Recent advances in both observations and numerical simulations of
star-forming regions have opened up the possibility of coupling these
two fields together. This thesis presents detailed radiative transfer
models created from large-scale simulations of star-forming molecular
clouds. The radiative transfer models are used to calculate synthetic
observables, which are compared directly to a number of observational
studies. The numerical simulations form several hundred
protostars, meaning that, for the first time, such a comparison can be
done in a truly statistical manner. The goal of this comparison is both
to benchmark the simulations by testing if observational results can
be reproduced, and to use the simulations to aid in the interpretation
of the observations.
The research deals with the earliest stages of star formation – the
protostellar phase – where the protostars are still embedded within
massive dusty envelopes with size-scales of roughly 0.1 pc1. We use
spectral energy distributions of the protostars in the simulation to
calculate evolutionary tracers, and find their distributions to match
the observations well, save for some optical depth issues that can be
traced back to the resolution of the simulation. We also study the
distribution of protostellar luminosities in the simulation, and find
that both median and spread matches the observed distribution well.
Both of these tests are important benchmarks of the simulation since
they show that the overall evolution of the protostars in the simulation
matches the observational results. We also study the occurrence
of circumstellar disks in the same simulation and find that they are
ubiquitous at all stages of the protostellar evolution.
A special emphasis is put on the study of protostellar accretion,
which may have important physical consequences for the evolution
of protostellar systems. The sublimation of CO-ice from dust grains
in the surrounding envelope can be used to trace accretion variability
in protostars, because the increased heating during an accretion
burst will cause the CO-ice to sublimate into the gas-phase where the
excess can be measured by telescopes. We recreate such observations
from a numerical simulation, and find that it is indeed possible to
trace accretion variability in such a manner, thereby confirming the
approach taken by an observational study. The synthetic observations
fail to reproduce the full spread of values seen in the real observations,
which can be traced back to a lack of accretion variability in the
simulation. This particular simulation does not include disk physics,
and we attribute this lack of accretion variability to that effect.
We also carry out an observational follow-up study, attempting tolink evidence of accretion bursts together with evidence of circumstellardisks. The study targets 20 embedded protostars in the Perseusmolecular cloud, and reveals plenty of evidence for variable accretionthrough observations of C¹⁸O (an optically thin isotopologue of CO).The study also reveals that low-luminosity protostars are more likelyto have enhanced C¹⁸O emission, which is interpreted to mean thatthe low-luminosity protostars are between accretion bursts, while thehigh-luminosity objects are those that are currently undergoing accretionbursts.