Undergraduate, Peer-Reviewed Science Journal
|
Issue 1, March 2003 Physical Sciences and Mathematics Starburst Populations in the Interacting Galaxies NGC 3395 and NGC 3396 Robert Bachilla University of Nevada at Las Vegas Advisor: Donna E. Weistrop, Ph.D. University of Nevada at Las Vegas
Abstract
Introduction
Tidal effects between the galaxies have distorted their shapes and initiated bursts of star formation. The star-forming regions we observe in NGC3395/3396 have sizes of about 1,000 light years, on average, and contain many clusters of newborn stars. Only a few of the largest, brightest regions are visible in Figure 1; they appear as red or yellow areas that are relatively small in comparison the size of each galaxy. NGC 3395 (right) is viewed face-on, while NGC 3396 (left) is seen nearly edge-on. The entire complex is positioned about 92-million light years away in the direction of Leo Minor. Starburst99 Models Models have been developed and widely used over the past few decades to rationalize the observed characteristics of star-forming regions. Of particular interest to our research are the Starburst99 (SB99) models developed by Leitherer et al. (1999). The SB99 models combine spectra of individual stars to predict the spectral appearance of a large number of stars as a function of time. Specifically, they predict the luminosity of a star-forming region as a function of wavelength and age, based on the number and variety of stars present in active star-forming regions. The model spectra are readily customized to observable parameters, such as redshift, extinction, and metallicity. In the present context, redshift describes the Doppler shift of light emitted from NGC3395/3396 toward the red end of the visible spectrum. Extinction describes the effects of photon scattering and absorption by interstellar gas and dust. Sources of extinction include gas and dust in both the Milky Way and the NGC3395/3396 galaxy pair. In terms of astronomy, "metals" are any elements heavier than helium. Metallicity then describes overall metal abundance of a region as a ratio of metal to hydrogen. All of these parameters are important to proper use of the SB99 models. Metallicity in particular must be determined by a second set of mathematical models. Metallicity Models In general, mathematical models of metallicity are based on emission line observations of nearby HII regions where independent methods exist to determine metal abundance. Such analysis has led researchers to construct robust metallicity models that may be extended for use in the spectra of distant galaxies. The mathematical models of McGaugh (1991, 1994) and Kobulnicky (1999) are particularly useful to our research. These metallicity models provide a relationship between the observed spectral emission lines of oxygen and the metallicity of an HII region. This is a very convenient relationship because the spectral lines of oxygen are among the most intense in the spectra of star-forming regions. Present Study In this analysis, we investigate 36 star-forming regions in NGC3395/3396 to determine the dominant stellar populations within each region. The 36 regions represent the largest and brightest areas of star formation in both galaxies, and therefore the most readily studied. In any particular star-forming region, there exists a varying combination of stars, HII regions, and other nebulae. Each one of these structures contributes photons to the overall observed spectrum. To distinguish the individual stellar populations in our observed star-forming regions, we compare our observed spectra to several SB99 models that are customized to match our observed parameters. By performing a best-fit analysis of our observed spectra and the SB99 models, we determine the age of our stellar populations and the dominant stellar species within each region. This information also allows us to establish a timeline of major episodes of star-formation within the interacting galaxy pair. This analysis of NGC3395/3396 is valuable to a broader understanding of star-formation as a result of galaxy interaction. It is part of a larger project led by Donna E. Weistrop at UNLV that aims to describe how star formation propagates during galaxy interaction. In brief, her team studies the effects of galaxy interaction on star formation in a number of different galactic environments and at different stages of merger maturity.
Approach
We distinguish the important characteristics of our spectra by examining the three main spectral components: absorption lines, emission lines, and continuum. Each object within the star-forming region contributes a particular component to the observed spectrum. For instance, stars within the star-forming regions are major contributors of absorption lines and continuum, while the luminous HII regions contribute strong emission lines, as well as a minor amount of continuum. The sharp Ha emission line at 6563 A provides a convenient gauge of the galaxy's redshift. The peak position of the Ha line may be measured using the SPLOT spectral analysis package in IRAF (Image Reduction and Analysis Facility). IRAF is a major analysis tool by astronomers, and is very useful in many types of applications. In our present research, we use IRAF to examine spectra and measure line positions in terms of wavelength and flux. We also use IRAF to manipulate the mathematical models and perform calculations, as described in future sections. Since we are interested in stellar populations, we remove the bright emission lines to better view the stellar features (absorption lines and continuum). The reasons for this are actually twofold. First, by isolating the absorption lines and continuum we can more easily notice distinguishing features in those parts of the spectra. Second, the SB99 continuum models do not contain emission features, so a best-fit analysis is more accurate without large-scale deviations in the observed continuum. We remove emission lines from the spectra using the SPLOT package in IRAF. The resulting spectra display more prominent and clearly identifiable stellar features. The isolated (emission line removed) continuum of a representative spectrum is shown in Figure 3.
