Photoluminescence by Interstellar Dust

Transcription

1 A Dissertation entitled Photoluminescence by Interstellar Dust by Uma Parvathy Vijh As partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics Advisor: Prof. Adolf N. Witt Graduate School The University of Toledo August 2005

2

3 An Abstract of Photoluminescence by Interstellar Dust Uma Parvathy Vijh Submitted in partial fulfillment of the requirements for the Doctor of Philosophy Degree in Physics The University of Toledo August 2005 In this dissertation, we report on our study of interstellar dust through the process of photoluminescence (PL). We present the discovery of a new band of dust PL, blue luminescence (BL) with λ peak 370 nm in the proto-planetary nebula known as the Red Rectangle (RR). We attribute this to fluorescence by small, 3-4-ringed polycyclic aromatic hydrocarbon (PAH) molecules. Further analysis reveals additional independent evidence for the presence of small PAHs in this nebula. Detection of BL using long-slit spectroscopic observations in other ordinary reflection nebulae suggests that the BL carrier is an ubiquitous component of the ISM and is not restricted to the particular environment of the RR. We present the spatial distribution of the BL in these nebulae and find that the BL is spatially correlated with IR emission structures attributed to aromatic emission features (AEFs), attributed to PAHs. The carrier of the dust-associated photoluminescence process causing the extended red emission (ERE), known now for over twenty five years, remains unidentified. We constrain the character of the ERE carrier by determining the wavelengths of the radiation that initiates the ERE λ < 118 nm. We note that under interstellar ii

4 conditions most PAH molecules are ionized to the di-cation stage by photons with E > 10.5 ev and that the electronic energy level structure of PAH di-cations is consistent with fluorescence in the wavelength band of the ERE. In the last few chapters of the dissertation we present first results from ongoing work: i) Using narrow-band imaging, we present the optical detection of the circumbinary disk of the RR in the light of the BL, and show that the morphology of the BL and ERE emissions in the RR nebula are almost mutually exclusive. It is very suggestive to attribute them to different ionization stages of the same family of carriers such as PAH molecules. ii) We also present a pure spectrum of the BL free of scattered light, resolved into seven molecular emission bands, superimposed upon a broad continuum. The relative intensity of the component bands varies with position within the nebula, suggesting an origin in a set of several related molecular species, most likely small PAHs. iii

5 Acknowledgments First and foremost I express my heartfelt gratitude for my advisor Prof. Adolf Witt for his guidance and support. His knowledge, erudition and enthusiasm for astrophysical problems has encouraged me and will continue to motivate me. I must also thank my parents for their support, their belief and pride in me. I hope I live up to their pride. I also thank my husband and friend Aarohi who in spite of having to go through graduate school and dissertation-writing/defense at almost the same time as myself has never failed to support and encourage me. I hope I have been as good a friend to him as he has to me. We would like to acknowledge Karl Gordon for his efforts towards the observations at Steward Observatory and Karl was also the PI of the HST proposal, observations from which lead to the determination of the ERE excitation wavelength. We also acknowledge Paul Sell who made the measurements for the above study as a NSF-REU student at the University of Toledo. We acknowledge Donald York and the observing team at Apache Point Observatory for the spectroscopic mapping and narrow-band imaging effort, preliminary results from which are reported is this dissertation. We would also like to thank Dr. Louis Allamandola for his valuable comments on our first discovery manuscript. We also thank David Malin and Hans van Winckel for providing us with blue images of the Red Rectangle, Laurent Verstraete and W. W. Jochims for supplying us with data on ionization cross sections of PAH molecules, and Lewis Hobbs, Theodore Snow and Donald York for constructive discussions about the Red Rectangle nebula and its central source. I would like to thank my committee members for their support and helpful comments. This research was made possible through a generous allocation of observing time at CTIO, KPNO, Steward Observatory and through grants from the US National Science Foundation. Financial support for this study was provided through NSF Grant AST to The University of Toledo. I would also like to acknowledge a CTIO thesis student travel grant, for travel to Chile. This research has made use of NASA s Astrophysics Data System (ADS) Bibliographic Services and the SIMBAD database, operated at CDS, Strasbourg, France, and also of the Aladin image server. Some of the data presented in this paper was obtained from the Multimission Archive at the Space Telescope Science Institute (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS Support for MAST for non-hst data is provided by the NASA Office of Space Science via grant NAG and by other grants and contracts. iv

6 Contents Abstract ii Acknowledgments iv Contents v List of Figures x List of Tables xiv 1 Introduction Dust Grains and Nanoparticles Luminescence by Dust Grains: Extended Red Emission Observational Techniques for ERE Detection Observational Constraints Models for the ERE Carrier PAHs in the ISM Unidentified Infrared Bands or the Aromatic Emission Features Optical Fluorescence v

7 1.3.3 Formation of PAHs in the ISM Outline Discovery of Blue Luminescence Introduction Target and Observations Results Line-depth Technique Identification Discussion Small PAHs in the Red Rectangle Introduction Observations Analysis and Results Blue Luminescence in the Red Rectangle Attenuation of HD Discussion Identification of the BL Carrier Spectral Variability of the BL Spatial Variation of the BL Attenuation of HD Conclusions vi

8 4 Detection of BL in Other Nebulae Introduction Observations Results Ced Ced NGC NGC Discussion The BL Spectrum Spatial Distribution of BL BL in Hydrogen Ionization Regions Survival vs. In-situ formation of BL Carriers Conclusions The Excitation of Extended Red Emission Introduction Observations and Data Reduction Results The ERE, H 2, and z-band Morphology of NW Filaments Determination of R V in NW Filaments Width of the ERE Filaments Wavelength of ERE Initiation vii

9 5.4 Discussion Impact on Existing ERE Models Possible New ERE Carriers Consistency Check: The Red Rectangle Consistency Check: The High- b Galactic Cirrus Conclusions Optical Emission Band Morphologies of the Red Rectangle Introduction Observations and Reductions Results and Discussion Blue Luminescence in the Red Rectangle Sharp Emission Feature at 5800 Å ERE Conclusions Spectral Characteristics of BL in the Red Rectangle Introduction Observations and Reductions Results The Band Structure of the BL Variation of BL Spectra with Position in the RR Variation of BL Intensities with Position in the RR Morphology of BL in the RR viii

10 7.4 Discussion Identification of the BL Carrier Temperature of the BL Carrier Conclusions Summary What did we learn? Future Work A Determination of BL Intensity at the Balmer Jump 207 References 225 ix

11 List of Figures 1-1 Interstellar dust emission Schematic energy level diagram of a photoluminescing system Line-depth technique Structures of some small PAHs UIR/AEF spectrum Jablonski diagram Solution dependence of PAH fluorescence Temperature dependence of PAH fluorescence Formation routes for PAHs in the ISM Variation of fluorescence peak with molecular size The RR nebula at different optical wavelengths Filling in of a Balmer line due to fluorescence Comparison of BL spectrum with solid-state and PAH candidates Blue Image of the RR with overlaid apertures Red HST image of the RR with overlaid apertures x

12 3-3 Spectral variation of the BL Ratio of Band-I BL to Band-I sc and normalized distributions Spatial distribution of the BL and correlations with other emissions Observed UV/optical SED of HD compared to model SED UV/optical attenuation curve for HD Attenuation curve for HD 44179, a different representation Ionization potential of PAHs as a function of molecular mass Image of Ced 201 with overlaid slits BL spectrum in Ced Image of Ced 112 with overlaid slits BL spectrum in Ced ISOCAM image of Ced 112 with overlaid apertures Band-I BL and Band-I sc distribution in Ced Ratio of Band-I BL to Band-I sc in Ced Image of NGC 5367 with overlaid slit Bl spectra in NGC Image of NGC 2023 with overlaid slit BL spectra in NGC Average BL spectra in Ced 112 and Ced 201 compared to BL in the RR Banded BL spectrum in Ced Band-I BL and Band-I sc in Ced 201 and Ced Ratio of Band-I BL to Band-I sc in Ced 201 and Ced xi

13 4-16 Average BL spectrum in NGC 2023 and NGC F606W WFPC2 image of NGC The ACS filter and the continuum-subtracted ERE and Hα images NICMOS images and continuum-subtracted ERE and H Four cuts through the ERE and H 2 images The density-dependence of dust optical depth of a PDR CCM extinction relations Histogram of ratios of filament widths Allowed ratio of extinctions The limits of ERE excitation First and second ionization potentials of PAHs Model spectra of representative PAH di-cations Image at 3934 Å with contours tracing the BL Image at 4050 Å with scattered light contours Å image divided by 5700 Å image Å image divided by 5700 Å image Å image divided by 5700 Å image Å image divided by 4050 Å image HST red image superimposed with slits and extraction windows Nebular spectra at different offset from the central star Normalized BL spectra from different nebular positions xii

14 7-4 Band-I BL and I sc distribution Deep blue image of the RR overlaid with extraction windows Fluorescence spectrum of anthracene xiii

15 List of Tables 2.1 Fluorescence Intensity at two positions in the RR nebula Details of objects observed, their illuminating stars and slits used Narrow-band observations of the RR nebula RR Observations at APO Molecular Bands in the BL spectrum of the Red Rectangle xiv

16 Chapter 1 Introduction Our universe is a dusty universe, and dust grains constitute the dominant form of solid matter in the universe. Dust is seen in a wide variety of astrophysical environments, ranging from circumstellar envelopes around cool red giants to supernova ejecta, from diffuse and dense interstellar clouds and star-forming regions to debris disks around main-sequence stars, from comets to interplanetary space to distant galaxies and quasars. Longward of the Lyman limit of hydrogen, grains represent the main source of continuum opacity within galaxies, resulting in absorption and scattering of starlight at UV/visible wavelengths and thermal emission in the IR and sub-mm range. In the high-z universe, dust grains obscure a large fraction of the ongoing star formation, while actually promoting star formation by inducing necessary conditions of low temperature and high density in molecular clouds. They play a central role as a photoelectric heating agent for the interstellar gas. Within the interstellar medium (ISM), dust grains constitute a reservoir for a significant fraction of all atoms more massive than helium and play the role of an efficient catalyst for 1

17 2 the formation of H 2. Interstellar dust grains span a wide range in sizes: from a few angstroms to a few micrometers. They play a vital role in the evolution of galaxies as an absorber, scatterer, and emitter of electromagnetic radiation. Dust grains act as drivers for the mass loss of evolved stars and are essential in the star and planet formation process. The most prominent visual effect of interstellar dust is as an agent shaping the spectacular appearance of dusty systems such as protostars, young stellar objects, evolved stars and galaxies. 1.1 Dust Grains and Nanoparticles Despite the importance of interstellar grains in many astrophysical processes, their nature, i.e. size distribution, their composition, their structure, as well as their total mass, is not well known. Interstellar nanoparticles form the smallest and the most numerous of these interstellar grains. The existence of this component has only been recognised within the past two decades. Nanoparticles, after absorbing star light, emit this absorbed energy not only as thermal radiation but also through specific vibrational bands and electronic photoluminescence transitions. This fact suggests the grains in the nanometer size range still retain their pure chemical and structural characteristics. The collective UV/optical absorption cross-section of nanoparticles is comparable to that of classical interstellar grains in the ISM. See Figure 1-1.They are most likely the structural building blocks of larger, composite grains and as such hold the key to

18 3 understanding the nature of interstellar grains in general. The study of interstellar nanoparticles is particularly promising, because they exhibit a wealth of observable diagnostics and in addition, their large surface/volume ratio makes them the dominant reservoir of solid surfaces in the ISM, enabling the formation of molecular hydrogen and other simple molecules as well as complex organic molecules which may lead to the origins of life. Figure 1-1 Different components of emission from interstellar dust (from Desert et al., 1990, fig.4)

19 4 1.2 Luminescence by Dust Grains: Extended Red Emission 1 Our knowledge of the existence of interstellar dust, its spatial distribution, its chemical composition and size distribution, and its mass relative to the rest of the interstellar medium is almost exclusively based upon observations of the interactions of interstellar dust grains with radiation. The processes studied in the past included mainly interstellar extinction and polarization of starlight at optical, near-ir and ultraviolet wavelengths, followed by studies of dust emission over a wide range of wavelengths from the near-ir to the sub-mm and microwave region. In this thesis we are reporting on the study of interstellar dust through the process of photoluminescence (PL), a process in which absorptions of photons at ultraviolet/optical wavelengths are followed by electronic transitions associated with the emission of longer-wavelength optical and near-ir photons. Interstellar dust in nebulae and in the diffuse interstellar medium of galaxies contains a component which responds to illumination by ultraviolet photons with efficient luminescence in the 500 nm to 1000 nm spectral range, known as Extended Red Emission (ERE) (see Witt & Vijh, 2004, for a detailed review). In the next section of this chapter we review the techniques of detection of the ERE and the constraints that any proposal for the ERE carrier must confront. Many models have been advanced over the past two decades to explain the ERE phe- 1 Based in part on Extended Red Emission: Photoluminescence by Interstellar Nanoparticles a review that appeared in Astrophysics of Dust 2004, ASP Conf. Proc. 309, Authors: A. N. Witt & U. P. Vijh

20 5 nomenon, but despite promising progress on several fronts, no completely satisfactory model for the ERE carrier/process exists at this time. 2 Series of rotational/vibrational transitions hν exc E 3 PL 1 Figure 1-2 Schematic energy level diagram of a photoluminescing system Figure 1-2 schematically illustrates the physics of a PL process. The ground state (1) represents the electronic ground state of a large molecule or molecular ion, or a state near the top of the valence band in a semiconductor particle. The excitation of the PL process under astrophysical conditions results from the absorption of a single UV/optical photon, leading to an electronic transition from state (1) to state (2). State (2) typically is a bound, high-lying vibrational/rotational level of the first or

21 6 second electronically excited state of a molecule or molecular ion, or a high state in the conduction band of a semiconductor particle. The excited system relaxes through a series of vibrational/rotational transitions until the electron finds itself in state (3), from where an optical electronic transition back to the ground state (1) is possible. In a polycyclic aromatic hydrocarbon (PAH) molecule, for example, state (3) can either be the lowest state in the singlet or triplet vibrational/rotational manifold of the first excited electronic level. The resulting emissions are referred to as fluorescence and phosphorescence, respectively. In a semiconductor particle, state (3) is a state near the lower edge of the conduction band, and the downward transition across the bandgap is simply referred to as photoluminescence. For a more complete review of PL in aromatic molecules and in semiconductor nanoparticles, the reader should consult the works of Birks (1970) and Yoffe (2001). Different photoluminescing systems can be characterized by the energy difference between the exciting and PL photons, known as the Stokes shift, and by the quantum yield or photon conversion efficiency. Here we define quantum yield as the ratio of the number of PL photons to the number of exciting photons needed to produce the photoluminescence. Most luminescing systems exhibit quantum yields of substantially less than 100%, but if the energy of the exciting photon is more than twice the energy of the PL photon, highly isolated molecules or molecular ions are expected to yield two (or more) PL photons per excitation, in a process known as Poincaré fluorescence (Leger et al., 1988). Thus, quantum yields well in excess of 100% may result in systems with large Stokes shifts under astrophysical conditions, although no specific identifications invoking this process have been made. In most natural

22 7 systems, ionization or the creation of defects either quenches or shifts the PL into another spectral region, e.g. the near-ir, which permits relaxation to occur without the emission of an optical photon, thus reducing the quantum yield to well below 100%. Early suggestions that the high surface brightness of reflection nebulae in the visible might be due in part to fluorescence by the nebular dust grains were advanced by Struve & Swings (1948) and by Aller (1956). Indeed, many minerals now thought to be part of the composition of interstellar grains do fluoresce under illumination by ultraviolet light or following particle bombardment (Koike et al., 2002). However, a first sensitive search for the existence of such fluorescence in several bright reflection nebulae at wavelengths shortward of 490 nm by Rush & Witt (1975) failed to reveal any detectable sign of dust fluorescence, leading to the conclusion that the nebular surface brightness at these wavelengths was a result of scattering with a relatively high grain albedo. The advent of new, sensitive detectors in the red region of the visible spectrum in the 1970 s brought the first detections of PL, although they were not immediately recognized as such. The detection of the broad luminescence band in the spectrum of the Red Rectangle nebula (Cohen, 1975) was followed by its analysis in terms of unusual dust scattering properties by Greenstein & Oke (1977), although the possibility of fluorescence was mentioned. Subsequently, with much better observational data, Schmidt et al. (1980) suggested a molecular-emission origin for the broad band of excess radiation in the Red Rectangle. The belief that this object was unique in exhibiting this excess radiation and the lack of a context of other sources showing

23 8 the same phenomenon clearly contributed to our inability to recognize the true nature of the luminescence process in the Red Rectangle. For example, the spectrum of red excess radiation detected in the Galactic dark nebula Lynds 1780 by Mattila (1979) appeared sufficiently different that its connection to the Red Rectangle was not recognized until several years later (Chlewicki & Laureijs, 1987). A breakthrough occurred in this field with the realization that dust luminescence in the red part of the spectrum, extended red emission or ERE, as it soon became known, was a common feature in many dusty environments that are illuminated by ultraviolet photons. Detections of the ERE in reflection nebulae (Witt et al., 1984a; Witt & Schild, 1986, 1988) soon led to the observation of ERE in high-latitude cirrus clouds (Guhathakurta & Tyson, 1989) and to the discovery of the presence of ERE in the continuum spectrum of planetary nebulae (Furton & Witt, 1990, 1992). Perrin & Sivan (1992) started a series of discoveries of the ERE band in HII regions with the Orion nebula, followed by the 30 Doradus nebula in the LMC (Darbon et al., 1998) and the HII region Sh 152 (Darbon et al., 2000). The ERE detections in other external galaxies include the halo of M82 (Perrin et al., 1995) and the prominent dust lane in NGC 4826, the Evil Eye galaxy (Pierini et al., 2002). The detection and measurement of ERE in the diffuse interstellar medium of the Milky Way Galaxy over a wide range of Galactic latitudes by Gordon et al. (1998) represents a particular milestone for three reasons. It demonstrated that the ERE carrier is a component of interstellar dust on a Galaxy-wide scale; it represented detections of the ERE intensities at a level three orders of magnitude fainter than observed in the Red Rectangle; and finally, it permitted the first reliable estimate of the quantum yield or

24 9 photon conversion efficiency of the ERE process Observational Techniques for ERE Detection Several factors contribute to the challenges one faces when attempting to observe ERE. First, ERE appears only in spatially extended objects, usually of relatively low surface brightness. Second, ERE is a very broad emission, spanning the wavelength range from 540 nm to at least 950 nm. And third, it always occurs in conjunction with other sources of diffuse emission or scattered light, most of which frequently are much brighter than the ERE. For example, in reflection nebulae the competing radiation is dust-scattered light; in HII regions dust-scattered light, atomic recombination continua as well as line emissions occur in the same spectral region, while in external galaxies integrated starlight is a contender as well. Attempts to measure ERE as part of the high-galactic-latitude background require careful subtractions of atmospheric foregrounds such as airglow and atmospheric scattered light, zodiacal light, integrated starlight and diffuse Galactic light, before the ERE can be isolated. As a result, different observational techniques must be applied to different types of sources. Color-Difference Technique The color-difference technique is an efficient way to probe the presence of ERE in extended objects, provided that broadband colors of the source are readily predictable in the absence of ERE, and the presence of sufficient ERE changes these colors. This is the case when the competing radiation is dust-scattered starlight, e.g. in reflection nebulae (RN) or in the the

25 10 diffuse Galactic background radiation. In an optically thin RN, the well-known wavelength dependence of the scattering optical depth leaves the nebular radiation bluer than that of the illuminating star throughout the optical spectrum. A simple approach for estimating the color difference between scattered and illuminating radiation is given by Witt (1985). If, therefore, the V-R color of a nebular region is redder than that of the star while the B-V color is bluer, this is a strong indication for the presence of ERE. Long-Slit Spectroscopy The spectroscopy of ERE in low-surface-brightness, extended sources is strongly affected by terrestrial airglow. Long-slit spectrographs provide the best approach to simultaneously record and subsequently subtract the foreground sky spectrum. Also, in sources whose spectrum is dominated by emission lines, e.g. HII regions and planetary nebulae, long-slit spectroscopy offers a possibility for studying the faint continuum, including the broad ERE band, between strong emission lines. The large extent of the ERE in wavelengths requires a combination of relatively low wavelength resolution and broad spectral coverage. Most of the work presented in this dissertation is based on this technique. Line-Depth Technique The presence of quasi-continuous luminescence may be difficult to discern, if it lacks a distinctive spectral feature, as is the case with the ERE band. In reflection nebulae, however, one can easily distinguish dust luminescence from scattered radiation by virtue of the fact that scattering faithfully reproduces relative line depths or equivalent widths of spectral lines present in

26 11 the spectrum of the illuminating source. The presence of a certain fraction of luminescence at the positions of such lines is revealed by a proportional reduction in the relative line depths in the nebular spectrum compared to corresponding relative line depths in the illuminating star, as illustrated in Figure 1-3. This method is particularly useful in the search for luminescence in the blue part of the spectrum, where strong hydrogen Balmer lines dominate the spectra of most reflection nebulae. It is important to note that the spectral resolution of both the stellar and the nebular spectrum must be identical to avoid a falsepositive result. This method was first employed by Rush & Witt (1975), who used photographic spectroscopy and narrow-band photoelectric photometry to place useful upper limits on the contribution made by possible dust photoluminescence to the observed surface brightness of several reflection nebulae. Figure 1-3 Illustration of the line-depth technique. In the nebular spectrum, consisting of scattered light and PL, the relative line depth (C N L N )/C N is smaller than the corresponding value (C S L S )/C S in the stellar spectrum. Unsharp Masking There exists a high degree of spatial variability of the ratio of

27 12 ERE to scattered light within a single nebula. In particular the ERE distribution seems to be in bright filamentary structures. These bright structures are a result of a favorable viewing geometry and a high local opacity for the stellar radiation responsible for ERE excitation, thus leading to exceptionally large column densities of ERE emitters, confined to a space with narrow lateral dimensions. The resulting ERE structures can be imaged with the unsharp masking technique (Malin & Zealey, 1979), which permits the simultaneous display of sharp features ranging in intensity over four orders of magnitude. Witt & Malin (1989) applied this technique with good success to NGC The location, width and orientation of numerous ERE filaments was revealed, enabling subsequent studies of spatial correlations with filamentary emissions by fluorescent molecular hydrogen (Field et al., 1998) and by CN (Fuente et al., 1995). Imaging Spectropolarimetry Scattered light in reflection nebulae is substantially linearly polarized at optical and near-ir wavelengths, typically at the 20% to 30% level. Watkin et al. (1991), through the use of imaging polarimetry, discovered that localized ERE filaments in NGC 7023 are associated with a significant reduction of linear polarization in the R- and I-bands, consistent with a dilution of polarization through the presence of unpolarized emission with an intensity equal to that of the ERE, as determined by independent spectroscopic and photometric techniques. This demonstration that ERE is unpolarized makes attempts to explain the ERE in terms of scattering with unusual dust charac-

28 13 teristics (e.g. Greenstein & Oke, 1977) unsustainable. Rather, the ERE must be the result of an unpolarized emission process Observational Constraints The identification of the ERE carrier remains an outstanding challenge. At this stage, an extensive set of observational data exist which any ERE carrier candidate must meet. Finding some material which produces approximately the correct spectrum of some ERE source is simply not sufficient. We now summarize the observational information about the ERE and try to deduce the resulting constraints for models. ERE is a PL process. The ERE has been observed in many astrophysical environments: reflection nebulae, HII regions, dark nebulae (illuminated by the ISRF), high-latitude Galactic diffuse ISM and cirrus, dusty ISM of external galaxies, carbon-rich planetary nebulae Carriers must consist of cosmically abundant refractory elements that are depleted from the gas phase in interstellar space and are capable of forming photoluminescent materials. Such elements include C, O, Si, Fe, and Mg. Carrier particles must survive under a wide range of interstellar and circumstellar conditions and have a Galaxy-wide distribution. The PL occurs in a broad, unstructured band with a peak wavelength varying between 600 nm and > 900 nm, within given sources and from source to source, in response to increasingly dense and hard illumination by UV photons. The width and the peak wavelength of the ERE band are positively correlated. Under conditions of low radiation density the ERE quantum yield is 10%. The ERE is unpolarized with an isotropic radiation pattern. The ERE peak wavelength and quantum yield are strongly affected by the density and hardness of the local UV photon field.

29 14 The ERE is observed in the absence of scattering and appears therefore not associated with sub-micron-sized scattering grains. The ERE appears to be uncorrelated with the strength of the interstellar 2175 Å extinction feature and the ubiquitous UIR emission bands. The ERE carrier particles appear to originate in such dust-forming environments as the proto-planetary Red Rectangle nebula and in C-rich planetary nebulae Models for the ERE Carrier None of the currently discussed unified models for interstellar dust (Draine, 2004; Dwek et al., 2004) specifically predict or account for the existence of ERE and its observed characteristics. Only one attempt (Zubko et al., 1999) has been made to simultaneously meet requirements posed by the wavelength dependence of extinction, abundance constraints, and ERE observations by including silicon nanoparticles as a small-grain component into standard dust models. Lack of adequate knowledge of the size-dependent dielectric functions of likely materials in the nanoparticle regime is currently preventing further work in this direction. Most of the past effort has been focused purely on identifying classes of carrier particles capable of producing the ERE. These have included carbon-based solids, such as hydrogenated amorphous carbon (HAC) (Duley, 1985), quenched carbonaceous composite (QCC) (Sakata et al., 1992), coal (Papoular et al., 1996), and bacterial pigments (Hoyle & Wickramasinghe, 1999), as well as large carbon-based molecules, such as PAHs (d Hendecourt et al., 1986; Leger et al., 1988), C 60 (Webster, 1993) and PAHclusters (Seahra & Duley, 1999). Non-carbon bearing ERE candidates such as silicon nanoparticles (Witt et al., 1998; Ledoux et al., 1998) and silicates (Koike et al., 2002)

30 15 have also received close attention. Hydrogenated Amorphous Carbon Hydrogenated amorphous carbon (HAC) is an organic refractory material consisting of a mixture of sp 2 and sp 3 coordinated hydrocarbons with a variable band-gap. The HAC ERE model attracted wide support early on, because not only did HAC exhibit PL in apparent agreement with observed ERE spectra (Witt & Schild, 1988), it also appears to be an interstellar dust component required to explain the interstellar 3.4 µm C-H stretch feature (Pendleton & Allamandola, 2002; Furton et al., 1999) and is widely employed as an integral component of models for interstellar dust. The early HAC PL spectra were obtained with exciting radiation near 500 nm wavelength and they displayed luminescence only in the red region with a spectrum similar to that of the ERE. Under UV illumination, however, the HAC PL spectrum extends well into the blue and near-uv, consistent with a band-gap of over 3 ev (Robertson, 1996; Rusli et al., 1996). Under astrophysical illumination conditions with ample UV photons, HAC would therefore exhibit PL in the 400 nm to 500 nm spectral range, contrary to observation. The HAC band-gap can be narrowed by irradiation or annealing, bringing it closer to the observational constraints, but this also results in a dramatic decrease of the PL quantum yield by three or four orders of magnitude. As a result, HAC is not able to reproduce the spectral characteristics and the observed quantum yield of the ERE simultaneously, and the HAC ERE model is no longer considered viable.

