astro-ph0112447/ms.tex
1: \documentclass[12pt, preprint]{aastex}
2: 
3: \newcommand{\mum}{\ifmmode{\mu m}\else{$\mu$m}\fi}
4: \newcommand{\iso}{{\em ISO}}
5: \newcommand{\iras}{{\em IRAS}}
6: 
7: \begin{document}
8: \title{Artifacts at 4.5 and 8.0 \mum\ in Short Wavelength Spectra from the
9: {\em Infrared Space Observatory}\footnote{Based on observations with
10: the {\em Infrared Space Observatory (ISO)}, an European Space Agency
11: (ESA) project with instruments funded by ESA Member States (especially
12: the Principal Investigator countries: France, Germany, the Netherlands,
13: and the United Kingdom) and with the participation of the Institute
14: of Space and Astronautical Science and the National Aeronautics and
15: Space Administration (NASA).}
16: }
17: 
18: \author{Stephan D. Price\altaffilmark{2}, G. C. 
19: Sloan\altaffilmark{3,~4},
20: Kathleen E. Kraemer\altaffilmark{2,~5}}
21: 
22: \altaffiltext{2}{Air Force Research Laboratory, Space Vehicles 
23: Directorate,
24:            29 Randolph Rd., Hanscom AFB, MA 01731-3010; 
25: 	steve.price@hanscom.af.mil, kathleen.kraemer@hanscom.af.mil}
26: \altaffiltext{3}{Institute for Scientific Research, Boston College,
27:            140 Commonwealth Ave., Chestnut Hill, MA 02467-3862;
28: 	sloan@ssa1.arc.nasa.gov}
29: \altaffiltext{4}{Infrared Spectrograph Science Center, Cornell 
30: University,
31:            Ithaca, NY 14853-6801}
32: \altaffiltext{5}{Institute for Astrophysical Research, Boston University,
33:            Boston, MA 02215}
34: 
35: \begin{abstract}
36: Spectra from the Short Wavelength Spectrometer 
37: (SWS) on \iso\ exhibit artifacts at 4.5 and 8~$\mu$m.  These 
38: artifacts appear in spectra from a recent data release, OLP 10.0, 
39: as spurious broad emission features in the spectra 
40: of stars earlier than $\sim$F0, such as $\alpha$ CMa.  Comparison of 
41: absolutely calibrated spectra of standard stars to corresponding 
42: spectra from the SWS reveals that these artifacts result from an 
43: underestimation of the strength of the CO and SiO molecular bands 
44: in the spectra of sources used as calibrators by the SWS. Although
45: OLP 10.0 was intended to be the final data release, these
46: findings have led to an additional release addressing this issue, OLP 10.1, 
47: which corrects the artifacts.
48: 
49: \end{abstract}
50: 
51: \keywords{astronomical data bases: miscellaneous --- methods: data
52: analysis --- techniques: spectroscopic}
53: 
54: 
55: \section{Introduction \label{sec.intro}}
56: 
57: The Short Wavelength Spectrometer (SWS) on the {\em Infrared Space
58: Observatory (ISO)} obtained approximately 1250 spectra covering the 
59: full 2.4--45~$\mu$m wavelength range at moderate resolution.  We are 
60: engaged in an ongoing project to classify these spectra \citep{ksp01}, 
61: to reprocess them, and to present them in a publically available 
62: database on-line (Sloan et al. 2002, in preparation).  SWS data are 
63: currently available in a partially processed form called Auto-Analysis 
64: Results (AARs).  The AARs have a fairly complex format \citep{lee01} 
65: that requires further processing before the data can be used for 
66: scientific analysis.  Our processing  method reduces these data to a 
67: single continuous 2.4 to 45~$\mu$m spectrum.
68: 
69: In assessing the quality of our new reprocessing algorithm, 
70: we compared the results for several infrared standards observed with 
71: the SWS to absolutely calibrated spectra from Cohen et al. (1992a, 
72: 1992b, 1995, 1996a, 1996b, 2001 (in preparation)).  These standards 
73: are based on synthetic spectra of the A0 dwarf $\alpha$ Lyr and the 
74: A1 dwarf $\alpha$ CMa, which serve as the reference standards in the
75: system.  Cohen et al. (1992a) describe the details of their method that
76: used high-quality ground-based and airborne photometry to normalize the
77: synthetic spectra to measured astronomical fluxes.  The synthetic
78: spectra are based on the models of \cite{kur79}, with updated
79: opacities and metallicities \citep{ck94}.
