Astronomy

How can ionized emission line flux decrease as a function of increasing metallicity or abundance?

How can ionized emission line flux decrease as a function of increasing metallicity or abundance?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

The chemical evolution of galaxies is an important way to learn about their formation and stellar/gaseous constituents. Many galaxies show narrow emission lines at optical wavelengths (3500-9000 Angstroms) from ionized elements (e.g., [O II] at 3727 Angstroms) and recombination Balmer lines (e.g., H$alpha$ at 6563 Angstroms). The origin of the ionized emission line flux can depend on the amount of the element present in the galaxy, and also on the nature of the ionizing source (i.e., the oft-mentioned "shape of the ionizing spectrum").

I have a few questions that I cannot seem to find clear answers to in textbooks, publications and review papers. My main reference for the following facts/questions is Kewley & Dopita 2002, ApJS, 142.

  1. What is the difference between metallicity and abundance, and what impact would these two parameters have on the ionized emission line flux from a galaxy (assuming that the ionizing spectrum is kept constant)?

  2. There are claims that some ionized lines (e.g., the auroral line [O III] at 4363 Angstroms) would become much weaker as the metallicity increases. Why? I would think that higher metallicity means more Oxygen and, assuming we keep the ionization spectrum fixed, more likely for more [O III] to exist. Should this not increase the [O III] 4363AA flux?

  3. Similarly, some excitation line ratios (e.g., ([O II] 3727 + [O III] 5007)/H$eta$, commonly known as $R_{23}$) are known to decrease with increasing abundance (measured via log(O/H)+12). Again, why? I would think if the metallicity or oxygen abundance is higher, then there is an increased chance of getting more [O II] and [O III] emission. Also, assuming that metallicity is higher, the amount of hydrogen should decrease, and this would decrease the $Heta$ flux in the $R_{23}$ ratio above, thus increasing the ratio at high abundance.


Metallicity and abundance

Metallicity

Without specifying a given metal, the term "metallicity" - abbreviated $Z$ - usually refers to the total metallicity of all elements, i.e. the mass fraction of all metals to the total mass of some ensemble of elements, e.g. a star, a cloud of gas, a galaxy, etc. (as usual, the term "metal" refers to all elements that are not hydrogen or helium). For instance, the mass of all metals in the Sun, divided by the Sun's mass, is 0.02: $$ Z_odot equiv frac{M_mathrm{C} + M_mathrm{N} + M_mathrm{O} + ldots}{M_odot} = 0.02. $$

Sometimes we talks about the metallicity of a given element, e.g. oxygen. The mass fraction of oxygen in the Sun is 0.005 (i.e. oxygen comprises 1/4 of all metals by mass), so we say $Z_mathrm{O} = 0.005$.

Unfortunately it is not uncommon to implicitly talk about the metallicity of an object, divided by Solar metallicity, such that a galaxy which has one-tenth of Sun's metallicity is said to have $Z=0.1$, rather than $Z=0.002$.

Abundance

The term "abundance" is only used for a single element. It basically expresses the same thing as metallicity, and is often used interchangeably, but is expressed in terms of the number $N$ of element nuclei, and as the ratio not to all nuclei but to hydrogen nuclei. For wacky historical reasons, we also take the logarithm and add a factor of 12. Taking again oxygen as an example, the mass fraction of 0.005 corresponds to a nuclei fraction of roughly $5 imes10^{-4}$, so we say that the abundance of oxygen is (e.g. Grevesse (2009)) $$ A(mathrm{O}) equiv log left( frac{N_mathrm{O}}{N_mathrm{H}} ight) + 12 = 8.7. $$

Metallicity of a given species vs. total metallicity

In general, the ratio of a given element to all metals is roughly constant. That is, various elements are produced by stars approximately by the same amount. But various processes may cause elements to exist in various forms. For instance, metals deplete to dust, but some elements tend not to form dust, e.g. Zn. For this reason, Zn is a better proxy of the total metallicity than, e.g. Mg, since half of the Mg may be locked up in dust.

Metals increases cooling

Elements also appear in various excitation states, which depend on various processes. The lines you mention, [O II] and [O III], arise from collisionally ionized oxygen, which subsequently recombines (in my first answer I wrote, wrongly, that it was excited), and thus depend on the temperature of the gas. As the metallicity of the gas in a galaxy increases, the ratio of the intensity of these lines to that of hydrogen lines (e.g. H$eta$) first increases, as expected. However, the increased metallicity also allows the gas to cool more efficiently. The reason is that metals have many levels through which electron can "cascade" down. If the electron recombines to the level where it was before, a photon of the same energy will be emitted, which itself may radiatively ionize another atom. But the many levels in metals makes de-excitation to intermediate level more probable, such that the electron cascades down, emitting several low-energy (infrared) photons, which are incapable of ionizing atoms and thus escape. The result is that energy leaves the system, i.e. the system is cooled.

This in turn means that, above a certain metallicity threshold - which is specific to a given species - the abundance of the collisionally excited lines begin to decrease. The following figure is taken from Stasińska (2002), and shows the turnover for the two oxygen lines:

This means that measuring the metallicity of a single species in general gives two solutions for the total metallicity. Luckily, as the turnover is different for different elements, measuring the metallicity for several species can constrain the total metallicity.


Frontiers in Astronomyand Space Sciences

The editor and reviewers' affiliations are the latest provided on their Loop research profiles and may not reflect their situation at the time of review.


  • Download Article
    • Download PDF
    • ReadCube
    • EPUB
    • XML (NLM)
    • Supplementary
      Material
    • EndNote
    • Reference Manager
    • Simple TEXT file
    • BibTex


    SHARE ON

    How can ionized emission line flux decrease as a function of increasing metallicity or abundance? - Astronomy

    Having described the ways that stellar population models are being made, with some caveats about their capabilities, I will give some examples of their use in this sectionFrole, indicating a number of issues that one has to be aware of when applying them. I will first discuss stellar population analysis on colors, and then on line indices or continuous spectra.

    Determining both an age and a metallicity of a galaxy, or even of an SSP, is tougher than it seems. Galaxy colors become redder as the galaxy ages, since more stars move to the giant branch, and also for increasing metallicities, since the effective temperatures of most stars decrease because of increasing opacities in the stellar photosphere. Colors and many line strengths in the optical basically depend on the temperature of the main sequence turnoff. The effects of increasing the age can be compensated for many observables by decreasing the metallicity. Worthey (1994) estimated that a factor of 3 increase in metallicity corresponds to a factor of 2 in age when using optical colors as age indicators, the so-called 2/3 rule. Optical colours are notoriously degenerated (see Fig. 1.7 in the red part of the diagram).

    There are, however, ways to break the degeneracy. Younger stellar populations with higher metallicities have a bluer contribution from the main sequence turnoff, and a redder one from the RGB. By using an optical color, together with a color that has a much higher relative sensitivity to the turnoff stars the degeneracy can be broken. This would be the case using colors such as UV - V, or Balmer line indices such as H or H. Equivalently, a combination of an optical color or line index which is strongly sensitive to the contributions of very cool giants would also break the degeneracy. Such colors would be e.g. V - K or J - K, or the CO index at 2.3 µm.

    Both methods are being applied. For spectra covering only a small range in wavelength, very sophisticated indices have been developed maximizing age-sensitivity while minimizing the sensitivity to metallicity (e.g. Vazdekis & Arimoto 1999). In general, assuming that the stars in a galaxy are not coeval, a blue spectrum will give a different mean age than a red spectrum, since colors/indices in the blue will be more sensitive to the younger stars etc. This is the so-called luminosity weighting of stellar populations (I prefer not to use the term light weighting, since this has other associations, in e.g. material sciences). When applying stellar population synthesis codes, one should always realize that one's results have been weighted with the luminosity of the stars, implying that the brightest stars give the impression to be more important than they really are, when one weights according to mass, the natural choice. As an example, in the UV, sometimes 90% of the light in a cluster is coming from one star (Landsman et al. 1998) (see Fig. 1.8). This means that the mass-weighted age of that cluster could in principle be very old, while the luminosity weighted value is close the value for that star, i.e. young. For the interpretation of galaxy ages this distinction between mass and luminosity weighting is particularly important.

    Figure 1.8. UIT image at

    In Chapter 1.3 we have found out that stellar populations in globular clusters are SSPs, and that populations in galaxies can be considered as linear combinations of SSPs. Recently, however, we have learned that the first assumption does not always hold. For a while it has been known that Cen, which up to now was considered to be a globular cluster, shows a spread in metallicity and possibly also age (Norris & Da Costa 1995). Conservative people could maintain for another 10 years that globular clusters have a single metallicity, by claiming that Cen is a galaxy, until recently Piotto et al. (2007), see Fig. 1.9, discovered multiple main sequences in the globular cluster NGC 2808. At the same time Mackey & Broby Nielsen found multiple main sequences in the LMC cluster NGC 1846. More clusters have been found later showing similar effects (e.g. Milone et al. 2008, Mackey et al. 2008). It is not clear yet what the reason is of these multiple branches. It could be that the He (or CNO) abundance is different, but also there might be a difference in age/metallicity. Spectroscopic studies here will have to show what really is happening.

    Color-magnitude diagrams can not only be used for globular clusters but also for galaxies in the rest of the Local Group. With HST it is possible to resolve individual stars below the Main Sequence Turnoff, and this way obtain exquisite star formation histories. In Fig. 1.10 I have reproduced a figure from the review of Tolstoy et al. (2009) with star formation histories of 3 dwarfs. An earlier, also excellent review, is by Mateo (1998). The star formation histories show that there are large variations between the galaxies of the local group, even between galaxies that have the same morphological classification (M32, NGC 205 and NGC 185). There are no galaxies for which we can exclude the presence of an underlying old population. Radial gradients in the populations of individual galaxies are seen as well. As mentioned before, more information about the abundance ratios in individual stars, giving information about star formation timescales, can be obtained from spectroscopy of bright giants in these galaxies.

    When one goes further away, one can only resolve stars on the Red Giant Branch and beyond. One can obtain the spread in metallicity, e.g. in Centaurus A (Harris et al. 1999), and the galaxies in the ANGST survey (Dalcanton et al. 2009). A great application of counting the stars on the RGB is to use these star counts to make maps of the stellar density in the outer parts of galaxies. This way people have found huge low surface brightness features linking M31 with its companions, including M33, probably remains from encounters between these galaxies (Ibata et al. 2001, McConnachie et al. 2009). For the spiral galaxy NGC 300 Bland-Hawthorn et al. (2005) have been able to measure the stellar surface brightness profile to a distance of 10 effective radii from the galaxy center in this way.

    In Fig. 1.11 a closeup is given of an RGB image of the disk of NGC 891, a nearby edge-on galaxy. What is clear are the many bright stars in the disk of the galaxy. Above it, many red filaments are seen. They are dust-lanes, seen up to large distances from the plane. In the lectures by Daniela Calzetti (this volume) you can see a lot of material about this dust, and how it extincts the light behind it. In Fig. 1.11 for example, further study shows that the blue stars seen in the left bottom corner are found in front of most of the disk of the galaxy, which itself is barely seen because of the extinction. In Fig. 1.12 one can see that the extinction is usually associated to spiral arms, and that it can be present to large radii. Here the extinction in a spiral disk is seen in front of an elliptical galaxy.

    Dust extinction is found predominantly in spiral galaxies of type Sab-Sc (e.g. Giovanelli et al. 1994). It is generally associated to molecular gas, and is stronger in larger (higher metallicity) galaxies. The UV energy absorbed by the dust is re-radiated in the IR and submm, responsible for a large fraction of the emission at those wavelengths. As far as stellar population synthesis is concerned the most important effect is that it reddens the colors using the dust extinction law (e.g. Cardelli et al. 1989). Reddening is strongest in the blue, and almost non-existent red-ward of 2 µm. In our Galaxy, it is impossible to see the Galactic Center in the optical, because of more than 20 magnitudes of extinction. However, in the infrared, at 2 micron, the extinction is only about 2 mag, so that observations there are easily possible. The ratio of reddening of dust in various colors is very similar to the effect of metallicity (and even age). This means that by simply measuring two colors, one cannot correct for the effects of extinction. For that more colors, or a spectrum are needed. It is therefore also not easy to measure the extinction from colors in a spiral galaxy. If one wants to do this, one can e.g. measure the amount of extinction statistically, by looking at the dependence on inclination. The color of a galaxy without extinction should not depend on inclination, while for an inclined galaxy the path length through the dust is longer, and thus the extinction. In Peletier et al. (1995) this technique is used to show that many nearby spiral galaxies in the B-band are optically thick within their central effective radius, but not in their outer parts. Certainly in bulges extinction plays a large role in many galaxies, and people analyzing the colors of external bulges should take this into account (e.g. Balcells & Peletier 1994).

    In the absence of dust color-color diagrams consisting of an optical and an optical-IR (or IR) color, or, e.g., of a UV-Opt and an optical color, can be used to separate the effects of age and metallicity. This method is popular for globular clusters in nearby galaxies, for which high quality spectra are difficult to get. As can be seen in Fig. 1.13, it is important that accurate observational colors are available. Such studies can maybe explain the bi-modality in globular cluster colors (Ashman & Zepf 1992). Chies-Santos et al. (2012), for a sample of 14 early-type galaxies, found that, although the optical color distributions of the globular clusters are bimodal, this is not always the case in the infrared (z-K). The authors explain this results with a non-linear color-metallicity conversion, but clearly state that better data are required to confirm their results. This means that colors of globular clusters are not well understood. The, up to recently, firm conclusion that globular clusters have a bimodal metallicity distribution is now up for discussion.

    The same color combination is also often used for galaxy research. In Fig. 1.13 (right) HST data are shown of the inner parts of a sample of early-type spiral galaxies (Peletier et al. 1999), on top of a grid of SSP models. Central colors are indicated with red, filled symbols, while the colors at 0.5 bulge effective radii are indicated in open, blue circles. These latter colors are calculated on the minor axis of the bulge, on the side not obscured by the galaxy disk. Interesting to see from this plot is that most galaxy centers lie, often far, away from the model grid, indicating considerable amounts of extinction AV of often more than 1 magnitude. The blue points cluster mostly together. Although the exact position of the model grid should be taken with caution (color differences are probably more reliable than exact colors, because of systematic effects in the models), this diagram seems to indicate that the bulges in this sample are old (mostly around 8-9 Gyr) with metallicities somewhat below solar.

    The fact that one has to determine such detailed colors leads to another problem: the models. Up to recently, the spectrophotometric quality of the stellar population models has not been very high. In Sánchez-Blázquez et al. (2006) it is shown that the colors of stars in the MILES sample are consistent with their spectra within 1.5%, something which cannot be said from e.g. the Stelib library (Le Borgne et al. 2003), used for the Bruzual & Charlot (2003) models. Vazdekis and his group have extended the MILES library with a subset of good spectra from the Indo-US library (Valdes et al. 2004), with the aim of making a stellar library with wavelengths up to 9500Å with good flux calibration. In Ricciardelli et al. (2012) they show that, when fitting well-calibrated SDSS data, there are problems fitting the g-r vs. r-i distribution of galaxies for galaxies with high velocity dispersions. In this paper it is discussed that probably -enhanced stellar population models are needed to solve this issue.

    The main difference between a spectrum of a star and one of a composite system, such as a galaxy, is that the composite spectrum is the sum of many stellar spectra, weighted by their flux at the particular wavelength, and shifted by their individual radial velocities. These velocity shifts are not to be discarded. In a large galaxy stars move through one another with a velocity dispersion of about 300 km/s. This means that every line in the spectrum is broadened by a Gaussian with such a dispersion, which means that abundances can not be measured any more from narrow lines from single transitions, since those are all blended. Abundances have to be obtained by fitting stellar population models with given abundances to the galaxy spectra. Velocity broadening cannot be avoided, and we have to live with broadened lines. The most common way to measure line strengths in composite spectra is by using a system of line indices. These indices are defined by three pass-bands: a feature band, and two continuum bands, and measured as equivalent width: the surface (in Å ) under the spectrum that is normalized using the continuum on both sides (see Worthey et al. 1994 for definitions of the Lick/IDS system. In the Lick/IDS system 21 indices were defined to measure the strongest stellar features in the spectrum in the optical at a resolution of about 9Å (Worthey et al. 1994). 4 more indices (2 H and 2 H indices) were added in 1997 by Worthey & Ottaviani. These indices were used by Trager et al. (2000) to determine stellar population parameters from Lick/IDS indices of many nearby galaxies. Later-on, many more indices were added by Serven et al. (2005). Other indices are available in the literature. Rose (1994) defined several indices with a one-sided continuum, mainly in the blue. Cenarro et al. (2001) defined indices in the region of the Ca II IR triplet, sometimes with with multiple continuum regions. Normally an index becomes larger as the absorption line becomes stronger. Some lines, however, like the H and H lines in the Lick system, are situated in such crowded regions, that their continuum fluxes are affected by metal abundances, and that the line index sometimes can be negative, even though H or H is found in absorption.

