String theory predicts heavy scalar fields called moduli, which set the size and shape of the hidden extra dimensions. After inflation these fields oscillate and briefly dominate the universe. When they decay, a few seconds before the light elements form, they reheat the universe to only a few MeV. This is attractive for dark matter, because the decay yields about the right abundance almost independently of earlier history. It is harmful for ordinary matter, because the decay dumps so much entropy that any earlier matter–antimatter imbalance is diluted to nothing. The escape route is to have the decay itself create both the baryon excess and the dark matter. Several explicit models do this, but each one reaches the observed values only for particular parameter choices. None of them tells us whether any model could do better, or whether the observed universe sits near a hard ceiling.
We propose to answer that question the way information theorists answer "how good can a code be." We will prove an outer bound, the most baryon and dark-matter asymmetry that any decay chain can produce given the modulus mass and its gravitational-strength couplings. We will then build the best explicit models, which give an inner bound, and report the multiplicative gap between the two. In coded caching, the outer bound is called a converse bound and the ratio is called the optimality gap. The deliverable is a "yield region" for moduli co-genesis, meaning the set of jointly achievable baryon and dark-matter yields, with a certified gap. It also gives a falsifiable consequence: a minimum modulus mass below which co-genesis is impossible for every model, not only the ones already written down. String geometry enters as the supplier of the decay rates and branching fractions that feed the bound.
Core Claim
The strongest explicit result is Confronting the moduli-induced lightest-superpartner problem. It shows that when a modulus of about 100 TeV decays into a hidden sector carrying its own asymmetry, the result can account for 100% of the dark matter. Symmetric hidden sectors give less than 10%, and hidden non-Abelian sectors give a few percent. The authors say the asymmetric case needs a "fortuitous conspiracy" of parameters. Our claim is that this conspiracy can be quantified. Whether the conspiracy is fine-tuning or near-necessity depends on how close the observed yields sit to the universal outer bound. If they sit close, every viable model is pushed to the same corner, and the coincidence becomes a prediction. If they sit far inside, then the explicit models are simply suboptimal, and better constructions exist.
The skeleton of the bound is standard cosmology. After the modulus $\phi$ of mass $m_\phi$ decays and reheats the universe to temperature $T_{RH}$, the number of decaying quanta per unit entropy is
$$Y_\phi = \frac{3\,T_{RH}}{4\,m_\phi}, \qquad Y_B \le \epsilon_{\rm eff}\, \mathrm{BR}_{\rm vis}\, \kappa\, Y_\phi ,$$
where $Y_B$ is the final baryon number per unit entropy. The factor $\epsilon_{\rm eff}$ is the net CP-violating asymmetry generated per decay chain, $\mathrm{BR}_{\rm vis}$ is the fraction of decays that reach the charge-carrying states, and $\kappa \le 1$ is the survival fraction after washout. A parallel expression holds for the dark-matter asymmetry $Y_\chi$.
The first relation follows from energy conservation at decay. For gravitational couplings, $T_{RH}$ grows as $m_\phi^{3/2}$, so $Y_\phi$ scales as $(m_\phi/M_{\rm Pl})^{1/2}$. The ceiling therefore depends on the modulus mass alone, once the other factors are bounded.
The research content lies in bounding $\epsilon_{\rm eff}$, $\mathrm{BR}_{\rm vis}$ and $\kappa$ jointly rather than separately. The main pieces are:
- a unitarity cap on CP asymmetries in multi-step chains;
- the requirement that intermediate states decay before nucleosynthesis while also surviving long enough to be out of equilibrium;
- the fact that a large branching fraction into a hidden sector also produces a symmetric dark-matter component, which drives annihilation to the fixed point of Nonthermal "WIMP miracle" and removes charge.
These coupled constraints are where the naive product bound should be loose, and where a tighter converse is worth proving.
Research Program
The first strand is the converse. We model the decay as a directed graph running from the modulus through intermediate states to the visible and dark sectors. Each edge carries a branching fraction and a CP phase, and each node carries a lifetime window. We then bound the maximal asymmetry flux through this graph. The information-theory literature supplies methodology rather than theorems. We are not claiming that coded-caching entropy inequalities apply to Boltzmann equations. What carries over is the discipline of proving outer bounds that account for coupling between cuts, and of reporting a multiplicative gap. In coded caching, naive cut-set bounds left gaps of 12 in Fundamental Limits of Caching. A New Converse Bound for Coded Caching reduced this to at most 4.7. It did so partly by treating the achievable quantity as a linear-fractional function and exploiting its quasiconvexity. Characterizing the Rate-Memory Tradeoff in Cache Networks Within a Factor of 2 reached a factor of 2.
