The Carnegie Supernova Project. I. Third Photometry Data Release of Low-redshift Type Ia Supernovae and Other White Dwarf Explosions

We present final natural-system optical (ugriBV) and near-infrared (YJH) photometry of 134 supernovae (SNe) with probable white dwarf progenitors that were observed in 2004–2009 as part of the first stage of the Carnegie Supernova Project (CSP-I). The sample consists of 123 Type Ia SNe, 5 Type Iax SNe, 2 super-Chandrasekhar SN candidates, 2 Type Ia SNe interacting with circumstellar matter, and 2 SN 2006bt-like events. The redshifts of the objects range from to 0.0835; the median redshift is 0.0241. For 120 (90%) of these SNe, near-infrared photometry was obtained. Average optical extinction coefficients and color terms are derived and demonstrated to be stable during the five CSP-I observing campaigns. Measurements of the CSP-I near-infrared bandpasses are also described, and near-infrared color terms are estimated through synthetic photometry of stellar atmosphere models. Optical and near-infrared magnitudes of local sequences of tertiary standard stars for each supernova are given, and a new calibration of Y-band magnitudes of the Persson et al. standards in the CSP-I natural system is presented.


Introduction
Type Ia supernovae (SNe, singular SN) are generally agreed to be the result of a carbonoxygen white dwarf that undergoes a thermonuclear runaway (Hoyle & Fowler 1960) owing to mass accretion in a binary system (Wheeler & Hansen 1971). The mechanism for the ignition of the degenerate material is thought to be tied to the interplay between the exploding white dwarf and its companion star. Potential progenitor systems are broadly categorized as "single degenerate" where the companion star is a main sequence, red giant, or helium star, or "double degenerate" where the system consists of two white dwarfs. Within this scheme, several triggering mechanisms have been proposed. The thermonuclear explosion can be triggered by the heat created during the dynamical merger of two white dwarfs after expelling angular momentum via gravitational radiation (e.g., Webbink 1984;Iben & Tutukov 1984). The explosion can also be triggered by compressional heating as the white dwarf accretes material from a degenerate or nondegenerate companion to close to the Chandrasekhar limit (e.g., Whelan & Iben 1973). A third mechanism involves the explosion of a sub-Chandrasekhar-mass white dwarf triggered by detonating a thin surface helium layer which, in turn, triggers a central detonation front (e.g., Nomoto 1982). A fourth mechanism might be a collision of two C-O white dwarfs in a triple-star system (Kushnir et al. 2013).
Currently, it is unclear whether the observed SN Ia population results from a combination of these explosion mechanisms or is largely dominated by one. The power-law dependence of the delay time between the birth of the progenitor system and the explosion as a SN Ia (the "delay-time distribution"; Maoz, Sharon, & Gal-Yam 2010) and the unsuccessful search for evidence of the companions to normal Type Ia SNe (e.g., see Li et al. 2011b;Schaefer & Pagnotta 2012;Olling et al. 2015) would seem to favor the doubledegenerate model, but some events, such as SN 2012cg (Marion et al. 2016) and SN 2017cbv (Hosseinzadeh et al. 2017) show a blue excess in their early-time light curves, indicative of nondegenerate companions. The rare SNe Ia that interact with circumstellar matter (CSM), such as SNe 2002ic (Hamuy et al. 2003) and PTF 11kx (Dilday et al. 2012), also favor a single-degenerate system. Type Ia SNe are important for their role in the chemical enrichment of the Universe (e.g., Nomoto, Kobayashi, & Tominaga 2013, and references therein). They also play a fundamental role in observational cosmology as luminous standardizable candles in the optical bands (e.g., Phillips 1993;Hamuy et al. 1996;Riess et al. 1996;Phillips et al. 1999) and as (essentially) standard candles at maximum light in the near-infrared (NIR) (Krisciunas, Phillips, & Suntzeff 2004;Krisciunas 2012;Phillips 2012, and references therein). The most precise current estimates for the value of the Hubble constant are based on SNe Ia (Riess et al. 2016, and references therein); moreover, Riess et al. (1998) and Perlmutter et al. (1999) used them to find that the Universe is currently expanding at an accelerating rate.
In this age of precision cosmology, observations of SNe Ia continue to play a crucial role (e.g., see Sullivan et al. 2011). Ironically, we are still faced with the situation that many more events have well-observed light curves at high redshifts (z > 0.1) than at low redshifts (Betoule et al. 2014). Since the SN Ia results are derived from a comparison of the peak magnitudes of distant and nearby events, the relatively heterogeneous quality of the low-redshift data directly affects the precision with which we are able to determine the nature of dark energy. Moreover, there are still legitimate concerns about systematic errors arising from the conversion of instrumental magnitudes into a uniform photometric system, calibration errors, the treatment of host galaxy dust reddening corrections, and evolutionary effects caused by differing ages or metallicities (Wood-Vasey et al. 2007;Freedman et al. 2009;Conley et al. 2011).
The Carnegie Supernova Project-I (CSP-I; Hamuy et al. 2006) was initiated to address these problems by creating a new dataset of low-redshift optical/NIR light curves of SNe Ia in a well-understood and stable photometric system. The use of NIR data provides several major advantages over optical wavelengths alone. First, color corrections caused by dust and any systematic errors associated with these are up to a factor of five smaller than at optical wavelengths (Krisciunas et al. 2000;Freedman et al. 2009). The combination of optical and NIR photometry also provides invaluable information on the shape of the host-galaxy dust reddening curve (Folatelli et al. 2010;Mandel et al. 2011;Burns et al. 2014). Finally, both theory and observations indicate that the rest-frame peak NIR magnitudes of SNe Ia exhibit a smaller intrinsic scatter (Kasen 2006;Kattner et al. 2012;Mandel et al. 2009) and require only minimal luminosity vs. decline-rate corrections.
The CSP-I was a five-year (2004)(2005)(2006)(2007)(2008)(2009) project funded by the National Science Foundation (NSF). It consisted of low-redshift (z 0.08) and high-redshift (0.1 < z < 0.7) components. Hamuy et al. (2006) presented an overview of the goals of the low-redshift portion of the project, the facilities at Las Campanas Observatory (LCO), and details of photometric calibration. It should be noted that the CSP-I also obtained observations of more than 100 low-redshift core-collapse SNe. Contreras et al. (2010, hereafter Paper 1) presented CSP-I photometry of 35 low-redshift SNe Ia, 25 of which were observed in the NIR. Analysis of the photometry of these objects is given by Folatelli et al. (2010). Stritzinger et al. (2011, hereafter Paper 2) presented CSP-I photometry of 50 more low-redshift SNe Ia, 45 of which were observed in the NIR. This sample included two super-Chandrasekhar candidates (Howell et al. 2006) and two SN 2006bt-like objects (Foley et al. 2010b). The high-redshift objects observed by the CSP-I in the rest-frame i band are discussed by Freedman et al. (2009).
In this paper, we present optical and NIR photometry of the final 49 SNe in the CSP-I low-redshift sample, including five members of the SN 2002cx-like subclass, also referred to as Type Iax SNe (see Foley et al. 2013), and two examples of the Type Ia-CSM subtype (Silverman et al. 2013). We provide updated optical and NIR photometry of the 85 previously published low-redshift SNe in the CSP-I sample since, in several cases, we have eliminated bad data points, improved the photometric calibrations, and obtained better host-galaxy reference images. This combined dataset represents the definitive version of the CSP-I photometry for low-redshift white dwarf SNe, and supersedes the light curves published in Papers 1 and 2, as well as those published for a few individual objects by Prieto et al. (2007), Phillips et al. (2007), Schweizer et al. (2008), Stritzinger et al. (2010), Taddia et al. (2012a), Stritzinger et al. (2014), and Gall et al. (2017). Other useful optical and near-IR observations of Type Ia SNe includes the photometry obtained by the Center for Astrophysics group (Hicken et al. 2009(Hicken et al. , 2012Friedman et al. 2015).

