[论文解读] Review of Hubble tension solutions with new SH0ES and SPT-3G data
本文利用最新的SH0ES和SPT-3G数据,评估了11种宇宙学模型以缓解哈勃张力问题,采用主动学习模拟器加速似然评估。研究发现,时变电子质量模型和Majoron模型在超过3σ的显著性水平下被排除,而早期暗能量和空间曲率模型将张力降低至1.0–2.9σ之间,但均未达到足以取代ΛCDM的统计显著性。
We present an updated analysis of eleven cosmological models that may help reduce the Hubble tension, which now reaches the $6σ$ level when considering the latest SH0ES measurement versus recent CMB and BAO data, assuming $Λ$CDM. Specifically, we look at five classical extensions of $Λ$CDM (with massive neutrinos, spatial curvature, free-streaming or self-interacting relativistic relics, or dynamical dark energy) and six elaborate models featuring either a time-varying electron mass, early dark energy or some non-trivial interactions in the neutrino sector triggered by a light Majoron. We improve over previous works in several ways. We include the latest data from the South Pole Telescope as well as the most recent measurement of the Hubble rate by the SH0ES collaboration. We treat the summed neutrino mass as a free parameter in most of our models, which reveals interesting degeneracies and constraints. We define additional metrics to assess the potential of a model to reduce or even solve the Hubble tension. We validate an emulator that uses active learning to train itself during each parameter inference run for any arbitrary model. We find that the time-varying electron mass and the Majoron models are now ruled out at more than $3σ$. Models with a time-varying electron mass and spatial curvature or with early dark energy reduce the tension to $1.0-2.9σ$. Nevertheless, none of the models considered in this work is favored with enough statistical significance to become the next concordance model of Cosmology.
研究动机与目标
- 评估ΛCDM的扩展是否能解决SH0ES与CMB/BAO测量之间6σ的哈勃张力。
- 将最新的SH0ES H₀ = 73.29 ± 0.90 km/s/Mpc和SPT-3G CMB数据纳入综合模型比较。
- 将总中微子质量视为自由参数,以揭示新的退化关系和约束条件。
- 开发并验证一种主动学习模拟器,以实现跨多样化宇宙学模型的高效似然评估。
- 提出新指标,以评估模型在统计显著性之外缓解哈勃张力的潜力。
提出的方法
- 采用贝叶斯框架结合MCMC采样,推断11种宇宙学模型(包括ΛCDM扩展和非标准模型)的后验分布。
- 使用CLASS玻尔兹曼代码计算CMB功率谱,并在每次MCMC运行期间实时训练主动学习模拟器,以降低计算成本。
- 每个链路的模拟器训练使用400–1700次CLASS调用,共执行60,000–90,000次评估,将完整模拟的计算量减少约99%。
- 采用BOBYQA算法并结合多次随机初始化以最小化χ²,利用模拟器平滑数值噪声。
- 通过对比使用和不使用模拟器时的后验等高线,验证模拟器的准确性,即使在非高斯后验情况下也表现出极佳的一致性。
- 引入新指标——DMAP和AIC,以评估模型在缓解哈勃张力方面的性能,超越标准贝叶斯证据。
![Figure 1 : A graphical representation of the first four columns of Table 6 . Left : The black dots and their intervals correspond to the mean value of $H_{0}$ and its 68%CL, respectively, for each model. The gold band is the 68% CL on $H_{0}$ from Planck [ 9 ] , while the light-salmon one correspond](https://ar5iv.labs.arxiv.org/html/2312.09814/assets/x1.png)
实验结果
研究问题
- RQ1在扩展的ΛCDM模型中,引入新的SH0ES和SPT-3G数据是否能降低哈勃张力的显著性?
- RQ2哪些宇宙学模型最有效地缓解哈勃张力,其缓解程度如何?
- RQ3主动学习模拟器在近似多样化模型(包括具有非高斯后验的模型)的CMB功率谱方面有多准确?
- RQ4基于当前数据,具有时变电子质量或轻量Majoron的模型是否是哈勃张力的可行解决方案?
- RQ5任何模型缓解哈勃张力的能力具有多高的统计显著性?是否存在可被视为下一任共识模型的模型?
主要发现
- 时变电子质量模型因与CMB和BAO数据存在强烈矛盾,在超过3σ的显著性水平下被排除。
- 涉及轻量Majoron和中微子区相互作用的Majoron模型同样在超过3σ的显著性水平下被排除。
- 将早期暗能量或空间曲率与时变电子质量结合的模型,将哈勃张力降低至1.0–2.9σ之间。
- 尽管有所缓解,但没有任何模型达到足够高的统计显著性,无法被视为宇宙学的下一任共识模型。
- 主动学习模拟器将计算成本降低了约99%,并能准确再现后验等高线,即使在高度非高斯的情况下亦然。
- 本研究确认,哈勃张力在所有测试模型中依然存在,且没有任何ΛCDM的扩展能提供统计上占优的解决方案。
![Figure 2 : Tension between the $H_{0}$ measurement by the SH0ES [ 25 ] collaboration and the marginalised $H_{0}$ posterior obtained in this work for each of the considered models and data combinations. The tension is quantified by $Q_{\text{MPCL}}$ , see eq. ( 4.5 ), and is given in units of $\sigm](https://ar5iv.labs.arxiv.org/html/2312.09814/assets/x2.png)
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。