1: %\documentstyle[10pt,emulateapj]{article}
2: %\documentstyle[12pt,aaspp4]{article}
3: %\documentclass{aastex}
4: %\documentclass[preprint2]{aastex}
5: \documentclass[12pt,preprint]{aastex}
6: %\documentclass[manuscript]{aastex}
7: %\singlespace
8: \def\p/{\mbox{$^1$}}
9: \def\pp/{\mbox{$^2$}}
10: \def\ppp/{\mbox{$^3$}}
11: \def\pppp/{\mbox{$^4$}}
12: \def\m/{\mbox{$^{-1}$}}
13: \def\mm/{\mbox{$^{-2}$}}
14: \def\mmm/{\mbox{$^{-3}$}}
15: \def\mmmm/{\mbox{$^{-4}$}}
16: \def\Ms/{\mbox{M$_\odot$}}
17: \def\ebv{\mbox{$E(4405-5495)$}}
18: \def\rv{\mbox{$R_{5495}$}}
19: \def\bt{\mbox{$B_T$}}
20: \def\vt{\mbox{$V_T$}}
21: \def\btvt{\mbox{$B_T-V_T$}}
22: \def\chimin{\mbox{$\chi^2_{\rm min}$}}
23: \newcommand{\sci}[2]{\mbox{$#1\cdot 10^{#2}$}}
24: \newcommand{\scipm}[3]{\mbox{$(#1\pm #2)\cdot 10^{#3}$}}
25: %stopzone
26: \slugcomment{Accepted at Publications of the Astronomical Society of the Pacific (June 2005 issue)}
27: \begin{document}
28:
29: \title{A cross-calibration between Tycho-2 photometry and HST spectrophotometry}
30: \shorttitle{A cross-calibration between Tycho-2 photometry and HST spectrophotometry}
31:
32: \author{J. Ma\'{\i}z Apell\'aniz\altaffilmark{1}}
33: \affil{Space Telescope Science Institute\altaffilmark{2}, 3700 San Martin
34: Drive, Baltimore, MD 21218, U.S.A.}
35: \email{jmaiz@stsci.edu}
36:
37: % ---------------affiliations of the science group -----------------------
38:
39: \altaffiltext{1}{Affiliated with the Space Telescope Division of the European
40: Space Agency, ESTEC, Noordwijk, Netherlands.}
41: \altaffiltext{2}{The Space Telescope Science Institute is operated by the
42: Association of Universities for Research in Astronomy, Inc. under NASA
43: contract No. NAS5-26555.}
44:
45: \begin{abstract}
46: I show that Tycho-2 photometry and HST spectrophotometry are accurate and stable enough to
47: obtain a precise cross-calibration by analyzing a well-calibrated sample of 256 stars observed
48: with both Hipparcos and HST.
49: Based on this analysis, I obtain the following photometric zero points in the Vega
50: magnitude system for Tycho-2: 0.020$\pm$0.001 (\btvt), 0.078$\pm$0.009 (\bt), and
51: 0.058$\pm$0.009 (\vt).
52: \end{abstract}
53:
54: \keywords{space vehicles: instruments --- stars: fundamental parameters ---
55: techniques: photometric --- techniques: spectroscopic}
56:
57: \section{Introduction}
58:
59: It is nowadays extremely common to analyze photometric observations by comparing them
60: with synthetic photometry derived from observed or theoretical spectral energy distributions
61: (SEDs). Such comparisons are done by establishing a magnitude system based on a reference
62: spectrum, be that of a real object, such as Vega, or a simple artificial spectrum, such as a
63: constant in $f_\lambda$ or $f_\nu$ \citep{synphot}. Each reference spectrum yields a different
64: magnitude system defined from the expression:
65:
66: \begin{equation}
67: m_{P} = -2.5\log_{10}\left(\frac{\int P(\lambda)f_{\lambda}(\lambda)\lambda\,d\lambda}
68: {\int P(\lambda)f_{\lambda{\rm,ref}}(\lambda)\lambda\,d\lambda}\right)
69: + {\rm ZP}_P,
70: \end{equation}
71:
72: \noindent where $P(\lambda)$ is the total-system sensitivity curve, $f_\lambda(\lambda)$ is the
73: SED of the object, $f_{\lambda{\rm,ref}}(\lambda)$ is the SED of the reference spectrum, and ZP$_P$
74: is the zero point for filter $P$. In order to attain accurate results when comparing measured
75: magnitudes or colors with SED models, it is necessary not only to have an accurate knowledge of
76: $f_{\lambda{\rm,Vega}}(\lambda)$ but also of $P(\lambda)$ and of ZP$_P$. For example,
77: \citet{BohlGill04} measured the spectrum of Vega and found that in the Johnson system (which
78: uses that star as a reference), ZP$_V$ = 0.026$\pm$0.008. One can also define a system where
79: ZP$_P$ = 0 for any filter, such as the VEGAMAG\footnote{Note that it is possible to derive
80: VEGAMAG magnitudes from SEDs using the Johnson throughputs but those values should be different
81: from published photometry due to the different zero points used.}, STMAG, or ABMAG systems
82: \citep{synphot}.
