Development of Imaging Fourier Transform for Astronomy by gjjur4356


									     Development of an Imaging Fourier Transform Spectrometer for
                        Frédéric Grandmont*a,b, Laurent Drissen**a, Gilles Joncas***a
                         Département de physique, Université Laval, bABB Bomem


We present an overview of the past and current development of the Imaging Fourier Transform Spectrometer (IFTS)
concept for ground telescopes produced in collaboration between ABB Bomem and Université Laval. This instrument
intends to produce spectra of variable resolutions up to R = λ/∆λ = 10 000 from the near UV to the near IR (350 nm to
900 nm). It is designed to fit the f/8 focus of the Mont Mégantic 1.6m optical telescope in Québec. The large number of
spatial elements (> 1 million pixels) is the novel aspect of this FTS design along with innovative metrology system.
Heritage from Next Generation Space Telescope (NGST) IFTS concept, The Lawrence Livermore National Laboratory
(LLNL)- ABB Bomem instrument and commercial ABB Bomem DA series FTS are reviewed. Techniques for
accurately servoing the moving mirror alignment to a value smaller than 0.1 arc second and position to sub nanometer
value are discussed. Also presented are results from the assembled interferometer sub-system.
Keywords: Imaging Spectrometer, Fourier Transform Spectroscopy, IFTS, Integral Field Unit

                                               1. INTRODUCTION
The NGST project in its early definition phases has given a strong push for the development of astronomical
instrumentation by providing new sources of funding in an area often considered less lucrative and hence of lower
interest for industrial partners. New alliances were formed with universities and institutions and the knowledge exchange
lead to very interesting new concepts. Many configurations of IFTS, a derivative of the classical Michelson
interferometer that has been used successfully in spectroscopy for decades, were proposed. The NGST platform was
well suited to benefit from such an integrated instrument capable of carrying many different science programs
simultaneously. The IFTS revealed itself to be the only true integrated instrument capable of performing 3-D spectro-
imagery without compromising waveband, spatial resolution and field of view (FOV) of the array detector. The IFTS
also presented a noble approach to prevent observational bias introduced by objects preselection required by standard
Multi-Object Spectrograph (MOS). The use of spectral folding technique along with preselection optical filters was also
proposed to obtain high resolution in restricted wavebands in a timely manner. All these aspects are being reused in the
ground version of the Canadian space concept presented for NGST. Additional challenges imposed by varying sky
transmission, presence of gravity and non-vacuum environment were addressed by different hardware/software

                                        2. A BRIEF HISTORY OF IFTS
The IFTS is a concept known for decades in astronomy. In fact, most FTS designs contain an image plane at the detector
to ensure that the FTS produce spectra relevant of a finite and known FOV. Migrating from single pixel FTS to IFTS
merely consist in adding pixels at the output. In the seventies, solar observatories already equipped with FTS have tested
multi-pixels detectors with mitigated succes1. Primitive version of IFTS were also introduced in military applications in
the mid 80's by slowly adding individual detectors at the focal plane output of commercial FT-IR. The first truly
operational IFTS in astronomy was put together in the early 90's2 as a modification to the popular Canada France Hawaii
Telescope (CFHT) FTS already in used since 1984. This instrument called BEAR produced three-dimensional data sets