    Lick indices are hard to measure. Not only do they require the observed (galaxy) spectrum to be convolved to exactly the right resolution, and do they require a rather uncertain correction for velocity broadening, they also need certain zero point corrections to make sure that the instrumental response of the observed spectrum has the same shape as the Lick/IDS in the 1980's, when the standard stars for the Lick system were observed. This is a tedious job, since the Lick system does not work with flux calibrated spectra, and requires that for every observational setup a number of Lick standards are observed. To improve the situation, a slightly modified line index system (LIS) has been defined by Vazdekis et al. (2010). It is based on the MILES stellar library, uses the same wavelength definitions as the Lick/IDS system, and is defined on flux calibrated spectra, so that it is much easier for people to use this backward compatible system. To make it possible to also use less broad indices, for e.g. globular clusters and dwarf galaxies, the LIS system has been defined for standard resolutions of 5, 8.4 and 14 Å FWHM.

    Although line indices are a good way to measure the strength of certain spectral features, there is, at present, no need any more to go through the indices, when comparing galaxy data with models, since one can directly fit the models to the data. Vazdekis (1999) already showed the power of this method, which dramatically can show regions of the spectrum where the models are inadequate (see Fig. 1.14). Of course, the stellar population models will have to be convolved with the correct LOSVD (Line of Sight Velocity Distribution), i.e. the broadening of the stars. The SAURON team (Sarzi et al. 2006) have applied full spectral fitting to Integral Field Spectroscopy of ellipticals and S0s, fitting the observed spectrum at every position in every galaxy with a set of SSP models, determining the LOSVD, and a best-fit stellar population model. They noticed that always large residuals occurred at the position of emission lines such as H and the [OIII] line at 5007Å. Realizing that they could at the same time also measure the velocity and broadening of the ionized gas, they then developed a method to fit at the same time velocity broadened SSP models and Gaussians representing the emission lines to the data. As a result, ionized gas could be found in 75% of the sample, much more than was previously known. This method is much more sensitive than e.g. methods that map emission lines using narrow-band filters. Of course, those latter methods can cover a larger area, at generally a higher spatial resolution.

    Separating absorption and emission is in particular important for lines which are at the same time important absorption and emission lines, such as the Balmer lines H and H. These lines, which are so important to determine ages of stellar populations, can easily be filled in by emission. Before full spectral fitting was possible assumptions were made that the H emission line strength was a constant fraction of the [OIII] 5007Å line. The maps of Sarzi et al. (2006, 2010) show, however, that this ratio varies from galaxy to galaxy. Methods like this are so powerful that Balmer absorption line strengths can be measured in spiral galaxies with strong emission lines (e.g. Falcón-Barroso et al. 2006, MacArthur et al. 2009).

    The most popular index-index diagram is the Mg b vs. H diagram. Here Mg b is mostly sensitive to metallicity, while H is mostly age-sensitive. Using this diagram it was found that massive galaxies have /Fe ratios higher than solar (Peletier 1989, Worthey et al. 1992). In Fig. 1.15 we show H against [MgFe50], a composite of the Lick indices Mg b and Fe 5015, from Kuntschner (2010). Here one sees how some relatively small differences between stellar population models can change the ages derived from these indices. Here the models of Schiavon (2007) and those of Thomas et al. (2003) are shown. The galaxies are from the SAURON sample of early-type galaxies, and are shown as lines, with a dot in the center. Using the models of Schiavon, the old galaxies (these are mostly the ellipticals and the massive S0's) are old everywhere, with metallicity decreasing outwards. If one uses the models of Thomas et al. (2003), the outer parts are older, with similar metallicity gradients. Note here that these ages are luminosity weighted ages, and that the younger galaxies probably consist of an older continuum with a young population superimposed.

    Measuring element abundance ratios from integrated spectra is much less straightforward than for individual stars. What we know is that [Mg/Fe] varies strongly as a function of galaxy velocity dispersion (a proxy of mass). We also know that -elements vary more as a function of than Fe (Sánchez-Blázquez et al. 2006). From a sample of galaxies in the Coma cluster Smith et al. (2009) and Graves & Schiavon (2008) claim that the abundance ratios Mg/Fe and Ca/Fe simultaneously decrease with Fe/H and increase with . These dependences can be explained by varying star formation time scales as a function of , and therefore different ratios of element enrichment by SN type II and Ia. For both C/Fe and N/Fe, no correlation with Fe/H is observed at fixed . This can be explained if these elements are produced primarily by low/intermediate mass stars, and hence on a similar time-scale to the Fe enrichment. The element abundance ratio trends with [Fe/H] are very similar to those in our Galaxy, which suggests a high degree of regularity in the chemical enrichment history of galaxies (Smith et al. 2009).

    The availability of full spectra makes it possible to recover the Star Formation History (SFH) in some detail. Here one can divide the efforts into 2 parts: efforts that fit the full spectrum, including UV and IR, using the far IR, originating mostly from dust emission, as well as the submm on one hand, and more detailed studies that determine the distribution of stars of various ages on the other. The most important result from the first kind of studies is the amount of hot, young stars, responsible for ionizing the gas end heating up the dust around them. For this work I refer to Daniela Calzetti's chapter in this book.

    As far as the analysis of photospheric light is concerned, there is a growing body of full spectrum fitting algorithms that are being developed for constraining and recovering the Star Formation History (e.g., MOPED - Panter et al. 2003 Starlight - Cid Fernandes et al. 2005 Steckmap - Ocvirk et al. 2006 Koleva et al. 2008). SED fitting works best for large wavelength ranges. It is based on a set of SSP models and an extinction curve, and fits at the same time the stellar population mix, the LOSVD and the amount of extinction.

    Using full spectrum fitting several independent bursts of star formation can be determined. The number of parameters that can be recovered from a spectrum depends strongly on the signal-to-noise ratio, wavelength coverage and presence or absence of a young population (Tojeiro et al. 2007). However, the results are strongly affected by the age-metallicity degeneracy, so interpreting the results is difficult. Also, there is a certain degeneracy between the number of components in the LOSVD, and the SFH. Some tests are shown in Ocvirk et al. (2008), who try to constrain at the same time the SFH and the LOSVD along the slit of the spiral galaxy NGC 4030. It is clear from this experiment that higher spectral resolution and S/N data is needed to obtain astrophysically reliable results which are not degenerate.

    On the other hand, such studies are very useful to determine whether a galaxy is well fit by a one-SSP model, and if this is not the case, what the relative mass fraction in the various components in the SFH is. That information is useful to understand the formation history, combining it with spatial information about e.g. interactions. I will give 2 examples here.

    The first is by Koleva et al. (2009). They derive star formation histories of dwarf ellipticals in the Fornax cluster using ESO-VLT data. To understand the formation of these galaxies, it is very important to know whether these star formation histories are extended or short-lived, and whether they are very diverse, as is the case in the Local Group. In the Fornax cluster the environment is different, with a stronger influence from the IGM. Usually the photometric images are featureless, so that little can be learned about the stellar populations from the morphological structure. The results show that the star formation histories are not very different from those in the Local Group, and vary from SSP-like to extended (Fig. 1.16).

    A second example is from MacArthur et al. (2009), who determined star formation histories in a sample of late-type spirals using long-slit high S/N Gemini/GMOS data. For these spirals, imaging already shows that they have composite stellar populations. This is confirmed and quantified by stellar population synthesis. The authors are able to derive star formation histories consisting of a number of logarithmically spaced SSPs. One of their most important conclusions is that, although young populations contribute a large fraction to the galaxy light of late-type bulges, in mass they are predominantly composed of old and metal rich SPs (at least a mass fraction of 80% ).

    In the previous sections we have discussed how to derive the star formation history of a stellar population at a given position in a galaxy. One should always remember that galaxies are morphological and dynamical entities, and that the stellar populations that one derives are the result of the formation and evolution of the galaxy, and therefore intimately related to the morphological/kinematical component that one studies. 2-dimensional spectroscopy is an ideal tool to connect stellar populations with morphology and dynamics. In the last decade many of the large galaxies in the nearby Universe have been studied using the SAURON instrument at La Palma. The NIR Integral Field Spectrograph Sinfoni at ESO's VLT is making a large impact in the field at galaxy formation at z

    2 (SINS survey - Förster-Schreiber et al. 2009). Many IFU surveys are being planned (e.g. CALIFA, Sánchez et al. 2012 VIRUS, etc.).

    The SAURON survey (de Zeeuw et al. 2002, Bacon et al. 2001) has shown that many early-type galaxies contain kinematically-defined central disks. H absorption maps often show that these disks contain stars that are younger than the stars in the main body of the galaxy. The connnection here between the stellar populations, the morphology and the kinematics shows us that these disks are formed later. Since most of the disks have angular momentum that has the same sign as that of the main galaxy, one thinks that the disks are formed from gas lost by stars in the galaxies themselves (Sarzi et al. 2006). Spiral galaxies, much more than elliptical galaxies, have several components, such as bulge, inner disk, outer disk, bars, rings, etc. Here studying the stellar populations together with morphology and dynamics much more enhances our understanding of all these components, and the galaxies as a whole. I will discuss here the inner regions of 2 spiral galaxies, in order of morphological type, from the SAURON study of Falcón-Barroso et al. (2006) and Peletier et al. (2007).

    The first one is the Sa galaxy NGC 3623 (Fig. 1.17). The inner regions of this galaxy mainly contain old stellar populations, as shown by the absorption line maps and confirmed by the unsharp masked HST image, an image which is a very good indicator of dust extinction. Since young stars are always accompanied by extinction, unsharp masked images are an efficient way to find younger stars. However, the presence of dust is not always sufficient to detect young stars. The radial velocity map shows an edge-on, rotating disk in the center, confirmed by a central drop in the velocity dispersion. This central disk contains old stellar populations, probably slightly more metal rich (see the age and metallicity maps).

    The next galaxy is also an Sab galaxy, NGC 4274 (Fig. 1.18). This galaxy seems not to be very different from the one before. The unsharp masked image shows a central spiral, which might be similar to the one in NGC 3623 (which is edge-on, so the spiral structure in the dust cannot be seen). This spiral is associated to a rotating feature, seen both in the velocity and the velocity dispersion maps. In both galaxies, ionized gas is present everywhere in the central regions (Falcón-Barroso et al. 2006). Different from NGC 3623, the stellar populations in the central disk in NGC 4274 are much younger than in the main body of the galaxy, as can be seen from the line strength maps, especially H absorption. If one now looks at the photometric decomposition (Peletier 2009), one sees that the part of the surface brightness profile of NGC 4274 that lies above the large exponential disk, i.e. the bulge, corresponds to the region of the inner disk, and is best fitted by a Sérsic profile with n = 1.3. In the case of NGC 3623 the so-defined bulge is much larger, and has a Sérsic index of n = 3.4. Kormendy & Kennicutt (2004) would call the bulge in NGC 4274 a pseudo-bulge, and the one in NGC 3623 a classical bulge. However, the comparison here shows that both objects are very similar, but that only the inner disk to elliptical bulge ratio in both galaxies is different. The study of other bulges in this sample shows that many early-type spiral galaxies contain central disks, with often young stellar populations in them.

    The study of the central stellar populations and dust can also be done very well using IRAC on the Spitzer Space Telescope. van der Wolk (2011) in his PhD Thesis presents color maps of these SAURON-selected spirals. Information about the ages of the stellar populations comes from the [3.6] - [4.5] maps (see Section 1.5.2). In Fig. 1.19 the [3.6] - [8.0] maps of the 2 galaxies are shown. These maps show the amount of warm dust (mainly due to PAHs). One can see that in NGC 3623 relatively little warm dust is present, consistent with the absorption line maps, while much more is present in NGC 4274.

    Spectra and colors of SSPs are fairly insensitive to the initial mass function (IMF), because most of the light comes from stars in a narrow mass interval around the mass of stars at the main sequence turnoff. On one hand this is good, because it allows modelers to produce predictions for the spectra of galaxies that are accurate at most wavelengths. However, the same effect makes it possible to hide a large amount of mass in the form of low mass stars in a stellar population, making the stellar mass-to-light ratio a badly constrained parameter. Colors and lines of galaxies can generally be fitted well with a Salpeter IMF (a power law function with x = 1.3, see above). However, the same observables can also be fitted with an IMF that flattens below a certain critical mass, e.g. the Chabrier (2003) IMF, which flattens off below 0.6 M , giving a M/L ratio which is a factor 2 lower.

    Until the end of the 1990's the uncertainties in the IMF were considered so important that estimates of stellar mass were rarely given. This changed with the influential paper by Bell & de Jong (2001), who showed that if one maximized the stellar mass in the disk when reproducing rotation curves of galaxies (the so-called maximum disk hypothesis) an IMF similar to the Salpeter IMF at the high-mass end with fewer low-mass stars, giving stellar M/L ratios 30% lower than the Salpeter value, was preferred. After this, it has become very common that stellar masses are given when fitting lines or colors of galaxies. In the important paper of Kauffmann et al. (2003), where stellar masses of many galaxies in the SDSS survey are calculated, the Kroupa (2001) IMF is used, a similar kind of IMF, and in only 2 sentences the authors mention that there can be systematic uncertainties in the derived stellar masses, as a result of the choice of the IMF.

    Although in the optical most features are only slightly sensitive to the IMF-slope, there are some, mainly in the infrared, which strongly depend on the dwarf to giant ratio, i.e. the IMF-slope. Examples are the Wing-Ford band at 0.99 µm, the Na I doublet at 8190 Å, and the Ca II IR triplet around 8600Å. These lines have been used by several authors to constrain the IMF-slope (e.g., Spinrad & Taylor 1971, Faber & French 1980, Carter et al. 1986, Schiavon et al. 2000, Cenarro et al. 2003), but the results have never been very conclusive. The most important reason for this is that telluric absorption lines make it hard to measure accurate line strengths in this region of the spectrum. The second reason is that it is not always straightforward to derive the IMF-slope from the observations. For example, in Fig. 1.20 Cenarro et al. showed, based on the strength of the Ca II IR triplet and a molecular TiO index, and on the anti-correlation between the strength of the Ca II IR triplet and the velocity dispersion, that the IMF slope in elliptical galaxies increases for larger galaxies. However, other solutions are possible, e.g., that the [Ca/Fe] abundance ratio becomes lower for more massive galaxies. One might also think about systematic errors in the stellar population models that are needed to establish the IMF-slope. For example, in the models that Cenarro et al. use, solar abundance ratios are used in the stellar evolutionary calculations, and the stellar library used mostly consists of stars in the solar neighborhood, which implies that here too the abundance ratios must be close to solar.

    Recently, van Dokkum & Conroy (2010, 2011), and Conroy & van Dokkum (2012) have revived this topic. Using new methods to better remove the atmospheric absorption lines and new models in the near-infrared, they present conclusions that the IMF-slope increases with increasing galaxy velocity dispersion (mass). For the largest galaxies the IMFs found are a bit steeper than Salpeter (x = 1.6). This implies that stellar masses inferred from stellar population analysis will have to be increased by a small factor, which will not be larger than 2. Although this is an important result, one should remember the caveat that such a result depends on the stellar population models, which for non-solar abundance ratios are not perfect yet. Similar remarks can be made about the recent paper of Ferreras et al. (2012), who confirm Conroy & van Dokkum's result using the Na doublet at 8200Å with stacked data of a large sample of SDSS galaxies, and of Smith et al. (2012) who use an area near the Wing-Ford band. Recently, Cappellari et al. (2012) claim that independent analysis based on stellar dynamical fits to 2-D kinematic fits to galaxies of the Atlas-3d survey confirms the IMF trends observed in stellar populations.

    If, when calculating stellar masses, one still doesn't want to depend on these estimates of the IMF slope, one can also use photometry or indices further to the infrared. For example, it is known that M/L ratios in the K-band vary little with stellar populations, since here the relative contribution to the light of dwarfs vs. giants is much smaller than in the optical. The same holds for the Spitzer [3.6] and [4.5] bands, which are still dominated by light from stars. Meidt et al. (2012) nicely show how stellar masses can be obtained from images in both these bands for spiral galaxies.