83:
84: The lack of an accurate
85: knowledge of the system sensitivities has plagued some ground-based systems such as Johnson's
86: $UBV$ \citep{Bessetal98}, mostly due to the use of different observing conditions, technologies,
87: and reduction techniques by different observers.
88: Modern space-based telescopes tend to be more uniform and stable and
89: usually produce better-quality photometry. ZP$_P$ is usually close to 0.0 but not exactly so,
90: the reason being the common use of
91: secondary calibrators (Vega is often too bright for a detector and/or
92: unaccesible from many sites to be observed).
93: ZP$_P$ thus needs to be measured \citep{Bessetal98,Coheetal03,BohlGill04}, since
94: assuming it is exactly zero can introduce systematic errors in the comparison between observed
95: magnitudes and reference SEDs.
96:
97: The Hipparcos mission \citep{ESA97} observed the full sky and yielded the Tycho
98: catalog, which is reasonably complete down to $V = 11.5$. The original Tycho catalog was
99: consequently reprocessed by \citet{Hogetal00a} to produce the Tycho-2 catalog, which contains
100: 2.5 million stars and is currently the most complete and accurate all-sky photometric survey in
101: the optical. The Tycho-2 catalog contains photometry in two optical bands, \bt\ and \vt,
102: whose sensitivities were analyzed by \citet{Bess00} and use Vega as the reference spectrum.
103: The careful processing of Tycho-2
104: photometry was described by \cite{Hogetal00b}, including the difffferent tests used to check for
105: possible systematic errors.
106:
107: HST spectrophotometry is calibrated using the method of \citet{ColiBohl94}, which
108: is based on a combination of Landolt $BV$ photometry, ground-based spectrophotometry, and SED
109: models. Four white dwarf stars observed with the Space Telescope Imaging Spectrograph (STIS)
110: are used as primary calibrators: G191B2B, GD71, GD153, and HZ43 \citep{Bohletal95}. The absolute
111: flux calibration has an accuracy of 4\% in the FUV and 2\% in
112: the optical \citep{Bohl00} but, given that the photometric repeatibility of STIS is
113: 0.2-0.4\% \citep{Bohletal01}, the relative flux calibration for colors derived from STIS spectra
114: is expected to be better than 2\% in the optical. The latest values for the
115: Advanced Camera for Surveys (ACS) photometric zero
116: points are also based on that same calibration \citep{DeMaetal04}.
117:
118: Given the usefulness of the Tycho-2 photometric database for a number of astronomical
119: studies and the accuracy of the absolute flux calibration of HST spectrophotometry, I
120: considered it important to calibrate the first with respect to the second one, as well as to check
121: the existence of possible problems in either of them. Ultimately, we want to obtain
122: the zero points for Tycho-2 magnitudes, ZP$_{B_T}$ and ZP$_{V_T}$, and color, ZP$_{B_T-V_T}$.
123:
124: \section{The data}
125:
126: In order to test the consistency of the Tycho-2 and HST calibrations we would ideally
127: need a large uniform sample of stars with \bt\ and \vt\ magnitudes that has also been observed
128: in the optical with an HST spectrograph. Such a sample has indeed been obtained for the Next
129: Generation Spectral Library (or NGSL, \citealt{Gregetal04}, see also
130: {\tt http://lifshitz.ucdavis.edu/\~{}mgregg/gregg/ngsl/ngsl.html}). Originally, the NGSL
131: intended to obtain low-resolution 1660-10\,200 \AA\ spectra for 600 stars with a wide range of
132: temperatures, gravities, and metallicities. Unfortunately, the failure of STIS in August 2004
133: ended the program before reaching that goal. However, 378 stars that are also included in the
134: Tycho-2 catalog had already been observed, and that number is large enough for our purposes.