Supernova Sample
In Figure 1 we present finder charts for the 134 SNe Ia comprising the low-redshift CSP-I white dwarf SN sample, indicating the positions of the SN and the local sequence of tertiary standard stars in each field (see §5.2). General properties of each SN and host galaxy are provided in Table 1. Two "targeted" searches, the Lick Observatory SN Search (LOSS; Filippenko et al. 2001;Leaman et al. 2011;Li et al. 2011a) with the 0.76 m Katzman Automatic Imaging Telescope (KAIT) and the Chilean Automatic Supernova Search (Pignata et al. 2009), accounted for 55% of the SNe selected for follow-up observations. Another 36% of the SNe in the sample were discovered by amateur astronomers, and the remaining 19% were drawn from two "untargeted" (sometimes referred to as "blind") searches: the Robotic Optical Transient Search Experiment (ROTSE-III; Akerlof et al. 2003)

and the Sloan Digital
The top panel of Figure 2 shows a histogram of the heliocentric radial velocities of the host galaxies of the 134 SNe in our sample. The redshifts range from z = 0.0037 (for SN 2010ae) to 0.0835 (for SN 2006fw). The median redshift is 0.0241, corresponding to a distance of 100 Mpc for a Hubble constant of 72 km s −1 Mpc −1 (Freedman et al. 2001). Table 2 summarizes spectroscopic classifications for the sample. The spectral subtype is listed, along with the epoch of the spectrum (relative to the time of B-band maximum) used to determine the spectral subtype. Also given are classifications in the Branch et al. (2006) and Wang et al. (2009) schemes using the same criteria as Folatelli et al. (2013). Photometric parameters for the subset of 123 SNe Ia are provided in Table 3. See §7.1 for details.

Imaging
Between 2004 and 2010, five 9-month CSP-I observing campaigns were carried out, each running from approximately September through May. During these campaigns, the vast majority of the optical imaging in the ugriBV bandpasses was obtained with the SITe3 CCD camera attached to the Las Campanas Observatory (LCO) Swope 1 m telescope. A limited amount of optical imaging was also taken with the Tek5 CCD camera on the LCO 2.5 m du Pont telescope.
NIR imaging of the CSP-I SNe during the first observing campaign was obtained exclusively with the Wide-Field IR Camera (WIRC) on the du Pont 2.5 m telescope , and some additional WIRC observations were carried out during campaigns 2-5. However, beginning with the second CSP-I campaign, a new imager built especially for the CSP-I, RetroCam, went into use on the Swope 1 m telescope and became the workhorse NIR camera for the remaining four campaigns.
Basic reductions of the optical and NIR images are discussed in detail in Paper 1. For the optical data, these consisted of electronic bias subtraction, flat-fielding, application of a linearity correction appropriate for the CCD, and an exposure-time correction that corrects for a shutter time delay. The individual dithered NIR images were corrected for electronic bias, detector linearity, pixel-to-pixel variations of the detector sensitivity, and sky background, and were then aligned and stacked to produce a final image.
Host-galaxy reference images were obtained a year or more after the last follow-up image. As described in Paper 1, most of the optical ugriBV reference images were obtained with the du Pont telescope and the Tek5 CCD camera using the same filters employed to take the original science images 18 . A smaller set of reference images was also taken with the du Pont telescope using a second CCD camera, SITe2, and a few were obtained using the Swope + SITe3 camera under good seeing conditions. For a small number of objects located far outside their host galaxies, subtraction of a reference image was unnecessary.
NIR Y JH host-galaxy reference images were obtained exclusively with WIRC on the du Pont telescope using similar filters as in RetroCam.

Filters
Precision photometry requires knowledge of the filter throughputs as a function of wavelength (e.g., Bessell 1990;Stubbs & Tonry 2006), so we devised an instrument incorporating a monochromator and calibrated detectors to precisely determine the response functions (telescope + filter + instrument) of the CSP-I bandpasses (Rheault et al. 2014). Paper 2 provides a detailed account of the measurement of the optical bandpasses, and in Appendix A we describe the calibration of the NIR bandpasses using the same instrument and similar techniques.
Repeated scans of the CSP-I ugriBV bandpasses show that the relative measurement errors in transmission are ∼ 1% or less. That is, the ratios as a function of wavelength of repeated scans of each individual filter fall within an envelope that is ±1%. Repeated scans of the Y JH bandpasses (both for the Swope + RetroCam and the du Pont + WIRC) indicate that each of these filters has been determined in a relative sense to a precision of 2-3%. Unfortunately, the throughput of the WIRC K s filter is highly uncertain beyond 2200 nm (2.2 µm) owing to the low power of the monochromator light source at these wavelengths and the rising thermal contamination at 2.3 µm. Nearly all of the K s -band observations made by CSP-I were obtained during the first observing campaign, and were published in Papers 1 and 2. However, due to the uncertainty in the K s filter response function, we have elected not to include any K s -band observations in this final data release paper. Those wishing to employ the CSP-I photometry for precision cosmology applications are advised not to use the K s -band measurements given in Papers 1 and 2. Figure 3 displays the optical and NIR bandpasses employed by the CSP-I after including atmospheric transmission typical of LCO. In constructing the optical filter bandpasses, we have assumed an airmass of 1.2, a value which corresponds to the mode of the airmasses of the standard-star observations used to calibrate the data. In Appendix B we test the validity of the final optical bandpasses by reproducing the measured color terms (see §6.1.1) via synthetic photometry performed on spectra of Landolt standards.

Photometric Reductions: Overview
In this and the following section we define the CSP-I natural photometric system and describe the methodology used to calibrate it. While this has been described in previous CSP-I publications, several changes have been made in our definitions and procedures. These changes affect the entire CSP-I sample, and as this is the final data release, we seek to make the procedure as clear as possible.

The CSP-I Natural System
Because of differences in instrument throughputs, photometry measured by different facilities will not agree. These differences are a strong function of color of the object and can therefore be taken into account through the use of color terms (e.g., see Harris et al. 1981). These color terms are typically measured empirically by observing a set of standard stars with a large range in color, and allow the observer to transform their instrumental photometry into the system in which the standards were measured.
The primary difficulty in dealing with supernova photometry is the fact that their spectral energy distributions (SEDs) are significantly different than those of the stars we use to calibrate. Supernova spectra also evolve significantly with time. Consequently, the color terms cannot be used on the SN magnitudes to transform them to a standard system. Instead, we adopt a natural system, in which the standard-star magnitudes are transformed to what we would measure through our own telescopes/instruments. There are several advantages to working in the natural system, as follows.
• If the system is stable (i.e., color terms do not vary significantly with time), nightly calibration of each filter does not rely on other filters to measure colors. This can be advantageous if time is short.
• Working in the natural system requires fewer standard-star measurements to obtain the nightly zero points, as the equations have one fewer unknown. In fact, the equations can be reduced to only one unknown, the nightly zero point (see §6).
• Photometry in the natural system is the "purest" form of the data and, given precise bandpass response function measurements, allows the CSP-I observations to be more readily combined with photometry in other photometric systems using S-corrections (Stritzinger et al. 2002;Krisciunas et al. 2003).
To compare photometry of SNe in host galaxies at different redshifts, precision Kcorrections must be calculated with the transmission functions used in the observations and not that of the standard system. Thus, one must back out the standard system color transformation to the natural system in order to do the K-correction.
Having introduced the natural system, we now proceed to describe in general terms the procedure used to measure and calibrate the photometry using standard stars.

Standard Stars
Observations of standard stars are required in order to calibrate the SN photometry. In this paper, we adopt the following nomenclature in referring to the different types of standard stars used by the CSP-I.
• Secondary standards. We employed observations of Landolt (1992) and Smith et al. (2002) standard stars to provide the fundamental calibration of the CSP-I optical photometry. The Landolt and Smith et al. stars are considered "secondary standards" since they were calibrated with respect to the primary standards Vega and BD+17 • 4708, respectively. In the NIR, the CSP-I photometry is calibrated with respect to the Persson et al. (1998, hereafter P98) secondary standards, which are tied to Vega.
• Tertiary standards. A "local sequence" of stars was established in each SN field in order to allow relative photometry of the SN to be measured. We refer to the local sequence stars as "tertiary standards" because they were calibrated via observations of secondary standards.

Supernova Photometry and Calibration
In order to measure photometry of the SNe accurately, the underlying host-galaxy light is first subtracted from each SN image using the host-galaxy reference images obtained after the SN has disappeared. The details of this procedure are discussed in Papers 1 and 2. DAOPhot (Stetson 1987) is then used to measure counts in our CCD frames for both the SN and the local sequence of tertiary standards using point-spread-function (PSF) photometry. For each tertiary standard, i, we measure a differential magnitude with respect to the SN, where e − SN and e − i are the photoelectrons measured for the SN and tertiary standards, respectively. The uncertainty σ (∆m SN,i ) is computed assuming Poisson statistics.
Once the tertiary standards have been calibrated to the natural system, the final magnitude of the SN can be computed as a weighted average, where m i are the calibrated magnitudes of the tertiary standards, and the weights w i are the inverse variance The uncertainty in the SN photometry is therefore which contains a variance term for the statistical uncertainty from photon counts as well as a systematic variance term that describes the uncertainty in each tertiary standard's absolute flux. This procedure is applied for each SN field in each filter. The remainder of this section deals with the determination of the calibrated magnitudes of the tertiary standard stars, m i .