*; phone 1 418 877-2944x318; fax 1 418 266-1422;; ABB
Bomem, 585 Boul. Charest est, Suite 300, Québec, QC G1K 9H4
**; phone 1 418 656-2131x5641; fax 1 418 656-2040;;
Université Laval, Québec, QC, Canada, G1K 7P4
***; phone 1 418 656-5982; fax 1 418 656-2040;;
Université Laval, Québec, QC, Canada, G1K 7P4
in the infrared that were unmatched in many ways3. Lately, the interest around the development of an integrated science
instrument for NGST capable of multi object spectroscopy along with wide field imagery revived the interest for the
IFTS4. This led to a project by LLNL to build a prototype imaging IFTS in the visible that operate from ground which
preliminary results5 were presented at the NGST science and exposition meeting along with 4 other IFTS proposals6,7,8,9.
A joint effort between LLNL and ABB Bomem then led to a second prototype instrument and has been turned into a
visiting instrument at Apache Point Observatory 3.5m telescope10.
In other fields of science using hyperspectral imagery such as medical, military and geoscience, the IFTS is already
positioned as a reference in instrumentation and commercial version are out on the market. The developments in detector
technology and improvements in digital electronic speed enabled FTS designers to increase the size of array detectors
without preventing the fast scanning of the Optical Path Difference (OPD) required for instrument stability. Now
military versions of 16 X 16 pixels can readily obtain spectrum rates of about 5 to 150 per second for the whole detector
depending on the number of points in the spectrum (250 to 8000 pts)11. Such IFTS require sampling the complete array
in the kHz range and terabits/s of data transfer are involved. For detector size of interest to astronomers (> 1K2 pixels),
read out rates typically of seconds to minutes are off by orders of magnitude. Classical fast scanning FTS found in the
commercial market for decades cannot be used, as their servo system cannot be slowed down to this level. Moreover,
astronomers often need to expose their detector to the scene for long periods of time, which suggests holding the
interferometric position of the moving mirror during the exposure. The "step scan" or "stop and stare" approach is used
to meet the requirements of this application. In this respect, the astronomy IFTS is different from what is typically found
on the market. Being stopped most of the time, these interferometers can easily accommodate any size of detector with
long readout time as long as the optical design allows the corresponding FOV. However an accurate servo system must
be designed to maintain the optical position of the moving mirror, especially in the visible waveband.
                                        3. THE IFTS CONFIGURATION
The suggested configuration for an astronomical IFTS follows from the desire to reduce to a minimum the number of
optical surfaces encountered by the science beam from telescope output to the camera. In addition, the correction of the
sky transmission fluctuations requires the collection of the total flux in the FOV and thus suggests a four port design (2
inputs and 2 outputs). Such configurations are usually achieved using cube corner retroreflectors or cat's eye mirrors. If
one wants to produce an interferometer that strongly modulates the visible light down to UV, the optical surfaces located
between the two beamsplitter passes must exhibit excellent surface figure (<λ/20) to allow even interference of the two
recombining wavefronts. The two mirror options mentioned previously complicate this task because of the additive
errors of all three reflections encountered in addition to optical shape consideration in the case of cat's eyes or
orthogonality issues in the case of cube corners. The flat mirror interferometer, besides providing one third of the
absorption of the others and improved wavefront accuracy, can also provide a four ports configuration when used off
axis as shown on Figure 1.
                                     Input from





                        OPD Piezo
                         Actuator                                                         Cold Shutter
                                                                                         or Deep Space
                                                                          Input #2
                                                              2         Output #2
                            DA Piezo                                                           Detector
                            Actuator          Moving                                              #2
                                                                         Fix Mirror

                                  Figure 1: Interferometer Optical Configuration
The current configuration uses this scheme at a 30 degree angle of incidence on the beamsplitter to minimise the
beamsplitter size in comparison to a 45 degree configuration. The off-axis angle of the incoming beam is chosen
perpendicular to the interferometer optical plane as shown in the computer aided design (CAD) drawing of Figure 2 in
order to further reduce the size of the round beamsplitter which in turn reduces the interferometer size. The two figure
views are approximately in the same orientation to facilitate the identification of parts. The incoming and outgoing
beams have been represented by hollow tubes running in and out of one of the interferometer port. The unit is sealed
with windows to prevent wind from destabilising the mirrors and bringing with them strong and rapidly changing
temperature gradients into the interferometric path. Vacuum operation would obviously help in solving air related OPD
problems but at the cost of increased weight and complexity of use. The stability in the sealed configuration was proven
to be sufficient for operation in the visible. The overall dimensions of the interferometer module are 50 X 25 X 16 cm.
These measures do not account for the metrology source and detection module.
                                            Figure 2: Interferometer Module
One imperative of the IFTS is to obtain spectrum with Signal to Noise Ratio (SNR) that benefit from all photons from
the source. Hence the two output ports must be imaged separately since their interferogram are complementary. The
current optical design aims at imaging both fields side by side on a single CCD as does the BEAR instrument12. The
fields are rectangular and should be of 7.5 X 15 arc min. This ensures that both exposition times are identical and will
resolve issues related to synchronization and noise level of distinct cameras. No obscuration is required to insert the laser
metrology beam in the interferometer. Hence adding the interferometer in the optical path running from the telescope to
the camera (given that I/O optics of focal reducer/enlarger are present) only add flat optics; 1-3 reflections and 6
substrates transmission. However, it forces the need for a second detector and hence increases detector noise

Problems of image ghosts can occur if wedge or angles of the flat transmissive substrates (beamsplitter/compensator,
interferometer windows) are not correctly adjusted. Secondary reflections are directed out of the detector using field stop
in front of the input optics or are precisely overlaid on main image. Seeing limited images were obtained using a similar
optical configuration in the LLNL-ABB Bomem instrument10.