135:
136: \begin{figure}
137: %\centerline{\includegraphics*[width=0.47\linewidth]{test5600n.ps}
138: % \includegraphics*[width=0.47\linewidth]{test5600c.ps}}
139: \centerline{\includegraphics*[width=0.47\linewidth]{fg1a.ps}
140: \includegraphics*[width=0.47\linewidth]{fg1b.ps}}
141: \caption{The 5600 \AA\ flux discontinuity as a function of \vt\ (photometric) without (left)
142: and with (right) CTE correction for the 378 stars in the NGSL sample. A single object is
143: located off the scale shown. The dashed lines indicate the limits considering for defining the
144: sample used to derive ZP$_{B_T-V_T}$}
145: \label{jump5600}
146: \end{figure}
147:
148: Each star was observed using three different STIS gratings, G230LB, G430L, and G750L.
149: Here, only the latter two will be used since those are the ones that span the Tycho-2 filters. The
150: NGSL spectra were obtained using the 52$\times$0.2 slit without a peak-up, which is not the ideal
151: configuration for absolute spectrophotometric calibration, the reason being that the slit is
152: narrow enough that if the star is not well centered, the aperture correction used may not be
153: accurate. In principle, an off-center positioning could also cause problems with the
154: relative spectrophotometric calibration if the aperture correction was strongly dependent on
155: wavelength due to variations in the PSF\footnote{Also, due to diffraction off the slit edge.}.
156: Fortunately, for G430L and G750L that is only a strong effect at the long-wavelength
157: part of the spectrum beyond \vt\ \citep{stis}. Therefore, the relative fluxes (and hence the
158: derived colors) should not be affected as long as the star did not shift
159: position significantly during the visit. Given that all exposures for a given star were obtained
160: in a single orbit (usually not filling it completely) and that the typical drift rate for HST
161: is less than 10 mas/hr \citep{stis}, indeed one would expect a shift of less than 10\% of a
162: STIS CCD pixel in the relative position of the star with respect to the slit between the first
163: and the last exposure. There are also two calibration issues that need to be addressed regarding
164: the fact that all exposures were obtained at the E1 position using a CCD subarray. The use of
165: a subarray requires the definition of a background region closer to the star than for a
166: measurement with a full array; I checked that this selection did not affect the measured
167: fluxes. Also, the location of the E1 position, closer to the readout point in the detector,
168: minimizes CTE effects but the CTE correction algorithm itself \citep{BohlGoud03} was not
169: originally designed to work with CCD subarrays, so a test should be carried out to check for its
170: validity.
171:
172: In order to explore the possible problems with the spectrophotometric accuracy
173: mentioned in the previous paragraph, I first calculated the effect of the CTE correction on the
174: derived spectrophotometric \bt\ and \vt\ magnitudes by extracting the spectra with and without
175: the correction and comparing the results. The correction turned out to be very small,
176: with an average of $-$0.010 magnitudes for \vt, $-$0.018 magnitudes for \bt, and $-$0.008
177: magnitudes for \btvt. As a further
178: check, I measured the value of the discontinuity between the G430L and G750L gratings by
179: calculating the ratio between the two fluxes in the 5550-5650 \AA, region, where both gratings
180: overlap in coverage. Results are shown in Fig.~\ref{jump5600}, both without applying the CTE
181: correction and applying it. As it can be seen, in both cases the majority of the data points are
182: clustered around a zero value for the jump, indicating the lack of a systematic effect.
183: However, for the non-CTE-correction case a small slope as a function of \vt\ is present, which
184: indicates that a non-negligible correction is indeed required. For the case with CTE correction,
185: the data have no dependency with \vt\ and have a distribution with mean of 0.12\% and standard
186: deviation of 1.83\%.
187:
188: Given the results in the previous paragraph, I adopted the following criteria for
189: selecting the NGSL sample:
190:
191: \begin{itemize}
192: \item The CTE correction was applied.
193: \item Only stars with a 5600 \AA\ jump of less than 2.5\% (in absolute value) were included.
194: \item Given that the spectrophotometry and the photometry used in this article were obtained
195: in different epochs, I eliminated variable stars. The criterion used was the presence
196: of a variability flag in the Tycho photometry.
197: \end{itemize}
198:
199: With the restrictions above, our final NGSL sample was reduced to 256 objects.