31 16 PAH Molecules and Ions Large PAH molecules and their ions are considered likely candidates for explaining the ubiquitous UIR emission bands (discussed in more detail in 1.3.1) in the ISM and in many of the nebular sources in which ERE is observed as well. Given that PAHs are also known to fluoresce efficiently (Berlman, 1965), their role as a potential source of ERE was proposed early on (d Hendecourt et al., 1986). Under collision-free conditions and with far-uv photons available for excitation, PAHs and their ions are even expected to yield several luminescence photons per single excitation, resulting in efficiencies of several 100% (Leger et al., 1988). However, neutral PAHs luminesce predominantly in the UV/optical wavelength range (300 nm nm) (Birks, 1973; Peaden et al., 1980), with similar efficiencies in the gas phase (Stockburger, 1973; Sgro et al., 2001; Reylé & Bréchignac, 2000) as in solution (Berlman, 1965). Studies of the charge state distribution of PAHs in different interstellar environments by Bakes & Tielens (1994) and Weingartner & Draine (2001) agree that a significant fraction of PAHs should still be neutral in radiation environments typical of reflection nebulae. A size distribution of PAH molecules would therefore reveal its presence by luminescing not only in the red ERE range but in the blue wavelength region as well. Laboratory studies of the PL characteristics of larger PAH molecules, especially when excited by far-uv photons in a collision-free environment, are needed to evaluate the likelihood that PAH molecules are the origin of the ERE. Relatively little is known about the fluorescence characteristics of PAH cations. The comparatively small energy difference ( 1 ev) between the electronic ground

32 17 state, D 0, and the first excited electronic state, D 1, in most PAH cations (Leach, 1987) causes their potential fluorescence to occur in the near-ir and not in the optical range (Crawford et al., 1985). PAH cations are reported to absorb predominantly at energies > 7.75 ev (< 160 nm) (Robinson et al., 1997), which results in an increased probability of internal conversion to highly excited vibrational levels of the electronic ground state D 0, from where relaxation occurs essentially by emission of IR photons (Leach, 1995). This suggests that quantum yields for fluorescence by PAH cations may be quite low. The dications are similar to neutral molecules having closedshell structures and singlet and triplet energy levels. Even less is known about the fluorescence properties of PAH dications. The absence of consistent spatial correlations between ERE and UIR band emissions attributed to PAHs is not by itself conclusive evidence against the PAH origin of ERE, because PAHs in different size ranges, degrees of hydrogenation, and charge states, all of which are expected to be environment-dependent, could be responsible for UIR bands and ERE to a different extent (Duley, 2001). PAH Clusters Spectroscopic data on fluorescence by specific PAH molecules are limited to systems with atomic weight < 500 amu or about 40 carbon atoms. Duley & Seahra (1998) suggested that interstellar PAHs exist as much larger molecules or in the form of stacks and aggregates of PAHs of up to 700 carbon atoms, which could be responsible for the interstellar 2175 Å absorption feature as well as the UIR band emissions. They also proposed (Seahra & Duley, 1999) that these same PAH clusters will pro-

33 18 duce PL emission with a principal band centered at 700 nm, which would account for the ERE. This model suffers from several weaknesses. The model predicts two emission side bands, with peaks at 500 nm and 1000 nm wavelength, to be associated with the main ERE band. A search for these additional bands in two of the strongest ERE filaments in NGC 7023 by Gordon et al. (2000) failed to produce evidence for either one. The model predicts a remarkably constant peak wavelength for the main ERE band for a wide range of size distributions of PAH clusters, contrary to the observational evidence. Finally, the model predicts correlations between the strengths of the 2175 Å band and the ERE band, which are not supported by observations. The most attractive feature of this model is the efficient utilization of the same dust component to explain several interstellar phenomena, namely the 2175 Å band, ERE, UIR bands. It would be highly desirable to obtain laboratory data on PAH clusters which could support the multiple claims made for this model (Duley, 2001). Other Carbonaceous Carrier Models The C 60 molecule exhibits a PL spectrum resembling that of the ERE in a few reflection nebulae, leading Webster (1993) to suggest C 60 as the source of ERE. However, sensitive searches for the presence of C 60 and C + 60 in interstellar space and in NGC 7023 (Snow & Seab, 1989; Moutou et al., 1999; Herbig, 2000) failed to find observable traces of this species. Furthermore, the measured quantum yield of C 60 PL is only (Kim et al., 1992), missing the ERE requirement by several orders of magnitude. Discharges through gas mixtures containing hydrocarbons result in non-volatile

34 19 carbonaceous residues with interesting PL characteristics. Wdowiak et al. (1989), using a gas mixture containing CO, CH 4, N 2, H 2 O, and Ar, with relative abundances approximating those believed to apply to the Red Rectangle, produced luminescing residues of high quantum yield. However, under UV illumination the PL spectrum consistently peaked at wavelengths < 600 nm, with substantial parts of the luminescence spectrum extending through the nm range. The spectral mismatch with the observations plus the presence of blue PL appear to rule out this particular candidate. Quenched carbonaceous composite (QCC), a filmy residue produced through a microwave discharge in a low-density methane plasma by Sakata et al. (1983), was shown to be a more promising ERE carrier candidate. This material exhibits PL spectra with variable peak wavelengths spanning the nm range, depending on the substrate temperature during deposition. Mass spectroscopy of QCC (Sakata et al., 1983) indicates that QCC does not contain large PAH molecules but appears to consist of a random agglomeration of 1- to 4-ring PAHs, small aliphatic molecules and radicals. When dissolved in liquid freon, QCC exhibits blue luminescence, reminiscent of the fluorescence of small PAHs. A more complete characterization of QCC through additional laboratory work would clearly be desirable. Silicon Nanoparticles In recent years, silicon nanoparticles (SNPs) (Witt et al., 1998; Ledoux et al., 1998, 2000, 2001, 2002) have emerged as an alternative to the carbonaceous ERE carrier candidates discussed above. The SNP model meets the observational con-

35 20 straints posed by ERE observations in respect to spectral variability and quantum yield better than most of the carbonaceous candidates. The photophysics of SNPs (Smith & Witt, 2002) also appears to be consistent with the observed variations of ERE peak wavelength and ERE quantum yield with environmental conditions. The balance between photo-ionization and recombination with free electrons, combined with the fact that charged SNPs are much less likely to luminesce, explains the variation of the ERE quantum yield with UV-radiation density, while the expected photo fragmentation of multiply charged SNPs leads to the erosion of the SNP size distribution, starting with the smallest sizes, and a corresponding shift to much longer ERE peak wavelengths in astrophysical environments with the highest radiation densities. Studies of the optical characteristics of SNPs (Amans et al., 2003) show the absorption coefficient still rising at the wavelength of 200 nm, where current measurements stop. This is consistent with ERE excitation requirements, but measurements at shorter wavelengths are urgently needed. The highly efficient PL exhibited by SNPs is the result of quantum confinement in nanocrystals and passivation of surface dangling bonds with atoms such as H, O, N, C, and Fe. Most existing laboratory studies have been done with O- and H-passivation, and it has been found that the surface composition affects the PL spectrum (Wolkin et al., 1999; Zhou et al., 2003), in the sense that SNPs with H- passivation and diameters of < 2.5 nm luminesce at blue and near-uv wavelengths, while O-passivated SNPs luminesce in the red only. Witt et al. (1998) suggested that interstellar SNPs would most likely be O-passivated. Once produced in a circumstellar outflow, SNPs with oxygen passivation would persist even in the ISM with abundant

36 21 H-atoms, because the Si-O bond is energetically two to three times stronger than the Si-H bond. Such particles would contribute to the widely observed 9.7 µm vibrational Si-O absorption band, generally referred to as the silicate feature. However, Li & Draine (2002) have pointed out that stochastically heated O-passivated SNPs would produce a 20 µm emission band well in excess of currently established observational limits. This objection might be overcome, if SNPs with O-passivation were either attached to or embedded in larger grains. Experiments, in which SNPs were produced by ion implantation in solids with subsequent annealing, have demonstrated that SNPs preserve their luminescent characteristics even when embedded (e.g. Iwayama et al., 2002). Another way to overcome the objection by Li & Draine (2002) is to consider other forms of passivation. For example, recent laboratory work (Mavi et al., 2003) has shown that passivation with Fe-atoms yields even more persistent and efficient photoluminescence in SNPs than O-passivation. Carbon should be able to provide passivation equally well, although experiments in this direction have yet to be done. The existence of Fe- or C-passivated SNPs could also overcome the observational constraint, which so far has restricted ERE detections to C-rich environments among planetary nebulae (Furton & Witt, 1992). The SNP model, while promising, also suffers from a number of loose ends. The study of the optical properties of SNPs must be pushed into the vacuum ultraviolet, the full range of possible passivation schemes must be explored, the process of SNP formation in stellar outflows must be examined, and the question of the long-term persistence of the PL of SNPs under astrophysical conditions over astronomical times scales must be investigated.

37 22 Thus the ERE is a well-characterized photo-luminescence property of dust existing in a wide range of astrophysical environments. The carrier of the ERE is a major component of interstellar dust: it intercepts 20% of the UV/optical photons in the diffuse ISM, if the intrinsic quantum yield is 50%, and a still higher fraction of the UV/optical photons, if the yield is lower than 50%. Yet, no current comprehensive model for interstellar dust explicitly accounts for the existence of the ERE. 1.3 PAHs in the ISM Polycyclic aromatic hydrocarbons (PAHs) are a class of very stable organic molecules made up of only carbon and hydrogen. These molecules are flat, with each carbon having three neighboring atoms much like graphite. The structures of a variety of representative PAHs can be seen in Figure 1-4. On earth, they are a standard product of combustion from automobiles and airplanes and some (such as benzo[a]pyrene) are present in charcoal broiled foods. These molecules are considered highly carcinogenic but they are also very common and stable in terrestrial environments.

38 23 Figure 1-4 Structures of some small PAHs Unidentified Infrared Bands or the Aromatic Emission Features PAHs are ubiquitous in space and form the class of largest known molecules in the ISM. The infrared spectra of a wide variety of objects associated with dust and gas including the diffuse interstellar medium (ISM), the edges of molecular clouds, reflection nebulae, young stellar objects, H ii regions, star forming regions, some C-rich Wolf-Rayet stars, post-agb stars, planetary nebulae, novae, normal galaxies, starburst galaxies, most ultra-luminous infra-red galaxies and AGNs are dominated by emission features at 3.3, 6.2, 7.7, 8.6, 11.2 and 12.7 µm (see Peeters et al., 2004, and

39 24 references therein). Approximately 20-30% of the Galactic IR radiation is emitted in these UIR bands and 10-15% of the interstellar carbon is contained in the UIR carriers (Snow & Witt, 1995), indicating that the carriers represent an abundant component of the ISM. Since the carriers of these features remained a mystery for almost a decade, the initial name for these features, the Unidentified Infrared (UIR) emission features, is still in use. We shall henceforth refer to these bands as the aromatic emission features (AEFs). The AEFs coincide with the vibrational modes characteristic of aromatic materials (Duley & Williams, 1981) and while the vibrational spectrum of (almost) any exclusively aromatic material can provide a global fit to the observed AEFs, they are generally attributed to mostly large aromatic hydrocarbon species. Figure 1-5 shows examples of rich AEF spectra. It should be noted that not all sources show all these emission features at the same time and their peak position and relative strength vary.

40 25 Figure 1-5 The ISO-SWS spectra of the planetary nebula NGC 7027 and the Photo- Dissociation region at the Orion Bar illustrate the richness and variety of the UIR/AEF spectrum. Also indicated are the aromatic mode identifications of the major UIR features (from Peeters et al., 2004) Optical Fluorescence Although the AEFs are attributed to PAHs, their specific sizes and ionization states remain elusive. The vibrational transitions responsible for the AEFs are largely independent of size, structure and ionization state of the molecule. However, electronic fluorescence, a transition from the upper excited level to the ground state, is

41 26 more specific (Reylé & Bréchignac, 2000). The processes which occur between the absorption and emission of light are best illustrated by a Jablonski diagram as shown in Figure 1-6. The S s represent singlet states and the T s triplet levels. Each electronic level has many vibrational and rotational levels, only vibrational levels are shown in the figure. The absorption of a photon leads to an electron being raised from the ground state to one of several vibrational levels of the upper excited states. With a non-zero probability, this electron loses its excess vibrational energy by relaxing to the first excited state with the emission of infra red photons. If there are no other competing channels of relaxation, the molecule fluoresces as the electron relaxes from the zero vibrational level of the first excited state to one of the vibrational levels of the ground state. Provided the lifetime of the excited singlet state S 1 is sufficiently long for the excited molecules to attain thermal equilibrium, the fluorescence emission occurs primarily from the lowest vibrational levels of S 1. The fluorescence spectrum thus displays the vibrational spacing of the ground state. These are fast transitions and the intensities of the bands are governed by the Franck-Condon principle. Because electronic excitations do not greatly alter nuclear geometry, the vibrational energy levels of the excited state are similar to those in the ground state, and the absorption and fluorescence spectra display symmetric patterns. Another general property of fluorescence is that the same fluorescence spectrum is generally observed irrespective of the excitation wavelength. This is known as Kasha s rule.

42 27 IC IR S 2 IR : infrared emission ISC : inter system crossing IC : internal conversion IR IC ISC absorption IR fluorescence S 1 IR ISC IR phosphorescence T 1 S 0 Figure 1-6 Jablonski diagram Laboratory Measurements of PAH Fluorescence Laboratory spectra of PAHs are mostly obtained in solution. Gas-phase spectra are more relevant for comparison with astronomical comparison, and there are significant differences between spectra obtained in different ways as demonstrated by Chi et al. (2001a,b). Figure 1-7 illustrates the differences in case of anthracene and pyrene. Spectra obtained at different temperatures also exhibit systematic differences as shown in Figure 1-8. Therefore only gas-phase spectra obtained at appropriate temperatures should be used for comparison with astronomical fluorescence spectra.

43 28 Figure 1-7 Fluorescence spectrum of anthracene and pyrene in gas-phase and in solution (from Chi et al. (2001a)) Figure 1-8 Fluorescence spectrum of anthracene and pyrene at different temperatures (from Chi et al. (2001a))

44 Formation of PAHs in the ISM Most of the carbon in the outflows from C-rich AGB stars is tied up in CO and C 2 H 2. Since CO is stable, acetylene and its radical derivatives are likely to be the dominant precursor molecules from which PAHs are formed. Figure 1-9 depicts two routes of PAH formation under different pressure/temperature conditions. Figure 1-9 Formation routes for PAHs in the ISM (from Keller, 1987)

45 Outline The chapters in this dissertation follow the format of papers in the Astrophysical Journal, with sections on introduction, observation, results and discussion. The references are consolidated at the end and footnotes to chapter titles provide information on the publication details. Chapter 2 presents the discovery of the blue luminescence (BL) emission in the Red Rectangle (RR) and its tentative identification as fluorescence by small, neutral PAHs. Chapter 3 presents details of further investigation of the BL in the RR, the spatial correlation of the BL distribution with the 3.3 µm PAH emission and a study of the attenuation characteristics of the central star in the RR. These provided further evidence for the identification of the BL as fluorescence from 3-, 4- ringed PAHs. In Chapter 4 we present the detection of BL in other ordinary reflection nebulae, establishing the BL as a more general photoluminescence phenomenon exhibited by interstellar dust. In Chapter 5 we present our investigation of the excitation wavelength of the ERE, which points to PAH dications as possible carriers of the ERE. In Chapter 6 we report on the results of a narrow-band imaging of the RR, which reveals the changing morphology of the nebula as probed by the different emissions and in Chapter 7 we present the detection of the resolved BL spectrum uncontaminated by scattered light at large offsets from the central star in the RR. Finally we summarize the main results of this work on Chapter 8.

46 Chapter 2 Discovery of Blue Luminescence in the Red Rectangle: Possible Fluorescence from Neutral Polycyclic Aromatic Hydrocarbon Molecules? 1 Here we report our discovery of a band of blue luminescence (BL) in the Red Rectangle (RR) nebula. This enigmatic proto-planetary nebula is also one of the brightest known sources of extended red emission as well as of unidentified infra- 1 Appeared in The Astrophysical Journal Letters 2004, Vol. 606, pg. 65, Authors: U. P. Vijh, A. N. Witt, & K. D. Gordon 31

47 32 red (UIR) band emissions. The spectrum of this newly discovered BL is most likely fluorescence from small neutral polycyclic aromatic hydrocarbon (PAH) molecules. PAH molecules are thought to be widely present in many interstellar and circumstellar environments in our galaxy as well as in other galaxies, and are considered likely carriers of the UIR-band emission. However, no specific PAH molecule has yet been identified in a source outside the solar system, as the set of mid-infra-red emission features attributed to these molecules between the wavelengths of 3.3 µm and 16.4 µm is largely insensitive to molecular sizes. In contrast, near-uv/blue fluorescence of PAHs is more specific as to size, structure, and charge state of a PAH molecule. If the carriers of this near-uv/blue fluorescence are PAHs, they are most likely neutral PAH molecules consisting of 3-4 aromatic rings such as anthracene (C 14 H 10 ) and pyrene (C 16 H 10 ). These small PAHs would then be the largest molecules specifically identified in the interstellar medium. 2.1 Introduction The family of emission bands at 3.3, 6.2, 7.7, 8.6, 11.2, & 12.7 µm (called the unidentified infra-red (UIR) bands) is found in almost all astrophysical environments including the diffuse interstellar medium (ISM), the edges of molecular clouds, reflection nebulae, young stellar objects, HII regions, star forming regions, some C-rich Wolf-Rayet stars, post-agb stars, planetary nebulae, novae, normal galaxies, starburst galaxies, most ultra-luminous infra-red galaxies and AGNs (see Peeters et al., 2004, and references therein). Approximately 20-30% of the Galac-

48 33 tic IR radiation is emitted in these UIR bands and 10-15% of the interstellar carbon is contained in the UIR carriers (Snow & Witt, 1995), indicating that the carriers represent an abundant component of the ISM. These UIR bands are the signatures of aromatic C-C and C-H fundamental vibrational and bending modes, and are generally attributed to a family of PAH molecules containing carbon atoms (Hony et al., 2001; Cook & Saykally, 1998; Leger & Puget, 1984; Verstraete et al., 2001; Allamandola et al., 1985; Sellgren, 1984). Although the presence of PAHs in space is widely accepted by most astronomers, their specific sizes and ionization states remain elusive. On absorption of a far-uv photon, a PAH molecule usually undergoes a transition to an upper electronic state. If the molecule undergoes iso-energetic transitions to highly vibrationally excited levels of the ground state, then the molecule relaxes through a series of IR emissions in the C-C and C-H vibrational and bending modes. These transitions are largely independent of size, structure and ionization state of the molecule. However, electronic fluorescence, a transition from the upper excited level to the ground state, is more specific (Reylé & Bréchignac, 2000). In particular, the wavelength of the first electronic transition seen in neutral PAHs is closely dependent on the size of the molecular species. In general, this fluorescence wavelength increases with the molecular weight of the molecule (Fig. 2-1). An observation of fluorescence in an astronomical source in the UV/visible range offers the possibility of estimating the size of the PAH molecules, which the observation of the UIR band emission in the same source does not.

49 peak λ em (Å) condensed PAHs non condensed PAHs Mol. wt. (a.m.u.) Figure 2-1 The first electronic transition in the fluorescence spectra (Berlman, 1965; Peaden et al., 1980; Ewald & Garrigues, 1985, 1988) of neutral PAH molecules as a function of their molecular weight. We distinguish between condensed and noncondensed PAHs, with condensed PAHs considered to be more stable under interstellar conditions. 2.2 Target and Observations The Red Rectangle (RR) (Cohen, 1975; van Winckel, 2003) is a unique protoplanetary nebula. Its bipolar structure is a result of mass loss from an evolved central binary star, HD 44179, directed by a circumbinary disk. Different aspects of the nebula become apparent in images at different wavelength regions shown in Figure 2-2. At blue wavelengths (3800 Å < λ < 4900 Å bottom panel) the nebular structure is dominated by a bright spherical blob of about 8 diameter, embedded in a faint rectangular envelope with approximate dimensions of The blue

50 35 spectrum of the nebula is dominated by dust-scattered light of the central A-type star. At red wavelengths (6220 Å) the nebula appears to have sharp radial structures, representing the projected walls of a bi-conical outflow cavity, as shown in the high-resolution Hubble Space Telescope WFPC2 image in the middle panel. The nebular spectrum in the red region (λ > 5400 Å) is dominated by Extended Red Emission (ERE) (Witt & Vijh, 2004; Witt & Boroson, 1990), following a spatial distribution (Schmidt & Witt, 1991) totally different from that of the dust-scattered light at blue wavelengths (λ < 5000 Å). The carrier of the ERE is still under investigation (Witt & Vijh, 2004). The top panel shows a high-resolution short-exposure WFPC2 near-ir image of the central source, demonstrating in fact that the star itself is not directly visible from Earth. Instead, we see two blobs of scattered light above and below a disc (Bujarrabal et al., 2003), which obscures the star at visible wavelengths. The RR nebula has the distinction of being the brightest known source of UIR band emission (Russell et al., 1978; Geballe et al., 1985) which is attributed to PAHs. In addition, the central star HD is thought to be a post-agb star in its first stage of evolution toward a planetary nebula. It is in this stage that substantial massloss takes place and a bi-polar structure develops. This stage in stellar evolution is considered a stage of active dust-production (Whittet, 2003). Therefore, the RR nebula is a likely location in which to find observational evidence for small PAH molecules, which are the building blocks for larger PAH structures (Keller, 1987) and carbon grains. In most other sources of UIR band emission, the emitting particles are a component of interstellar dust, which is likely to have experienced substantial

51 36 processing by UV radiation, interstellar shocks and chemical modification over a very long time span of existence since leaving the environment of its original formation. It is well established that PAH molecules with less than 40 C-atoms are unlikely to survive the harsh radiation conditions of interstellar space (Jochims et al., 1999), while they may well be abundant in the more benign, UV-poor radiation field of HD 44179, especially when partially shielded by the opaque disk apparent in Figure 2-2.

52 AAT Blue WFPC2 F622W WFPC2 F814W 80" 30" 2.5" Figure 2-2 The RR nebula. These images illustrate the observable structures of the RR nebula at different optical wavelengths and different angular scales. The left-most panel shows a blue image (3800 Å Å), obtained using the 4-m AAT. The middle panel shows a HST WFPC2 image centered at a wavelength of 6220 Å. The geometry of the central illuminating source, obscured from direct visibility by an optically opaque circumstellar disc, is shown in the right-most panel, obtained with the HST WFPC2 at a waveband centered at 8140 Å. The angular scales of each of the three frames are indicated on the bottom. In all panels north is to the right and east to the top. The orientation and width of the two CTIO spectrograph slits are indicated on both the bottom and middle panels, and are labeled CTIO 1 and CTIO 2, respectively. 37

53 38 We obtained low-resolution, long slit spectra of the RR with the R-C (Cassegrain) spectrograph at the Cerro Tololo Inter-American Observatory (CTIO) 1.5-m telescope. These observations were made on March 26, 2003, using a 2. 5 wide slit, 7. 7 long. The grating #09 with 300 l mm 1, blazed at 4000 Å, provided a 8.6 Å resolution and a spectral coverage of 2600 Å. Using a CuSO 4 filter to select the grating s first order, the setup covered a wavelength range of Å including the range from H β to the Balmer discontinuity. The Loral IK CCD detector yielded a spatial scale of 1.3 pixel 1 along the slit. We used a coronographic decker assembly to minimize scattered light from the star while probing as much of the inner nebula as possible. All observations were made using the full extent of the 7. 7 long slit to get simultaneous sky observations. We obtained spectra at two nebular locations, 2. 5 and 5 south of the central star HD with the slit in east-west direction. The nebular exposures were bracketed by exposures of the central star. Individual exposures were limited to 5-10 minutes on the nebula and 1 minute for the star, and 4-5 exposures were obtained for each orientation. Data Reductions were carried out with IRAF 2.12 EXPORT, and all spectra were flux calibrated via observation of standard stars. 2.3 Results Line-depth Technique We used the line-depth technique to initially detect and measure the fluorescence spectrum at each of the two nebular positions. This technique allows the identifi-

54 39 cation and measurement of any continuous emission in a reflected spectrum, such as represented by the blue spectrum of the RR, relying upon the comparison of the depths of nebular and stellar spectral absorption lines. It works particularly well with the strong hydrogen Balmer lines at 6562, 4861, 4340, 4101, 3970, 3889, 3835, 3798, 3770 Å and so on. In the absence of any emission processes such as continuous fluorescence by PAH molecules, the relative line-depths in the reflected spectrum would be identical to the corresponding ones in the spectrum of the illuminating star, assuming that the wavelength-dependent scattering properties of the dust, e.g albedo and phase function asymmetry, remain unchanged over the few-angstrom width of each individual line. The presence of any continuous emission manifests itself by the lines in the reflected spectrum having smaller line depths than those in the illuminating source spectrum. The illuminating source in this case is the star HD 44179, having an ideal spectral type A with strong Balmer lines. Also, at T eff 9000 K the star is not likely to produce Balmer line emission in the surrounding nebula. The decrease in the depth of a line can be directly related to the amount of underlying continuous emission (in this case, the fluorescence intensity). As the star is enshrouded by the disk and as the stellar spectrum is actually seen in reflected light, if there were any fluorescence in this zero-offset spectrum, the line-depth technique would reveal any additional fluorescence at the offset positions in the nebula. As an example we show in Figure 2-3 the normalized spectra at the zero offset position and at an offset 2. 5 south, 3. 9 east of H ζ at 3889 Å. One can see that the line-depth at the nebular position is substantially diminished.