The fact that the two output ports of the interferometer are imaged on separate detector sections is critical for solving the
sky transmission issues. Our instrument takes typically minutes to hours to acquire 1 datacube (interferograms for each
pixel) depending on exposure time and number of steps. Sky transmission fluctuations will create a variation in the
recorded intensity at each step that is not related to the mirror position hence interference. The Fourier transform
calculation used to obtain spectra from interferograms assumes that all intensity variations are related to interference and
produce the spectrum accordingly. It is then of paramount importance to correct those variations before "Fourier
transforming" the data at the risk of producing ghosts in the spectrum. This "renormalisation" of the interferograms is
performed using the visibility function obtained from the addition of both detector signals. Since the total energy from a
point in the FOV must end up on one detector or the other, the total intensity that reaches the instrument at each step can
be retrieved for each frame. Intrinsic source variations can also be corrected to a certain degree using this technique as
long as the spectral characteristics are stable. This method was used successfully with the LLNL - ABB Bomem
The flat mirror configuration has the strong disadvantage of requiring a very precise alignment throughout the mirror
scan. Passive alignment methods are rejected since the scan mechanisms would have to maintain the 0.05 arc sec
required for modulation of near UV light over the 1 cm mechanical travel. The sweep mechanism used is based on a
piezo actuated porchswing system that passively maintains alignment to approximately 2 arc second. A Dynamic
Alignment (DA) system similar to the one used in the ABB Bomem DA series, flat mirror FTS commercialized for more
than 20 years, has been modified to operate in step scan mode. The active system must compensate for mirror
misalignment during scan as well as perturbation caused by gravity vector reorientation when tracking stars. Two piezo
stacks mounted on the back of the moving mirror support can modify its angle in orthogonal axis over 1.6 arc minutes.

The alignment is not the only parameter on which precise control must be exerted. Sampling UV wavelengths of 300
nm at Nyquist frequency suggest taking 1 CCD exposure every 150 nm of OPD, hence every 75 nm of mirror
displacement (light round trip to the mirror doubles the physical movement). For comparison purposes, an aluminium
rod of 20 cm (typical distance separating mirrors and beamsplitter) takes 0.16°K variation to produce the same
movement in thermal expansion. Concept study for the NGST6 suggested that positioning accuracy between 1 and 5 %
of the step size would be required to minimize noise contribution attributable to sampling jitter. This leaves us with
positioning requirements in the range of sub-nanometers! A precise positioning system must be designed to evenly
sample the interferogram but also to maintain the mirror position during CCD exposure subjected to gravity and thermal
perturbation. The base of an innovative metrology system able to achieve such resolution were set at Université Laval
and successfully implemented by ABB Bomem in a risk reduction prototype in 1999 produced under financing of the
Canadian Space Agency (CSA). A complete IFTS instrument was assembled by the LLNL using the prototype
interferometer. The results obtained with the prototype confirmed the viability of the servo system concept. A few
modifications were added such as increasing the length of the piezos and the speed of the servo loop. The Figure 3
shows the diagram of the positioning system.

        On Telescope
                                                       Science                                         25 Mb/s Link
                                                       Beam Output                        ADC 16bit       (LVDS)
                                                                                           1 MHz
                         CCD Camera

                             Beamsplitter                                                    ADC              DIO
                             Compensator                                                                     PCI card

                                                                              IR Detectors              SERVO CPU
                                                                                                         P-III 850 MHz
         Science Beam Input                                               Alignment Piezos
                                                                                                            PC/104 card
                                                                          OPD Piezos

                        Metrology                         Moving Mirror                        6m

                         Beam                                                                  Cable
                                    Fix Mirror                                                           Controller
                                                                             In Control        12 m
                                                                               Room            Cable   In Telescope
          Laser diode
            1.55 µm