Tertiary Standard Calibration
In the natural system, the calibrated magnitudes of the tertiary standards are determined relative to the natural magnitudes of secondary standards observed on photometric nights. For a set of photometric nights (j), on which a set of secondary standards (k) is measured, the estimate of the magnitude of the tertiary standard (i) in a particular filter (λ) is where m nat,k,λ is the natural-system magnitude of the secondary standard k, ∆m i,j,k,λ is the differential magnitude (see Eq. 1) between the tertiary standard i and the secondary standard k on night j, X i,j,λ and X j,k,λ are the respective airmasses of the tertiary and secondary standards, and k λ is the extinction coefficient. The weights (w j,k,λ ) are the inverse variances where σ (∆m i,j,k,λ ) is calculated assuming Poisson errors and σ m nat,k,λ is taken from the published standard photometry. The uncertainty, σ (m i,λ ), is analogous to Eq. 4: Note that because this is a natural system, there is no color term in Eq. 5 and hence no dependence on the color of the tertiary standards. The differential magnitudes, ∆m i,j,k,λ , are measured using aperture photometry as this was found to be more robust for the secondary standard stars, which tend to be significantly brighter than the tertiary standards. In the optical we used an aperture of 7 , while for the NIR, we used an aperture of 5 . A sky annulus of inner radius 9 and 2 width was used to estimate the sky level for both the optical and NIR.
The final ingredients are the natural-system magnitudes for the secondary standards, m nat,λ . As discussed in §5.1, color terms are used to transform the standard magnitudes of these stars into the natural-system magnitudes that would be measured through our telescopes. The form of these transformations is assumed to be linear with color, where λ is the color term and C λ is the associated color based on the standard magnitudes.
As an example, for λ = B, we choose C λ = (B − V ). It is important to emphasize that because these color terms are only ever used to compute m nat,λ , it is the range of colors of the secondary standards used for calibration that determines their relative importance. In other words, we are forcing the zero point of the natural system and standard system to be the same at zero color.
Technically, each telescope/instrument used by the CSP-I will have its own color terms and hence its own set of natural magnitudes for the secondary standards. In the next section, we describe each of these in detail.

Photometric Reductions: Details
Our natural system is defined by Eq. 8. If all published secondary standards had zero color, then the definition of the natural system would be trivial. However, the published standard stars have a range of colors. To use all the published standards to define the natural system we must calculate color terms, as given in Eq. 8, which transform the published standard system into a table of the same stars with natural-system photometry.
If something goes wrong with the photometric system and the transmission functions change, then the color terms in Eq. 8 will change. However, for program objects that are stars, the natural system will remain well defined because stars were used to define the standard to natural systems. This is not true for SNe because they have different SEDs. Thus, an important sanity check on our reductions is to see if the color terms vary over time. Provided that the transmission functions do not change, the color terms should never change. However, we must keep track of any variations of the color terms to verify that the natural system is stable.
6.1. Optical Photometry 6.1.1. Swope + SITe3 We define the transformation of the instrumental ugribv magnitudes into the natural system through the following equations: where u g r i BV correspond to magnitudes in the standard system. The color terms ( λ ) are measured in the manner described below. The magnitudes of the secondary photometric standards of Landolt (1992) and Smith et al. (2002) are thus used to calculate new magnitudes of these stars in the natural photometric system of the Swope telescope using the above equations.
On photometric nights, we can solve for these color terms based on observations of the secondary standards. To do this, we fit the instrumental magnitudes with the following equations: Note that these equations differ slightly from those defined in Eqs. 1-6 of Hamuy et al. (2006) in that the colors on the right-hand side of the equations are in the standard system and not the instrumental system.
The calibration strategy adopted by the CSP-I for the optical imaging obtained with the Swope telescope was to observe a minimum of eight secondary standard stars over a range of airmass during one photometric night every week. During the course of the CSP-I, different team members would use IRAF 19 tools and procedures to fit these observations to Eqs. 15-20 to obtain the nightly extinction coefficients, color terms, and zero points for each band. For this final data release, we have redone the nightly measurements of the extinction coefficients, color terms, and zero points in a uniform manner using a more sophisticated, noninteractive method that accounts for outliers and provides more realistic error bars. In detail, we used a Mixture Model Markov Chain Monte Carlo (MCMC) fitting procedure (as in Hogg, Bovy, & Lang 2010), which includes a photometric model, a Gaussian model for the outliers, an extra variance term, and a q parameter accounting for the fraction of the data points that fit the photometric model. The MCMC modeling is specified as follows: is the photometric model for the observed instrumental magnitudes, and is the Gaussian normal distribution model for outliers. The seven-parameter likelihood function, L(k λ , λ , ζ λ , σ 2 e , q, µ, σ), is then expressed as where N is the number of standard-star observations in one photometric night in one filter.
In this model, m i and σ i correspond to instrumental magnitude and error bars; k λ , λ , and ζ λ are the nightly extinction coefficients, color terms, and zero points (respectively) for filter λ; C λ is the color associated with the color term; and σ is the standard deviation of the Gaussian error distribution (for outliers) centered on µ. The extra variance, σ 2 extra , is an additional error term added to every single measurement. This is necessary because a single bright secondary standard star typically has an uncertainty due to photon statistics of only a few millimagnitudes, while the zero-point dispersion for a good night is no better than 0.01 mag. Finally, q represents the fraction of the data that fits the photometric model, while 1 − q is the fraction that can be considered as outliers. A handful of nights with values of q < 0.8 in different filters was discarded as likely to have been nonphotometric. Figure 4 displays nightly values of the atmospheric extinction coefficients in ugriBV for the Swope + SITe3 camera derived with this MCMC model over the five CSP-I campaigns.
Histograms of the collected extinction-coefficient measurements are shown in the right-hand part of each panel. Figure 5 shows a similar plot of the color terms over the five campaigns. In neither of these figures is there evidence for significant secular changes in the extinction coefficients or color terms.
The nightly photometric zero points for the Swope + SITe3 camera are shown in Figure 6. The obvious zig-zag pattern arises from the accumulation of dust and aerosols between the two washings of the primary mirror (marked by the red arrows) that occurred during the CSP-I observing campaigns. Smaller dips in sensitivity are observed around mid-February 2006 (JD 2,453,780) and mid-March 2008 (JD 2,454,540). Similar dips are visible during the summer months in the zero points measured by Burki et al. (1995) between November 1975 and August 1994 at the neighboring La Silla Observatory, and we speculate that these are associated with an increase in atmospheric haze that occurs due to the inversion layer being generally higher at that time of the year. Interestingly, these dips do not appear to be accompanied by significant changes in the extinction coefficients and color terms.
The demonstrated stability of the nightly extinction coefficients and color terms over all five CSP-I campaigns justifies adopting average color terms and extinction coefficients for the final photometric reductions. This reduces the problem to solving for nightly zero points only. In this way, just a handful of secondary standard star observations is needed to calibrate the natural photometry for the local sequence tertiary standards observed during the same night. The final mean extinction coefficients and color terms adopted for the five CSP-I campaigns are given in Table 4. Using these mean color terms, natural-system magnitudes for the Smith et al. (2002) and Landolt (1992) secondary standards were calculated via Eqs. 9-14. These, in turn, were used to derive magnitudes in the natural system of the local sequences of tertiary standards in each of the SN fields.
Final u g r i BV magnitudes of the local sequences of tertiary standards for all 134 SNe are listed in Table 5. Note that these are given in the standard system (i.e., as calculated using Eqs. 9-14) in order to facilitate their usage by others. In all cases, these magnitudes are based on observations made on at least three different photometric nights, and the accompanying uncertainties are weighted averages of the errors computed from these multiple measurements. 20 As discussed in Paper 2, on 14 January 2006 (unless otherwise noted, UT dates are used throughout this paper; JD 2,453,749) the original V filter ("LC-3014") used at the Swope telescope was broken and subsequently replaced by another V filter ("LC-3009"). However, after a few nights of use, it was determined that the replacement filter had a significantly different color term compared to the original. This filter was replaced on 25 January 2006 (JD 2,453,760) with a third filter ("LC-9844"), which was used for the remainder of the CSP-I campaigns. Although the bandpass of the LC-9844 filter is slightly broader than that of the LC-3014 filter (see Figure 3), observations at the telescope as well as synthetic photometry showed the color terms to be the same to within ∼ 0.002. Hence, we adopted the same natural magnitudes of the local sequences of tertiary standards for observations made in both of these filters. However, the color term of the LC-3009 filter was sufficiently different that we have treated separately the reduction of the smaller number of observations obtained with this filter. Table 4 gives the mean color term for V -band transformations for the LC-3014 and LC-9844 filters, whose value is −0.058. For the small amount of V -band photometry obtained with the LC-3009 filter, we assume the color term of −0.044 derived in Paper 2.