55 Normalised Intensity λ (Å) S, 3.9 E Figure 2-3 Normalized spectra from two positions, zero offset and 2. 5 south, 3. 9 east at 3889 Å, illustrating the filling in of the Balmer line due to fluorescence. Error bars are photon-statistics errors. In the presence of any continuous emission the relative line depth in the nebula R N = (C N L N )/C N is smaller than the relative line depth R S = (C S L S )/C S of the same line in the illuminating star, where C and L refer to continuum and line center intensities, respectively, and where the subscripts N and S refer to the nebular and stellar spectrum, respectively. The fluorescence intensity relative to the scattered nebular continuum can therefore be expressed as R S /R N 1 (see Witt & Vijh, 2004, for a review). Given the nature of our measuring technique, we can determine the fluorescence intensity only at the wavelengths of the resolvable Balmer lines as well as the Balmer discontinuity, i.e. at ten wavelength positions with our spectral resolution in our

56 41 spectral range. By measuring the line-depths of each of the Balmer lines in the star and the nebula, we extracted the fluorescence intensities at those wavelengths. Table 2.1 shows the fluorescence intensities at the two offsets: 2. 5 south, 3. 9 east and 5 south, 6. 5 east. Independent observations by the authors using the 90-in. Bok telescope (Steward Observatory) and the B&C spectrograph show corroborating evidence for this UV/blue fluorescence in the RR (U. Vijh, A. Witt, & K. Gordon, in preparation). Table 2.1 Fluorescence Intensity at two positions in the RR nebula. Wavelength Intensity (erg cm 2 s 1 Å 1 sr 1 ) 2. 5 south, 3. 9 east 5 south, 6. 5 east ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Errors are combined photon-statistics errors Identification Hydrogenated amorphous carbon (HAC), silicon-nanoparticles (SNP), SiC grains, and small PAHs are all known to luminesce at blue wavelengths and have also been considered as likely candidates to explain the ERE (Witt & Vijh, 2004). Figure 2-4

57 42 shows fluorescence spectra of HAC, SNP and SiC (Li et al., 2000; Huang et al., 2002; Liao et al., 2002; Belomoin et al., 2000; Patrone et al., 2000). It reveals that the observed BL in the RR is unlike the luminescence spectrum of HACs, SNPs, or SiC. Given the exceptionally strong UIR-band emission in the RR, the most likely source of the newly discovered BL may then be fluorescence by PAH molecules. The nebular BL spectrum suggests a peak in the fluorescence intensity corresponding the first electronic transition S 1 - S 0 near 3750 Å. Comparison with Figure 2-1 suggests PAH molecules in the mass range of 170 to 270 amu as the likely sources. Shown in Figure 2-4 are comparisons with laboratory fluorescence spectra (Chi et al., 2001a,b; Reylé & Bréchignac, 2000) of gas-phase PAHs in that mass range with the BL spectrum of the RR at an offset of 2. 5 south, 3. 9 east.

58 43 Normalized Intensity RR 2.5"S, 3.9"E HAC film from CH 3 + HAC film H capped SNP SNP Patrone SiC Normalized Intensity Wavelength λ (Å) RR 2.5"S, 3.9"E Pyrene Anthracene Phenanthrene Chrysene Naphthalene Wavelength λ (Å) Figure 2-4 The UV/blue fluorescence spectra of the RR nebula compared to PAH and solid-state luminescence. (a), comparison with some solid-state candidates and (b), comparison with small neutral PAHs. The intensities are normalized for easy comparison of the spectra with existing laboratory fluorescence spectra. The dashed lines representing the RR fluorescence spectrum are drawn to guide the eye (measurements exist only at specific wavelengths indicated by solid circles).

59 44 Laboratory spectra obtained with PAHs in the gas-phase are more likely to be comparable to astrophysical spectra. These gas-phase spectra show considerable variation with temperature, becoming smoother, losing band-structure and the peak shifting to longer wavelengths with increasing temperatures (Chi et al., 2001a,b). Though anthracene suggests a good match to the observed spectrum, we do not have sufficient spectral resolution to completely rule out other possible candidates, as the measurement technique works only at the discrete wavelengths of the Balmer lines. 2.4 Discussion Another important factor which should be considered when identifying possible carriers is fluorescence efficiency. Although many different fluorescing species may exist with similar abundances, the spectra of the most efficient ones will be dominant. Laboratory studies have shown that on comparison of emission rates per molecule, from PAH molecules placed at 1 AU from the Sun, anthracene and pyrene show times more efficiency than naphthalene (C 10 H 8 ) and phenanthrene (C 14 H 10 ) (Bréchignac & Hermine, 1994). Additional support for small PAHs as sources of the observed BL is found in the spatial correlation between the BL and the PAH 3.3 µm emission in the RR and the presence of the distinct PAH ionization discontinuity in the spectrum of the central source, HD (U. Vijh. A. Witt, & K. Gordon, in preparation).

60 Chapter 3 Small PAHs in the Red Rectangle 1 Following our initial discovery of blue luminescence in the spectrum of the Red Rectangle (RR) and its identification as fluorescence by small three- to four-ringed polycyclic aromatic hydrocarbon (PAH) molecules, we report on the spatial correlation between the blue luminescence and the 3.3 µm emission, commonly attributed to small, neutral PAH molecules, and on the newly-derived UV/optical attenuation curve for the central source of the RR, HD Both results provide strong additional evidence for the presence of small PAH molecules with masses of less than 250 amu in the RR, which supports the attribution of the blue luminescence to fluorescence by the same molecules. We contrast the excellent spatial correlation of the two former emissions with the distinctly different spatial distribution of the extended red emission (ERE) and of the dust-scattered light within the RR. The UV/optical attenuation curve of the central star is unlike any interstellar extinction curve and is 1 Appeared in The Astrophysical Journal, Vol. 619, pg. 368, Authors: U. P. Vijh, A. N. Witt, & K. D. Gordon 45

61 46 interpreted as resulting from circumstellar opacity alone. Major contributions to this opacity are absorptions in broad bands in the mid-uv, contributing to the electronic excitation of the luminescing PAH molecules, and a sharp ionization discontinuity near 7.5 ev in the far-uv, which places a sharp upper limit on the masses of the PAH molecules that are responsible for this absorption. The strength of the far-uv absorption leads to an abundance of the PAH molecules of 10 5 relative to hydrogen in the RR. Such small PAHs are perhaps unique to the environment in the RR, where they are shielded from harsh radiation by the dense circumstellar disk. 3.1 Introduction Recently, we reported the discovery of blue luminescence (BL)(Vijh et al., 2004) in the spectrum of the bipolar, proto-planetary Red Rectangle (RR) nebula (Cohen, 1975; Cohen et al., 2004). The emission, peaking near 378 nm, is contained in a band of FWHM 45 nm. The band-integrated intensity of the BL is comparable to that of the scattered light intensity and of the intensity of the extended red emission (ERE), which is also exceptionally strong in this object (Schmidt et al., 1980; Witt & Boroson, 1990). This suggests that the BL is produced by a carrier of substantial abundance in the gas/dust medium in the RR. Through a comparison of the observed spectrum with laboratory photoluminescence spectra of a collection of likely dust components and with gas-phase fluorescence spectra of polycyclic aromatic hydrocarbon (PAH) molecules, Vijh et al. (2004) identified neutral three- and four-ring PAH molecules such as pyrene, anthracene

62 47 and phenanthrene as the most likely sources of the BL. This identification also appears consistent with several other important facts. Firstly, the RR is the brightest known source of emission in the unidentified infrared (UIR) features (Russell et al., 1978; Geballe et al., 1985), which are commonly attributed to PAH molecules and ions (Bakes et al., 2004); this suggests high relative abundances and optimal excitation conditions for such molecules. Also, as a post-agb star (van Winckel, 2003), the carbon-rich central star of the RR, HD 44179, is in an active mass-losing and dust-producing stage of its life, suggesting that PAH molecules are currently condensing in the carbon-rich outflow (Keller, 1987; Cherchneff et al., 1992). This includes, in particular, smaller PAH molecules that are able to exist in the relatively benign radiation environment of the RR, but which would not be viable after being ejected into the much harsher radiation field of interstellar space, where three- and four-ringed PAHs are not expected to survive (Jochims et al., 1994, 1999). And finally, anthracene and pyrene have exceptionally high fluorescence efficiencies, which would cause their spectra to dominate any fluorescence spectrum of a mixture of PAH molecules (Bréchignac & Hermine, 1994). If confirmed by further observations, the detection of three- and four-ringed PAH molecules would represent the detection of the largest specifically identified molecules so far observed outside the solar system. In this paper we report new observations and analysis dealing with the spatial distribution of the BL in the RR nebula. We show that the spatial distribution of the BL is distinctly different from that of the scattered light distribution in the same region of the spectrum and also distinctly different from that of the ERE. Significantly, there is an exceptionally close correlation between the distribution of

63 48 the BL and the distribution of the UIR-band emission at 3.3 µm, attributed to the C H stretch transition in neutral PAH molecules, which provides strong support for the identification of the BL as fluorescence from relatively small PAH molecules. In the second part of this paper we discuss the newly-derived attenuation curve for the central star, HD The attenuation is shown to be circumstellar rather than interstellar in nature and to be dominated by absorptions attributable to electronic bound-bound and bound-free transitions in small PAH molecules. 3.2 Observations Low-resolution, long slit spectra of the RR were obtained with the R-C (Cassegrain) spectrograph at the Cerro Tololo Inter-American Observatory (CTIO) 1.5 m telescope. These observations were made on 2003 March 26 and 29, using a 2. 5 wide slit, 7. 7 long. For observations in the blue ( nm), grating #09 with 300 l mm 1 was used which is blazed at 4000 Å, provides a 8.6 Å resolution and a spectral coverage of 260 nm. Using a CuSO 4 filter to select the grating s first order, the setup covered a wavelength range of nm, including the range from H β to the Balmer discontinuity. Grating #11 with 158 l mm 1, blazed at 8000 Å, with a 16.4 Å resolution was used in the first order with a GG495 cut-on filter for the red observations with a wavelength coverage from nm. The Loral IK CCD detector yielded a spatial scale of 1. 3 pixel 1 along the slit. A coronographic decker assembly was used to minimize scattered light from the star while probing as much of the inner nebula as possible. All observations were made using the full extent of the 7. 7 long slit to

64 49 get simultaneous sky observations. Spectra were taken at two nebular locations, 2. 5 and 5 south of the central star HD with the slit in E-W direction (PA = 90 ). The nebular exposures were bracketed by exposures of the central star. Individual exposures were limited to 5-10 minutes on the nebula and 1 minute for the star, and 4-5 exposures were obtained for each orientation. Data reductions were carried out with IRAF 2.12 EXPORT, and all spectra were flux calibrated via observation of standard stars. The 3.3 µm data are reproduced from Kerr et al. (1999). The spectroscopic observations were made at the United Kingdom Infrared Telescope (UKIRT) on 1995 December 11. The spectra were taken with a slit 90 long, 1. 2 wide, placed 5 south of the central star HD (PA = -85 ). For the FUV spectral energy distribution (SED) of HD archival observations from the International Ultraviolet Explorer (IUE) LWP 22416, SWP 38188, and LWR were used. David Malin (David Malin Images) provided us with one of the few existing deep blue images of the RR for the comparison of our spatially resolved spectra with morphological details of the nebular structure. The image, reproduced in Figure 3-1, was taken by David Malin with the AAT at the Anglo-Australian Observatory. The exposure time was 15 min, the detector was a Kodak IIa-O plate (blue sensitive) with a GG385 cut-on filter. The effective bandpass extends from 390 nm to 480 nm. A red image of the RR taken with the HST has been reproduced to aid in the correlation of the observed features with the nebular structure.

65 Analysis and Results Blue Luminescence in the Red Rectangle Nebular spectra were extracted in two-pixel bins along the two slits with 7 apertures along the 2. 5 south slit and 8 apertures along the 5 south slit with a spatial scale of 2. 6 per aperture. Thus, each spectrum represents a region of the nebula. Figure 3-1 shows the blue image of the nebula overlaid with the BL apertures and the 3.3 µm slit and Figure 3-2 is a red image overlaid with the central ERE apertures. The line-depth technique (Vijh et al., 2004; Witt & Vijh, 2004) was used to detect and measure the BL at the positions of the hydrogen Balmer lines for each spectrum. Measurement of the Balmer discontinuity in the nebular spectra compared to that in the stellar spectrum is used to infer the BL at the Balmer discontinuity (see Appendix A for details). Measurement of the equivalent widths of the absorption lines could also be used for a similar analysis, however a small uncertainty in the determination of the continuum levels over the steep spectral shape results in a relatively large uncertainty in the equivalent width. In contrast the errors in the line-depth measurements are essentially limited to the photon noise in the spectrum. The co-added nebular spectra had a signal-to-noise ratio per pixel of 10 in the outermost regions to 250 closer to the central star in the blue and a range of in the red region of the spectrum. The stellar spectra had a signal-to-noise ratio per pixel of 500 in both spectral ranges.

66 51 Figure 3-1 Blue image of the Red Rectangle with overlaid extraction apertures. The size of the image is The white dot is an overlaid pixel to indicate the position of the central source. The dashed lines indicate the slit used for the 3.3 µm observation. North is to the top, east to the left.

67 52 Figure 3-2 Red image of the RR with overlaid extraction apertures. The size of the image is The HST image (Credits: NASA, ESA, Hans Van Winckel (Catholic University of Leuven, Belgium), and Martin Cohen (University of California, Berkeley)) has been reproduced to aid the identification of the nebular structure with the extracted apertures. North is to the top, east to the left. Spectral Variability in the BL The BL spectrum is not identical in different regions of the nebula and certain trends can be noted. The primary peak in the spectrum is near 378 nm (Vijh et al., 2004, Fig. 4). A secondary peak around 397 nm starts to develop as we probe regions farther from the central star, more prominently along the 5 south slit. Also, farther

68 53 from the star, the BL intensities at longer wavelengths start to become stronger compared to the peak intensity. Interestingly, spectra that probe regions inside the conical outflow indicate that the BL in these regions has another peak at much shorter wavelengths, shortward of 360 nm. Figure 3-3 depicts the BL spectrum at three such representative locations: 2. 5 south, 7. 8 east, a position in the outer regions of the nebula; 2. 5 south, 2. 6 east, a region close to the central source, inside the bipolar outflow; 5 south, 5. 2 west, a region along the farther slit just outside the cone wall. Based on the existing correlation between the wavelength of peak intensity of the fluorescence spectrum and the molecular size of PAH molecules (Vijh et al., 2004), these trends can be interpreted to suggest a size variation in the BL carrier at different locations in the nebula: closer to the central source we see an emergence of a smaller population of PAHs and farther out we see increasing contributions from larger emitters. Given the limited spectral resolution inherent in our detection technique and only fair spatial resolution these trends are at best an indication of the size distribution of the BL emitters. We defer a more complete analysis of the spectral variation of the BL spectrum until after completion of an ongoing program in collaboration with D.G. York which maps the complete RR nebula with dense spatial coverage using the 3.5-m telescope at Apache Point Observatory.

69 ".5s 7".8e 2".5s 2".6e 5"s 5".2w Normalised Intensities λ (Å) Figure 3-3 Spectral variation of the BL. The solid line traces the intensity at 2. 5 south, 7. 8 east, the dashed line at 2. 5 south, 2. 6 east and the dash-dotted line at 5 south, 5. 2 west positions. Solid circles indicate wavelength positions associated with the 2. 5 south slit and solid squares indicate wavelength positions associated with the 5 south slit. Intensities are normalized to unity at the peak positions. Gradients of BL/Scattered Light Figure 3-4 (upper panel) shows the distribution of the ratio of band-integrated BL to the total scattered light in the same band at various positions along the two slits. Figure 3-4 (lower panel) illustrates the spatial distribution of the BL and the scattered light intensities on a normalized scale. Due to the anisotropy of the phase function, scattered light is strongly forward directed and falls off steeply as we probe regions where our line of sight penetrates regions of the nebula where larger scattering angles predominate. The BL, on the other hand is due to fluorescence and is an isotropic emission and thus falls off less steeply. Thus, the spatial variation of the BL and

70 55 scattered light along the two slits reveals the isotropic nature of the BL. Therefore, the ratio I BL /I Scat increases away from the star, whereas the actual intensities fall away from the star as the density of emitters and the exciting radiation both decrease. It is equally illuminating to note that the BL intensities have similar profiles along the two slits, whereas the distribution of the scattered light it quite different; along the slit closer to the star, the scattered light falls off more steeply than along the second slit as the relative change in the distance from the central star varies more slowly in the latter case.

71 10 2.5s-BL/Scat 5s-BL/Scat 56 BL/Scattered Light offset (arcsec) 2.5s-BL 5s-BL 2.5s-Scat 5s-Scat Normalised Intensities offset (arcsec) Figure 3-4 Ratio of Band-I BL to Band-I sc and normalized distributions. upper panel: Ratio of band-integrated BL to scattered light along the two slits. Solid line traces the ratio along the 2. 5 south slit and the dashed line that along the 5 south slit. lower panel: Normalized BL and scattered light intensities. Intensities are normalized to unity at 2. 6 east position. In both figures, solid circles denote spatial positions along the 2. 5 south slit and the solid squares those along the 5 south slit. Positive and negative offsets indicate positions to the west and east of the star HD respectively.

72 57 Correlations with Other Emissions Figure 3-5 displays the spatial distribution of the band-integrated intensities of ERE and the BL along the two slits and the 3.3 µm UIR-band emission along a 5 south (PA=-85 ) slit (Fig. 3-1 and 3-2 show the apertures). The 3.3 µm intensities (obtained from Kerr et al. (1999)) have been normalized to the intensity of the BL at the 0 east-west, 5 south position. This figure reveals a number of interesting facts. The 3.3 µm UIR-band emission and the BL are almost perfectly correlated, which we will discuss in more detail in the next paragraph. The spatial distribution of the BL is distinctly different from that of the ERE, the ERE is strongly peaked close to the central source and on the X-shaped walls of the outflow cone while the BL has a broader, more diffuse distribution. The 3.3 µm emission correlates exceptionally well with the BL along the 5 south slit, and the small differences are probably attributable to the slightly different widths and PA of the two slits (see Fig. 3-1). The 3.3 µm emission feature belongs to the distinctive set of mid-infrared emission features which have been attributed to emission from PAH molecules, which become vibrationally excited upon absorption of a UV photon and subsequently relax with the emission of IR photons (Bakes et al., 2004; Schutte et al., 1993), in addition to electronic relaxation through fluorescence or phosphorescence transitions. Most of the vibrational transitions are insensitive to molecular sizes, but the 3.3 µm emission attributed to C H stretch from PAH molecules is sensitive to the size of the molecular species. PAHs having 20 carbon atoms show 3.3 µm emission intensities 50 times stronger than that from PAHs

73 58 having 100 carbon atoms (Schutte et al., 1993). Earlier observations of the RR showed a pronounced lack of similarity between the 3.3 µm and the 11.3 µm surface brightness distributions (Bregman et al., 1993), the latter being attributed to C H out-of-plane bending modes in relatively large PAH molecules. Furthermore, the 3.3 µm emission traces neutral PAH species (Bakes et al., 2004), and becomes extremely weak in ionized PAHs. The fact that the BL follows the 3.3 µm distribution so closely indicates that the BL emitters and the 3.3 µm emitters are similar small, neutral PAHs and thus strengthens our identification of the BL as fluorescence by small neutral PAHs. The ERE, BL and the 3.3 µm emission are all isotropic emissions and yet the spatial distributions of the ERE is dissimilar to those of the BL and the 3.3 µm emission. The ERE falls off much more steeply than the other two that have a broader distribution. The ERE is largely confined to the bipolar, X-shaped structure of the nebula (see also Schmidt & Witt, 1991, Fig. 2,), whereas the BL is the dominant emission in the outer regions. The variation in the spatial distributions of these different emissions must be attributed the distribution of the carriers and the respective exciting radiations. The BL carriers, most likely small PAHs, are probably ionized inside the bipolar outflow cone, where they are exposed to direct illumination from the central RR source. Thus fluorescence in these regions is effectively quenched. Farther out, outside the cones, the molecules/carriers are shielded from the far-uv ionizing radiation but still receive through scattering the mid-uv exciting radiation and thus emit with fairly high intensities. The ERE carriers on the other hand seem more robust as the ERE intensities are high closer to the star inside the bipolar outflow cone and

74 59 along the walls (X-shaped structure). This could, in fact be interpreted as pointing toward some unknown ionized species as the carrier of the ERE. Outside the outflow region the ERE intensities drop by over three orders of magnitude, suggesting that either the carriers of the ERE do not exist in such regions or if they do exist they do not receive the exciting radiation needed to produce the ERE.

75 60 1 Band-ERE-2".5s Band-BL-2".5s 0.1 E W Intensities e offset (arcsec) Band-ERE-5"s Band-BL-5"s 3.3-5"s 0.1 E W Intensities e offset (arcsec) Figure 3-5 Spatial distribution of the BL and correlations with other emissions. Bandintegrated intensities of BL and ERE at positions along the two slits (upper panel: along the 2. 5 south slit and lower panel: along the 5 south slit) are plotted. The 3.3 µm emission is normalized to the value of the BL at 5 south, zero offset position. Filled circles indicate positions along the 2. 5 south slit and filled squares those along the 5 south slit. The intensities are band-integrated and are given in units of ergs cm 2 s 1 sr 1 for the BL and the ERE.

76 Attenuation of HD Visual Reddening and Attenuation The central source of the RR, HD 44179, is a single-line spectroscopic binary star (Waelkens et al., 1992, 1996; Van Winckel et al., 1995; Men shchikov et al., 2002) that is totally hidden from direct view at optical wavelengths by a circum-binary disk seen nearly edge on. Diffraction-limited imaging (Cohen et al., 2004) reveals the central source as a pair of scattered-light lobes above and below the disk, directing light from the hidden central stars via scattering by dust toward the Earth. The photometrically dominant component of the binary was classified as spectral type B9/A0 (Cohen, 1975) originally, but is now recognized as a post-agb star with an effective temperature near 8000 K and a highly metal-deficient atmosphere typical of post-agb stars (van Winckel, 2003; Men shchikov et al., 2002). The low metal abundance, e.g. [Fe/H] = -3, gives this star the spectroscopic appearance of a much hotter B9/A0 star, but with an intrinsic (B V ) color characterized by its actual effective temperature. By convolving the appropriate Kurucz atmosphere model (T eff = 8250 K, log g = 1.5, [Fe/H] = -3.0) with the B and V pass bands, we determine an intrinsic (B V ) = mag for the photometrically dominant component of HD We note that the nature of the secondary component is not known directly from observation, but its inferred mass of 0.35 M and the fact that the innermost region of the RR contains a small H ii region suggest that it is most likely a hot white dwarf with a luminosity of less than 2% of that of the primary (Men shchikov et al., 2002).

77 62 The observed (B V ) color of HD has been inferred from broadband photometry of the RR by Cohen (1975) as 0.39 mag., leading to an estimated color excess of E(B V) = 0.35 mag. The peculiar nature of HD 44179, its location within an optically thick disk, and its uncertain distance make it difficult to estimate the total visual attenuation of its light as seen from Earth. Here we adopt the latest model (Men shchikov et al., 2002) which assigns a distance of 710 pc and a total luminosity of 6050 L to the central object, with uncertainties of 10% and 20%, respectively. Given a visual brightness of V = 8.83 (Cohen, 1975), we then estimate a visual attenuation of A V = 4.2 mag. Here, we deliberately distinguish between attenuation and extinction. By extinction we understand the partial obscuration of direct star light due to absorption and scattering by dust along the direct line-of-sight to the star, while attenuation measures the reduction of observable flux due to absorption and scattering in a complex geometry, where the residual observable flux may indeed only be scattered light re-directed into the line of sight while the direct light is totally absorbed, as is the case in the RR. The attenuation of the light of HD is characterized by a ratio R V = A V /E(B V ) = 11.9, much larger than the average value of R V 3.1 encountered with interstellar extinction. The attenuation in the UV/visible portion of the spectrum of HD is therefore much grayer than typical interstellar extinction, an effect attributed in the past to extinction by larger grains in the RR (Men shchikov et al., 2002). We consider it more likely that the large value of R V arises from the fact that only scattered light from HD can be observed at optical wavelengths. Upon first scattering, scattered light is initially bluer than that

78 63 of the illumination source, but upon transfer through an optically thick medium and multiple scattering, scattered light also reddens but at a slower rate than expected from the optical path length covered, as shown in the example of the colors of reflection nebulae by Witt (1985). The high value of the estimated visual attenuation of A V = 4.2 mag and the fact that scattered light must be re-directed by 90 require that multiple scattering is the dominant process for the transfer of optical radiation from the center of the RR. UV/Optical Attenuation Curve In Figure 3-6 we display the observed (attenuated) UV/optical SED of HD over the wavelength interval from 130 nm to 600 nm. We produced this SED by combining the observed flux distributions in the near- and far-uv obtained from the IUE archive (LWP 22416, SWP 38188, and LWR 04273) with our optical SED obtained from CTIO spectroscopy. No adjustments or corrections were applied; the continuity between the ground-based and space-based observations is almost perfect, indicating that the two sets of absolute calibrations are in excellent accord. Also shown in Figure 3-6 is a model SED for a metal-deficient, giant-star using a Kurucz (1993) stellar atmosphere for [Fe/H] = -3, log g = 1.5 and T eff = 8250 K, which we propose as a suitable representation of the intrinsic SED for the photometrically dominant component of HD The model SED is normalized to the observed SED at V. The effective temperature of this model atmosphere is higher than that proposed by Men shchikov et al. (2002) by 500 K, but our choice was dictated by the requirement that the model and the observed SED should have the same relative

79 64 strength of the Balmer discontinuity such that the ratio of these two SEDs would be continuous across the Balmer discontinuity. Model SEDs with temperatures differing from 8250 K by as little as 250 K exhibit noticeable discontinuities, when divided into the observed SED.