                                                 Figure 3: Servo System Diagram
A laser source of wavelength chosen outside the CCD sensitivity region is directed on axis into the interferometer. 64
infrared detectors at the output collect the laser light and are digitized locally at a global rate of 10 kHz. Those values are
sent over a high speed digital communication link to a single board computer located nearby the telescope feet. A C
coded servo loop computes the angle and position of the moving mirror with respect to the fix mirror. This information
is used to correct the piezos length via digital to analog converters that drive the piezo amplifiers. This servo loop was
tested to run in the kilohertz range on a Pentium III 850 MHz. The high speed is most useful for keeping track of the
absolute OPD in case of fast perturbations. The fact that the servo computer knows at all time the absolute OPD in the
interferometer enable to use various sampling approaches that can be of great interest to astronomers.
To obtain a specific spectrum (spectral region and resolution) with an FTS, one must decide before hand the sampling
interval (step size) and the number of samples to be taken in the interferogram. Classically, interferogram are sampled
symmetrically around the Zero Path Difference (ZPD). In broad band observation, only a small portion of the
interferogram (around ZPD) is strongly modulated. Hence if a very large set of steps is foreseen for a given
measurement, bad weather appearing in the middle of the measurement could results in useless data sets if all taken in
weakly modulated region. However, starting from ZPD and building the data sets by acquiring samples from each side
alternatively, as shown in Figure 4, can results in a useful data set even if it cannot be completed. In this case the spectral
resolution of the measurement would be reduced accordingly but the spectrum would have a strong SNR.

                             Sampling Points

                                         ...   4     3    1    2    5    6    ...

                                         Figure 4: Alternative Sampling Method
Another advantage of knowing the absolute OPD at all time is the selection of the sampling interval. In the servo
computer code, the OPD is stored as a float number. This means that any value of step size can be selected in order to
tailor the spectral range of the acquisition. When used in conjunction with preselection optical filter and spectral folding
technique, FTS can produce spectra of various finite wavebands. Figure 5 shows examples of the various wavebands that
can be selected using the proper step size and aliasing order. The lower equations on the graph enable to calculate the
waveband (∆λ) according to a desired central wavelength (λc) and the order of the alias chosen (n). The upper one relates
the chosen waveband setting to the step size (λéch) required to properly sample the interferogram. For any of these
waveband selections, an optical filter must be obtained to transmit light within the exact band or a smaller one. The
extinction ratio of the preselection filter in the rest of the CCD sensitivity region is primordial to ensure that the aliased
spectrum does not contain residual transmitted signal from other wavelengths. One could also think of using "camel
back" type band pass filter which could select sub waveband from two different alias. This could be most useful for
generating datacube in two distinct spectral regions of interest separated by a large unused continuum instead of
performing two separate measurements with half the acquisition time.
                                                                                ∆λ = 2/n(n+1) λéch


                                                                                     ∆λ = λc/(n+½)
           ∆ λ (nm)



                      200                                                                          1st folding
                                                                                                   2nd folding
                      100                                                                          3rd folding
                                                                                                   4th folding
                                                                                                   5th folding
                        200       400            600           800            1000          1200            1400
                                                          Central λ (nm)

                                         Figure 5: Waveband Selection Chart
                                           5. LAB TESTING RESULTS
The interferometer module has been fully assembled and is currently under testing. All substrates used were
characterised on ZYGO at various temperatures once mounted in their support to ensure that the FTS will offer
diffraction limited performance and will not affect the Point Spread Function (PSF) of the telescope during observation.

One very important part of the design that needs accurate testing is the beamsplitter coating efficiency in the band of
interest. This is because the beamsplitter is the component that imposes the limit on the waveband of the IFTS although
the current coating efficiency band is very close to the CCD one. The waveband of the instrument is defined accurately
since later on in the design process, the optical design is tuned accondingly for the reduction of the chromatic aberration
of the input/output optic system if refractive optics are chosen. The current beamsplitter consist in a 20 layers dielectric
coating laid on a 2 cm thick UV grade fused sillica plate. The ensemble has near zero absorption and achieves near 45 –
55 % reflection from 350 – 900 nm. Figure 6 shows the measurement result of the reflection curve which waviness is
typical of such multi-layer coatings. This ensures modulation efficiency losses due to splitting efficiency to be around
1% in the waveband.
                                      Figure 6: Beamsplitter Transmission Curve
The interferometer has been tested to modulate monochromatic light evenly over its full 2 cm of OPD leading to a
theoretical spectral resolution of 0.6 cm-1 FWHM (R = 48000 @ 350nm to R = 21000 @ 800nm). Effects of FOV and
pixel size along with positioning accuracy of the mirror in the telescope environment will potentially degrade this value.
This part of the test program is awaiting the completion of the optical design.

