JSTOR

2.1. Optical Imaging

The vast majority of the CSP optical imaging is being obtained with the Swope 1 m f/7 telescope Direct CCD Camera, which uses a pixel, 15 μm pixel⁻¹ SITe CCD. This detector has a readout noise of 6.6 e⁻ and a gain of 2.5 e⁻ ADU⁻¹ (where ADU means analog‐to‐digital converter unit). To speed up the CCD readout and save disk space, we limit the readout to pixels. At a scale of 0 435 pixel⁻¹, this corresponds to a field of view (FOV) of . The quality of images obtained at the Swope telescope ranges from ∼1 to 2 FWHM, with an average seeing of 1 3.

For our optical passbands we utilize Sloan Digital Sky Survey (SDSS) u g r i filters, in addition to Johnson B and V filters (Fukugita et al. 1996; Bessell 1990). We have chosen SDSS filters because we feel it is likely that this will be the dominant photometric filter set for the next decade due to the all‐sky photometric maps of the SDSS (north) and the Mount Stromlo and Siding Spring Observatories SkyMapper3 project (south). We decided not to use the z filter because the red side of the filter is determined by the CCD quantum efficiency (QE) and not solely by the filter bandpass. The CCD red response is very temperature sensitive, causing the combined filter‐plus‐detector throughput to be variable. The z bandpass also overlaps with a saturated H₂O band at 9300 Å, adding further uncertainty to its total throughput and effective wavelength. The B and V filters were included in our program to sample the spectral region covered by the g filter with somewhat narrower filters and to facilitate comparison of our results with historical data sets.

Our SDSS filters were manufactured by Asahi Spectra Company, Ltd.,4 as specified in Table 2. We have synthesized natural‐system passbands by combining the filter transmissions with the CCD QE, two aluminum reflectivity curves (one for the primary and another for the secondary mirror), and an atmospheric transmission spectrum. The resulting SDSS bandpasses are shown in the top panel of Figure 1, along with the standard bandpasses (normalized at maximum) for the USNO 40 inch (1 m) telescope. This comparison reveals an excellent match between the two systems, with the exception that our i bandpass is somewhat narrower than that used at the USNO; the effective wavelength is, however, unchanged. The bottom panel of Figure 1 compares the passbands for our Harris B and V filters with the Johnson B and V passbands described by Bessell (1990). It is clear that our instrumental system provides a reasonable match to the Johnson system.

TABLE 2 Specifications for the SDSS and Johnson Filters

Open New Window

(82 KB)

Fig. 1.— Top: Natural‐system synthetic bandpasses for the Swope u g r i SDSS filters (dotted lines) and the USNO 40 inch telescope standard bandpasses (solid lines) used for the establishment of the SDSS photometric system (normalized to 100% at maximum). Bottom: Natural‐system bandpasses for the Swope Harris B and V filters (dotted lines) and the B and V Bessell (1990) bandpasses (divided by a linear function in wavelength and normalized to 100% at maximum).

Open New Window

Since one of our goals is to eliminate systematic errors in the SN magnitudes caused by differences between the instrumental and the standard bandpasses, we have started a program of regular measurements of the transmission of our filters using a spectrometer at LCO. Similarly, we are planning in the near future to improve the measurement of the QE curve for the Swope CCD using equipment available for this purpose at the CTIO laboratories. Our natural‐system bandpasses will be regularly updated as we improve our measurements of the CCD QE, the mirror reflectivities, and other optical elements of our instrument. All of these will be posted on the CSP World Wide Web site and made available through the NASA/IPAC Extragalactic Database (NED).

Some BVI imaging has also been obtained with the Wide Field Reimaging CCD Camera (WFCCD) on the 2.5 m f/7.5 du Pont telescope. This uses a pixel Tektronix CCD with 24 μm pixels. A subraster is used to observe a FOV at a scale of 0 774 pixel⁻¹.

Likewise, a small amount of BVR imaging has been taken with the Low‐Dispersion Survey Spectrograph (LDSS‐2; Allington‐Smith et al. 1994) on the 6.5 m f/11 Magellan Clay telescope. Until the end of 2004, this instrument employed a SITe CCD for which we limited the readout to a section of pixels containing a circular FOV of 6 4 diameter. The pixel size of 15 μm corresponds to a scale of 0 38 pixel⁻¹. The BVR filter specifications for this instrument are given in Table 2. In 2005 February, the LDSS‐2 instrument underwent a significant upgrade that involved replacement of both the optics and the detector. The new CCD is an STA0500A detector with pixels that, in combination with the new optics, has a scale of 0 19 pixel⁻¹. The upgraded instrument, now called LDSS‐3, retains the same B filter used in LDSS‐2, but the V and R filters were replaced by SDSS g , r , i , and z filters. Although we have not yet employed LDSS‐3 for the CSP, we expect to do so occasionally in the future.

