SPECTROPHOTOMETRY OF PLANETARY NEBULAE - IOPscience

SPECTROPHOTOMETRY OF PLANETARY NEBULAE IN THE BULGE OF M31

Martin M. Roth,1,2

Thomas Becker,1,2

and Andreas Kelz1

Astrophysikalisches Institut Potsdam, An der Sternwarte 16, D-14482 Potsdam, Germany; [email protected]

and

Jurgen Schmoll1,2,3

Astronomical Instrument Group, Department of Physics, University of Durham, Rochester Building, South Road, Durham DH1 3LE, UK;

[email protected]

Received 2003 September 7; accepted 2003 November 19

ABSTRACT

We introduce crowded-field integral field (3D) spectrophotometry as a useful technique for the study ofresolved stellar populations in nearby galaxies. The spectroscopy of individual extragalactic stars, which is nowfeasible with efficient instruments and large telescopes, is confronted with the observational challenge of ac-curately subtracting the bright, spatially and wavelength-dependent nonuniform background of the underlyinggalaxy. As a methodological test, we present a pilot study with selected extragalactic planetary nebulae (XPNe)in the bulge of M31, demonstrating how 3D spectroscopy is able to improve the limited accuracy of backgroundsubtraction that one would normally obtain with classical slit spectroscopy. It is shown that because of theabsence of slit effects, 3D spectroscopy is a most suitable technique for spectrophometry. We present spectra andline intensities for five XPNe in M31, obtained with the MPFS instrument at the Russian 6 m Bolshoi TeleskopAzimutal’nij, INTEGRAL at the William Herschel Telescope , and PMAS at the Calar Alto 3.5 m telescope. Theresults for two of our targets, for which data are available in the literature, are compared with previouslypublished emission-line intensities. The three remaining PNe have been observed spectroscopically for the firsttime. One object is shown to be a previously misidentified supernova remnant. Our monochromatic H� maps arecompared with direct Fabry-Perot and narrowband filter images of the bulge of M31, verifying the presence offilamentary emission of the interstellar medium in the vicinity of our objects. We present an example of a flux-calibrated and continuum-subtracted filament spectrum and demonstrate how the interstellar medium componentintroduces systematic errors in the measurement of faint diagnostic PN emission lines when conventionalobserving techniques are employed. It is shown how these errors can be eliminated with 3D spectroscopy, usingthe full two-dimensional spatial information and point-spread function (PSF) fitting techniques. Using 3D spectraof bright standard stars, we demonstrate that the PSF is sampled with high accuracy, providing a centroidingprecision at the milliarcsecond level. Crowded-field 3D spectrophotometry and the use of PSF fitting techniquesis suggested as the method of choice for a number of similar observational problems, including luminous stars innearby galaxies, supernovae, QSO host galaxies, gravitationally lensed QSOs, and others.

Subject headings: galaxies: individual (M31) — instrumentation: spectrographs — planetary nebulae: general —techniques: spectroscopic

1. INTRODUCTION

Understanding the formation and evolution of galaxies isone of the most prominent goals of modern cosmology. Aftera long-lasting discussion about the best evolutionary scenariosfor elliptical galaxies, for example, either according to theclassical picture of instantaneous dissipative collapse (Larson1974) or rather in the hierarchical scenario through gradualmergers of spirals (White 1980), it seems today that the latterview is the more widely accepted one. The observational evi-dence that nearby ellipticals show peculiarities such ascounterrotating cores, rings, gas disks, and recent star for-mation can be explained by merger and accretion processes

over an extended period of time. On the theoretical side, nu-merical N-body simulations combined with the smoothedparticle hydrodynamics (SPH) technique in a �CDM cos-mology are capable of reproducing observed photometric andkinematical properties of elliptical galaxies fairly well, thussupporting the hierarchical picture, despite some remainingdiscrepancies (e.g., Meza et al. 2003). On the other hand,Chiosi & Carraro 2002 presented SPH simulations producingelliptical galaxies, which are compatible with monolithiccollapse, so that the problem is not yet finally solved.

Among the classical tests for the validity of any of thetheoretical pictures, the integrated-light measurement ofbroadband colors and of absorption-line indices play an im-portant role. The observations can be compared with inte-grated light stellar population models in order to distinguishbetween different star formation histories and chemical en-richment (e.g., Saglia et al. 2000; Mehlert et al. 2000).

However, the presence of dust and ionized gas in the in-terstellar medium of elliptical galaxies (Goudfrooij et al. 1994;Bregman, Hogg, & Roberts 1992) has motivated the question,

1 Visiting Astronomer, German-Spanish Astronomical Centre, Calar Alto,operated by the Max Planck Institute for Astronomy, Heidelberg, jointly withthe Spanish National Commission for Astronomy.

2 Visiting Astronomer, Special Astrophysical Observatory, Selentchuk,Russia.

3 Visiting Astronomer, Isaac Newton Group of Telescopes, La Palma.

531

The Astrophysical Journal, 603:531–547, 2004 March 10

# 2004. The American Astronomical Society. All rights reserved. Printed in U.S.A.

whether or to what extent the observed gradients are real (dustaffecting the surface photometry through extinction, gaseousemission filling in the absorption-line profiles of H�, Mgb,Mg2). Moreover, systematic errors of long-slit spectroscopydue to the effects of a variable point-spread function (PSF) ofthe spectrograph optics along the slit have to be taken intoaccount. For a discussion of these problems, see Mehlert et al.(2000).

From a different point of view, Freeman & Bland-Hawthorn(2002) advocate using what they termed ‘‘near-field cosmol-ogy’’ to probe the six-dimensional phase space and chemicalcompositions of very many individual stars in order to unraveldetails of the star formation history of the Milky Way and thusthe signatures of its formation and evolution.

An alternative approach, now rapidly producing newabundance data and reaching out to significantly larger dis-tances, is based on the spectroscopic analysis of individualresolved luminous stars in nearby galaxies, e.g., M31 (Smarttet al. 2001; Venn et al. 2000), NGC 6822 (Venn et al. 2001),M33 (Monteverde et al. 1997), or NGC 300 (Urbaneja et al.2003). Using the ESO-VLT+FORS, the feasibility of highsignal-to-noise stellar spectroscopy even beyond the localgroup was demonstrated by Bresolin et al. (2001), who mea-sured seven supergiants of spectral types B, A, and F withV � 20:5 in NGC 3621, a galaxy that is located at a distanceof 6.7 Mpc. For a review on extragalactic stellar spectroscopy,see Kudritzki (1998).

Currently the only way to measure individual abundancesfrom old or intermediate-age stars in galaxies more distantthan the Magellanic Clouds is through the emission-linespectra of extragalactic planetary nebulae (XPNe; see Walshet al. 2000). This approach has some similarities with thestandard method of measuring abundance gradients from in-dividual H ii regions in the disk of spiral galaxies (Shaver et al.1983; Zaritsky, Kennicutt, & Huchra 1994). As opposed to H ii

regions, XPN metallicities can be derived in a homogeneousway for galaxies of any Hubble type and on all scales ofgalactocentric distances. The task of obtaining abundancegradients out to large radii, where a low surface-brightnessprecludes to measure reliable colors or absorption-line indices,can be addressed with XPNe, providing important constraintsfor galactic evolution models (Worthey 1999). Moreover,since radial velocities of XPNe are measurable out to severaleffective radii, they are potentially useful for probing thegravitational potential of galaxies (Mendez et al. 2001) and fortracing merger events (Hui et al. 1995; Durrell et al. 2003).Recently, XPNe and an H ii region have been detected in theintracluster space of the Virgo cluster, giving an excellentopportunity to study the properties of this unique stellarpopulation and, potentially, their star formation history andmetallicity (Arnaboldi et al. 2002; Gerhard et al. 2002;Feldmeier et al. 2003).

When compared with integrated light studies, the analysisof single objects has the advantage that the abundance deter-minations are based on detailed physical models and thatan assessment of individual errors is possible. The study ofresolved stellar populations in galaxies out to the distance ofthe Virgo Cluster has become a major science case for theproposed new generation of extremely large telescopes(Hawarden et al. 2003; Najita & Strom 2002).4

However, the spectroscopy of individual point sourcesin nearby galaxies suffers generally from source confusion,either from the continuum light of unresolved stars with asmall angular separation from the target or from the emission-line spectra of H ii regions and diffuse nebulosities of theinterstellar medium (ISM), or from both components at thesame time. A common feature of these two background com-ponents is that their spectral and spatial distribution is uncor-related and inhomogeneous, making it difficult to provideaccurate background estimates when one is using conventionalslit spectrographs.Roth et al. (1997) proposed the development of crowded-

field integral field (3D) spectroscopy as a superior method fordisentangling the spectrum of a target point source from thebackground in a crowded-field environment. Integral Beld(referred to as 3D) spectroscopy is an established but not yetentirely common, and therefore relatively new observingtechnique. It provides simultaneously spectra for many spatialelements over a two-dimensional field-of-view (see reviewby Boulesteix 2002). 3D spectroscopy has been used verysuccessfully for kinematic mapping and integrated light stellarpopulation studies of galaxies (de Zeeuw et al. 2002). Ourreason for observing point sources may not be immediatelyobvious, given the technical effort to generate and record upto thousands of spectra in any single exposure. The mainargument, however, is that the knowledge of the PSF of astellar object can be used to apply PSF-fitting techniques, thusdiscriminating the source against the background. Such atechnique would be analogous to PSF-fitting CCD photome-try, which has been so successful for the construction ofglobular clusters color-magnitude diagrams and the photo-metric study of resolved stellar populations in nearby galaxies(Mateo 1998).To this end, the design of the new integral field spectro-

photometer PMAS was put forward (Roth et al. 2000a), fea-turing a relatively small field of view (FOV; maximum 32�32 element lenslet array), but a unique wavelength coverage(4096 pixels) and a good response down to the UV, comparedwith other 3D instruments (Boulesteix 2002). First PMASobservations of point sources in M31, which are relevant forthe problems discussed below, were presented by Roth et al.2002a, 2002b.In this paper we discuss a pilot study for the novel ob-

serving technique of crowded-field 3D spectrophotometry forthe observation of XPNe in the local group galaxy M31. At adistance of 770 kpc toward M31 (Madore & Freedman 1991),an angular separation of 100 corresponds to a projected linearscale of 3.7 pc, making each XPN appear as a point source forground-based observations. In M31, more than 400 XPNcandidates have been detected with on-band/off-band imagingtechniques in the prominent [O iii] k5007 emission line(Ciardullo et al. 1989, henceforth CJFN89), enabling us toreadily select targets from a large range of apparent brightness(m5007 � 20:5 : : : 24:5) and location over the host galaxy,facilitating telescope pointing and identification and allowingus to investigate in detail faint diagnostic lines, which tend tosuffer from poor background-subtraction accuracy. In a futurepaper (Fabrika et al. 2004) we shall present a similar study for3D observations of luminous stars in M33.The paper is organized as follows: In x 2, we review the

peculiar observational problems for the spectroscopy of XPNeand explain, how 3D spectrophotometry can be used toovercome these problems. Section 3 presents a brief overviewof 3D techniques and the instruments that we have used for

4 See R. F. G. Wyse, K. Olsen, M. Rich, & R. O’Connell 2000, in TheGSMT Book: Report of the Stellar Populations Panel, available at http://www.aura-nio.noao.edu/book/index.html.

