinement in an entirely different measurement.
Why lithium is so unyielding becomes clear from the nuclear chain. The first 3 minutes of the Universe climbed a ladder where every rung was fusion with a proton or neutron, and at mass 5 the ladder ran into an abyss: no stable nucleus with 5 particles exists. Lithium-7 was synthesized by a detour—beryllium-7 accumulated first, then later captured an electron and turned into lithium-7. And that is crucial, because the fate of the “lithium-7 + beryllium-7” pair depended on how many free neutrons were around and how long the era in which they still existed lasted.
Hence the temptation: what if we tweak the history of expansion itself, so that less lithium is born? Almost all exotic scenarios work along those lines, and this year one of them became noticeably more concrete.
A group from Korea and Japan proposed a two-stage version with 2 unstable particles. First, a majoron—a hypothetical particle with a lifetime from 10 seconds to 10,000 seconds—decays predominantly into neutrinos. This slows the conversion of neutrons into protons and shifts the balance so that less lithium and beryllium form. The snag is that deuterium simultaneously becomes more abundant than observations allow. Then the second particle enters the game, an axion-like one with a lifetime from 100,000 seconds, decaying into photons: its light later photodissociates the excess deuterium without undoing the work already done on lithium.
The scheme is elegant, and the authors honestly call it a proof of concept rather than an answer: it has been tailored to the problem. But it shows what shape the answer must take. Lithium cannot be removed on its own—you have to move several elements at once in a coordinated bundle, because deuterium sits beside it and shifts with any intervention. There is simply no free play in this system: it is a fork where every attempt to touch one element echoes in another.
There is another genre of solution—radical cosmology. Last autumn, A&A published a note attempting to extinguish the lithium discrepancy by abandoning the principle on which all BBN physics rests: it assumes that in that era the thermal energy of matter and radiation scaled differently. Formally, it all works.
But the price is immediately visible in the numbers: baryon density in that version must be 10–38 percent, against the accepted 4.9 percent in ΛCDM. In other words, the cost is a direct contradiction with the microwave background—the very thing from which the density is derived. The cure kills the patient more surely than the disease.
For a long time, the lithium problem seemed not only vertical—the plateau level is below prediction—but horizontal as well. Individual stars with the very lowest metallicity—poorer than one-thousandth of the Sun’s—fell out of the plateau: their lithium settled below the general level, while the star SDSS J102915+172927, at a metallicity of −4.9, showed truly extreme depletion despite an utterly ordinary assortment of other elements. This was called the plateau meltdown—its collapse at the metal-poor end.
Then it emerged that this anomaly probably has a dull explanation. Ultra-low-metallicity stars are often carbon-rich: their atmospheres have been enriched with carbon from a companion, and the less lithium such a star has, the more carbon it contains. In other words, they lost their lithium not as primordial witnesses but as descendants of stars with turbulent pasts—they have nothing to do with the original abundance in the young Galaxy. Exclude carbon-rich stars from the sample, and the plateau extends farther down without bending.
Confirmation came in June: a homogeneous analysis of 103 ultra-low-metallicity stars using LAMOST and Subaru data—in this work, the abundances of 12 elements were recalculated in non-local thermodynamic equilibrium—found a slight positive slope in the plateau rather than a bend. If it is real, the question of the plateau’s collapse goes away—but so does the question of whether it represents a pure primordial abundance: a slope means the Galaxy had time to add a little lithium of its own.
At the same time, the second lithium problem, mentioned less often, was disposed of. It concerned the isotope lithium-6: its primordial abundance should be tiny—on the order of 2 ten-thousandths of a percent of lithium-7—yet observations long found much more of it, spawning a separate story about exotic sources. It turned out the isotope is not present in old stars at all. Lithium-6 was sought with the high-resolution infrared spectrograph ESPRESSO and found in none of the observed stars; what had previously been taken for its signature is produced by convective asymmetry in the line—the boundary between modeling and artifact.
If nuclear physics is not to blame and the first minutes have nothing to do with it, the other end remains: the stars eat the lithium themselves. This is now considered the main direction, and a major review bringing the picture together came out in February.
It yields an important negative result. Only a few years ago, a popular version held that lithium in the Galactic halo had been destroyed by the very first, massive stars, which scrubbed the interstellar medium clean before the observed dwarfs formed. But the plateau turned out to have a universality that kills this version: the same lithium level has been found in Omega Centauri—the stripped core of an absorbed dwarf galaxy—in the debris of Gaia-Ophiuchus, which collided with the Milky Way around 10 billion years ago, and in the separate S2 stellar stream. Different galaxies with different histories of star formation show the same value. No local devourer of the interstellar medium can produce that kind of universality—and such a scenario would also have generated an excess of carbon, nitrogen, and oxygen, which is not observed in halo stars.
So the mechanism must be internal and must work the same way in every star. And here begins the territory where everything rests on rotation. Lithium does not make its own way into the hot zone in stellar atmospheres—it is slowly dragged there by mixing, and mixing comes from how a star loses rotation through its magnetic wind. The more strongly the outer layers are braked, the steeper the rotational-velocity gradient between core and surface, the stronger the circulation and turbulence, the deeper the lithium disappears.
Hence the famous lithium dip—a narrow gap in the temperature distribution around 6700 kelvin, where lithium is especially scarce. And here the literature contradicts itself in the details: some versions of the models reproduce the hot edge of this dip well and overdo depletion at the cool edge, while others do exactly the reverse.
A careful way around this fork—artificial mixing tuned to suppress lithium settling—predicts a loss of roughly 0.2 to 0.5 dex from the initial abundance. A shift of exactly the required scale. But the parameters of this mixing are not derived from first principles; they are fitted, and the theory pays a practical price for that: all such fitted models require stars to have a great deal of mixing, which in turn ought to spin up the interiors. Helioseismology, meanwhile, shows that the Sun’s interior rotates almost as a single body, without the expected acceleration of the core. The fit to lithium runs against data that lithium does not have.
There remains one final test that could be made—and that is where things do not come together. If stars really destroy lithium, the reverse effect should be visible: where lithium has gone missing, there should be more of it in the interstellar medium. There is no precise measurement of that medium, and a conference talk last year included an honest caveat: stellar depletion is the likely answer, but if current estimates of lithium abundance in interstellar gas hold up, they will rule out that version again. The circle closes around a single element, right on the boundary of what can be measured.
For 40 years, the lithium problem has looked like a puzzle with no solution because every solution is already taken somewhere else. Nuclear rates have been checked experimentally and will not move. Exotic physics requires fitting the answer and echoes in deuterium or in baryon densities from the microwave background. Stellar depletion works numerically, but rests on fitted parameters and runs into helioseismology. The most likely position today is the stars, but it cannot be confirmed until a reliable, independent measurement of lithium in the interstellar medium appears—and no such measurement exists yet.
That is the strangeness of the situation. Cosmology has several predictions that can be tested directly in the laboratory with a precision unreachable for most astrophysical conclusions—and one of them disagrees with the sky so stubbornly that, instead of an answer, a list of candidates has accumulated, each known in advance to be incomplete. Lithium remains the only place where the first minutes of the Universe and the life of an individual star cannot divide a single line between them.