    Stellar population synthesis in the UV is less well developed, because of various reasons. First of all, the amount of data available is limited, since it all has to come from space. Secondly, interpretation is complicated, since a few hot stars can over-shine all other stars, making it very difficult to obtain information on the not so young stellar populations.

    Burstein et al. (1988) published a large number of IUE-spectra of early-type galaxies. Their main result was a relation between the 1550 - V color (1550 is here a passband with effective wavelength 1550Å ) and the optical Mg2 index. Massive galaxies with large Mg2 index have a very blue 1550 - V color. This effect, the so-called UV-upturn, is probably due to extreme horizontal branch stars, but can also have other reasons (see O'Connell 1999 and Yi 2008 for reviews). With new and better quality GALEX-data, Bureau et al. (2011) show that this effect is present only when the optical H index is low, which implies that from the optical spectrum there is no evidence for any young stellar populations (see Fig. 1.21.

    Very little has been done on the analysis of line strength indices. This is surprising, since the UV is particularly important for the analysis of high redshift spectra. Recently, Maraston et al. (2009) published some stellar population models based on the IUE-stellar library of Fanelli et al. (1992). A problem with this empirical library is, that its range in metallicity is small. However, in this region empirical stars are probably more reliable than synthetic spectra, due to difficulties treating the effect of stellar winds that affect the photospheric lines of massive stars. There are still considerable differences between using the Fanelli library and a high-resolution version of the Kurucz library of stellar spectra (Rodríguez-Merino et al. 2005) in the Maraston models.

    There are ongoing efforts to develop a stellar library from HST/STIS stars, providing higher S/N and higher resolution spectra than IUE covering a much larger parameter space (the NGSL library - Gregg et al. 2006). This library has not been incorporated into any stellar population models yet, although it has been characterized and stellar parameters have been homogenized in Koleva & Vazdekis (2012).

    Just like the UV, the near-IR has also not been studied very much. While broadband colors are predicted by many stellar population models, very few spectrophotometric models are available. The problem has been mentioned before. The NIR is dominated by evolved stellar populations, i.e., RGB and especially AGB, of which the number and lifetimes are not well known, since they are so short-lived that good statistics cannot be obtained from globular and open cluster HR-diagrams. Furthermore, AGB stars lose large amounts of mass, making their lifetimes and also their spectra uncertain. On top of that, they are highly variable. Spectrophotometric models at a resolution of

    1100 are available from Mouhcine & Lançon (2002). They are based on about 100 observed stars from Lançon & Wood (2000) for static luminous red stars, stars from Lançon & Mouhcine (2002) for oxygen rich and carbon rich LPVs, and the theoretical library of Lejeune et al. (1997, based on Kurucz models) in all other cases. Conroy & van Dokkum (2012) recently made some models using the IRTF library (Rayner et al. 2009, Cushing et al. 2005). At low resolution (

    50Å ), there are models from Maraston (2005) and Charlot & Bruzual (Version of 2007, unpublished), based on theoretical atmospheres, and only tested in the broad bands J, H and K. Maraston also presents some low resolution indices. The problem with the models at present is that only the broad band fluxes have been tested well using clusters and galaxies, but that detailed testing of line indices or narrow band fluxes is still lacking. For example, in a recent paper, Lyubenova et al. (2012) showed that globular clusters cannot be fitted by the models of Maraston (2005) models in the C2 - DCO diagram. C2 indicates the line strength of a feature at 1.77 µm (Maraston 2005), while DCO is an index measuring the strength of the CO band head at 2.29 µm (Mármol-Queralto et al. 2008). The problem here is a lack of Carbon stars in the models, stars of

    1 Gyr (Lançon et al. 1999). It indicates that making models of the TP-AGB phase is very difficult (see also Marigo 2008). The situation might be improving soon. Better data of nearby galaxies and clusters are becoming available (e.g. Lyubenova et al. 2012, Silva et al. 2008, Mármol-Queralto et al. 2009). And better stellar libraries are expected (e.g., the X-Shooter library (Chen et al. 2011)). By comparing data with models, we will learn where the models should be improved, up to the moment that the NIR will give useful constraints to galaxy evolution theories.

    In Fig. 1.22 one sees the nearby dwarf elliptical NGC 205 in 2 bands. In the redder band, F814W, the giants can be distinguished much more easily from the underlying mass of fainter stars than in F555W. One can imagine that if this galaxy is placed at larger distances, one can see the individual giants up to larger distances in F814W. One can use the noise map, obtained after removing a smooth model of the galaxy, as a measure of the galaxy distance. Even more, since this noise map strongly depends on the number of bright giants and supergiants, one can use the noise characteristics, or the surface brightness fluctuations as a way to characterize the stellar populations in a galaxy.

    A review about surface brightness fluctuations as a stellar population indicator is given in Blakeslee (2009). It shows that the method can be used well for determining distances in early-type galaxies (giants or dwarfs), but that the use for stellar population analysis is still limited to the optical. In the near-IR there are considerable discrepancies between surface brightness fluctuations predicted by models and the observations (Lee et al. 2009). With the advent of new, large telescopes, this work will undoubtedly become more important in the future. *****


    Imaging FTS: a different approach to integral field spectroscopy.

    The vast majority of astronomical imaging spectrometers (or integral field spectrographs) are based on dispersive approaches. We present in this paper the most recent technical developments and some scientific results based on another approach, imaging Fourier transform spectroscopy (iFTS), which has been given a strong boost during the past decade, mostly because of enormous improvements in digital imaging capabilities, computer power, and servo control systems. A large number of research programs will certainlybenefit from an instrument capable of simultaneously obtaining spatially resolved spectra on extended areas (

    10 arcminutes) with a 100% filling factor, seeing-limited spatial resolution, and a flexible spectral resolution up to R

    [10.sup.4], and iFTS is very promising in that regard. Single-pixel FTS has been in regular use in commercial applications, remote sensing of the Earth's atmosphere, and astronomy, mostly in the infrared and sub-mm domain. However, by using appropriate optical configurations, fast readout CCD detectors, and especially improved metrology and servo systems, it is possible to transform the traditional one-pixel infrared FTS into a truly integral field spectrometer for the visible range.

    The most noteworthy scientific result from the use of the FTS approach in astronomy is the accurate measurement of the cosmic microwave background radiation spectral distribution, by the FIRAS instrument on the COBE satellite [1] and the COBRA rocket experiment [2]. On a completely different field, high resolution spectra of late-type stars were obtained on a regular basis with the FTS at Kitt Peak's Mayall telescope [3, 4]. A high-resolution FTS was one of the first instruments to be attached to the Canada-France-Hawaii Telescope and was widely used on a large variety of planetary and stellar programs [5, 6]. Combined with an imaging system in the early 1990s, it was renamed BEAR [7] and provided integral field spectra of a variety of objects such as planetary nebulae, massive star clusters, and star-forming regions in a 24-arcsecond field of view [8]. More recent examples include SPIRE-FTS, one of three instruments to fly on ESA's Herschel Space Observatory [9], a far-infrared FTS on the Japanese satellite AKARI, a mid-IR FTS (CIRS) on the Cassini spacecraft, and, for the near IR, PFS on Mars Express with a copy on Venus Express. The development of imaging FTS in astronomy was given a strong incentive during the early definition phases of the NGST (now known as the James Webb Space Telescope): astronomers supported by the three participating space agencies (NASA, ESA, and the Canadian Space Agency) presented studies of imaging FTS at the NGST Instrumentation meeting in Hyannis in 1999 [10-12]. None of these concepts however was included in the final instrument suite of the telescope. Our involvement in the iFTS endeavour is a direct consequence of this meeting. More recently, Boulanger et al. [13] proposed the design of a 1.2 m space telescope, H2EX, equipped with a wide-field imaging FTS specifically aimed at studying molecular hydrogen in the Universe. The advantages and disadvantages of the imaging FTS technique, as well as the relative merit of different approaches to 3D imagery, are discussed by Bennett [14]. An earlier version of the present paper, more complete in terms of technical explanations, is presented by Drissen et al. [15]. A recent review of the imaging FTS concept, with some historical perspective, a detailed comparison between the various imaging spectroscopy concepts and technical details not discussed in the present paper, is presented by Maillard et al. [16].

    An astronomical imaging Fourier transform spectrometer is basically a Michelson interferometer inserted into the collimated beam of an astronomical camera system, equipped with two detectors. The Michelson interferometer consists of a beamsplitter used to separate the incoming beam into two equal parts, two mirrors on which the halves of the original beam are reflected back, a moving mechanism to adjust the position and orientation of one of the mirrors (the other mirror is fixed), and a metrology system (IR laser and detector) to monitor the mirror alignment. All wavelengths from the field are simultaneously transmitted to either one or both of the interferometer outputs in which the array detectors sit. By moving one of its two mirrors, the interferometer thus configured therefore modulates the scene intensity between the two outputs instead of spectrally filtering it. This configuration results in a large light gathering power since no light is lost except through items common to any optical design: substrate transmission, coatings efficiency, and quantum efficiency of the detectors. All photons from the source can hence be recorded at each exposure provided that both complementary outputs of the interferometer are recorded. This requires a modification to the "standard" Michelson configuration in which half the light goes back to the source: the incoming light enters the interferometer at an angle allowing the two output beams to be physically separated. A CCD detector is then attached to each of the two output optics collecting the light from the interferometer (see Figure 6 in [15]).

    While in most FTSs targeting very bright sources the interferometer's mirror is moved at a regular, servoed speed, the weak signal from astronomical sources requires a step-scan approach. The interferogram cube is obtained through the acquisition of a series of short exposure images with the two CCDs. At each step, one of the two mirrors in the interferometer is moved by a very short distance (between 175 nm and

    5 [micro]m, depending on the spectral resolution and waveband chosen). The signal at each pixel is modulated, as a function of the mirror position, by a pattern which depends on the spectral content of the source each of the detectors' pixels is recording at each step a signal complementary to the corresponding pixel on the other detector. The vector composed of such a pixel recording is called an interferogram and is uniquely determined by the spectral content of the incoming light. The sum of the two images acquired at each step by the two detectors is then identical to a single image obtained with a "normal" camera. Spectral information for every pixel is recovered through a Discrete Fourier Transform (DFT or FFT) through the interferogram cube which can at any time during acquisition be turned into a spectral cube since each image contains information covering the whole waveband. The inclusion of additional exposures to an interferogram cube simply refines the meshing of the output spectra (spectral resolution). Thus, with an iFTS, spectral resolution is proportional to the total optical path difference (OPD) between the two arms of the interferometer scanned between the first and the last image of the data cube this OPD needs to be properly sampled through a series of mirror displacementsat predetermined sequential interference positions. Once the data cube has been obtained and its individual images corrected for instrumental artifacts (bias, Flatfield, like for any imager), the Fourier transform of each pixel's interferogram produces a wavelength-calibrated data cube. A by-product of the spectral data cubes is therefore a deep panchromatic image (within the limits of the filter used--see below) of the targets. Figure 1 summarizes the data acquisition with an iFTS and Figure 2 shows a tangible example.

    2.1. Technical Challenges. Like every imager, an iFTS must include high transmission optics producing high quality panchromatic images across the entire waveband covered by the instrument. But this does not translate alone into a good spectroscopic performance. In order to perform well on this aspect, a good modulation efficiency is also required. The performance of an FTS is thus characterized by its modulation efficiency (ME), that is, the capability of the interferometer to modulate the incident light:

    ME = I(modulated light)/I(incoming light). (1)

    The modulation efficiency can be viewed as an analog to the grating efficiency in dispersive spectrographs. In the worst case scenario, where the modulation efficiency is zero, the light from the source is recorded on the detector but the interferogram is a straight line and no spectral information can be extracted from it.

    This efficiency depends on a multitude of factors, the most technically challenging being the following.

    (1) The surface quality of the optical components in the interferometer (mirrors and beamsplitter): at a given wavelength, ME is lowered by a decreased surface quality it is therefore more and more difficult to obtain a good ME as we move from the infrared to the ultraviolet (most FTSs available today indeed work at infrared and sub-mm wavelengths) the number of reflections within the interferometer plays a major role in the global ME. Mirrors with a surface quality of [lambda]/20 (peak-to-valley) are commercially available for a reasonable price, but large [lambda]/30 mirrors must be custom made and are therefore much more expensive. In the case of a flat mirror design such as SpIOMM (see below), the ME at 350 nm doubles (from 35% to 70%) as the mirror and beamsplitter surface quality improves from [lambda]/20 to [lambda]/30 at 800 nm, the improvement is not as large (from 85% to 92%). Moreover, even if the mirror substrate is of high enough quality, any error in the coating deposit or any tension caused by the mechanical parts used to maintain the mirror within the interferometer can ruin the initial surface figure and dramatically reduce the modulation efficiency, especially in the blue part of the spectral range.

    (2) The mirror alignment and the stability of the OPD during an exposure: both crucially depend on the quality of the metrology and the servo system, which represents the highest technical challenge a visible-band iFTS faces. In order for the beams from the two arms to interfere properly, the two mirrors need to be very well aligned. The smallest deviation, in any direction, from the correct angle between the two mirrors reduces the spatial coherence (interference) of the two beams as they recombine. Again, this effect is more obvious at short wavelengths. A deviation of only 1.5 microradian from perfect alignment can decrease the ME by up to 25% at 350 nm. The DFT assumes that all data points of the interferogram vector are acquired at equidistant OPD intervals. Deviations from this assumption result in an increased noise level or artifacts in the resulting spectra. Even if the mirror steps are perfectly equidistant, a jitter of the OPD during an exposure, with a standard deviation as low as 10 nm (caused, e.g., by telescope vibrations transmitted through the structure of the instrument and uncompensated by the servo system) also decreases the ME by a significant amount, especially in the near UV. Therefore, monitoring the distance between the two mirrors as well as their alignment many thousand times per second, combined with a fast correction of any deviation, is required to ensure a constant, high modulation efficiency.

    Another factor that must be taken into account is the dead time due to the CCD readout. Since an interferogram is acquired through a series of a few hundred short exposure images, the time required to read the CCD (typically 10 s in the case of SpIOMM) lowers the global efficiency of the instrument. Recent improvement in CCD technology reduces this dead time to a minimum,

    2 s. Typical exposure times vary between 15 s for bright Galactic targets in the red to 120 s for galaxies in the blue. CCD readout noise (

    3-10 e) is usually unimportant as the photon noise from the source or the night sky background dominates.

    2.2. Use of Filters with an iFTS. A spectrum of an extended source covering the entire visible range (say, 350-700 nm) could be obtained with an iFTS without the use of a filter, the only constraints being the optics transmission and the detector's quantum efficiency (which, in this wavelength range, are both excellent). However, two properties of this instrument favor the use of filters: the need to properly sample the interferogram to reach the required spectral resolution, and the distributed photon noise.

    As mentioned above, the spectral resolution of a data cube is set by the maximum distance travelled by the moving mirror of the interferometer between the first and the last image of the cube. But one cannot simply take an image at a given mirror position and then move the mirror far from its initial position and hope to get a good resolution. The total optical path difference needs to be properly sampled, the step size being determined by the shortest wavelength and the total wavelength range to be covered by the cube. Using filters to restrict the total wavelength range and spectral folding (or aliasing) techniques allows increasing the mirror step length and thus the number of mirror steps for a given spectral resolution, at the expense of the total wavelength coverage. A simplistic example will help clarify this: say that we would like to use the bright lines H[beta], [OIII] 4959, 5007, H[alpha], [NII] 6548, 6584 and [SII] 6717, and 6731 to characterise an HII region. The minimum spectral resolution required in the red (R

    1200) is set by the need to separate the [SII] doublet and in the blue (R

    200) to separate H[beta] from [OIII]. Getting R = 1200 with an unfiltered cube covering the entire range allowed by a visible-band iFTS would require

    1500 mirror steps, which would limit the individual exposures to

    6 seconds for a 4-hour integration (taking the CCD readout time into account). Using two filters to isolate the blue and red line groups with the same spectral resolution requires only 120 steps and therefore allows much deeper individual images at each step for the same total time spent on the target.

    The second reason to use filters is to reduce photon noise, caused by the well-known multiplex disadvantage of the FTS [16], which, in some cases, counterbalances its clear multiplex advantage. With a dispersive spectrograph, photon noise at a given wavelength is only due to the total number of photons (from the source and the sky) at this particular wavelength. But with an iFTS (this is also the case for traditional, single-pixel FTS), photons from the entire wavelength range allowed by the optics are detected at each step (wavelengths are not filtered, but modulated) and distributed equally amongst all wavelengths after the Fourier transform. In an iFTS cube of an HII region, photon noise from a bright [OI] 5577 night-sky line affects the much fainter nebular [NII] 5755 or HeI 6678 lines, which is not the case for a spectrum of the same object obtained with a dispersive spectrograph. This also explains why an iFTS is at its best targeting emission-line objects, for which the continuum is rather low compared to the strengths of the emission lines. In some cases (very low surface brightness features such as the ionized stripping tails in galaxy clusters, distant Lya galaxies), the night sky continuum sets the detection limit.