For galaxy subtraction, we need template images after the SN has faded. These will be obtained beginning in 2005 November with the Direct CCD Camera on the 2.5 m du Pont telescope. The detector is the same one used with the WFCCD and covers a FOV of with a scale of 0 259 pixel⁻¹. With a typical image quality of 0 7 (FWHM) at the du Pont telescope, this instrument is ideal for acquiring high‐quality deep templates. The filter set will be the same as that used on the 1 m telescope.

2.2. Infrared Imaging

During the first CSP campaign, we obtained 85% of the NIR images with the Wide Field Infrared Camera (WIRC) on the 2.5 m du Pont telescope (Persson et al. 2002), but this situation will change when a new NIR camera, RetroCam, becomes available at the Swope telescope for the second campaign (see § 6). WIRC is equipped with four pixel Rockwell HAWAII‐1 HgCdTe arrays forming a square footprint with a 175 center‐to‐center spacing in the reimaged telescope focal plane. Each array covers a FOV of ∼ with a 0 196 pixel⁻¹ scale. For SN imaging we use the two best quality detectors of the four (detectors 2 and 3). Departures from linearity in measured flux are 1% below 30,000 e⁻, growing to about the 5% level at ∼60,000 e⁻ (see § 4.2). The typical image quality for this instrument falls between 0 5 and 0 8 (FWHM). We observe in the Y, J, H, and K_s passbands, using filters manufactured by Barr Associates, Ltd. (Y, J, H), and Optical Coating Laboratory, Inc. (K_s).

Some NIR data have also been obtained using the Persson Auxiliary Nasmyth Infrared Camera (PANIC) mounted on the 6.5 m Baade telescope (Martini et al. 2004). This instrument has a single pixel HgCdTe HAWAII‐1 detector covering a FOV at a scale of 0 125 pixel⁻¹. The nonlinear behavior of the detector is similar to that of the WIRC arrays. The active optics on the primary mirror produce a typical image quality of 0 3–0 6 (FWHM). The Y, J, H, and K_s filters are similar to those employed with WIRC.

Figure 2 shows the total system transmission curves for WIRC and PANIC, with the YJHK_s filters. The plotted curves include atmospheric transmission, throughput of every optical element in the telescope and instrument, and detector QE. The graph shows excellent agreement between the WIRC and PANIC systems.

(91 KB)

Fig. 2.— Total transmission curves for YJHK_s in WIRC and PANIC. The curves include the Earth’s atmosphere, the telescope and instrument optical elements, and the detector QE.

Open New Window

2.3. Spectroscopy

The majority of our spectroscopic observations have been obtained with the 2.5 m du Pont telescope using the WFCCD instrument in its spectroscopic long‐slit mode. A 400 line mm⁻¹ blue grism is employed with the Tektronix pixel CCD to provide a wavelength coverage from 3800 to 9200 Å at a dispersion of 3.0 Å pixel⁻¹. For the ∼1 6 slit width used for the SN observations, this setup gives a FWHM resolution of ∼6.0 Å. For a mag object in clear conditions, a signal‐to‐noise ratio (S/N) per pixel of 50 at 5000 Å is typically achieved for an exposure time of 1200 s.

When the WFCCD is not available on the 2.5 m telescope due to block scheduling constraints, it has been possible to obtain some spectra with the Las Campanas Modular Spectrograph. This instrument uses a SITe pixel CCD with 15 μm pixel⁻¹ and a 300 line mm⁻¹ grating (blazed at 5000 Å). The resulting spectral coverage is ∼3800–7300 Å at a dispersion of 2.45 Å pixel⁻¹. A slit width of 1 is used for the SN observations, which gives a FWHM resolution of ∼7 Å. Compared to the WFCCD, this instrument has lower efficiency and longer exposures are required.

Occasionally during the first CSP campaign, we were scheduled single nights at the 6.5 m Magellan Clay telescope with LDSS‐2. For these observations, a 300 line mm⁻¹ grism blazed at 5000 Å was employed, providing wavelength coverage of 3600–9000 Å at a dispersion of 5.3 Å pixel⁻¹. For a 1 slit, this translates to a FWHM resolution of ∼13.5 Å.