80 65 2.5e 12 Observed SED Model SED at 8250 K 2e e 12 Flux 1e 12 5e e λ (Å) Figure 3-6 Observed UV/optical SED of HD compared to model SED. The observed SED has been obtained by combining the archived flux distributions in the near- and far-uv from IUE and optical flux distribution obtained by our CTIO spectroscopic measurements. The model SED is from a Kurucz (1993) stellar atmosphere for [Fe/H]=-3, log g =1.5, and T eff =8250 K, normalized to the observed SED at the V-band. The stellar fluxes are displayed in units of ergs cm 1 s 1 Å 1.

81 66 Figure 3-7 displays the wavelength dependence of the UV/optical attenuation suffered by the light of HD The attenuation curve is normalized to E(B V ) = 1, and it is shown in comparison with the average Galactic interstellar extinction curve with identical normalization. Several distinct differences between the two curves are immediately apparent. With the slopes forced to coincide in the B V range, the observed attenuation curve rises more steeply in the near-uv compared to the interstellar extinction curve. There is no sign of the nm absorption peak; instead we see a very much broader attenuation hump reaching its peak near λ 1 5 µm 1. Finally, near λ µm 1 the attenuation curve exhibits a sharp discontinuous rise, which is much more abrupt than the gentle rise shown by the extinction curve.

82 67 12 HD Attenuation Avg. IS extinction pyrene 10 8 E(λ - V)/E(B - V) 6 4 λ β λ β 2 λ α λ p λ -1 (µm) -1 Figure 3-7 UV/optical attenuation curve for HD The average interstellar extinction curve is shown for comparison with the solid line. Also shown is the expected contribution of PAH ionization to the far-uv part of the attenuation curve by pyrene as a representative example with a dashed line. Indicated schematically are the four characteristic PAH absorption bands, the α, para, β and the β. The curves are normalized to E(B V )=1.

83 68 As is well known, the perception of anomalous attenuation/extinction curves is strongly dependent upon their normalization (Fitzpatrick, 2004). In Figure 3-8, therefore, we display the attenuation curve for HD normalized at V, scaled to A V = 1.0, and compare it to Galactic extinction curves with R V = 3.1, representative of the diffuse ISM, and R V = 5.5, representative of a molecular cloud environment and a grain size distribution dominated by larger grains. While the overall level of UV attenuation in HD relative to the attenuation at longer wavelengths is roughly comparable to the R V = 5.5 extinction curve and consistent with earlier findings of large dust grains in the RR (Men shchikov et al., 1998; Jura et al., 1997), the two features that make the HD attenuation truly unique remain clearly visible: the absence of the nm absorption peak and the sharp discontinuous rise in attenuation at λ µm 1.

84 HD Attenuation Avg. IS extinction, R V = 3.1 CCM extinction curve, R V = A λ /A V λ -1 (µm) -1 Figure 3-8 The UV/optical attenuation curve for HD 44179, normalized to A V = 1 mag. For comparison, two Galactic extinction curves for R V = 3.1 and R V = 5.5 are shown as well.

85 70 In two subsequent sections we will interpret the features just described in more detail. Given the overall appearance of the attenuation curve for HD 44179, we can conclude, however, that the suggestion by Men shchikov et al. (2002) that most, if not all, of the reddening of HD is caused by interstellar instead of nebular material intrinsic to the RR is almost certainly incorrect. The hydrogen column density (N(H I) cm 2 ) implied by the equivalent width of the 578 nm diffuse interstellar band in the spectrum of HD (Hobbs et al., 2004) translates to an extinction of A V < 0.05 mag. Also noteworthy is the absence of narrow, interstellar absorption lines Na i and Ca ii in the spectrum of HD (Hobbs et al., 2004). The probability that the line-of-sight to one of the most unusual objects in the Galaxy should also be characterized by one of the most unusual interstellar extinction curves ever observed is vanishingly small. Given the large internal optical depth of the RR implied by the visual attenuation found in 3.3.2, we suggest that the unique characteristics of the attenuation curve of HD must be explained in terms of the optical characteristics of the molecular and dust constituents of the RR. Far-UV Rise The abrupt and discontinuous rise in the observed attenuation curve beginning at λ 1 6 µm 1, corresponding to a wavelength of 167 nm or a photon energy of 7.4 ev, is not characteristic of dust extinction but rather suggests the onset of strong absorption, e.g. the absorption associated with the ionization discontinuity in small PAH molecules, which occurs at just this energy. Verstraete et al. (1990) published detailed ionization cross-sections for two PAH molecules, pyrene and coronene, which

86 71 exhibit a sharp onset at λ µm 1, similar to the observed rise in the HD attenuation curve. When normalized to cross-section per C-atom, the absolute values of the ionization cross-sections and their wavelength dependencies are almost indistinguishable. A typical value is cm 2 C-atom 1 at 143 nm or λ µm 1. The work by Jochims et al. (1996) shows similar ionization cross-section spectra for naphthalene, azulene, anthracene, phenanthrene, and benz(a)anthracene, all with an sharp onset near 7.5 ev, corresponding to λ µm 1. However, the ionization potential (IP) of PAH molecules is generally size-dependent in a manner similar to the size-dependence of the peak wavelength of fluorescence (Vijh et al., 2004). We demonstrate this dependence of the IP on molecular size by plotting known IP values (Eilfeld & Schmidt, 1981) against their molecular mass in Figure 3-9. This graph illustrates that the rise in the far-uv attenuation curve is consistent with the ionization of PAHs, provided their molecular mass is not greater than about 250 amu. A similar upper limit of the masses and sizes of PAH molecules was also deduced by Vijh et al. (2004) on the basis of the observed blue fluorescence seen in the RR. The attenuation curve suggests that absorption by molecules of this size and smaller could contribute to the sudden far-uv rise but that molecules of larger size cannot be present with significant abundance close to the central star; if they were, the discontinuity would occur at longer wavelengths.

87 72 9 Acenes & Benzologs Perylene, Coronene & Bianthene Series Terrylene & Peropyrene Series Pyrene Series Calulated Values 8 Ionisation Potential (ev) 7 Red Rectangle 6 5 I.P. Graphite Molecular Wt. (amu) Figure 3-9 Ionization potential of different families of PAHs as a function of molecular mass. Also shown for reference are the IP of graphite (Clar & Schmidt, 1975; Burns & Yelke, 1969) and the ionization discontinuity apparent in the attenuation curve of the central source of the RR.

88 73 An important question is whether the observed increase in attenuation in the RR over the wavelength interval from λ µm 1 to λ µm 1 is in agreement with the likely carbon column density and the measured ionization crosssection at λ µm 1. The observed increase in attenuation over this interval is 2.0 mag, when reduced to the actual color excess of E(B V ) = 0.35 of HD 44179, corresponding to an optical depth increase by τ = With an ionization crosssection σ = cm 2 C atom 1 at 143 nm wavelength, the corresponding carbon column density is N C = cm 2 for carbon atoms tied up in small PAH molecules with masses less than 250 amu. If the visual attenuation of 4.2 mag is associated with a mixture of gas and dust with relative elemental abundances as found in the Sun, then N(H i + 2H 2 )/A V = cm 2 mag 1 (Bohlin et al., 1978), and N(H i + 2H 2 ) = cm 2. With a solar carbon abundance ratio of N C /N H = , this leads to a total expected column density of carbon in all forms of N C = cm 2. Hence, in order to produce the observed far-uv rise by ionization of small PAHs, we require that about 50% of all carbon atoms reside in small PAHs, if normal solar abundances prevail. Thus, the strength as well as the wavelength dependence of the observed far-uv rise in the attenuation curve are fully consistent with an origin in the ionization of small PAH molecules. In addition, HD is recognized as a carbon-rich post-agb star whose relative carbon abundance must exceed that of the Sun by factors of two to six (Cherchneff & Barker, 1992), making the case even stronger. In Figure 3-7 we have indicated the expected contribution of PAH ionization to the far-uv part of the attenuation curve for pyrene as an representative example. The fact that the observed rise is slightly steeper than the

89 74 individual curve of a specific PAH molecule can be understood easily, if still smaller PAH molecules with higher IP values contribute to the absorption at the shorter wavelengths. Mid-UV Hump Given the facts that blue fluorescence from PAH molecules with an intensity comparable to that of the ERE contributes strongly to the the nebular emission of the RR and that the ionization of these molecules in the inner part of the RR contributes strongly to the far-uv attenuation at λ 1 6 µm 1, it appears most likely to us that the mid-uv hump in the attenuation curve is produced by absorption related to electronic bound-bound transitions in these same molecules. With a sharp onset of the fluorescence near 360 nm, the absorption powering this emission must occur at shorter wavelengths. According to Clar (1972), the absorption spectra of PAH molecules consist of four characteristic bands, the α, para, β, and β bands, located at progressively shorter wavelengths in the mid-uv. They are schematically indicated in the mid-uv portion of Figure 3-7, with arrows starting at the band edges and the height indicating the relative strengths of the bands (See also Schmidt, 1977). The β-band which typically occurs at a wavelength 1/1.35 of that of the α band, is typically two orders of magnitude stronger than the α-band, while the intermediate para-band is typically one order of magnitude stronger than the α-band. The wavelength positions of these bands are closely related to the IPs of the corresponding molecules (Clar & Schmidt, 1975, 1977, 1979), so that a continuum of sizes of small PAH molecules with IP > 7.4 ev can be expected to produce a broad, quasi-continuous absorption band as ob-

90 75 served in the attenuation curve of HD Verstraete & Leger (1992) produced a generic mid- and far-uv absorption spectrum for interstellar PAH molecules, which exhibits the mid-uv hump and the far-uv rise so clearly apparent in the attenuation curve of HD Discussion Identification of the BL Carrier When we first attributed the BL in the RR to fluorescence by three- and fourringed neutral PAH molecules (Vijh et al., 2004), the identification of the likely carrier was tentative for several reasons. Our observational technique for separating the BL spectrum from the underlying scattered light spectrum produces measurements of the BL intensities only at the positions of the hydrogen Balmer lines, yielding data with very limited and uneven spectral resolution that is incapable of revealing fine spectral features needed for specific identifications of individual molecular species. In addition, the fluorescence spectra of several likely molecular PAH species occupy nearly the same wavelength interval, contributing to the observed spectrum in proportion to their relative abundances and their fluorescence quantum yields. Finally, the gasphase fluorescence spectra of PAHs vary as a function of temperature, with subbands losing their sharpness with increasing temperature (Chi et al., 2001a). For these reasons, the identification of the carrier of the BL can only be made in part on the basis of spectral comparison and must rely largely upon other supporting evidence

91 76 such as presented in this paper. There are prospects for improving the spectral resolution of the BL spectrum by using other techniques for separating the BL spectrum from the scattered light spectrum. One such technique is spectropolarimetry. It relies on the fact that the scattered nebular light is highly linearly polarized (Perkins et al., 1981) while the BL is expected to be unpolarized. The presence of the BL in the composite spectrum of the RR would then be revealed by a proportional reduction in the linear polarization of the combined light. The only two published sets of spectropolarimetric observations at optical wavelengths of the RR nebula (Schmidt et al., 1980; Reese & Sitko, 1996) do indeed reveal a sudden reduction of the linear polarization shortward of 410 nm consistent with the onset of the BL band. Unfortunately, the existing data lack the spectral resolution and adequate spectral coverage at shorter wavelengths that would be necessary. A program of spectropolarimetric observations of the RR is being planned. At this stage, the strongest supporting evidence for the correctness of our identification of the carrier of the BL are the close spatial correlation with the 3.3 µm C H stretch emission that is highly specific to small, neutral PAH molecules, and the evidence obtained from the analysis of the attenuation curve of the central source that exhibits strong features identifiable with electronic-band absorptions and an ionization discontinuity specifically pointing toward PAH molecules with masses of 250 amu or less.

92 Spectral Variability of the BL Our initial results showing evidence of systematic spectral variability of the BL with position within the bi-polar structure of the RR are intriguing but clearly need more observations with the highest possible spatial and spectral resolution. Such observations would have the potential of probing the variation in the dominant sizes of the PAH molecules as a function of distance from the central source and as a function of position within and outside of the bi-polar outflow cones and within the shadow of the circum-source disk. Our initial results are consistent with an increase in the size of the most strongly fluorescing PAH molecules with increasing distance from the central source, with a superimposed azimuthal dependence with respect to the outflow axis of the RR. This could reflect the growth of the molecules with increasing distance, but it could also be a result of the different excitation requirements for the fluorescence of PAH molecules of different sizes. The wavelengths of the absorption bands of PAH molecules are related to their respective ionization potentials (Clar & Schmidt, 1975, 1977, 1979) and occur at shorter wavelengths for smaller PAHs, and at longer wavelengths for larger PAHs. As a consequence, exciting radiation for the smallest PAHs may be lacking at increased distances as a result of the high opacity of the intervening nebular material Spatial Variation of the BL The close spatial correlation between the BL and the 3.3 µm emission along two nearly identically placed slits, positioned 5 south of HD is remarkable in it-

93 78 self by supporting the interpretation that both emissions are produced by the same molecules. It is further remarkable in that both emissions show a distinctly different distribution than the distribution of the scattered light in the RR. The latter is produced by dust with a strongly forward-directed scattering phase function, which when coupled with an embedded source, produces a strongly peaked brightness distribution of the resulting scattered light. This is clearly apparent when viewing the blue image of the RR (Fig. 3-1), which is dominated by a bright circular source of about 8 diameter containing most of the scattered light. The PAH fluorescence, both electronic in the form of the BL as well as vibrational in the form of the 3.3 µm emission, by contrast is much less centrally peaked, consistent with an isotropic emission process for both. It is further remarkable that the BL brightness distribution does not reflect the crossing of the X-shaped arms or whiskers of the RR, which are so prominent at red wavelengths (Cohen et al., 2004). This indicates that the BL is neither associated with the interior of the outflow cones nor with the walls of the outflow cavity but rather is associated with a population of emitters residing in a shell surrounding the inner bi-polar structure. This would place them into the shadow of the circum-source disk and thus outside the reach of ionizing radiation, consistent with our attribution of the BL to neutral PAHs. The difference between the spatial distributions of the BL and the ERE is striking. It has been known for some time that the ERE is strongly concentrated in the X-shaped arms of the RR (Schmidt & Witt, 1991), which are thought to be the projections of the walls of the bi-polar outflow cavity upon the plane of the sky. Our data leave little doubt that the two emitters involved require totally different excitation

94 79 conditions. Given the ubiquity of the ERE in a wide range of interstellar environments (Witt & Vijh, 2004), it is not likely that the concentration of the ERE in the walls of the RR outflow cavity is the result of an abundance enhancement but rather reflects the specific excitation requirement of the ERE process. As reviewed by Witt & Vijh (2004), there are several independent indicators suggesting that photon with energies E > 7.25 ev are required to initiate the ERE. These energies are much higher than those required for the excitation of fluorescence in neutral PAH molecules and the number of such high-energy photons are much more limited in the SED of HD Thus, they can barely penetrate the walls of the outflow cavity before their number is exhausted, thus limiting the spatial extent of the ERE-producing emitters. The topic of the ERE excitation will be the subject of another publication by the present authors Attenuation of HD The attenuation curve of HD appears highly unusual when compared to the wavelength dependence of interstellar extinction, but it appears to be fully consistent both qualitatively and quantitatively with expectations, if the scattered light of HD has to penetrate a large column density (A V = 4.2 mag) of circumstellar material filled with neutral PAH molecules with atomic masses less than 250 a.m.u. With a column density of carbon in the form of small PAH molecules estimated from the strength of the far-uv absorption related to PAH ionization, we can estimate a molecular abundance of such PAH molecules with typically 16 carbon atoms

95 80 per molecule relative to hydrogen in the RR of about This agrees well with the canonical abundances of PAH molecules derived from the analysis of the strength of the mid-ir aromatic emission features, commonly referred to as the unidentified infrared emission bands (Leger & d Hendecourt, 1987). This makes PAH molecules the most abundant species of interstellar molecules after H 2 and CO. The fact that despite the strong indications for a presence of small PAH molecules there is no clear counterpart for the interstellar nm absorption feature in the attenuation curve of HD 44179, suggests that proposals to associate the nm feature with PAH absorption in current dust models should be viewed with caution. The current observations appear to prohibit at least a role for small PAHs with masses < 250 amu. This still leaves the possibility that larger PAH structures could be involved in contributing to the interstellar nm feature, because they would have the advantage of greater thermodynamic stability under interstellar conditions. The small PAH molecules apparent in the RR are able to form and survive in the relatively benign radiation environment of the RR, especially in the shadow of the optically thick dust ring blocking direct radiation from the central source. It is in these parts of the RR nebula where the BL intensity is strongly enhanced in comparison to the scattered radiation and the ERE. These small PAH molecules, however, are not likely to survive the harsher radiation conditions prevailing in the diffuse interstellar medium (Jochims et al., 1994, 1999), where they would be subject to photo-dissociation. As a consequence, it does not appear likely that BL with spectral characteristics and intensities similar to those observed in the RR will be seen in dusty star-forming regions such as reflection nebulae and H ii regions (Rush & Witt, 1975).

96 81 The conditions deduced for the RR environment also places interesting constraints on the carriers of the diffuse interstellar bands (DIBs). Current ideas for the carriers of these much-studied but still unidentified interstellar absorption features concentrate on carbonaceous molecules, their ions in particular (see Snow, 1995, for a review). PAH ions have been considered as the most likely carriers by a number of investigators. Given the strong indicators for a large column density of PAH molecules and their ions in the RR, it is remarkable that only one DIB λ5780 is detected in the spectrum of HD (Hobbs et al., 2004). T. P. Snow (2004; private communication) has proposed that this fact might be explained, if DIBs are produced by PAH cations with masses well in excess of 250 a.m.u. Another fact that must be taken into consideration is the absence of absorptions resulting from the D 2 D 0 transition of anthracene + (708.76±0.13 nm), naphthalene + (670.73±0.06 nm) (Sukhorukov, 2004), and acenaphthene + (646.3 nm)(biennier et al., 2003) which might be expected in the spectrum of HD 44179, if these cations are present in the light path. Some of these discrepancies could be resolved by considering the ionization of these PAHs to the dication stage. 3.5 Conclusions We summarize our conclusions: 1. Our spatially resolved observations of the BL in the RR have yielded the first tentative evidence that the detailed spectrum of the BL is spatially variable, consistent with a change in the dominant size of the fluorescing molecules with

97 82 position relative to the central source. 2. The spatial distribution of the band-integrated BL in the RR differs fundamentally from those of the dust-scattered radiation and of the ERE, consistent with an origin from a nebular volume located mainly outside the outflow cavity and concentrated in the shadow of the disk obscuring the central source. This supports our earlier conclusion that the BL is produced by fluorescence from small, neutral PAH molecules. 3. There exists an excellent spatial correlation between the BL and the distribution of 3.3 µm C H stretch emission from PAHs. The latter emission is predominantly produced by small, neutral PAH molecules, and the close spatial correlation supports the suggestion that the BL has its origin in the similar molecules. 4. By combining our observations with existing IUE data for HD 44179, we have determined the UV/optical attenuation curve for the central source in the RR. The attenuation curve is characterized by an exceptionally high value of R V = A V /E(B V ) = 11.9, a remarkable absence of the nm absorption bump, the latter being a characteristic of Galactic interstellar extinction, and a strong, discontinuous rise in attenuation near λ µm 1. These unusual characteristics suggest strongly that the reddening observed in the light from HD is a result of radiative transfer in the RR and not due to interstellar extinction. 5. The far-uv rise in the attenuation curve of HD is qualitatively and quanti-

98 83 tatively different from the much more gradual far-uv rise in Galactic extinction curves. The rise is, however, consistent with the onset of photo-ionization of small PAH molecules with masses of less than 250 amu, if the abundance of such PAH molecules relative to hydrogen is about This represents a fully independent confirmation of the presence of small PAH molecules in the RR environment, consistent with our conclusion regarding the origin of the BL in the RR. 6. The broad mid-uv hump in the attenuation curve of HD whose appearance is quite unlike that of the familiar nm interstellar absorption band, can be explained satisfactorily as resulting from a superposition of the (typically) four mid-uv absorption bands characteristic of neutral PAH molecules. An observable consequence of this absorption and the resulting electronic excitation is de-excitation via fluorescence, which we indeed observe. 7. We do not expect that BL with comparable relative intensities as seen in the RR will be seen in star-forming regions such as reflection nebulae and H ii regions, where the mid-ir UIR band emissions are otherwise strong. The small PAH molecules required by the BL are not expected to survive in the radiation fields typical of star-forming regions. Note added in proof in the Journal: Recent observations of other reflection nebulae have produced evidence for the presence of BL in some instances, although at lower relative intensities compared to the RR (see next Chapter for details).

99 Chapter 4 Blue Luminescence and the Presence of Small Neutral PAHs in the ISM 1 Blue Luminescence (BL) was first discovered in a proto-planetary nebula, the Red Rectangle (RR) surrounding the post-agb star HD BL has been attributed to fluorescence by small, 3-4 ringed neutral polycyclic aromatic hydrocarbon (PAH) molecules, and was thought to be unique to the RR environment, where such small molecules are actively being produced and shielded from the harsh interstellar radiation by a dense circumstellar disk. In this paper we present the BL spectrum detected in several ordinary reflection nebulae illuminated by stars having temperatures between 10,000 23,000 K. All these nebulae are known to also exhibit the infrared 1 To appear in The Astrophysical Journal November 2005, Vol 633, Authors: U. P. Vijh, A. N. Witt, & K. D. Gordon 84

100 85 emission features called aromatic emission features (AEFs) attributed to large PAHs. We present the spatial distribution of the BL in these nebulae. In the case of Ced 112, the BL is spatially correlated with mid-ir emission structures attributed to AEFs. These observations provide evidence for grain processing and possibly for in-situ formation of small grains and large molecules from larger aggregates. Most importantly, the detection of BL in these ordinary reflection nebulae suggests that the BL carrier is an ubiquitous component of the ISM and is not restricted to the particular environment of the RR. 4.1 Introduction We discovered the presence of small, 3-4 ringed, neutral polycyclic aromatic hydrocarbons (PAHs) in the Red Rectangle (RR) through their electronic fluorescence, called Blue Luminescence (BL) (λ peak 3750 Å) (Vijh et al., 2004, 2005b). The observed spectrum was attributed to small PAHs on the basis of a comparison with laboratory spectra of PAH molecules fluorescing in the gas phase (Vijh et al., 2004), the spatial correlation between the BL and the 3.3 µm emission, commonly attributed to small, neutral PAH molecules, and the newly-derived UV/optical attenuation curve for the central source of the RR, HD (Vijh et al., 2005b). The latter two results provided strong independent evidence for the presence of small PAH molecules with masses of less than 250 amu in the RR, which supports the attribution of the blue luminescence to fluorescence by the same molecules. Such fluorescence is excited by the absorption of near-ultraviolet photons, followed by the relaxation to the lowest level

101 86 of the first excited electronic state via a series of mid-ir vibrational transitions. The subsequent electronic transitions to various vibrational levels of the electronic ground state are then observed as optical fluorescence, i.e. the BL. The BL is much more specific in terms of size and ionization state of the emitter than vibrational transitions observed in the mid-ir (Reylé & Bréchignac, 2000). In particular, the wavelength of the most energetic electronic transition seen in neutral PAHs is closely dependent on the size of the molecular species. In general, this fluorescence wavelength increases with the molecular weight and size of the molecule (Vijh et al., 2004). Other emission bands that are most commonly attributed to PAHs are the family of features at 3.3, 6.2, 7.7, 8.6, 11.3, & 12.7 µm (referred to as aromatic emission features (AEF) or unidentified infra-red (UIR) bands) (Hony et al., 2001; Cook & Saykally, 1998; Leger & Puget, 1984; Verstraete et al., 2001; Allamandola et al., 1985; Sellgren, 1984). These AEFs are the signatures of aromatic C-C and C-H fundamental vibrational and bending modes and the specific sizes and ionization states of the PAH molecules or ions cannot be ascertained from these spectra alone. On absorption of a far-uv photon, a PAH molecule or ion usually undergoes a transition to an upper electronic state. If the molecule or ion undergoes iso-energetic transitions to highly vibrationally excited levels of the ground state, then the molecule or ion relaxes entirely through a series of IR emissions in the C-C and C-H vibrational and bending modes. These transitions are largely independent of size, structure and ionization state of the molecule and are identified with the AEFs. The AEFs are found in almost all astrophysical environments including the diffuse interstellar medium (ISM), the edges of molecular clouds, reflection nebulae, young stellar objects, HII regions, star form-

102 87 ing regions, some C-rich Wolf-Rayet stars, post-agb stars, planetary nebulae, novae, normal galaxies, starburst galaxies, most ultra-luminous infra-red galaxies and AGNs (see Peeters et al., 2004, and references therein). The presence of the AEF carriers in such a wide range of environments suggests that they must include a wide range of sizes of aromatic structures in several ionization states. Therefore, an observation of BL in astronomical sources in the UV/visible range is of particular value because it offers the possibility of tracing the specific presence of small, most likely neutral PAHs in the ISM. The central star in the RR (where the BL was first discovered), HD44179, is a post-agb star that is in an active dust-producing stage. Small dust grains and PAH molecules in a wide range of sizes are currently being produced in its circumstellar outflow environment. The discovery of BL in such an environment was therefore not a surprise. However, several detailed investigations (Jochims et al., 1999; Le Page et al., 2003, and references therein)(see also Allain et al., 1996a,b) have led to the conclusion that PAH molecules with fewer than 30 carbon atoms, i.e. the ones detected in the RR, do not survive in the interstellar environment, once they are ejected from their place of formation. Photo-dissociation by the interstellar radiation field is expected to limit their life time severely. However, as mentioned earlier, the AEFs are seen in almost all astrophysical environments including those with harsh interstellar radiation fields. Do the small PAHs that produce the BL survive under interstellar conditions or are the AEFs primarily due to larger PAHs? The presence or absence of BL in environments with differing levels of UV radiation can help answer these questions. Presence of BL preferentially traces the small, neutral PAH

