*2.3. Motivation for New Studies*

As of the early 2000s, our knowledge of the LMC's WR population was thought to be relatively complete thanks to the work of Breysacher's BAT99 catalog [39]. However, other galaxies of the Local Group, namely M31 and M33, still lacked galaxy-wide surveys. Figure 2 shows the observed WC/WN ratio compared to the 2005 Geneva Evolutionary Group's model predictions [1]. (These were the first complete set of models at different metallicities which included the important effect of rotation.) Notice first that the observed relative number of WCs to WNs increases with metallicity. This is exactly what we would expect given single-star evolution because higher metallicity environments will allow more WCs to form. This increase in ratio vs. metallicity is additionally what the models predict. However, a comparison between the models and the observations show that the relative number of predicted WRs is not consistent between the two. Additionally, the models do a particularly poor job of predicting the WC to WN ratio at higher metallicities, such as in M31 and M33.

**Figure 2.** The state of our knowledge of the WC/WN ratio vs metallicity in the mid 2000s. The points are from the 1998 Massey and Johnson summary [73]. The solid curve shows the predictions based upon the 2005 Geneva evolutionary models that included rotation for the first time [1]. Note that, while both show an increase in the WC/WN ratio with metallicity, there is a large discrepancy between the observed results and model predictions at higher metallicity values. Recall that NGC 6822 contains only four WRs (all of WN-type) and the SMC only 12 WRs (one of which is a WC/WO), thus deviations from the models for these two galaxies are not significant.

Clearly, a problem existed, but was it a failing of the models or observations (or both)? Given the complexities of modeling the physics at the end of a massive star's life, it made sense that there could be some deficiencies in the models. However, there were a few reasons that suggested that the observations were actually at fault. For one, as discussed above, there was still no galaxy-wide targeted

survey of WRs in the LMC, M31 or M33; only the SMC had been well covered by the Massey and Duffy survey. The vast majority of WRs that had been discovered within those galaxies had been discovered either by accident or as part of a survey of a limited portion of the galaxy. Additionally, crowding of tight OB associations (where we expect to find the vast majority of WRs) makes finding even bright, strong-lined WRs difficult. Thus, telescopes with more resolving power could help disentangle the tightly-packed regions. Finally, and perhaps more importantly, there is a strong observational bias towards detecting WC-type stars over WNs.

The basis for this observational bias is shown in Figure 3. The strongest emission feature in WCs is nearly 4× stronger than the strongest line in WNs, making WNs much more difficult to detect than WCs of similar brightness [74]. (More accurately, this is an issue of line fluxes; see treatments in [73] and [38].) Thus, while a galaxy (or catalog such as BAT99) might be complete for WC-type stars, there might be a number of missing WNs since their emission lines are so much weaker. The exclusion of these stars would bias the WC to WN ratio to higher values, much like we see when we compare the relative number of WRs observed to that predicted by the Geneva Evolutionary models. Indeed, this was particularly a problem for M31. The ratio of 2.2 shown in Figure 2 is the galaxy-wide average for M31, including the older photographic work; if, instead, one used only the eight CCD fields, this value would drop to 0.9 [73], giving strong credence to selection effects being responsible for the problem.

**Figure 3.** Line strengths of galactic and LMC WRs. The red histogram shows the line strengths (measured as the log of the equivalent width) for a WN's strongest emission line, HII *λ*4686. The blue histogram shows the same for the WC's strongest emission line, CIII/IV *λ*4650. The WC's strongest line is up to 4× stronger than the WN's strongest line making WC stars much easier to detect.

The lack of a galaxy-wide survey for M31 or M33 as well as the possibility of crowding and a strong observational bias against WN stars led us to conduct our own survey for WRs in M31 and M33.

### **3. New Era of Discoveries**

As discussed above, as of 2005, the observed WC/WN ratio was quite poorly aligned with the theoretical predictions at higher metallicities. Thus, M31 and M33 were two ideal regions to study. M31 has the highest metallicity of the Local Group galaxies at log(O/H) + 12 = 8.9 [12,75]. M33 has a strong metallicity gradient going from log(O/H) + 12 = 8.3 in the outer regions up to log(O/H)+12 = 8.7 in the inner regions [76]. Thus, these two galaxies presented the perfect opportunity to re-examine the differences between theory and observations.

In 1985, Massey and Armandroff had pioneered the use of interference filter imaging with CCDs to identify WR candidates [60]. However, the small size of the CCDs available at that time limited

the area that could be covered and the large read-noise limited the sensitivity. An equally large problem, however, was the use of photometry to identify candidates. This method was far superior to "blinking by eye," as had been used in the photographic studies by [52,70,71], and allowed "statistically significant" candidates to be identified. However, the fraction of false positives was overwhelming, simply given the large number of stars involved.

In the mid-2000s, large format CCD Mosaic cameras came along, such as those implemented on the Kitt Peak and Cerro Tololo 4-meter telescopes. CCDs now have read-noises of 3 e- rather than 100 e-, and these mosaic cameras made it practical to cover all of M31 and M33 in a finite number of fields. Equally importantly, supernova and transient searches had required the development of the powerful technique of image subtraction, where the the PSFs were matched between two images, and one image subtracted from another to identify images. We took advantage of both of these improvements in conducting our own searches.