Approximately 10% of the CSP spectroscopy has been carried out with the CTIO 1.5 m telescope in the service mode of the SMARTS consortium.5 We use the facility Ritchey‐Chrétien (RC) Cassegrain Spectrograph equipped with a pixel Loral CCD, usually at , which gives a dispersion of 5.7 Å pixel⁻¹, a wavelength coverage of 3000–10100 Å, and a FWHM resolution of ∼14 Å.

Note that for none of the above spectrographs do we employ a filter to block second‐order light, which means that there could be second‐order contamination redward of about 7000 Å. This effect will be quantified in a future paper.

3.1. Optical Imaging

Optical imaging with the 1 m Swope telescope would typically begin during the daytime by taking bias frames, as well as dome flats for each of the g r i BV filters with exposure times chosen to achieve ∼23,000 e⁻ per pixel. Immediately after sunset, we would observe the twilight sky with the u filter, followed by sky flats with two or three additional filters. Generally, three to five images were obtained per filter, making sure to offset the telescope between the individual images. The telescope was then focused using the V filter. For the remaining five filters, we would use previously derived focus offsets relative to the V filter.

During the night, SNe were observed according to the priorities previously assigned by the head of the optical‐imaging working group, who determined exposure times based on the source brightness (typically ) and our requirement to achieve 3% photometry. In general we would observe the SNe in all six filters, taking one image per filter. However, given that we usually had more objects to observe than telescope time available, we would drop the 3% precision requirement for the faintest objects and, in such cases, limit the exposure time to 900 s. When the resulting photometric precision reached 0.2 mag, we discontinued the observations through that filter. This would happen first in the u filter, typically at a magnitude of ∼21.

Johnson and SDSS standards (Landolt 1992; Smith et al. 2002) were observed regularly during photometric nights. At the beginning of the first campaign, few science targets were available so we observed 11–13 standard fields per night. As the campaign unfolded and more targets became active, we limited these observations to 4–5 fields. In all cases, we spread the observations of the standards over an air mass range of 1–2 for extinction determinations. The exposure times were 3–120 s depending on the filter used and the brightnesses of the stars. The purpose of these observations was to calibrate local standard stars in the field of each SN so that later on we could do differential photometry of the SN relative to them.

During cloudy nights, we performed detailed shutter timing and linearity tests on the Swope telescope. The results can be found in Appendices A and B.

The observations with WFCCD and LDSS‐2 were performed in a way similar to those with the Swope telescope, except that we did not observe standard stars. We expect to observe standard fields in the future with the purpose of deriving color terms for the filters used. We did not measure the CCD responses and the shutter corrections for these instruments, but we expect to do so in the future for the WFCCD (the LDSS‐2 is no longer operational).

3.2. Infrared Imaging

Calibration images for WIRC and PANIC were usually taken just after closing the dome at the end of the night. Typically the observer would obtain 15 dark frames with exposure times matching those of the science objects, 15 dome flats per filter with the dome lights on (with ∼20,000 e⁻ pixel⁻¹), followed by 15 dome flats with the lights off. Final dome flats were constructed from the “lights‐on” images minus the “lights‐off” images. This procedure was found to be preferable, especially at K_s, to using dark‐subtracted lights‐on dome flats since contributions to the flux from additive sources were removed correctly. The latter sources include the camera window, telescope mirrors, and scattered thermal emission, which illuminate the detector differently than does the flat‐field screen.

For WIRC, we obtained twilight flats on some nights to test the illumination quality of the dome flats. These tests showed agreement between twilight and dome flats within 1% for all four NIR passbands. However, for PANIC it was necessary to take twilight flats for illumination corrections to bring the photometric flatness to better than 2%. We took twilight flats every night in sets of at least five exposures per filter, offsetting the telescope between exposures and controlling the flux level to be in the linear regime. We used dome flats for both WIRC and PANIC to build bad‐pixel masks for each night.

For the WIRC observations, we preselected the telescope pointing for each SN such that there were a number of good local calibration stars in the field. Thus the SN was not always centered in the detector. We started observing each SN by placing it on detector 2 and taking a sequence of dithered exposures in one of the filters. Usually we used nine dither positions in a square pattern, but when the SN was bright enough we used only five positions. We set the exposure time of the individual images to 20, 30, or 45 s, depending on the brightness of the SN. Occasionally, for the faintest SNe we repeated the exposures at each dither position. Once the dither sequence was completed, we would offset the SN field to detector 3 and repeat the sequence for the same filter. This procedure allowed us to obtain sky images suitable for subtraction for each of the two detectors while observing the object on the other detector. We usually observed with all four filters, although in the case of faint SNe, we dropped the K_s filter because of the large uncertainties introduced by the high background levels that characterize this band. Typical times spent on each target, including overhead, were between 1 and 2 hr, with effective exposures between 200 and 800 s in each filter.