                                    6. PROPOSED SCIENCE PROGRAMS
The IFTS advantages reside mostly in the wide field imaging spectroscopy mode. Three science programs with different
spectral resolutions will be scheduled at the telescope to determine the usefulness of the IFTS and its ability to provide
high quality astronomical data:

1. Low resolution full spectral range
The IFTS could in principle advantageously simulate the use of broad-band filters (such as Johnson’s or Thuan-Gunn’s)
to obtain color-magnitude diagrams of stellar clusters. In principle, only low-resolution interferograms (R~10) are
necessary to obtain the required information, but the calibration of the data and its comparison with color-magnitude
diagrams obtained with standard filters will probably require higher resolution. As a first step, we plan to observe a few
well-known clusters for which photometry has been obtained with different sets of standard filters.

2. Mid resolution full spectral range
At a resolution R~500-1000, important diagnostic nebular lines can be studied in extended objects. In particular, narrow-
band filter imaging and slitlet spectroscopy have been widely used to measure the spatial variations of important
physical parameters, such as interstellar reddening, electron temperature and chemical abundances in nebulae and
galaxies. Both techniques have their drawbacks13: the use of interference filters cover a wide area, but a large number of
filters must be used to cover all the important lines, extreme care must be taken with the calibration, and only face-on
galaxies can be observed because of the differential rotation of the galaxies causing the emission lines to shift outside of
the filter bandpass. Multi-object spectrographs, such as the CFHT/MOS14 allow to obtain a relatively large number of
well-calibrated spectra across galaxies, but still cover a very small fraction of the entire field of view. The IFTS,
covering a large FOV and wide spectral range, will be used to measure chemical abundance gradients in a sample of
spiral and irregular galaxies; the data will be compared with those of conventional techniques and will be used to infer
the evolutionary histories of galaxies15,16. The full spectral range from [OII] 372.7 nm to [SII] 673.1 nm is required for
this work, with a resolution sufficient to resolve the [NII] lines at 658.4 nm and 654.8 nm from Hα at 656.3 nm. Density
and temperature fluctuations within single, Galactic HII regions will also be studied.

3. High resolution narrow band
While the use of the full field of view may be the main advantage of the IFTS, it will also allow moderately high
resolution (R~10000) spectra of bright objects to be obtained. This is the kind of resolution provided by our standard
spectrograph at the observatory, and the use of a single instrument to perform two completely different scientific
programs in the same night without changing the instrument could be advantageous. To test this specific instrumental
mode, we will obtain spectra of selected emission lines in bright Wolf-Rayet stars to distinguish individual sub-peaks
caused by the clumpiness of their strong winds17.

                                                 7. CONCLUSION
Instrument structure design is currently underway as well as Input/Output optics for large field imaging on the CCD.
This probably represents one of the biggest challenges of the IFTS since for pan chromatic spectro-imagery no
refocusing of the telescope is possible given that all wavelengths are acquired simultaneously. Hence a wide field
imaging system with chromatic aberration correction for a very large waveband must be designed. The use of reflective
optics has been considered for smaller FOV. Table 1 summaries some of the IFTS characteristics.

                    IFTS Characteristic                                           Value
                    Mass                                         < 100 kg (FTS = 10 kg)
                    Dimension                                    < 130 X 60 X 60 cm
                    Waveband                                     350 – 900 nm
                    Input Beam F/#                               ƒ/8
                    Total FOV                                    ~ 7 X 14 arc min
                    Pixel FOV                                    0.65 X 0.65 arc sec
                    Detector Type                                Liquid Nitrogen Cooled CCD
                    Maximum Path Difference                      ± 1 cm from ZPD
                    Spectral Resolution Limit (R = λ/∆λ)         Up to R = 48000 @ 350nm and
                                                                 R = 21000 @ 800 nm
                    Servo Loop Speed                             > 1 kHz
                    Maximum Moving Speed (with absolute          0.5 mm/s in OPD
                    OPD tracking)
                    Dead Time Between Steps (stepping +          2 seconds (limited by CCD readout)
                    reading previous frame)
                                            Table 1: IFTS Characteristics

Funding is secured for all subsequent development phases of the IFTS. First light at the Mégantic Telescope is expected
in the fall of 2003. We are very anxious to compare the outcoming results with the ones gathered with the LLNL – ABB
Bomem instrument.

This work is supported by many organizations other than Université Laval and ABB Bomem. The CSA supported the
development of the interferometer itself. The rest of the instrument and study work was financed by the National Science
and Engineering Research Council (NSERC) of Canada, the "Fond Québécois de Recherche sur la Nature et les
Technologie" (FCAR), the Canadian Institute for Photonic Innovations (CIPI) and the Canadian Foundation for
Innovation (CFI). This work also benefited largely from the LLNL-ABB Bomem IFTS demonstration effort. Some of
the current design original ideas were generated among this group. A special thanks to Ron Wurtz, Ed Wishnow,
Sebastien Blais-Ouellet, Dennis Carr and Kem Cook for their help.
1.    M. J. S., Belton, S. T. Ridgway, "The Spectroplanetograph; or a Two-dimensional Imaging Fourier Spectrometer",
      Bulletin of the AAS, Vol. 10, p.547, 1978
2.    Maillard, J.P., Simons, D., ESO Conf. On Progress in Telescope and Instrumentation Technologies, No 42, p. 733,
3.    Maillard, J. P.; Joblin, C.; Mitchell, G. F.; Vauglin, I.; Cox, P., S106: morphology and excitation of PAH, H2 and H
      from ISO, C10µ and BEAR, The Universe as Seen by ISO. Eds. P. Cox & M. F. Kessler. ESA-SP 427, 1999
4.    Graham, J. R., Abrams, M., Bennett, C., Carr, J., Cook, K., Dey, A., Najita, J., Wishnow, E.H., The Performance
      and Scientific Rationale for an Infrared Imaging Fourier Transform Spectrograph on a Large Space Telescope,
      PASP, Volume 110, Issue 752, pp. 1205-1215, 1998
5.    Wurtz, R., Cook, K.H., Bennett, C.L., Bixler, J., Carr, D., and Wishnow, E.H., Ground-Based Demonstration of
      Imaging Fourier Transform Spectrometry and Techniques, ASP Conf.Ser.207: Next Generation Space Telescope
      Science and Technology, p. 203, 2000
6.    Graham, James, IFIRS: An Imaging Fourier Transform Spectrometer for the NGST, ASP Conf.Ser.207: Next
      Generation Space Telescope Science and Technology, p. 240, 2000
7.    S. L. Morris, A.Villemaire, F.Grandmont, L. Moreau, J. Ouellette, "A Canadian IFTS for the NGST", ASP
      Conference Series, Vol. 207, P. 276, 2000
8.    Posselt, W., Maillard, J.P., Wright, G., NIRCAM_IFTS: Imaging Fourier Transform Spectrometer for NGST, ASP
      Conf.Ser.207: Next Generation Space Telescope Science and Technology, p. 303, 2000
9.    Maillard, J.P., Comparison of Two Concepts of Imaging FTS, ASP Conf.Ser.207: Next Generation Space Telescope
      Science and Technology, p. 479, 2000
10.   R. Wurtz, E. Wishnow, S. Blais-Ouellette, K. Cook, D. Carr, F. Grandmont, I. Lewis, C. Stubbs, LIFTS: an Imaging
      Fourier Transform Spectrograph for Astronomy, ASP Conf. Ser, 2002
11.   A. Villemaire et al., "High-resolution airborne imaging spectrometer", Proc. SPIE Vol. 3436, p. 924-930, 1998
12.   Maillard, J.P., 3-D Spectroscopy with a Fourier Transform Spectrometer, 3D Optical Spectroscopic Methods in
      Astronomy, ASP Conf. Ser., Vol. 71, p. 316, 1995
13.   Y. Dutil, J.-R. Roy, "The Hazards of Imaging Spectrophotometry with Interference Filters", The Astronomical
      Journal, Vol. 122, p. 1644, 2001
14.   L. Drissen, J.-R. Roy, A. F. J. Moffat, M. M. Shara, "The Ionizing Star Clusters of Giant H II Regions in NGC
      2403", The Astronomical Journal, Vol. 117, p. 1249, 1999
15.   Y. Dutil, J.-R. Roy, "Chemical Evidence for Morphological Evolution of Spiral Galaxies", The Astrophysical
      Journal, Vol. 516, p. 62, 1999
16.   D. Devost, J.-R. Roy, L. Drissen, "The Ionized Gas in the Aftermath of a Starburst: The Case of NGC 1569", The
      Astrophysical Journal, Vol. 482, p. 765, 1997
17.   S. Lépine, A. F. J. Moffat, "Wind Inhomogeneities in Wolf-Rayet Stars. II. Investigation of Emission-Line Profile
      Variations", The Astrophysical Journal, Vol. 514, p. 909, 1999

To top