ROTH ET AL.532 Vol. 603

this work. We describe our observations in x 4 and data re-duction/analysis in xx 5, 6. We discuss our results in x 7 andconclude with a summary in x 8.

2. OBSERVATIONAL PROBLEMS

2.1. Previous Work

XPNe are ideal tracers of intermediate-age and old extra-galactic stellar populations, because their hot central stars areamong the most luminous stars in the Hetrzsprung-Russelldiagram, emitting their radiation predominantly in the UV. Asubstantial fraction (of order 10%) of the total luminosity is re-emitted by the surrounding nebula in a prominent emission-line spectrum, which gives enough contrast (for the brightlines) to detect the object as a point source against the brightbackground of unresolved stars of the parent galaxy. Practicalapplications of this property consisted in narrowband imagingspectrophotometry, centered on the bright emission line of[O iii] k5007 and the construction of PN luminosity functions(PNLF) for the purpose of distance determinations (see reviewby Ciardullo 2003). Approximately 5000 XPNe in more than40 galaxies have been identified to date (Ford, Peng, &Freeman 2002).

Prompted by the motivation described in x 1, severalauthors have conducted spectroscopic observations of indi-vidual XPNe in nearby galaxies and derived abundances fromthe observed emission-line intensities. Jacoby & Ciardullo(1999, henceforth JC99) obtained spectra of 15 XPNe in M31using the Kitt Peak 4 m telescope, equipped with the RCspectrograph and a multislit mask, with total exposure times of6.5 and 6 hr, for two setups, respectively. Their paper gives anexcellent account of the technical and observational issues tobe addressed below, and we will use it as our major reference.Richer et al. (1999, henceforth RSM99) observed 30 XPNe inM31 and nine XPNe in M32. Walsh et al. 1999 investigatedfive objects in NGC 5128 (the Cen-A galaxy). In the bestcases, and as far as the data quality allowed, photoionizationmodels were computed and matched to the observed spectra.From the constraints imposed on these models, abundances, inparticular [O/H], were derived. A wealth of data exists forMagellanic Cloud objects, which are an order of magnitudecloser and therefore much easier to observe than those inM31 and other more distant galaxies. The LMC study ofDopita et al. 1997 is an example for the potential of XPNeto investigate the chemical evolution of stellar populations.More recently, observations of XPNe in M33, obtained withthe multiobject fiber spectrograph at the William HerschelTelescope (WHT), were presented by Magrini et al. (2003).We shall compare our new 3D results with these previouspublications in x 7.

Although we have pointed out that central stars of XPNe areintrinsically luminous and that the prominent [O iii] k5007 linereflecting this brightness is relatively easy to detect, thecomparison of XPN spectra with ionization models and sub-sequent abundance determinations requires the detection ofmuch fainter lines, e.g., [O iii] k4363, which is important formeasuring the electron temperature Te. For a k4363 line in-tensity of e.g., 1% of k5007 (depending on Te), spectroscopywith a 3.5 m telescope will yield a total rate of �0.25 photonss�1 for an XPN of m5007 ¼ 21:0. At a distance modulus of24.6, this corresponds to an object near the bright cutoff of thePNLF. In the case of M31, if this object is less than 20 awayfrom the nucleus, the surface brightness is �V � 19 (Hoessel& Melnick 1980), contributing typically more than 0.7

photons s�1 within an aperture of 100 diameter and a spectralresolution of 3 A. Obviously, at this level of emission-lineintensity, the measurements are background-limited and re-quire exposure times of many hours to determine the spatiallyvariable surface brightness distribution of unresolved starsaccurately enough. In most real cases, the [O iii] k4363 isfainter and more difficult to measure than in our simple ex-ample. Jacoby & Kaler 1993, henceforth JK93) have de-scribed the effects of crowding for their observations of XPNein the Magellanic Clouds and summarized the following ob-servational problems.

2.1.1. Slit Losses

JK93 present an instructive diagram (their Fig. 1), allowingus to immediately read off the percentage of light loss as afunction of slit width relative to the seeing FWHM, dependingon centering errors of the star with respect to the slit. A slitwidth identical to the seeing FWHMwill in the ideal case resultin slit losses of roughly 25%, increasing to 30%, 50%, and 70%for a slit offset of 0.25, 0.50, and 0.75 (in units of seeing diskFWHM), respectively. Opening the slit will reduce slit losses,allowing for guiding errors and less precise pointing. How-ever, the spectral resolution will be degraded, and again theamount of background contamination will increase. For agiven background surface brightness it is possible to deter-mine a slit width with an optimized signal-to-noise ratio (S/N)of the detected flux. Since total exposure times are spanningmany hours, and because of the variable nature of seeing(Sarazin & Tokovinin 2002), it would be impractical to con-stantly adjust the slit width to the optimum. In addition, a slitwidth optimized at a certain wavelength may not be optimalat another wavelength, e.g., when there are strong absorption-or emission-line features in the background.

2.1.2. Seeing

Spectrophotometric measurements covering a large spectralrange, e.g., from [O ii] k3727 to [S ii] k6731, are affected byan increasing size of the seeing disk from the red to the blue(Fried 1966). In order to avoid substantial slit losses in theblue, and, resulting from this effect, systematic wavelength-dependent flux errors, JK93 advise a combination of expo-sures with wide and narrow slit width settings, thus providinga calibration for these errors while retaining the requiredspectral resolution and reduced background contamination forthe measurement of faint lines. In our application, the overallefficiency would drop prohibitively low because of the in-crease of exposure times.

2.1.3. Atmospheric Refraction

The long exposure times that are generally required for faintXPNe result in observations over a large range of air mass,i.e., exposures at large zenith distance cannot be avoided.Atmospheric dispersion produces an elongated chromatic fo-cal plane image, whose length is a function of zenith distance,leading to systematic errors if some light is not collected inthe slit. In multislit spectrographs the problem cannot be cir-cumvented with the standard strategy of adjusting the slitorientation along the parallactic angle when the slit maskshave a mechanical fixed orientation (JC99). The alternative toopening the slit larger than what would be necessary on thebasis of the seeing disk automatically results in a penalty ofincreased background contamination (see below). An instru-mental concept to counteract this problem is the atmospheric

3D SPECTROPHOTOMETRY OF XPNe IN M31 533No. 2, 2004

dispersion compensator (ADC; e.g., Avila, Rupprecht, &Beckers 1997), which, however, introduces light losses and isnot always applicable.

2.1.4. Background Contamination

The presence of a bright, both spatially and wavelength-dependent variable background of the host galaxy is a mostserious complication for accurate emission-line spectropho-tometry of XPNe. The following considerations are not re-stricted to these objects. They also hold true in analogousways for stars or any other point sources in galaxies, e.g., bluesupergiants, novae, supernovae, H ii regions.

The galaxy background surface brightness distribution iscomposed of unresolved stars and contributions from the ISM,either in the form of dust (extinction) or as gaseous emission,which may exhibit both diffuse or filamentary characteristics.Jacoby, Ford, & Ciardullo 1985, for example, have mapped inH� a prominent spiral-shaped gaseous component in the bulgeof M31, using a mosaic of CCD images. Long-slit spectros-copy described by Ciardullo et al. 1988 reveals strong emissionin [S ii] kk6717, 6731 and [N ii] kk6548, 6583—unfavorablycoinciding with diagnostic emission lines of XPNe.

The accuracy that can be obtained from an interpolation ofthe background along the slit on both sides next to the targetappears to be limited. JK93 present an example (LMC J22;their Figs. 2 and 3) that demonstrates the severe effects arisingfrom a bright nearby star and from the ubiquitous diffuse ISMemission. Similar problems have obviously also affected theextraction of the spectrum of PN 5601 in NGC 5128 of Walshet al. 1999, as one can deduce from the skewed appearance ofthe continuum (their Figs. 2 and 3). Some of the spectra plottedin RSM99 and JC99 show systematic deviations of the averagecontinuum from zero, indicative of errors in the backgroundsubtraction (e.g., PN 6, PN 24, and PN 25 in M32 and PN 12,PN 23, and PN 28 in M31 [RSM99]; CJFN 23, 125, 177, 179,and 470, and FJCHP 51 in M31 [JC99]). While this is notnecessarily evidence for errors in the emission-line intensities,one cannot rule out the possibility that, e.g., H� or one of theother Balmer lines could have been affected in the sense thatthe absorption line present in the background was not properlytaken into account, leading to an underestimate of the intensityof this emission line. We will show that the discrepancy in the[O iii]/H� line ratio for PN 29 in M31 reported by JC99 be-tween their result and the one of RSM99 can be explained inthis way. Such systematic errors would be particularly trou-blesome since an estimate of the interstellar reddening is af-fected by the accuracy of the Balmer decrement measurement.

Another severe problem is related to gaseous emission ofthe ISM. Let us consider, for example, a filament extendingperpendicular to the slit and stretching across the point sourceunder study. This feature will escape completely undetectedand will inevitably cause a systematic increase of some di-agnostic line intensities of the target. Other less extreme sit-uations may occur when again the true flux contribution of thefilament cannot be estimated precisely enough because onlytwo measurements next to the object along the slit are avail-able for an interpolation. Again, PN 29 in the study of JC99 isan example in which an overestimate of the flux and an un-usual intensity ratio of [S ii] kk6717, 6731 has led tounphysical predictions for the electron density in this object.This case is discussed in more detail in x 6. Other examples inthe sample of JC99 are the faint XPN CJFN 470 and CJFN455, where the [S ii], [O ii] and [N ii] line intensities areprobably overestimates.

