Sound-horizon reduction (early dark energy, ΔN_eff, varying mₑ) versus ACT DR6, SPT-3G and DESI DR2
Harder than the ladder half: ACT DR6 and SPT-3G found no preference for EDE, varying constants or modified recombination, so success needs a new model class rather than more data; it sits just above the old Hubble tension (50) and below primordial B-modes (55). Leverage 50 matches the old Hubble tension because a confirmed pre-recombination modification would reshape early-universe cosmology, but a firm exclusion only adds bounds. Application value 7, as for the cosmological constant problem.
The problem
Can a change to the early universe fix the Hubble tension without breaking any other measurement?
- 1CMB measures angles, needs a ruler length
- 2Shorter ruler would raise inferred H₀
- 3Add early energy, particles or new constants
- 4Must still fit all other data
- 5Newest data disfavour most such models
The idea. The CMB and baryon acoustic oscillations (BAO) measure angles. To turn an angle into H₀, they need the true length of a standard ruler, the comoving sound horizon r_d: how far sound waves traveled in the photon–baryon plasma before the baryons decoupled from light. The CMB gives angular sizes very precisely, so if r_d were smaller by 5 to 7%, the inferred H₀ would be larger, about 72 to 73 km/s/Mpc, matching the local ladder.
Built from
earlier problems → this oneThe problems this one is built from: progress or tools it needs (builds on), special cases that come first (generalizes), and weaker results on the way (implies). On the right, what builds on it.
- Start with
- Then
- Sound-horizon reduction (early dark energy, ΔN_eff, varying mₑ) versus ACT DR6, SPT-3G and DESI DR2This problem
Path to understanding
foundations → problemThe research entries this problem rests on, in the order to read them. Each links to its explainer.
- Start with
- Step 2
- Step 3
- ΛCDM, the standard model of cosmologyTheory · host
- Then
- Sound-horizon reduction (early dark energy, ΔN_eff, varying mₑ) versus ACT DR6, SPT-3G and DESI DR2The problem
State of the art
Early dark energy (Poulin et al., arXiv:1811.04083) fit Planck with H₀ ≈ 71–72, ACT DR4-era analyses showed mild preference (Hill et al., arXiv:2109.04451), but ACT DR6 found no preference for early dark energy, varying constants or modified recombination, and N_eff = 2.86 ± 0.13 (arXiv:2503.14454), and profile-likelihood analyses indicate that much of the earlier posterior weight is a prior-volume effect. SPT-3G D1 alone gives H₀ = 66.66 ± 0.60, and the CMB-SPA combination of SPT-3G, ACT and Planck gives 67.19 ± 0.38 in ΛCDM, with a 2.8σ CMB–DESI DR2 BAO discrepancy and mild preferences for modified recombination or a varying electron mass in a non-flat universe when BAO are added (arXiv:2506.20707, June 2025). ACT DR6 with Planck and DESI DR2 BAO gives 68.43 ± 0.27 (arXiv:2503.14452). DESI DR2 BAO (arXiv:2503.14738, March 2025) pins the late-time expansion and the product H₀·r_d. Typical EDE fits raise the matter density and spectral index and worsen the fit to weak-lensing S₈ and to DESI BAO. The June 2026 decade review (Cai & Wang, arXiv:2606.20434) concludes that most standard fixes are tightly constrained and considers interacting dark energy as an alternative. No model class is accepted as resolving the tension. SPT-3G D1 and ACT DR6 (2025) are the latest CMB releases as of October 2026: SPT-3G Ext-10k exists only as forecasts (arXiv:2510.24669) and the Simons Observatory had published commissioning papers but no cosmology.
Realistic applications
A confirmed sound-horizon modification would be new physics before recombination (a light scalar sector or extra radiation), constraining axion-like and dark-sector model building. A firm exclusion converts into bounds on f_EDE, ΔN_eff and mₑ variation. The inference methods (profile likelihoods, emulators) are reusable across cosmological extensions. There is no near-term practical application.
What counts as solved
A model class is excluded when profile-likelihood and Bayesian fits to the full CMB+BAO+SN+lensing data bound f_EDE (or its analogue) below the value needed for H₀ > 71 at ≥3σ in two independent pipelines. A model is a candidate resolution when it fits all those data, including S₈, with a likelihood improvement over ΛCDM of at least 3σ after the extra parameters and has a physical, technically natural origin.
How it gets verified
ReproductionA training run, experiment or computation rerun by someone else.
A model class is excluded when profile-likelihood and Bayesian fits to the full CMB, BAO, SN and lensing data bound its early-dark-energy fraction (or analogue) below the value needed for H0 > 71 at 3 sigma or more in two independent pipelines. The spec must name the datasets and likelihoods (including ACT DR6, SPT-3G and DESI DR2), the samplers, convergence criteria and the chi-square improvement over LCDM that makes a model a candidate. Fits run on public data, so anyone can rerun them.
Claims under verification
+ add a claimNo claimed results on the Verification board for this problem yet.
Where the work is
Mostly model-building and Bayesian inference. Compute is Boltzmann-code runs and emulators. The empirical share is the next CMB and BAO data.
Bottlenecks
ranked by votes- #1TheoryNo technically natural pre-recombination model
EDE needs a tuned potential and a coincidence of its decay with equality, and no UV-motivated realisation fits all data. Without one the result is a fit, not an explanation.
Community grades
Sign in to add your own grade.
Connected problems
Progress
- Jun 18, 2026Hubble tension decade reviewsource
Cai & Wang review: standard fixes tightly constrained, tension unresolved.
- Jun 25, 2025SPT-3G D1 and CMB-SPAsource
Combined SPT, ACT and Planck CMB gives H₀ = 67.19 ± 0.38 in ΛCDM.
- Mar 18, 2025ACT DR6 and DESI DR2 releasesource
ACT DR6 finds no preference for EDE; DESI DR2 BAO tightens late-time expansion.
- Nov 9, 2018Early dark energy proposedsource
Poulin et al. show an EDE component fits Planck with higher H₀.