    Within the limits imposed by the compromise between spectral coverage and resolution, once a filter has been selected the user can choose its preferred spectral resolution, from R = 1 (panchromatic image, which is anyway a byproduct of all data cubes) to the maximum limit imposed by the interferometer's architecture (typically a few times 104), and tailor it for each object.

    2.3. SpIOMM, an iFTS Prototype. In order to demonstrate the capabilities of a wide-field iFTS working in the visible band, our group has designed and built SpIOMM (Spectrometre Imageur de l'Observatoire du Mont Megantic) at Universite Laval [18] in close collaboration with ABB-Analytical (formerly Bomem), a Quebec City-based company, and the Institut National d'Optique (INO). The primary objective of any astronomical instrument development being to address a science case, the design of SpIOMM was optimized to feed our science projects on the interstellar medium, late stages of stellar evolution, star formation, and galaxy evolution and which could not be obtained by existing instruments. This instrument, attached to the 1.6 m telescope of the Observatoire du Mont-Megantic, is capable of obtaining seeing-limited, spatially resolved spectra of extended sources in filter-selected bandbases of the visible (350-900 nm) in a 127 x127 field of view with a spectral resolution R

    10-25000. It offers a very large contiguous field of view with 100% filling factor, resulting in millions of spectra per data cube. During its first few years of operations, only one output port was recorded, with a 1340 x 1300, 0.55"-pixel LN-cooled CCD, with a readout time of 8 seconds. We recently added a 2k x 2k CCD to its second output port.

    SpIOMM's modulation efficiency is very good (85%) in the red, as measured with nightly observations of a He-Ne laser data cube, also used to spectrally calibrate the science cubes. The optical quality of its mirror and beamsplitter (A/20, peak-to-valley), as well as the prototype metrology system implemented, does not allow us to obtain exquisite modulation efficiency in the near UV (

    25%), so most of our data cubes are obtained in the 450-700 nm range. Nevertheless, some cubes were obtained in the U band, as shown in Figure 3.

    Over the course of the past few years, we have obtained data cubes of HII regions [19], planetary nebulae [20], Wolf-Rayet bubbles, and unique 3D views of young supernova remnants [21,22], as well as a sample of nearby spiral galaxies (next section). Exposure times vary from 7 seconds per step for the observations of bright nebulae in the red filter to 90 seconds per step for the observations of galaxies in the blue filter. A typical data cube therefore requires a total exposure time between one and five hours. The technical and scientific progress of SpIOMM have been described in a series of SPIE papers to which the reader is referred for more details [23-26]. Applications of SpIOMM to the field of nearby galaxies are presented in the next sections.

    3. Nearby Spiral Galaxies: Emission Lines

    The new possibilities offered by iFTS to observe extragalactic HII regions will greatly improve our understanding of galaxy evolution. With a high spatial resolution, SpIOMM allows us to observe several hundred star forming regions simultaneously over all the structural components of a galaxy. The excellence of the statistics and the systematic characterization achieved for the HII regions will help to derive metallicity gradients and gain knowledge about the different mechanisms that drive star formation. In this paper, NGC628 data cubes are going to be used as a typical example of SpIOMMs capabilities.

    3.1. Observations and Data Reduction. Two filters have been used, a blue filter covering the wavelength domain from 475 to 515 nm and a red one from 650 to 680 nm. They include several useful emission lines: H[beta], [OIII][lambda][lambda]4959, 5007, [NIII][lambda][lambda]6548, 6584, H[alpha], HeI[lambda]6678, and [SII][lambda][lambda]6717, 6731. The spectral resolution (R = [lambda]/[DELTA][lambda]) of the data cubes is

    1600 in the red. A binning of the pixels was performed to achieve a pixel size of 1.07". We applied the basic CCD corrections (bias, darks, and flats) to the data cubes before correcting for sky transparency variations. A He-Ne laser was used for the wavelength calibration. The flux calibration was performed using data cubes of the standard star HD74721 in both filters. Finally, the average spectrum of the night sky was obtained from thousands of pixels around the galaxy and subtracted from the galaxy spectra (see Figure 4(a)).

    3.2. Spectral Analysis. The ionized gas emission lines are fitted with two different techniques. One is based on a pixel-by-pixel Gaussian fit of spectral lines using the fithi routine from the IDL MAMDLIB library, available from the following web site: http://www.cita.utoronto.ca/

    mamd/mamdlib.html. Figure 4(b) shows an example of the result for one pixel in an HII region.

    The other method consists in using the HIIphot code [27] to define the HII regions contours and then fit the global spectrum (sum of the pixels in the same region) of each HII regions using the fithi routine. Figure 4(c) shows an example of an HII region contours identified with HIIphot over the H[alpha] image. The parameters specified within HIIphot make sure that all the H[alpha] flux is included in the HII region contour map.

    At the time of writing, no correction has been applied yet for Balmer absorption lines. Long slit spectra obtained along the galactic radius are going to be used to correct for the impact of the absorption on H[beta] and H[alpha] lines. In the case of NGC628, absorption in H[beta] and H[alpha] is not negligible only in the very center of the galaxy (inner 1 kpc). For the moment with no correction for the absorption, the derived parameters using H[alpha] line are really close to reality and for the parameters using H[beta] line, we are only taking into account regions with radial position higher than 1 kpc. Detection limit for one pixel within an HII region is

    6x[10.sup.-17] erg [s.sup.-1] [cm.sup.-2] in the H[alpha] line. Figure 4(d) shows an example of a faint pixel spectrum.

    3.3. Results. The internal dust extinction is calculated using the theoretical ratio H[alpha]/H[beta]

    2.87 for HII regions at 10000 K with [R.sub.V] = 3.1 (the Milky Way extinction on the line of sight was first taken into account). The internal E(B - V) map of NGC628 is shown in Figure 5(a) using the pixel-by-pixel analysis. The total Ha flux derived with SpIOMM for NGC 628 is 1.39 x [10.sup.-11] erg [s.sup.-1] [cm.sup.-2]. This value is in good agreement with other studies. For example, with a smaller field of view, Sanchez et al. ([28] 6' x 6) and Kennicutt et al. ([29] 6.4' x 6.4) found a value of 1.13 x [10.sup.-11] erg [s.sup.-1] [cm.sup.-2] and 1.04 x [10.sup.-11] erg [s.sup.-1] [cm.sup.-2], respectively, and with a larger field of view, Hoopes et al. ([30] 29' x 29') obtained a value of 1.51 x [10.sup.-11] erg [s.sup.-1] [cm.sup.-2].

    The dust-corrected pixel-by-pixel H[alpha] flux map of NGC628 is shown in Figure 5(b). Its total dust-corrected H[alpha] luminosity is then 3.56 x [10.sup.41] erg [s.sup.-1] (assuming a distance of 9.7 Mpc). The global star formation rate of NGC628, using Kennicutt's [31] formula (SFR [[M.sub.[dot encircle]] [yr.sup.-1]] = 7.9 x [10.sup.-42] [L.sub.H[alpha]] [erg [s.sup.-1]]), is then SFR = 2.8 [M.sub.[dot encircle]] [yr.sup.-1]. This is somewhat higher than the 2.4 [M.sub.[dot encircle]] [yr.sup.-1] estimated by Sanchez et al. [28] but the difference is explained by the larger number of HII regions revealed by SpIOMM in the external regions of the galaxy.

    Using the N2 and the O3N2 metallicity indicators defined in [32] (PP04), we derived the metallicity of the HII regions identified with HIIphot as described above. As shown in Figure 6(a), a linear fit using the N2 indicator and all the HII regions identified gives the relation 12 + log(O/H) = 8.687-0.020 r[kpc]. A very similar relation, 12 + log(O/H) = 8.689 - 0.022 r[kpc], is obtained using the N2 indicator for a subsample of HII regions also characterized with the O3N2 indicator (the 4[sigma] detections of the OIII[lambda][5007] lines have been successful for only 134 regions). As shown in Figure 6(b), if the O3N2 indicator is used, we find a steeper gradient, 12 + log(0/H) = 8.780 - 0.033 r[kpc]. Contrary to the O3N2 indicator, the N2 indicator has a local maximum in its metallicity distribution [33] that, we suspect, may introduce more dispersion and uncertainty in the N2 gradient.

    A big advantage of an instrument like SpIOMM is the possibility to perform a detailed spatial analysis of the emission lines. The emission lines studied are affected by changes in the ionization parameter and the electron temperature [33] as we move away from the HII regions. Furthermore, a component in emission may come from the galaxy diffuse ionized gas (DIG [34]). The origin for the DIG is still complex (background ionization from older stellar populations, stellar winds, AGN, shocks). Among others, the DIG can contribute to low ionization species, like NII, SII, H[alpha], H[beta], and OII [35, 36]. In NGC628, we see systematic changes in line ratios as a function of the distance from an H[alpha] emission peak: in the case of a relatively low metallicity (

    8.4) region we find an increasing [NII][lambda]6584/H[alpha] ratio and a decreasing [OIII][lambda]5007/H[beta] ratio (see the examples in Figure 7 for NGC 628), while for a relatively high metallicity (

    8.8) region, these line ratio variations are inverted (see the examples in Figure 8 for NGC 628). These behaviors are to be discussed in terms of the DIG contribution and variation in the HII region conditions, but they clearly point out here to the importance of the spatial resolution when defining and characterizing HII regions and observing accurate metallicity gradient. As shown by the pixel-by-pixel O3N2 metallicity map of NGC 628 (Figure 9), if one uses a larger aperture to study the HII regions, it will have an impact on the abundance estimate for each region and also on the galaxy metallicity gradient. DIG effects on the characterization of extragalactic HII regions are going to be discussed in future publications.

    4. Elliptical Galaxies: Absorption Lines

    4.1. Absorption Lines Extraction. The vast majority of commercially available FTS use this technique to measure the strength of absorption lines in their target spectra. So nothing in the iFTS concept prevents it from obtaining spectra of continuum sources with absorption features. However, the multiplex disadvantage discussed above increases the noise level for a given spectral element compared to dispersive spectroscopy. This distributed noiserecorded in the interferogram (photon noise from the entire bandpass contributes to the recorded signal at each mirror step) is transferred to the spectrum after the FFT.

    Figure 10 shows a simplified example of distributed noise. Given two sources of equal total flux, one with emission and one with absorption, it is clear that the interferograms are utterly different even if the mean value is the same. The emission line shows large variation in its interferogram while they fade quickly for the absorption line. In this example, only a [square root of N] photon noise is applied and, since the mean flux is the same for both lines, the average level of noise is about the same. This in turn produces a level of noise in the spectra that is also about the same for both. It is then easy to see that if one were to extract the profile of both lines, the level of noise compared to the flux in the line would be much higher for the absorption feature. For the same mean flux in the bandwidth covered by the filter, spectra dominated by emission lines will show a greater SNR than spectra of continuum sources with absorption lines. Stellar spectra have been nevertheless obtained on a regular basis with SpIOMM some are shown in Figure 11. In fact, all SpIOMM data cubes include stars for which spectra can be extracted.

    It is then obvious that an iFTS is not the ideal instrument to observe individual stars. However, the wide field of view provided by SpIOMM allows observations of hundreds of stars simultaneously, as well as extended absorption line sources, which compensates for the multiplex disadvantage. Cubes of open clusters were obtained, and a technique was developed to optimize the extraction of the stellar interferograms, taking into account seeing variations during the cube acquisition results will be presented elsewhere.

    4.2. The Case ofM87. To quantify SpIOMM's ability to provide scientifically useful absorption spectra of an extended source, we targeted the giant elliptical galaxy M87 using SpIOMM's custom V filter (538-649 nm). Our primary goal was to compare SpIOMM's data with those obtained from direct imagery and long-slit spectroscopy. Using IRAF's STSDAS package, we first extracted luminosity profiles and isophotes from the panchromatic image obtained by combining all slices from the spectral cube (see Figure 12(a)). Our data being in excellent agreement with multiband imagery from Liu et al. [37], we then extracted spectra along isophotal lines at different galactocentric radii, shown in Figure 12(b). The spectra show many interesting features, the two most important being a molecular TiO band on the red side and the sodium doublet line (NaD) in the center near 5900 [Angstrom].

    In order to compare our data with those obtained by Davidge [17] with a long-slit spectrograph at the Canada-France-Hawaii 3.6 m Telescope, we extracted the NaD index as defined by Worthey and Ottaviani [38] as well as this line's equivalent width determined from a fitted Gaussian profile. These values are compared in Figure 13, which shows a clear negative gradient with increasing galactocentric radius. Our Lick NaD index values are slightly lower than Davidge's, on average, but this is mostly due to our lower spectral resolution (1.2 nm compared to 0.6 nm). Correction suggested by Vazdekis et al. [39] to compensate for this would add about 0.4 [Angstrom] to our values which would bring them closer to Davidge's but would not affect the slope of the gradient which is similar.

    The extraction of scientific data from absorption features in an extended object like a galaxy is proving to be feasible. Both visual and spectral data obtained on the source with the imaging FTS SpIOMM are in good agreement with the literature. Stability is of the essence during the observation to ensure the best possible SNR on the absorption features. This is particularly important for SITELLE, which should enable much more detailed and precise study of absorption features due to its more stable design compared with SpIOMM.

    5. Next Steps: SITELLE and SNAGS

    5.1. SITELLE. SpIOMM has been a wonderful prototype to work with and to learn from, special thanks to our privileged access to the well-equipped 1.6 m telescope at the Observatoire du Mont Megantic, where observing conditions can be harsh, especially in winter when temperatures can drop to -30[degrees]C. But the full advantages of the iFTS technology can only be attained with a more robust instrument installed on a large telescope with excellent sky conditions. Our team therefore designed and built SITELLE, an iFTS accepted as a guest instrument by the Canada-France-Hawaii Telescope (CFHT). All the lessons learned from our regular use of SpIOMM have been implemented in the design of SITELLE [40, 41] and its data reduction software, ORBS [42]. SITELLE's field of view and maximum spectral resolution (resp., 11' x 11' and R

    25000) are very similar to SpIOMM's, but its performance will be greatly enhanced, especially in the near UV, thanks to very high quality optics within the interferometer cavity ([lambda]/30), more robust metrology, servo mechanism, and structural stiffness leading to very low mirror jitter (-10 nm RMS), as well as high QE, and low read noise 2k x 2k CCDs by e2v. Thanks to four readout amplifiers on each CCD, readout time is

    3 s, thereby increasing the overall efficiency of the instrument. The overall throughput of SITELLE, including modulation efficiency and detectors' quantum efficiency, is shown in Figure 14. Spatial sampling is also improved (0.32" pixels) to match the excellent seeing provided by the CFHT. Sky brightness on Mauna Kea is also remarkably low, significantly improving the detectability of faint sources. If all these factors are taken into account, we estimate that SITELLE at the CFHT will be about 15 times more sensitive at 350 nm and 6 times more at 700 nm than SpIOMM at OMM.

    At the time of writing, SITELLE is in the final phases of integration and testing, for a planned delivery at the CFHT in early 2014. First light, commissioning, and science verification will occur soon after. As a guest instrument, SITELLE will be accessible to all CFHT partners. The long list of projects presented at a workshop organized in May 2013 (http://www.craq-astro.ca/sitelle/talk.php) has demonstrated the interest of the CFHT community (and beyond) for this instrument in areas as diverse as comets, star clusters, nearby galaxies, and high redshift Lyman- emitters. As demonstrated by Graham et al. [43] in a paper describing the rationale for equipping a space-based IR telescope with an iFTS, such an instrument is a very powerful tool to study distant galaxies in a relatively unbiased way. At very low flux levels, sky background sets the detection limits of emission-line sources visible-band sky background on Mauna Kea being very similar to that of IR background in space, SITELLE can be seen as the ground-based version of Graham's proposed iFTS. The power of SITELLE in this type of study is that it will sample the redshift space uniformly on a wide field, allowing unbiased spectroscopic redshift determination and line profile analysis on several galaxies per cube. In order to follow up on the SpIOMM work presented in this paper, the next section summarizes the plans for a survey of nearby galaxies, SNAGS.