There are advantages and weaknesses to this technique. The advantage is that the nearby galaxies tend to be very large, making it generally impossible to use the median of the dithered SN+galaxy data on a single detector to form a clean sky image with no print‐through. The sequence of off‐source dithered exposures (obtained when the SN and galaxy move to the other detector), on the other hand, produces a very clean sky image. The disadvantage is that sometimes the sky may be taken as much as 15 minutes before or after the SN+galaxy exposure, and the NIR sky can change over this time. Thus the sky images, while very clean, may have larger residual subtraction features because of changing sky emission levels.

On the photometric WIRC nights, we observed three to five standard stars from Persson et al. (1998) on detector 2. For observing efficiency, we based the photometric transformations to the standard system on data obtained with this single detector. Standard stars were observed with five dither positions and three repetitions at each position. We chose stars with J fainter than 11 mag to avoid saturation and fixed the exposure time of individual images to 5 s, resulting in total exposure times of 75 s for each filter. The photometric transformations derived from these observations were applied to comparison stars in the SN fields (observed only with detector 2) to obtain precise magnitudes for these local standards.

For our PANIC observations, we selected a slightly different pointing for each SN than that used for WIRC in order to include the maximum number of comparison stars in the smaller FOV of PANIC. Observations were made with five dither positions and two or three exposures per position, depending on the SN brightness. The exposure times for individual images were 10, 20, or 30 s, with total exposure times ranging between 100 and 450 s. We followed each of these sequences by a sky sequence after offsetting the telescope by 1 or 2 in any direction, making sure that the host galaxy fell outside of the sky field.

3.3. Spectroscopy

Our spectroscopic observations typically started during the afternoon by taking bias frames, followed by a series of dispersed dome flats with wide (∼7 ) and narrow (∼1 ) slits. In addition, for the WFCCD and LDSS‐2 instruments, dome flats were obtained for the respective imaging filters (BVI for the WFCCD and BVR for LDSS‐2). We adjusted the lamp voltage and the exposure times to obtain 10,000 e⁻ pixel⁻¹ in our flat‐field direct images and a maximum of 15,000 e⁻ pixel⁻¹ in the spectroscopic flats.

We observed SNe with the narrow slit aligned along the parallactic angle (Filippenko 1982) according to a priority list built for every night. Total exposure times with the WFCCD and the Modular Spectrograph varied between 900 and 2700 s, and between 180 and 900 s with LDSS‐2. For each SN, we divided the total exposure into three independent integrations for cosmic‐ray rejection. In between the first two or last two exposures, an image of a comparison lamp for wavelength calibration was taken. During the night we observed at least two spectrophotometric standards (Hamuy et al. 1992, 1994a) with the wide slit. A few weeks after the beginning of the first campaign, we decided to include one observation per night of a telluric standard (Bessell 1999) with high S/N and the same narrow slit used for the observations of the SNe. Since data from the CTIO 1.5 m telescope were obtained in service mode, we were not able to obtain such a calibration with that instrument.

4.1. Optical Imaging

To accelerate and automate the reduction of the optical imaging, we developed a custom IRAF7 script package to handle the Swope 1 m data. For all nights, we first corrected the exposure time for the shutter timing error (see Appendix A). Next we processed the images through bias subtraction, nonlinearity corrections (see Appendix B), and flat‐field division. For a given filter, we constructed the flat field by dividing the median‐filtered combined sky flat by the median‐filtered combined dome flat, heavily smoothing this ratio, and multiplying this illumination correction into the dome flat. This procedure removed a small gradient of ∼1%. When a nightly sky flat could not be obtained we used one from a previous night. Random checks of science images taken during the night typically showed that the sky level across the images varied by less than 1%. Although fringing is present in the i band we did not attempt to remove it from our frames. Since our i ‐band instrumental magnitudes of the local standards in the SN fields can be transformed to the standard system (as explained below) to better than 0.015 mag, this implies that neglecting the fringing correction introduces an error of ≲0.015 mag.

Next we used the IRAF DAOPHOT package to compute instrumental magnitudes for the standard stars with an aperture of 7 in radius (the same used in the establishment of the standard system) and a sky annulus located 7 –9 from the star. We computed instrumental errors using a Poisson model based on the noise parameters of the CCD. This model provided realistic errors for faint stars but excessively small errors for bright stars. We adopted 0.015 mag as the floor to the calculated errors, based on the observed dispersion in the transformation between instrumental and standard magnitudes of bright stars (see below). To derive the transformation of the instrumental magnitudes into the standard system, we used the following (Harris et al. 1981): In these equations u g r i BV (left‐hand side) are the published magnitudes in the standard system (Landolt 1992; Smith et al. 2002), ugribv (right‐hand side) correspond to the natural‐system magnitudes, k_i is the extinction coefficient, x_i is the effective air mass, ct_i is the color term, and zp_i is the zero point for filter i.