du Pont + Tek5
As already mentioned, owing to its larger aperture and better delivered image quality, the 2.5 m du Pont telescope was used during the CSP-I to obtain host-galaxy reference images in ugriBV using the facility Tek5 CCD camera. A small amount of SN followup imaging was also obtained with this telescope/instrument combination. Unfortunately, precise measurements of the filter response functions with the Tek5 camera were not carried out, and this camera has since been decommissioned. Nevertheless, it is possible to estimate the color terms of this system using the local sequences of tertiary standards (established with the Swope + SITe3 camera) in the fields of the SNe observed with the du Pont + Tek5 camera.
To carry out this experiment we chose two objects, SNe 2007ab and 2008O, that were observed in both the Swope + SITe3 and du Pont + Tek5 systems. Both SNe are at relatively low Galactic latitudes with many foreground stars in their fields. Natural-system magnitudes in the ugriBV bandpasses were measured for the 100 brightest stars in each field using all of the Swope + SITe3 images calibrated by the respective local sequence of tertiary standards. The range in (B − V ) colors covered by these stars was +0.2 to +1.5 mag for SN 2007ab, and +0.4 to +1.2 mag for SN 2008O.
SN 2007ab was observed on one night with the du Pont + Tek5 camera, and SN 2008O on four nights. Instrumental magnitudes were measured for the same 100 field stars in each of the images taken on these nights, and differences (∆m) were calculated with respect to the Swope+SITe3 natural system magnitudes: If the response functions for a given filter are identical between the Swope SITe3 and du Pont Tek5 cameras, we would expect ∆m to be a constant. On the other hand, if the response functions are significantly different, we would expect to detect a relative color term as well.
We therefore analyzed the observations by fitting the model Here, the color C λ is in the natural system and depends on the filters as per Eqs. 15-20. For example, ∆B = b (B − V ) + ζ b for the B band.
For the griBV filters, we find that the color term is within 1-2σ of zero. The color term for the u filter is also consistent with zero to ∼ 2σ, but these results are of lower confidence since this filter was utilized only one night for each SN.
Based on these results, it is justified to assume that the SN photometry obtained in the griBV filters with the du Pont + Tek5 camera is on the same natural system as the Swope + SITe3 camera. It also seems likely that any difference between the u bandpasses is small. We have therefore opted to calibrate the SN photometry obtained with the du Pont + Tek5 camera using the natural system tertiary standard star magnitudes, mean extinction coefficients, and mean color terms measured with the Swope + SITe3 camera.

Swope + RetroCam
The Swope + RetroCam Y JH bandpasses are shown in Figure 3. On the night of 2008 December 8 (JD 2,454,808), the observer detected a change in the J-band dome flat-field images, suggesting either contamination or that the filter might be starting to delaminate. The decision was taken to replace the suspect filter, and this was accomplished approximately a month later. The last observations made with the original filter, which we will refer to as "J RC1 ", were obtained on 2009 January 2 (JD 2,454,833). Observations with the replacement filter, which we call "J RC2 ", began on 2009 January 15 (JD 2,454,846). It was eventually determined that the change in the J RC1 filter was due to contamination, and this problem affected the J RC1 observations obtained between JD 2,454,808 and 2,454,833. Although we have no evidence that the contamination significantly changed the bandpass of the J RC1 filter and have therefore included these observations in this paper, we caution the reader that the reliability of these observations is less certain than that of the other J-band photometry published in this paper.
In Papers 1 and 2, we neglected any color terms that might exist in transforming J and H measurements made by the CSP-I to the P98 photometric system. In order to check this assumption, we have reproduced the P98 bandpasses by combining the filter transmission data and typical NICMOS3 quantum-efficiency curve given by these authors with two aluminum reflectivity curves (one for the primary and another for the secondary mirror) and an atmospheric transmission spectrum typical of LCO. The resulting response functions are plotted in red in Figure 3.
The (J − H) colors of the P98 secondary standard stars used by the CSP-I to calibrate both the Swope and du Pont NIR observations range from only +0.19 to +0.35 mag. This is too small of a color range to measure the NIR color terms, and so we must resort to synthetic photometry of model atmospheres to estimate these. We downloaded the Castelli & Kurucz (2003) atmosphere models for a range of stellar parameters. For each model spectrum, we then computed synthetic photometry for a range of reddenings (E(B − V ) = 0.0 to 2.5 mag), and plotted the differences between the P98 magnitudes and the RetroCam and WIRC J and H magnitudes as a function of the (J − H) P98 color (see Figure 7). Linear fits to these data yield the following: In Appendix C, we present du Pont + RetroCam observations of P98 standards covering a much wider range of colors that validate the accuracy of this procedure.
Although the effect of the color terms in Eqs. 26-28 is less than 0.01 mag over the small range of color of the P98 standards, it is a systematic effect and so we use them to transform the P98 magnitudes to the natural system.
The Y photometric band was introduced by Hillenbrand et al. (2002). Hamuy et al. (2006) calculated synthetic (Y − K s ) and (J − K s ) colors from Kurucz model atmosphere spectra using the estimated filter response functions for the Magellan 6.5 m Baade telescope "PANIC" NIR imager. These values were fitted with a fifth-order polynomial with the requirement that (Y −K s ) = 0.0 when (J −K s ) = 0.0 mag, consistent with the definition that α Lyr (Vega) has magnitudes of zero at all NIR wavelengths (Elias et al. 1982). This relation was then used to compute Y -band magnitudes from J and K s for all of the P98 secondary standards. In Appendix D, we repeat this exercise using the measured Swope + RetroCam Y -band response function along with the J and K s filter response functions we have derived for the P98 standards.

du Pont + WIRC
Color terms for the du Pont + WIRC system were calculated from synthetic photometry of the Castelli & Kurucz (2003) stellar atmosphere models in the manner described previously for the Swope + RetroCam. We find We note that these color terms are nearly identical to those for the J RC2 and H RC filters (cf. Eqs. 27 and 28). Figure 3 shows that the du Pont + WIRC Y bandpass cuts off more rapidly at blue wavelengths than is the case for the Swope + RetroCam Y filter. We again employ synthetic photometry to evaluate the color term required to transform the du Pont + WIRC Y -band secondary standard star observations to the Swope + RetroCam system. This gives Over the range of (J − H) colors of the local sequence tertiary standards, this color term is too large to be ignored. This means that for the Y band, we must work in two different natural systems: that of the Swope + RetroCam, which we adopt as the "standard" system, and that of the du Pont + WIRC.