    5.2. SNAGS. Powerful constraints on models of galactic evolution and the dynamical processes that transform galaxies are derived from homogeneous determinations of chemical abundances in individual gaseous nebulae, the distribution of their stellar populations in terms of age and metallicity, and the gaseous and stellar kinematics. With its efficient and versatile 3D spectroscopic capabilities over a large field-ofview, SITELLE offers an opportunity to study these evolution constraints in a significant number of large, nearby disk galaxies. Recent examples of using integral field capabilities to study nearby disk galaxies show the full potential of this technique to assess nebular abundances and stellar metallicity [28]. We are currently planning to a conduct a survey (SNAGS) of a sample of

    75 large (D > 37), nearby spirals with SITELLE at CFHT. Our main objectives are (1) to determine the spatial distribution of nebular abundances across their disk and assess radial, azimuthal, and local variations within different galaxy components (i.e., bars, arm and interarm regions, and outer disk), (2) to reconstruct their star formation history through stellar populations, and (3) to map the gaseous and stellar kinematics and study the diverse dynamical processes governing large-scale star formation and mixing within their disks.

    As demonstrated by other papers in this volume, SNAGS is not the first project aiming at integral field spectroscopy of a large sample of nearby galaxies. One of the earliest and most successful is the SAURON survey, mostly concentrating on early-type galaxies [44, 45]. But by far the most similar programs are CALIFA [46-48] and VENGA [49, 50]. The science objectives of SNAGS will be distinguished from those of the other two projects by the main advantages of its instrumental setting: significantly higher spatial resolution (limited only by CFHT's exceptional image quality and SITELLE's pixel size of 0.32") with 100% filling factor higher spectral resolution (R

    2000) which, combined with a very precise wavelength calibration inherent to the iFTS concept, will allow detailed kinematics studies on very small scales very wide field of view which, combined with SITELLE's high throughput from the near UV across the visible range and Mauna Kea's dark sky, will allow us to probe the outermost regions of galaxies.

    Our preliminary sample of galaxies includes objects of different morphologies, masses, and environment. For instance, several objects with bars will be observed to investigate the potential role of nonasymmetrical components in triggering/quenching star formation and in mixing chemical elements in galaxy disks. Other objects with star formation in their very outer disks (e.g., identified from GALEX observations) were selected to evaluate large-scale breaks in radial abundance gradients and to study chemical processes at the periphery of these disks [51]. SNAGS will obtain spectrophotometrically calibrated data cubes within three spectral ranges to cover the strong lines from [OII][lambda]3727 to [SII][lambda][lambda]6717, 6731. Nebular abundances will be determined using several line-ratio methods and calibrations (e.g., R23, N2Ha, and O3N2, see Kewley and Zahid [52] for a review) complementary observations on larger telescopes will be obtained for a smaller sample of galaxies to derive the abundances through "direct" auroral line measurements in order to evaluate the "indirect" methods. Gas kinematics on a small-scale will be studied using moderate resolution (R

    4000) for all three spectral ranges. Current studies so far (e.g., CALIFA and VENGA) have concentrated on global properties (abundance gradients) in galaxies. SNAGS will also assess small-scale variations caused by multiphase stellar winds, enrichment by starburst clusters in peculiar evolutionary stages, and dynamical processes. SITELLE is an ideal instrument to concentrate on these small-scale (<100 pc) variations and establish conditions under which they take place. A pilot study for SNAGS will be carried out during the science verification phase for the instrument planned for early 2014, with the survey itself starting later on after final commissioning.

    We have presented in this paper the iFTS as a viable approach to integral field spectroscopy. The regular use of SpIOMM and preliminary tests performed with SITELLE have demonstrated that the technical challenges imposed by the very stringent requirements on the optical quality of an iFTS components as well as its metrology and servo system have been overcome and now they allow us to reach a high modulation efficiency in the visible and the near UV.

    The science niche for the iFTS approach must take into account its main advantages (very wide field of view, high throughput, seeing-limited image quality, and flexible spectral resolution) and disadvantages (spectrally distributed noise, and necessary compromise between spectral coverage and resolution) compared with standard integral field spectrographs. Although the original, single-pixel FTS was classically known for its ability to obtain very high spectral resolution (R up to 105), the "sweet-spot" of the astronomical iFTS clearly sits in the observations of extended emission-line sources at low-to-moderate values of R. One can however still exploit the high spectral resolution capability of an iFTS by using narrow-band filters and use it for absorption line studies as well.

    The arrival of SITELLE on a 4 m class telescope (CFHT) under exceptional skies, where it will be used for a wide variety of science projects, will better define the iFTS niche and its capabilities.

    The authors declare that there is no conflict of interests regarding the publication of this paper.

    We would like to acknowledge financial contributions from the Canadian Foundation for Innovation, the Canadian Space Agency, the Natural Sciences and Engineering Council of Canada, the Fonds Quebecois de la Recherche sur la Nature et les Technologies, the Canada-France-Hawaii Telescope, and Universite Laval. We also thank Ghislain Turcotte, Bernard Malenfant, and Pierre-Luc Levesque for their help at the telescope and the CFHT team led by Marc Baril for their exquisite work on the detector system for SITELLE and its implementation at the telescope.

    [1] J. C. Mather, E. S. Cheng, R. E. Eplee Jr. et al., "A preliminary measurement of the cosmic microwave background spectrum by the cosmic background explorer (COBE) satellite," The Astrophysical Journal, vol. 354, pp. L37-L40, 1990.

    [2] H. P. Gush, M. Halpern, and E. H. Wishnow, "Rocket measurement of the cosmic-background-radiation mm-wave spectrum," Physical Review Letters, vol. 65, no. 5, pp. 537-540, 1990.

    [3] N. Scoville, D. N. B. Hall, S. G. Kleinmann, and S. F. Ridgway, "Detection of CO band emission in the Becklin-Neugebauer object," The Astrophysical Journal, vol. 232, pp. L121-L124, 1979.

    [4] S. T. Ridgway, D. F. Carbon, D. N. B. Hall, and J. Jewell, "An atlas of late-type stellar spectra, 2400-2778 inverse centimeters," The Astrophysical Journal, Supplement Series, vol. 54, pp. 177-209, 1984.

    [5] A. A. Chalabaev and J. P. Maillard, "Near-infrared spectroscopy of gamma cassiopeiae--constraints on the velocity field in the envelope," The Astrophysical Journal, vol. 294, pp. 640-645, 1985.

    [6] J. P. Maillard, J. Crovisier, T. Encrenaz, and M. Combes, "The spectrum of comet p/Halley between 0.9-MICRONS and 2.5MICRONS," Astronomy and Astrophysics, vol. 187, p. 398, 1987

    [7] J. P. Maillard and D. Simons, "Progress in telescope and instrumentation technologies," in Proceedings of the ESO Conference Workshop, p. 733, 1992.

    [8] T. Paumard, J.-P. Maillard, and M. Morris, "Kinematic and structural analysis of the minispiral in the galactic center from BEAR spectro-imagery," Astronomy and Astrophysics, vol. 426, no. 1, pp. 81-96, 2004.

    [9] R. Wu, E. T. Polehampton, M. Etxaluze et al., "Observing extended sources with the Herschel SPIRE fourier transform spectrometer," Astronomy and Astrophysics, vol. 556, article A116, 11 pages, 2013.

    [10] W. Posselt, J.-P. Maillard, and G. Wright, "NIRCAM-IFTS: imaging fourier transform spectrometer for NGST," in NGST Science and Technology Exposition, vol. 207 of Astronomical Society of the Pacific Conference Series, p. 303, 2000.

    [11] J. R. Graham, "IFIRS: an imaging Fourier transform spectrometer for the next generation space telescope," in NGST Science and Technology Exposition, vol. 207 of Astronomical Society of the Pacific Conference Series, p. 240, 2000.

    [12] S. L. Morris, J. Ouellette, A. Villemaire, F. Grandmont, and L. Moreau, "A Canadian IFTS for the NGST," in NGST Science and Technology Exposition, vol. 207 of Astronomical Society of the Pacific Conference Series, p. 276, 2000.

    [13] F. Boulanger, J. P Maillard, P Appleton et al., "The molecular hydrogen explorer H2EX," Experimental Astronomy, vol. 23, no. 1, pp. 277-302, 2008.

    [14] C. L. Bennett, "Critical comparison of 3-D imaging approaches (review)," in Imaging the Universe in Three Dimensions: Astrophysics with Advanced Multi-Wavelength Imaging Devices, vol. 195 of Astronomical Society of the Pacific Conference Series, p. 58, 2000.

    [15] L. Drissen, A. P Bernier, M. Charlebois, A. Alarie, F. Grandmont, and J. Mandar, "Imaging Fourier transform spectroscopy for astronomy," in Fourier Transforms--New Analytical Approaches and FTIR Strategies, G. Nikolic, Ed., chapter 23, InTech, Rijeka, Croatia, 2011.

    [16] J. P Maillard, L. Drissen, F. Grandmont, and S. Thibault, "Integral wide-field spectroscopy in astronomy: the imaging FTS solution," Experimental Astronomy, vol. 35, no. 3, pp. 527-559, 2013.

    [17] T. J. Davidge, "Spectroscopic gradients in bright elliptical galaxies," The Astronomical Journal,vol. 103, no. 5,pp. 1512-1525, 1992.

    [18] F. Grandmont, Developpement dun spectrometre imageur a transformee de Fourier pour lastronomie [Ph.D. thesis], Universite Laval, 2007

    [19] D. Lagrois, G. Joncas, and L. Drissen, "Diagnostic line ratios in the IC1805 optical gas complex," Monthly Notices of the Royal Astronomical Society, vol. 420, no. 3, pp. 2280-2299, 2012.

    [20] D. Lagrois, L. Drissen, and G. Joncas, Submitted to Monthly Notices of the Royal Astronomical Society.

    [21] M. Charlebois, L. Drissen, A.-P Bernier, F. Grandmont, and L. Binette, "A hyperspectral view of the crab nebula," The Astronomical Journal, vol. 139, no. 5, pp. 2083-2096, 2010.

    [22] A. Alarie, A. Bilodeau, and L. Drissen, Monthly Notices of the Royal Astronomical Society. In press.

    [23] F. Grandmont, L. Drissen, and G. Joncas, "Development of an imaging Fourier transform spectrometer for astronomy," in Specialized Optical Developments in Astronomy, vol. 4842 of Proceedings of SPIE, pp. 392-401, Waikoloa, Hawaii, USA, August 2002.

    [24] A.-P. Bernier, F. Grandmont, J.-F. Rochon, M. Charlebois, and L. Drissen, "First results and current development of SpIOMM: an imaging Fourier transform spectrometer for astronomy," in Ground-Based and Airborne Instrumentation for Astronomy, vol. 6269 of Proceedings of SPIE, p. 135, Orlando, Fla, USA, May 2006.

    [25] A.-P. Bernier, L. Drissen, P Charlebois et al., "Science results from the imaging Fourier transform spectrometer SpIOMM," in 2nd Ground-Based and Airborne Instrumentation for Astronomy, vol. 7014 of Proceedings of SPIE, Marseille, France, June 2008.

    [26] L. Drissen, A. Alarie, T. Martin et al., "New scientific results with SpIOMM: a testbed for CFHT's imaging Fourier transform spectrometer SITELLE," in 4th Ground-Based and Airborne Instrumentation for Astronomy, vol. 8446 of Proceedings of SPIE, Amsterdam, The Netherlands, July 2012.

    [27] D. A. Thilker, R. Braun, and R. A. M. Walterbos, "HIIphot: automated photometry of H II regions applied to M51," The Astronomical Journal, vol. 120, no. 6, pp. 3070-3087, 2000.

    [28] S. F. Sanchez, F. F. Rosales-Ortega, R. C. Kennicutt et al., "PPAK wide-field integral field spectroscopy of NGC 628--I. The largest spectroscopic mosaic on a single galaxy," Monthly Notices of the Royal Astronomical Society, vol. 410, no. 1, pp. 313-340, 2011.

    [29] R. C. Kennicutt, J. C. Lee, J. G. Funes, S. Sakai, and S. Akiyama, "An H[alpha] imaging survey of galaxies in the local 11 Mpc volume," The Astrophysical Journal, Supplement Series, vol. 178, no. 2, pp. 247-279, 2008.

    [30] C. G. Hoopes, R. A. M. Walterbos, and G. D. Bothun, "Farultraviolet and Ha imaging of nearby spiral galaxies: the OB stellar population in the diffuse ionized gas," The Astrophysical Journal Letters, vol. 559, no. 2, pp. 878-891, 2001.

    [31] R. C. Kennicutt, "Star formation in galaxies along the hubble sequence," Annual Review of Astronomy and Astrophysics, vol. 36, pp. 189-231, 1998.

    [32] M. Pettini and B. E. J. Pagel, "[O III]/[N II] as an abundance indicator at high redshift," Monthly Notices of the Royal Astronomical Society, vol. 348, no. 3, pp. L59-L63, 2004.

    [33] L. J. Kewley and M. A. Dopita, "Using strong lines to estimate abundances in extragalactic HII regions and starburst galaxies," The Astrophysical Journal, Supplement Series, vol. 142, no. 1, pp. 35-52, 2002.

    [34] G. A. Blanc, A. Heiderman, K. Gebhardt, N. J. Evans II, and J. Adams, "The spatially resolved star formation law from integral field spectroscopy: VIRUS-P observations of NGC 5194," The Astrophysical Journal, vol. 704, no. 1, pp. 842-862, 2009.

    [35] L. M. Haffner, R. J. Reynolds, and S. L. Tufte, "WHAM observations of Ha, [S II], and [N II] toward the Orion and perseus arms: probing the physical conditions of the warm ionized medium," The Astrophysical Journal, vol. 523, no. 1, pp. 223-233, 1999.

    [36] B. Otte, R. J. Reynolds, J. S. Gallagher III, and A. M. N. Ferguson, "Searching for additional heating: [O II] emission in the diffuse ionized gas of NGC 891, NGC 4631, and NGC 3079," The Astrophysical Journal, vol. 560, no. 1, pp. 207-221, 2001.

    [37] Y. Liu, X. Zhou, J. Ma, Y. Yang, J. Li, and J. Chen, "Deep surface photometry of M87 with 13 optical bands," The Astrophysical Journal, vol. 129, no. 6, pp. 2628-2635, 2005.

    [38] G. Worthey and D. L. Ottaviani, "Hy and Hd absorption features in stars and stellar populations," The Astrophysical Journal, Supplement Series, vol. 111, no. 2, pp. 377-386, 1997.

    [39] A. Vazdekis, P. Sanchez-Blazquez, J. Falcon-Barroso et al., "Evolutionary stellar population synthesis with MILES--I. The base models and a new line index system," Monthly Notices of the Royal Astronomical Society, vol. 404, no. 4, pp. 1639-1671, 2010.

    [40] L. Drissen, A.-P. Bernier, L. Rousseau-Nepton et al., "SITELLE: a wide-field imaging Fourier transform spectrometer for the Canada-France-Hawaii telescope," in 3rd Ground-Based and Airborne Instrumentation for Astronomy, vol. 7735 of Proceedings of SPIE, July 2010.

    [41] F. Grandmont, L. Drissen, J. Mandar, S. Thibault, and M. Baril, "Final design of SITELLE: a wide-field imaging Fourier transform spectrometer for the Canada-France-Hawaii telescope," in 4th Ground-Based and Airborne Instrumentation for Astronomy, vol. 8446 of Proceedings of SPIE, October 2012.

    [42] T. Martin, L. Drissen, and G. Joncas, "ORBS: a data reduction software for the imaging Fourier transform spectrometers SpIOMM and SITELLE," in 2nd Software and Cyberinfrastructure for Astronomy, vol. 8451 of Proceedings of the SPIE, Amsterdam, The Netherlands, July 2012.

    [43] J. R. Graham, M. Abrams, C. Bennett et al., "The performance and scientific rationale for an infrared imaging Fourier transform spectrograph on a large space telescope," Publications of the Astronomical Society of the Pacific, vol. 110, pp. 1205-1215, 1998.

    [44] R. Bacon, Y. Copin, G. Monnet et al., "The SAURON project--I. The panoramic integral-field spectrograph," Monthly Notices of the Royal Astronomical Society, vol. 326, no. 1, pp. 23-35, 2001.

    [45] H. Kuntschner, E. Emsellem, R. Bacon et al., "The SAURON project--XVII. Stellar population analysis of the absorption line strength maps of 48 early-type galaxies," Monthly Notices of the Royal Astronomical Society, vol. 408, no. 1, pp. 97-132, 2010.

    [46] E. Marmol-Queralto, S. F. Sanchez, R. A. Marino et al., "Integral field spectroscopy of a sample of nearby galaxies I. Sample, observations, and data reduction," Astronomy and Astrophysics, vol. 534, article A8, 2011.

    [47] S. Sanchez, F. F. Rosales-Ortega, R. A. Marino et al., "Integral field spectroscopy of a sample of nearby galaxies II. Properties of the HII regions," Astronomy and Astrophysics, vol. 546, article A2, 28 pages, 2012.

    [48] J. Iglesias-Paramo, J. M. Valchez, L. Galbany et al., "Aperture corrections for disk galaxy properties derived from the CALIFA survey: balmer emission lines in spiral galaxies," Astronomy and Astrophysics, vol. 553, article L7, 5 pages, 2013.

    [49] G. A. Blanc, A. Heiderman, K. Gebhardt, N. J. Evans II, and J. Adams, "The spatially resolved star formation law from integral field spectroscopy: VIRUS-P observations of NGC 5194," The Astrophysical Journal, vol. 704, no. 1, p. 842, 2009.

    [50] G. A. Blanc, T. Weinzirl, M. Song et al., "The VIRUS-P exploration of nearby galaxies (VENGA): survey design, data processing, and spectral analysis methods," The Astrophysical Journal, vol. 145, no. 5, p. 138, 2013.

    [51] F. Bresolin, "Clues on the rejuvenation of the S0 galaxy NGC 404 from the chemical abundance of its outer disk," The Astrophysical Journal, vol. 772, no. 2, p. L23, 2013.

    [52] L. Kewley and H. J. Zahid, "Strong-line methods for oxygen abundance determinations in giant HII regions and emission line galaxies: methods and uncertainties," Mapping Oxygen in the Universe, 2012, http://www.iac.es/congreso/oxygenmap/pages/electronic-presentations.php.

    Laurent Drissen, (1) Laurie Rousseau-Nepton, (1) Sebastien Lavoie, (1) Carmelle Robert, (1) Thomas Martin, (1) Pierre Martin, (2) Julie Mandar, (3) and Frederic Grandmont (3)

    (1) Departement de Physique, de Genie Physique et d'Optique, Universite Laval and Centre de Recherche en Astrophysique du Quebec (CRAQ), 1045 Avenue de la Medecine, Quebec, QC, Canada G1V 0A6

    (2) Department of Physics and Astronomy, University of Hawaii at Hilo, 200 W. Kawili Street, Hilo, HI 96720-4091, USA

    (3) ABB Analytical, 585 Boulevard CharestEst, Suite 300, Quebec, QC, Canada G1K9H4

    Correspondence should be addressed to Laurent Drissen [email protected]

    Received 15 November 2013 Accepted 24 February 2014 Published 22 April 2014


    3 Results of observations

    Fig. 1 shows the location of six spectrograms taken at four slit positions, designated PA102, PA187, PA304, and PA347 in accordance with their position angles. The two panels of the figure show the H α images of the galaxy’s star-forming regions. The upper image was taken with the 2.1-m KPNO telescope within the framework of the SINGS survey (Kennicutt et al., 2003) and the lower image was taken by the HST (application number 10522) and adopted from the Hubble Legacy Archive 1 1 1 http://hla.stsci.edu/ .

    Figure 1: The H α +continuum images of the Ho II galaxy (panel ‘a’ shows the image taken with the 2.1-m KPNO telescope and panel ‘b’, archival HST image (F658N/ACS WFC)). Panel ‘a’ gives the designations of all H ii regions discussed in this paper (denoted by the HSK numbers from the catalogue of Hodge et al. 1994 ), shows the position of the spectrograph slit PA347 used in our observations and the position of slit #2 in observations of Croxall et al. (2009) . Panel ‘b’ shows the positions of the spectrograph slits in our observations the short white line segments indicate the locations of the Croxall et al. (2009) spectrograms marked by their numbers in accordance with the above paper.

    In the north–south direction the PA187 spectrogram crosses the following H ii regions (named according to the catalogue of Hodge et al. 1994 ): HSK 71, HSK 73, HSK 70 (the nebula surrounding an ultraluminous X-ray source – ULX Holmberg II X-1), HSK 69, and HSK 67. The PA102 slit crosses the H ii regions HSK 32, HSK 41, HSK 45, HSK 56, HSK 59, and HSK 73. The PA347 slit covers the western chain of H ii regions: HSK 7, HSK 15, HSK 17, HSK 16, HSK 20, and HSK 26. The PA304 spectrogram is least informative, because it crosses faint H ii regions HSK 65, HSK 61, HSK 57, HSK 31, and HSK 25.

    Figure 2: Examples of spectra taken with different SCORPIO/SCORPIO-2 grisms. The top panel shows the region between − 21.4 and − 17.3 arcsec along the PA187 slit (the central part of the HSK73 nebula, which exhibits bright emission in the He ii λ 4686 line) observed with the VPHG1200B grism the middle panel shows the region between 40.7 and 45.0 arcsec along the PA102 slit (the HSK45 nebula) observed with the VPHG1200R grism, and the bottom panel shows the region between 94.6 and 98.2 along the PA347 slit (the HSK7 nebula) observed with the [email protected] grism.

    Fig. 2 shows examples of spectrograms taken with different grisms. The top panel shows the integrated spectrum of the part of the HSK73 H ii region crossed by the PA187 slit (positions ranging from − 21.4 to − 17.3 arcsec along the slit) taken with the VPHG1200B grism. It is the region where we found the brightest He ii λ 4686 emission line (see section 4 ). The middle panel shows the integrated spectrum of the eastern boundary of HSK45, which is the brightest nebula in the galaxy, taken with the VPHG1200R grism (the PA102 spectrogram, positions ranging from 40.7 to 45.0 arcsec). The bottom panel shows the integrated spectrum of the HSK7 region taken with the [email protected] grism (the PA347 spectrogram, positions ranging from 94.6 to 98.2 arcsec). The spectrum of this region is of the highest quality among all the spectra taken with this grism within the framework of this study (in the vicinity of the [O ii ] λ 3727 and [O iii ] λ 4363 lines its signal-to-noise ratio is as high as 6–7).

    Due to the wide spectral range of the [email protected] grism there is possible second order contamination on the spectra obtained with this grism at wavelengths redder than 7000 Å. But consideration of the continuum at this range (see, for example, bottom panel of the Fig. 2 ) shows that possible systematic errors of emission line intensity measurements is no more than 12% and lies in the uncertainties reported further.

    To ensure the homogeneity of the inferred gas metallicities in the galaxy, we also use the relative line intensities reported by Lee et al. (2003) and Croxall et al. (2009) . Lee et al. (2003) do not specify the locations of the spectra and only give the names of the H ii regions studied. Figure 1 shows the locations of the spectra taken by Croxall et al. (2009) (we adopt the coordinates from the author-corrected astro-ph electronic version of the paper). The spectrograms are numbered in accordance with Croxall et al. (2009) .

    Table 2 lists our estimated relative emission-line intensities in the individual H ii regions. We determined these intensities by integrating the spectrum along the part of the nebula crossed by the slit. The intensities are measured relative to the H β line intensity assuming that I ( H β ) = 100 .

    We obtained two spectrograms for the PA187 slit. One of them, like in the case of PA304 and PA347, covers the entire spectral range from the blue to the red part, whereas the other one contains only the blue part of the spectrum (see Table 1 ). The two spectrograms were taken on different nights with varying seeing using different instruments equipped with different CCDs. Hence a question naturally arises whether the two spectra are on the same intensity scale. We compared the line intensities in the blue part of the spectrum for each H ii region as estimated from both spectrograms and found their intensity scales to be proportional to each other. Hence albeit the line intensities measured by different spectrograms differ from each other, the ratios of the emission line intensities to the H β flux are the same for both spectra within the observational errors. That is why in the case of PA187 Table 2 lists relative line intensities measured either by the ‘wide-range’ spectrogram taken with the [email protected] grism or by the spectrogram taken with the VPHG1200B grism depending on the line wavelength. The intensities measured by spectra taken with the ‘wide-range’ [email protected] grism are used for lines located on the red side of [O iii ] λ 5007 , and those measured by the spectra taken with the VPHG1200B grism are used for lines in the blue part of the spectrum, where the sensitivity of the [email protected] grism decreases.

    The situation with PA102 is more complicated, because in this case two spectrograms have been taken for the blue and red parts of the spectrum and their spectral ranges do not overlap, preventing a similar comparison. Therefore to correctly estimate the ratios of line intensities on the red side of [O iii ] λ 5007 to H β , we calculated the ratios of the corresponding line intensities to H α and used the ‘theoretical’ I ( H α ) / I ( H β ) = 2.809 ratio for the average electron temperature T e = 13000 K (Osterbrock & Ferland, 2006) .

    The line intensity ratios listed in Table 2 are reddening corrected. We determined the E ( B − V ) colour index from the Balmer decrement using the theoretical intensity ratios I ( H α ) / I ( H β ) = 2.809 and I ( H γ ) / I ( H β ) = 0.4745 for the electron temperature T e = 13000 K (Osterbrock & Ferland, 2006) and the extinction curve of Cardelli et al. (1989) as parametrised by Fitzpatrick (1999) . For the PA102 slit we used only the I ( H γ ) / I ( H β ) ratio. The resulting E ( B − V ) extinction values are listed in Table 3.

    Our inferred colour excesses E ( B − V ) are greater than the E ( B − V ) = 0.06 ± 0.04 estimate reported by Croxall et al. (2009) . We obtained a similar extinction value E ( B − V ) = 0.03 towards Holmberg II from the extinction map based on the data of the infrared sky survey (Schlegel et al., 1998) .

    The mean colour index averaged over all our spectrograms is E ( B − V ) = 0.17 ± 0.08 . It differs from the estimate of Croxall et al. (2009) and this can be due to the use of different extinction laws. Our inferred E ( B − V ) value possibly corresponds to the stronger local extinction, because all the bright H ii regions studied are located in the direction of the maximum H i column density in the ‘walls’ of a giant cavity, and, most likely, are partially embedded in this dense supershell.

    3.1 Estimation of the electron density and temperature.

    Our electron density n e and electron temperature T e estimates for a number of H ii regions in Holmberg II have large uncertainties.

    Table 3 lists the electron densities n e for each H ii region derived from the [S ii ] λ 6717 / λ 6731 line intensity ratio. This line intensity ratio proved to be very close to the limiting value for low densities and therefore we could determine n e in a number of H ii regions only very approximately.

    We determined the electron temperatures in H ii regions in terms of the so-called two-zone model assuming that T e in the low and high ionization zones is the same for all ions whose radiation emerges from these regions. Correspondingly, we assume that the temperature in the high-ionization region (for the O 2 + and Ne 2 + ions) is equal to T e (O iii ), and the temperature in the low-ionization region (for the O + , N + , and S + ions) is equal to T e (O ii ). We also adopt the common extension of the model proposed by Garnett (1992) , which assumes that the temperature inside the region of the S 2 + and Ar 2 + emission is equal to the temperature T e (S iii ).

    We calculate the T e (O iii ) temperature from the [O iii ] ( λ 4959 + λ 5007 )/[O iii ] λ 4363 line intensity ratio (Osterbrock & Ferland, 2006) . However, because of the low intensity of the [O iii ] λ 4363 line we could determine T e (O iii ) accurately enough only for 10 H ii regions (including two different parts of the HSK73 nebula) by the relation of Pilyugin et al. (2010) :

    where Q 3 = I [ O I I I ] ( λ 4959 + λ 5007 ) / I [ O I I I ] λ 4363 and t = 10 − 4 T e ( O iii ).

    We estimate the T e (O ii ) temperature in low-ionization regions by the [O ii ] ( λ 3727 + λ 3729 )/[O ii ] ( λ 7320 + λ 7330 ) line intensity ratio using the following relation of Pilyugin et al. (2009) :

    where x = 10 − 4 n e t − 1 / 2 2 , t 2 = 10 − 4 T e (O ii ), and Q 2 = I [ O I I ] ( λ 3727 + λ 3729 ) / I [ O I I ] ( λ 7320 + λ 7330 ) .

    While determining T e (O ii ) we had to address a number of problems. First, in observations of nearby objects, such as Holmberg II, the [O ii ] λ 7320 + λ 7330 emission feature falls within the domain of strong atmospheric hydroxyl absorption lines, which reduce the accuracy of the intensity measurements for these lines and hence that of the inferred T e ( O ii ). Our failure to properly subtract the contribution of the air glow lines in the spectra of some H ii regions makes the corresponding [O ii ] λ 7320 + λ 7330 intensity estimates less reliable.

    Second, the sensitivity of the CCD used in half of the observations decreases sharply in the blue part of the spectrum. This prevented us from measuring the [O ii ] λ 3727 + λ 3729 lines accurately enough in the PA304 and PA347 spectra, making it impossible to calculate T e (O ii ). In the case of the PA102 and PA187 spectra we derived the temperature in the low-ionization regions only in four H ii regions, where the signal-to-noise ratio in the [O ii ] λ 7320 + λ 7330 lines was greater than 4.

    We could thus ‘directly’ determine the electron temperatures both in the low- and high-ionization regions only in HSK45, HSK67, HSK71, and HSK73. However, even in these nebulae the inferred T e (O ii ) estimates are not accurate and most likely represent the upper temperature limits for the corresponding regions. We therefore used empirical methods to infer the temperatures in low-ionization regions in the cases where we could determine T e (O iii ).

    Many methods are known for the empirical determination of electron temperature in low-ionization regions from the known T e (O iii ) temperature. Hägele et al. (2008) compared some of the methods used to estimate the temperature in the O + emission region. Pérez-Montero & Díaz (2003) pointed out that this temperature is strongly dependent not only on T e (O iii ), but also on the electron density n e . This imposes strong constraints on the applicability of the empirical T e (O ii ) on T e (O iii ) dependences. In particular, we did not use the dependence of T e (O ii ) on T e (O iii ) found by Pérez-Montero & Díaz (2003) for three electron density values, n e = 10 , 100 , and 500 cm − 3 , because of the above-mentioned low accuracy of the estimated n e .

    Stasińska (1980, 1990) proposed a method, which gained widespread popularity. Izotov et al. (2006) proposed to determine the temperature in regions with ionization higher and lower than in the O 2 + emission region using a technique depending on the gas metallicity. Furthermore, Pilyugin et al. (2009) proposed an empirical formula for estimating T e (O ii ) and T e (N ii ) from T e (O iii ). These authors point out that the T e (N ii ) value should be preferred for the low-ionization regions because of its lower dispersion. López-Sánchez et al. (2012) proposed a calibration based on the model spectra of H ii regions.

    We used all the above mentioned methods to determine the temperatures in the low-ionization regions. We found the calibration of López-Sánchez et al. (2012) to be the most consistent with the estimate determined from the spectra of four H ii regions, which represents an upper limit for the T e (O ii ) temperature in these nebulae. Hereafter throughout this paper we adopt the estimates obtained using the above technique obtained by the relation:

    The temperatures T e (O ii ) estimated by other methods are much higher than our inferred upper constraints for four H ii regions.

    The dependence of the temperature in the S 2 + emission region on T e (O iii ) has a much lower scatter than the corresponding dependence for low-ionization regions (see, e.g., Hägele et al. 2008 ). We estimated T e (S iii ) using the equation proposed by Izotov et al. (2006) for different H ii region metallicities:

    t 3 = − 1.085 + t × ( 2.320 − 0.420 t ) , f o r 12 + log ( O / H ) ≃ 7.2 , = − 1.276 + t × ( 2.645 − 0.546 t ) , f o r 12 + log ( O / H ) ≃ 7.6 , = 2.967 + t × ( − 2.261 + 1.605 t ) , f o r 12 + log ( O / H ) ≃ 8.2 ,

    where t 3 = 10 − 4 T e (S iii ), t = 10 − 4 T e (O iii ).

    Table 3 lists the adopted electron temperatures in the regions of different ionization.

    3.2 Estimation of the gas metallicity

    Currently, many methods are used to estimate the abundances of chemical elements and primarily that of oxygen, which determines the metallicity of the interstellar medium. The most popular is the so-called ‘direct’, or T e method, which allows elemental abundances to be estimated from the forbidden-line intensity ratios provided that the electron temperature T e is known in the region where the corresponding emission line forms. However, we cannot always use this method since we have direct spectra-based electron temperature estimates in the zones of different ionization solely for four H ii regions. Only in 10 regions the temperatures in low-ionization zones were estimated using empirical methods. We therefore used the T e method to estimate the relative abundances of the O + , O 2 + , N + , S + , S 2 + , Ar 2 + , and Ne 2 + ions in these regions wherever the signal-to-noise ratio for the corresponding ion lines was greater than 3.

    We estimated the relative ion abundances using the relations from studies based on modern atomic data: we used the equations from Pilyugin et al. (2010) to compute the abundances of O + and O 2 + ions, and those from Izotov et al. (2006) , to compute the N + , S + , S 2 + , Ar 2 + , and Ne 2 + ionic abundances. For the unobserved ionization stages we calculated the ionization correction factors ( I C F ) by the relations adopted from Izotov et al. (2006) , which allowed us to determine the abundances of oxygen, sulphur, argon, and neon.

    Among the popular metallicity determination methods there are those based on the intensity ratios of bright emission lines. These include the so-called ‘empirical’ methods calibrated by H ii regions with bona fide oxygen abundance estimates and methods based on theoretical photoionization models. In this paper we use six such methods:

    Figure 3: The H ii region HSK45. The central part of the figure shows the HST H α image together with the position of the PA102 slit. The left, right, and bottom panels show the distribution of relative emission-line intensities, electron temperature, and oxygen abundance, respectively, along the portion of the slit crossing the HSK45 region.

    The PT05 method (Pilyugin & Thuan, 2005) , where the oxygen abundance is fitted by a function of R 2 = I [ O I I ] λ 3727 + λ 3729 / I H β , R 3 = I [ O I I I ] λ 4959 + λ 5007 / I H β , R 23 = R 2 + R 3 , and the excitation parameter P = R 3 / ( R 3 + R 2 ) . It is one of the most widely used empirical methods with the applicability domain confined to the 0.55 < P < 1 interval

    The ONS method (Pilyugin et al., 2010) , which allows the oxygen and nitrogen abundances to be determined as a function of parameters R 2 , R 3 , N 2 = I [ N I I ] λ 6548 + λ 6583 / I H β , S 2 = I [ S I I ] λ 6717 + λ 6731 / I H β , and P

    The ON method (Pilyugin et al., 2010) , which is similar to the ONS method, but does not require the knowledge of S 2

    The NS method (Pilyugin & Mattsson, 2011) , which does not require [O ii ] λ 3727 + λ 3729 intensity measurements and allows the oxygen and nitrogen abundances to be determined as functions of parameters R 3 , N 2 , and S 2

    The PP04 method (Pettini & Pagel, 2004) , which allows the relative oxygen abundance to be determined from the parameter O 3 N 2 = log [ ( I [ O I I I ] λ 5007 / I H β ) / ( I [ N I I ] λ 6583 / I H α ) ] . This method is practically extinction independent it works in the − 1 < O 3 N 2 < 1.9 interval

    The KK04 method ( Kewley & Dopita 2002 with the new parametrisation by Kobulnicky & Kewley 2004 ) based on theoretical photoionization models, which can be used to determine the oxygen abundance as a function of parameter R 23 and ionization parameter q , which, in turn, can be obtained from the O 32 = I [ O I I I ] λ 4959 + λ 5007 / I [ O I I ] λ 3727 + λ 3729 line intensity ratio.

    All the above methods can be used to determine the oxygen abundances, 12 + log ( O / H ) in the H ii regions. According to the authors of original publications, these methods are accurate to about 0.1 d e x (the PP04 method is less accurate, its error is of about 0.2 − 0.25 d e x the ON, NS, and ONS methods have smaller errors – of about 0.075 d e x ). The ON, NS, and ONS methods can also be used to determine the nitrogen abundance and hence the log ( N / O ) abundance ratio, which is important for chemical evolution models of galaxies.

    The chemical abundances of the Holmberg II H ii regions were earlier estimated by Masegosa et al. (1991) Lee et al. (2003) , and Croxall et al. (2009) . Moustakas et al. (2010) summarised these results and reported the galaxy-averaged abundances and average abundances at different galactocentric distances. We use all these data to compare with our results.

    Croxall et al. (2009) and Lee et al. (2003) report relative line intensities, which we used to determine the elemental abundances by applying all the methods employed in this paper.

    Table 3 lists the resulting estimates of abundances of chemical elements and individual ions for the H ii regions based on our observations and line intensities reported by Croxall et al. (2009) and Lee et al. (2003) . Table 3 gives only the formal measurement errors, which do not include the uncertainties of the each method itself.

    Abundances obtained for similar slit positions based on our observations and those reported in previous studies mostly agree well with each other within the quoted observational errors. The oxygen abundances reported by Croxall et al. (2009) and Lee et al. (2003) slightly differ from those calculated in this study using the fluxes reported by the above authors. This discrepancy is due to the use of different atomic data. For some H ii regions, the abundances determined in this paper do not match those obtained by Croxall et al. (2009) and Lee et al. (2003) because we observed different parts of the nebulae, which show significant abundance variations (see Section 4.1 ).


    How can ionized emission line flux decrease as a function of increasing metallicity or abundance? - Astronomy

    In this section, we will concentrate on the impact of the physical input parameters on He abundance determinations. We will discuss the necessity of obtaining accurate line strengths and the limitations in doing so. We pay special attention to reddening as determined by H I line ratios. The uncertainties in this correction are particularly important as they feed into the uncertainty in all of the subsequent He I line strength determinations. We will also discuss determination of the electron temperature and density.

    With the advent of large format, linear CCD array detectors in the last decade, we are in the best position ever to obtain spectra of emission line objects with the quality and accuracy necessary for helium abundance measurements. While it may seem unnecessary to discuss the measurement of emission line strengths here, this work starts with the assumption that the spectra have been properly calibrated and that errors associated with that calibration have been taken into account. Targets and standard stars should both be observed close to the parallactic angle in order to minimize atmospheric differential refraction (Filippenko 1982). It is important to observe several standard stars (preferably from the HST spectrophotometric standards of Oke 1990). These standard stars are believed to be reliable to about 1% across the optical spectrum, and thus, this sets a fundamental minimum level of uncertainty in any observed emission line ratio. Observations of both red and blue stars allows a check on the possibility of second-order contamination of the spectrum. Typically, one-dimensional spectra are extracted from long-slit (2-D) observations. Special care needs to be taken setting the extraction aperture width and the aperture should be sufficiently wide that small alignment errors do not give rise to systematic errors (this comes at a cost in signal/noise, but ensures photometric fidelity). Given these potential uncertainties, it is unreasonable to record errors of less than one percent in emission line ratios, regardless of the total number of photons recorded. Of course, multiple independent measurements of the same target provide the best estimates of true observational errors, and existing measurements of this type confirm this minimum error estimate (Skillman et al. 1994).

    It should also be noted that it is imperative to integrate under the emission line profile (as opposed to fitting the line with a Gaussian profile). Fitting procedures can introduce systematic differences between high signal/noise and low signal/noise lines. Given the dynamic range of the H I and He I emission lines required to produce an accurate He/H abundance (e.g., the faint He I line 6678 is about 1% of H and He I 4026 is less than 2% of H), any systematic error between measuring strong and faint lines will have dramatic results. A special challenge is presented by the presence of underlying stellar absorption. The underlying absorption is generally broader than the emission, so quite often, when observed at a resolution of a few Angstroms or better, the H I or He I emission line is sitting in an absorption trough. Measuring all H I and He I emission lines in a consistent manner is important to obtaining a good solution for both the emission strength and the underlying absorption (see next section). Measurements at maximum resolution possible (while still measuring all lines simultaneously) are preferred.

    Because (1) we know the theoretical emissivities of the recombination lines of H I (e.g., Hummer & Storey 1987), (2) the ratios of the H I recombination lines in emission are relatively insensitive to the physical conditions of the gas (i.e., electron temperature and density), and (3) there are a number of H I recombination lines spread through the optical spectrum, it is possible to use the observed line ratios to solve for the line-of-sight reddening of the spectrum (cf., Osterbrock 1989). If one assumes a reddening law ( f (), e.g., Seaton 1979), in principle, it is possible to solve for the extinction as a function of wavelength by measuring a single pair of H I recombination lines. Values of C(H), the logarithmic reddening correction at H, can be derived from:

    where I () is the intrinsic line intensity and F () is the observed line flux corrected for atmospheric extinction. Assuming a reddening law introduces a degree of uncertainty. Studies in our Galaxy have shown that the reddening law exhibits large variations between different lines of sight, but these variations are most important in the ultraviolet (Cardelli, Clayton, & Mathis 1989). Additionally, the total measured extinction can have both Galactic and extragalactic components (and note the added complexity of the shift in wavelength for the reddening law for systems at significant redshift). Note that it is typical that no error is associated with the assumption of a reddening law. Davidson & Kinman (1985) point out that tying the He I emission lines to the nearest pair of bracketing H I lines significantly reduces the impact of assuming a reddening law (i.e., "the interpolation advantage"), but it is unlikely that there is absolutely no error incurred with this assumption.

    Underlying stellar absorption will affect the ratios of individual H I line pairs, so, in practice, it is best to measure several H I recombination lines. One can then solve for both the reddening and the stellar absorption underlying the emission lines (e.g., Shields & Searle 1978 Skillman 1985). It is generally assumed that the underlying absorption for the brightest Balmer H I lines is constant in terms of equivalent width. It is not clear how much error is incurred through this assumption, and inspection of stellar spectra shows that it is unlikely to be true for the fainter Balmer emission lines (e.g., H8, H9, and higher). However, one has the observational check of comparing these corrected lines to their theoretical values.

    We recommend solving for the reddening and the underlying absorption by minimizing the differences between the observed and theoretical ratios for the three Balmer line ratios H / H, H / H, and H / H. Both H7 and H8 are blended with other emission lines, so they cannot be used for this purpose. While the H9 and H10 lines are often not observed with sufficient accuracy to constrain the reddening and absorption, in high quality spectra, the relative strengths of H9 and H10 provide a check on the derived solutions. In Appendix A we describe our method of using a 2 minimization routine to determine the best values of C(H), the underlying equivalent width of hydrogen absorption ( a HI ), and their associated errors.

    Figure 1 is presented for instructional purposes. It shows a comparison of the observed and corrected hydrogen Balmer emission line ratios for three synthetic cases. In constructing this figure, synthetic H I Balmer emission line spectra were calculated assuming an electron temperature (18,000 K), density (100 cm -3 ), and H equivalent width (100 Å). Balmer emission line ratios were derived for three different combinations of reddening and absorption. All emission lines and equivalent widths were given uncertainties of 2%. In the first case, the spectrum was calculated assuming reddening and no underlying absorption. The second case assumes underlying absorption and no reddening. The third case has both. The open circles show the deviations of the original synthesized spectra from the theoretical ratios in terms of the synthesized uncertainties (2% for all lines). Note that reddening and underlying absorption induce corrections in the same direction for all three line ratios, i.e., the H / H line ratio increases for increased reddening and underlying absorption and the bluer Balmer line ratios all decrease for both effects. This covariance results in a degeneracy, thereby decreasing the diagnostic power of the corrections as we will show.

    The filled circles in Figure 1 show the results of using the 2 minimization routine described in Appendix A. If such a minimization is used, then the 2 should be reported. This allows one to make an independent check on the validity of the magnitude of the emission line uncertainties. As one can see, the minimization procedure accurately reproduces the assumed input parameters. In case 1, the minimization found C(H) = 0.10 ± 0.03 and a HI = 0.00 ± 0.57. Similarly, for the other two cases, we find C(H) = 0.00 ± 0.03, a HI = 2.00 ± 0.59 and C(H) = 0.10 ± 0.03, a HI = 2.00 ± 0.59 respectively. In all three cases, since the data are synthetic, the 2 values for the solutions are vanishingly small. Appendix B discusses cases from the literature where the 2 values are quite large, indicating either a problem with the original spectrum, an underestimate of the emission line uncertainties, or both.

    As a test to determine the appropriateness of the uncertainties for the values of C(H) and a HI as produced by the 2 minimization, we have run Monte Carlo simulations of the hydrogen Balmer ratios. The Monte Carlo procedure is described in Appendix A. Figure 2 shows the results of Monte Carlo simulations of solutions for the reddening and underlying absorption from hydrogen Balmer emission line ratios for three synthetic cases based on the input parameters of case 3 of Figure 1. That is, the original input spectra had reddening with C(H) = 0.1 and a HI = 2 Å. For these values of C(H) and a HI , we have run the Monte Carlo for three choices of EW(H) = 50, 100, and 200 Å. (EW(H) = 100 was used in Figure 1). Let us first concentrate on the results shown in the bottom panel of Figure 2. Each small point is the minimization solution derived from a different realization of the same input spectrum (with 2% errors in both emission line flux and equivalent width). The large open point with error bars shows the mean result with 1 errors derived from the 2 solution from the original synthetic spectrum. The large filled point with error bars shows the mean result with 1 errors derived from the dispersion in the Monte Carlo solutions. Note that the covariance of the two parameters leads to error ellipses. The Monte Carlo simulations find the correct solutions, but the error bars appropriate to these solutions are significantly larger than the errors inferred from the single 2 minimization. In this case there is a small offset in the mean solutions (mostly due to the fact that solutions with negative values are not allowed). In the bottom panel, the errors in C(H) are 29% larger and the errors in a HI are about 61% larger for the Monte Carlo simulations compared to the single 2 minimization.

    The middle and top panels of Figure 2 show cases for decreasing emission line equivalent width. Note that, given the input assumptions, the constraints on the underlying absorption are stronger in absolute terms for the lower emission line equivalent width cases. In all three cases, the 2 minimization errors are smaller than those produced by the Monte Carlo simulations. For the middle panel, the differences are 41% for C(H) and 80% for a HI , while for the top panel, the differences are 46% for C(H) and 86% for a HI .

    These test cases have shown that the errors in C(H) and the underlying stellar absorption can be underestimated by simply using the output from a 2 minimization routine, and that Monte Carlo simulations can be used to give a more realistic estimate of the errors. Based on this experience, we recommend that the best way to determine the true uncertainties in the derived values of C(H) and a HI is to run Monte Carlo simulations of the hydrogen Balmer ratios. Simply running a 2 minimization will underestimate the uncertainty (due, in large part, to the covariance of the two parameters being solved for). Since the reddening correction must be applied to the He I lines as well, this uncertainty will propagate into the final estimation of the He abundance. This uncertainty, we find, is too large to be ignored.

    If He I lines are observed at a given wavelength , their intensities relative to H after the reddening correction is given by eq. (1). The ratios I () / I ( H ) can then be used self-consistently to determine the He abundance and the physical parameters describing the HII regions, after the effects of collisional excitation, florescence, and underlying absorption as described in the next section. We can quantify the contribution to the overall He abundance uncertainty due to the reddening correction by propagating the error in eq. (1). Ignoring all other uncertainties in X R () = I () / I ( H ), we would write

    In the examples discussed above, C(H)

    0.04 (from the Monte Carlo), and values of f are 0.237, 0.208, 0.109, -0.225, -0.345, -0.396, for He lines at 3889, 4026, 4471, 5876, 6678, 7065, respectively. For the bluer lines, this correction alone is 1 - 2% and must be added in quadrature to any other observational errors in X R . For the redder lines, this uncertainty is 3 - 4%. This represents the minimum uncertainty which must be included in the individual He I emission line strengths relative to H. Note that these errors alone equal or exceed the 2% errors in the individual line strengths assumed for this exercise. However, the magnitude of the reddening error terms for the red lines can be reduced if these lines are compared directly to H. If the corrected H / H ratio is identical to the theoretical ratio, then it would be allowable to include only the uncertainty in the reddening difference between H and the red He I emission line. On the other hand, it is frequently the case that the corrected H / H ratio is significantly different from the theoretical ratio.

    Finally, we should note that an additional complication is the possibility that, in the highest temperature (lowest metallicity) nebulae, the H line may be collisionally enhanced (Davidson & Kinman 1985 Skillman & Kennicutt 1993). In their detailed modeling of I Zw 18, Stasinska & Schaerer (1999) have found this to be an important effect (of order 7% enhancement in H). If this is not accounted for, this has the effect of artificially increasing the determined reddening (and thus, artificially decreasing the helium abundances measured from the lines redward of H (e.g., 5876, 6678) and increasing the helium abundances measured from lines blueward of H (e.g., 4471). More work along the lines Stasinska & Schaerer (1999) with photoionization modeling of high temperature nebulae is needed to determine whether this effect is common in these low metallicity regions.

    Since the temperature is governed by the balance between the heating and cooling processes, and since the cooling is governed by different ionic species in different radial zones, one expects different ions to have different mean temperatures (cf. Stasinska 1990 Garnett 1992). While this is best treated with a complete photoionization model, a reasonable compromise is to treat the spectrum as if it arose in two different temperature zones, roughly corresponding to the O + and O + + zones. Since the oxygen ions play a dominant role in the cooling, this is a reasonable thing to do. Deriving temperatures in the high ionization zone generally consists of measuring the highly temperature sensitive ratio of the emission from the "auroral line" of [O III] (4363) relative to the emission from the "nebular lines" of [O III] (4959,5007). Temperatures for the low ionization zone are usually derived from photoionization modeling (e.g., PSTE) although it is possible to derive temperatures in the low ionization zone from the [O II] I(7320 + 7330) / I(3726 + 3729) ratio and a similar ratio for [S II] (e.g., González-Delgado et al. 1994 PPR).

    Note that, to date, usually only the temperature from the high ionization zone is used to derive the He abundance, and the He which resides in the low ionization zone is generally not dealt with in a self-consistent manner. To estimate the potential size of this effect, we can look at the data for SBS1159+545 from IT98. In SBS1159+545, 19% of the oxygen is in the O + state (and thus 81% in the O + + state). Assuming all of the gas to be at a temperature of 18,400 K (the [O III] temperature), a 5876 / H ratio of 0.101 ± 0.002 yields a helium abundance of 0.0855 (before reduction to account for collisional enhancement and in agreement with IT98). Assuming 81% of the gas to be at the [O III] temperature of 18,400 K and 19% of the gas to be at the [O II] temperature of 15,200 K results in a helium abundance of 0.0848, or a difference of 0.8%. While this is a small difference, it is not negligible when compared to the reported uncertainty in the measurement. Curiously, including the effects of collisional enhancement almost perfectly cancels this effect for the reported density of 110 cm -3 ( y + = 0.0815 treated as a single temperature zone and y + = 0.0811 treated as two temperature zones for this object). Thus, using a lower temperature for the y + in the O + zone can increase or decrease the helium abundance depending on the density. The main point here is that the temperatures used for the two zones and the helium abundance should be treated consistently (as emphasized by PPR).

    Steigman, Viegas, & Gruenwald (1997) have investigated the effect of internal temperature fluctuations on the derived helium abundances and find this to be important in the high temperature regime. The presence of temperature fluctuations, when analyzed assuming no temperature fluctuations, results in underestimating both the oxygen and helium abundances (here only [S II] densities are used, which are typically higher than the densities derived from He I lines). Assuming a range of relatively large temperature fluctuations (with a maximum of 4000K) results in an overall shift in the derived primordial helium abundance of about 3%. Steigman et al. have argued that, in absence of constraints on the temperature fluctuations, the errors should be increased to account for this uncertainty.

    Peimbert, Peimbert, & Ruiz (2000) have shown that the different temperature dependences of the He I emission lines can be used to solve for the density, temperature, and helium abundance simultaneously and self-consistently. They point out that photoionization modeling consistently shows that the electron temperature derived from the [O III] lines is always an upper limit to the average temperature for the He I emission, and thus, assuming the [O III] temperature will always produce an upper limit to the true helium abundance. Here we will not explore the possibility of adding the electron temperature as a free parameter to our minimization routines. This is, in part, because the main motivations are to explore the method promoted by IT98, to explore the possibility of handling the effects of underlying absorption, and also, because one of our main conclusions, that Monte Carlo modeling is required for a true estimation of the errors will be true regardless of the minimization parameters. Nonetheless, this is a very important result with the implication that most helium abundances reported in the literature to date are really upper limits .

    The average density can be derived by measuring the relative intensities of two collisionally excited lines which arise from a split upper level. In the "low density regime" collisional de-excitation is unimportant and all excitations are followed by emission of a photon. The ratio of the fluxes then simply reflects the ratio of the statistical weights of the two levels. In the "high density regime", where the level populations are held at the ratio of their statistical weights, the emission ratio becomes the ratio of the product of the statistical weights and the radiation transition probabilities. In the intermediate regime, near the "critical density" the line ratios are excellent density diagnostics. The best known is that of [S II] 6717 / 6731 which is sensitive in the range from 10 2 to 10 4 cm -3 and can be observed at moderate spectral resolution. At higher spectral resolution, one can use several other line pairs (e.g., [O II] 3726 / 3729).

    In order to convert these line ratios into densities, one needs to know the energy level separations, the statistical weights of the levels, and the radiative and collisional excitation and de-excitation rates. Fortunately, one can use the five-level atom program originally written by De Robertis, Dufour, & Hunt (1987) which has been made generally available within IRAF (1) by Shaw & Dufour (1995). This program has the additional great advantage that the authors have promised to keep the input atomic data updated.

    As emphasized by ITL94, ITL97 and IT98, the [S II] line ratio suffers from two problems as a density diagnostic: (1) it is measuring the density is the low ionization zone, which may apply to less than 10% of the emission in a low metallicity giant HII region, and (2) it is relatively insensitive to density below about 100 cm -3 . Since the collisional excitation of the He I lines is important at the 1% level down to densities as low as 10 cm -3 , the [S II] lines are not ideal density indicators (cf. Izotov et al. 1999), and deriving densities directly from the He I lines is, in principle, preferable. This is discussed further in ڊ. However, calculating the density from the [S II] lines (and other collisionally excited lines) provides an excellent consistency check on the density derived from the He I lines.

    1 IRAF is distributed by the National Optical Astronomy Observatories, which is operated by the Association of Universities for Research in Astronomy, Inc. (AURA) under cooperative agreement with the National Science Foundation. Back. *****


    Dosimetry

    IV.B.7 Beta Sensitivity of TLD Materials

    All TLD materials in common use are sensitive to beta radiation. Theoretically, it would be possible to use TLDs for routine beta dosimetry. However, accurate assessment of beta-radiation absorbed dose is difficult to achieve with the personnel-monitoring devices currently available. Most of the conventional dosimeters were designed to detect the penetrating component of the radiation field and estimates of the nonpenetrating component radiation field and estimates of the nonpenetrating component are often obtained through the use of algorithms derived from calibrations in standard radiation fields.

    The inability to perform accurate beta dosimetry can be attributed to a number of factors. These include the spectral energy distribution of the radiation, the low penetrating nature of the radiation, the wide energy range of beta-emitters encountered in the work environment, the influence of backscatter and attenuation in the badge components, and the lack of suitable calibration sources and techniques. An additional and very important factor is the steeply sloped energy-response curve exhibited by most TLD materials. For example, if the relative response per unit exposure of LiF TLD-100 to the 2.2-MeV beta particles from 90 Y is assumed to be 1.0, the response is ∼0.2 for beta radiation from 204 Tl (0.76 MeV) and 90 Sr (0.55 MeV). The decrease in response continues, reaching ∼0.08 for 99 Tc (0.29 MeV) and ∼0.04 for 35 S (0.17 MeV).

    Within the last few years a number of personnel-monitoring badges have been designed that are intended to measure the beta component of the radiation field more accurately. These dosimeters are usually multielement (i.e., four or more TLDs) and require fairly sophisticated algorithms to obtain an estimate for the beta dose. Many attempts to produce an ultrathin TLD have been reported. Often, a TLD powder is mixed with a polyethylene base so that a thin but flexible dosimeter can be produced. At least one commercial badge has a thin (0.015-in.-thick) TLD incorporated in it strictly as a beta dosimeter. However, enclosing the dosimeter in some sort of badge or holder usually defeats the purpose of the beta dosimeter.

    At this point, the development of TLD dosimeters specifically for use in beta-radiation fields has not progressed past these techniques described. Other, more sophisticated techniques are under study. These include the implantation of materials such as carbon into the crystal to alter its response and new reading techniques using lasers. However, it can be concluded that the errors associated with personnel beta-radiation monitoring may be quite large and improvements urgently needed.


    An atlas of star-forming galaxy equivalent widths

    We present an atlas of starburst galaxy emission lines spanning 10 orders of magnitude in ionizing flux and 7 orders of magnitude in hydrogen number density. Coupling SEDs from Starburst99 with photoionization calculations from Cloudy, we track 96 emission lines from 977 Å to (205

    upmumbox) which are common to nebular regions, have been observed in H II regions, and serve as useful diagnostic lines. Each simulation grid displays equivalent widths and contains (sim1.5 imes10^<4>) photoionization models calculated by supplying a spectral energy distribution, chemical abundances, dust content, and gas metallicity (ranging from (0.2

    Z _) ). Our simulations will prove useful in starburst emission line data analysis, especially regarding local starburst galaxies that show high ionization emission lines. One sample application of our atlas predicts that C IV (lambda) 1549 will serve as a useful diagnostic emission line of vigorous star formation for coming James Webb Space Telescope observations predicting a peak equivalent width of approximately 316 Å.


    Tidally disrupted dusty clumps as the origin of broad emission lines in active galactic nuclei

    Type 1 active galactic nuclei display broad emission lines, which are regarded as arising from photoionized gas moving in the gravitational potential of a supermassive black hole 1,2 . However, the origin of this broad-line region gas is unresolved so far 1,2,3 . Another component is the dusty torus 4 beyond the broad-line region—probably an assembly of discrete clumps 5,6,7 —which can hide the region from some viewing angles and make them observationally appear as type 2 objects. Here, we report that these clumps moving within the dust sublimation radius, such as the molecular cloud G2 discovered in the Galactic Centre 8 , will be tidally disrupted by the black hole, resulting in some gas becoming bound at smaller radii while other gas is ejected and returns to the torus. The clumps fulfill necessary conditions to be photoionized 9 . Specific dynamical components of tidally disrupted clumps include spiral-in gas as inflow, circularized gas and ejecta as outflow. We calculated various profiles of emission lines from these clouds, and found that they generally agree with Hβ profiles of Palomar–Green quasars 10 . We found that the asymmetry, shape and shift of the profiles strongly depend on [O iii ] luminosity, which we interpret as a proxy of dusty torus angles. Tidally disrupted clumps from the torus may represent the source of the broad-line region gas.

    The mass of individual clumps can be estimated by the tidal disruption condition at the dust sublimation radius given by the inner edge of torus, (_<< m>>approx 0.4phantom< ule<0em><0ex>>_<45>^<1/2>_<1,500>^<-2.6>,< m>) , where L 45 = L UV/10 45 erg s −1 and T 1,500 = T sub/1,500 K is the dust sublimation temperature 6,11,12 . The tidal disruption happens at or inside the Roche limit of D tid = (M /M C) 1/3 R C, where M is the black hole mass, M C is the clump mass and R C = (M C/πN H m p) 1/2 is the size, where N H is the column density and m p is the proton mass. Dust-free clouds are disrupted when D tidD sub, and therefore the largest dust-free clouds are given by (_/_approx 2.4phantom< ule<0em><0ex>>^<3>_<8>_<1,500>^<-15.6>_<24>^<3>) , their size is (_=5.1 imes 1<0>^<13>^<3/2>_<8>^<1/2>_<24>_<1,500>^<-7.8>) cm and their density (_=1.5 imes 1<0>^<10>^<-3/2>_<8>^<-1/2>_<1,500>^<7.8>phantom< ule<0em><0ex>>c^<-3>) , where M is the earth mass, () = L 45/M 8, (_<8>=_<ullet >/1<0>^<8>_) and N 24 = N H/10 24 cm −2 . For active galactic nuclei (AGNs) with (_<ullet >/_=1<0>^ <6>sim 1<0>^<9>) and () = 1, we have (_/_approx 0.02 sim 24.0) . This estimate shows that the typical properties of the captured clumps generally fulfill the photoionization condition for broad emission lines of AGNs and quasars 9 . Moreover, we realize that (M C, R C, n C) are very sensitive to T sub, implying that the properties of broad-line regions (BLR) generally depend on processes of dust production in galactic nuclei. Note that () = 1 corresponds to Eddington ratios of L Bol/L Edd ≈ 0.35, where L Bol ≈ 5L UV is used 13 . In the Galactic Centre, G2 with a mass of 3M captured by the central black hole 8 is well in this range of M C, lending support to the idea that such captures could be common in other galactic centres.


    Did stripped stars re-ionize the Universe?

    Mathieu Renzo, a PhD student at the University of Amsterdam.

    Today’s astrobite is a guest post, written by Mathieu Renzo. Matheiu is a third year PhD student at the University of Amsterdam, working on the evolution of massive stars in binaries using both population synthesis and structure models. He enjoys traveling, and takes advantage of all the opportunities to go anywhere that come with the PhD.

    Authors: Y. Götberg, S. E. de Mink, J. H. Groh

    First Author’s Institution: Institute for Astronomy, University of Amsterday, The Netherlands

    Status: Published by A&A, open access

    One long standing puzzle about the evolution of the Universe is the epoch of reionization. After the recombination of electrons and protons that created the cosmic microwave background, the universe was opaque and light could not get through. But roughly

    500 million years after the Big Bang, something turned on and started producing ionizing radiation, causing protons and electrons to re-separate. This ended the “cosmic dark ages” and made the universe transparent again.

    It is generally believed that this “something” was the population of massive stars (although AGN are also considered). Most stars have a thick hydrogen-rich envelope, which absorbs the ionizing photons they produce, but stars more massive than

    30 solar masses can lose to stellar winds their entire hydrogen-rich envelope. By doing so, these stars reveal their hot helium-rich core and become Wolf-Rayet stars, which can release ionizing radiation. However, current estimates of the massive star populations in the early Universe can hardly produce enough ionizing photons from Wolf-Rayet stars to explain the epoch of re-ionization.

    In today’s paper, Götberg et al. propose another way to remove the ionizing-radiation-blocking envelope from stars: mass transfer in binaries. In the local universe, the vast majority of massive stars are in binaries. Interactions with a companion can change the properties of stars. In particular, it is expected that

    30% of massive stars will lose their hydrogen-rich envelope to their companion after the end of their main sequence phase, when they expand to become giants. By losing their envelope, these stars also expose their helium core, and become “stripped stars”. Stripped stars might have contributed or even dominated the epoch of re-ionization, if massive stars preferred to be in binaries also in the early universe.

    Figure 1: Evolution on the HR diagram of a single 12 solar mass star (A-B-C-gray track), and the same star in a binary (colored line, from A to H). The single star never enters the shaded region, where a significant amount of ionizing radiation is emitted. Instead, the binary donor spends all its core helium burning duration there, roughly 10% of its total lifetime. The binary mass transfer phase is marked by the black contour, and the color of the binary track indicates the core helium abundance: long live phases correspond to where the colors change faster. Source: Figure 2 from today’s paper.

    This paper presents an exploratory calculation of the evolution of massive stars in binaries, both their internal structure (c.f. Fig. 1), using a stellar evolution code, and the “look”, i.e. the spectrum (c.f. Fig. 2), using a radiative transfer code of the donor star in a representative example of a binary. They explore how both vary as a function of the metallicity of the stars, i.e.

    the total amount of elements heavier than helium. As you look back in cosmic history, the metallicity tends to decrease, so stars in the early Universe had a much lower metallicity than the stars in our galaxy.

    They find that stripped stars at all metallicities can produce a significant amount of ionizing radiation. They do find that at lower metallicities a larger fraction of the hydrogen envelope is retained, but because of the high effective temperature achieved, that hydrogen is fully ionized, which means it does not block the ionizing radiation.

    Stripped stars can be much less massive than typical Wolf-Rayet stars, but still just as hot and ionizing. Therefore, because of binary interactions, stars that would never have provided ionizing radiation were they single, can emit up to

    80% of their flux in ionizing radiation (cf. Fig. 2). Moreover, because of their lower mass, these stripped stars live longer and can outnumber typical Wolf-Rayet stars. And finally, the stripped stars come later: the binary interaction removing the hydrogen envelope is delayed by the entire main sequence lifetime of the donor. This is helpful, because it leaves time for stellar winds and supernovae from nearby stars to clear the gas around the (soon to become) stripped star and make sure the ionizing photons are not blocked in its vicinity.

    Figure 2: Spectral energy distribution of a stripped star at solar metallicity (filled purple curve), with overplotted putative companions. Colored bars on the top indicate the spectral range accessible to present-day instruments, dashed gray bars indicate the spectral range of decommissioned instruments. Some spectral lines can be seen to outshine the continuum of companion in wavelengths accessible from the ground. Source: Figure 13 from today’s paper.

    So, in summary: Yes! The stripped stars are lower mass version of Wolf-Rayet stars, they can only be produced in binaries and they come later, are more numerous, and live longer than Wolf-Rayet stars. Because of all these effects combined, they might have had an important role in the beginning of the “age of enlightenment” of the universe.