TL;DR
This paper introduces a novel Bayesian framework for jointly analyzing the power spectrum (PS) and voxel intensity distribution (VID) of CO emission to constrain the CO luminosity function. By combining these two complementary observables, the researchers demonstrate a significant improvement in precision, reducing uncertainties by up to 58% and 30% compared to using either observable alone, which is crucial for future CO intensity mapping experiments like COMAP.
What Problem Does This Paper Address?
Mapping the universe's large-scale structure and its evolution is a cornerstone of modern cosmology. CO intensity mapping experiments, like the upcoming COMAP, aim to achieve this by observing the faint, diffuse emission from carbon monoxide (CO) molecules across vast cosmic volumes. A key scientific goal for these experiments is to precisely measure the CO luminosity function (LF), which describes how many CO-emitting galaxies exist at different luminosity levels. This function is a vital probe of galaxy formation and evolution.
However, extracting the CO LF from noisy, complex intensity maps is challenging. Traditional methods often rely on either the power spectrum (PS), which captures the statistical distribution of fluctuations on different scales, or the voxel intensity distribution (VID), which describes the probability distribution of brightness temperatures in individual voxels. While each provides valuable information, they are sensitive to different aspects of the CO signal and are affected by different sources of noise and cosmic variance. The problem this paper addresses is how to optimally combine these complementary observables to achieve the most precise and robust constraints on the CO luminosity function.
Methods
The authors develop a Bayesian framework utilizing a Markov Chain Monte Carlo (MCMC) sampler to estimate the CO luminosity function. The core of their approach lies in constructing a joint likelihood that incorporates both the power spectrum (PS) and the voxel intensity distribution (VID).
Key methodological aspects include:
- Fast, Semi-Realistic Simulations: To handle the complex likelihood and covariance matrices, the researchers rely on computationally efficient simulations. These simulations are built upon "peak patch" dark matter halo catalogs, which are known for their speed, and are coupled with an empirical CO luminosity ($L_{CO}$) to dark matter halo mass ($M_{halo}$) model from Li et al. (2016). This allows for the generation of a large number of realistic mock CO intensity maps.
- Joint Covariance Matrix Estimation: A crucial step is the estimation of the joint covariance matrix between the PS and VID. This matrix accounts for correlations both within each observable (e.g., between different PS bins) and between the PS and VID themselves. They achieve this by averaging over a large number of simulated maps.
- Approximation for MCMC: To make the MCMC computationally tractable, they employ an approximation for the covariance matrix that scales with model parameters based on the independent variance. This allows the covariance matrix to be updated efficiently at each step of the MCMC chain.
- Gaussian Likelihood: For both the PS and VID, a Gaussian likelihood is assumed. While the authors acknowledge that the VID might deviate from Gaussianity at high temperature bins due to discrete counts, they argue that this approximation is generally valid and can be mitigated by adjusting bin widths.
Datasets
The study utilizes simulated datasets generated to mimic the COMAP experiment.
- Dark Matter Halo Catalogs: "Peak patch" cosmological simulations are used to generate dark matter halo catalogs. These simulations are designed to be computationally fast, allowing for the creation of many independent realizations of the cosmic structure.
- Mock CO Intensity Maps: These halo catalogs are then used in conjunction with the $L_{CO}(M_{halo})$ model from Li et al. (2016) to generate mock CO intensity maps. These maps are further convolved with instrumental beam profiles and have noise added to simulate the COMAP observations for both Phase 1 and Phase 2.
No new observational datasets were used; the study focuses on forecasting capabilities based on simulations.
Key Insights and Ideas
- Complementarity of PS and VID: The PS and VID are not redundant; they probe different aspects of the CO signal and are sensitive to different luminosity ranges. The VID is particularly powerful at constraining the high-luminosity end of the CO LF, while the PS is more effective at the low-luminosity end.
- The Power of Joint Analysis: Combining PS and VID offers a significant advantage over using either observable in isolation. This synergy arises from their complementary sensitivities, leading to a more comprehensive picture of the CO emission.
- Simulation-Based Inference is Key: The ability to generate fast, semi-realistic simulations is critical for building the necessary covariance matrices and performing the Bayesian analysis. This approach is generalizable to other random field observations with complex likelihoods.
- Model Dependence: The precision of the constraints is inherently tied to the accuracy of the underlying CO luminosity model. While the current model provides a good starting point, future improvements in simulations and models will be crucial.
Key Results
The primary quantitative result is the significant improvement in the precision of CO luminosity function constraints achieved by the joint PS+VID analysis:
- Uncertainty Reduction: On average, the joint analysis reduces the uncertainties (in dex) in the CO luminosity function by 58% for the PS and 30% for the VID, when compared to using each observable separately.
- COMAP Phase 1 & 2 Forecasts: The study provides forecasts for both COMAP Phase 1 and Phase 2, demonstrating the expected precision achievable by the experiment.
- Parameter Constraints: The analysis also provides posterior distributions for the parameters of the Li et al. (2016) $L_{CO}(M_{halo})$ model. Notably, the parameters governing the CO luminosity-infrared luminosity relation ($\alpha$ and $\beta$) are shown to be significantly better constrained by the joint analysis.
Limitations and Future Work
- Model Dependence: The accuracy of the results is contingent on the fidelity of the $L_{CO}(M_{halo})$ model. If the true CO emission is more complex than the model predicts, the derived constraints may not be reliable.
- Minimum Halo Mass: The simulations used a minimum dark matter halo mass of $2.5 \times 10^{10} M_{\odot}$. While the authors note that this mass range contributes a small fraction of the CO signal, exploring lower mass thresholds could be important for investigating low-luminosity cutoffs in the CO LF.
- Instrumental Systematics and Foreground Contamination: The current analysis neglects instrumental systematics and astrophysical foregrounds. The authors suggest that these can be incorporated in future work once more realistic instrument models are available.
- Generalizability: The authors emphasize that the method is generally applicable to any random field with a complicated likelihood, provided fast simulation procedures are available. Future work could explore applying this framework to other cosmological probes or different astrophysical functions.
- Redshift Evolution: The current analysis assumes a simplified approach by neglecting redshift evolution of observables. For future, higher signal-to-noise experiments, analyzing different redshift bins independently could be beneficial.
Summary updated