NIR Natural-System Photometry
From the above, we conclude that the color terms for the J RC2 and J WIRC filters are sufficiently similar that we can average them and, therefore, create a single natural system for all of the tertiary standards and SNe observed. Likewise, the color terms for the H RC and H WIRC filters are nearly identical, and the photometry obtained with them can also be considered on the same natural system. However, the J RC1 color term differs considerably with respect to those of the other two J filters, and therefore defines its own natural system. Likewise, the color term for the Y WIRC filter compared to Y RC is too large to be ignored.
We therefore adopt the following equations to transform the secondary standard star magnitudes to the natural systems in Y JH for the Swope + RetroCam: where the Y RC magnitudes are taken from Appendix D, and the J P98 and H P98 magnitudes are from P98.
For each photometric night where secondary standard stars were observed, the NIR photometric equations are then simplified to where y, j, and h correspond to the instrumental magnitudes; and k y , k j , and k h are extinction coefficients. Figure 8 displays nightly values of the atmospheric extinction coefficients in Y JH over the five CSP-I campaigns for both the Swope + RetroCam and du Pont + WIRC systems. These were derived using the MCMC fitting procedure described in §6.1.1 from observations of typically 2-10 secondary standards per night. Histograms of the collected extinction-coefficient measurements are shown at the right-hand side of each panel. No significant difference is observed between the two telescope + camera systems, so we can combine the observations. The resulting mean extinction coefficients are given in Table 4.
As was found to be the case for the optical bandpasses, the stability of the extinction coefficients during the five CSP-I observing campaigns is such that these average values can be adopted, leaving only the nightly zero points in Eqs. 37-39 to be determined.
Thirteen SN fields were not observed for the requisite minimum of three photometric nights. In order to improve the photometric calibration of the tertiary standards for these fields, we devised a "hybrid" calibration whereby calibrated tertiary standards from one field are used to calibrate the tertiary standards in another field that is observed on the same night under photometric conditions, but when secondary standards were not observed. In this case, we use a modified version of Eq. 5: where k now refers to calibrated tertiary standards of different SN fields observed in filter λ on the same photometric night (j).
In brief, this procedure worked as follows.
• A catalog of tertiary standard stars was produced from 126 SN fields (90 SNe Ia and 36 SNe of other types) calibrated on a minimum of four photometric nights in each of the three NIR filters.
• This catalog of tertiary standards was then used to measure an alternative set of zero points for each night of NIR imaging during the five CSP-I campaigns.
• These new zero points were then filtered to include only those nights where (1) a minimum of three SN fields with calibrated tertiary standards was observed, (2) a minimum continuous span of three hours of imaging was obtained, and (3) a maximum dispersion of 0.08 mag in the nightly zero point as calculated from the observations of the tertiary standards was observed that night. This latter criterion is similar to that used in filtering the photometric nights chosen for calibrating the tertiary standard stars using the P98 secondary standards, but it should be noted that the typical dispersion in zero points for photometric nights was significantly less (0.02-0.03 mag).
In Figures 9 and 10, the zero points calculated using only the P98 secondary standard stars are plotted as a function of time for the du Pont + WIRC and the Swope + RetroCam, respectively. Shown for comparison are the zero points obtained using the hybrid method described above. Note that the agreement is generally excellent, although the uncertainties in the zero points derived in the hybrid method are generally larger since the local sequence stars are typically 3-4 mag fainter than the P98 secondary standards. The hybrid method provides potential photometric zero points for an additional 52 nights in Y , 44 nights in J, and 40 nights in H for the du Pont + WIRC observations, and an additional 154 nights in Y , 139 nights in J, and 123 nights in H for the Swope + RetroCam data. Nevertheless, we used only those nights that allowed us to improve the calibration of the thirteen SN fields.

Filter Contamination
Close inspection of Figure 10 reveals a faster-than-expected change in the zero-point evolution of the RetroCam Y and H filters during campaign 3, producing large breaks between the end of campaign 3 and the beginning of campaign 4. These discontinuities do not correspond to when the primary mirror was washed (indicated by the vertical gray lines in Figure 10). A similar problem is observed with the H band during campaign 4, where the zero point decreases by nearly 1 mag between the washing of the primary mirror and the end of the campaign, as opposed to the much smaller changes observed for the Y and J filters over the same period. This behavior suggests slowly increasing contamination of the filters. To test this hypothesis, we plot with dashed blue lines in Figure 10 the dates that the RetroCam dewar was warmed up, pumped, and then cooled down again. The recovery of the Y and H zero-point values in campaign 3, and the H zero point in campaign 4 is seen to coincide with when the dewar was pumped, consistent with the contamination hypothesis. Comparison of flat fields taken during campaigns 3 and 4 provide further evidence for slowly changing contamination seen as a radial pattern of increasing counts from the center to the edges of the filter that disappears when the dewar is pumped.
To examine the effect of this changing contamination on the photometry, we used observations of stars in the fields of several SNe at low Galactic latitudes. The observations of the Type IIn SN 2006jd ) from both campaigns 3 and 4 were used, supplemented by observations of the Type Ia SNe 2007hj and 2007on, and the Type II SNe 2008M and 2008ag carried out during campaign 4. Figure 11 shows an example of the magnitude differences in the Y , J, and H-band photometry of stars in the field of SN 2008ag between images obtained on 20 October 2007 and 5 April 2008. For the Y and J bands in campaign 4, the magnitude differences are consistent with zero over the entire detector. In contrast, the H filter shows clear evidence of a radial gradient amounting to ∼ 0.014 mag per 100 pixels as measured from the center of the detector. However, as the lower-right panel of Figure 11 shows, there is no evidence that the filter bandpass itself was changed significantly by the contamination since both blue and red stars show the same radial-gradients. Y and H-band observations in campaign 3 show a similar radial gradient effect.
Unfortunately, these contamination problems were not recognized during the course of the CSP-I campaigns. For most SNe, the error in the photometry due to the contamination is relatively small (0.02-0.03 mag), but systematic; we must therefore correct for the effect. Fortunately, the growth of the contamination was nearly linear in time. This is illustrated in Figure 12 where magnitude differences in H-band photometry of stars in the field of SN 2008ag are plotted at five epochs between 20 February 2008 and 20 June 2008 with respect to observations made on 16 February 2008. Fitting these trends with straight lines provides a recipe for correcting the SN and tertiary standard photometry, with the correction being a function of both time and the (x, y) coordinates of the SN or standard on the RetroCam detector. Specifically, we fit the slope measurements as a function of time by the relation where JD is the Julian date of the observation and p(JD end ) is the slope (measured in units of mag per 100 pixels) on the Julian date at the end of the period of contamination, JD end . The formula for calculating the correction to the photometry of the tertiary standards and SN in an image take on any Julian date during the period of contamination is then where d radial is the radial distance in pixels from of the star or SN from the center of the image, and the constant 256/100 makes the average of the magnitude corrections for each image approximately zero.
These corrections were applied to photometry obtained with the Swope + RetroCam as follows:  Table 6. Note that the J and H magnitudes are given in the standard P98 system, whereas the Y magnitudes are in the natural system of the Swope + RetroCam (which we have adopted as the "standard" system). The accompanying uncertainties are the dispersions of the multiple measurements of each sequence star.

Final Light Curves
Final optical and NIR photometry of the 134 SNe in the CSP-I sample is given in Tables 7-12. Tables 7 and 8 give the ugriBV photometry in the natural systems of the Swope + SITe3 and du Pont + Tek5 (respectively), and Table 9 gives the small amount of Vband photometry obtained in the natural system of the LC-3009 filter at the Swope. NIR photometry of 120 SNe in the natural systems of the Swope + RetroCam and du Pont + WIRC is found in Tables 10, 11, and 12. On those occasions when more than one NIR measurement is given for an object on a given night, it is because the WIRC used on the du Pont telescope images the SN location on more than one chip. Rather than averaging the measurements, we give the individual values.

Type Ia SNe
Plots of the individual light curves of the Type Ia SNe 21 in the CSP-I sample are displayed in Figure 13 along with fits (solid red lines) using SNooPy (Burns et al. 2011). Photometric parameters derived from the SNooPy fits are provided in Table 3. In some cases we can directly measure the epoch of B-band maximum, T max (B), and the B-band decline rate, ∆m 15 (B). The latter is defined as the number of magnitudes the object faded in the first 15 days since the time of B-band maximum, and has long been known to correlate with the absolute magnitudes of SNe Ia at maximum light (Phillips 1993). Often, however, photometric coverage is not optimal for direct measurements, and it is more robust to estimate T max (B) and the decline rate using a family of light-curve templates. Hence, for each object in Table 3, we have used the "max model" method of SNooPy to calculate template-derived estimates of the epoch of B-band maximum and the decline-rate parameter, ∆m 15 (B), that we denote as T max (template) and ∆m 15 (template), respectively. We also give the dimensionless "stretch BV parameter," s BV , which is equal to ∆T BV /(30 d), where ∆T BV is the number of days since T (B max ) that a supernova's (B − V ) color reaches its maximum value (Burns et al. 2014). Burns et al. discuss the advantages of this new parameter over the usual decline-rate parameter, especially for rapidly declining light curves. In particular, plots of reddening-corrected colors vs. s BV show low root-mean-square scatter, allowing a more definitive characterization of the photometric properties of SNe Ia.
One "normal" SN Ia listed in Table 3 that we cannot fit using just CSP-I photometry is SN 2006dd, because the CSP-I data cover only the post-maximum linear decline. We refer the reader to Stritzinger et al. (2010), which contains pre-maximum, maximum-light, and post-maximum photometry of this SN obtained with the CTIO 1.3 m telescope using its dual optical/NIR imager ANDICAM.
The middle panel of Figure 2 shows a histogram of the values of the B-band decline rate, ∆m 15 (B), as obtained from the template fits. The bottom panel of this figure shows a histogram of "stretch BV " values.