80: 
81: Secondary standards are added to the system by observing their spectra 
82: in conjunction with that of the primaries so that atmospheric, 
83: telescopic, and instrumental transients can be removed.  The spectra
84: for the secondary standards are then obtained by dividing the observed
85: secondary by the observed reference standard and multiplying by the
86: assumed spectrum (the model) for the reference, degraded to match the
87: spectral resolution of the instrument used:
88: 
89: \begin{eqnarray}
90:   S_{b,final}(\lambda) =
91:   \frac{S_{b,obs}(\lambda)}{S_{a,obs}(\lambda)}
92:  S_{a,assumed}(\lambda),
93: \label{eqn.1}
94: \end{eqnarray}
95: 
96: \noindent where the subscript $a$ refers to the reference standard,
97: and the subscript $b$ refers to the secondary.  This is the standard
98: method used by spectroscopists at ground-based telescopes to 
99: calibrate a program source by ratioing its spectrum to that of a 
100: standard star, preferably secured at an airmass matching that of the
101: target.  In effect, Equation \ref{eqn.1} transfers the 
102: quality of the synthetic A star model from the reference to the new 
103: standard, whatever its spectral type.
104: 
105: Cohen et al. (1992b, 1995, 1996a, 1996b, 2001) have applied this
106: method to create absolutely calibrated composite spectra for 13
107: infrared standards, using spectra from ground-based telescopes, the
108: Kuiper Airborne Observatory (KAO) and the {\em Infrared Astronomical
109: Satellite (IRAS)} \citep[further details of the process can be found in 
110: ][]{cwb92a, cww92b}. \cite{cwb96a} added $\alpha^{1}$ Cen (G2 V) as an
111: alternative reference standard for the southern hemisphere to give a
112: total of 3 reference standards as well as the 13 secondaries.  Most
113: of the secondary standards are giants with spectral classes later
114: than K0 and were chosen for their intrinsic brightness; sources later 
115: than approximately M3 are avoided due to the possibility of  their 
116: variability \citep{eg97}.
117: 
118: The preferred infrared reference standard is Sirius ($\alpha$ CMa),
119: due to its brightness and its dust-free spectrum beyond 20~$\mu$m.
120: Figure 1 compares our derived SWS spectrum of this source to the
121: Kurucz model presented by Cohen et al. (1992a).  Deviations between
122: the two occur in the vicinity of 4.5~$\mu$m and again at 8.0~$\mu$m.
123: We propose an explanation for the origin of this discrepancy in this 
124: Letter and discuss the implications for the calibration of the SWS 
125: database and the impact on calibration of future infrared missions.
126: 
127: 
128: \section{Data Format and Analysis \label{sec.data}}
129: 
130: Data from the SWS are publically available in AAR format.  This
131: format corrects the spectra, as far as possible, for the relative
132: spectral response of the detectors, differences in gains between
133: individual detectors and a variety of other issues, as described by
134: \cite{lee01}.  The files are organized in 12 spectral segments,
135: ranging from 1 to 4 for each of the 4 different detector bands.
136: Each spectral segment contains interleaved spectra from the 12 
137: individual detectors, scanned in the direction of both increasing 
138: and decreasing wavelength, giving a total of 24 separate spectra 
139: for each spectral segment.  Thus the processing of the full 
140: wavelength range from an AAR file requires the combination of 288 
141: individual spectra into one.
142: 
143: We start with data from the Off-line Processing
144: (OLP) pipeline 10.0, available in June 2001, which was intended to be 
145: the final release. This version includes the latest attempts to remove
146: the memory effect from the Si:As data (4--12~$\mu$m) (Kester et al.
147: 2001).  Although these attempts have significantly improved the data,
148: residual effects remain, and one can see that the spectra presented
149: here tend to diverge from the expected result as the data approach
150: 12~$\mu$m.  However, we have high confidence in the quality of the
151: data at shorter wavelengths, where the flux levels from the sources
152: are much stronger and where the discrepancies between the SWS data
153: and the reference spectra occur.