At the beginning of the first campaign, we observed a standards field hourly (∼20 stars during the night), which allowed us to solve for all unknowns (ct_i, zp_i, k_i). As the campaign progressed the pressure to observe SNe increased, leaving less time for measurements of standards. We adopted the approach of measuring only five to eight standard stars over a wide range of air mass (∼10 stars during the night). With this small number of stars we fixed the color term to the average value, solving only for the extinction coefficient and zero point.

During the first observing campaign, we obtained weighted least‐squares solutions for 53 clear nights. Dispersions around best‐fit transformations were ∼0.02–0.04 mag in u and ∼0.01–0.02 mag in other filters. This clearly demonstrates the excellent quality of the LCO site for this photometric program. Figure 3 shows the extinction coefficients as a function of time for all filters. While the scatter was only ∼0.03 in g r i BV, the dispersion was 3 times greater in the u band, which is the most sensitive to extinction variations from night to night.

(99 KB)

Fig. 3.— Extinction coefficients for the u g r i BV filters as a function of time. The average coefficient is plotted with a horizontal line, and its numerical value is indicated in each panel, along with the rms scatter.

Open New Window

Color terms are presented in Figure 4. No obvious secular change was seen over this 250 day span in any filter, even though the primary mirror was washed on 2005 April 6 UT. While the color terms were close to zero (≲0.02) for g r i , the magnitudes of these terms in u BV were greater (∼0.04–0.06), indicating small but nonnegligible differences between the instrumental and standard system.

(92 KB)

Fig. 4.— Color terms for the u g r i BV filters as a function of time. The average color term is plotted with a horizontal line, and its numerical value is indicated in each panel, along with the rms scatter. The vertical dotted line shows the time when the primary mirror of the 1 m telescope was washed.

Open New Window

To calibrate the local standards around the SN, we started by selecting the 5–10 brightest stars in every SN frame. For each frame, we used these stars to compute aperture photometry and derive the magnitude correction between a small aperture (generally 3 ) and the 7 aperture employed for the Johnson+SDSS standards. The aperture correction was typically ∼0.03–0.1 mag. We assumed there were no field effects in the aperture corrections. Then we computed photometry for 10–15 field stars through the small aperture and corrected to the large aperture, bringing the natural‐system photometry to the 7 radius aperture. While this correction added some complication to the procedure, it improved the statistical accuracy of the fainter stars.

Finally, we used equations (1)–(6) to derive magnitudes in the standard system for the local standards. The uncertainty in the resulting magnitudes was the Poisson error in the instrumental magnitudes (assuming a minimum error of 0.015 mag). Here we neglected the uncertainty in the aperture correction, which was always less than 0.01 mag (otherwise we excluded such measurements). From the multiple measurements obtained on different clear nights we took weighted averages, yielding a photometric sequence of secondary standards around each SN with uncertainties as small as 0.004 mag in the individual magnitudes.

We then performed differential photometry of the SN relative to the local standards. The great advantage of this approach is that, since the SN is observed simultaneously with the field stars, the magnitude differences within a CCD frame are, to first order, immune to the passage of clouds. To improve the instrumental precision, we performed point‐spread function (PSF) photometry. For this purpose, we employed all the stars of the photometric sequence to determine an average PSF for every CCD image, and we fitted the resulting PSF to the SN and the standards to a radius of 3 . We converted the instrumental magnitudes to the standard system using equations (1)–(6), assuming that the extinction effects (k_ix_i) were simple additive constants that were absorbed, to first order, by the zero point. As shown in Figure 4, the color terms had no secular variations and we adopted average color terms, solving only for the photometric zero points. The final uncertainty in the SN magnitude was the instrumental error in the PSF fit (assuming a minimum of 0.015 mag, as explained above). We neglected errors due to the transformation to the standard system since the uncertainty in the color term and in the zero point are well below 0.015 mag.

The extraction of the SN magnitude from a CCD frame is generally uncertain because the SN resides in a host galaxy with unknown features at the SN position. Thus, the light curves computed so far should be considered preliminary. As soon as we obtain galaxy templates with the 2.5 m telescope, we will use image subtraction to remove the image of the galaxy and obtain definitive light curves. This procedure will follow that described by Hamuy et al. (1994b), which consists of (1) using the task geomap to determine the coordinate transformation between the two images (assuming a two‐dimensional linear polynomial) and using the task geotran to register the template to the SN+galaxy image; (2) using the task psfmatch to find the two‐dimensional difference kernel that, when convolved with the template, matches the PSF of the SN+galaxy image; (3) using the task linmatch to match the flux scale of the template to that of the SN+galaxy image; (4) subtracting the modified template from the SN+galaxy image; and (5) extracting a small box around the SN from the subtracted image and inserting it in the original SN+galaxy image.

The custom script package developed for the Swope telescope was used also to process the WFCCD and LDSS‐2 data through the flat‐field division. We have not attempted yet to compute SN magnitudes for these instruments because the lack of standard observations has prevented us from calculating color terms. This will be remedied during the second campaign.

4.2. Infrared Imaging

We reduced both WIRC and PANIC data using software pipelines. These pipelines are a combination of IRAF scripts following the steps of (1) linearity correction, (2) dark combination and subtraction, (3) flat‐field combination and division, (4) bad‐pixel mask production, (5) sky image computation and subtraction, and (6) combination of dithered frames into final stacked images. Both pipelines work in a similar fashion, with slight differences due to differing observing procedures.

In the first step, we applied a predetermined linearity correction law for the HAWAII‐1 detectors to every pixel value above 16,000 e⁻: where I is the observed number of ADUs, I_corr is the corrected value, and the coefficients are set to , , and . Next, we normalized the dark‐subtracted dome flats by the median. We considered pixel values outside the range 0.6–1.67 to be unrealistic, replacing them by 1 in the normalized flat and marking them in the nightly bad‐pixel mask. If sky flats were taken (as was usually the case for PANIC), we median combined them and normalized the result to an average value of 1.0, replacing pixel values outside the range 0.2–5 by 1. We checked the results for possible residual stars. In the case of WIRC, we preferred sky flats (when available) to dome flats and we used them to check the illumination quality of the latter, which was usually good to within 1% for all filters. In the case of PANIC, dome flats showed illumination gradients of about 5%–10% across the image. Therefore, when sky flats could not be obtained, we used ones from a previous night. We stored all combined darks, flats, and bad‐pixel masks of a night for future use or examination.

Sky frames to be subtracted from individual object frames were then constructed. We combined off‐target images to produce sky frames in a two‐step process. First, we computed a sky frame by directly averaging all off‐target images. We subtracted this first sky frame from the individual off‐target images. We detected objects in the subtracted frames and masked them in the original frames. We recalculated the sky frame, scaling the images to a common mode. We scaled the sky to the mode of the individual (dithered) object frames and subtracted it from them. We computed also the mode of the object frame in a two‐step process to eliminate the contribution of sources in the image. We obtained the final image by aligning and averaging the individual object frames. In this process, we preserved all pixels except those marked in the bad‐pixel mask. Because the images may have been taken in nonphotometric conditions, we trimmed the data to only the regions of complete overlap. We generally obtained two images per SN and filter with WIRC (one from detector 2 and another from detector 3), and one image with PANIC.

Once the pipelines had produced stacked science images, we proceeded with the photometric measurements. Every time standard stars were observed with WIRC, we measured instrumental magnitudes through a standard aperture of 5 in radius (Persson et al. 1998), with a sky annulus at 5 –7 from the star. The magnitudes carried associated statistical uncertainties based on a Poisson model of the noise. As with the optical photometry, these estimates turned out to be unrealistically small when the objects were bright. Based on the dispersion found in the photometric solutions (see below), we estimate the minimum error in a single measurement to be 0.02 mag. This was our adopted minimum uncertainty in the instrumental magnitudes.

We used the instrumental magnitudes of standard stars to solve for the photometric transformation of the night, defined by the following equations: Here YJHK_s are the magnitudes in the standard system (the JHK_s magnitudes published by Persson et al. [1998] and the Y magnitudes given in Appendix C), yjhk are the corresponding instrumental magnitudes, k_i is the extinction coefficient, x_i is the effective air mass, and zp_i is the zero point for filter i. We fixed the extinction coefficients for JHK_s to the canonical values given by Persson et al. (1998): , , and . In the case of Y, we used the same value as for J ( ). We employed a weighted least‐squares method to find the zero points zp_i. Typical dispersions around the fits were ∼0.02–0.03 mag for all filters. On six nights, we observed 4–5 standards spread over air mass 1–2 in order to also fit the extinction coefficients. We found no significant deviations from the nominal values for JHK_s. In the Y band we found an average of , in agreement with the value of given by Hillenbrand et al. (2002).