Type Iax SNe
Type Iax SNe are spectroscopically similar to Type Ia SNe that are more luminous than average because they show high-ionization lines such as Fe III, but have lower maximum-light velocities and fainter absolute magnitudes for their light-curve decline rates (Foley et al. 2013). The prototype of this subclass is SN 2002cx (Filippenko 2003;Li et al. 2003). Plots of the individual light curves of the five SNe Iax SNe in the CSP-I sample (SNe 2005hk, 2008ae, 2008ha, 2009J, and 2010ae) are displayed in Figure 14. Preliminary photometry of SN 2005hk was published by Phillips et al. (2007). Stritzinger et al. (2014Stritzinger et al. ( , 2015 have published updated photometry of SN 2005hk and new photometry of SNe 2008ha and 2010ae.

Other Subtypes
Two objects observed by the CSP-I, SNe 2007if and 2009dc, are candidates for the super-Chandrasekhar ("SC") subtype (Howell et al. 2006). SNe 2006bt and 2006ot are members of the SN 2006bt-like subclass (Foley et al. 2010b, Paper 2). Two other events (SNe 2005gj and2008J) belong to the rare Type Ia-CSM subtype (Silverman et al. 2013). The light curves of these six SNe are shown in Figure 15.

Conclusions
In this paper we have presented the third and final data release of optical and NIR photometry of the 134 nearby (0.004 z 0.08) white dwarf SNe observed during the CSP-I. This sample consists of 123 Type Ia SNe, 5 Type Iax SNe, 2 super-Chandrasekhar candidates, 2 Type Ia-CSM SNe, and 2 SN 2006bt-like events. NIR photometry was obtained for 90% of these SNe. Optical spectroscopy has already been published for approximately two-thirds of the SNe Ia in the sample, and the remaining spectra are currently being prepared for publication. In addition to providing a new set of light curves of low-redshift SNe Ia in a stable, well-characterized photometric system for cosmological studies, the combined CSP-I dataset is allowing us to improve dust extinction corrections for SNe Ia (Burns et al. 2014). The excellent precision and high cadence of the CSP-I observations also facilitate detailed analysis of the light curves, leading to a deeper understanding of the physics of thermonuclear events (e.g., Höflich et al. 2010Höflich et al. , 2017. Over the course of the CSP-I, more than 100 core-collapse SNe were observed. Photometry of 7 SNe IIn has already been presented by Stritzinger et al. (2012) and Taddia et al. (2013). In an accompanying paper to this one , the final data release of optical and NIR photometry of 34 stripped-envelope core-collapse SNe is presented. Publication of optical and NIR photometry of 83 SNe II observed during the course of the CSP-I is also in preparation. Preliminary V -band light curves for this sample have already been published by Anderson et al. (2014), and CSP-I observations of two SN 1987A-like events were presented by Taddia et al. (2012b). Extensive optical spectroscopy of many of these core-collapse SNe was also obtained, and is currently being prepared for publication.
In 2011, we began a second phase of the CSP to obtain optical and NIR observations of SNe Ia in the smooth Hubble flow. Over a four-year period, light curves were obtained for nearly 200 SNe Ia, ∼ 100 of which were at 0.03 z 0.10. NIR spectra were also obtained of more than 150 SNe Ia. This dataset, which we plan to publish over the next three years, in combination with the CSP-I light curves published in the present paper, should provide a definitive test of the ultimate precision of SNe Ia as cosmological standard candles.

Electronic Access
To obtain an electronic copy of the photometry of any of the SNe included in this paper, the reader is directed to the CSP website at http://csp.obs.carnegiescience.edu/. (At the time of the posting of this preprint the tar ball of the photometry was not yet posted at this website.) Also available at this website are the optical spectra of CSP-I SNe Ia published by Folatelli et al. (2013). This paper is dedicated to the memory of our dear colleague, Wojtek Krzeminski (1933Krzeminski ( -2017, who played an important role in the early history of Las Campanas Observatory and who, during his retirement, obtained many of the observations presented in this paper. The CSP particularly thanks the mountain staff of the Las Campanas Observatory for their assistance throughout the duration of our observational program, and to Jim Hughes and Skip Schaller for computer support. Special thanks are due to Allyn Smith and Douglas Tucker for allowing us to publish their u g r i magnitudes of P177D and P330E (and to Dan Scolnic for leading us to Allyn and Douglas). This project was supported by NSF under grants AST-0306969, AST-0908886, AST-0607438, and AST-1008343. M. Stritzinger We thank the Lick Observatory staff for their assistance with the operation of KAIT. LOSS, which discovered many of the SNe studied here, has been supported by many grants from the NSF (most recently AST-0908886 and AST-1211916), the TABASGO Foundation, US Department of Energy SciDAC grant DE-FC02-06ER41453, and US Department of Energy grant DE-FG02-08ER41563. KAIT and its ongoing operation were made possible by donations from Sun Microsystems, Inc., the Hewlett-Packard Company, AutoScope Corporation, Lick Observatory, the NSF, the University of California, the Sylvia & Jim Katzman Foundation, and the TABASGO Foundation. We give particular thanks to Russ Genet, who made KAIT possible with his initial special gift, and the TABASGO Foundation, without which this work would not have been completed.

A. CSP-I Near-Infrared Bandpasses
Paper 2 provided a detailed description of the calibration of the CSP-I optical bandpasses. The setup consisted of a monochromator that allowed the throughput of the entire telescope plus detector system to be measured in situ without having to rely on multiple calibrations (filters, windows, aluminum reflections, detector quantum efficiency) multiplied together. Here we provide a summary of the calibration of the NIR bandpasses using the same monochromator.
The calibration of the two NIR cameras used for CSP-I was carried out in late-July 2010 (Swope + RetroCam) and early-August 2010 (du Pont + WIRC). The measurements were made on at least two different nights for each filter to ensure that the method was repeatable. The monochromator system uses a fiber that splits, sending 90% of the power to the dome-flat screen and 10% to a "witness screen." The witness screen was placed in a dark box to prevent ambient light from reaching it.
Two germanium detectors were used, each 10 mm in diameter, which were calibrated in the lab at Texas A&M University using a NIST traceable Gentec calibrated photodiode. The Ge detectors are sensitive from 900 to 1600 nm, and were used only for the calibration of the Y -band filter (900 to 1100 nm). To calibrate the three longer-wavelength CSP-I bandpasses (J, H, and K s ), a 2 mm diameter InGaAs detector was purchased and shipped to the National Research Council of Canada for calibration prior to its use in Chile.
Due to the lower light levels produced by this system in the NIR and because of the smaller area of the InGaAs detector, all of the actual calibrations were done using the 35 cm by 35 cm witness screen made of the same material as the dome-flat screen. When taking Y -band data, one Ge detector detected photons from the dome-flat screen while the other Ge detector simultaneously detected photons from the witness screen in the dark box. The two detectors in the box were placed about 10 cm from the witness screen.
The Y -band signal at the witness screen was about 50 times stronger than the signal from the dome-flat screen. The voltage of the Ge photodiode (as a function of wavelength) measured from the dome-flat screen and scaled by a factor of ∼ 50 matches, within the errors of measurement, the voltage of the other Ge photodiode used to measure the witness screen. Thus, we are confident that the witness screen can give reliable results for the longer-wavelength bandpasses.
A challenge inherent to IR measurements is the presence of background thermal drifts occurring on timescales of seconds. Also, during the day and on nights with moonlight, even with the dome closed there is some ambient light in the dome, which is relevant to the Y -band calibration described here.
To minimize the problem of thermal drifts and residual light in the dome, we took data using the following method. For each wavelength we obtained two 30 s "Dark" images and two 30 s "Light" images. A "Dark" image is taken with the light off and a "Light" image is taken with the light on. We subtract one "Dark" image from one "Light" and average the two net values to get a single measurement. Additionally, RetroCam on the Swope telescope exhibited severe image persistence, so we took a short (6 s) "Dark" exposure after each "Light" one to clear the camera detector.
The photodiode detectors are not temperature stabilized, so the amplifier offsets and the background drift significantly during the 30 s required for an exposure. They are relatively stable on a 1 s timescale. The error caused by the drift is much more important than the noise in the detector. It is then better to take shorter integration times (∼ 1 s) to avoid drift problems even if we sacrifice a bit of averaging of the noise.
Since the output of the lamp is very stable (< 1%) over a period of hours, we assumed that the output was constant over the 30 s exposure time and only measured the amplitude just before opening the shutter to take a camera image. Before each exposure we cycled the light on and off 5 times for a 10 s period to obtain 5 values of the amplitude, which we averaged to calculate the power at the detector.
The normalized transmission curves for each filter were obtained using a subsection (x and y from 60 to 964 pixels) of chip 2 in WIRC on the du Pont telescope. For RetroCam on the Swope telescope we used the same subsection of its chip. In general, we are confident that the measured transmission is accurate to 2-3% of the peak transmission.
We have normalized each filter for each camera separately. We estimate that the transfer from the dome-flat screen to the witness screen and the relative photodiode calibration uncertainties are 1% and 2%, respectively, for the Swope and du Pont cameras. For the Swope + RetroCam, the relative photodiode noise level is 0.5% for the Y JH filters, as is the relative noise level of the camera. The total uncertainty in each filter is obtained from the square root of the sum of the individual components added in quadrature, or 2.3%. For the du Pont + WIRC, the relative photodiode noise level and the relative noise level of the camera is 1-2% for Y JH. For these filters the total uncertainty is 2.6% in Y and J, and 3.2% in H.
We have also investigated the focal-plane uniformity of the filters in the Swope and du Pont cameras. Each filter was scanned twice on at least two separate nights. The scans were performed with a wavelength step of 5 or 10 nm. The analysis involved dividing each chip into four quadrants and comparing the relative response curves for each quadrant. At worst there is a 1.5 nm shift in the filter cutons and cutoffs as a function of quadrant and chip. This effect would be negligible unless a very narrow emission line happens to fall at that exact wavelength.