103 88 molecules, as the larger molecules, even if they were neutral, have generally lower fluorescence efficiencies, and ionized PAH molecules do not fluoresce at optical wavelengths. Small, partially dehydrogenated PAH radicals or radical ions having a singlet ground state, however, could contribute to the BL as well (W.W. Duley, Private communication). To test the viability of such small PAH structures in the general ISM far from C-rich stellar outflows, we obtained long-slit spectra of several reflection nebulae and star-forming regions where AEFs and other emission features like the extended red emission (ERE) have been detected. The temperatures of the illuminating stars in these regions cover a range of K, and thus offer a chance to study the distribution of small PAH molecules as a function of environment. 4.2 Observations The objects observed for this investigation are listed in Table 7.1. The observations of the nebula Ced 201 were made at the Steward Observatory 2.3-m Bok telescope on 2002 November 5 and those for NGC 2023, NGC 5367 and Ced 112 were made at the Cerro Tololo Inter-American Observatory (CTIO) 1.5-m telescope on 2003 March 26 and 28. At the Bok telescope we used the Boller and Chivens spectrograph with a grating having 300 lines mm 1, blazed at 376 nm in the first order with a L3800 cut-on filter. The detector used was a CCD (ccd20) yielding a spatial scale of 0. 8 pixel 1. The slit used was 4 long and 4. 5 wide. At the CTIO 1.5-m we used the R-C spectrograph with the grating #09 having 300 lines mm 1, blazed at 400 nm in the first order with a CuSO 4 filter. The detector used was the Loral

104 89 IK CCD and the spectral coverage was from 340 nm to nm. The slit used was 7 long and 2. 5 wide and yielded a spatial scale of 1. 3 pixel 1. A coronographic decker assembly was used to minimize scattered light from the star while probing as much of the inner nebula as possible at CTIO. For all observations the nebular exposures were bracketed by exposures of the illuminating stars and the long-slit enabled simultaneous exposures of the nebula and the sky. Table 7.1 gives the details of exposure times and slit positions. Data reductions were carried out with IRAF 2.12 EXPORT, and all spectra were flux-calibrated via observation of standard stars. Table 4.1 Details of objects observed, their illuminating stars and slits used. Object Observatory Illuminating Star Slit Positions (Offset) Exposures (s) Name Spectral Type Temperature (K) Ced 201 RN Bok 2.3 m BD B9.5 V 10, north, 20 south Ced 112 RN CTIO 1.5 m HD B9 V 10,500 0,6 north, 6 south Ced 112 RN CTIO 1.5 m HD B9 V 10, north NGC 5367 RN CTIO 1.5 m Her 4636 Herbig Ae/Be 18,700 5 north NGC 2023 RN, PDR CTIO 1.5 m HD B1.5 V 23, Results Ced 201 Ced 201 is a rather compact object at a distance of 420 pc (Casey, 1991), on the edge of a molecular cloud. It is excited by the B9.5V star BD Witt et al. (1987) note that the radial velocity of this star differs from that of the molecular cloud by 11.7±3.0 km s 1 so that Ced 201 is probably the result of an accidental encounter of the star with the molecular cloud, while for most other reflection nebulae the exciting star was born in situ. An arc-like structure located between the star and

105 90 the denser parts of the cloud to the north and the north-east at about 18 from the star may represent a shock due to the supersonic motion of the star. The mid-ir ISO spectra of Ced 201 give evidence for transformation of very small carbonaceous grains into the carriers of the AEFs, due to the radiation field of the illuminating star and/or to shock waves created by its motion (Cesarsky et al., 2000). Ced 201 exhibits ERE (Witt & Boroson, 1990) and AEFs (Cesarsky et al., 2000). Figure 4-1 DSS.J.POSSII image of Ced 201, overlaid with our slits showing the extracted apertures. The white central pixel indicates the position of the star BD The dash at the bottom left is 10. The right-most extraction window on the northern slit captures the spectrum of the arc-like structure seen on this image. We obtained nebular spectra along two slits (PA 90 ), 15 north and 20 south of the central star. Figure 4-1 shows an optical image of the nebula overlaid with the

106 91 extracted apertures. We extracted 9 spectra along the 15 north slit and 6 spectra along the 20 south slit each covering 68.4 sq. arcsec of the nebula. BL was measured using the line-depth technique (Vijh et al., 2004) at each of the six hydrogen Balmer lines from Hβ to Hη in each of these apertures. The line-depth technique detects the BL intensity at the wavelength positions of strong absorption lines by virtue of the fact that the presence of BL reduces the line depths in the dust-scattered nebular spectrum compared to the depths of the identical lines in the spectrum of the illuminating star. Figure 4-2 shows representative BL spectra in four such apertures and the ratio of BL to the continuum in the same apertures. The use of a cuton filter at 3800 Å restricted the measurement of the last Balmer line measured to 3835 Å. Longward of 3800 Å these spectra are similar to the BL spectra seen in the RR (Vijh et al., 2004, 2005b) and other nebulae discussed in further sections of this paper. The spectra from different offsets do differ, as was also noted by Vijh et al. (2005b) and can probably be attributed to differing radiation environments. Given the poor spectral resolution of BL spectra measured using the line-depth technique, in this paper, we will not characterize the differences in the spectra but will simply refer to them as BL spectra.

107 92 2.1e e 05 20"S, 15.2"E 20"S, 0" 15"N, 7.6"E 15"N, 7.6"W I BL (erg cm 2 s 1 Å 1 sr 1 ) 1.5e e e e e e "S, 15.2"E 20"S, 0" 15"N, 7.6"E 15"N, 7.6"W 0.24 I BL /I sc wavelength (Å) Figure 4-2 BL spectra (upper panel) and ratio of BL to the continuum (lower panel) at 4 different positions in Ced 201.

108 Ced 112 Ced 112 is a prominent reflection nebula in the cloud Chamelion I at a distance of 160 pc (Whittet et al., 1997). It is illuminated by the pre-main sequence Herbig AeBe star HD 97300, spectral type B9 (Rydgren, 1980) with a luminosity of 35 L (van den Ancker et al., 1997). Its pre-main sequence evolutionary status as a Herbig AeBe star is based on indirect clues, namely the association with a reflection nebulosity, the presence of an infrared excess at λ > 5 µm, and its location on the ZAMS in the HR diagram (Whittet et al., 1997). There is no evidence of significant infrared excess at wavelengths shorter than 5 µm nor of Hα emission. HD is very likely a relatively old object among HAeBe stars, representative of the latest stages of the pre-main sequence evolution. There is general agreement that the star is an embedded member of the young stellar population and is thus situated at a distance coincident with that of the cloud itself. Infrared emission from the nebula Ced 112 is dominated by AEFs centered at 6.2, 7.7, 8.6, 11.3, and 12.7 µm with very small contributions from continuum emission at longer wavelengths (Siebenmorgen et al., 1998). Siebenmorgen et al. (1998) also report the detection of an extended, ring-like structure around HD 97300, whose emission is dominated by the AEFs.

109 94 Figure 4-3 AAO.R.DSS2.038 image of Ced 112, overlaid with our slits showing the extracted apertures. The central, white pixel indicates the position of the star HD The dash at the bottom left is 10. We obtained nebular spectra along four slits (PA 90 ), 0 offset, 6 north, 6 south and 10 north of the central star. Figure 4-3 is an optical image of the nebula overlaid with the extracted apertures. We extracted 10 spectra along the 6 north and 6 south slits and 6 spectra along the slit through the center (0 offset) each covering sq. arcsec of the nebula. Along the 10 north slit we extracted 4 spectra each covering 32.5 sq. arcsec. BL was measured using the line-depth technique in each of these spectra. Along the 6 south slit we extracted 14 spectra with finer apertures

110 95 each covering sq. arcsec to enable comparison with ISOCAM data discussed in the next section. Figure 4-4 show representative BL spectra in three such apertures and the ratio of BL to scattered light in the same apertures.

111 96 2.4e e 05 6"S, "E 6"S, 8.125"W 6"S, "W I BL (erg cm 2 s 1 Å 1 sr 1 ) 1.8e e e e e e e "S, "E 6"S, 8.125"W 6"S, "W 0.80 I BL /I sc wavelength (Å) Figure 4-4 BL spectra (upper panel) and ratio of BL to the continuum (lower panel) at 4 different positions in Ced 112. These spectra correspond to extractions along the slit shown in Figure 4-5.

112 97 Spatial Correlation of BL with mid-ir emission As mentioned earlier, Siebenmorgen et al. (1998) reported the detection of an extended, ring-like structure around HD in the mid-ir images of the nebula taken with ISOCAM. They obtained images in four narrow-band filters centered at 6.0 µm (lw4), 6.8 µm (lw5), 11.3 µm (lw8) and 14.9 µm (lw9). All four images show an extended emission centered on the star and an elliptical ring of size around it. They also report that the spectra taken with the circular variable filter between 5.8 and 13.8 µm roughly aligned along a line intersecting the star, the emission minimum, and the ring seen in the SE direction (PA = ) show strong AEF features. Our 6 south offset slit extends over the nebula and the ring. Figure 4-5 shows the lw9 image overlaid with our slit, with the extraction apertures marked and Figure 4-6 shows a plot of the distribution of the band-integrated BL and dust-scattered light in the apertures along the slit. The integrated BL clearly shows enhancements in the regions where the slit crosses the ring on both sides, while the scattered light does not. It is also interesting to note that in the spectra shown by Siebenmorgen et al. (1998) the flux at 7.8 µm between two positions on the ring and outside away from the star varies by a factor of 2.4 while the band integrated BL intensity varies by a factor of 2.6 for similar locations. Figure 4-7 shows the distribution of the ratio of the band-integrated BL to the band-integrated scattered light. The enhancement of the BL at the positions of the ring is clearly seen in this representation as well.

113 Figure 4-5 ISOCAM image (lw9) of Ced 112 from Siebenmorgen et al. (1998) overlaid with our slit showing the extraction apertures. The white pixel is overlaid to indicate the position of the central star HD The arrows indicate where the slit crosses the ring-like structure seen in the image. 98

114 Band-BL-6" S Band-Scat-6" S 0.01 Intensities offset (arcsec) Figure 4-6 Band-integrated BL and scattered light intensities in the apertures indicated in Figure 4-5. A central-blocker used in the slit prevented measurements close to zero offset. Offsets are along the slit, with zero offset being exactly 6 south of the star.

115 Band-I BL /Band-I sc offset (arcsec) Figure 4-7 Ratio of Band-I BL to Band-I sc in the apertures indicated in Figure 4-5. Offsets are along the east-west, 6 south slit with zero being exactly south of the star NGC 5367 NGC 5367 surrounds the double star Her 4636 in the head of the prominent Cometary Globule CG 12 and lies at high galactic latitude (b = +21 ). It is an example of a relatively isolated low-to-intermediate-mass star formation region (van Hill et al., 1975), which is thought to have formed as a result of a nearby supernova event million years ago (Williams et al., 1977), and has converted about 20% of its gas mass into stars (White, 1993). The most probable distance to NGC 5367 is 630 pc, and the mass of the surrounding molecular cloud is 120 M (Williams et al., 1977). Williams et al. (1977), using optical and infrared photometry, showed the likeliest

116 101 masses of the two stars in the binary system to be 4.5 and 8 M, with spectral type B7 and B4, respectively, with the latter being surrounded by a dust-shell, emitting as a 1600 K blackbody. Reipurth & Zinnecker (1993) also obtained spectra of the two stars (called N and S for north and south) as part of their visual pre-main sequence binary program. They show that the northern component has strong Hα emission and the southern component, the B7 star, has no such emission, and they conclude that both are probably Herbig Ae/Be stars. Vieira et al. (2003) also include Her 4636 in their study of Herbig Ae/Be stars and estimate a rotational velocity of 158 km s 1 and an effective temperature of 18,700 K.

117 102 Figure 4-8 AAO.R.DSS2.325 image of NGC 5367, overlaid with our slits showing the extracted apertures. The central, white pixel indicates the position of the star Her The dash at the bottom left is 20. We obtained nebular spectra 5 north of the illuminating stars. Figure 4-8 shows an optical image of the nebula overlaid with the extracted apertures. We extracted 8 spectra each covering sq. arcsec of the nebula along the slit. BL was measured using the line-depth technique in each of these apertures. A combined spectrum of both the stars was used for the stellar measurements. Figure 4-9 show representative BL spectra in three such apertures and the ratio of BL to scattered light in the same apertures. The red rise in the spectra is a result of emission from hydrogen recom-

118 103 bination filling in the hydrogen Balmer lines, in addition to the BL. Characteristic hydrogen recombination spectra alone would result in a spectral intensity steadily decreasing with decreasing wavelength. Not only do these spectra show a peak at λ 370 nm they also show a secondary peak at λ 400 nm, which is a feature seen in many BL spectra.

119 e e 06 5"N, 24.3"E 5"N, 28.6"W 5"N, 41.6"W I BL (erg cm 2 s 1 Å 1 sr 1 ) 3.0e e e e "N, 24.3"E 5"N, 28.6"W 5"N, 41.6"W 0.80 I BL /I sc wavelength (Å) Figure 4-9 BL spectra (upper panel) and ratio of BL to the continuum (lower panel) at three different positions in NGC These spectra are contaminated with emission from hydrogen recombination, but the peak 400 nm, 377 nm and the rise towards the Balmer discontinuity at λ= 357 nm are indicative of BL.

120 NGC 2023 NGC 2023 is a bright reflection nebula which is a part of the Orion L1630 molecular cloud at a distance of 450 to 500 pc (de Boer, 1983). This nebula, also known as vdb 52, is illuminated by the bright B1.5 V star HD (Racine, 1968). NGC 2023 is an example of photo-dissociation regions (PDRs) that exist on the surfaces of molecular clouds in the vicinity of hot, young stars. The interaction of the newborn stars with their parental molecular environment leads to intense emission of fine-structure lines of carbon and oxygen and of H 2 ro-vibrational transitions. NGC 2023 has been well studied at many wavelengths, including the mid- and near IR region showing AEFs attributed to PAHs (Sellgren et al., 1985; Verstraete et al., 2001; Moutou et al., 1999). NGC 2023 also exhibits ERE (Gorodetskii & Rozhkovskii, 1978; Witt et al., 1984b; Witt & Schild, 1988).

121 106 Figure 4-10 SERC.ER.DSS2.768 image of NGC 2023, overlaid with our slits showing the extracted apertures. The central, white pixel indicates the position of the central star HD The dash at the bottom left is 20. We obtained nebular spectra with a slit (PA 90 ) at 0 offset with a central decker blocking out the star. Figure 4-10 shows an optical image of the nebula overlaid with the extracted apertures. We also obtained spectra of the star HD We extracted 5 spectra each covering 32.5 sq. arcsec of the nebula along the slit. BL was measured using the line-depth technique in each of these apertures. Figure 4-11 show representative BL spectra in three such apertures and the ratio of BL to scattered light in the same apertures. As in NGC 5367, the red rise in the spectra is a result

122 107 of hydrogen recombination filling in the hydrogen Balmer lines in addition to the BL. In this case the origin of this recombination emission is the rather extended region of photo-ionized hydrogen gas overlying the Orion constellation. There is no measurement of BL at Hβ because the nebular Hβ (λ 4861 Å) emission more than fills the scattered stellar Hβ absorption line. Characteristic recombination spectra alone would result in a spectrum steadily decreasing with decreasing wavelength. These spectra show a upturn at λ 400 nm and a pronounced peak at λ 380 nm, which we interpret as evidence for the presence of BL.

123 e e 05 0", 26"E 0", 13"E 0", 13"W I BL (erg cm 2 s 1 Å 1 sr 1 ) 1.2e e e e ", 26"E 0", 13"E 0", 13"W 0.80 I BL /I sc wavelength (Å) Figure 4-11 BL spectra (upper panel) and ratio of BL to the continuum (lower panel) at three different positions in NGC These spectra are heavily contaminated with emission from hydrogen recombination, but the peak 380 nm and the rise towards the Balmer discontinuity at λ = 357 nm are indicative of BL.

124 Discussion We have presented optical long-slit spectra of reflection nebulae illuminated by stars having a range of temperatures between 10,000 23,000 K, and thus covering a wide range of radiation environments. The long-slit technique enables us to extract many individual spectra in each nebula that differ in local radiation density and probably size-distribution of emission carriers but do not differ considerably in chemical composition or grain-processing histories. Comparison of BL spectra from different nebulae will help us study the evolution of the carriers with changing environments and also see what effects different grain processing have on the BL emitters The BL Spectrum The BL spectra in different apertures along the slit in the same nebula varied in intensity with offset from the respective stars but were fairly similar in spectral shape, and therefore we produced one average spectrum for each of the slits. This allows us to identify differences between the BL spectrum in different nebulae and find possible correlations to varying environments and local ISM histories. Figure 4-12 shows the average BL spectrum from the two slits in Ced 201, four slits in Ced 112 and two BL spectra from the RR (reported earlier in Vijh et al. (2004)). In this section we only discuss Ced 201 and Ced 112. The other two nebulae NGC 2023 and NGC 5367 will be discussed in a later section ( 4.4.3). The RR BL spectrum from the slit 2. 5 south offset has been scaled down by a factor of 5 to aid comparison. Four differences in these spectra may be noted. One, the BL intensity in the RR-2. 5

125 110 south is about 10 times higher than in Ced 112 but the RR-5 south is similar in intensity. Two, the RR spectra have peaks at λ 377 nm similar to Ced 112. Three, the RR spectra have a long-wavelength component that is absent in Ced 112, but present in Ced 201. And finally, the Ced 201 BL spectrum is unlike that of Ced 112, but could be similar to that of the RR, given the short wavelength rise, but lack of data short-ward of 380 nm makes it difficult to classify the spectrum. All the spectra also show a secondary peak at λ 400 nm, which could be an indication of a population of efficient fluorescing molecules of slightly larger sizes. The nebular spectra in Ced 201 show some interesting band-like emission features, which are not present in the stellar spectrum and therefore must be nebular in origin. Figure 4-13 shows the nebular spectrum at 15 north, 38 west divided by the stellar spectrum. This resulting spectrum is free of any stellar features and the peaks at the positions of the Balmer lines indicate filling in of the nebular lines by the BL. The two broad bands at λ 4050 Å and λ 4500 Å are new features and are likely to be signatures of slightly larger PAHs (> 14 C atoms) than the 3-4 ringed ones that produce the blue peak at λ 377 nm.

126 111 I BL (erg cm 2 s 1 Å 1 sr 1 ) 1.4e e 05 1e 05 8e 06 6e 06 4e 06 RR 2.5"S offset ( 0.2) RR 5"S offset Ced 112 0" offset Ced 112 6"N offset Ced 112 6"S offset Ced "N offset Ced "N offset Ced "S offset 2e wavelength (Å) Figure 4-12 Average BL spectra from the four slits in Ced 112, and two slits in Ced 201 compared to two spectra from the RR.

127 Ced "N, 38"W F neb /F * wavelength (Å) Figure 4-13 Ratio of the nebular spectrum to the stellar spectrum in Ced 201 from a position 15 north and 38 west of the central star. The two broad bands at λ 4050 Åand λ 4500 Å are new features and are likely to be signatures of slightly larger PAHs (> 14 C atoms). Kemper et al. (1999) model the far-ir and sub-millimeter observations of Ced 201 and conclude that heating by the ionization of PAHs (carbonaceous particles with linear size > 15 Å) is required to explain the observed properties of Ced 201. Creation of PAHs that produce the observed mid-ir features by processing of small carbonaceous grains is also invoked in this nebula (Cesarsky et al., 2000). Cesarsky et al. (2000) suggest that some of carbonaceous material that emits the mid-ir continuum and broad bands far from the star is processed through the effect of the star that moves through the molecular cloud, producing AEF carriers: the continuum is only 2 times

128 113 fainter 12 from the star than close to the star, demonstrating the partial disappearance of its carriers near the star while the AEFs become very strong close to the star. They also infer that these carbonaceous grains must be very small (radius of the order of 1 nm) since they are heated transiently by visible photons to the temperatures of 250 K necessary to emit the observed mid-ir features. It is also interesting to note that Witt et al. (1987) find from UV/visible scattering studies of Ced 201 that the grains responsible for the visible scattering have a narrow size distribution skewed towards big, wavelength - sized grains. This is mostly seen within about 20 from the star. On the other hand the exceptionally strong 2175 Å extinction band observed in the spectrum of BD , the illuminating star, suggests the presence of many smaller carbonaceous grains. All this is evidence for grain processing by the radiation of the star, or/and by shattering of grains by a shock wave associated with its motion. The banded spectrum (Figure 4-13) is of the arc-like structure seen toward the NW in the DSS image of the nebula shown in Figure 4-1. This arc is probably a result of a shock and we suggest that the BL carriers are produced in situ from the larger AEF carriers being photo-dissociated in the stellar UV-field Spatial Distribution of BL We observed little variation in spectral shape in the BL spectra extracted from different apertures along a given slit, in all the nebulae, but considerable variation in the intensity of both the BL and the scattered light. We plot band-integrated BL and scattered light intensities as a function of offset from the central illuminating star

129 114 for each slit position. We integrate from Hβ to the Balmer discontinuity (486 nm 357 nm) for both the BL and the scattered light in case of Ced 112 and till nm at the lower wavelength limit for Ced 201 because of the wavelength filter that was used for those observations. Figure 4-14 shows band-integrated BL and scattered light intensities as a function of offset from the central star along the six slits in the two nebulae. The distribution of BL in Ced 112 seems to be flatter than that in Ced 201, but the overall BL-intensities in Ced 201 are about five times higher than those in Ced 112. Compared to the maximum intensity of the BL in the RR (2. 5 south slit) the intensities in Ced 112 and Ced 201 are 20 times and 4 times lower respectively.

130 Ced201 15"N BL 15"N sc Ced201 20"S BL 20"S sc Intensities (erg cm 2 s 1 sr 1 ) Ced112 0" BL 0" sc Ced112 6"S BL 6"S sc Ced112 6"N BL 6"N sc Ced112 10"N BL 10"N sc offset (arcsec) Figure 4-14 Band-integrated BL and scattered light intensities along the various slits in Ced 201 (upper panel) and Ced 112 (middle and lower panel).

131 Ced201 15"N Ced201 20"S Band I BL /Band I Sc Ced112 0" 6"S 6"N 10"N offset (arcsec) Figure 4-15 Ratio of the band-integrated BL to band-integrated scattered light in Ced 201 (top panel) and Ced 112 (bottom panel). On large scales the BL distribution is uniform compared to the scattered light in Ced 112 but shows asymmetries in Ced 201.

132 117 Our BL observations of Ced 112 also trace out the ring of of AEF emitters seen in mid-ir observations by Siebenmorgen et al. (1998). However, taking the complete distribution over all four slits into account, the large-scale distribution of both BL and scattered light is fairly uniform. This is also borne out by the distribution of the ratio of the band-integrated BL to the scattered light intensity shown in Figure 4-15 (bottom panel). The ratio of BL to scattered light remains fairly constant over large areas of the nebula. The spatial distribution of the BL in Ced 201 is interesting. The BL seems asymmetrically distributed with respect to the star; this is primarily due to the asymmetry in the nebulosity itself. Figure 4-1 shows that the nebula is much more extended to the east than the west, toward the molecular cloud that the star is ploughing into. BL can be detected up to 80 into the cloud, though the intensity falls off rapidly. On the west the BL and the scattered light both fall off more quickly, but where the 15 north slit crosses the arc-like shock-structure, the nebular spectrum shows the distinct banded features discussed earlier in This complex distribution is true of other tracers as well. The CO(2 1) map of the molecular cloud is distinctly asymmetric with double peaked profiles to the north and single peaked profiles to the south whereas the AEF spectra on the other hand are remarkably symmetric (Cesarsky et al., 2000). The relative intensity of BL to the scattered light is shown in the upper two panels of Figure The forward-directed scattered light falls off much more rapidly than the more isotropic BL, and thus the ratio rises at larger offsets.

133 BL in Hydrogen Ionization Regions NGC 5367 and NGC 2023 are both nebulae where the detected spectrum is heavily contaminated by emission from hydrogen recombination. Again as the individual spectra in each slit do not vary in spectral character we present an average BL spectrum for each of the slits in the two nebulae in Figure The two show similar red spectra with intensities decreasing towards the higher-order Balmer lines characteristic of recombination emission from ionized hydrogen but with peaks slightly short-ward of λ 380 nm. The NGC 2023 spectrum is more severely contaminated than NGC 5367 because of a stronger ionizing radiation field, and the NGC 5367 spectrum also shows a secondary peak at λ 400 nm similar to other BL spectra, in addition to the peak short-ward of 380 nm. The measurement at the Balmer-jump (λ = 357 nm) in the other BL spectra is a measure of difference in absorption, but in these spectra it is a measure of emission from ionized hydrogen. Nevertheless, the peak in the spectrum at λ < 380 nm and the secondary feature at 400 nm are signatures of BL, albeit marginal.

