Introducing PxP: A Population Synthesis Framework for Predicting YSO Properties

Abstract

The most direct method of measuring the star formation rate is with young stellar objects (YSOs), but this requires high-resolution observations and high-quality models. Using the latest YSO radiation transfer and stellar evolution models, we have developed a population synthesis code that generates model YSO populations that can be observed by JWST. We combine these model populations with principal component analysis (PCA) and maximum likelihood fitting to create a complete framework for predicting the age and mass of YSO populations. We dub this combination of Population synthesis and PCA, PxP, and show that it is effective at predicting mass and age with self-fitting tests. We apply PxP to the Spitzer identified YSOs in N44 and find a mass of (1.1+-0.1)*10^4 M_sun and an age of 0.74^{+0.06}_{-0.03} Myr, consistent with previous work. Next, we identify 112 YSO candidates in the archival JWST observations of NGC 604. Applying PxP to this newly identified population we find a mass of (2.2+-0.2)*10^4 M_sun and an age of 0.62+-0.01 Myr. This first look at this framework demonstrates its effectiveness with a specific set of models and leaves clear opportunities for future exploration. PxP allows us to directly determine the recent (<3~Myr) star formation history, giving an unprecedented look at the effect of the large-scale environment on individual star formation.

Figures (13)

• Figure 1: Illustration of the population synthesis hierarchy proposed. The starting point is the age and mass of the cloud-scale population. Both individual YSOs and unresolved clusters are then added until the total mass is reached. Individual YSO SEDs are determined by following the left path and unresolved cluster SEDs are determined by following the right path.
• Figure 2: Example output population SEDs from the synthesis code, where each coloured line represent a single source (isolated YSO or unresolved cluster) in the population. All populations are $10^3$ M$_\odot$, and the points show the synthetic JWST photometry with colours and connecting lines only to guide the eye. The top two panels show runs with two different ages, while the bottom two panels show two populations with the same properties. The notable differences between populations with the same properties illustrate the need for dimensionality reduction and a large grid of generated models to find the most likely properties of a real population.
• Figure 3: Illustration of how we go from many iterations of a cloud-scale population with the same parameters into a single grid of PDFs. The right population has three of its SEDs highlighted (red, blue, and green lines), and where those SEDs are translated onto the PCA grid is marked with matching coloured points. This process is then repeated for each age, mass and extinction.
• Figure 4: Illustration of how we fit a real population to the best PDF grid.
• Figure 5: The results of the self-fitting tests to determine the accuracy of PxP's age predictions. The newly generated populations were fit by PxP for four ages (green, blue, red, and purple) and five population masses. The points show the median age produced by PxP for each population with a small offset to input mass for clarity. The solid line shows the average median produced by PxP and the dotted line shows the input age used to generate the population. The shaded region shows the average interquartile range produced by PxP. The results of the self-fitting tests show that the age estimations of PxP are reasonable for populations above $\sim10^4$ M$_\odot$.