B. Optical Color Terms
To test the accuracy of the response functions of the CSP-I optical bandpasses shown in Figure 3, we computed color terms using a subset of the stars from the spectrophotometric atlas of Landolt standards published by Stritzinger et al. (2005). Eighteen of the stars in this atlas are in the list of Smith et al. (2002) u g r i standards. Of these, one star (SA98-653) does not have sufficient wavelength coverage to include the u and g bands, and we suspect that another (SA104-598) is variable. The results of synthetic photometry of the sixteen remaining stars are shown in Figure 16. In each plot, the abscissa is the color from the published standard-star magnitudes, and the ordinate is the difference of the magnitude calculated via synthetic photometry using the bandpasses in Figure 3 and the published standard-star magnitude. The red triangles correspond to the 16 stars in the Stritzinger et al. (2005) atlas, and the dashed lines are the best fits to these points. The slopes of these fits correspond to the color terms, and the values are indicated in red. The slopes of the observed color terms are indicated by the solid black lines, with the values shown in black.
The histograms in each plot correspond to the colors of the Smith et al. (2002) and Landolt (1992) standard stars observed routinely by CSP-I.
In general, the agreement between the measured and computed color terms is good. For the u, r, i, B, and V bands, the color terms agree to better than 1σ. For the g band, the agreement is within 2σ. Considering the relatively small number of stars used for this test, and the fact that these cover a somewhat smaller range of color than the actual standards used routinely by CSP-I, we consider the results of this test to be consistent with the observations.

C. NIR Color Terms
In principle, we can check the color terms derived in §6.2.1 for the J RC2 and H filters of the Swope + RetroCam through observations of the red stars listed in Table 3 of P98. Unfortunately, although some of the red stars were observed on a few nights during the CSP-I campaigns, these data were not reduced at the time they were taken, and a subsequent disk crash made it impossible to recover them. However, since mid-2011, RetroCam has been in use on the du Pont telescope, and observations of several of the P98 red stars were made in late-December 2015. Figure 17 shows synthetic synthetic photometry in J RC2 and H of the Castelli & Kurucz (2003) atmosphere models for a range of stellar parameters and reddenings (E(B − V ) = 0.0 to 2.5 mag). Here the differences between the P98 magnitudes and the RetroCam magnitudes are plotted as a function of the (J − H) P98 color. The filter response functions used for these calculations were measured in 2013 November for RetroCam on the du Pont telescope using the same monochromator and setup described in Appendix A. The predicted color terms of RetroCam on the du Pont telescope and on the Swope telescope are very similar, as might be expected. The black symbols in Figure 17 are our observations of several P98 standards, along with three stars with (J − H) P98 > 1.5 mag taken from the P98 red-star list that correspond to reddened M giants (typically M3 III) in Bok globules in the Coal Sack. The three red stars plotted with red symbols are young stellar objects (YSOs) from the P98 list. The latter stars are not representative of the Castelli & Kurucz (2003) models used for calculating the synthetic photometry, and are also often variable. 22 Hence, we do not include these stars in the fits shown in Figure 17.
Comparing the synthetic photometry of the models with the observations (excluding the YSOs), we find excellent agreement between the predicted and observed color terms in J and H. This gives us confidence in the reliability of the color terms calculated for the Swope + RetroCam in Eqs. 27 and 28 of §6.2.1.

D. Y -Band Photometric Standards
Hamuy et al. (2006, Appendix C) describe how we derived Y -band magnitudes for the NIR standards of P98. In brief, we used Kurucz (1991) model atmospheres, the P98 filter functions, J P98 and K P98 , and our best estimate of the transmission of our Y filter to compute synthetic (Y − K s,P98 ) colors as a function of synthetic (J P98 − K s,P 98 ) colors. Since then, we have scanned our NIR filters (see Appendix A) and improved stellar atmosphere models (Castelli & Kurucz 2003) have become available. Hence, in this Appendix we rederive Yband magnitudes for the NIR standards of P98. Note that the photometric zero points for the NIR filters are computed assuming Vega colors are all zero. Figure 18 shows the results for the Swope + RetroCam Y band. The grey circles correspond to synthetic photometry of model dwarf-star atmospheres with nearly solar metallicity (log(g) > 4.0, −0.1 < log(Z/Z ) < 0.1, where g is the local acceleration of gravity in cm s −2 and Z is the abundance of elements heavier than helium). The gray region indicates the color range spanned by the P98 standards used by the CSP-I. The green solid line is a fifth-order polynomial fit to the data, The blue dashed line is the fit from Hamuy et al. (2006). As with Hamuy et al. (2006), we do not allow a constant offset, forcing the polynomial through Y − K s,P98 = J P98 − K s,P98 = 0 mag. However, we note that the synthetic colors of the (Castelli & Kurucz 2003) model atmospheres do not have zero NIR colors for an A0 V star. This can be seen in the lower panel of Figure 18, where we have subtracted a linear fit (solid red line) to the points in order to better visualize the difference between the old and new fits. We could adjust the NIR zero points to make all synthetic NIR colors zero at the expense of Vega acquiring nonzero colors, but choose not to do this in order to be more consistent with our previous natural system.
In Table 13 we give the final Y -band photometric values of most of the P98 standard stars. We include values for stars considerably further north than we can observe at Las Campanas Observatory, in case northern hemisphere observers require them.

E. Photometric Zero Points
The zero points of a photometric system are necessary for computing transformations to other photometric systems (commonly referred to as S-corrections) as well as computing cross-band K-corrections within the same photometric system. The definition of the magnitude of a source with SED f λ measured by an instrument with response F λ is given by where f * λ is some standard SED (e.g., Vega) and m * is its magnitude through the instrument defined by F λ . Here and in Eq. E2 below c and h are the speed of light and Planck's constant, respectively. The numerator within the log function is the observed photon flux detected by the CCD, while the denominator is the photon flux of the standard SED through the same instrument and is generally not observed. Defining the zero point as ζ = m * + 2.5 log 10 1 ch F λ f * λ λdλ , we clearly need three pieces of information to compute the zero point: the total instrument response, the standard SED, and the magnitude m * . As mentioned previously, the CSP-I has directly measured all components of F λ except the atmosphere. This leaves the standard SED and value of m * .
The CSP-I used three sets of secondary standards to calibrate our photometry: Landolt (1992) for BV , Smith et al. (2002) for ugri, and P98 for Y JH. These standards, more than anything else, define our photometric zero point. However, we also require a high-fidelity SED that covers the wavelength range of our filters, and such SEDs generally do not exist for these standards. Stritzinger et al. (2005) give SEDs at optical wavelengths for 18 Landolt (1992) and Smith et al. (2002) standards. Ultimately, the Landolt (1992) and P98 standards are tied to α Lyr while the Smith et al. (2002) standards are tied to BD+17 • 4708, both of which have accurate SEDs (Bohlin & Gilliland 2004a,b), so we use these to compute our zero points.
Lastly, we need the value of m * for each instrument and SED combination. We begin with the standard magnitudes of each star in the system for which it was defined and use our color terms (see Table 4) to compute the magnitudes that would have been observed through our instrument. The adopted standard magnitudes and transformed natural-system magnitudes are listed in Table 14. For B and V , we adopt the standard magnitudes for α Lyr from Fukugita et al. (1996). For Y JH, we adopt zero magnitudes for α Lyr to be consistent with Elias et al. (1982). For ugri, we adopt the standard magnitudes of BD+17 • 4708 from Smith et al. (2002).
The reader should note that Bohlin & Landolt (2015) present evidence that BD+17 • 4708 is slightly variable. From 1986 to 1991 this star brightened by ∼ 0.04 mag in multiple optical bands. Following the suggestion of Bohlin & Landolt (2015), we have also calculated zero points in ugriBV using the standard and natural-system magnitudes of the CALSPEC standards P177D and P330E. For B and V , we use the values measured by Bohlin & Landolt (2015), while for ugri, we adopt unpublished measurements made by Allyn Smith in the USNO u g r i z standard system of Smith et al. (2002).
In Table 15 we give corresponding zero points calculated with the SEDs of the four primary calibration standards. The agreement in zero points between P177D, P330E, and α Lyr for B and V is excellent, as is also the case between P177D, P330E, and BD+17 • 4708 for ugri.
It is worth pointing out that we have two different networks of standard stars in the optical. As such, there is no guarantee that the BV zero points will be consistent with the ugri zero points in an absolute sense. To investigate this, we compute synthetic photometry of the Pickles (1998) stellar library and compare the (B − g) and (V − g) colors with the observed colors of stars common to Landolt (1992) and Smith et al. (2002). The results, shown in Figure 19, indicate a systematic offset between the synthetic and observed (B − g) and (V −g) colors of ∼ 0.03 mag. This could easily be fixed by adding the appropriate offsets to the zero points of the B and V filters, thereby bringing both systems into alignment. However, this introduces a reliance on the Pickles (1998) spectra and presupposes that it is the B and V magnitudes that should be adjusted, when the problem could just as easily be with the ugri zero points. Hence, we prefer not to apply a zero-point correction to either system. Nevertheless, users of the CSP-I photometry should be aware of this inconsistency.

F. AB Magnitude Offsets
According to Eq. 7 of Fukugita et al. (1996), a broad-band AB magnitude is defined as m AB = −2.5 log 10 d(log 10 ν)f ν F ν d(log 10 ν)F ν − 48.6 , where f ν is the flux per unit frequency of the object expressed in units of ergs s −1 cm −2 Hz −1 , and F ν is the response function of the filter. Since AB magnitudes are directly related to physical units (Oke & Gunn 1983), they offer a straight-forward way of transforming magnitudes to flux densities.
To convert from natural magnitudes to AB magnitudes, we need to solve for an offset for each filter such that Combining equations F1 and F2 with the definition of our CSP natural magnitudes, it can be shown that offset = 16.847 − ζ + 2.5 log 10 F λ dλ λ , where ζ is the zero point of the filter. The zero points of the CSP natural magnitudes are derived in Appendix E and are listed in Table 15. Table 16 shows offsets calculated with equation F4 for all of the CSP-I filters . Once the offsets have been applied, the flux in each band is given by: f ν ν = 10 −0.4(m AB +48.6) erg s −1 cm −2 Hz −1 . (F5) where f ν ν is the weighted average of f ν with weight function F (ν) · ν −1 . Note that equation F5 is not the proper inverse of equation F1. One cannot derive a precise monochromatic flux from an AB magnitude, especially for objects such as supernovae that have SEDs very different from the stars used in the fundamental spectrophotometric system. This is discussed in detail by Brown et al. (2016).   b Type Ia subtypes from SNID (Blondin & Tonry 2007). Members of Type Iax and Ia+CSM subtypes are listed as "Iax" and "Ia+CSM", respectively; super-Chandrasekhar candidates are denoted by "SC"; SN 2006bt-like are listed as "06bt-like".
c Wang et al. (2009) classification. N = normal, HV = high velocity, 91bg = SN 1991bg-like. d CN = core normal, CL = cool, SS = shallow silicon, BL = broad line (Branch et al. 2006). e Number of nights with optical photometry.   b See Eqs. 9-14 or 15-20 for which standard colors are used in combination with these coefficients to obtain the color correction terms for the optical photometry. V -band photometry obtained with the LC-3009 filter and given in Table 9 (029) Note. - Table 5 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content.
a All photometry is measured in magnitudes. Values in parentheses are 1σ uncertainties. This photometry of field stars is in the two standard systems, not the natural system.  (095) 0.000 (000) Note. - Table 6 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content. a All photometry is measured in magnitudes. Values in parentheses are 1σ uncertainties. The JH photometry presented here is in the P98 system. The Y -band photometry is in the natural system.  (006) Note. - Table 7 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content.     (034) Note. -Table 10 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content. a The photometry presented here was obtained with the J-band filter RC1, which was used prior to 15 January 2009. All photometry is measured in magnitudes. Values in parentheses are 1σ uncertainties.  (068) 0.000(000) 0.000 (000) Note. -Table 11 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content. a The photometry presented here was obtained with the J-band filter RC2, which was used after 15 January 2009. All photometry is measured in magnitudes. Values in parentheses are 1σ uncertainties.  (000) Note. -Table 12 is published in its entirety in the electronic edition of the journal. A portion is shown here for guidance regarding its form and content. a All photometry is measured in magnitudes. Values in parentheses are 1σ uncertainties.        The optical bandpasses shown in the upper portion of the figure were measured using a monochromator as described in Paper 2, and have then been multiplied by an atmospheric absorption and extinction spectrum typical of LCO for an airmass of 1.2. The NIR bandpasses shown in the lower half of the figure were determined with the same monochromator (see Appendix A for details), and also include atmospheric absorption appropriate for LCO. Note that three different V filters were used with the Swope telescope + SITe2 CCD camera during the course of the CSP-I (see §6.1.1 for details), and two J filters were utilized at the Swope with RetroCam (see §6.2.1).
The dashed red lines show the response functions for the J and H filters employed by P98. These were derived by combining the filter transmission and detector quantum efficiency data given by these authors with two aluminum reflections and an atmospheric transmission function appropriate to LCO.    color (green and red points, respectively). The differences between synthetic photometry in the natural system of the du Pont + WIRC J filter and the P98 J bandpass are also shown (blue points). The slopes give the J-band color terms (CT) for the Swope + RetroCam and the du Pont + WIRC systems using synthetic photometry derived from Castelli & Kurucz (2003) atmosphere models covering a range of stellar parameters and reddenings (E(B − V ) = 0.0 to 2.5 mag).   an ordinate (y-axis) range of 6 mag. Note that x = 0 corresponds to the time of B-band maximum. Best-fitting SNooPy fits using the "max model" mode are shown for each SN.
As the Y -band photometry obtained with two different cameras has different color terms, we color code the Y -band photometry obtained with the du Pont telescope and WIRC with orange points. Starting 15 January 2009 the J-band photometry was obtained with the Swope telescope and RetroCam using the J RC2 filter. Those data points are color coded orange. Note -all 123 light curves for this figure will be reproduced in the electronic edition of the journal.   The dashed lines are the best fits to the red triangles; the slope of each fit is listed numerically in the sub-plots using a red font. The slopes corresponding to the measured mean color terms (see Table 4) are indicated by the solid black lines; the numerical values of these are reproduced in the sub-plots using a black font. The histograms in each plot correspond to the colors of the standard stars observed routinely by the CSP-I.  Table 3 of P98 obtained with RetroCam on the duPont telescope. The differences between the P98 and observed magnitudes are plotted versus the (J − H) P98 color. The red stars plotted with black symbols correspond to reddened M giants in the Coal Sack, while those plotted with red symbols are YSOs. The solid and dashed lines are fits to the observations excluding and including (respectively) the YSOs. (Top:) The swath of grey points shows the differences between synthetic photometry in the natural system of the Swope + RetroCam J RC2 bandpass and the P98 J filter for Castelli & Kurucz (2003) atmosphere models covering a range of stellar parameters and reddenings (E(B − V ) = 0.0 to 2.5 mag), plotted versus the (J − H) P98 color. (Bottom:) Same, but for the H band. colors. The gray circles correspond to model dwarf star atmospheres with nearly solar metallicity. The red and green solid lines are linear and fifth-order polynomial fits to these data, while the blue dashed line is the relation given by Hamuy et al. (2006). The gray shaded region indicates the color range of the P98 standards used by the CSP-I.  Landolt (1992) and Smith et al. (2002).