154: 
155: Sloan et al. (2002) will present details of the algorithm used to
156: combine the 288 individual spectra into one.  We have tested for the
157: possibility that the algorithm might be responsible for the artifacts
158: seen in Figure \ref{fig.1} by processing the AAR data manually with 
159: the Infrared Spectral Analysis Package (ISAP) following the standard 
160: procedures.  The two methods produce similar results, which indicates 
161: that the problem is intrinsic to the spectrum released by \iso\ and 
162: not to a specific method of processing.  
163: The artifact peaking at 4.5~$\mu$m also appears in Figures 6 and 7 of 
164: \cite{dec01}, although the wavelength stretch makes it more difficult 
165: to notice.
166: 
167: \begin{figure*}
168: \plotone{f1.ps}
169: \caption{A comparison of SWS data for $\alpha$ CMa ({\it dotted
170: line}) with the Kurucz model used by Cohen et al. 
171: (1992a) ({\it solid line}).  The spectra have been normalized between 
172: 6 and 7~$\mu$m and are plotted in units of $\lambda^4 F_{\lambda}$, 
173: which clearly shows deviations from the reference spectrum at the CO 
174: and SiO bands.  The bottom panel plots the difference between the
175: two as a percentage of the flux density of the Kurucz model.}
176: \label{fig.1}
177: \end{figure*}
178: 
179: 
180: \section{Discussion \label{sec.disc}}
181: 
182: The divergence between the SWS spectrum of $\alpha$ CMa and the
183: Kurucz model is characterized by two broad ``emission'' features, at
184: $\sim$4--5~\mum\ and $\sim$8~\mum.  (The rise at longer wavelengths
185: is caused by the memory effect mentioned in Sec. \ref{sec.data}.)
186: The feature at 8~$\mu$m is reminiscent of the 8~$\mu$m emission
187: artifact found in the data from the Low Resolution Spectrometer (LRS)
188: aboard \iras\ by Cohen et al. (1992b).  The LRS database was
189: originally calibrated assuming that $\alpha$ Tau could be
190: characterized as a $10^4$ K blackbody.  However, since the spectrum
191: of $\alpha$ Tau (a K5 III) contains significant absorption from the
192: SiO fundamental band at 8~$\mu$m, this error propagated to the entire
193: LRS database, producing an apparent emission feature in the spectra
194: of $\alpha$ CMa, $\alpha$ Lyr, and other early-type stars.
195: 
196: Fifteen stars have served as spectral calibrators for the SWS 
197: \citep{sch96,lee01,shi01}.  Ten of them are K and M giants, which
198: show strong molecular absorption bands in their spectra.  The
199: dominant bands peak at wavelengths of $\sim$2.4~\mum\ (CO overtone),
200: $\sim$4.1~$\mu$m (SiO overtone), $\sim$4.5~$\mu$m (CO fundamental,
201: blended with the weaker SiO overtone), and $\sim$8.0~$\mu$m (SiO
202: fundamental).  For the assumed spectrum of each of these stars in OLP 10.0, 
203: the SWS instrument team chose to use synthetic models of the stars from 
204: 2 to 12~$\mu$m. For longer wavelengths, a composite spectrum 
205: from Cohen et al. where available or a template spectrum from 
206: \cite{cwc99} where no composite existed were used.  The two sets of
207: data were spliced together at 12~$\mu$m \citep{sch96,lee01,shi01}. 
208: Previous pipeline releases used the Cohen composites and templates for the
209: entire wavelength range.
210: 
211: The K5 giant $\gamma$ Dra is a typical example of an SWS calibrator.
212: Figure \ref{fig.2} compares the synthetic spectrum of this source
213: \citep{dec00,dec01} to the observationally calibrated composite
214: spectrum \citep{cwc96b}.  The differences between the two spectra 
215: (the lower panel) are similar to the artifacts seen in Figure 
216: \ref{fig.1}.  \cite{dec01} compares the composite spectra of several 
217: other SWS calibrators to the synthetic spectra (in her Figure 9), 
218: and most of them show similar deviations\footnote{\cite{dec01} notes the 
219: discrepancy between the model and the composite spectrum of 
220: $\gamma$ Dra, but incorrectly attributes it to the substitution of data of 
221: $\alpha$ Tau for $\gamma$ Dra in this wavelength regime.  