The spectrum in Figure 3 shows several important characteristics that are common to the other spectra in NGC3395/3396. One notable characteristic is the steady increase of continuum emission toward shorter wavelengths. Stars can be modeled as blackbodies in most cases because their continuum emission is similar to blackbody radiation. This relationship provides us with some evidence about the region's stellar populations. Some of the most outstanding features in the optical region of the continuum are found between 3400 A and 4200 A. Here we find the Balmer discontinuity and the Balmer absorption lines. The Balmer discontinuity, or "Balmer jump" appears as a steep decline in the spectrum toward shorter wavelengths just after ~3700 A. This feature develops because photons of wavelength 3647 A or shorter have sufficient energy to ionize hydrogen from its n=2 level, while photons at longer wavelengths do not. In the spectra for NGC3395/3396, the Balmer jump limit occurs at a redshifted value of about 3667 A. Some of the important Balmer absorption lines also appear between 3400 A and 4200 A. We obtain a closer view of these absorption features in Figure 4.
In this region of the spectrum, we identify the Balmer series absorption lines, a result of downward electron transitions that terminate at the hydrogen n=2 orbital. They are especially important in this context because their intensities reveal information about the region's dominant stellar populations. Balmer absorption is strongest in A-type stars, which have an upper-range main sequence lifetime of about 600 Myr. B-type stars also show prominent Balmer absorption lines. O-type stars display weaker Balmer absorption lines, but they are still present. In the range <3200 A, the detector response begins to deteriorate, and noise dominates the spectra. The same effect appears at the opposite end of the spectra at about 8000 A. These areas mark the limits of the optical wavelengths, the extent of Earth's optical window, and the detector's response. Metallicity Even though HII regions contain a small number of oxygen atoms, the emission lines of ionized oxygen are among the strongest peaks found in an optical spectrum (see figure 2). The prominence of [OII] and [OIII] emission makes oxygen a convenient source to determine a region's metallicity, X/H, or the ratio of all elements to hydrogen (excluding helium). The electron transitions that produce [OII] and [OIII] emission have a low probability of occurrence, so they are known as "forbidden transitions" and produce "forbidden" emission lines. Here, the brackets ([…]) indicate that the emission is a result of forbidden transitions. The strength of forbidden emission lines varies according to the electron temperature within the HII region. Metal ions of oxygen, sulfur, and nitrogen act as coolants that reduce the mean kinetic energy (temperature) of the electron gas by removing energy from the region. Metal ions such as OII and OIII effectively cool an HII region because the excitation potentials of their low-lying levels are on the order of kT, leaving them susceptible to collisional excitation. Some time after being collisionally excited, an ion de-excites and emits a photon. This photon then escapes the HII region, carrying away its kinetic energy and reducing the overall electron temperature of the gas. In this way, metal ions make a significant contribution to radiative cooling despite their low abundances. To demonstrate the process of radiative cooling in more detail, we consider the energy level diagrams for ionized oxygen, [OII] and [OIII], shown in Ffigure 5.
To illustrate the process, we consider a stationary OIII ion in the path of a free electron with energy hn >>2.5kT. The free electron gives some (or all) of its kinetic energy to excite the oxygen ion. A bound electron initially in the ground state of [OIII] reaches the metastable state m1 and remains there for a relatively long time. Eventually the bound electron spontaneously de-excites into the ground state, emitting a photon of wavelength 4959 A or 5007 A. The forbidden photon then escapes the HII region. The net effect is a removal of kinetic energy from the region and a reduction in temperature. In this analysis, we use the forbidden line fluxes for [OII] and [OIII] to determine the metallicity of a star-forming region, as proposed by McGaugh (1991) and Kobulnicky (1999). There are several steps in this process. First, we must determine the theoretical oxygen abundance of each region using the observed oxygen emission lines [OII] and [OIII]. We then use our calculations of model oxygen abundance to derive the overall metallicity of each region. We use the metallicity indicator R23 to calculate the relative strength of oxygen emission to Hb emission. The expression is given by where [OII], [OIII], and Hb are the extinction corrected line fluxes for wavelengths l3727, ll4959 and 5007, and l4861, respectively. In the spectra for NGC 3395 and NGC 3396, we observe strong, sharp emission lines of [OII] and [OIII]. As a result, peak positions and fluxes are readily measured using the SPLOT spectral analysis package in IRAF. To be useful in this context, the emission lines are corrected for extinction. In this case, data sets for extinction corrected line fluxes were available from a previous research project. Figure 6 shows the model relationship between the oxygen abundance, log(O/H), and R23, as proposed by McGaugh (1991, 94) and Kobulnicky (1999).
The upper branch in Figure 6 represents conditions in the most metal-rich HII regions. R23 is a minimum where the model oxygen abundance is greatest, [12+log(O/H)] ~9.1. Here, the high metal abundance produces efficient cooling in the HII region, reducing the electron temperature and suppressing further collision events. As a result, the [OII] and [OIII] emission lines are absent from the spectrum, or weak, at best. Further toward the right, R23reaches its maximum as the oxygen abundance decreases to [12+log(O/H)] ~8.4. Here, fewer metals are available to cool the region, and collisions occur more frequently, leading to strong, observable emission lines of [OII] and [OIII]. The maximum value for R23 also represents a turnaround in the model relationship. R23 steadily decreases as smaller numbers of metal ions are present in the region. This trend in R23 continues to zero as the diminishing population of metal ions produce weaker, less observable emission lines. In our calculations, we use the polynomial equations for the curves in Figure 6 shown below (Kobulnicky 1999): LOWER BRANCH: UPPER BRANCH: - y(0.0047 - 0.0221x - 0.102x2 - 0.0817x3 - 0.00717x4) where and The ionization parameter, log(O32), describes the relative number of OIII to OII ions within the region. It is also a measure of the number density of ionizing photons. As seen in Figure 6, the O32 parameter has a different effect on each branch of the model. In the upper branch, a smaller O32 ratio slightly reduces the model oxygen abundance for a given R23. The O32 ratio has a greater and more uniform effect on the lower branch, where a smaller ratio indicates a higher model oxygen abundance of about 1% for any value of R23. To determine the oxygen abundance for each region in NGC3395/3396, we must know beforehand which model branch (lower or upper) to use in our calculations. The lower and upper model branches correspond to eq(2) and eq(3), respectively. Here, we use the ratio [NII]/[OII] to indicate the appropriate branch. HII regions in high surface-brightness galaxies generally have log([NII]/[OII])3 -1. These bright HII regions are metal rich, while HII regions in low surface-brightness galaxies are usually metal poor (McGaugh 1991). McGaugh (1991) presents a model abundance grid that relates oxygen abundance to the overall metallicity, Z, in an HII region. In short, McGaugh's abundance grid allows us to convert our calculated oxygen abundance, log(O/H), to metallicity, X/H, where X represents all elements heavier than helium. The model abundance grid is reproduced for our purposes in Table 1.
By linear interpolation, we have added several rows to McGaugh's original grid to include the several metallicity parameters in the Starburst99 continuum models. Model Spectra To investigate the stellar populations in NGC3395/3396, we use the SB99 model spectral energy distributions developed by Leitherer, et al. (1999). The SB99 continuum models predict the observed properties of regions undergoing active star-formation. Specifically, they plot luminosity (erg/s/A) as a function of wavelength for ages in increments between 1 and 900 Myr. There are several different models available for various environmental conditions. Each model depends on three main parameters: 1) Initial Mass Function, 2) Metallicity, and 3) Star Formation Process. Each parameter is described below.
IRAF Analysis and Computer Routines All methods described below are performed using standard IRAF packages. We begin our analysis using the SPLOT package to remove the bright emission lines and atmospheric features from the observed spectra. This allows us to isolate the stellar components of the spectra, namely the absorption lines and continuum. We then choose several representative spectra from each galaxy and identify all outstanding features. The spectra are especially rich in Balmer absorption and may also include visible lines of calcium, magnesium, and occasionally helium. Next, we use the SYNPHOT package to convolve the Starburst99 models to match the observed redshift and reddening (extinction) characteristics of NGC 3395/3396. We know from previous analysis that NGC 3395 and NGC 3396 both have redshifts of 0.005. Each region has a different value for reddening in between 0.1 and 0.4 (rounded to one significant figure). However, research by other astronomers shows that reddening in the continuum is about 1/2 the reddening determined from emission line analysis. While there is no conclusive explanation for this, it may be an effect of the dusty HII regions where emission lines are produced. Stars that produce the spectral continuum reside in less-dusty regions of space, which may account for a decrease in reddening. To proceed with our analysis, we identify a value for 1/2 reddening in each region and convolve the model set accordingly. As a result, we develop three separate model sets for comparison to the observed spectra, each at a redshift of 0.005, with reddening in 0.1 intervals between 0 and 0.2 (see Table A). Next, we separate the spectra from each galaxy into three wavelength ranges, or "lbands": (3200-3900 A), (3902-6200 A), and (6202-8000 A). This step is necessary to identify the stellar populations that dominate each region of the spectrum. It is necessary to normalize the observed spectra and the models because they have different units. The observed spectra are in flux units, "erg/s/A/cm2." and the models are in luminosity units, "erg/s/A." Flux and luminosity differ by a constant 4pR2, where R is distance-in this case about 92 million light years. Thus, by normalizing the observed spectra, we eliminate the R dependence. Once we normalize the observed spectra and models, we perform a best-fit analysis using the FITGRID package to determine which models match the observed spectra. A best-fit criteria is based on the least sum-of-squares difference between the observations and models.
Results and Interpretations which relates wavelength l (in meters) to temperature T (in K). For example, a G-type star like the Sun has a temperature of about 6000 K, so its peak wavelength intensity occurs at about 5000 A. Clearly, the continuum in Figure 3 reaches maximum intensity somewhere in the ultraviolet or near-ultraviolet wavelengths. Therefore, we are certain that cool M, K, or G-type stars do not dominate this region. The continuum's steady increase into shorter wavelengths indicates that either A, B or O-type stars, with temperatures 10,000, 25,000, and 50,000 K, respectively, dominate the region. Since we do not have spectra in the UV and near-UV, we cannot say anything more definite about the wavelength of peak emission. Our investigation of the Balmer absorption lines reaffirms our expectations that O, B, and A-type stars dominate the region. The massive O and B stars are most directly responsible for the bright appearances of our star-forming regions. They are the dominant cause of HII region luminosity. Metallicy Using the log(O/H) and log(R23) relationship in Figure 6 and the corresponding polynomial equations (eqs(2)-(5)), we calculate log(R23) in a range of values between 0.6 and 0.9. Our measured values for log(O32) are distributed fairly evenly between -0.7 and 0.1. Our measured values of log([NII]/[OII]) vs. log(R23) are shown in Figure 7 (also see Table A).
In Figure 7, the observed [NII]/[OII] line ratios fall safely above -1. This indicates that NGC 3395 and NGC 3396 are metal-rich galaxies of high surface-brightness. Therefore, we use the metal rich upper branch to calculate oxygen abundance from our observed spectra. We determine uncertainty for each data point in Figure 7 by examining the regions for which we have more than one observation. For example, we have two independent spectra of region 3395#12. Ideally, we should calculate an identical [NII]/[OII] flux ratio for each of the two spectra, since they are of the same region. Hence, any discrepancy in the calculated flux ratios reveals an approximate uncertainty in our measurements. Our spectra contain four regions with two independent spectra each: 3395#2, 3395#12, 3395#15, and 3396#26. Calculations of log([NII]/[OII]) from these recurring observations show a maximum discrepancy of 7% in region 3395#12. We assume that all data points have similar uncertainty. Using eq(3), we plot model oxygen abundance vs. observed R23 in Figure 8.
Calculations of 12+log(O/H) show a maximum uncertainty of 0.7% in region 3395#12 and also in 3395#15. We assume that all data points have similar uncertainty. Figure 8 shows that our observed values for R23 correspond to a range of model oxygen abundance between 8.4 and 8.8 (see Table A). The regions in NGC 3995 have an average oxygen abundance of 8.56, while the regions in NGC 3396 have an average of 8.60. Consider the model abundance grid in Table 1, and recall that the R23 data in Figure 8 gives 12+log(O/H) = 8.56±0.7% and 8.60±0.7% for NGC 3395 and NGC 3396, respectively. Note that McGaugh represents oxygen abundance as log(O/H), so we must subtract 12 from our data for proper comparison to the model grid. We then obtain log(O/H) = -3.44±0.7% and -3.40±0.7%. According to Table 1, these values most closely correspond to a metallicity Z=0.008. SB99 Model Spectra To summarize our model parameters, we use an IMF of M-2.35, an instantaneous star formation process, and a metallicity of Z=0.008. Sample model spectra based on these parameters are shown in Figure 9, below. Three different ages are shown for comparison:
The SB99 models follow an expected trend according to their age. The 5-million-year-old (5 Myr) model has a steep incline toward shorter wavelengths and small Balmer absorption lines. This model suggests a dominant presence of hot, bright, young stars of spectral type B and O. In the 50 Myr models, the incline is less steep, and the Balmer absorption lines are more pronounced. By 50 Myr, all the O stars have died out, along with the most massive B stars. A-type stars and the less massive species of B stars dominate the region (in brightness). These populations display prominent Balmer absorption lines, as the models indicate. Out of the three models shown, the 500 Myr model shows the strongest Balmer absorption of all. Its moderately steep incline and strong Balmer absorption suggests a dominant population of A-type stars, the least massive species of B-type stars, and massive F-type stars. Stellar Populations Table B shows our key results of stellar age populations in all 36 star-forming regions. Figure 10 shows example results from one best-fit analysis of observed spectra to the SB99 models.
In both galaxies, results from lband 1 (3200-3900 A) and lband 3 (6202-8000 A) show a population of young stars less than 20 million years old (see Table B). In both bands, but especially in lband 3, there is evidence of an even younger population less than 10 Myr. These results suggest a major episode of star-formation in the recent past. The existence of these young populations is confirmed by Hancock et al. in the pending Astrophysical Journal article "Star Forming Regions in the UV Bright Interacting Galaxies NCG 3395 and NGC 3396." In lband 2 (3902-6200 A), the ages are distributed over a larger range, but it is still possible to distinguish at least one older population. The foremost older population is between 30 and 90 Myr, and there is some evidence of an even older population greater than 100 Myr. Results from all three bands suggest three major star-forming episodes within the last few hundred million years. A concise description of follows: lBand 1: 32 regions have ages <20 Myr. Of these regions, 7 have ages <10 Myr. Two regions have an age of 40 Myr. lBand 2: 7 regions have ages <20 Myr. 19 regions have ages between 30 and 90 Myr, and 8 regions have ages >100 Myr. lBand 3: 32 regions have ages <20 Myr. Of these regions, 24 have ages <10 Myr. One region is 90 Myr, and another is 900 Myr. A histogram distribution of our observed age populations is shown in Figure 11.
Discussion
Acknowledgements
References Kobulnicky, H., et al. (April 1999) On Measuring Nebular Chemical Abundances in Distant Galaxies Using Global Emission-Line Spectra. Astrophysical Journal vol. 514:544-557. Leitherer C., et al. (July 1999) Starburst99: Synthesis Models for Galaxies with Active Star Formation. Astrophysical Journal Supplement vol. 123:3-40. McGaugh, S. (October 1991) HII Region Abundances: Model Oxygen Line Ratios. Astrophysical Journal vol. 380:140-150. McGaugh, S. (May 1994) Oxygen Abundances in Low Surface Brightness Disk Galaxies. Astrophysical Journal vol. 426:135-149. Osterbrock, Donald E. Astrophysics of Gaseous Nebulae and Active Galactic Nuclei. Sausalito: University Science Books, 1989. Swihart, Thomas L. Astrophysics and Stellar Astronomy. New York: John Wiley & Sons, 1968. Zeilik, Michael and Stephan A. Gregory. Astronomy and Astrophysics. Fourth ed. Fort Worth: Saunders College Publishing, 1998. Journal of Young
Investigators. 2003. Volume Seven.
Copyright © 2003 by Robert Bachilla and JYI. All rights reserved. |
JYI is supported by: The National Science Foundation, The Burroughs Wellcome Fund, Glaxo Wellcome Inc., Science Magazine, Science's Next Wave, Swarthmore College, Duke University, Georgetown University, and many others.