2.2. Advances of Integral Field Spectroscopy

3D spectroscopy, contrary to conventional slit spectroscopy,produces data cubes with full two-dimensional spatial infor-mation within the FOV of the instrument. Figure 1 shows anidealized data cube with a square field of view, square spatialelements (‘‘spaxels’’) along coordinates (x, y), and spectraarranged in columns (k). From this entity, one can retrievemonochromatic maps at certain wavelengths (data cube slices)or co-added maps equivalent to broad- or narrowband images,and single spectra (columns), as well as co-added spectra forany desired region, i.e., digital aperture within the FOV. Westress that with proper spatial sampling it is possible to recordintensity profiles of monochromatic maps. For point sources,this is equivalent to measuring the wavelength-dependent PSFof the target. Conversely, one can use a priori knowledge ofthe stellar PSF to fit this profile in any data cube slice at thepredicted centroid position and derive a flux estimate even inthe presence of a complex and spatially variable backgroundsurface brightness distribution.Considering the observational problems outlined in x 2 and

comparing with spectrographs possessing slits of any geom-etry, we note that there are a number of advantages, whichbecome important especially for background-limited spec-troscopy of faint objects:

1. 3D spectroscopy is insensitive to pointing and guidingerrors when observing point sources. XPNe at the distance ofM31 have equivalent V magnitudes fainter than 21. Near thenucleus, they are normally invisible for broadband TV acqui-sition cameras on 4 m class telescopes. With 3D spectroscopy,the slit losses outlined in x 2.1.1 are no longer of concern.2. Since there is no fixed physical aperture, an optimal

digital aperture can be chosen throughout the data cube, thusoptimizing the S/N at any given wavelength and solving theproblem of x 2.1.2.3. In a data cube, differential atmospheric refraction is ob-

served as a translation in (x, y) as a function of k (Arribas et al.1999). Provided that the point source is not located too close tothe edge of the field in the direction of translation, it willalways be contained in the FOV and available for analysis,solving problem x 2.1.3.4. A most serious drawback of slit spectroscopy is that there

are only two regions along the slit next to the source, which canbe used to estimate the background at the location of the target,setting, e.g., the detection limit for Ly� emitters at 8–10 mclass telescopes to a few 10�17 ergs cm�2 s�1 (Morris et al.

Fig. 1.—Data cube, generated with integral field spectroscopy


2002). This limit is dictated by systematic errors of slit ge-ometry and other instrumental effects, even when the spatialdistribution of the night-sky background can be assumed to beconstant (Glazebrook & Bland-Hawthorne 2001). The situationfor the observation of our XPN is considerably worse: thesurface brightness of unresolved stars near the nucleus ofM31 is up to 7 mag brighter than the continuum of the night-sky background, and spatially variable on different scales.Because of the composition of the underlying stellar popula-tion, the spatially variable surface brightness distribution mayexhibit a variance in absorption-line profiles depending on thecontribution from stars of different spectral types. In addition,there is a spatially uncorrelated pattern of emission-line spectraemerging from H ii regions, supernova remnants, the diffuseISM, and other nebulosities (see Fig. 2).

Contrary to slit spectroscopy, 3D spectroscopy provides thefull two-dimensional spatial information at any wavelength inthe vicinity of the target. We will show that fitting a PSF to theXPN image at any wavelength slice of the data cube allows us toprovide the best possible discrimination of the point sourceagainst the variable background, thus reducing systematicerrors, which are otherwise unavoidable for slit of fiber spec-trographs. The details of this procedure are explained in x 6.

3. INSTRUMENTATION

3D instruments have been developed on the basis of dif-ferent principles: fiber bundles (e.g., Barden & Wade 1988),micropupil lens arrays (e.g., Bacon et al. 1995), micropupillens arrays coupled to fibers (e.g., Allington-Smith, Content,& Haynes 1998), and slicers (e.g., Weitzel et al. 1996). PMASis based on the fiber-coupled lens array design (see detailsbelow). In order to prepare for the use of PMAS alreadyduring its construction phase, and for reasons of data reduc-tion software development, a preliminary observing programwas conducted with two other existing 3D instruments.

3.1. MPFS

MPFS5 is a 3D spectroscopy instrument featuring the lensarray–fiber bundle design. It has been in operation since 1997in the prime focus of the 6 m Bolshoi Teleskop Azimutal’nij(BTA) in Selentchuk, Russia (Si’lchenko & Afanasiev 2000). Ithas a 16� 16 element lens array of square lenses, with a spatialsampling of 0B5, 0B75, and 100 at different magnifications, re-spectively. The 256 spectra are imaged onto a thinned 1K � 1KSITe TK1024 CCD using a f /1.2 Schmidt-Cassegrain camera.

5 See http://www.sao.ru/~gafan/devices/mpfs/mpfs_main.htm.

Fig. 2.—H� narrowband image of the central 60 of the Andromeda galaxy M31, obtained with Calar Alto 3.5 m prime focus camera+Fabry-Perot. The continuumhas been subtracted using an off-band image to reveal the complex and fuzzy emission structure of the ISM. The tiny dark spots are XPN candidates, while spotswith white substructures are indicative of residuals from imperfectly subtracted continuum sources (stars). The insert shows a broadband Digitized Sky Surveyimage with an outline of the H� frame, illustrating the rapid increase of continuum surface brightness toward the nucleus. North is to the top, and east to the left.


The total efficiency (excluding the atmosphere) is about 5%at 500 nm, dropping steeply toward the blue. From a varietyof gratings, dispersions of 1B35–5 A pixel�1 can be selected.

3.2. INTEGRAL

INTEGRAL6 is a fiber bundle type of 3D spectrographat the 4.2 m WHT, on La Palma (Arribas et al. 1998). Oneof three different bundles with 0B45, 0B9, and 2B7 projectedfiber diameters and with 205, 219, and 135 fibers, respect-ively, are available to accommodate different observationalrequirements in terms of spatial resolution and FOV. The fibersare packed in a hexagonal arrangement, filling a rectangularfield. The fiber bundles are coupled to WYFFOS, the bench-mounted fiber spectrograph on the Nasmyth platform of theWHT. The detector is a thinned 1K � 1K CCD at the internalfocus of the spectrograph Schmidt camera. Different gratingswith dispersions between 0.35 and 6.2 A pixel�1 can be used.

3.3. PMAS

PMAS7 is a dedicated 3D instrument with a lens array of16� 16 square elements in the present configuration, coupledby means of a fiber bundle to a fully refractive fiber spectro-graph, which is currently equipped with a 2K � 4K thinnedCCD (SITe ST002A), providing 2048 spectral bins. A 2�2K � 4K mosaic CCD, which was commissioned recently,increases the free spectral range to 4096 spectral bins. Alongthe spatial direction on the CCD, the current setup providesmore than twice the space that would be required for the totalof 256 fiber spectra. The present fiber bundle has been con-servatively manufactured with 100 �m diameter, high OH�

doped fibers for good UV transmission. A future upgrade with50–60 �m diameter fibers will replace the existing integralfield unit (IFU) with a 32� 32 element array. The opticalsystem is based on UV-transparent media and provides goodtransmission in the blue. A unique feature of PMAS is theinternal A&G camera, equipped with a LN2-cooled, blue-sensitive SITe TK1024 CCD, giving images with a scale of0.2 arcsec pixel�1 and a FOVof 3:4� 3:4 arcmin2. The cameracan be used with broad- and narrowband filters of the standardCalar Alto filter inventory. For a more detailed description, seeRoth et al. 2000a and Kelz, Roth, & Becker 2003.

4. OBSERVATIONS

In order to understand the problems of real 3D data and forthe purpose of developing the P3d data reduction software(Becker 2002) well before the commissioning of PMAS, weconducted a preliminary program of observations with otherexisting instruments (MPFS, INTEGRAL). We selected ourtargets from the list of CJFN89, supplemented with narrow-band H� images for the purpose of finding objects thatappeared to be critical in terms of background subtraction. TheH� frames were kindly made available to us prior to publi-cation by T. Soffner, University of Munich. The images weretaken on 1994 September 10, with the Calar Alto 3.5 m primefocus CCD camera, equipped with a Fabry-Perot etalon. Theetalon has a passband of �9 A FWHM and can be scanned inwavelength, thus providing flux and radial velocity informa-tion at the same time (Meisenheimer & Hippelein 1992). Anexample of three co-added frames (exposure time 3� 1000 s)for a 60 diameter FOV, centered on the nucleus, is shown inFigure 2. From the on-band image, a corresponding off-band

image has been subtracted, revealing the emission-line com-ponent with the XPNe as point sources, residuals of less thanperfectly subtracted stars and the prominent spiral-like ap-pearance of the ISM emission near the nucleus and throughoutthe bulge of M31. This feature has been known for some time(Munch 1960). The first optical images were obtained byJacoby et al. (1985). Spectra by Ciardullo et al. (1988) revealemission in H� and the strong low-excitation lines of [N ii]kk6548, 6583 and [S ii] kk6717, 6731, which are responsiblefor some of the background subtraction problems reported byRC99. This fact is illustrated further by the thumbnail imagesof our targets in Figure 3, where the fuzzy and filamentarystructure of the emission-line background becomes quiteobvious.

Fig. 3.—Calar Alto CCD images of XPNe in H� , compared with recon-structed narrowband maps from 3D data cubes. Left: 3.5 m prime focus Fabry-Perot images, except for PN 276 (2.2 m, CAFOS+narrowband filter). Right:data cube maps, obtained with MPFS (PN 27) and PMAS (PN 29, PN 56/181,and PN 276). Central wavelengths are at H� , except for PN 56/181 ([O iii]k5007), which was observed only in the blue. Note that the extended wings ofthe 9 A FWHM Fabry-Perot transmission profile introduce a nonnegligiblecontamination from [N ii] kk6548, 6583, which was mimicked in our 3Dmaps. The scale is 0.5 arcsec pixel�1. North is up, and east to the left.

6 See http://www.ing.iac.es/~bgarcia/integral/html/integral_home.html.7 See http://www.aip.de/groups/opti/pmas/OptI_pmas.html.


A log of our observations quoting the different instrumentsused is given in Table 1. Note that 3D spectra were takenwith different wavelength coverage, spectral resolution, andspatial sampling, depending on the instrument and prevailingobserving conditions. The object identification numbers wereadopted from CJFN89. Each target observation was accom-panied by standard star observations and bias, continuum,and spectral line lamp calibration frames, which, for bestaccuracy, where normally taken before and after a series oftarget exposures at any pointing for MPFS and PMAS, inorder to minimize flexure effects. INTEGRAL, employing thebench-mounted WYFFOS spectrograph, is insensitive toflexure.

5. DATA REDUCTION

The data reduction for all observations, except for MPFSdata of 1997, was performed using the P3d package of Becker(2002). P3d is an IDL code that was originally developed forPMAS but has been adapted to 3D data from other instruments(SPIRAL, INTEGRAL). The 1997 data were reduced by J. S.using a modified version of the dofiber task of IRAF.8 This taskinvolved a considerable amount of interactive work and re-quired controling each of the�256 spectra individually for anysingle CCD frame, whereas the P3d program works automat-ically once a set of parameters has been set appropriately. Thevarious steps of the data reduction procedure are as follows.The raw CCD frames are first bias-subtracted and cleaned fromcosmic-ray events. As an option, P3d allows application of acorrection for pixel-to-pixel CCD response nonuniformity,known as the standard flat-field correction for direct-imagingapplications. This correction is performed using grossly defo-cussed continuum flat-field exposures and corrects only forsmall scale pixel-to-pixel variations. Note that for 3D fiberspectroscopy, there are different levels of response calibration:(1) detector pixel nonuniformity, (2) total spaxel-to-spaxelthroughput variation (caused by lens array defects and dif-fraction, variation in fiber transmission, and coupling), and(3) the wavelength-dependent nonuniformity of spaxel-to-spaxel response, which is also linked to the wavelength-dependent variation of the detector quantum efficiency.