222: In the header to the file containing the composite spectral data for 
223: $\gamma$ Dra, \cite{cwc96b} state that they calibrated this spectral 
224: region using spectral ratios of $\gamma$ Dra to $\alpha$ Boo 
225: obtained on the KAO.}.  It is highly unlikely 
226: that an instrumental effect (such as the memory effect) would 
227: manifest itself precisely at the wavelengths of these photospheric 
228: absorption bands.
229: 
230: The coincidence of the emission artifacts apparent in Figure 
231: \ref{fig.1} with the molecular bands in K and M stars and the
232: similarity of the artifacts to the deviations between the model 
233: and composite spectrum in Figure \ref{fig.2} lead us to 
234: hypothesize that the artifacts result from an underestimate of the 
235: depth of these bands in the synthetic spectra.
236: 
237: Figure \ref{fig.3} compares the data from the SWS to both the
238: model and composite spectrum of $\gamma$ Dra.  Although not perfect, 
239: {\em the SWS spectrum, as calibrated in OLP 10.0,  matches the 
240: model of $\gamma$ Dra significantly better than the Cohen composite}.  
241: Additional spectra of $\gamma$ 
242: Dra from the SWS database, taken at different times during the 
243: mission and at different spectral resolutions, as well as spectra
244: of other cool calibrators, produce similar results.  
245: 
246: If the synthetic spectra used to calibrate the SWS generally 
247: underestimate the strength of the molecular bands, then this 
248: miscalibration should propagate to the entire SWS database.  As 
249: Figure \ref{fig.4} shows, the artifacts at 4.5 and 8~$\mu$m are 
250: readily seen in bright spectra from early-type stars, including
251: the reference standard $\alpha$ Lyr, which should show neither 
252: emission nor absorption from molecular bands.  In the spectrum
253: of the average giant with no composite spectrum for comparison,
254: the effect of the miscalibration will be more subtle and more
255: difficult to recognize, since it will only reduce the apparent
256: depth of the absorption bands.
257: 
258: In Figure \ref{fig.3} the difference between the SWS data and the 
259: composite spectrum for $\gamma$ Dra exhibits the largest excess 
260: between $\sim$4.0 and 4.7~\mum.  While an excess clearly exists at 
261: these wavelengths in the spectra of the hot stars in Figure
262: \ref{fig.4}, the confidence level of the magnitude of this excess
263: shown in Figure \ref{fig.3} is not high.  Because the atmosphere is
264: almost opaque in the core of the CO$_2$ band at 4.3~\mum, even at
265: aircraft altitudes,  the uncertainty in the spectral composite is 
266: typically $\sim$6\% from 4.22 to 4.58 \mum\  and peaks at 10\% at 
267: 4.22 \mum\ \citep{cwc96b}. Roughly half of the composites, including
268: $\gamma$ Dra, have a data gap at $\sim$4.2-4.4 \mum\ due to the 
269: atmospheric band.  The spectral artifact is, however, much broader than
270: the data gap and larger than the uncertainties can account for.
271: The uncertainties in the composite spectrum  at the other feature are 
272: $\la$2\%.
273: 
274: To reduce the noise from a single comparison, we  calculated
275: the weighted average of the ratios of all of the spectra common
276: between the sources used to spectroscopically calibrate the SWS
277: and the composite and reference spectra from Cohen et al.  This
278: weighted mean, displayed in Figure \ref{fig.5}, provides an
279: estimate for the modifications needed in the OLP 10.0 data.  It also
280: provides an assessment of the reliability of the synthetic spectra
281: relative to the absolutely calibrated composite
282: spectra.
283: 
284: In light of our findings, the SWS team has revised their calibration
285: strategy and the \iso\ Data Center has agreed to an additional data
286: release, OLP 10.1.  By basing the new spectral calibration only on 
287: stars earlier than K, the worst of the discrepancy between the 
288: synthetic spectra and the Cohen spectra should be mitigated. The 
289: Decin models agree with the Cohen composites, templates, or Kurucz 
290: models (as appropriate) for the five stars (two As, one F, and two Gs) 
291: which meet this criterion (R. Shipman 2001, private communication).