Our dark-matter-to-baryon ratio $\Omega_{\rm DM}/\Omega_B$ is exactly such a ratio of yields weighted by masses. We will use the same quasiconvexity argument to optimize it over the edge parameters. The result is a two-dimensional outer region in the $(Y_B, Y_\chi)$ plane for each $m_\phi$, rather than a single number.
The second strand is achievability: explicit models that push toward the boundary. We will implement three constructions and solve their Boltzmann equations through the modulus-dominated era.
- Asymmetric hidden sector. This is the asymmetric $U(1)_x$ sector of Confronting the moduli-induced lightest-superpartner problem, taken as published.
- Assimilation with a dynamical asymmetry. In the assimilation mechanism of Dark matter assimilation into the baryon asymmetry, singlet dark matter is destroyed by scattering off heavy states $\Psi$ that carry the baryon asymmetry. That paper simply assumes the initial $\Psi$ asymmetry. We will instead generate it from CP-violating modulus decay, which closes a gap the authors leave open.
- Two-sector leptogenesis during modulus domination. This is the leptogenesis of Light dark matter from leptogenesis, in which heavy sterile neutrinos decay into both the visible and dark sectors. Here the sterile neutrinos will be produced by modulus decay during the modulus-dominated era, rather than thermally.
The third strand is geometry. The branching fractions in the bound are not free parameters in a string compactification. Whether the modulus decays mostly into the visible sector or a hidden sector depends on where those sectors live in the extra dimensions: on the large bulk volume or on small localized cycles. We will take the Large Volume Scenario geometries catalogued in Necessary conditions on Calabi-Yau manifolds for large volume vacua. That scan found 418 manifolds with a single large four-cycle among 4,434 toric hypersurfaces, and 50 ($h^{1,1}=3$) plus 106 ($h^{1,1}=4$) of them admit the stricter "Strong Cheese" form. Where their intersection data fix the ratios of coupling strengths between bulk and localized cycles, we will map each geometry to a point on the $m_\phi$ axis with bounded branching fractions. We then ask what fraction of these geometries lies in the region where the outer bound permits the observed yields.
The $G_2$ construction of G2-MSSM serves as a second, M-theory data point. It predicts 10–100 TeV moduli and wino dark matter but leaves the baryon asymmetry unaddressed.
Baselines
The naive reference point is decoupled high-scale baryogenesis followed by modulus dilution. The asymmetry is set early and then divided by the entropy released at decay. Nonthermal "WIMP miracle" flags this as the unresolved problem, and it should fail by orders of magnitude across the 10–100 TeV window. We will compute that dilution factor with the same Boltzmann code, the same modulus decay widths and the same $T_{RH}$ used for all other models. It anchors the bottom of the yield plane.
The strongest existing method is the asymmetric hidden $U(1)_x$ model of Confronting the moduli-induced lightest-superpartner problem. Its reported result is that it accounts for 100% of the observed dark matter density. It is the only published moduli-decay model that does so while evading indirect-detection limits. The file lists no baryon yield for this model: no reported result found on Lacuna for the co-generated baryon yield in the asymmetric $U(1)_x$ scenario.
We will reproduce that model and scan its parameters with the same grid resolution and the same constraints applied to our own constructions. The constraints are nucleosynthesis timing, dark radiation limits, and the requirement $\Omega h^2 \approx 0.1$. We will record the largest jointly achievable $(Y_B, Y_\chi)$ as a function of $m_\phi$. An explicit construction "beats" it if it reaches the observed ratio $\Omega_{\rm DM} \approx 5\,\Omega_B$ over a strictly larger fraction of scanned parameter volume at equal scan budget, or at a lower modulus mass.
For the assimilation model the reported benchmark is $\Omega_{\rm DM}h^2 \simeq 0.1$ for dark-matter masses of 100–400 GeV with TeV-scale $\Psi$ Dark matter assimilation into the baryon asymmetry. Our dynamical version must recover this band once the $\Psi$ asymmetry is generated rather than assumed.