After these first basic procedures, a critical step consisted inthe tracing of spectra, which became difficult when there waslittle flux and thus resulted in confusion for automatic pro-cedures, in particular near the edge of a spectral range whenthe response was dropping steeply. This is where the dofiberroutine is normally required to correct interactively for mis-identifications and errors. P3d includes a geometrical modelbased on calibration exposures, which is cross-correlated andmatched with the observed set of spectra for any single ex-posure in order to avoid this complication. For the case ofPMAS, Figure 4 in Roth et al. 2002c shows how these spectraare arranged on the detector, with comfortably wide interordergaps (14 pixels distance), which is an unusual arrangementcompared with most other 3D instruments. In order to makeefficient use of detector estate, one would normally try tosqueeze as many spectra as possible onto the CCD. Theinevitable overlap of adjacent spectra to a degree that dependson the distance of spectra and optical quality of the spectro-graph optics leads to ‘‘crosstalk’’ with effects as discussed byRoth et al. 2000c and Becker, Roth, & Schmoll 2000 andinvestigated in detail by Becker (2002); but see also Alling-ton-Smith & Content 1998. We shall further discuss this issuebelow.

P3d allows for two different modes of extraction, which isthe next step in the data reduction process. The standard modeis to define an extraction swath along the traces and integratethe contained flux for any spectral bin. This is the procedurethat is also employed in a subset of P3d routines, forming thePMAS online data reduction software for use at the telescope.Swath extraction fails badly when the spacing of spectra isvery dense, in particular for applications where one is inter-ested in faint spectral signatures on a bright background.Such features are likely to be completely wiped out in thepresence of crosstalk. This difficult situation was encounteredwith our MPFS data sets. In order to correct for crosstalk, P3dprovides a profile-fitting mode that can be applied when amodel for the cross-dispersion profile can be constructed. Thedetails of this procedure are described in a separate paper(T. Becker 2004, in preparation). Profile-fitting also allowsus to correct for scattered light, which has been a considerablecomplication for our MPFS spectra in the blue (Fig. 4 in Rothet al. 2000c).

After extraction, a wavelength calibration is applied usingspectral line lamp exposures. This is another step where MPFSwith dofiber required human interaction and correction for

8 IRAF is distributed by the National Optical Astronomy Observatories,which is operated by the Association of Universities for Research in As-tronomy, Inc. (AURA) under cooperative agreement with the National ScienceFoundation.

TABLE 1

Log of Observations

Date Instrument Object Exposure Seeing Grating

�k(A) A pixel�1 Flux Standard(s)

1997 Nov 6................ MPFS PN 29 3 � 1200 1.5 V600 4000–6750 2.65 BD+284211, Feige 24

1997 Nov 6................ MPFS PN276 2 � 900 1.5 V600 4000–6750 2.65 Hiltner-600

1998 Sep 18............... MPFS PN276 2 � 1200 1.1 V600 4270–6910 2.6 BD+284211, Feige 24

1998 Sep 21............... MPFS PN 29 8 � 1200 1.5–2.0 V600 4210–6850 2.6 Feige 24

1998 Sep 27............... MPFS PN27 4 � 1800 2.0–2.5 V600 4210–6850 2.6 BD+284211

1998 Sep 28............... MPFS PN27 4 � 1800 2.5–3.0 V600 4210–6850 2.6 BD+284211

1998 Dec 26 .............. INTEGRAL PN276 3 � 1800 1.5 R600B 3540–6660 3.01 SP0305+261, G191–B2B

2001 Oct 23 ............... PMAS PN276 3 � 1200 1.5 V600 3550–5200 1.6 HR718

2001 Oct 24 ............... PMAS PN 29 9 � 1200 1.1 V600 5600–7200 1.6 HR153, HR1544

2001 Oct 26 ............... PMAS PN 29 8 � 1800 1.4 V600 3550–5200 1.6 HR153, HR1544

2002 Aug 31.............. PMAS PN56 4 � 1800 1.0 U600 3440–5100 1.6 HD192281

2002 Sep 1................. PMAS PN56 9 � 1200 1.1 U600 3440–5100 1.6 HD192281, HD19445

2002 Sep 1................. PMAS PN181 5 � 1200 1.1 U600 3440–5100 1.6 HD192281, HD19445


errors. P3d uses calibration lamp template spectra resulting ina robust procedure with no need for interactive corrections.

The set of extracted and wavelength-calibrated spectra wasthen corrected for spaxel-to-spaxel (or fiber-to-fiber) varia-tions by normalizing to a continuum flat-field exposure. Afinal correction of the smooth wavelength-dependent spaxel-to-spaxel response variation of order 3%–5% rms over theentire wavelength range was provided by means of a twilight-sky flat-field exposure for the MPFS (1998) and PMAS frames(‘‘fiber flat’’), omitting any disturbing atmospheric features.Using standard star exposures, the set of spectra was thenweighted with the instrumental and atmospheric sensitivityfunction to produce the final flux-calibrated data cube.

The resulting data set consists of a two-dimensional frameof stacked spectra, which we found to be a convenient ar-rangement for inspection with a visualization tool. An exam-ple for the appearance of this tool, which allows one to plotmaps, single or co-added spectra for mouse-selected digitalapertures, is shown in Roth et al. 2002b. The data were finallywritten to disk in a FITS compatible format whose specifica-tion was developed by the Euro3D consortium (Walsh & Roth2002).

6. DATA ANALYSIS AND RESULTS

This section is organized as follows: In xx 6.1–6.3 wediscuss general properties of our reduced 3D data, the pro-cedure of merging single exposures into a final data cube, ourtechniques of background subtraction, 3D spectrophotometry,and the final steps to derive the flux-calibrated and dereddenedXPN spectra. In xx 7.1–7.5 we present the results for indi-vidual objects. If not indicated otherwise, the software usedfor the various steps was developed under IDL and writtenby T. B.

6.1. 3D Point-Spread Function

A crucial test as to whether or not the proposed PSF fittingtechnique would be useful for 3D data consisted in the anal-ysis of standard star data cubes, which were well-enoughexposed to provide good S/N in any quasi-monochromaticslice (Fig. 4, upper right plot). The slices were treated likenormal, albeit tiny, CCD images, and a Moffat function wasfitted independently to the frame of each wavelength bin. Asan instructive example, the result for a particularly poor ex-posure of the standard star HR 1544 is shown in Figure 4,chosen for the purpose of demonstrating that it is indeedpossible to determine an accurate and well-defined PSF. Thedata cube was obtained with the P3d_online data reductionroutine without profile fitting. In the two panels in the middlewe plot the resulting FWHM fit values in X and Y versuswavelength, and in the bottom panels the correspondingcentroid offsets in X and Y with respect to a wavelength in themiddle of the free spectral range. A map of the mean residualof the PSF fit is displayed in the upper left panel.

First of all, we note that the centroid is not located at a fixedposition, but moves monotonically with wavelength in X and Y.The standard star was observed at an hour angle of 1h04m andat air mass 1.177, with an expected offset due to differentialatmospheric refraction of 0B18 between the wavelengths of5700 and 7300 A, according to Filippenko (1982). The am-bient parameters used for this estimate are a temperature of9.5

�C, a relative humidity of 33%, and an atmospheric pressure

of 789.8 mbar. The same offset of 0B18 is obtained in our datacube when we add the X and Y offsets in quadrature (a spatialelement corresponding to 0B5).

Second, it is seen that the scatter about a mean curve is verylow: a best-fit low-order polynomial yields an rms scatter of0.0015 and 0.0018 spaxels in X and Y, respectively, indicatingthat the PSF is an extremely sensitive measure of position(� 0B001). The obvious flaw near 5800 A, introducing asystematic error of �0B1 over a small wavelength interval, isentirely due to a local detector blemish that could not be flat-fielded out. The flaw vanishes completely in exposures wherethe star is centered on another region of the IFU. Likewise, weshow that similar systematic errors at a few isolated wave-length bins near the cores of the H� and other absorption linesare caused by small differential wavelength calibration errorsbetween different sets of spectra and the effects of spectralrebinning.Third, we observe that the PSF is nonsymmetric, as seen in

the residual image and FWHM plots, indicating an ellipticalelongation and an extended horizontal feature at a level of 1%peak intensity. This result is instructive in two ways: (1) thehorizontal feature illustrates the effect of crosstalk, introducedbecause of the presence of the extended wings of the fiberspectrograph PSF. The effect amounts to typically 0.5% peakintensity and vanishes when the profile-fitting extraction isapplied; and (2) the elongation indicates an optical aberrationof the 3.5 m telescope, which was in fact identified asdecentering coma and astigmatism due to a misalignment of

Fig. 4.—Moffat function PSF analysis of standard star data cube slices. Theupper left panel shows the mean residual of the fit, with a gray-scale stretchfrom 0% (white) to +1% (black) of peak intensity. For explanation see text.


the primary mirror, and which was corrected during a main-tenance period shortly after our 2002 August run.9

Finally, we note that the expected decrease of seeing disksize with wavelength is clearly seen in the plots of FWHM(Fried 1966).

6.2. Co-adding Data Cubes, Atmospheric Refraction

The measurement of faint line intensities of XPNe at thedistance of M31 using a 4 m class telescope requires totalexposure times of order several hours. The combination ofdata cubes taken over such a period of time is complicated bythe fact that each exposure is affected by a different amount ofdifferential atmospheric refraction, i.e., shift of monochro-matic slices with wavelength. To counteract this effect, weshifted the slices at the wavelength of interest to a commonreference system, which was compensated for refraction, us-ing the formulae of Filippenko (1982). Moreover, the effectof a different amount of atmospheric extinction as a functionof air mass had to be taken into account when co-adding alldata sets to form the final data cube. The set of PMASexposures taken in 2001 October was suffering from flexure ofthe telescope guiding camera, which required yet anotheradjustment. The net effect of flexure, which was confirmedindependently with two different daytime tests at the tele-scope, resulted in an overall shift of order 0B5–100 hr�1, thusrequiring shifting these data cubes to compensate both fordifferential refraction and flexure. Note that this problemdisappeared after the commissioning of the PMAS A&Gcamera, which is internal to the instrument and replaces thetelescope TV-Guider system. The necessity to correct for theshifts, however, was an interesting exercise for 3D mosaicexposures of areas larger than provided by the FOV of theIFU. The maps shown in Figure 3 therefore cover a larger areathan normally obtained from the instrumental FOV alone.

We note in passing that the continuous offsets between theexposures resemble the familiar dithering technique for directimaging, with the potential advantage of improving the sam-pling of the PSF and reducing residual errors of the spaxel-to-spaxel nonuniformity correction (Wisotzki et al. 2003).

6.3. 3D Spectrophotometry

The last step to create the spectra of our targets consisted inmeasuring the flux within a digital aperture or using a PSF fitaround the centroid of the fully corrected, final data cube at allwavelengths.