Note that we assumed zero color terms since the instrument detector and filters were essentially the same as in Persson et al. (1998). We will verify and monitor this assumption during the course of future campaigns.

With the photometric solution for the night, we proceeded to calibrate the local standard stars in each SN field observed with detector 2 of WIRC. We started by measuring magnitudes of bright, isolated stars through apertures of 1 –5 in every stacked science frame and deriving the magnitude correction between a 2 and a 5 radius aperture. Typical aperture corrections ranged between 0.02 and 0.1 mag. We kept this correction below 0.1 mag by increasing the size of the small aperture, if necessary. Since the uncertainty in the aperture correction was generally ∼0.02 mag, the increase in the statistical uncertainty on the 2 measured magnitude was marginal and certainly lower than that introduced by increasing the aperture to 5 . We measured magnitudes for a number of comparison stars (between 3 and 15) in each SN field through the small aperture and corrected them to the 5 aperture. We transformed these instrumental magnitudes to the standard system using equations (8)–(11). The weighted average of the resulting magnitudes for all comparison stars obtained on different photometric nights was calculated and checked for consistency. We then computed SN magnitudes relative to comparison stars in a manner identical to that used for the optical data.

4.3. Spectroscopy

The first step in the spectroscopic reductions was to combine the bias and flat‐field images and then process the science frames through overscan subtraction, trimming, bias correction, and flat‐fielding. Next, a general wavelength calibration was derived for the night, which we used later as a starting point to obtain specific calibrations from the comparison lamp images obtained at the position of each SN. Our wavelength calibrations were typically characterized by an rms scatter of 0.1 pixel.

The sensitivity function for the night was then calculated. This procedure consisted of extracting one‐dimensional spectra of the flux standards, applying the wavelength calibration, dividing them by the one‐dimensional telluric standard spectrum (after eliminating by hand the weak intrinsic spectral features of such stars), and calculating a response curve based on the flux values measured for the standard stars. As pointed out by Bessell (1990; see also Matheson et al. 2000), the division by the telluric spectrum has the following advantages: (1) it effectively removes the high‐frequency (10–250 pixel wavelength) wiggles in the spectrum that are introduced by typical flat‐field continuum sources and that can have a higher frequency than the flux point spacing, especially in the red; (2) the response curve can be fit accurately with a low‐order function; and (3) most of the telluric features disappear, except those which are strongly saturated (like the oxygen band near 7600 Å). The latter effect is very important in revealing the calcium, oxygen, and carbon features in the red.

Extraction of the SN spectra was accomplished using a window whose width was chosen depending on the seeing and the uniformity of the underlying galaxy distribution. Typically the window included ∼90% of the SN flux and excluded the wings of the stellar profile where the contribution of the galaxy introduced unwanted noise. We used two adjacent windows along the spatial direction to estimate the sky level and the contribution of the galaxy in the SN aperture. In general, we tried to select these background windows as close as possible to the SN aperture, but the criteria for their selection varied from case to case depending on the nature of the adjacent background. A cubic spline was used to interpolate the background at the SN position. In most cases this interpolation provided an accurate estimate of the actual background, but occasionally the background was so nonuniform that some contamination of the host galaxy in the one‐dimensional SN spectrum was unavoidable.

The extracted one‐dimensional SN spectrum was then wavelength calibrated using the comparison lamp exposure taken at the SN position. Finally, the wavelength‐calibrated spectra were divided by the one‐dimensional telluric spectrum, and the sensitivity function was applied. Multiple observations of the same SN were combined into a final spectrum using a median‐filter algorithm to remove deviant pixels caused by cosmic rays.

5.1. Optical Imaging

During the first campaign, we obtained 7852 science optical images and established photometric sequences around all of the 38 SNe included in our follow‐up observations. Final light curves must await the eventual subtraction of template images, but we present here preliminary light curves obtained with the Swope telescope of two SNe that did not suffer significant contamination from their hosts. While we chose these SNe for minimal contamination, the effect is bound to appear at some point when the SNe became very faint. Nevertheless, these examples illustrate the excellent photometric quality of the CSP light curves.

Figure 5 shows the u g r i BV light curves of the Type Ia SN 2004eo. We observed this object for ∼80 days starting ∼10 days before maximum light with a cadence rarely achieved in previous studies. To estimate the precision of our photometry we fit a Légendre polynomial to the light curves with the lowest possible order but making sure to eliminate systematic residuals. In this case, the scatter around the fits amounts to 0.028 (order 8), 0.012 (order 9), 0.006 (order 12), 0.012 (order 15), 0.010 (order 9), and 0.008 (order 10) mag in u , g , r , i , B, and V, respectively. These dispersions can be taken as an empirical and realistic estimate of the random error in a single observation.