The ablation that isolates our key ingredient is the one-line product bound. It is the displayed inequality with each factor maximized independently: $\epsilon_{\rm eff}$ at its unitarity limit, $\mathrm{BR}_{\rm vis}=1$ and $\kappa=1$. The refined converse, which couples the factors, counts as a contribution only if it lies measurably below this naive ceiling. Our threshold is a factor of 2 or more somewhere in the 10–100 TeV window. It also counts if it changes the qualitative conclusion, namely the minimum viable modulus mass.
Why not just…?
Why not just scan the explicit Boltzmann models, as Blinov et al. already did? A scan says what a given model family can reach, not what any model can reach. So a scan alone cannot distinguish a fine-tuned coincidence from a near-universal ceiling. This alternative is a real threat, because if the best explicit models already sit near the naive bound, the scan is the whole story. The ablation settles it. If the gap between the best inner model and the one-line bound is already below about 2 everywhere, we will report that and treat the contribution as the achievability results and the geometric map, not the converse.
Why not just use the one-line product bound and skip the converse machinery? The one-line bound ignores that pushing more decays into a hidden sector also creates a symmetric dark component, which annihilates and removes charge. It also ignores that CP phases large enough to matter force couplings that shorten lifetimes toward equilibrium. In caching, exactly this kind of neglected coupling between cuts inflated the gap from roughly 4.7 to 12 A New Converse Bound for Coded Caching. Whether the same happens here is an empirical question, and the ablation in the Baselines section answers it directly.
Why not just generate the baryon asymmetry after the modulus decays? At a reheating temperature of a few MeV, the electroweak processes that convert lepton number into baryon number are long inactive. Any post-decay mechanism must therefore violate baryon number directly at low energy, which runs into proton-decay and meson constraints. The assimilation paper already struggles to fit its late-decay window Dark matter assimilation into the baryon asymmetry. The converse covers this route automatically, because it applies to any decay chain whose last step precedes nucleosynthesis.
Why not just import coded-caching entropy inequalities wholesale? Han's inequality and entropy sub-modularity, as used in Improved Approximation of Storage-Rate Tradeoff for Caching with Multiple Demands, bound Shannon entropies of random variables. Baryon number is a conserved charge, not a Shannon entropy. We borrow the proof strategy: coupled cuts, quasiconvex optimization of ratios, and multiplicative-gap reporting. We do not borrow the inequalities themselves, and we will not claim otherwise.
Proposed Experiments
The first experiment computes the naive and refined outer bounds on $Y_B$ and on the $(Y_B, Y_\chi)$ region across modulus masses of 10–100 TeV, the window singled out by Nonthermal "WIMP miracle" and Non-thermal dark matter and the moduli problem in string frameworks. We overlay the observed baryon yield of roughly $10^{-10}$ and the relic density $\Omega h^2 \approx 0.1$. The decisive output is the modulus mass below which the refined bound falls under the observed baryon yield. If that crossing lies inside the window, co-genesis becomes a prediction of the minimum modulus mass for every such model.
The second experiment runs the three inner-bound constructions and the decoupled baseline with identical constraints. We report the multiplicative gap to the refined outer bound at each mass, together with the fraction of parameter volume reproducing $\Omega_{\rm DM} \approx 5\,\Omega_B$. Success means one of two things: a gap within a small constant (we target at most about 3) across the window, or a demonstration that no construction closes a larger gap. The latter would identify where new model-building is needed.
The third experiment maps the 418 Swiss-cheese geometries and the 156 Strong Cheese cases onto the bound. We keep those whose intersection data constrain the branching fractions, and report the fraction that permits the observed yields. If that fraction is near zero for Large Volume Scenario geometries but not for the $G_2$ case, that is a sharp geometric discriminator. If it is substantial, the "fortuitous conspiracy" is not fortuitous: it is what these geometries generically produce.
References
- Confronting the moduli-induced lightest-superpartner problem
- Nonthermal “WIMP miracle”
- Fundamental Limits of Caching
- A new converse bound for coded caching
- Characterizing the rate-memory tradeoff in cache networks within a factor of 2
- Dark matter assimilation into the baryon asymmetry
- Light dark matter from leptogenesis
- Necessary conditions on Calabi-Yau manifolds for large volume vacua
- <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:msub><mml:mi>G</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math>-MSSM: An<mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline"><mml:mi>M</mml:mi></mml:math>theory motivated model of particle physics
- Improved Approximation of Storage-Rate Tradeoff for Caching With Multiple Demands
- Non-thermal dark matter and the moduli problem in string frameworks