In order to do this accurately enough, one has to subtract atwo-dimensional model of the background surface brightnessdistribution, since contrary to the night-sky emission, this lightis anything but constant over the field of view. In our case, thebackground consists of the two dominant components: stellarcontinuum and diffuse/filamentary ISM emission. As discussedin more detail in x 7, the latter component presented a peculiarproblem, since it is difficult to distinguish from the continuum,being responsible for intensity overestimates of some weakXPNe emission lines in conventional spectrophotometry.

A two-step procedure was devised to accomplish the mod-eling and subsequent removal of these components. The XPNcase (as opposed to stars) is quite favorable, in that the objectcontinuum (typically 10�3 � IH� A

�1) is completely negligiblein comparison with the background intensity of the galaxy. Theentire data cube consists essentially of background light, except

for the few wavelengths with appreciable XPN emission-lineflux. Using a normalization scheme for all of the monochro-matic slices to take out the continuum spectral features, it waspossible to (1) create continuum maps as a function of wave-length, (2) enhance the contrast of the ISM emission-line fila-ments, and (3) create a map of the emission component. Thebackground correction was finally completed by subtracting thetwo components from the original data cube. The technicaldetails of this procedure are described in Becker (2002).

Although we cannot completely rule out the possibilitythat at the wavelengths of strong background absorption- oremission-line features the chance alignment of PSF-compatibleanomalies with the XPNe, i.e., unrelated point sources, wouldremain undetected and cause systematic errors of the XPNline fluxes, we expect that the probability of such chancealignments is small. In general we find that our continuumsubtraction yields flat spectra which are free of obvious re-siduals (Fig. 5), contrary to the examples discussed in x 2.1.

Because of the relatively high contrast of the [O iii] k5007line in an almost featureless region of the background con-tinuum, the centroid was defined to an accuracy of a fractionof a spaxel (e.g., 0B1 in the case of PN 29), and the flux wasobtained directly from aperture photometry. Using a standardstar curve-of-growth analysis, an aperture correction wasapplied to account for the flux lost outside of the digital ap-erture. The PSF at this wavelength was determined as a two-dimensional Gaussian and used as a template for the other,fainter emission lines with a correction for the wavelength-dependent variation of the seeing FWHM. Since the mea-surements are background shot-noise dominated, a more so-phisticated PSF characterization was unnecessary.

The final procedure was then to fit the wavelength-depen-dent model PSF to each data cube slice, constrained by thecentroid of the bright [O iii] feature. Note that the fit allows fornegative fluxes, which, in the presence of noise, avoids apositive bias of the detection probability. Note also that theabsence of any limiting apertures (slit effects) justifies the termspectrophotometry.

6.4. Two-Channel Deconvolution

As a more sophisticated method to separate a point sourcefrom the background, we used the two-channel Richardson-Lucy deconvolution algorithm cplucy (Lucy 1994), which isavailable within IRAF. It is optimized for improved photo-metric performance and avoids the artifacts and oscillationsthat are known for the standard RL-Algorithm (Hook & Lucy1993). The two-channel method makes use of the preciseknowledge of the location of one or several point sources anditerates the point-source channel separately from anotherchannel, which is processing the smoothly varying back-ground. We applied the algorithm to our data cubes such thateach monochromatic slice was subject to an independentcplucy run. The PSF is defined by one of the bright lines, e.g.,[O iii] k5007, corrected for the weakly varying FWHM as afunction of wavelength as determined from standard starexposures. The flux of the point-source detection in each slicein the end created the final spectrum. Details of the applicationof the two-channel deconvolution method to the peculiarcase of a data cube, the treatment of a photometric bias in thecase of very low background intensity levels, and various testsare described in more detail in Becker (2002).

As a drawback, we note that the procedure of fitting anoptimized PSF to the centroid of an XPN in each wavelengthbin (i.e., monochromatic map), is forced to produce detections

9 See also U. Thiele 2003, Calar Alto Newsletter 5, available at http://www.caha.es/thiele/Newsletter/align35.html.


with flux �0. The resulting spectra have nonnormal noisedistributions. They were preferentially used for confusion-limited cases and to determine upper limits for faint, back-ground-limited lines.

6.5. Dereddened Line Intensities

The emission-line intensities were measured after flux-calibrating the extracted XPN spectra with standard starexposures by fitting Gaussians at the nominal wavelengths,corrected for the radial velocity shift known from the bright[O iii] line. This step was performed within the ESO-MIDASdata reduction package and through use of a line fitting toolkindly provided by A. Schwope (AIP). We have assessed thestatistical errors of these fits by implanting artificial emission

lines with known intensities at several (typically �100) dif-ferent wavelengths near the original line into the data, fittingthese simulated lines and evaluating the standard error of asingle measurement from the variance of the whole set. Thedereddening based on the Balmer decrement followed thestandard procedure as described by RC99.The final results for our objects are listed in Table 2. The

errors for typical line intensities (faint/intermediate/brightlines) are listed as S/N values at the end of the table.

7. DISCUSSION

7.1. PN 29

This is the best-studied object of our sample. It was ob-served in three different observing runs with the MPFS and

Fig. 5.—Fully reduced and flux-calibrated spectra of our XPN sample. The spectral bin width for each spectrum (A pixel�1) is listed in Table 1.


PMAS instruments. We had selected PN 29 initially for the1997 MPFS run because of its relatively high brightnessamong the CJFN89 XPNe (m5007 ¼ 21:01), being locatednevertheless close enough to the nucleus of M31 to study thesystematic effects of different background subtraction tech-niques (13200 northeast from nucleus). Coincidently, PN 29 isone of the objects in common of the JC99 and RSM99 sam-ples, where the two sets of measurements produced conflictingresults, which can now be reconciled with our new 3Dobservations. JC99 give a detailed account of their measure-ment and the difficulties to match the observed line intensitieswith a physically reasonable photoionization model. The datapresented two major problems: first, their [O iii] k5007/H�line ratio of 14.8 is significantly different from the value of22.2 reported by RSM99. The discrepancy is completely in-compatible with the respective error estimates, so that thedifference remained a mystery. Second, the [S ii] line ratioI(k6717)/I(k6731) with a value of 1.76 is at the low-densitylimit, which required invoking a two-zone ionization model in

order to match the observed line intensities. Since, as a result,the sulfur abundance of this model was unusually high(½S=H� ¼ þ0:5), the authors considered the possibility of anartifact due to the weakness of the sulfur lines and the difficultbackground subtraction.

Concerning the first problem, an example of our new data isshown in Figure 6. The spectrum was obtained from thecombined data cube of a total of four 1800 s exposures using a600 lp mm�1 grating (1.65 A pixel�1) by co-adding fourspectra within a square digital aperture of 1� 1 arcsec2. Nextto [O iii] k5007 and k4959, H� is seen in emission on the bluewing of the stellar background absorption line. The thick lineis the spectrum of the background alone, derived as the av-erage from an annulus of 50 spaxels of radius ¼ 300 around thecentral aperture and scaled to a best match of the targetspectrum. Because of the large number of contributing spectra,the background spectrum has very little noise. Although atfirst glance the agreement with the continuum of the XPNaperture appears to be satisfactory, a closer inspection reveals

TABLE 2

Reddening-corrected Line Strengths

k(A) Ion PN 27a PN 29a PN 29b PN 56b PN 181b PN 276a PN 276c PN 276b

3727........... [O ii] . . . . . . <26 <38 <83 . . . . . . 820

3869........... [Ne iii] . . . . . . 145 26 95 . . . 84 71

3889........... He i, H i . . . . . . <15 <29 <74 . . . . . . 29

3969........... [Ne iii], H i . . . . . . 56 <26 <71 . . . 32 40

4101........... H� . . . . . . 40 17 <67 . . . . . . 31

4340........... H� 47 48 43 46 <64 49 51 47

4363........... [O iii] 37 . . . 19 27 <64 19 29 12

4686........... He ii . . . 23 19 <19 <62 . . . . . . 11

4861........... H� 100 100 100 100 100 100 100 100

4959........... [O iii] 542 612 565 399 368 117 122 118

5007........... [O iii] 1645 2030 1698 1282 1121 365 376 345

5755........... [N ii] . . . . . . <5 . . . . . . 10 10 . . .

5876........... He i 17 12 12 . . . . . . 11 12 . . .

6300........... [O i] . . . . . . 10 . . . . . . 69 83 . . .6360........... [O i] . . . . . . <5 . . . . . . 24 26 . . .

6548........... [N ii] . . . 16 15 . . . . . . 124 157 . . .

6563........... H� 286 286 286 . . . . . . 304 310 . . .

6583........... [N ii] 13 48 35 . . . . . . 380 381 . . .6678........... He i . . . . . . 15 . . . . . . . . . . . . . . .

6717........... [S ii] . . . 18 <5 . . . . . . 193 180 . . .

6731........... [S ii] . . . 30 <5 . . . . . . 188 176 . . .7065........... He i . . . . . . <5 . . . . . . . . . . . . . . .

7135........... [Ar iii] . . . . . . 12 . . . . . . . . . . . . . . .

Uncertainties (in %)

4363........... [O iii] . . . . . . 14.3 9.4 . . . . . . . . . 19.6

4861........... H� . . . 9.1 3.3 2.5 6.6 . . . . . . 2.3

5007........... [O iii] . . . 0.7 0.6 0.4 0.8 . . . . . . 0.7

[O iii] k5007 Magnitude and Logarithmic Extinction

m5007.......... 20.81 20.95 21.01 20.96: 21.88: 19.99: 20.13: 20.45

c................. 0.34 0.41 0.30 0.33 0.33 . . . . . . . . .

Notes.—Dereddened line intensities are tabulated in units of IðH�Þ ¼ 100. The instrument used for each measurement is indicated by footnotes.Typical uncertainties for lines of low, intermediate, and high intensity as derived from fitting emission-line profiles to [O iii] k4363, H�, and [O iii]k5007 are listed in the lower part of the table. Faint lines for which we did not succeed in obtaining a reliable fit are indicated with 5 � upper limits. Thelast two rows give the measured monochromatic magnitudes m5007, and the logarithmic extinction c. The discrepancies in m5007 among the differentPN 276 measurements are attributed to nonphotometric conditions during the MPFS and INTEGRAL observations.

a MPFS.b PMAS.c INTEGRAL.


that a measurement of the XPN H� line intensity dependssensitively on the details of the background definition. Obvi-ously, in order to arrive at an accurate line intensity estimate, itis necessary to properly account for the slope of the absorp-tion-line profile in the process of background subtraction.