134 119 Intensities (erg cm 2 s 1 Å 1 sr 1 ) 1.2e 05 1e 05 8e 06 6e 06 4e 06 2e e 06 1e 06 NGC " offset 5e 07 NGC "N offset wavelength (Å) Figure 4-16 Average BL spectrum in NGC 2023 (top panel) and NGC 5367 (bottom panel). Due to contamination by emission from hydrogen recombination, these are only marginal detections. Observation of extended emission, especially in the blue region of the spectrum, is challenging because of competing contribution from scattered light, and emission from hydrogen recombination succeeds in swamping it even further. We would advise other observers to avoid regions with ionized hydrogen in future searches of BL Survival vs. In-situ formation of BL Carriers BL was first discovered in the RR, a proto-planetary nebula surrounding the evolved post-agb star HD This is a region of active dust production, and therefore it is not surprising to see BL from small molecules with high fluorescence

135 120 efficiencies like anthracene or pyrene, provided there are environments where such species can remain neutral. The regions in the shadow of the circum-binary disk of the central binary provides just such a sheltered environment, and this is where the BL in the RR is predominantly observed. Therefore, a bottom-up production mechanism where the PAH molecules are built from smaller carbon structures is feasible in the RR. Such small PAH molecules are not expected to survive in the ISM, being photo-dissociated under interstellar radiation conditions (Jochims et al., 1999; Le Page et al., 2003, and references therein). Therefore we did not expect to see BL in other nebulae and predicted its absence in Vijh et al. (2005b). However, this prediction is valid only if the only possible means for the existence of these molecules in the other nebulae was travel through the ISM: time-scales of such travel are much longer than the expected molecular lifetimes in such environments. Therefore, the detection of BL in other, non-carbon-rich nebulae came as a surprise. However, the formation of PAH molecules through a top-down process emerges as a likely alternative formation mechanism of small PAHs in these locations. The presence of hydrogenated amorphous carbon (HAC) solids in interstellar space is well established through the observation of the interstellar absorption feature at 3.4 µm (see Pendleton, 2004, for a recent review). In an important laboratory investigation, Scott & Duley (1996) demonstrated that heating of HAC solids will lead to the evolution of aromatic structures, distinguished by their characteristic 3.29 µm C-H stretch feature. Subsequently, Scott et al. (1997) showed that HAC solids will decompose under intense UV radiation into an extended sequence of aromatic carbon clusters with a wide mass range, with small molecules (< 20 atoms) appearing first

136 121 under the lowest-fluence conditions. These laboratory investigations followed earlier theoretical work by Duley et al. (1992) and Taylor & Williams (1993), suggesting that even smaller hydrocarbons, in particular CH +, could be produced through the erosion of carbonaceous dust in shocks, cloud-intercloud interfaces and PDRs. We envision that a similar process could be responsible for the appearance of small PAH molecules in reflection nebulae. Furthermore, larger PAH molecules that are able to survive conditions in the diffuse ISM are likely to undergo photo-fragmentation, as recently suggested by Rapacioli et al. (2005) to explain the variation in the sizes of of AEF emitters across a PDR in the reflection nebula NGC We conclude, therefore, that the most likely source of the small PAHs in reflection nebulae, detected through the BL, are larger carbonaceous structures such as HAC grains or large, stable PAH clusters. The likely cause is a process of photo-dissociation and photo-fragmentation, driven by the strongly enhanced radiation fields of the nearby early-type illuminating stars. There is other ancillary evidence for grain processing in these environments, in particular in Ced 201 (Cesarsky et al., 2000). New molecules are being formed as the star BD rams into the molecular cloud, exposing previously sheltered environments to UV radiation. In this case, we are witnessing a dynamic process where these molecules are both being created from larger species and destroyed by impinging UV radiation. In answer to the question posed in the introduction: Do small PAHs survive in the ISM? or are the AEFs primarily produced by larger PAHs?: Yes, small PAHs survive as parts of larger entities and are liberated again in environments of enhanced radiation density. They do exist in regions where AEFs are observed and should be taken into account in models that

137 122 try to explain the AEFs. However, since BL is strongly biased toward small PAH molecules, the detection of a BL signal does not preclude the presence of a much wider distribution of sizes of aromatic structures. 4.5 Conclusions We have presented here BL spectra in reflection nebulae illuminated by stars having temperatures between 10,000 23,000 K. Of the nebulae studied, Ced 201 and NGC 2023 have previously been shown to exhibit ERE as well (Witt & Boroson, 1990; Witt et al., 1984b; Witt & Schild, 1988). ERE has also been seen in the other two nebulae NGC 5367 and Ced 112 (U. Vijh et al., in preparation). Many of these nebulae also show the AEFs, attributed to PAHs. Our BL spectra which we have previously identified as fluorescence from small, neutral PAHs provide evidence that these small molecules are present in the ISM. Given that these small PAHs (with less than 20 carbon atoms) are not expected to survive harsh interstellar conditions, we believe that the BL spectra provide evidence for the production of these small molecules from larger structures by the impinging UV radiation from the central stars.

138 Chapter 5 The Excitation of Extended Red Emission: New Constraints on its Carrier from HST Observations of NGC ,2 The carrier of the dust-associated photoluminescence process causing the extended red emission (ERE) in many dusty interstellar environments remains unidentified. 1 Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS These observations are associated with program # Submitted to The Astrophysical Journal, May 2005, Authors: A. N. Witt, K. D. Gordon, U. P. Vijh, P. H. Sell, T. L. Smith, & Rui-Hua Xie 123

139 124 Several competing models are more or less able to match the observed broad, unstructured ERE band. We now constrain the character of the ERE carrier further by determining the wavelengths of the radiation that initiates the ERE. Using the imaging capabilities of the Hubble Space Telescope, we have resolved the width of narrow ERE filaments appearing on the surfaces of externally illuminated molecular clouds in the bright reflection nebula NGC 7023 and compared them with the depth of penetration of radiation of known wavelengths into the same cloud surfaces. We identify photons with wavelengths shortward of 118 nm as the source of ERE initiation, not to be confused with ERE excitation. There are strong indications from the well-studied ERE in the Red Rectangle nebula and in the high- b Galactic cirrus that the photon flux with wavelengths shortward of 118 nm is too small to actually excite the observed ERE, even with 100% quantum efficiency. We conclude, therefore, that ERE excitation results from a two-step process. The first step, involving far-uv photons with E > 10.5 ev, leads to the creation of the ERE carrier, most likely through photo-ionization or photo-dissociation of an existing precursor. The second step, involving more abundant near-uv/optical photons, consists of the optical pumping of the previously created carrier, followed by subsequent de-excitation via photoluminescence. The latter process can occur many times for a single particle, depending upon the lifetime of the ERE carrier in its active state. While none of the previously proposed ERE models can match these new constraints, we note that under interstellar conditions most polycyclic aromatic hydrocarbon (PAH) molecules are ionized to the di-cation stage by photons with E > 10.5 ev and that the electronic energy level structure of PAH di-cations is consistent with fluorescence in the wave-

140 125 length band of the ERE. Therefore, PAH di-cations deserve further study as potential carriers of the ERE. 5.1 Introduction Extended Red Emission (ERE) results from a dust-related optical photoluminescence process in the interstellar medium. It was first observed in the peculiar bi-polar Red Rectangle nebula (Cohen, 1975; Schmidt et al., 1980). Soon after its initial discovery, ERE was shown to be present in many other dusty interstellar environments, albeit at mostly lower intensities, e.g. in reflection nebulae (Witt et al., 1984b; Witt & Boroson, 1990), HII regions (Perrin & Sivan, 1992; Darbon et al., 2000), carbon-rich planetary nebulae (Furton & Witt, 1990, 1992), as well as the diffuse interstellar medium of the Milky Way (Gordon et al., 1998) and other galaxies (Perrin et al., 1995; Pierini et al., 2002). The ERE intensity in many dusty sources is proportional to the local density of the illuminating radiation field. This provides a strong argument in favor of the suggestion that the ERE is photondriven, i.e. the ERE is a photoluminescence process. The spectroscopic signature of the ERE is a broad (60 to 100 nm FWHM), unstructured emission band, typically extending from 540 nm to beyond 900 nm in wavelength. The peak wavelength of the ERE band varies from somewhat longward of 600 nm to beyond 800 nm in response to varying environmental conditions, in particular the density of the illuminating ultraviolet (UV) radiation field (Smith & Witt, 2002). This is a defining characteristic of the ERE, which distinguishes it from other emission features with essentially in-

141 126 variable emission wavelengths. At high Galactic latitudes, the intensity of the ERE is comparable to that of the dust-scattered diffuse galactic light (DGL), which has led to an estimated lower limit of the ERE quantum yield of 10 ± 3% and the conclusion that the ERE carrier must be a major contributor to the absorption part of interstellar extinction at UV/visible wavelengths in the Milky Way (Gordon et al., 1998). Its ubiquitous presence in radiation environments ranging in strength over more than five orders of magnitude testifies to the relative robustness of the ERE carrier. More details about ERE studies done over the past three decades may be found in a recent review by Witt & Vijh (2004). None of the current models for interstellar dust (e.g. Draine, 2004; Zubko et al., 2004) predict ERE, nor do they provide a satisfactory post-facto explanation for the observed characteristics of the ERE. As reviewed by Witt & Vijh (2004), numerous ad-hoc models have been proposed to account for the ERE. Most of them suggest ERE carriers in the form of large molecular structures or nanometer-sized grains, in which electronic excitations by shorter-wavelength photons are followed by efficient electronic radiative transitions across a bandgap of < 2 ev. Well-studied examples of such processes are fluorescence and phosphorescence in organic molecules and the photoluminescence in semiconductor nanoparticles. Such particles may be able to meet the constraints posed by the observed spectral characteristics, including their environment-dependent variations, and by the inferred quantum efficiency (Smith & Witt, 2002). However, all carrier models proposed so far suffer from the following dilemma. The excitation of fluorescence or photoluminescence generally occurs with opti-

142 127 mum efficiency at energies just slightly above the bandgap energy, i.e. photons at wavelengths just shortward of the ERE band should be most suitable for the initiation of the ERE. The difference between the optimum excitation energy and the energy at which the luminescence emerges, the so-called Stokes shift, is typically of the order of 1 ev or less in typical photoluminescent agents. At first glance, this does not appear to be the case for the ERE in view of two previous studies of the excitation requirements of the ERE. The first study by Witt & Schild (1985), based upon multi-band surface brightness photometry of several reflection nebulae, concluded that the ratio of the intra-nebular absorption of the ERE-exciting radiation, A(exc), to the color excess E(B-V) suffered by the intra-nebular radiation reaching the same nebular location had values A(exc)/E(B-V) 7-9. When related to the observed UV/optical extinction curves of the illuminating stars for the three nebulae included in that study, it was concluded that the required ratio of A(exc)/E(B-V) could only be found in the immediate vicinity of the nm absorption peak of the respective extinction curves or at wavelengths shortward of 150 nm, where the far- UV rise in the extinction again provides the required opacity. While the multi-valued nature of the extinction curve prevented a unique estimate of the excitation energy by this method, the results indicated that ultraviolet rather than optical photons are responsible for ERE excitation. The second study related to ERE excitation by Darbon et al. (1999) compiled the records of positive ERE detections as well as the results of negative searches for ERE in a large number of nebulae and related them to the effective temperatures of the exciting stars. While ERE is present in abundance (with few exceptions) among

143 128 nebulae illuminated by stars with T eff > 10, 000 K, not a single nebula with stars of T eff < 7000 K exhibits ERE, although dust is present with significant optical depths in the nebulae associated with the latter group of stars. Stars with T eff = 7000 K have spectral energy distributions (SEDs) with a far-uv cutoff near 170 nm, corresponding to an energy of 7.3 ev, while stars with T eff = 10, 000 K have their corresponding far-uv flux cutoff near 110 nm, corresponding to photon energies of 11.2 ev (Kurucz et al., 1974). The results of Darbon et al. (1999) can thus be interpreted to mean that photons with energies somewhere within the range between 7.3 ev and 11.2 ev represent the minimum photon energy required to initiate the ERE process. The results of Witt & Schild (1985) are fully consistent with this conclusion. However, the Darbon et al. (1999) results eliminate the nm absorption peak of the interstellar extinction curve as a possible source of ERE excitation, because T eff = 7000 K stellar atmospheres still have significant flux at this wavelength, yet they are not found to be able to excite ERE. Further supporting observations come from the more recent work of Darbon et al. (2000), who compared the spatial distributions of the ERE and the emissions in various unidentified infrared emission (UIR) bands, attributed to aromatic hydrocarbon structures, in the HII region Sh 152. While the UIR bands are strongest at the interfaces between the ionized volume of the nebula and beyond, the ERE is strongly correlated with the distribution of the ionized hydrogen located in the central region of the HII region. This provides evidence for the fact that far-uv photons are needed for the initiation of the ERE as well as proof of the robust nature of the ERE carrier, while the aromatic hydrocarbon

144 129 structures can apparently be successfully excited by radiation at longer wavelengths, which is able to penetrate into the nebular environments outside the ionization front. This latter picture is also supported by the recent findings by Smith et al. (2004) that the 3.3 µm emitter does not require UV photons for excitation. The results discussed above suggest a Stokes shift in excess of 6 ev for the ERE process, if in fact the far-uv photons required to initiate the ERE are also the source of the actual ERE excitation. Another, more likely possibility is that the far-uv photons simply create and maintain the ERE emitter, for example by photo-ionization or -dissociation of a precursor, which could then be capable of photoluminescence upon excitation by abundant lower-energy photons. The ERE excitation could thus be a two-step process, with a far-uv photon required to produce the carrier, a particle capable of strong absorption in the optical/near-uv region in which lower-energyphotons closer in energy to the 2 ev band gap would excite the actual ERE. To decide among these possibilities, it is essential to determine the wavelength of the ERE initiation as accurately as possible. This must be followed by a comparison of the number density of photons with wavelengths shortward of this limit in a given system with the number density of ERE photons generated within the system. If the latter is greatly in excess of the former, a two-step excitation process is the most likely explanation. The observations reported in this chapter provide a basis for a determination of the critical wavelength of ERE initiation. We employ the fact that the ERE appears in narrow filamentary structures (e.g. Witt & Malin, 1989) on the surfaces of molecular cloud clumps in reflection nebulae, in particular in the bright reflection nebula NGC We assume that the physical width of a cloud edge as seen in

145 130 the light of the ERE is determined by the depth of penetration of the photons giving rise to this particular nebular emission. We will compare these penetration depths with those of photons of known wavelengths giving rise to the appearance of the same cloud edges in the light of photo-excited 1-0 S(1) H 2 vibrational fluorescence at 2.12 µm(lemaire et al., 1996; Field et al., 1998; Takami et al., 2000; An & Sellgren, 2003) and in the light of scattered radiation in the z-band. In NGC 7023, the former is the result of photo-excitation in the Lyman and Werner bands in H 2 (Takami et al., 2000), which occurs near 110 nm and shortward in wavelength, while the latter is the result of simple scattering by dust at the effective wavelength of observation near 900 nm. The penetration depths of the respective exciting radiations are inversely proportional to the respective extinction coefficients for the exciting radiations, with local gas/dust densities canceling when ratios are considered. The unknown wavelength region of ERE initiation can then be estimated from the derived ratios of extinction coefficients, taking into account the opacity sources present in the nebular environment. 5.2 of this chapter presents the observations and reductions; in 5.3 we present the results. The discussion of the implications of our results are contained in 5.4. In particular, we apply the wavelength constraint for ERE initiation to two well-studied ERE sources to assess the likelihood of a two-step excitation process for the ERE. This is followed by a set of conclusions in 5.5.

146 Observations and Data Reduction Figure 5-1 Location of the regions shown in Figs. 5-2 and 5-3 on a F606W WFPC2 image of NGC 7023 (Gordon et al., 2000).

147 132 The prominent northwest filament in NGC 7023 (see Figure 5-1) was imaged with the Advanced Camera for Surveys (ACS) and the Near Infrared Camera and Multi Object Spectrometer (NICMOS) on the Hubble Space Telescope (HST) as part of HST program #9741. The ACS images were taken in F475W, F625W, F850LP, and F656N which correspond to SDSS g, SDSS r, SDSS z, and Hα filters. The centers of the ACS images were offset to avoid the bright central star of NGC The three broad-band ACS images were taken to allow for the creation of a continuumsubtracted ERE image and the narrow band F656N filter to measure the contribution from scattered Hα emission to this ERE image. The origin of the scattered Hα radiation is the strong Hα emission line in the spectrum of the central illuminating star, HD The NICMOS images were taken with the NIC2 camera and the F212N and F215N filters which measure the 1-0 S(1) H 2 emission line wavelength range and associated red continuum at µm and 2.15 µm respectively. The observations in each ACS filter were taken split between two images with a large enough dither to fill in the gap between the two chips. The total exposure times were 1000, 1000, 1200, and 1560 s for the F475W, F625W, F850LP, and F658N filters, respectively. The two images per filter were combined with the online multidrizzle and additionally processed using LACOSMIC (van Dokkum, 2001) to identify cosmic rays which were not removed with multidrizzle. Residual cosmic rays are present in the final mosaics, especially in the gap between the two ACS chips where we only have a single measurement. These residual cosmic rays do not affect our measurements as they rely on structures much larger than a cosmic ray.

148 133 N W SDSS g ACS F475W SDSS r ACS F625W SDSS z ACS F850LP ACS F658N ERE H alpha Figure 5-2 A region, rotated from North, and centered on 21 h 01 m , (2000) is shown for the four observed ACS filters and continuum-subtracted ERE and Hα images. This region encompasses the northwest filament of NGC A region extracted from the ACS images focusing on the northwest filament is shown in Fig. 5-2 for all four observed filters and the continuum-subtracted ERE and Hα images. The ERE image was created by first creating a continuum image due to dust scattered light using a linear combination of the F475W and F850LP images. These two filters have central wavelengths (474 nm and 905 nm, respectively) to the blue and red of the ERE peak seen in this filament which is at 650 nm

149 134 (Gordon et al., 2000). The resulting continuum image was subtracted from the observed F625W image to produce the ERE image. The success of the continuum subtraction can be seen by the much reduced strength of the clump southwest of the filament (brightest extended source in all four observed ACS bands). This clump is composed of only scattered light as it has a linearly decreasing spectrum across the three broad ACS bands as measured in a pixel box centered on the clump. The Hα image was created by subtracting the F625W image from the F658N image after scaling the F625W image by the ratio of fluxes of stars measured in both images. The prominence of ERE in this filament is clearly seen as the F625W image is much sharper than the F475W and F850LP images. If the F625W image were dominated by scattered light and not ERE, it would have an intermediate appearance between the F475W and F850LP. One possible reason for this sharper appearance could be Hα emission. This is not the case as the sharpness of the continuum-subtracted Hα image more closely matches the F425W and F850LP images than the F625W or ERE images. Note that the Hα image has the same morphology as the scatteredlight dominated F425W and F850LP images. This is because the central star of NGC 7023, HD , has a strong Hα emission line (Gordon et al., 2000). Thus, the continuum-subtracted Hα image traces the scattered Hα light of the central star, not Hα emission from extended gas in the nebula.

150 135 NICMOS F212N NICMOS F215N ERE H2 ACS F850LP Figure 5-3 A region, rotated from North, and centered on 21 h 01 m , (2000) is shown for the two observed NICMOS filters, continuum-subtracted ERE, continuum-subtracted H 2, and ACS F850LP images. This region encompasses the bright portion of the northwest filament of NGC The observations in the two NICMOS filters were taken split between 4 images with a spiral dither pattern. The total exposure times were 352 s in both the F212N and F215N filters. The 4 separate images were combined using custom scripts to allow for careful masking of bad pixels and the coronographic spot. A single star was present in each image allowing for accurate registration between images and to the ACS observations. The mosaics of the F212N and F215N observations are shown in the top row of Fig. 5-3 displayed with the same scale. It is clear from these two images that this filament has very strong H 2 emission which was known from existing near-infrared spectroscopy (Gordon et al., 2000). The H 2 image was created by subtracting the calibrated F215N image from the calibrated F212N image.

151 136 The accuracy of the subtraction of the continuum can be easily seen as the stellar image disappears completely as well as the diffraction spikes from the central star of NGC In addition to the NICMOS images, the continuum-subtracted ERE and ACS F850LP images are shown for the same region in Fig Results The ERE, H 2, and z-band Morphology of NW Filaments We examined the filamentary morphology of the NW PDR in NGC 7023 on two spatial scales. The first scale is defined by the more limited field of view of NICMOS and focuses on the complex, multi-filament structure in the center part of the NW PDR in NGC 7023 shown in Figure 5-3. Our observations of the 1-0 S(1) emission from molecular hydrogen are limited to this field of view, and we are presenting this image together with the ERE and z-band images limited to this field. The wider field of view of the ACS defines the second scale shown in Figure 5-2; it includes the southern and northern extensions of the filaments in Figure 5-3, for which we can compare the ERE structure only with the corresponding structure in the z-band. The data shown in Figures 5-2 and 5-3 represent the highest-resolution images of the NGC 7023 NW PDR to date. The location of these images within the larger extent of NGC 7023 is indicated in Figure 5-1. The structure of the NW PDR shown in the more limited field of NICMOS (Fig-

152 137 ure 5-3) is highly complex and consists of a series of narrow filaments with typical widths of 1 when viewed in the light of ERE and the light of H 2. These filaments are fully resolved in both images. The angular resolution is marginally higher in the ACS images compared to the NICMOS images. This is based on comparisons of stellar images which yielded ratios FWHM(ACS)/FWHM(NICMOS) = 0.93, with stellar images having diameters with FWHM of and respectively. When viewed in the z-band, which is dominated by dust-scattered light, the same structures appear much more diffuse, although they were imaged with the same angular resolution. This comparison suggests that the sharpness of the structures as seen in ERE and H 2 is not a result of a limited physical extent of the structures, e.g. thin sheets viewed edge-on, but that they have a physical width that is wider than indicated by the ERE and H 2 filament images. The narrow ERE and H 2 filaments, therefore, may represent just the edges of more extended molecular cloud clumps illuminated by the central B3Ve star in NGC 7023, HD

153 138 Case 4 Case 1 ERE H Case 3 Case 2 Figure 5-4 Four cuts through the ERE (solid line) and H 2 (dotted line) images are given illustrating the 4 cases seen. These cases are 1) near-perfect match between ERE and H 2 filaments, 2) evidence for H 2 self-shielding, 3) H 2 filaments broader than ERE, and 4) H 2 filaments narrower and in front of ERE filaments. The x-axes of the graphs increase in the direction of propagation of the exciting radiation. There is independent evidence from the interferometric HCO + observations by Fuente et al. (1996) for the existence of at least two but possibly four high-density molecular clumps in our field of view that can be distinguished on the basis of their differing radial velocities. In particular, the extended filament that begins on the

154 139 right-hand side of Figure 5-2 and turns upward, passing near the embedded star (star K in the map of An & Sellgren, 2003), appears to be a coherent structure moving at 2.4 km s 1, while the group of narrow filaments shown in Figure 5-3 closest to the illuminating source belongs mostly to a structure moving at 4.0 km s 1. The apparent complexity of the filamentary structure of the NW PDR is most likely the result of the line-of-sight superposition of several independently moving molecular clouds, whose surfaces facing HD are currently being photo-dissociated. A comparison of the morphology of the H 2 and ERE filaments in Figure 5-3 reveals two significant facts. One, the ERE filaments appear to be about as sharp as the corresponding H 2 filaments. We will investigate this aspect further in 5.3.3; for now we can conclude that the radiation initiating the H 2 fluorescence and the ERE, respectively, apparently faces similar optical depths per unit mass of molecular cloud environment. Two, the spatial correlation between the distribution of the ERE and the H 2 emission in Figure 5-3 is not exceptionally strong. ERE is present wherever H 2 emission is seen, but there are bright ERE filaments, especially in the upper part of Figure 5-3, for which the corresponding H 2 features are either much weaker or absent altogether. A similar qualitative relationship between ERE and H 2 emission has previously been noted by Field et al. (1994) in a study of the reflection nebula NGC 2023 and was confirmed in a more detailed study of NGC 7023 by Lemaire et al. (1996). When comparing the detailed profile shapes and positions of ERE and H 2 filaments, we can identify four classes, illustrated in Figure 5-4. These can be described as follows. Case 1 represents locations where ERE and H 2 exhibit nearly identical

155 140 structures. Out of 36 cuts examined, seven fall into this category. Class 2 includes cases (17/36) where the first H 2 filament is very narrow and recessed with respect to the corresponding ERE filament. A second ERE filament is not matched by H 2. We believe these to be instances of H 2 -self-shielding. Case 3 represents instances (9/36) where the H 2 filament is broader than the corresponding ERE filament, while Case 4 illustrates the rather rare (3/36) circumstance where the H 2 filament is narrower than the ERE filament and peaks slightly in front of the ERE filament. The near-absence of H 2 emission (Case 2) in regions exhibiting strong ERE cannot be explained by differences in line-of-sight extinction. We just concluded that both types of emission appear to be initiated by photons of similar energy; the resulting outgoing 1-0 S(1) H 2 emission faces an opacity only 1/8 as large as the resulting ERE. Hence, if one type of emission should be missing due to extinction, it would be the ERE, not the H 2 emission, contrary to our observations. As noted by Field et al. (1994) and Lemaire et al. (1996), this result provides a strong argument against models in which the ERE carrier is initiated chemically by hydrogenation of carbonaceous dust in the hot atomic hydrogen gas found in H 2 photo-dissociation fronts (Witt & Schild, 1988; Duley & Williams, 1990). Lack of carriers also cannot be the explanation for the absence of H 2 luminescence; H 2 molecules are the principal constituents of molecular clouds in the ISM.

156 141 Figure 5-5 The density-dependence of the dust optical depth of a PDR (solid line) and the efficiency of converting far-uv photons into near-ir 1-0 S(1) H 2 photons (dotted line) is shown. The NGC 7023 NW PDR involves molecular gas with densities ranging from n(h 2 ) 10 4 cm 3 to n(h 2 ) 10 6 cm 3 (An & Sellgren, 2003). We notice that ERE filaments in the NW PDR that lack H 2 counterparts of corresponding surface brightness tend to be broader than those that have bright H 2 counterparts. If this broader, more diffuse appearance is a reflection of the penetration depth of the radiation initiating the ERE, this then indicates that the density in the ERE filaments lacking H 2 counterparts is substantially lower. Under these conditions, the UV pumping efficiency as measured by the number of 1-0 S(1) photons emitted per incident photon in the wavelength range nm is significantly reduced (Draine & Bertoldi, 1996), because dust is absorbing an increasing fraction of the photons needed for the excitation of H 2. This is because molecular hydrogen is unable to be re-formed as quickly under lower-density conditions, the rate being proportional

157 142 to the square of the density. As we had seen earlier, the ERE carrier is part of the dust and a likely competitor for photons in the far-uv range. Thus, with competition lacking from molecular hydrogen, the ERE carrier is able to absorb a greater fraction of the available photons under lower-density conditions. We illustrate the dependence of the dust optical depth of a PDR and the efficiency of converting absorbed far-uv photons into near-ir photons in the 1-0 S(1) H 2 transition in Figure 5-5, calculated from relations provided by Draine & Bertoldi (1996). The dust optical depth of the PDR is measured from the front of the molecular cloud to the point at which 50% of the H 2 is photo-dissociated. It appears likely, therefore, that the lack of a detailed spatial correlation between ERE filaments and the 1-0 S(1) emission of H 2 is a result of density variations by about one order of magnitude between high-density filaments, where both ERE and H 2 are bright, and lower density cloud faces, where the ERE is dominant Determination of R V in NW Filaments When viewing the filaments in the light of different emission mechanisms, e.g. the fluorescent 1-0 S(1) vibrational emission of H 2 at 2.12 µm and the much more diffuse appearance of the same structures in the light of dust-scattered light in the z-band (Figure 5-3), we are comparing the relative penetration into the cloud surfaces of the respective exciting radiations. In the case of the fluorescent H 2 emission, the exciting radiation is Lyman and Werner band photons absorbed by H 2 molecules at wavelengths λ nm. The opacity restricting the transfer of such photons

158 143 into the cloud faces is most certainly extinction by dust and, depending on local density conditions, self-shielding by H 2 (Draine & Bertoldi, 1996). In the z-band, dust extinction is the only significant opacity source. The relative penetration depth along a given line-of-sight from the central star into a molecular cloud is determined by the inverse ratio of these two opacities. Taking into account the possible contribution by H 2 self-shielding to the opacity, the ratio of the widths of the H 2 and the z filaments at identical locations, then, is an upper limit to the ratio of the dust opacities at the two effective wavelengths involved, nm and 900 nm, respectively. This ratio can now be used to identify the appropriate dust extinction curve for the molecular cloud fronts in the NGC 7023 NW PDR. Figure 5-6 Extinction curves generated from the CCM R V -dependent relationship are shown for a range of R V values.

