The Primordial Lithium Problem

Summary

In 1982 François and Monique Spite found that the oldest, most metal-poor stars in the galactic halo all show about the same amount of lithium, regardless of how metal-poor they are. A flat line like that looks like a primordial value: the lithium the universe made, before stars began changing the inventory.

Standard Big Bang nucleosynthesis, run at the baryon density the Planck satellite measured, predicts about three times that much. The same calculation gets deuterium and helium right, and gets them right tightly. Three recent calculations, from two groups using different codes, put the prediction between about 4.7 and 5.6 parts in ten billion. A 2020 analysis calls the problem "more acute", and the nuclear escape route has been narrowed by measurement.

What is documented

The plateau. Spite and Spite reported in 1982 a near-constant lithium abundance across halo dwarfs of widely varying metallicity, deducing a primordial value of lithium to hydrogen of about 1.12 parts in ten billion. The benchmark modern measurement, from 28 halo dwarfs observed with the Very Large Telescope, puts the plateau at an abundance of 2.199 plus or minus 0.086 on the astronomers’ logarithmic scale, which is about 1.58 parts in ten billion.

The prediction. Three recent calculations at the Planck baryon density give lithium-7 to hydrogen as 4.68 plus or minus 0.67 parts in ten billion (2016), 5.623 plus or minus 0.247 (2018), and 4.72 plus or minus 0.72 (2020). The 2016 and 2020 values come from one group, sharing most of their authors; the 2018 value comes from a second group using a different code.

The size of the gap. The papers state it as a factor. "About a factor of three", says the 2016 review. "A factor of 3.6", says the 2018 one. A 2011 review gives the range as three to four, and states the significance as 4.2 to 5.3 sigma. The 2016 review describes the predicted and observed likelihoods as "essentially disjoint".

The rest of the calculation works. The 2016 calculation predicts deuterium to hydrogen at 2.58 plus or minus 0.13 parts in a hundred thousand, and a helium mass fraction of 0.24709 plus or minus 0.00025, both consistent with observation. The 2020 paper notes the deuterium constraints are now very tight, and says for that reason "the lithium problem remains, and indeed is more acute".

The nuclear route has been tested. In 2016 the n_TOF collaboration at CERN measured the beryllium-7 neutron-alpha reaction across a wide energy range. The result hints "at a minor role of this reaction in BBN, leaving the long-standing cosmological lithium problem unsolved". Separately, the 2011 review notes that lowering the helium-3 alpha capture rate by the factor needed is ruled out by agreement with solar neutrino measurements. What survives is a set of unmeasured resonances, which the 2011 review describes as experimentally accessible.

The plateau is not perfectly flat. The 2010 survey reports a meltdown below a metallicity of minus three: the plateau stops holding, the scatter roughly doubles toward lower metallicity, and a positive correlation appears between lithium and metallicity at two to three sigma. The 2011 review notes that the error budget on the measured abundance is dominated by systematics, principally the stellar temperature scale, rather than by statistics.

The baryon density that drives the prediction. From Planck, a baryon density parameter of 0.0224 plus or minus 0.0001. The prediction depends on it, and it is now measured precisely.

Leading explanations

The stars destroyed it (Korn and colleagues, 2006; hypothesis, untested). That atomic diffusion and settling in the thin convective envelopes of old metal-poor dwarfs removed lithium from the visible surface, so the plateau understates the gas these stars formed from, and nucleosynthesis is untouched. If it holds, it would dissolve the problem without changing any cosmology. The authors hedge in their own title, calling it a probable solution, and print no corrected primordial value. The 2011 review calls the astrophysical route "not ruled out, but strongly constrained", and warns that "it is entirely possible that the lithium problem cannot be resolved astrophysically".

A wrong or missing nuclear reaction rate (rejected for the measured channels; open for the rest). Either a standard cross section is mismeasured or an unmeasured resonance opens a channel that destroys beryllium-7, which is what most primordial lithium-7 is made from. One candidate reaction, beryllium-7 with a neutron giving two alpha particles, was measured in 2016 and found to play a minor role. The remaining resonances have not been measured.

New physics during or after nucleosynthesis (hypothesis, untested and penalised). Late-decaying or annihilating relic particles, bound-state effects, varying constants or a non-standard expansion history could reduce the lithium yield. The 2016 review says the mismatch is "perhaps pointing to new physics". The difficulty is that fixes which buy lithium cost deuterium, and deuterium is now measured tightly enough that the trade has got worse rather than better. The 2011 review reports that decaying-particle fits require fine tuning and reach only marginal significance.

The plateau is galactic, not primordial (Miranda, 2025; hypothesis, untested). That the plateau is produced by early stellar yields and galactic chemical evolution rather than by nucleosynthesis directly, so there is no discrepancy to explain. A single recent paper, whose stated aim is to show exactly that.

Current status

Unsolved. The conflict is stable, and a 2020 analysis calls it "more acute". Three recent calculations from two groups put the prediction between 4.68 and 5.623 parts in ten billion against a measured plateau of about 1.6, a factor of roughly three; a 2011 review puts the significance at four to five sigma. Two nuclear fixes have been narrowed by experiment and observation: one candidate reaction was measured and found to play a minor role, and the alternative rate fix is blocked by agreement with solar neutrino measurements. New-physics fixes buy lithium at the cost of deuterium, which is now measured tightly enough to make that trade worse. Stellar depletion would leave the cosmology untouched, and the observed meltdown at the lowest metallicities is consistent with it, but no paper the Registry retrieved claims that case is closed. Four things would change this record: a stellar-structure calculation reproducing both the plateau and the meltdown from a corrected initial abundance matching the prediction; a laboratory measurement of the remaining beryllium-7 destruction resonances; a lithium measurement in pristine gas rather than in a stellar atmosphere, which would decide between the stars destroying it and the early universe overproducing it; or a firm, uncontested lithium-6 detection, which would make the anomaly harder to pin on depletion alone.

Last reviewed: November 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.