We have simulated the background subtraction of conven-tional slit spectroscopy by co-adding flux in virtual horizontaland vertical slitlets adjacent to the point source, as indicatedby the rectangles in the H� map of Figure 7. Linear interpo-lation within each slitlet pair yields the spectra shown in thelower left panel of Figure 6 (thin solid line: horizontal slit;dash-dotted line: vertical slit). By way of inspecting thecontinuum map (Fig. 7), the offset between the two curves isreadily explained: the vertical slitlet pair samples, on average,a somewhat brighter background surface brightness than thehorizontal slit does, while the XPN centroid is located near alocal minimum of the continuum background surface bright-ness distribution. Therefore, the slit-based values overestimatethe background at the location of the point source, which isprobably also why the JC99 continuum spectrum of PN 29exhibits an overall negative offset from zero. An attempt tobring the two former curves into agreement by scaling them toa best match with the source spectrum outside of the H�feature, however, fails in the core of the absorption profile: thevertical slit predicts a significantly shallower absorption thanthe horizontal slit does (Fig. 6, lower right panel ). The dif-ference can be made plausible by realizing that the back-ground is increased by a diffuse emission component, whichpartially fills in the absorption profile (H� map in Fig. 7)—a familiar problem known also from integrated light popula-tion studies (see remarks in x 1). In our exercise, the horizontalslitlet pair is less effected by the ISM filament than the vertical

one, thus producing the deeper absorption trough. Numeri-cally, the difference translates into [O iii] k5007/H� line ratiosof 12.0 and 19.5, respectively.Although it is unlikely that our simulation is capable of

reproducing exactly the measurements of JC99 and RSM99,respectively, we can offer a sensible explanation for the dis-crepancy between their H� results: most likely the differencewas caused by a different setup of the slit orientation. Fromthis experiment it becomes clear that the simple slit approachis insufficient to provide an accurate background modelas required for a reliable determination of the contaminatedXPN H� line intensity. In contrast, the result of our PSF fitting

Fig. 6.—Background subtraction in H�. Top: Overview of a spectral regionnear 5000 A. Emission-line spectrum of PN 29 (thin line), superposed on thebackground continuum of unresolved stars (thick line). Bottom panels: Sim-ulating slit spectroscopy. Lower left: co-added flux from 3� 3 spaxels centeredon the XPN (thick line), background interpolation from simulated slit withhorizontal (thin solid line) and vertical orientation (dash-dotted line). Lowerright: The same spectra after renormalization to match the continuum nearH�; note the discrepancies in the core of the absorption line. The dashed line isthe continuum as determined from the two-channel cplucy algorithm.

Fig. 7.—Monochromatic maps of PN 29 in [O iii], H� , and H�, in thecontinuum and several low/high excitation forbidden emission lines. Thepositional offset between the upper and lower four maps is due to two differentpointings, taken in the blue and in the red, respectively. Except for the lowerright panel, the continuum has been subtracted. The overplotted centroids ofthe corresponding PSFs in H� and [O iii] k5007 (white contour) are indicatedto guide the eye, in particular for the faint emission lines. The gray-scale andcontour levels are set to arbitrary logarithmic scales in order to enhance faintfeatures in the background surface brightness distribution (white ¼ f aint,black ¼ bright). Scale: 0.5 arcsec spaxel�1. North is up, and east is left.


procedure with cplucy is plotted as the thick line in the bottompanels of Figure 6, the dashed line indicating the backgroundchannel alone. The [O iii] k5007/H� line ratio from thisanalysis is 17.0. An important consequence of the smaller H�intensity compared with JC99 is an increase of our value forthe logarithmic extinction from c ¼ 0:14 to c ¼ 0:30.

Concerning the problem of the [S ii] doublet kk6717, 6731,which appeared to be unusually bright [Iðk6717Þ ¼ 51] in thedata of JC99, these lines are not even detected in our spec-trum. They must be significantly fainter than this value, with aconservative upper limit of Iðk6717Þ � 5. A closer inspectionof the [S ii] map in Figure 7 shows that the XPN is embeddedin a fuzzy structure of background emission, which mostprobably has caused the overestimate suspected by JC99. Wenote in particular that at the location of the PSF centroid de-termined from H� , no local maximum is visible in [S ii].Furthermore, our map in [O ii] kk3727, 3729 also does notreveal a trace of any local maximum, contrary to the nearby[Ne iii] k3869, which is well defined. The 5 � upper limit for[O ii] is Iðkk3727; 3729Þ � 26.

These findings do not come entirely as a surprise. The ISMspectra of M31 presented by Ciardullo et al. 1989 show strongemission in the low-excitation lines of [S ii] kk6717, 6731 and[N ii] kk6548, 6583, with line strengths comparable to H� .After careful modeling and subtraction of the continuum, wehave produced a flux-calibrated spectrum of the gaseousemission background component (Fig. 8), which compareswell with the result of Ciardullo et al. 1989. Note that theorder of magnitude of the [S ii] k6717 flux within a 200 slit iscompatible with the PN 29 line intensity of 4� 10�16 ergscm�2 s�1 measured by JC99, which the authors had suspectedto be affected by background contamination. We also confirmthat the discrepant [S ii] kk6717, 6731 line ratio of 1.76 is dueto the background filament and is not intrinsic to PN 29.

In order to highlight the sensitivity to background con-tamination, we have plotted the JC99 and RSM99 line in-tensities versus our PMAS results, normalized to Iðk5007Þ ¼100 (Fig. 9). First of all, there is general agreement with JC99for the high excitation and Balmer lines within the error bars,except for [Ne iii] k3967, which we measure a factor of 0.8fainter. Since our spectrum is somewhat noisier because of ashorter total exposure time (4 hr as compared with 6.5) and

probably a higher blue atmospheric extinction at the site,10 wecannot rule out an error. We note, however, that RSM99 quotea value even lower than ours (0:6� JC99), so that we do notconsider the difference to be significant. More importantly,there is a striking disagreement for the low-excitation lines[O ii] k3727, [N ii] kk6548, 6583, and [S ii] kk6717, 6731, forall of which we derive significantly lower values (or upperlimits, respectively). We thus confirm that indeed the sus-pected background contamination is present not only for [S ii],but also for other ions, including [O ii].

7.2. PN 27

This object is located 15200 east of the nucleus and is amongthe brightest of the CJFN89 survey (m5007 ¼ 20:85). JC99were able to obtain an excellent spectrum, allowing them toconstruct one of the best-constrained models of their XPNsample. A large number of well-determined emission lines, inparticular the presence of a bright [O iii] k4363, providedsufficient confidence in the resulting model fit. Nevertheless,JC99 note the general faintness of the low-ionization linessuch as [N ii] and [S ii], the latter of which are just marginally

Fig. 8.—Continuum-subtract background spectrum near PN 29Fig. 9.—Comparison of PN 29 line intensities from JC99 and RSM99 vs.

PMAS results, normalized to Iðk5007Þ ¼ 100.

10 See U. Hopp & M. Fernandez, 2002, Calar Alto Newsletter 4, availableat http://www.caha.es/newsletter/news02a/boletin4.html.


present, giving an indication of a line ratio close to the lowdensity limit. Our MPFS spectrum, resulting from eightexposures with a total exposure time of 4 hr under mediocreseeing conditions (200–300), is consistent with JC99 as far asthe brighter lines are concerned (Table 2). For the fainter linessuch as [S ii] kk6717, 6731, the S/N is not sufficient to drawany firm conclusions. Nevertheless, the measurable [N ii]k6583 line presented us with a very good opportunity to applythe cplucy deconvolution technique, which uses our a prioriknowledge of the location and shape of the PSF as a criterionfor the separation of the point-source flux from the back-ground. At the wavelength of this emission line, the mapshows again a background emission filament that partiallyoverlaps the contours of the point source as derived from thebright nearby H� line (Fig. 10, upper panels). We applied thecplucy algorithm to investigate how much the filamentaryintensity is affecting a conventional measurement of the point-source emission at the locus of PN 27 and to what degree it ispossible to separate the two components. The lower panels ofFigure 10 show separate [N ii] k6583 maps for the pointsource and the background channel, respectively, after pro-cessing with cplucy. The ISM filament, which is also seenin the direct FP image in Figure 3, is very bright in proportionto the XPN, and, owing to its vicinity to the centroid ofthe point source, practically impossible to detect with the verylimited spatial resolution of a slit. Even our data cube aperturespectrophotometry using a background annulus resulted inan overestimate of the XPN emission in this line: 3:1�10�16 ergs cm�2 s�1. The upper limit derived with cplucyamounts to ð1:2 � 0:3Þ � 10�16 ergs cm�2 s�1. This resultis only half the line strength reported by JC99 and RSM99.The important conclusion from this experiment is that slitspectroscopy is indeed critically subject to systematic errorsin crowded-field environments and that 3D spectroscopy

provides a good tool to overcome these limitations, as onewould have expected from the long-standing experience withcrowded-field CCD photometry. As for the particular targetPN 27, a more accurate measurement of [N ii] k6583 andother faint emission lines would clearly benefit from an 8–10 mclass telescope with superior image quality, e.g., the GMOS-IFU at the Gemini-North Telescope (Allington-Smith et al.2000).

7.3. PN 56

PN 56 is another test case with suspected low-ionizationbackground contamination, but bright enough in [O iii] toprovide a well-defined PSF (m5007 ¼ 20:94). The XPN islocated 12000 southwest of the nucleus. The suspicious fila-ment extends in an arclike distribution to the west/southwest,also affecting another nearby object (PN 181) that is observedsimultaneously in the same FOV (Fig. 3). A third surfacebrightness maximum in H� is seen about 900 southwest ofPN 56. The somewhat irregular outline of the map in Figure 3is the result of dithering, allowing us to cover a larger FOVthan provided by the IFU alone. The observations were per-formed with PMAS during commissioning of the U600 grat-ing. Because of poor weather conditions (clouds), only spectrain the blue were recorded. The spectrum of PN 56 (Fig. 5)exhibits a relatively bright H� of Ið½O iii�k5007Þ=Iðk4861Þ ¼12:8 and the lines of H�, H�, [O iii] k4363, and [Ne iii] k3869.For other lines, such as He ii k4686, [O ii] kk3727, 3729,and [Ne iii] k3968, we were merely able to derive upperlimits.

7.4. PN 181

PN 181 is located 4B5 west of PN 56. It is the faintest of ourobjects, with m5007 ¼ 21:85. It is embedded in an environmentwith significant filamentary emission. Since the target wasincluded merely in a fraction of the exposures contributingto the mosaic with PN 56, the total integration time is only6000 s. The quality of the exposure was also affected by poortransmission of the atmosphere and clouds. The spectrum isdominated by photon shot noise of the background such thatonly [O iii] kk5007, 4959, H�, and [Ne iii] k3968 are detectedwith confidence (Fig. 5). At the 2 � level, there are hints of H�and H�, but we refrain from further analysis at this unsufficientlevel of S/N.

7.5. PN 276

Spectra of this object were secured for the first time by J. S.during the 1997 MPFS run at the 6 m BTA (Schmoll 2001). PN276 had been selected after a series of nondetections of othertargets due to pointing problems, since it is a relatively brightobject (m5007 ¼ 20:48), which was expected to be more easilyrecognized on the raw CCD frame. It is also located farther outin the disk (rnucl ¼ 82000) and is therefore less subject tobackground-subtraction problems. We remind the reader thatdespite the aforementioned advantage of 3D spectroscopy withregard to pointing accuracy, direct target acquisition using astandard broadband TV guiding camera is extremely difficult,if not impossible, because of the lack of contrast, thus de-manding blind offset pointing (see Fig. 2 in Roth et al. 2002c).PN 276 was also considered interesting because of the op-portunity of observing simultaneously two objects within thesame FOV (PN 276 and PN 277 in Ford & Jacoby 1978, of

Fig. 10.—Monochromatic maps of PN 27 in H� and [N ii] k 6583. The twolower frames show how the stellar profile can be separated from the contin-uum using the cplucy two-channel deconvolution algorithm.


which the latter, however, was no longer listed in CJFN89).Already from the inspection of monochromatic maps that wereproduced by the MPFS online data reduction software, a sus-picious extended triangular shape of the object became obvi-ous. Considering the spatial extension and the spectrum withatypically bright low-excitation lines (e.g., [O i] k6300), itbecame quickly clear that this object could not be explained asa single XPN or the superposition of two XPNe with smallangular separation. As an interesting comparative test case,PN 276 was subsequently reobserved also with MPFS in 1998,with INTEGRAL in 1998, and with PMAS in 2002. Theresulting flux-calibrated spectra are shown in Figure 11. Whilethe 1998 observations with MPFS and INTEGRAL wereobtained with similar spectral resolution and wavelengthcoverage, the PMAS observations were chosen to extend fur-ther into the blue at twice the dispersion. Although this spec-trum becomes increasingly noisy toward the atmosphericcutoff, the extraordinary strength of ½O ii�Iðk3727; 3729Þ ¼820 is clearly visible. The fitted line intensities for each ob-servation are listed in Table 2, showing good agreement be-tween the different instruments.

In an attempt to understand the physical nature of this ob-ject, the XPN hypothesis can clearly be ruled out. First, the[S ii] doublet kk6717, 6731 line ratio of 1.03 implies anelectron density of �400 e� cm�3, which is low for a typicalPN. Second, Te � 55000 K, derived from the [O iii] line ratio[I(k4959)+I(k5007)] /I(k4363) is too high. Third, the spatialextension of approximately 500 � 400 is incompatible with thetypical angular size of an XPN at a distance of 770 kpc(<0B1). For an H ii region, again, the electron temperatureis too high. The spectrum matches very well the generalsignature of supernova remnants (SNR) as explained, e.g., inOsterbrock (1989): strong [S ii] kk6717, 6731, [O ii] kk3727,

3729, and [O i] kk6300, 6364, all of which are indicative ofshock-excitation rather than photoionization. The high value ofTe is typical for SNRs, as well as the low electron density. Withlog ½IðH�Þ=Ið½N ii�Þ� ¼ �0:22 versus log IðH�Þ=Ið½S ii�Þ½ � ¼�0:1 in the diagnostic diagram of Sabbadin & D’Odorico(1976), PN 276 is located unambiguously in the SNR region,far away from H ii regions or PNe. The projected size of15� 12 pc2, confirmed with our direct CAFOS images in[O iii] k5007 and H� (Fig. 3), is also plausible only for anSNR. Finally, the coincidence with an X-ray source at thesame position, detected in the ROSAT PSPC survey of M31(Supper et al. 1997) and flagged as possible SNR (ID249)provides further evidence that PN 276 is in fact a SNR: theposition of 042m16F93, +41�18041B5 (1975), as listed inCJFN89, differs by only 2B7 from the ROSAT position of0h42m16F9, +41�18038B8, precessed from 2000 to 1975, inaccord with the PSPC error circle of 500 and the astrometricaccuracy of �100 quoted in CJFN89.

8. SUMMARY AND CONCLUSIONS

Our pilot study has demonstrated the usefulness of integralfield (3D) spectroscopy for spectrophotometry of faint plane-tary nebulae in M31. We have observed five objects, two ofwhich (PN 27 and PN 29) could be compared with previouslypublished data. Of the three new spectra, one reveals that theobject (PN 276) had been misclassified as an XPN. Spatialextension, characteristic properties of the spectrum, and posi-tional coincidence with a previously detected X-ray source in-dicate with a high level of confidence that the source is a SNR.

Thanks to the full two-dimensional spatial information ofthe observing technique, we make explicit use of the a prioriknowledge of the PSF, which is conveniently determined fromthe bright XPN [O iii] k5007 emission line. The PSF is used as

Fig. 11.—Spectra of the SNR PN 276, observed with MPFS, PMAS, and INTEGRAL. The spectral bin width for each spectrum (A pixel�1) is listed in Table 1.


a constraint to disentangle the point-source flux from con-taminating contributions that are due to diffuse or filamentarygaseous emission in the bulge of M31. We find that the low-excitation XPN line intensities [O ii] I(kk3727, 3729), [N ii]kk6548, 6583, and [S ii] kk6717, 6731 tend to be over-estimated in previous measurements with classical spectro-graphs. We have measured the emission-line intensity of theISM component and find that the fluxes corresponding to atypical spectrograph aperture are compatible with the level ofbackground contamination required to explain the systematicdiscrepancies in line intensities, which have been reported inthe literature. Simulated slit spectroscopy from a data cube forPN 29 allowed us to produce arbitrary line ratios for I([O iii]5007 A)/I(H�), suggesting an explanation for the discrepancyreported by JC99 and RSM99. We have employed successfullythe cplucy deconvolution algorithm to achieve the best sepa-ration of point sources from a highly complex background,both in terms of spectral and of spatial variation.

A fundamental conclusion is that the limited geometry ofconventional slit spectroscopy is ill-suited to solve the prob-lem of background subtraction of faint sources embedded incomplex, high surface brightness crowded fields. Our resultssupport and quantify the previously suspected notion that suchmeasurements suffer significantly from systematic errors. 3Dspectroscopy offers the unique opportunity to make full use ofthe two-dimensional spatial and one-dimensional spectral in-formation of a data cube to construct the best possible back-ground model and eliminate these problems. At the otherextreme, single-fiber multiobject spectroscopy in high surfacebrightness regions is obviously a very limited technique andmost vulnerable to systematic errors since there is no practicalway to provide reliable background estimates in the immediatevicinity of the point source (but see Durrell et al. 2003 forexcellent results from a radial velocity survey based on [O iii]of XPNe in M51). It must be stressed, however, that 3Dspectroscopy does not necessarily provide dramatic improve-ments in S/N for photon-starved observations but ratherallows one to eliminate the systematic errors discussed in thispaper.

As a side remark, we note that, in contrast to the con-clusions of Allington-Smith & Content (1998), residualcrosstalk between adjacent spectra on the CCD is potentiallydangerous for PSF-fitting algorithms, since the net effect is adistortion of the PSF (see Fig. 4), which must be expected tovary with FOV, wavelength, and time. Analogous to the rea-soning of Mateo (1998) concerned with CCD photometry, thedistortion is not important for the fitting of faint stars; how-ever, it is very much so for applications requiring a gooddefinition of the PSF (see Wisotzki et al. 2003).

In continuation of our successful pilot study, we are pur-suing a guaranteed observing time program targeting XPNe inM31 and other local group galaxies. The analysis of ourpresent and the expected new data by comparison with ioni-zation models following the recipe described by JC99 will bepublished in a forthcoming paper.

We have begun to use 3D spectrophotometry for a number ofother applications that benefit from the evaluation of spatialinformation and the PSF, e.g., spectrophotometry of SN2002er(Christensen et al. 2003; see also Aldering et al. 2002 for adescription of the SNAP experiment), the quadruple-lensedQSO HE 0435�1223 (Wisotzki et al. 2003), and the opticalcounterparts of superluminous X-ray sources (I. Lehmannet al. 2004, in preparation), luminous blue variable candidatesin M33 (S. N. Fabrika et al. 2004, in preparation). Although

our XPN study was facilitated by the availability of thebright emission in [O iii] k5007, which is a rather fortunate caseto determine a well-defined PSF, we stress that the PSF fittingtechnique, in particular when using the two-channel cplucyalgorithm in combination with high-resolution HST images,opens a new window for solving the difficult observationalproblems of resolved stellar populations in nearby galaxies.Although an upgrade of the present IFU with a 32� 32

element lens array is underway, increasing the diameter of theFOV by a factor of 2, the multiplex advantage of PMAS forthe rather sparsely distributed XPNe in nearby galaxies is verylimited. This fact is a direct consequence of the baseline de-sign, which favors wavelength coverage over the maximumnumber of spectra in terms of detector space, making PMAS arather unique instrument. Normally, PMAS can observe onlyone target at a time. With a typical rate of one object per night,the efficiency may appear prohibitively low. On the otherhand, we note that in a recent paper by Magrini et al. (2003),out of 42 XPN candidates in M33, observed during two nightswith the WYFFOS multiobject fiber spectrograph at WHT,only three spectra of were found to be of sufficient quality foran abundance analysis.Clearly, the spectrophotometry of our XPN targets would

benefit substantially from the light-collecting power and im-age quality of a modern 8–10 m class telescope, pushing thesource confusion problem to fainter limits and providingbetter S/N for the weak lines that are so crucial for the plas-madiagnostic analysis ([O iii] k4363, [S ii] kk6717, 6731,etc.). Multiobject spectroscopy of supergiant stars has alreadybeen demonstrated to be feasible out to distances beyond thelocal group with reasonable integration times on the order of�10 hr (Bresolin et al. 2001). The single-target approach ofour pilot study would be unrealistically expensive. However,the use of a multiobject deployable IFU instrument in com-bination with the techniques described in this paper wouldeventually become an extremely powerful tool for crowded-field spectroscopy and the study of resolved stellar pop-ulations in local galaxies, one of the key science cases of theproposed new generation of extremely large telescopes withdiameters of 30–100 m. As a promising alternative strategy,complex 3D instruments with arcminute fields of view and oforder 105–106 spatial elements are currently under investiga-tion for the 2nd Generation Instrumentation Plan at the ESO-VLT (Bacon et al. 2002; Morris et al. 2002).

Part of this work was supported by the DeutscheForschungsgemeinschaft (DFG) under grant HA1850/10-3and by the German Verbundforschung des BMBF, grant05AL9BA1. The authors would like to thank Till Soffner,University of Munich, and Hans Hippelein, MPIA Heidelberg,for providing us with the H� Fabry-Perot image of M31. M.M. R., T. B., and J. S. are indebted to Victor Afanasiev andSeguei Dodonov, Special Astrophysical Observatory inSelentchuk, Russia, for their hospitality and help with MPFSobservations at the 6 m BTA, for an introduction into 3D datareduction, and for fruitful discussions and insight into integralfield spectroscopy. We are thankful to Evencio Mediavilla forgenerously providing us with INTEGRAL observing time. J. S.would like to thank Evencio Mediavilla and Stefano Ciroifor introducing him into IFU data reduction techniques underIRAF. The excellent support of Calar Alto staff duringthe PMAS commissing phase and in operation is gratefullyacknowledged.