159 144 Figure 5-7 Histogram of the ratios FWHM(z)/FWHM(H 2 ) and FWHM(z)/FWHM(ERE) of filaments widths. Cardelli et al. (1989, hereafter CCM) showed that the wide range of extinction curves encountered in different galactic environments can be reproduced as a family of functions dependent upon a single parameter, the ratio of total-to-selective extinction, R V, as illustrated in Figure 5-6. The low-density diffuse ISM in the Milky Way is most commonly represented by CCM curves with R V = 3.1 while dense molecular cloud material exhibits CCM extinction curves with R V in the range from 5 to 6. As long as we adopt the CCM formalism, these curves can be distinguished from each other by measuring a single ratio of extinctions at two widely separated wavelengths. As we noted before, such extinction ratios can be estimated from the relative penetration depth at the respective wavelengths. We measured the FWHM of the H 2 and z-band filaments (Figure 5-3) at 22 positions where such measurements could be carried out

160 145 without interference from blends by closely spaced H 2 filaments, whose z-counterparts would merge into a single profile. A histogram of the measurements of the ratio of FWHM(z)/FWHM(H 2 ) is shown in Figure 5-7 by a dotted line. The average of the measured ratios of FWHM(z)/FWHW(H 2 ) is 2.74 ± In Figure 5-8 we show the ratio of the dust extinction cross sections at nm and 900 nm, the latter being the effective wavelength of the z-band, as a function of R V for CCM extinction curves, and we have entered our measured ratio as a horizontal bar. We conclude from this figure that a CCM extinction curve with R V = is a good estimate for the dust extinction curve in the molecular clouds making up the NW PDR in NGC If H 2 self-shielding is a significant contributor to the total extinction for the H 2 -exciting radiation, this value of R V is a lower limit. The fact that R V = 5.62 is very typical of molecular cloud environments (R V = 5.5) suggest, however, that H 2 self-shielding is a minor contributor to the opacity for the H 2 -exciting radiation in the filaments measured in this experiment Width of the ERE Filaments Employing the same rationale as in the previous section, we now compare the relative widths of the ERE and z filaments in Figures 5-2 and 5-3 in identical locations. A total of 41 measurements were made in the larger field (Fig. 5-2) and 26 such measurements were made in the smaller field (Fig. 5-3). A histogram of the resulting ratios FWHM(z)/FWHM(ERE) is shown in Figure 5-7 (solid line). The two sets of measurements resulted in average ratios of FWHM(z)/FWHM(ERE) of 2.52 ± 0.65

161 146 and 2.32 ± 0.64, respectively, with a weighted average of 2.40 ± Comparing this value to the corresponding ratio of FWHM(z)/FWHM(H 2 ) = 2.74±0.61, we conclude that on average the ERE filaments are about 10% broader than the corresponding H 2 filaments, if they are present, and that the opacity facing the radiation that initiates the ERE is, therefore, about 10% less than the opacity encountered by the radiation that excites the H S(1) radiation. Figure 5-8 The ratio of extinctions at nm and 900 nm is plotted as a solid line. The allowed region from the measurement of the FWHM(z)/FWHM(H 2 ) ratio is shown as the shaded region Wavelength of ERE Initiation In order to identify the wavelengths of the radiation that appears to be required for ERE initiation, we plotted a CCM extinction curve for R V = 5.62, plus two dotted extinction curves representing the uncertainty limits in the R V value, found to be ap-

162 147 propriate for the dust in the molecular material in NGC 7023, in Figure 5-9. Assuming a ratio A(z)/A(V) = 0.59 for the R V = 5.62 case, we place our measured ratio of FWHM(z)/FWHM(ERE) onto this graph as a horizontal line at A(λ)/A(V) = This constraint implies that only those wavelength ranges in which the extinction exceeds this horizontal limit exhibit adequate dust opacity to be consistent with our data in Figure 5-7. Resulting from the multi-valued nature of the extinction curve, this identifies two spectral regions: the range from 250 nm to 189 nm, where the horizontal line crosses the nm bump in the extinction curve, and the region shortward of 118 nm in the far-uv, where the far-uv rise of the extinction curve reaches values in excess of our opacity constraint. Thus, similar to the investigation by Witt & Schild (1985), our method alone does not result in the identification of a unique spectral region responsible for ERE initiation, but it clearly excludes near-uv or optical radiation slightly shortward of the ERE band from consideration.

163 148 Figure 5-9 The limits on the ERE excitation wavelength are shown on top of a CCM R V = 5.62 extinction curve. The constraint imposed by the ERE filament widths is discussed in 5.3.3; the constraints due to the limits on the effective temperatures of the exciting stars are discussed in 5.3.4; and the two vertical lines labeled < 100% and < 50% related to the insufficiency of far-uv photons for the actual excitation of the ERE are discussed in We can now narrow down the choice between the two UV spectral ranges identified above by applying the results of Darbon et al. (1999) as a further constraint. In short, Darbon et al. (1999) found that stars with T eff 10, 000 K can excite ERE and very commonly do, while stars with T eff 7000 K do not. We entered the FUV cut-offs of the SEDs of atmospheres with these two temperatures in Figure 5-9 as two vertical lines. The long-wavelength limit of the spectral region in which ERE is initiated radiatively is defined by the point at which our applicable extinction curve enters from below the cross-hatched area limited by the three constraint lines. For R V = 5.62, this wavelength is 118 nm, corresponding to a photon energy near 10.5 ev. Therefore,

164 149 the application of the constraints by Darbon et al. (1999) eliminates the nm extinction band as the source of excitation of the ERE. Our new constraint based upon the penetration of the exciting radiation into the cloud face considerably narrows the acceptable wavelength range for ERE initiation allowed by the Darbon et al. (1999) results, however. Both sets of constraints are consistent with the possibility that photons with wavelengths between 118 nm and 91.2 nm, the latter being set by absorption by HI, can contribute to ERE initiation. 5.4 Discussion Having obtained a fairly accurate estimate of the wavelengths of the photons needed to initiate the ERE will be critical for determining further details of the ERE process as well as the ERE carrier. As we stated earlier, our present observations do not allow us to determine whether photons with E > 10.5 ev are required to excite the ERE carrier, with (at best) one ERE photon resulting from each absorption of each far-uv photon, or whether the presence of far-uv photons is merely required for the creation of the ERE carrier, e.g. by either photo-ionization or photo-dissociation of an ERE-precursor, followed by pumping of the newly created ERE carrier by lowerenergy optical photons. However, this issue can be resolved directly by counting the number of photons available in the 118 nm to 91.2 nm range that are absorbed in a given system and comparing this to the number of ERE photons emitted. We will return to this question in and

165 Impact on Existing ERE Models While the nature of the ERE as a dust-related interstellar photoluminescence process has been undisputed, many diverse proposals have been advanced for possible identifications of the ERE carrier (see Witt & Vijh, 2004, for a review). With the determination of the minimum wavelength of the radiation required for ERE initiation, many of these proposals can now be ruled out. In discussing these proposals below, we will focus primarily on the excitation requirements and not on respective failures to meet other observational constraints. Neutral PAH Molecules d Hendecourt et al. (1986) proposed that fluorescence, i.e. electronic transitions from the S 1 state to vibrational levels of the S 0 ground state, or phosphorescence, i.e. electronic transitions from the T 1 state to vibrational levels of the S 0 ground state, in neutral polycyclic aromatic hydrocarbons (PAHs) could be responsible for the ERE. The same species are considered the likely source of the ubiquitous mid- IR aromatic emission features (AEF) (Bakes et al., 2004; Onaka, 2004; Peeters et al., 2004). In the laboratory, many PAH molecules are known to fluoresce efficiently at optical wavelengths. However, as noted by d Hendecourt et al. (1986), PAH fluorescence generally occurs at much shorter wavelengths than the ERE band while phosphorescence frequently occurs in the same spectral region as the ERE. Although the T 1 - S 0 transitions are forbidden, isolated PAH molecules in the ISM could conceivably be strong sources of phosphorescence, provided that the T 1 state can attain

166 151 high populations. The T 1 state, which is always at a lower energy compared to the S 1 state in neutral PAH molecules, is populated by intersystem crossings from excited states of the singlet manifold to highly-excited vibrational levels of the T 1 state. This implies, therefore, that the excitation of phosphorescence occurs via the same absorption bands that excite the fluorescence in these molecules. The absorption spectra of neutral PAH molecules are well known; their absorption bands are found immediately shortward in wavelength of where their S 1 - S 0 fluorescence occurs, with the strongest bands found generally near 200 nm. Thus, the wavelengths of radiation required for the excitation of PAH fluorescence and PAH phosphorescence are two to three times longer than what was found in the present investigation. Recently, neutral PAH fluorescence (Vijh et al., 2004, 2005b) has been detected in the Red Rectangle, an object in which ERE is present as well, and the spatial distributions of the two emissions were found to be distinctly different. This is only further support for our conclusion that neutral PAH phosphorescence is not the source of ERE. PAH Clusters Seahra & Duley (1999) suggested that PAH clusters in the form of stacks or aggregates of PAH molecules with up to 700 carbon atoms could produce ERE-like photoluminescence in a band centered near 700 nm. The same authors (Duley & Seahra, 1998) had previously proposed that the same PAH clusters could be responsible for the interstellar absorption band at nm as well as the mid-ir AEF bands. The energy for both the ERE and the AEF in this model was to come from the absorption in the nm band. While the present penetration study does not exclude this

167 152 possibility, the application of the Darbon et al. (1999) constraint (Figure 5-9) does. The PAH cluster model is therefore not supported by the ERE excitation results. Hydrogenated Amorphous Carbon Photoluminescence by grains consisting of hydrogenated amorphous carbon (HAC) or coated by HAC mantles was an early candidate for the ERE process (Duley, 1985; Witt & Schild, 1988; Duley & Williams, 1990; Jones et al., 1990). With a band gap of over 3 ev (Robertson, 1996), HAC not only exhibits efficient photoluminescence at wavelengths much shorter than observed in ERE, it also is efficiently excited at optical/near-uv wavelengths (Watanabe et al., 1982), which is in conflict with our present results. The latter objection can also be raised against a material closely related to HAC, referred to as quenched carbonaceous composite (QCC) (Sakata et al., 1992). While the band gap of various HAC materials can be reduced by a variety of treatments, the most efficient excitation occurs always just shortward of the emission band resulting from HAC photoluminescence. This puts these models in serious conflict with our new excitation constraints. Silicon Nanoparticles An ERE model that received intensive study in recent years involves photoluminescence by oxygen-passivated silicon nanoparticles (SNPs), containing between about 200 and 6000 silicon atoms (Witt et al., 1998; Ledoux et al., 1998, 2000, 2001, 2002; Smith & Witt, 2002; Li & Draine, 2002). The band gap of SNPs is size dependent due to quantum confinement and with suitable limits on the nanoparticle sizes it

168 153 readily matches the observed wavelengths of the ERE band. However, again as in all instances of typical semiconductor photoluminescence, the most effective excitation occurs at energies just slightly above the band gap energies, i.e. slightly above 2.3 ev. In addition, SNPs are photo-ionized when exposed to photons with energies in excess of 5.1 ev (Fuke et al., 1993), with the result that photoluminescence from subsequent excitations is quenched (Nirmal & Brus, 1999; Smith & Witt, 2002). Consequently, our new constraints regarding ERE initialization also do not favor the SNP model Possible New ERE Carriers All previously considered ERE carrier models appear to be inconsistent with the requirement that photons with energies in excess of 10.5 ev are required to initiate the ERE in astrophysical environments. There are several reasons for assuming that the energy limit of 10.5 ev is not directly related to the actual excitation of the ERE. Below, we will argue that the number density of photons with energies in excess of 10.5 ev in typical ERE-producing environments is insufficient to account for the number of ERE photons emitted from those environments, if the E > 10.5 ev photons were to be the direct source of excitation, even if a photon conversion efficiency of 100% is assumed. A possible solution to this dilemma is provided by models in which the E > 10.5 ev photons simply create the ERE carrier, with a small number of FUV photons required for its maintenance. The ERE carrier thus created needs to have the ability for efficient photoluminescence in the nm wavelength range and strong absorption throughout the optical/uv regions of the spectrum, where ample

169 154 photons for the pumping of the ERE are available. Furthermore, the carrier species involved in such a process must be abundant in the ISM in order to account for the contribution of the ERE carrier to the absorption in the ISM, amounting to about 10% of the photons of the Galactic interstellar radiation field in the nm wavelength range (Gordon et al., 1998).

170 I 1 I H Ionization Opacity 13 Ionization Potential (ev) Dust Extinction Opacity + H 2 Self-shielding Opacity Dust Extinction Opacity Molecular Wt. (amu) Figure 5-10 First ionization potentials of PAH molecules (filled circles) (Eilfeld & Schmidt, 1981) and their corresponding cations (open circles) (Leach, 1996) versus molecular weights. The three horizontal lines correspond to the cutoff of the far-uv radiation field due to H-ionization opacity (top), the onset of the energy regime in which dust extinction and H 2 self-shielding opacity compete for photons (middle), and the lower limit on the energy for ERE initiation (lowest), respectively.

171 156 In terms of abundance and total cross section, interstellar PAHs meet these requirements. However, as shown in 5.4.1, neutral PAHs cannot account for the observed ERE. By contrast, doubly ionized PAH molecules offer interesting prospects. In Figure 5-10 we have plotted the first ionization potentials of neutral PAH molecules and of PAH mono-cations as a function of the size of the molecules as measured by their molecular weights. We note that most PAHs with up to 50 carbon atoms can be readily ionized into the mono-cation stage with photons with energies in the 6 ev to 8 ev range, and can subsequently be sent into the di-cation state with photons in the energy range from 10.5 to 13.6 ev. We should, therefore, expect to find PAHs in the di-cation state in environments characterized by the presence of neutral or molecular hydrogen, as previously suggested by Leach (1987, 1996). PAH di-cations are closedshell systems similar to neutral PAH molecules with S 1 - S 0 transitions occurring at optical wavelengths longer than the corresponding transitions in neutral molecules. Fluorescence could thus be expected in the ERE range. The viability of such a model requires laboratory spectroscopy on PAH di-cations that could reveal the presence of optical fluorescence, following the excitation through near-uv/optical absorptions. No such experimental data exist at this time.

172 157 optical absorption (a.u.) anthracene++ coronene++ pyrene++ terrylene++ ovalene photon energy (ev) Figure 5-11 Model spectra of representative PAH di-cations. We can, however, obtain some general idea of the expected spectroscopic properties of PAH di-cations, using density-functional theory modeling calculations (TD- DFT) (Onida et al., 2002). This theory incorporates electronic screening and relevant correlation effects (Xie et al., 2004) for vertical electronic excitations in the groundstate geometry and thus represents a fully ab initio formalism for computing excited states. Over the past decade, TD-DFT has given promising results for finite systems (atoms, molecules, clusters) (Xie et al., 2004; Onida et al., 2002; Xie et al., 2005). In this work, we use the TD-DFT implemented in the Gaussian 03 package (Frisch, 2003). Based on the ground-state structures for all clusters, which were optimized by using DFT (Kohn & Sham, 1965) with the nonlocal hybrid B3LYP functional (Becke, 1993), we compute the excited energies of both singlet and triplet states (Lewis & Kasha, 1944), the oscillator strengths, and the optical absorption gap (gen-

173 158 erally defined by the energy of the first dipole-allowed transition for finite systems (Xie et al., 2004)). To ensure the accuracy of our calculations, we choose 6-31G(d,p) Pople-type basis set which adds polarization functions to the atoms (Xie et al., 2005). In Figure 5-11 we show the composite absorption spectra of five PAH di-cations, including results for pyrene++ and anthracene ++, whose likely presence in the Red Rectangle (HD 44179) has been demonstrated by both their blue fluorescence as well as their spectroscopic signature of the first ionization (Vijh et al., 2004, 2005b). When compared with available experimental data for neutral and once-ionized PAH species, these types of calculations agree remarkably well with experimental measurement. The spectra in Figure 5-11 show that representative PAH di-cations absorb strongly in the energy range from 2 ev to 7.5 ev, suggesting a characteristic band gap of 2 ev. Photoluminescence across this band gap could be responsible for the ERE. A still controversial issue is the question of the stability of twice-ionized PAH molecules. A recent study of the fragmentation pathways of the di-cation of benzene by Rosi et al. (2004) revealed an unexpectedly high degree of stability. This result suggests that the di-cations of multi-ringed PAHs could exhibit a still higher degree of stability than benzene and thus play a substantial role in interstellar chemistry and photophysics Consistency Check: The Red Rectangle The two principal results emerging from the study of the excitation of the ERE in the NGC 7023 NW PDR are: (1) the ERE initiation requires photons with energies

174 159 > 10.5 ev, e.g. for the creation of the ERE carrier; (2) the ERE process is most likely a two-step process, in which the first step creates the carrier via ionization or dissociation, requiring in excess of 10.5 ev, and the second step produces the actual excitation of the ERE carrier via longer-wavelength photons. In this section we will examine the Red Rectangle, one of the two environments where ERE is being produced with very high efficiency, the other being the high- b Galactic cirrus. Two aspects are of greatest significance, the morphology of ERE sources and the ERE energetics. The Red Rectangle is a well-studied bi-polar protoplanetary nebula with a circum-stellar disk obscuring the central source (Men shchikov et al., 2002; Cohen et al., 2004). The central source consists of a L 6000 L AGB star with a T eff 8250 K atmosphere (Vijh et al., 2005b) and a low-mass (M 0.35 M ) hot white dwarf of unknown luminosity, which could amount to at most 100 L (Men shchikov et al., 2002; Driebe et al., 1998). This upper limit to the the white dwarf s luminosity would apply only, however, during the 10 5 yr duration of the stellar core s contraction from the AGB region of the Hertzsprung-Russell diagram to the top of the white dwarf cooling sequence. Once the stellar core has reached a size corresponding to the white dwarf mass-radius relation, a more typical luminosity of a T eff = 60,000 K hot white dwarf is about 6.5 L. The relevance of these luminosity limits will become apparent presently. Spatially, the ERE in the Red Rectangle is confined to geometrically thin regions within the walls of the bi-conical outflow cavities (Schmidt & Witt, 1991; Cohen et al., 2004, Vijh et al., in prep.), leading to the appearance of the X-shaped bipolar structure, while scattering in the 500 to 350 nm wavelength range produces a spherical

175 160 reflection nebula devoid of the bipolar structure (Cohen et al., 2004). The confinement of the ERE to the walls of the outflow cavity is in agreement with our finding of an initiation of the ERE by high-energy UV photons. These photons, stemming predominantly from the hot white dwarf, are spatially confined to the outflow cones and its walls by the optically thick circum-binary disk and the steeply rising far-uv opacity of the nebular dust in the Red Rectangle (Vijh et al., 2005b). At wavelengths longward of 170 nm (E< 7.3 ev), the nebular opacity is greatly reduced. Thus, the mid-uv, near-uv and optical photons, which are produced almost entirely by the cooler but extremely luminous AGB star, are able to penetrate the walls of the outflow cavity and produce the nearly spherical blue reflection nebula component of the Red Rectangle (Cohen et al., 2004). They are also able to efficiently pump the ERE carriers produced in the thin wall regions of the outflow cones. The ERE energetics of the Red Rectangle is particularly useful in supporting the concept of a two-step process of ERE excitation. Schmidt et al. (1980) reported the band-integrated ERE luminosity of the entire Red Rectangle nebula as L ERE 0.6 (D/280 pc) 2 L. Using the currently accepted distance of 710 pc (Men shchikov et al., 2002; Hobbs et al., 2004), we arrive at an estimate of L ERE 3.9 L for the Red Rectangle. This value is uncorrected for attenuation by the optically thick circum-binary disk surrounding the stellar sources as well as for any interstellar extinction along the line of sight. Given the unusual color characteristics of the RR and the non-isotropic morphology of the RR, these attenuations are difficult to estimate. The light of the central AGB star alone suffers an attenuation in excess of A V = 4 magnitudes (Vijh et al., 2005b). We, therefore, consider it a conservative estimate

176 161 that the total ERE in the RR suffers an attenuation by about 1 magnitude, resulting in a total estimated ERE luminosity of L ERE 10 L. If photons with E > 10.5 ev were responsible for the actual excitation of the ERE in a single-step process, and if the excitation process were 100% efficient, with one ERE photon of 1.9 ev produced for every absorption of a E> 10.5 ev photon, the total luminosity in E > 10.5 ev photons required for this excitation would be L FUV > 55 L. Thus, with somewhat more realistic assumptions about the ERE excitation efficiency of possibly 30% to 50%, we would require well in excess of 100 L in far-uv photons to produce the observed ERE luminosity of the Red Rectangle, well in excess of what is likely available from the hot white dwarf companion. If, by contrast, the far-uv photons emitted by the hot white dwarf companion of the central AGB star are used solely for creating and maintaining the ERE emitters, the AGB star (T eff 8250 K; L 6000 L ) has more than enough energy to excite L ERE 10 L with mid-uv and optical photons. Assuming a Stokes shift of 1 ev typical for photoluminescence, absorption of an excitation energy of less than 15 L would suffice to generate L ERE 10 L, assuming 100% efficiency. (Gordon et al., 1998) have estimated the ERE photon conversion efficiency to be > 10%, which would lead to an upper limit for the ERE generating luminosity of 150 L. This is still only about 2.5% of the luminosity of the AGB star. Thus, a two-step process of ERE excitation is entirely feasible, given the energetics of the Red Rectangle nebula system.

177 Consistency Check: The High- b Galactic Cirrus Next to the Red Rectangle, the high- b Galactic cirrus is another environment in which the intensity of the ERE is roughly equal to that of the dust-scattered light, for very similar reasons. First, in both instances the scattering geometry involves mainly large-angle scattering, which does not lead to high scattered-light intensities. Second, the illuminating radiation field in both cases contains both a strong far-uv component and a strong optical component (see Witt & Johnson, 1973, for a spectrum of the interstellar radiation field), a condition which appears to be associated with highly efficient production of ERE. The interstellar medium giving rise to the high- b Galactic cirrus is optically thin in the far-uv, where the ERE is initiated, as well as at the ERE wavelength range itself. This, coupled with a reasonably accurate knowledge of the spectrum of the exciting interstellar radiation field (Mathis et al., 1983), allows a reliable estimate of the ERE intensity per hydrogen atom along an average line of sight to be made, as well as an estimate of a lower limit of the photon conversion efficiency of the ERE process (Gordon et al., 1998). Gordon et al. (1998) found an ERE intensity of (1.43 ± 0.31) ergs s 1 Å 1 sr 1 H-atom 1 for the Galactic cirrus at b > 20, independently confirmed by Szomoru & Guhathakurta (1998). With the well-established observed relation between hydrogen column density and extinction in the Galaxy (Diplas & Savage, 1994), Gordon et al. (1998) estimated that (10 ± 3)% of all photons that are absorbed from the interstellar radiation field by interstellar dust must be absorbed by the ERE carrier to produce the observed ERE intensity with a conversion efficiency of 100%

178 163 on a photon-per-photon basis. If, as our current results suggest, far-uv photons are required to initiate the ERE, we must first consider the far-uv portion of the interstellar radiation field. If we consume photons through absorption by the ERE carrier, starting at the 91.2 nm limit set by the interstellar hydrogen opacity and proceed toward longer wavelengths, we find that we must use all photons shortward of 130 nm in wavelength or with energies down to 9.5 ev. This limit is indicated by a vertical dashed line in Figure 5-9, labeled 100%. We note that this limit violates the ERE initiation limit of 10.5 ev set by the opacity constraint found in NGC If we assume a more realistic conversion efficiency of 50% for the photons absorbed by the ERE carriers, all photons absorbed between 91.2 nm and 185 nm in wavelength would be required to generate the observed ERE intensity. This second limit is shown in Figure 5-9 as a dashed vertical line labeled 50%. Therefore, in close analogy to the case of the Red Rectangle discussed earlier, we find that that the interstellar radiation field does not contain enough photons with energies E > 10.5 ev to pump any ERE carriers sufficiently to generate the ERE intensities observed from the high- b Galactic cirrus. Again, the problem would meet with a simple solution, if the high-energy photons are used solely to produce and maintain a population of ERE carriers which can then be pumped by the far more abundant optical/near-uv photons of the interstellar radiation field. Thus, a twostep process of ERE excitation is also being suggested by the data available for the high- b diffuse interstellar medium.