200:
201: In order to provide a further check on the possible effects of systematic aperture and
202: CTE effects I used a control sample by analyzing the spectrophotometric standards of
203: \citet{Bohletal01} that have Tycho-2 photometry with magnitudes brighter than 12.0 and
204: uncertainties of 0.06 magnitudes or less. Those stars were observed with STIS using the
205: 52$\times$2 slit and with the full CCD array, so the problems mentioned above (slit centering
206: and validity of the CTE correction algorithm) should not be
207: relevant. Four of the \citet{Bohletal01} stars satisfy the above requirements for both
208: \bt\ and \vt\ (BD +17 4708, BD +28, 4211, BD +75 325, HD 93521) and an additional three
209: do so only for \bt\ (AGK + 81 266, Feige 34, HZ 44).
210:
211: The stars in both samples were analyzed using a synthetic photometry code created
212: for \citet{Maiz04c} and their spectrophotometric \bt\ and \vt\ magnitudes calculated.
213:
214: \section{Results}
215:
216: \begin{figure}
217: %\centerline{\includegraphics*[width=\linewidth]{B_T-V_T_test.ps}}
218: \centerline{\includegraphics*[width=\linewidth]{fg2.ps}}
219: \caption{Comparison between photometric and spectrophotometric \btvt\ colors as a function
220: of photometric \btvt\ for the two samples. The error bars represent the photometric
221: uncertainties and the horizontal line marks the proposed ZP$_{B_T-V_T}$.}
222: \label{btvtplot1}
223: \end{figure}
224:
225: \begin{figure}
226: %\centerline{\includegraphics*[width=0.47\linewidth]{B_T-V_T_histo.ps}
227: % \includegraphics*[width=0.47\linewidth]{B_T-V_T_histonorm.ps}}
228: \centerline{\includegraphics*[width=0.47\linewidth]{fg3a.ps}
229: \includegraphics*[width=0.47\linewidth]{fg3b.ps}}
230: \caption{Histograms for the comparison between photometric and spectrophotometric \btvt\ colors
231: for the NGSL sample. (left) Regular histogram. (right) Histogram for the data shifted by the
232: proposed ZP$_{B_T-V_T}$ and normalized by the individual uncertainties. A Gaussian distribution
233: with $\mu=0$ and $\sigma=1$ is overplotted for comparison.}
234: \label{btvtplot2}
235: \end{figure}
236:
237: I start by analyzing the relative flux calibration of our data. I show in
238: Fig.~\ref{btvtplot1} the difference between the photometric (from the Tycho-2 catalog) and
239: spectrophotometric (computed from STIS spectra) values for the \btvt\ of each of the stars in
240: our samples as a function of the photometric \btvt. No general trend is observed as a function
241: of color and the data are symmetrically distributed around a central value, which I take to be
242: ZP$_{B_T-V_T}$. I measured ZP$_{B_T-V_T}$ by calculating the weighted mean using
243: $1/\sigma^2_{B_T-V_T}$ as weights and found it to be 0.020$\pm$0.001 magnitudes.
244: I show the histogram for the \btvt\ data in Fig.~\ref{btvtplot2}, both in absolute and
245: relative (corrected for ZP$_{B_T-V_T}$ and dividing each point by its photometric uncertainty)
246: terms. The second histogram has a median of $1.2\cdot 10^{-5}$ and a standard deviation of 1.04
247: and the distribution is very well approximated by a normalized Gaussian. All of the above implies
248: that an accurate cross-calibration of colors vs. relative fluxes between Tycho-2 photometry and
249: HST spectrophotometry is possible in principle without having to invoke e.g. modifications in
250: the Tycho filter sensitivities or the STIS calibration. Furthermore, given that the normalized
251: histogram has a standard deviation only slightly larger than 1.0, the largest source of
252: deviations from the expected value originates in the photometry, not in the spectrophotometry.
253: Since the mean photometric $\sigma_{B_T-V_T}$ = 0.025 magnitudes, the accuracy of the
254: spectrophotometrically-derived Tycho-2 colors must be better than 1\%, which agrees with the
255: published value for the STIS photometric repeatibility \citet{Bohletal01}.
256:
257: Only four objects in the Bohlin sample have both \bt\ and \vt\ photometry.
