Star Clusters

One concern with star clusters as a source of contamination is the possibility of ``hiding'' a luminous dusty star in a dusty star cluster. To explore this problem we estimated what the SED of the star cluster containing $\eta$Car would look like if it were located in one of the targeted galaxies. We combined the SED of $\eta$Car from Humphreys & Davidson (1994) with the SED of the Carina nebula from Smith & Brooks (2007) to produce an SED of the entire complex (Figure6). The combined SED is clearly a multi-component SED, which is not typical of our candidates. Moreover, the Carina nebula is roughly $\sim 2.5\arcdeg$ in extent (Smith & Brooks2007), which at the distance of M33 becomes $\sim 20\farcs$ and would be easily resolved by IRAC. Even at the distance of NGC7793 it would still subtend $\sim 5\farcs$ and be resolved. At all these distances it would be very easily resolved by HST or JWST (Gardner et al.2006).

In Figures 2 and 3, we show the mid-IR luminosity $L_{mIR}$, SED slope $a$, and fraction $f$ of $\eta$Car (``$\eta$''), the Carina nebula excluding $\eta$Car (``$\eta -$''), and the entire complex including $\eta$Car (``$\eta +$''). It is apparent from these figures that even if the Carina nebula was not resolved: (1) we would select analogs of $\eta$Car and unresolved dusty stellar complexes hosting such analogs, (2) while it is close, we would not select a stellar complex that is identical to the Carina nebula excluding $\eta$Car, and (3) there are no sources with $L_{mIR}$, $a$, and $f$ comparable to $\eta$Car in M81. Indeed, this last point is true for each galaxy we studied.

There are, however, far more compact star clusters among the candidates such as M33-5, M33-8 and M81-10 (see Section3) where HST images are required to recognize their spatial extent. Even in these cases it is unlikely we would lose a candidate. First, it would require a ``conspiracy'' of a sort, namely that the SED of the hotter circumstellar dust around the star (with characteristic $T\sim 400$K and $L_{bol}\sim few \times 10^6\,L_{\odot}$) seamlessly merges with the colder SED of the interstellar dust (with characteristic $T\sim100$K and $L_{bol}\sim 10^7\,L_{\odot}$) in the cluster. Typically we find that this leads to SEDs with ``bumps'' which we do not observe.

Possibly more constraining is the requirement that for a compact cluster to hide an $\eta$Car analog it must still contain large amounts of interstellar gas and dust several million years after the cluster formed to allow for the time that even the most massive stars require to evolve away from the main sequence. However, a cluster sufficiently luminous to hide an $\eta$Car analog must host many luminous stars with strong UV radiation fields and winds, which will likely clear the cluster of gas and dust needed to produce strong mid-IR emission. For example, 30Dor, which harbors stars possibly as massive as $300\,M_{\odot}$ and is about $1.5\times 10^6\,$years old (e.g., Crowther et al.2010), is a weak source of $8\,\micron$ emission (see, e.g., Figure1 of Zhang & Stanek2012).