179 Conclusions We summarize our conclusions: 1. We have imaged sharp molecular cloud edges in the NW PDR of the reflection nebula NGC 7023 in the light of the ERE band and in the light of emission from the H S(1) transition, using the HST ACS and HST NICMOS instruments, respectively. The cloud edges appear as narrow filaments with widths as small as 0.3 in both bands, while the appearance of the same cloud edges is broader and more diffuse in images showing dust-scattered radiation at Hα and in the z-band. 2. In general, every H 2 filament is matched by a corresponding ERE filament, while the reverse correlation does not hold to the same degree. We interpret this onesided correlation as resulting from a density-dependent conversion efficiency of far-uv photons into near-ir 1-0 S(1) photons. In lower-density regions, absorption of far-uv photons by dust dominates over the absorption by H 2 molecules, leading to ERE filaments without corresponding H 2 filaments, while in higher-density regions absorptions of far-uv photons by dust and by H 2 molecules are comparable. 3. By comparing the widths of H 2 filaments with the corresponding widths of the cloud edges in the z-band, we estimated the wavelength-dependence of the dust extinction in the molecular clouds comprising the NW PDR of NGC Assuming that the extinction follows the R V -dependent relationship developed

180 165 by CCM, we found a value of R V By comparing the widths of the ERE filaments with the corresponding widths of the cloud edges in the z-band, we found the ERE filaments on average about 10% wider than the corresponding H 2 filaments. This indicates that the dust opacity encountered by photons that initiate the ERE is only slightly smaller than the dust opacity encountered by photons exciting the H 2 molecules. Given that the H 2 emission in NGC 7023 is excited by far-uv photons with λ > nm, the CCM-extinction curve of R V 5.6 identifies two regions of the spectrum where the dust opacity matches the required value, the range around the mid-uv extinction bump (250 nm to 189 nm) and the region shortward of 118 nm. 5. We used the observation by Darbon et al. (1999) that stars with T eff 7000 K do not excite ERE to eliminate the possibility that ERE is initiated by mid- UV photons in the 250 nm to 186 nm range. This leaves far-uv photons at wavelengths shortward of 118 nm ( 10.5 ev) as the ultimate source of ERE excitation. 6. We showed that all existing ERE models fail in the face of this new requirement. 7. With the examples of two environments where the ERE excitation is highly efficient and where the number of ERE photons has been determined, we demonstrated that the number of far-uv photons with energies in excess of 10.5 ev is insufficient to generate the number of observed ERE photons, assuming reasonable values for the photon conversion efficiencies.

181 We concluded that the ERE excitation must therefore be a two-step process. The first step requires E > 10.5 ev photons to produce the actual ERE carrier, probably by a process of photo-ionization or photo-dissociation of a suitable precursor. A modest far-uv flux would then suffice to maintain a population of ERE carriers thus created. The second step of the ERE excitation then requires efficient pumping of the ERE carrier by abundant optical/near-uv photons with energies above the ERE band-gap of 2 ev. 9. A possible ERE carrier must therefore be a system that (1) has an ionization potential in excess of 10.5 ev, that (2) exhibits strong absorption in the optical/near-uv spectral region, and that (3) is capable of efficient photoluminescence in the wavelength range of the ERE band. 10. We suggest that PAH dications with masses 500 amu appear to meet the first two of these three requirements. Laboratory experiments designed to study the possibility of fluorescence by PAH dications are needed to test their potential as ERE sources.

182 Chapter 6 Optical Emission Band Morphologies of the Red Rectangle Introduction The optical spectrum of the Red Rectangle (RR) nebula consists of several components of comparable intensities, and it is therefore possible to use narrow-band imaging to isolate the morphological substructures where particular emissions are dominant. i) Dust-scattered radiation: to first order, this spectrum is a somewhat bluer version of that of the more luminous star in the central binary system, HD 44179, a T eff 8250 K A-giant AGB star. ii) Blue luminescence (BL) band: this emission 1 To be submitted to The Astrophysical Journal Letters, July 2005, Authors: U. P. Vijh, A. N. Witt, D. G. York, V. V. Dwarkadas, B. E. Woodgate, & P. Palunas 167

183 168 band with a peak near 380 nm (FWHM 45 nm) has been attributed to fluorescence by neutral PAH molecules with C-atoms. (Vijh et al., 2004, 2005b) iii) Extended red emission (ERE): this band peaks near 670 nm (FWHM 180 nm) and it is the primary cause for the unusual red color of the RR. The carrier of the ERE is yet unidentified (Witt & Vijh, 2004). iv) Sharp red emission features: this is a set of relatively narrow emission bands due to as yet unidentified molecules. They appear in a wavelength range largely coincident with that of the ERE band (Van Winckel et al., 2002). It has also been suggested that some of these bands are emission counterparts of corresponding Diffuse Interstellar Bands (DIBs). Spectroscopic studies have demonstrated that the intensities of the emission bands ii), iii), and iv) are comparable to those of the underlying scattered light continuum.. This opens the possibility of constraining the physical state(s) of the respective emitters, because the almost-edge-on geometry of the bi-polar outflow allows one to associate specific physical conditions with regard to density, radiation field, and ionization equilibrium with specific nebular regions. 6.2 Observations and Reductions The data were obtained at the Apache Point Observatory on January 7, 8, 9, 10 and using the Goddard Fabry Perot on the 3.5-m telescope. Images of the RR were obtained using narrow-band filters centered at 3934 Å ( FWHM 27 Å), 4050 Å (FWHM 57 Å), 4596 Å (FWHM 57 Å), 5700 Å (FWHM 135 Å), 6400 Å (FWHM 135 Å). The etalon was used to obtain images centered at 5800 Å

184 169 and 5855 Å, each with a 15 Å effective width. The nights were partially cloudy and only relative fluxes can be ascertained from these images. Data reduction was carried out using standard IRAF tasks. All images were trimmed to a size of pixels or The darks, twilight flats and object frames were corrected using an average zero frame. The flats and object frames were then dark-corrected using a combined dark frame. The flat frames for each filter were obtained by moving 30 between successive exposures to avoid coherent addition of bright stars. These flat frames were then median-combined for each filter. Dark corrected object frames were then flat fielded using the combined twilight flat for each filter. Object exposures were taken in sets of five, in a pentagram pattern of pointings. Object frames with the best seeing were selected and median-combined using tasks imalign and imcombine. Table 6.1 shows the exposure times for each filter. Stars in the frames were used to register and shift the images: this worked for all filters except the 3934 Å filter. As no stars other than the central extended source were visible on these frames, the center of the extended source was used to shift and co-add the images. This approach was shown to be correct, as one of the brightest stars in the field now appeared in the co-added image albeit with extremely low counts. All images were further cropped to pixel or to best show the RR. Images were scaled using counts of bright stars before computing the ratio images discussed in the next section.

185 170 Table 6.1 Narrow-band observations of the RR nebula. Filter Filter width (Å) Exposure (s) Comment BL band , Scattered Scattered Scattered + ERE Sharp emission feature Sharp emission feature , ERE + Scattered 6.3 Results and Discussion Blue Luminescence in the Red Rectangle The two narrow-band images at 3934 Å and 4050 Å show the appearance of the RR in the wavelength region of the BL band. The former is much closer to the peak of the BL emission, while the latter covers the long-wavelength tail of the BL band and is therefore comparatively more dominated by the scattered light component of the spectrum. Figure 6-1 is an image of the RR with the 3934 filter superimposed with contours. The most remarkable feature of these images is the fact that the inner isophotes of the 3934 Å image are consistently elongated in the E-W direction, apparently coincident with the orientation of the disk surrounding HD A check on stellar images in the 3934 Å frames confirmed that this elongation is real and not the result of guiding errors in the individual 900 s exposures. This image confirms results from long-slit spectroscopy at CTIO (Vijh et al., 2005b) which showed that

186 171 the BL emission exhibits a strong preference for the parts of the RR that are located in the shadow of the disk. This is consistent with the identification of the BL as originating from fluorescence by small (3- to 4-ring) neutral PAH molecules. Such molecules are photo-excited by stellar photons of 4 ev energy, which can diffuse into the outer disk regions via scattering. The same molecules would be photo-ionized by 7.5 ev photons, which are present in regions of the RR with direct lines-of-sight to the central sources, e.g. the outflow cones. We believe, therefore, that we are looking at emission from the outer portions of the optically thick disk surrounding the central stars in the RR, i.e regions where the local radiation field causes the ionization equilibrium to be preferentially neutral. We note that the isophotes of the 4050 Å image, shown in Figure 6-2 are distinctly circular in the inner region of the image, in contrast to the elongated 3934 Å isophotes. This reflects the fact that dustscattered light is dominant here. The outer isophotes of the 4050 Å image become almost square, with the sides being parallel to the polar axis of the RR. While the BL image at 3934 Å shows slight intensity enhancements at the base of the outflow walls on the W-side of the nebula, no such enhancements are apparent in the 4050 Å image, despite the fact that this is a deeper exposure. The dimensions of the RR disk implied by the narrow-band imaging observations are in excellent agreement with the results of Bujarrabal et al. (2003) who mapped this disk in CO emission with the IRAM interferometer array. Their map shows the disk having an angular extent of 9. When de-convolved with their beam pattern, the inferred disk diameter is 5.2. The 3934 Å data show elongated isophotes over an extent of 6.5, which is in perfect agreement, taking into account the 1 seeing disk of the optical data.

187 172 A perplexing aspect of the 3934 Å image is the orientation of the disk (PA 90 ) compared to the orientation of the polar axis of the RR (PA 11 ). If the disk were exactly perpendicular to the polar axis, a PA of 101 would have been expected. However, HST images of the innermost portions of the circum-binary disk of the RR (Cohen et al., 2004, Fig. 3) also seem to indicate that the disk is not perpendicular to the nebular outflow axis, and in fact that it is oriented in E-W direction. If confirmed, we have independently determined the orientation of the RR disk.

188 Figure 6-1 Image at 3934 Å with overlaid contours showing elongated profiles along the disk tracing the BL distribution. 173

189 174 Figure 6-2 Image at 4050 Å with overlaid contours showing circular scattered light profiles Sharp Emission Feature at 5800 Å The 5700 Å image contains mainly dust-scattered nebular light and a contribution from ERE, mainly in the outlying regions. The narrow-band 5800 Å and 5855 Å images contain a major contribution from the 5797 Å and 5855 Å sharp emission features, in addition to scattered light and a slightly larger contribution of ERE com-

190 175 pared to the 5704 Å image. The 5700 Å image is slightly overexposed, leading to bleeding from the stellar image and to the appearance of diffraction spikes that are not visible on the other images. The 5800/5700 and the 5855/5700 ratio images, Figures 6-3 and 6-4, show that there is excess emission at 5800 Å and 5855 Å in two locations to either side of the central source, on a line approximately perpendicular to the nebular axis. The X-shaped dark markings on in the ratio images corresponding to the walls of the outflow cones are the result of a systematic displacement of the intensity maxima of the ERE and the sharp emission features. The narrow emission features reach peak intensities slightly inside the cones with respect to the ERE maxima (Schmidt & Witt, 1991). The rung-like structures inside the cone across the polar axis of the nebula also appear dark on these ratio images. This experiment should definitely be repeated with images of of comparable depth and neither overexposed at the center. With those conditions, it would be possible to derive significant quantitative information about the spatial variation of the 5800 Å band intensity.

191 Figure Å image divided by 5700 Å image. 176

192 177 Figure Å image divided by 5700 Å image ERE The 6400 Å image is dominated by the ERE. When compared to images at wavelengths < 5000 Å, this image demonstrates that the X-shaped structures (variably referred to as arms or whiskers in the RR literature) are a result of the ERE

193 178 band. We interpret these structures as the walls of the internally illuminated bi-polar outflow cavities, seen in projection. When divided by the 5700 Å continuum image, the resulting 6400/5700 ratio image, Figure 6-5 convincingly demonstrates that the ERE is preferentially emitted by material in and within the walls of the outflow cavities, and is extremely weak outside these regions. This is demonstrated even more clearly in the ratio image 6403/4050, Figure 6-6. This is basically a display of the ratio of ERE to dust-scattered radiation in the RR. We can interpret the relative confinement of the ERE to the outflow cavities and their walls in one of two ways: either the carrier of the ERE exists only in the outflow cavities and its walls, or the excitation requirements for the ERE restrict this emission process to regions which can receive the required photons from the central sources. Both of these ways can actually be combined in the light of the results of a recently completed investigation (Witt et al., 2005) of the ERE in the reflection nebula NGC This study concluded that the ERE results from a two-step process: Step 1 Creation of the ERE carrier. Ionization of the ERE carrier progenitor by photons with energies 10.5 ev < E < 13.6 ev creating the actual ERE carrier. This step can proceed only inside the outflow cavity and its walls with a direct line of sight to the central stars, in particular to the hot white dwarf companion of HD The latter object is the suspected source of the far-uv photons in the RR. Step 2 Pumping of the ERE carrier, followed by fluorescence. Excitation of the carrier ion by optical/near-uv photons from the bright A-giant HD followed

194 179 by fluorescence in the ERE band. While the latter photons permeate other parts of the nebula as well, they can excite the ERE only in those places, where the earlier ionization has created the carriers. Neutral PAHs, such as those seen in the RR disk shadow will also absorb these optical/near-uv photons and respond with blue fluorescence. If PAHs are also responsible for the ERE, the ionization potential constraint found in NGC 7023 suggests that PAHs are ionized to the di-cation stage before they luminesce in the ERE band. Quantum-chemical calculations of the electronic energy level structure of PAH di-cations show the presence of strong absorption bands shortward of 6000 Å. Whether or not PAH di-cations fluoresce in the ERE band (> 6000 Å) is currently subject of laboratory investigations.

195 Figure Å image divided by 5700 Å image. 180

196 181 Figure Å image divided by 4050 Å image. 6.4 Conclusions 1. The morphology of the BL and ERE emissions in the RR nebula are almost mutually exclusive. 2. As the total intensities from these two emissions when summed over the nebula

197 182 are of comparable magnitude, it is very suggestive to attribute them to different ionization stages of the same family of carriers such as PAH molecules. 3. We have presented the optical detection of the circum-binary disk of the RR in the light of the BL. 4. The sharp emission features at 5800 Å and 5855 Å have spatial distributions similar to the ERE, but they reach peak intensities inside the cones in front of the ERE and show excess emission in the disk. Lastly, since the BL most likely arises in small neutral PAH molecules, the detection of the disk in the light of the BL places in question earlier claims by Waters et al. (1998) that the disk in the Red Rectangle is made up of oxygen-rich material produced in an earlier mass-ejection epoch, while the bi-polar outflow lobes are carbon-rich matter produced during the more recent outflow.

198 Chapter 7 Spectral Characteristics of Blue Luminescence in the Red Rectangle 1 We report the successful resolution of the spectrum of the recently discovered blue luminescence in the Red Rectangle into a spectrum consisting of seven molecular emission bands, superimposed upon a broad continuum. The relative intensity of the component bands varies with position within the nebula, suggesting an origin in a set of several related molecular species. The band widths and positions resemble those of fluorescence spectra of neutral aromatic molecules with between three and seven carbon rings. The successful identification of the carrier molecules will require gas-phase fluorescence spectra of a suitable range of aromatic molecules, which avoid 1 Submitted to The Astrophysical Journal Letters June 2005, Authors: U. P. Vijh, A. N. Witt, D. G. York, J. C. Barentine, R. McMillan, T. P. Snow, & D..F. Malin 183

199 184 the solvent-dependent wavelength shifts affecting the great bulk of currently available laboratory spectra. 7.1 Introduction The Red Rectangle (RR) is a well-known proto-planetary nebula in which the mass-losing stellar source, HD 44179, is surrounded by an optically thick disk, seen almost edge-on from Earth. Constrained by the massive disk, the mass outflow from the central post-agb star has created a bi-polar nebula of exceptional symmetry (Cohen et al., 2004). The presence of the disk prevents direct star light from reaching the Earth while the nebular matter above and below the disk is being fully illuminated by the stellar radiation. Given that dust-scattered light is subject to a strongly forward-directed phase function, relatively little dust-scattered light from the RR is reaching the Earth. On the other hand, nebular emissions excited by the stellar illumination are radiated isotropically and are therefore comparatively intense and easily observable. For this reason and because the carbon-rich RR environment is actively forming dust and molecules, the RR has been a most productive laboratory for the discovery of optical emissions from these nebular constituents. Prominent examples are the broad extended red emission (ERE) (Cohen, 1975; Schmidt et al., 1980), the sharp molecular emission features near 580 nm (Schmidt & Witt, 1991; Van Winckel et al., 2002), and, most recently, the blue luminescence (BL) reported by Vijh et al. (2004, 2005b). Following the discovery of BL in the RR, Vijh et al. (2005a) also detected BL in several other ordinary reflection nebulae, albeit with

200 185 much lower intensities. This suggests that the BL carrier is a ubiquitous component of the interstellar medium and is not restricted to the particular environment of the RR. The identification of the carrier of the BL is currently still a matter of controversy. While Vijh et al. (2005b) have advanced several observational arguments which favor an explanation in terms of fluorescence by small, neutral polycyclic aromatic hydrocarbon molecules, Nayfeh et al. (2005) have proposed small silicon nanoparticles as the carrier of the BL. The main reason why it is difficult to decide among these two (and possibly other) alternatives rests with the method with which the initial detection of the BL spectrum was accomplished. The BL detection was achieved with the Balmer line-depth technique (Vijh et al., 2004), which permits the determination of the BL intensities in the presence of dominant dust-scattered light only at the wavelengths of the Balmer absorption lines. This results in a poorly resolved spectral profile with insufficient detail for a definitive identification. In this Letter we are reporting on a new observational effort to gain a more complete view of the BL spectrum in the RR and its variation with position in the nebula. We are taking advantage of the fact shown by Vijh et al. (2005b) that the gradient of the dust-scattered light in the RR with offset from the center is much steeper than the gradient in the BL, a direct consequence of the forwarddirected phase function for the former and the isotropic emission pattern for the latter. By obtaining deeply exposed nebular spectra at larger angular offsets we have succeeded for the first time in obtaining well-resolved BL spectra without any significant contribution from dust-scattered light.

201 186 In 7.2 we will summarize the observations presented in this paper. The results will be presented in 7.3, where we will compare BL spectra obtained at different locations within the RR and where we will relate these results to a deep blue image of the RR obtained with the 3.9-m Anglo-Australian Telescope to gain insight into the spatial distribution of the BL-emitting molecules. In 7.4 we will address the complex issue of identification of the BL carrier(s) and we will summarize our conclusions in Observations and Reductions The observations of the RR and HD were made at the Apache Point Observatory 3.5-m telescope, using the dual imaging spectrograph (DIS). Table 7.1 gives details of the observations. The DIS is a medium dispersion double spectrograph with two independent collimators and cameras for the blue and red sides. In this letter we report on the results from the blue observations only. The high-blue grating has 1200 lines mm 1 with a central wavelength of 440 nm and a linear dispersion of 0.62 Å pixel 1. The blue detector was a Marconi back illuminated CCD yielding a pixel scale of 0.42 pixel 1. A slit 2. 5 long and 1. 5 wide was used. All observations of the central star were carried out at the parallactic angle. Nebular observations with the slit passing through the star on Nov 26 were also carried out at the parallactic angle which translated to a mean PA of -24. On April 02, the nebula was observed with a slit at PA 90, at a 2. 5 southern offset to obtain tie-in data with previous observations made at CTIO (reported in Vijh et al., 2004, 2005b)).

202 187 Table 7.1 RR Observations at APO Date Object Slit Position Exposure (s) Nov 26, 04 RR a Par. angle Apr 02, 05 HD44179 Par. angle 2 120,2 60 Apr 02, 05 RR 2. 5 south, PA Data reductions were carried out with IRAF 2.12 EXPORT. Spectra of standard stars were obtained on both nights, but as the observing conditions were nonphotometric, spectra are not calibrated to absolute flux, only to relative flux. In contrast, the CTIO observations, with which we will compare these data, were obtained under photometric conditions and at a much lower airmass. The Nov. observations were corrected for differential atmospheric refraction by using the stellar observation as the trace. At CTIO, the airmass was low and as the slit used was wider (2. 5) differential refraction was not an issue. These observations were used to correct the data from the E-W slit for differential refraction to a limited degree. We extracted 20 spectra along each slit each covering of the nebula. The long-slit enabled simultaneous sky observation which was subtracted out. Figure 7-1 shows a high resolution HST image of the nebula superimposed with the two slits and shows where the extracted spectra shown in Figure 7-2 were obtained.

203 Figure 7-1 HST Red image with superimposed slit images at PA -24 and PA 90. The image is 30 30, north is to the top and east to the left. The boxes indicate apertures, spectra from which are shown in Figure 7-2. The sharp nebular structures in this image are primarily due to ERE. 188

204 Results The Band Structure of the BL In Figure 7-2, we illustrate the changing composition of the nebular light in the RR with increasing offset from the center by displaying three nebular spectra, extracted from the PA -24 slit data at offsets 1.89 south, 4.41 south, and south. These data cover a range in intensity of the nebular light of a factor of The presence of strong Balmer absorption lines in the top spectrum illustrates the dominance of dust-scattered light originating in the AIII star located in the central cavity of the RR disk. The second spectrum, down in intensity by a factor of about six, begins to show the increasing dominance of several broad emission bands with only a weak contribution from dust-scattered light, evidenced by the greatly reduced relative depth of the Balmer absorption lines. The lowest spectrum at an offset of (9.78 south, 4.36 east) shows only a spectrum of emission bands superimposed upon a broad continuum without any measurable contribution from scattered light. North of the star along the slit, a similar progression of spectra are seen as well.

205 190 Figure 7-2 Nebular spectra at different offsets from the star along the PA -24 slit. These illustrate the changing composition of the nebular light in the RR with increasing offset from the center. Vertical bars at the bottom indicate the positions of the bands in the BL spectrum tabulated in Table 7.2. In the last spectrum free of scattered light, we detect the presence of seven distinct bands. Their wavelengths and energies are listed in Table 7.2. We identify these seven bands as the components of the BL.

206 191 Table 7.2 Molecular Bands in the BL spectrum of the Red Rectangle Band# Peak Wavelength (nm) Energy (cm 1 ) , , , , , , ,495.2 With the peak wavelengths of the BL bands known, we again examine the RR spectrum at 1.89 south, and we notice that the presence of bands 3, 4, 6, and 7 is clearly recognizable by eye. The banded structure of the BL spectrum does, therefore, persist over almost the entire range of offsets along the PA-24 slit Variation of BL Spectra with Position in the RR The pure BL spectrum is observable only in the outer fringes of the RR; at positions closer to the nebular center we need to separate the scattered light component from the BL, which is possible with the present data only through the line-depth technique (Vijh et al., 2004). In order to obtain comparable data for all parts of the RR, we evaluate the BL intensities at the wavelengths of the Balmer lines in those spectra which are free of scattered light. In Figure 7-3 we display several BL spectra obtained during the present observing program at APO as well as during the previous observing program at CTIO (Vijh et al., 2004, 2005b). The spectra are normalized at Balmer Hη ( nm) to aid inter-comparison. Note that the spectral coverage

207 192 of the DIS at APO does not extend to as short a wavelength as the CTIO spectrograph. Principal uncertainties in the data stem from the uncertainty of placing the continuum level for the line-depth measurements and from differential atmospheric refraction, which affected the APO spectrum at (2.5 south; 4.41 west), and is corrected to only a limited degree. We will, therefore, only highlight certain qualitative trends apparent in these data. Figure 7-3 Several BL spectra obtained during the present observing program at APO as well as during the previous observing program at CTIO (Vijh et al., 2004, 2005b). The spectra are normalized at Balmer Hη ( nm) to aid inter-comparison. Vertical bars at the bottom indicate the positions of the bands in the BL spectrum tabulated in Table 7.2. We notice high BL intensities at wavelengths shortward of Hη at small southern but large western offsets, where the BL was initially detected. This suggest that the

208 193 emission in band #1 (see Tab. 7.2) is dominant in these locations. These positions are well outside the walls of the RR outflow cones, i.e. in the disk-shadow region. All spectra appear to exhibit a shoulder or secondary maximum in the nm wavelength range. There is also an increase in emission at longer wavelengths with increasing offset. Given the band-nature of the underlying BL spectrum, we interpret these trends in terms of varying relative strengths of the seven BL bands. This may be interpreted as observational support for a model that attributes the BL bands not to a single molecule but rather to a set of two or more molecular carriers. A shift to increasing emission in the longer-wavelengths bands would be consistent with an increasing molecular size with distance from the center Variation of BL Intensities with Position in the RR In Figure 7-4 we are comparing the band-integrated (360 nm nm) BL intensities along the PA-24 APO slit with the band-integrated intensities of dust-scattered radiation. For large offsets, where there is very little scattered light we show intensities calculated in two ways: un-corrected for scattered light, by integrating the nebular spectra, and two: over-correcting for scattered light by assuming all the light at Hβ is scattered. As observed previously (Vijh et al., 2004, 2005b; Cohen et al., 2004), the scattered-light halo of the RR dominates in the innermost parts of the nebula as a result of the strongly forward-directed phase function coupled with an embedded source. The BL varies with distance more gradually and extends to larger angular offsets than the scattered light.

209 Band-BL: line-depths Band-BL: sc over-corrected Band-BL: sc un-corrected Band-sc Intensities (erg cm -2 s -1 sr -1 ) offset (arcsec) Figure 7-4 Band-integrated BL and the scattered light distribution along the PA -24 slit. The central star s position is at 0 offset, negative offsets indicate positions in the southern part of the slit Morphology of BL in the RR In Figure 7-5 we display a deep blue image of the RR taken by D.F. Malin with the 3.9-m Anglo-Australian telescope. The effective spectral range covered by the plate/filter combination in this image is about 385 nm nm, and therefore includes six of the brightest BL bands. The slit positions at which nebular spectra are shown in Figure 7-3 are indicated as white boxes. The positions showing essentially pure BL band spectra without contributions from scattered light are located outside the circular central halo. This makes it very likely that the faint, fuzzy nebular extension

210 195 surrounding the central scattered light halo derives its luminosity primarily from BL. This extended distribution clearly lacks the sharply defined structure visible in the light of the ERE (Fig. 7-1). The BL and the ERE, while comparable in intensity, are not correlated in their spatial distributions. The ERE is enhanced strongly on and near the walls of the outflow cones of the RR and defines the distinctive X-shaped appearance of the RR in red-wavelength images of the RR shown in Figure 7-1. We believe that the fluorescing molecules of the BL are physically located outside the mantles of the RR outflow cones where they are protected from direct illumination by the central post-agb giant and its hot He-white dwarf companion. This would be consistent with the attribution of the BL to fluorescence by neutral molecules.

211 Figure 7-5 Deep blue image of the RR taken by D.F. Malin with the 3.9-m Anglo- Australian telescope. The image is 30 30, north is to the top and east to the left. White boxes indicate slit positions at which nebular spectra (shown in Fig. 7-3) have been extracted. 196