(68 KB)

Fig. 5.— The u g r i BV light curves of the Type Ia SN 2004eo.

Open New Window

In Figure 6 we present optical light curves for the Type II plateau SN 2004fx. A prediscovery image on 2004 October 22, where the SN was not visible, indicates that the SN was caught no later than 20 days after explosion. Our first observations confirm that the SN was still quite blue at that point and that it grew progressively redder. We observed the SN on a weekly basis through the plateau phase for ∼80 days, every other night for the next 20 days as it evolved faster, and less frequently again during the linear phase. The scatter in u about a fourth‐order Légendre polynomial is 0.10 mag, but this is dominated by the last few points where the statistical uncertainties are much greater. When the four points with the greatest residuals are removed, the scatter drops to 0.044 mag. In the other bands, the scatter during the plateau phase around an order 10 Légendre polynomial varies between 0.02 and 0.03 mag. During the late‐time linear phase the dispersion around a linear fit varies between 0.03 and 0.10 mag in g r i BV. Part of this scatter is caused by variable amounts of galaxy contamination, as the seeing varied during the observations. We expect the final light curves to be much more uniform at these late epochs.

(72 KB)

Fig. 6.— The u g r i BV light curves of the Type II‐P SN 2004fx.

Open New Window

5.2. Infrared Imaging

During the first campaign, we obtained thousands of individual NIR images. The reduction process produced 1449 calibrated mosaics. We established photometric sequences around all of the 24 SNe included in our follow‐up observations. Our NIR light curves are preliminary, awaiting the final template images and subtraction.

The NIR photometry is presented in the YJHK_s light curves for two representative cases. Figure 7 presents the Type Ia SN 2004ey, while Figure 8 shows the Type II SN 2004fx. In both cases, we show for comparison the i ‐band light curves from our own optical follow‐up program.

(66 KB)

Fig. 7.— YJHK_s light curves of the Type Ia SN 2004ey. The i ‐band light curve is also shown to guide the eye.

Open New Window

(63 KB)

Fig. 8.— YJH light curves of the Type II‐P SN 2004fx. The i ‐band light curve is also shown to guide the eye.

Open New Window

The light curves show the typical sampling achieved of one point every 5–7 days, with gaps of up to 15 days during lunar “dark runs” when WIRC was not available at the 2.5 m du Pont telescope. By exchanging time with other programs, we were able to improve the sampling of some light curves, especially in the cases of SNe Ia around maximum light, to one point every 1–2 days. For SNe II, we relaxed the frequency of observations to one point every ∼10 days, enough to follow the slow plateau evolution. From the observations reduced independently for detectors 2 and 3 of WIRC, we are able to estimate the precision of the measurements. In cases of high S/N, where photon uncertainties can be neglected, the deviation between the two points is ∼0.02–0.03 mag in the YJH bands, consistent with expectations.

The distinctive secondary maximum of SNe Ia is clearly seen in Figure 7, occurring about 20 days after first maximum. This secondary maximum is remarkably prominent in Y, even surpassing the brightness of the first maximum. The NIR behavior of the Type II SN is characterized by a slow yet steady luminosity increase during the plateau phase, a feature previously seen in SN 1999em (Hamuy et al. 2001).

5.3. Spectroscopy

During the first CSP campaign, we obtained a total of 213 optical spectra that are fully reduced. During this period we provided spectral classification in the IAU Circulars for 27 SNe.

Figure 9 displays the spectroscopic evolution of the Type Ia SN 2004ef starting 8 days before maximum light for a period of 47 days. The telluric features are quite evident in these spectra because we did not obtain telluric standard observations for this SN.

(79 KB)

Fig. 9.— Spectroscopic evolution of the Type Ia SN 2004ef. The telluric features are indicated with the Earth symbol (⊕). (Oke & Gunn 1983).

Open New Window

Figure 10 shows the temporal evolution of the spectrum of the Type II SN 2004fx. Unlike the case for SN 2004ef, these spectra were divided by a telluric standard. Most telluric features disappear and only small residuals can be seen at the strongest absorptions (like the A band near 7600 Å).

(102 KB)

Fig. 10.— Spectroscopic evolution of the Type II‐P SN 2004fx. The telluric features are indicated with the Earth symbol (⊕). (Oke & Gunn 1983).

Open New Window