258: Of course, such a low number is not enough to do
259: accurate statistics but we can see in Fig.~\ref{btvtplot1} that the results from that sample are
260: consistent with the derived ZP$_{B_T-V_T}$ and with the absence of a color dependence for the
261: cross-calibration. One can argue that the two stars with \btvt\ around $-$0.3 (BD +75 325 and
262: HD 93521) are slightly above the ZP$_{B_T-V_T}$ line and, hence, may indicate the beginning of a
263: small departure from the general rule for very blue objects. However, it can also be a
264: fluctuation caused by small-number statistics, which is reinforced by the reversal of the trend
265: in the bluest of the stars in the Bohlin sample, BD +28 4211. In any case, even if the effect
266: turned out to be real, it would be very small and affect only extremely blue stars.
267:
268: \begin{figure}
269: %\centerline{\includegraphics*[width=0.47\linewidth]{B_T_test.ps}
270: % \includegraphics*[width=0.47\linewidth]{V_T_test.ps}}
271: %\centerline{\includegraphics*[width=0.47\linewidth]{fg4a.ps}
272: % \includegraphics*[width=0.47\linewidth]{fg4b.ps}}
273: \centerline{\includegraphics*[width=1.00\linewidth]{fg4.ps}}
274: \caption{Comparison between photometric and spectrophotometric \bt\ (left) and \vt\ (right)
275: magnitudes as a function of the photometric values for the two samples. The error bars
276: represent the photometric uncertainties and the horizontal line marks the proposed
277: ZP$_{B_T}$ (left) and ZP$_{V_T}$ (right). The asymmetric scatter of values below the line is
278: due to light loss at the slit due to poor centering (see text for details).}
279: \label{btvtplot3}
280: \end{figure}
281:
282: Next, I deal with the absolute calibration of our data. I show in Fig.~\ref{btvtplot3}
283: the difference between the photometric and spectrophotometric values for both \bt\ (left panel)
284: and \vt\ (right panel). Once again, no significant trend is observable as a function of \bt\ or
285: \vt, thus enabling the cross-calibration of the two systems. However, a difference is readily
286: apparent between the relative and absolute plots for the NGSL sample: for the absolute case
287: the data are also concentrated around a central point but a tail
288: towards the negative in the distributions of $B_{T,\rm phot}-B_{T,\rm spec}$ and
289: $V_{T,\rm phot}-V_{T,\rm spec}$ is clearly present. Such a behavior is the expected one if some
290: of the stars are not well centered on the 52$\times$0.2 slit, as previously mentioned. Under
291: such circumstances it is not possible to derive a value for either ZP$_{B_T}$ or ZP$_{V_T}$
292: with a precision of the order of 0.01 magnitudes or better. Therefore, for that task I decided
293: to use instead the Bohlin sample, which was obtained using the wider 52$\times$2 slit. Given
294: the high precision achieved for ZP$_{B_T-V_T}$, it is a better strategy to measure just one of
295: the two absolute ZP values and then use the color zero point to obtain the other one.
296: Considering that I have more stars with \bt\ data than with \vt\ data (seven vs. four), I
297: choose the first one for the absolute flux calibration. A weighted mean for those seven stars
298: in the Bohlin sample yields ZP$_{B_T}$ = 0.078$\pm$0.009 magnitudes. Applying ZP$_{B_T-V_T}$,
299: we then obtain ZP$_{V_T}$ = 0.058$\pm$0.009 magnitudes. As it can be seen in
300: Fig.~\ref{btvtplot3}, those values are compatible with the observed distributions in
301: $B_{T,\rm phot}-B_{T,\rm spec}$ and $V_{T,\rm phot}-V_{T,\rm spec}$ for the NGSL
302: sample if one discards the negative-values tail.
303:
304: I have obtained a precise cross-calibration between Tycho-2 photometry and HST
305: spectrophotometry which, combined with the independent absolute flux calibrations of both
306: instruments, should allow for accurate comparisons between Tycho-2 photometry and synthetic
307: photometry generated from SED models. Furthermore, the use of the zero points presented here
308: should also allow combinations of Tycho-2 photometry with that of other surveys with precise
309: zero points, such as 2MASS \citep{Coheetal03}.
310:
311: \acknowledgments
312:
313: I would like to thank Rodolfo Barb\'a for fruitful conversations on this topic and Ralph Bohlin,
314: Santiago Arribas, and an anonymous referee for useful comments on the manuscript.
315:
316: \bibliographystyle{apj}
317: \bibliography{general}
318:
319: \end{document}
320:
321: