*3.1. Identification of Candidate WRs*

Searching for candidate WRs was done using the same method in both galaxies as is detailed in [77,78]. Overall, the method combines photometric observations using an interference filter system with image subtraction and photometry for candidate detection.

Thanks to the WR's strong emission lines, they're relatively simple to detect using the appropriately designed interference filters. Taft Armandroff and Massey used spectrophotometry for WR and non-WR to design a three-filter system that was optimized identifying WRs in the optical [60]. All three filters have ∼50 Å wide bandpasses, with one centered on the strongest optical line in a WC's spectrum, CIII/IV *λ*4650 ("WC" filter), another centered on the strongest optical line in a WC's spectrum, HeII *λ*4686 ("WN" filter) and a third on the neighboring continuum at *λ*4750 ("CT" filter). (Placement of the continuum filter to the red of the emissionline filters is crucial; otherwise, red stars show up as candidates.) The bandpasses are shown placed atop the spectrum of both an LMC WC- and WN-type WR in Figure 4. This filter set was used by [60] to search for WRs in the Local Group galaxy dwarfs NGC 6822 and IC 1613, as well as two small test regions of M33. Such work was then extended to selected regions of M33 [73] and M31 [72], and for the galaxy-wide survey of the SMC [49] discussed above. With these interference filter images in hand, there are two main methods of determining stars that are brighter in the on-band filters (WC and WN) vs. in the continuum (CT). The first is using image subtraction and the second is using photometry.

**Figure 4.** Filter bandpasses of WN, WC and CT filters. The WN and WC filters are centered on the strongest lines of the WC and WN-type WRs while the CT is centered on the neighboring continuum; this figure was adapted from [79].

As mentioned above, image subtraction has been used with grea<sup>t</sup> success to detect small brightness changes between on and off band photometry by the supernovae community [80]. Simply subtracting the CT from the WC filter should yield candidate WCs while subtracting the CT from the WN filter should yield candidate WNs. However, seeing variability and small changes in pixel scales across the images turn this simple idea into a complex problem and thus cross-convolution methods and point-spread fitting techniques must be used. Example programs include the Astronomical Image Subtraction by Cross-Convolution program [81] and High Order Transform of PSF and Template Subtraction (HOTPANTS) [82]. An example resulting image is shown in Figure 5 where the background stars have been subtracted out and the candidate WRs are left behind.

**Figure 5.** WR-detection through image subtraction. Three known WRs are outlined in red dashed circles. After subtracting the continuum filter from the WN filter, the resulting image shows three WRs as black stars. This method was used to search for candidate WRs; this figure is from [78].

As discussed above, most WRs are formed in dense OB associations (in fact, Neugent and Massey found that 80% of the WRs in M33 were found in OB associations [78] with only 2% being truly isolated). This dictates the need for crowded field photometry to determine the magnitude differences between the WC-CT and WN-CT filters. Armandroff and Massey had adopted Peter Stetson's DAOPHOT crowded field photometry software [83], with subsequent modifications and porting to IRAF [84]. Careful matching in crowded regions must be performed by eye. Photometry is obtained for all the stars on each on-band exposure (WC, WN), and then matched with the photometry for the same stars on the CT exposure. A zero-point adjustment is then made so that the average difference was zero, and then stars that were more than 3*σ* brighter on either the WC or WN filter exposure when compared to the continuum exposure can be identified.