292: 
293: 
294: 
295: 
296: 
297: \begin{figure*}
298: \plotone{f2.ps}
299: \caption{A comparison between the model of $\gamma$ Dra used to
300: calibrate the SWS data ({\it thin line}) and the composite spectrum
301: calibrated by Cohen et al. (1996b) ({\it thick line}).  The spectra
302: have been normalized between 6 and 7~$\mu$m and the model resampled to
303: match the composite.  The data for the model have been adapted from 
304: Fig. 8 and 9 in \cite{dec01}.}
305: \label{fig.2}
306: \end{figure*}
307: 
308: \begin{figure*}
309: \plotone{f3.ps}
310: \caption{Comparing SWS data for $\gamma$ Dra ({\it dotted line}) with 
311: calibrator spectra.  The upper panel is as in Fig. 2, but with a 
312: spectrum from the SWS superimposed ({\it dotted line}).  The lower 
313: panel compares the difference between the SWS data and the model ({\it 
314: thin line}) to the difference between the SWS data and the composite
315: from Cohen et al. (1996b) ({\it thick line}).}
316: \label{fig.3}
317: \end{figure*}
318: 
319: \begin{figure*}
320: \plotone{f4.ps}
321: \caption{Several SWS spectra of bright early-type naked stars.
322: Most have been multiplied by a constant (given in parentheses).
323: The bottom spectrum is the Kurucz model of $\alpha$ Lyr used by
324: \cite{cwb92a}.}
325: \label{fig.4}
326: \end{figure*}
327: 
328: \begin{figure*}
329: \plotone{f5.ps}
330: \caption{The weighted mean of the deviations between the SWS
331: spectra of the standard stars in common with the composites
332: and models of Cohen et al.  Both the SWS data and the
333: composites were resampled to 0.1~\mum\ spacing and placed on 
334: the same wavelength grid. Error bars are one sigma.}
335: \label{fig.5}
336: \end{figure*}
337: 
338: 
339: \section{Conclusions}
340: 
341: We have identified spectral artifacts in the vicinity of the CO
342: and SiO bands of late-type standard stars that appear in all OLP 10.0
343: spectra.  These artifacts constitute a known systematic bias over
344: a fairly large spectral range, and removing them is relatively
345: straightforward. While OLP 10.0 was to have been
346: the final version, an additional release, OLP 10.1, has been made
347: to correct the database based on our findings.
348: 
349: The SWS calibrations between 2 and 12~$\mu$m were predominantly based on 
350: stellar atmospheric models of cool stars.  These models have been used 
351: with great success in analysis of physical properties of cool stars 
352: \citep[e.g.][]{dwe00}.  Model spectra of A stars, which contain only 
353: atomic lines and no molecular bands, appear to be well founded. Synthetic
354: spectra can also achieve higher spectral resolution than that afforded by
355: the composites. However, 
356: the artifacts in the SWS database that mirror the dominant molecular 
357: absorption features in the 4--10~$\mu$m range indicate that synthetic 
358: spectra of cool giants require further progress before they can be 
359: used for definitive spectral calibration. 
360: 
361: The Infrared Spectrograph (IRS) on the upcoming Space Infrared
362: Telescope Facility (SIRTF) faces issues similar to those encountered by
363: the SWS team, with the added complication that all of the commonly
364: used spectral standards will be far too bright for use on the IRS.
365: Until the commission of the Stratospheric Observatory for Infrared
366: Astronomy (SOFIA), it will prove difficult to observationally
367: calibrate standards faint enough for use by the IRS.  Until this
368: happens, the IRS will need to rely on models for spectral calibration.
369: If models of later stars are required, improvements in the production 
370: of synthetic spectra will also be necessary.
371: 
372: 
373: \acknowledgements
374: We wish to thank Russ Shipman for many useful discussions regarding 
375: the calibration process.  Thijs de Graauw provided support for this
376: work through the USAF/SRON MOU.   One author (KEK) was supported by 
377: the National Research Council via a Research Associateship through 
378: the Air Force Office of Scientific Research.  This research has made 
379: use of NASA's Astrophysics Data System Abstract Service.
