Image: Hubble Space Telescope/WFC3,
Credits: A. M. Ghez, M. R. Morris, A. Stolte
Timescales for starburst cluster formation
One question in the starburst cluster project came as a surprise -
we found we could answer a long-discussed cluster formation timescale
question beyond our original expectation:
Do starburst clusters form in true "bursts"?
Is the phrase "starburst" justified? Or do massive, dense clusters form over
extended periods of time?
A related question is: Are starburst clusters dynamically stable?
Are starburst clusters close to virial equilibrium or do they expand or collapse?
Together with our colleagues at the Max-Planck-Institute for Astronomy in Heidelberg,
we could also analyse the rich data set covering the spiral arm clusters:
No age spread in starbursts
In a PhD thesis conducted under supervision of Dr. Wolfgang Brandner
Natalia Kudryavtseva was able to show that the spiral-arm clusters
NGC 3603 YC and Westerlund 1 formed almost instantaneously. The time span
for formation of both clusters is estimated to be less than 10% of the
cluster ages of 2 and 4 Myr, respectively. Using the proper motion
membership samples of stars in each cluster, the young, pre-main sequence
population is tightly constrained. Especially in NGC 3603, the population
is so tight that the age spread is expected to be less than 100,000 years.
This implies that the young star cluster with its tens of thousands of stars
formed in about onehundred thousand years - universally, a very short timescale
Colour-magnitude diagrams of Westerlund 1 and NGC 3603 YC
(Kudryavtseva et al. 2012, Fig. 1).
The green asterisks of the young population in Westerlund 1 indicate proper motion-selected
cluster members. In the colour-magnitude diagram of NGC 3603 YC, only cluster member
candidates are shown. The dense, lower populations are the pre-main sequence stars, while
the dense vertically aligned population is the hydrogen-burning main sequence in both
clusters. The blue lines represent the best-fitting isochrones, suggesting an age of
5 Myr for Westerlund 1 and 2 Myr for NGC 3603.
The transition region marked in both diagrams is a strong indicator for
the cluster age. The width and location of the transition region changes quickly
with time. If multiple generations of stars with various ages are present, the
transition region becomes wider and broader with large scatter, because the
different evolutionary stages of the young stars are mixed.
Kudryavtseva et al. (2012), Natalia used the shape of the transition region
to derive limits on the age spread. The tight and sparse transition regions
suggest that the age spread in both clusters has to be very small.
A statistical analysis of these CMDs (inside the area of each red box) confirmed
this impression. In both clusters, the maximum possible spread between the ages
of the young cluster members is less than 10% of the cluster age.
Likelyhood distribution for stellar evolution models with different ages
(Kudryavtseva et al. 2012, Figs. 2 and 3).
The likelyhoods are derived from fitting isochrones to the red boxed regions
in the CMDs (see
Kudryavtseva et al. 2012 for details). From this set of
stellar evolution models, the most likely age for Westerlund 1 is 5 Myr with
a possible spread of 400,000 years, and 2 Myr with a spread of less than 100,000
years in NGC 3603. In both cases, the width in the age distribution is consistent
with the uncertainties in the photometries and a possible binary contribution.
This implies that no physical "real" age spread is needed to explain the width
of each age distribution.
Here, the proper motion membership proved vital for constraining the age
spread and hence the timescale of cluster birth. Without membership, the
colour-magnitude diagrams might appear very wide and scattered, especially
in the transition region between the pre-main sequence and the hydrogen-burning
zero-age main sequence (see also
Rochau et al. 2010 and
With the capability to reject the majority of field
stars and keep only the proper motion cluster members, this transition region
became very clean and narrow in both clusters. From these colour-magnitude
diagrams, Natalia was then able to derive the maximum spread in the age of
each population using statistical methods.
Conclusion: The age spread in both spiral arm starburst clusters is found
to be less than 10% of the cluster age. The statistical analysis showed that
it was critical to employ proper-motion membership to derive clean transition
regions and pre-main sequence distributions to constrain the age spreads.
The observed 10% possible age variation can be explained by photometric
errors and the contribution of binaries to the width of the pre-main sequence
transition region. Hence, no true age spread is needed to explain the young
cluster population. The two starbursts Westerlund 1 and NGC 3603 appear to
have been born in "bursts of star formation" indeed.
Internal velocity dispersion & virial equilibrium
A second result that could only be obtained with a good cluster membership
sample was the derivation of the internal velocity dispersion.
This spread in orbital velocities is related to the virial mass of each
cluster if the cluster is in virial equilibrium. For NGC 3603 YC, Boyke Rochau
found a velocity disperion of 4.5 km/s, which suggests that the dynamical
mass is in very good agreement with the photometric mass of the cluster,
and hence that NGC 3603 YC is in virial equilibrium
(Rochau et al. 2010 and
A similar result was found for the Arches cluster with a central velocity
dispersion of about 6 km/s
Clarkson et al. 2012).
If a cluster is in virial equilibrium, this implies that the cluster is
neither rapidly expanding nor collapsing. The cluster can still dynamically
evolve slowly with time, for instance because of the segregation of high-mass
stars to the cluster core and the evaporation of low-mass stars near the
cluster outskirts. These mechanisms also lead to changes in the cluster extent
and the core radius, but this dynamical evolution is much slower than the
revirialisation of a cluster out of virial equilibrium.
Conclusion: The proper-motion membership samples allowed the derivation
of the central, internal velocity dispersion of starburst clusters for
the first time. The clusters seem to be dynamically stable:
neither the Arches nor NGC 3603 YC are rapidly expanding or collapsing.
Because of mass segregation, where the high-mass stars sink to the center
and loose orbital energy to the low-mass population, the central velocity
dispersions are lower limits to the overall velocity dispersion in each
starburst cluster. A mean, global velocity dispersion can only be derived
when including the low-mass stars at larger radii, which is not feasible
with current instruments. Nevertheless, the comparison with dynamical
simulations shows that the central velocity dispersion is expected to
be between 50% and 75% of the global, mean velocity dispersion in these
clusters (Olczak et al., work in progress).