REFERENCES

Aldering, G., et al. 2002, Proc. SPIE, 4835, 146Allington-Smith, J., & Content, R. 1998, PASP, 110, 1216Allington-Smith, J. R., Content, R., Dodsworth, G. N., Murray, G. J., Ren, D.,Robertson, D. J., Turner, J. E., & Webster, J. 2000, Proc. SPIE, 4008, 1172

Allington-Smith, J. R., Content, R., & Haynes, R. 1998, Proc. SPIE, 3355, 196Arnaboldi, M., et al. 2002, AJ, 123, 760Arribas, S., Cavaller, L., et al. 1998a, in ASP Conf. Ser. 152, Fiber Optics inAstronomy III, ed. S. Arribas, E. Mediavilla, & F. Watson (San Francisco:ASP), 149

Arribas, S., Mediavilla, E., Garcia-Lorenzo, B., del Burgo, C., & Fuensalida,J. J. 1999, A&AS, 136, 189

Avila, G., Rupprecht, G., & Beckers, J. M. 1997, Proc. SPIE, 2871, 1135Bacon, R., et al. 1995, A&AS, 113, 347———. 2002, in Scientific Drivers for ESO Future VLT/VLTI Instrumentation,ed. J. Bergeron & G. Monnet (Berlin: Springer), 108

Barden, S. C., & Wade, R. A. 1988, in ASP Conf. Ser. 3, Fiber Optics inAstronomy, ed. S. Arribas, E. Mediavilla, & F. Watson (San Francisco: ASP),113

Becker, T. 2002, Ph.D thesis, Univ. PotsdamBecker, T., Roth, M. M.,& Schmoll, J. 2000, in ASP Conf. Ser. 195, Imagingthe Universe in Three Dimensions, ed. W. van Breugel, & J. Bland-Hawthorn(San Francisco: ASP), 544

Boulesteix, J. 2002, in ASP Conf. Proc. 282, Galaxies: The Third Dimension,ed. M. Rosado, L. Binette, & L. Arias (San Francisco: ASP), 374

Bregman, J. N., Hogg, D. E., & Roberts, M. S. 1992, ApJ, 387, 484Bresolin, F., Kudritzki, R. P., Mendez, R. H., & Przybilla, N. 2001, ApJ,548, L159

Chiosi, C., & Carraro, G. 2002, MNRAS, 335, 335Christensen, L., Becker, T., Jahnke, K., Kelz, A., Roth, M. M., Sanchez, S. F.,& Wisotzki, L. 2003, A&A, 401, 479

Ciardullo, R. 2003, preprint (astro-ph/0301279)Ciardullo, R., Jacoby, G. H., Ford, H. C., & Neill, J. D. 1989, ApJ, 339, 53Ciardullo, R., Rubin, V. C., Ford, W. K., Jacoby, G. H., & Ford, H.C. 1988, AJ,95, 438

de Zeeuw, et al., 2002, MNRAS, 329, 513Dopita, M., et al. 1997, ApJ, 474, 188Durrell, P. R., Mihos, J. C., Feldmeier, J. J., Jacoby, G. H., & Ciardullo, R.2003, ApJ, 582, 170

Fabrika, F., et al. 2004, A&A, submittedFeldmeier, J. J., Ciardullo, R., Jacoby, G. H., & Durrell, P. R. 2003, ApJS,145, 65

Filippenko, A. V. 1982, PASP, 94, 715Ford, H., & Jacoby, G. H. 1978, ApJS, 38, 351Ford, H., Peng, E., & Freeman, K. 2002, in ASP Conf. Ser. 273, The Dynamics,Structure, & History of Galaxies, ed. G. S. Da Costa & H. Jerjen (SanFrancisco: ASP), 41

Freeman, K., & Bland-Hawthron, J. 2002, ARA&A, 40, 487Fried, D. L. 1966, J. Opt. Soc. Am., 56, 1372Gerhard, O., Arnaboldi, M., Freeman, K. C., & Okamura, S. 2002, ApJ, 580,L121

Glazebrook, K., & Bland-Hawthorn, J. 2001, PASP, 113, 197Goudfrooij, P., de Jong, T., Hansen, L., & Norgaard-Nielsen, H. U. 1994,MNRAS, 271, 833

Hawarden, T. G., et al. 2003, Proc. SPIE, 4840, 299Hoessel, J. G., & Melnick, J. 1980, A&A, 84, 317Hook, R. N., & Lucy, L. B. 1993, in ESO Conf. Proc. 44, Science with the HST,ed. P. Benvenuti & E. Schreier (ESO: Garching), 245

Hui, X., Ford, H. C., Freeman, K. C., & Dopita, M. A. 1995, ApJ, 449, 592Jacoby, G. H., & Ciardullo, R. 1999, ApJ, 515, 169Jacoby, G. H., Ford, H., & Ciardullo, R. 1985, ApJ, 290, 136Jacoby, G. H., & Kaler, J. B. 1993, ApJ, 417, 209Kelz, A., Roth, M. M., & Becker, T. 2003, Proc. SPIE, 4841, 1057Kudritzki R. P. 1998, in Stellar Astrophysics for the Local Group, ed. A.Aparicio, A. Herrero, & F. Sanchez (New York: Cambridge Univ. Press), 149

Larson, R. B. 1974, MNRAS, 166, 585Lucy, L. B. 1994, in The Restoration of HST Images and Spectra II, ed. R.Hanisch & R. White (Baltimore: STScI), 79

Madore, B. F., & Freedman, W. 1991, PASP, 103, 933Magrini, L., Perinotto, M., Corradi, R. L. M., & Mampaso, A. 2003, A&A,400, 511

Mateo, M. 1998, in Stellar Astrophysics for the Local Group, ed. A. Aparicio,A. Herrero, & F. Sanchez (New York: Cambridge Univ. Press), 407

Mehlert, D., Saglia, R. P., Bender, R., & Wegner, G. 2000, A&AS, 141, 449Meisenheimer, K., & Hippelein, H. 1992, A&A, 264, 455Mendez, R. H., Riffeser, A., Kudritzki, R.-P., Matthias, M., Freeman, K. C.,Arnaboldi, M., Capaccioli, M., & Gerhard, O. E. 2001, ApJ, 563, 135

Meza, A., Navarro, J. F., Steinmetz, M., & Eke, V. R. 2003, ApJ, 590, 619Monteverde, M. I., Herrero, A., Lennon, D. J., & Kudritzki, R.-P. 1997, ApJ,474, L107

Morris, S., Content, R., Sharples, R., Bower, R., Davies, R., & Baugh, C. 2002,in Scientific Drivers for ESO Future VLT/VLTI Instrumentation, ed.J. Bergeron & G. Monnet (Springer: Berlin), 99

Munch, G. 1960, AJ, 65, 55Najita, J., & Strom, S. E. 2002, Proc. SPIE, 4835, 1Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae and Active GalacticNuclei (Mill Valley: University Science Books)

Richer, M. G., Stasinska, G., & McCall, M. L. 1999, A&AS, 135, 203Roth, M. M., Becker, T., Bohm, P., & Kelz, A. 2002a, in Scientific Drivers forESO Future VLT/VLTI Instrumentation, ed. J. Bergeron & G. Monnet(Berlin: Springer), 136

Roth, M. M., Becker, T., Bohm, P., & Kelz, A. 2002b, in ASP Conf. Ser. 282,Galaxies: the Third Dimension, ed. M. Rosaldo, L. Binette, & L. Arias (SanFrancisco: ASP), 411

Roth, M. M., Becker, T., & Kelz, A. 2002b, ASP Conf. Ser. 282: Galaxies: theThird Dimension, ed. M. Rosaldo, L. Binette, & L. Arias (San Francisco:ASP), 403

Roth, M. M., Becker, T., & Schmoll, J. 2000b, in ASP Conf. Ser. 195, Imagingthe Universe in Three Dimensions, ed. W. van Breugel & J. Bland-Hawthorn(San Francisco: ASP), 122

Roth, M. M., Seydack, M., Bauer, S., & Laux, U. 1997, Proc. SPIE, 2871, 1235Roth, M. M., et al. 2000a, Proc. SPIE, 4008, 277Sabbadin, F., & D’Odorico, S. 1976, A&A, 49, 119Saglia, R. P., Maraston, C., Greggio, L., Bender, R., & Ziegler, B. 2000, A&A,360, 911

Sarazin, M., & Tokovinin, A. 2002, in ESO Conf. Proc. 58, Beyond Con-ventional Adaptive Optics, ed. E. Vernet, R. Ragazzoni, S. Esposito, &N. Hubin, ESO Conference andWorkshop Proceedings (Garching: ESO), 321

Schmoll, J. 2001, Ph.D. thesis, Univ. PotsdamShaver, P. A., McGee, R. X., Netwon, L. M., Danks, A. C., & Pottasch, S. R.1983, MNRAS, 204, 53

Si’lchenko, O., & Afanasiev, V. 2000, A&A, 364, 479Smartt, S. J., Crowther, P. A., Dufton, P. L., Lennon, D. J., Kudritzki, R. P.,Herrero, A., McCarthy, J. K., & Bresolin, F. 2001, MNRAS, 325, 257

Supper, R., Hasinger, G., Pietsch, W., Trumper, J., Jain, A., Magnier, E.A.,Lewin, & W. H. G., von Paradijs, J. 1997, A&A, 317, 328

Urbaneja, M. A., Herrero, A., Bresolin, F., Kudritzki, R.-P., Gieren, W., & Puls,J. 2003, ApJ, 584, L73

Venn, K. A., et al. 2001, ApJ, 547, 765Venn, K., McCarthy, J. K., Lennon, D. J., Pryzbilla, N., Kudritzki, R. P., &Lemke, M. 2000, ApJ, 541, 610

Walsh, J. R., Jacoby, G. H., Peletier, R. F., & Walton, N. A. 2000, Proc. SPIE,4005, 131

Walsh, J. R., & Roth, M. M. 2002, Messenger, 109, 54Walsh, J. R., Walton, N.A., Jacoby, G. H., & Peletier, R. F. 1999, A&A,346, 753

Weitzel, L., Krabbe, A., Kroker, H., Thatte, N., Tacconi-Garman, L. E.,Cameron, M., & Genzel, R. 1996, A&AS, 119, 531

White, S. D. M. 1980, MNRAS, 191, 1Wisotzki, L., Becker, T., Christensen, L., Helms, A., Jahnke, K., Kelz, A., Roth,M.M., & Sanchez, S.F. 2003, A&A, 408, 455

Worthey, G. 1999, in Chemical Evolution from Zero to High Redshift, ed. J. R.Walsh & M. R. Rosa (Springer: Heidelberg), 168

Zaritsky, D., Kennicutt, R. C., & Huchra, J. P. 1994, ApJ, 420, 87


SPECTROPHOTOMETRY OF PLANETARY NEBULAE - IOPscience

Documents

Transcript of SPECTROPHOTOMETRY OF PLANETARY NEBULAE - IOPscience