380: 
381: \clearpage
382: \begin{thebibliography}{}
383: 
384: \bibitem[Castelli \& Kurucz (1994)]{ck94} Castelli, F., \& Kurucz, R.
385: L. 1994, \aap, 281, 817
386: 
387: \bibitem[Cohen et al. (2001)]{cmh01} Cohen, M., Megeath, T.G.,
388: Hammersley, P.L., Martin-Luis, F., \& Stauffer, J. 2001, \aj, submitted
389: 
390: \bibitem[Cohen et al. (1992a)]{cwb92a} Cohen, M., Walker, R. G.,
391: Barlow, M. J., \& Deacon, J. R. 1992a, \aj, 104, 1650 
392: 
393: \bibitem[Cohen et al. (1999)]{cwc99} Cohen, M., Walker, R. G.,
394: Carter, B., \& Noguchi, K. 1999, \aj, 117, 1864
395: 
396: \bibitem[Cohen et al. (1992b)]{cww92b} Cohen, M., Walker, R. G., \&
397: Witteborn, F. C. 1992b, \aj, 104, 2030
398: 
399: \bibitem[Cohen et al. (1996a)]{cwb96a} Cohen, M., Witteborn, F. C.,
400: Bregman, J. D., Wooden, D. H., Salama, A., \& Metcalfe, L. 1996a,
401: \aj, 112, 241
402: 
403: \bibitem[Cohen et al. (1996b)]{cwc96b} Cohen, M., Witteborn, F. C.,
404: Carbon, D. F., Davies, J. K., Wooden, D. H., \& Bregman, J. D. 1996b,
405: \aj, 112, 2274
406: 
407: \bibitem[Cohen et al. (1995)]{cww95} Cohen, M., Witteborn, F.C. ,
408: Walker, R. G., Bregman, J. D., \& Wooden, D.H. 1995, \aj, 110, 275
409: 
410: \bibitem[Decin (2000)]{dec00} Decin, L. 2000, Ph. D. Thesis,
411: University of Leuven
412: 
413: \bibitem[Decin (2001)]{dec01} Decin, L. 2001, in The Calibration
414: Legacy of the ISO Mission, ed. L. Metcalf \& M. F. Kessler, (ESA SP-481), in press
415: 
416: \bibitem[Decin et al. (2000)]{dwe00} Decin, L., Waelkens, C.,
417: Eriksoon, K., Gustafsson, B., Plez, B., Sauval, A. J., Van Assche,
418: W., \& Vandenbussche, B. 2000, \aap, 364, 137
419: 
420: \bibitem[Eyer \& Grenon (1997)]{eg97} Eyer, L., \& Grenon, M. 1997, in
421: Hipparcos - Venice '97, (ESA SP-402), 467
422: 
423: \bibitem[Kester et al. (2001)]{kfl01} Kester, D., Fouks, B., \&
424: Lahuis, F. 2001, in The Calibration Legacy of the ISO Mission, ed. L.
425: Metcalf \& M. F. Kessler, (ESA SP-481), in press
426: 
427: \bibitem[Kraemer et al. (2001)]{ksp01} Kraemer, K. E., Sloan, G. C.,
428: Price, S. D., \& Walker, H. J. 2001, \apj, submitted
429: 
430: \bibitem[Kurucz (1979)]{kur79} Kurucz, R. L. 1979, \apjs, 40, 1
431: 
432: \bibitem[Leech et al. (2001)]{lee01} Leech, K., et al. 2001, in The
433: ISO Handbook Version 1.2
434: 
435: \bibitem[Schaeidt et al. (1996)]{sch96} Schaeidt, S. G., et al. 1996,
436: \aap, 315, L55
437: 
438: \bibitem[Shipman et al. (2001)]{shi01} Shipman, R. F., et al. 2001,
439: in The Calibration Legacy of the ISO Mission, ed. L. Metcalf \& M. F.
440: K. Kessler, (ESA SP-481), in press
441: 
442: 
443: 
444: \end{thebibliography}
445: 
446: 
447: 
448: 
449: 
450: \end{document}
451: 
452: 
453: 
454: 
455: