Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.astro.louisville.edu/moore/wispi/pasp2000.pdf Äàòà èçìåíåíèÿ: Mon May 10 10:16:59 2010 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 19:56:49 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: çàêîí âèíà

PUBLICATIONS OF THE ASTRONOMICAL SOCIETY OF THE PACIFIC, 112 : 801 õ 808, 2000 June ( 2000. The Astronomical Society of the Paciïc. All rights reserved. Printed in U.S.A.

A Wide-Field Spectral Imager
JOHN F. KIELKOPF AND PAMELA M. GRAHAM
Department of Physics, University of Louisville, Louisville, KY 40292 ; john=aurora.physics.louisville.edu, pam=nimbus.physics.louisville.edu Received 1998 June 30 ; accepted 2000 March 1

ABSTRACT. A spectroscopic imager was designed and constructed to provide a new capability for measuring the spectrum of the sky across wide ïelds of view. The instrument uses a high-speed optical system to form an image on the entrance slit of a matching stigmatic spectrograph, which then disperses a spectrum of a strip of the sky onto a charge-coupled device (CCD). A separate direct imaging system controls pointing and tracking, and both instruments share in tandem a computer-controlled mounting which permits automated data acquisition. The demonstrated optical performance is a 4 A ] 11@@ single pixel resolution in a 4000 A ] 3¡ ïeld. Absolute calibration on Vega gives a response at 6563 A that is 7 .2 times the CCD readout noise at an integrated exposure of 1 photon cm~2 A~1. Sample airglow, nebu lar, stellar, and cometary spectra illustrate the detection of extended emission features with ÿuxes from 5 to 2000 R. A measurement of zenith brightness for Na and Hg emission lines from urban artiïcial lighting is tabulated.

1. INTRODUCTION The wide-ïeld spectral imager (WISPI) we describe here was built to measure the emission spectra of extended sources with low surface brightness. Conventional telescopes equipped with nebular spectrographs have such a limited ïeld of view that large-scale structure escapes detection. With this in mind, we designed an instrument speciïcally for optical spectroscopy with spatial resolution over ïelds of up to 7¡. Recently a few observational programs have made use of the idea that the detectability of faint extended emission in the night sky is limited not by aperture but by f-ratio (Meaburn 1976). Fast camera lenses, narrowband Ha ïlters, and charge-coupled device (CCD) detectors combine to give an image ïeld of several degrees with 0¡ resolution and a .1 sensitivity to 1 R or less (Gaustad et al. 1997 ; Hentges & McCullough 1998 ; Gaustad, McCullough, & Van Buren 1996). The Wisconsin Ha Mapper (WHAM) uses a 0.6 m telescope and a Fabry-Perot spectrometer to image the sky in a 1¡ beam with 12 km s~1 velocity resolution in a 200 km s~1 spectral window from 4800 to 7200 A (Reynolds et al. 1998). Integration over the beamwidth and discrimination with high spectral resolution have given detections of high velocity cloud Ha emission lines as faint as 0.06 R (Tufte, Reynolds, & Haner 1998). Our interest, however, is to record a comprehensive spectrum and in the same exposure achieve angular resolution of better than 1@ across a ïeld of view of several degrees. Galactic cirrus, for example, is recognizable over ïelds of several degrees (McCullough 1997 ; Szomoru & Guhatha801

kurta 1998) but has ïlamentary structure as small as 30A (Sandage 1976). Szomoru & Guhathakurta (1998) report that the optical spectroscopic surface brightness of the extended red emission from the diuse interstellar medium is about 2 orders of magnitude less than the OH night sky emission, which makes its measurement challenging, especially with a narrow ïeld of view. WISPI is a fast imaging system coupled to a long-slit nebular spectrograph and a low-noise detector, designed to observe faint extended sources emitting line spectra. We will describe it brieÿy here and illustrate its performance with several examples.

2. DESIGN WISPI can be considered two matched instruments, a front-end imaging system with a stop at the image plane and a stigmatic spectrograph which uses that stop as its entrance slit. Like a "" long-slit îî spectrograph used with a large telescope, WISPI produces an image which is spatially resolved perpendicular to the direction of dispersion. This concept is constrained by the goal of maintaining the fastest possible optical system throughout. The need for a long physical slit length to give a wide ïeld of view, while optimizing resolution in the two-dimensional image plane, led us to consider using highly corrected transmission optics for the imaging components. Multielement 35 mm photographic camera lenses yield point-source images comparable to the pixel size of typical scientiïc CCDs. Since they are available with a large selection of apertures and focal lengths, we took advantage of these o-the-shelf com-


802

KIELKOPF & GRAHAM gen. Readout electronics are in a separate module, connected to the dewar by a 10 m cable and to a camera control computer by a fast serial link. The packaged detector system used in WISPI was manufactured by Princeton Instruments (now Roper Scientiïc). The pixel size and overall dimensions of this detector are matched well to the optical performance of the Nikon ED-IF series of Nikor camera lenses. These employ extra low dispersion (ED) glass, typically in the front elements, and have a mechanically stable internal focus (IF) adjustment. Although they are designed to be used primarily in the visible, our tests of ïve dierent samples showed that they are sufficiently well corrected to focus within 2 pixels up to 11000 A. Although the best focus setting in the blue violet is shifted from the best setting in the infrared, a band 4000 A wide may be focused in one image. This bandwidth is an eective compromise between the needs for broad

ponents in WISPIîs design. An overview of the optical layout is shown in Figure 1. It is convenient to explain the factors that led to this arrangement by starting with the detector and working forward, because detector size sets the instrument scale and determines the choice of lenses and grating. When the instrument was built, the largest low-noise detector we could obtain was a Tektronix 1024 ] 1024 thinned, back-illuminated CCD with 24 km square pixels. Its intrinsic response spans from 3800 A in the ultraviolet to 11000 A in the near-infrared, with a peak quantum effi ciency of 80% at 7000 A. Although the detector in WISPI has a proprietary ÿuorescent overcoating that extends its short-wavelength response to below 2000 A in the far ultraviolet, the optical system limits useful response to the band from 4000 to 11000 A. The detector is mounted in a universal orientation dewar and is cooled by liquid nitro-

Nikor lens 400 mm focal length 143 mm diameter Nikor lens 180 mm focal length 64 mm diameter Princeton Instruments 1024x1024 CCD SBIG 320x240 ST5 CCD

Nikor lens 180 mm

Nikor lens 200 mm focal length 100 mm diameter Milton Roy 300 groove/mm grating 102x128 mm 7620 A blaze

FIG. 1.õOptical layout of the Wide-Field Spectroscopic Imager (WISPI). Light enters through the 400 mm focal length f/2.8 objective and forms an image of the sky on the bilateral adjustable slit. It is collimated and then dispersed by a 300 groove mm~1 grating. The spectrum and the sky are imaged on the CCD detector.

2000 PASP, 112 : 801õ808


WIDE-FIELD SPECTROSCOPY coverage and for resolution adequate to separate lines in complex spectra. The spectrum from 4000 A to 11000 A is covered in two overlapping exposures. We selected a 200 mm f/2 lens for the spectrograph camera because it was the fastest long focal length lens available of the ED-IF type. It has 10 imaging components protected by an additional ÿat optical element. In total there are 11 components and 18 air-glass surfaces. The antireÿection coatings are eective, and we do not notice multiple reÿections or serious scattered light in the images. Its mechanical design provides support at a built-in foot near its center of mass and at the standard Nikor lens mount, which makes a bayonet connection to the CCD camera. A focusing ring remains accessible and can be adjusted without disturbing the other system components. For 1024 pixels along the dispersion direction, the reciprocal dispersion must be approximately 4 A pixel~1,or 160 A mm~1, in the spectral image plane to cover 4000 A. This ïxes the ratio of grating groove spacing to the camera focal length, without uniquely determining either one. When a 200 mm focal length lens is used in the spectrograph camera, the desired scale of 4 A pixel~1 is achieved with a 300 groove mm~1 grating. We use a 128 mm wide ] 102 mm high grating from Milton Roy (formerly Bausch and Lomb, now Spectronic Instruments) with a 7620 A blaze. The height of the grating is matched to the

803

diameter of the spectrograph camera entrance pupil. The width of the grating is greater than the projected beam diameter. The minimum angle possible between collimator axis and camera axis is 28¡, given the sizes of both lenses. The grating is centered on the intersection of the optical axes and mounted on a rotating stage which has a ïne adjustment for center wavelength. To preserve ïnal focalplane image quality, we placed the grating close to the point which makes the spectrograph telecentric on the detector side. The grating has 90% efficiency at its blaze angle, and in ïrst order from 5000 to 11000 A the response stays above the 50% grating efficiency points. In diraction grating instruments such as this one, the ïrst order image at any wavelength contains light from half that wavelength in the second order. When, for example, the spectrograph is to reach a maximum of 10000 A in the ïrst order, wavelengths less than 5000 A must be excluded. Combined eects of atmospheric transmission, optical glass transparency, and grating efficiency sharply delimit useful transmission at 4000 A without a separate ïlter, so that wavelengths up to 8000 A may be covered in the ïrst order without an order separation ïlter. A Wratten 21 gelatin ïlter with 95% transmission above 5500 A is inserted in the optical path as a long-pass cuto ïlter for imaging above 8000 A. Because beam shape changes at the grating, and because the grating sends dierent wavelengths to the spectrograph camera at varying angles, we require a camera lens of larger diameter than the beamwidth. Since we use an f/2 lens with an aperture of 100 mm as the camera, a smaller 180 mm focal length ED-IF f/2.8 Nikor lens is used for the collimator. The choice of collimator lens was made to optimize coupling between the imaging optics and the spectrograph optics, while providing a roughly 1 : 1 map of the slit to the
1500 Vega 1250 cm å ) Flux (photons s
-1

1000

-1

-2

750 O2 500 H O 250 H
2

H2O

0 4500

O 5000 5500 6000 6500 (å) 7000 7500

2

8000

FIG. 2.õThe mounted Wide-Field Spectroscopic Imager. The detector dewar is on the near side.

FIG. 3.õA spectrum of Vega recorded in a 10 s exposure on 1997 March 11. The image was processed to remove readout bias, corrected for a ÿat ïeld, and calibrated against a standard ÿux for Vega (Tug et al. 1977).

2000 PASP, 112 : 801õ808


804

KIELKOPF & GRAHAM
8.0
400 H NGC 7000

Response (ADU pixel / photons cm å )

-1

-2

300

6.0
Flux Difference (R) 200

-1

4.0

100 [NII]

[NII]

[SII]

2.0

He 0

0.0 4000

-100 6350

6450

6550 (å)

6650

6750

6850

5000

6000 (å)

7000

8000

9000

FIG. 4.õInstrumental response derived from spectra of Vega compared to absolute measurements (Tug et al. 1977).

FIG. 6.õDetailed dierence spectrum of the North American Nebula with most background features removed. Two regions from the image used to make Fig. 5 were subtracted : one with strong Ha and one with mostly night sky emission. The residual noise is of the order of 1 R.

image plane. This lens has eight elements with 12 antireÿection-coated air-glass surfaces. The slit was made by Spex Industries (now Jobin Yvon). It is bilaterally adjustable from 2 km to 2.5 mm, with a 25 mm height. A machined aluminum cylinder which houses the slit is ÿanged on each end to accept the bayonet connector of a Nikor lens. To make a rigid coaxial assembly, the collimator and objective lenses are then mounted back to back on opposite sides of the slit, with the collimator supported also by a machined clamp that holds it ïrmly at its
1500 Hg Na 1250 NGC 7000

front element. WISPIîs primary structural component consists of this assembly attached to a 20 cm ] 1 m aluminum channel baseplate. The choice of imaging optics is constrained only in that the lens system must supply an f/2.8 beam over a 25 mm ïeld. An eight-element 400 mm f/2.8 ED-IF Nikor lens gives an image scale of 11A pixel~1 at the detector and a ïeld of .1 view of 3¡ along the slit. The front two elements are ED .2 glass protected by a ÿat window so there are a total of nine components with 14 antireÿection-coated air-glass surfaces. This 400 mm lens is mounted by a supporting foot close to
60000 Twilight 16 Mar 1997 11:07 UT

50000

Scattered blue skylight

1000 Flux (R) Na [OIII] [OI] Na 500 Na H Na 250 Na Na [OI]

H [NII]
Flux (R)

40000

750

Hg

30000

20000

NaD ~2000 R Twilight Airglow

[SII] OH

OH
10000

0 4500

5000

5500

6000 6500 (å)

7000

7500

8000

0 4000

5000

6000 (å)

7000

8000

9000

FIG. 5.õSpectrum of the North American Nebula (NGC 7000) also showing airglow and night sky lines derived from an image recorded in a 1200 s exposure on 1997 March 11. This spectrum was extracted through a region showing emission in Ha and has been bias and ÿat-ïeld corrected and calibrated by comparison to Vega. Typical suburban night sky lines from Na and Hg lighting are indicated and listed in Table 1.

FIG. 7.õTwilight sky spectrum on the morning of 1997 March 16 at 11 : 07 UT showing the twilight sodium airglow. Telluric absorption lines of O and H O are prominent in the near-infrared, and Rayleigh scattering 2 2 enhances the blue. The center of the ïeld was 24¡ above the northeastern horizon at an azimuth of 54¡. Sunrise was 44 minutes later at azimuth 92¡.

2000 PASP, 112 : 801õ808


WIDE-FIELD SPECTROSCOPY

805

FIG. 8.õSpectral image of comet Hale-Bopp recorded in a 600 s exposure at 01 : 43 UT on 1997 April 1. This fully corrected image is centered on the Na D lines, and covers a ïeld 1000 A ] 0¡ . The Na tail is the prominent emission line indicated, and the dust tail is the vertical continuum down the center of the .8 image.

its center of gravity and the standard camera mount coupled to the slit housing at the rear. A light and dust shield covers the assembled instrument. Much wider ïelds are accommodated without changing the spectral resolution by substituting the imaging lens for shorter focal lengths. For example, with a 180 mm focal length objective that matches the collimator exactly, the ïeld of view is 7¡ with a scale of 24A pixel~1 in the ïnal focal plane. .8 The telescope is mounted with its slit oriented northsouth on a German equatorial. Precision 25 cm diameter 360 tooth worm gears manufactured by Edwin Byers Co. are on both axes. They are driven by Aerotech Unidex microstepping motors programmed for 3600 steps per revolution, or 1A per step. The motor controllers are serially interfaced to the telescope computer and maintain an accumulated step count which is the oset in arcseconds from a

reference position. The controllers are also able to run short programs that are used for tracking. The right ascension program maintains a constant sidereal rate which is increased or decreased in response to a feedback loop from the guiding camera, while the declination program moves the telescope incrementally when declination tracking corrections are needed. The motor controllers operate in response to commands from the telescope computer, but when the telescope is tracking they interact with the guiding camera, directly subject to an interrupt from the telescope controller. Field acquisition and guiding feedback come from a tandem camera with a Peltier-cooled CCD. The guiding camera platform is attached to the frame of the spectral imager, but it has an adjustable calibrated oset if needed. Normally we use a short focal length camera lens on a

2000 PASP, 112 : 801õ808


806

KIELKOPF & GRAHAM
2000 1-0 Hale-Bopp Ion Tail 27 Mar 1997 01:36 UT

3. PERFORMANCE While WISPI was designed to study extended sources where its fast optical system is an advantage, it works well on bright stars. The stellar image formed by the objective is of the order of 1 pixel in the ïnal focal plane, and when the slit is open any star imaged within it produces a spectrum. This mode permits a measurement of the throughput without losses at the slit. It is used to produce a ÿux calibration by comparing the measured spectrum of Vega with the absolute measurement of Tug, White, & Lockwood (1977). This instrumental response calibration may be applied to any stellar spectrum, but a correction for losses at the slit or for tracking error has to be applied when the slit is narrow. A representative ÿux-calibrated slit spectrum of Vega is shown in Figure 3. The rms noise in this spectrum is of the order of 2 photons s~1 cm~2 A~1 at 6000 A where the signal is roughly 800. The instrumental response function derived from Vega is shown in Figure 4, given in units of photons s~1 cm~2 A~1 above the atmosphere. The response to a stellar source may be used to transform the signal at each pixel in the spectral image into a ÿux in rayleighs, where 1 R is 106 photons cm~2 s~1 radiated into all directions by an extended source. As a check, the ÿux for Ha in the North American Nebula (NGC 7000) is used as an extended source standard (Scherb 1981). A spectrum from that region calibrated in this way is in Figure 5. The observed Ha signal from this image is 800 R above background compared to the value 850 ^ 50 R given by Scherb (1981). The spectrum in Figure 5 appears dominated by other contributions from suburban scattered light at our observing site and from airglow. Airglow at night is well characterized (Chamberlain 1961, p. 345 ; Roach & Gordon 1973 ; Allen 1976, p. 134). An atlas at comparable spectral resolution is available (Broadfoot & Kendall 1968), and a detailed high-resolution study of the OH emission spectrum has been made by Osterbrock et al. (1996). The ÿuxes we observe are consistent with those reported previously and support the calibration of the instrument using Vega. Table 1 lists for reference the prominent urban lighting night sky lines which we routinely see with their wavelengths and ÿuxes above the Na continuum. These values were extracted from a zenith exposure on a typical moonless night, recorded at Moore Observatory, an isolated site 22 km from the center of Louisville, Kentucky. All of the observed lines are attributed to Hg vapor and high-pressure Na lamps. Wavelengths for Hg are from the critical compilation of Crosswhite (1972) and for Na are from the wavelength tables of Striganov & Sventitskii (1968, p. 231). Low-pressure Na is not used in this area, and consequently all exposures show the very broad reversed Na D line. The ÿuxes at the peaks of the broad continuum are given for this line in the table, but most of the line strength is redistributed into the far wings and not represented by this measure+

Flux (R)

1000

2-1 3-2 3-3 0-1

1-1

1500

CO

0 4000

5000

6000 (å)

[O I] 7000

500

2-3 + H2O

8000

9000

FIG. 9.õSpectrum of comet Hale-Bopp recorded in a 600 s exposure at 01 : 36 UT on 1997 March 27. The data shown are extracted from the ion tail region with night sky background subtraction and ÿux calibration on Vega.

Santa Barbara Instrument Group (SBIG) ST-5 camera as the guider. The ST-5 incorporates a Texas Instruments TC255 CCD with 320 ] 240 10 km square pixels. This is convenient because a lens with a focal length of 180 mm provides 11A pixel~1, almost the same angular scale per .5 pixel as the 400 mm lens provides on the spectroscopic CCD. The ST-5 tracking software identiïes tracking errors to subpixel accuracy. In this mode the acquisition ïeld is 1¡ ] 0¡76, smaller than the ïeld of the spectroscopic tele.0 . scope, but large enough to identify the target. A photograph of the instrument is in Figure 2. The cameras used by WISPI run on a dedicated Windows 98 or Windows NT system with vendor-supplied software. The camera-control computer uses a Linux system ïle server and exports its display to remote XWindows systems using a Virtual Network Computing (VNC) server. VNC is a Java-based utility that enables remote network computers with Java-enabled Web browsers to assume control of remote Windows or Windows NT systems on a TCP/IP network.1 The WISPI pointing and tracking control is based on a Forth system developed for a 0.5 m telescope (Kielkopf & Hinkle 1987). For the work shown here, it was run with Laboratory Microsystems WinForth on the camera control computer. A WISPI control program for Linux is under development.
õõõõõõõõõõõõõõõ 1 VNC is free software developed by AT&T Laboratories, Cambridge, and is available from http ://www.uk.research.att.com/software.html.

2000 PASP, 112 : 801õ808


WIDE-FIELD SPECTROSCOPY
TABLE 1 URBAN LIGHTING NIGHT SKY SPECTRAL LINES (4490õ8500 A) j (A) 4816.07 5460.73 5769.60 5789.66 5790.66 4494.18 4497.66 4541.63 4545.19 4664.81 4668.56 4747.94 4751.82 4978.54 4982.81 5148.84 5153.40 5682.63 5688.19 5688.20 5889.95 5895.92 6154.23 6160.75 7373.23 7373.49 7809.78 7810.24 8183.26 8194.79 8194.82 Peak Flux (R) \3 1140 440 400 165 \40 175 105 575 150 Blend Blend Blend Blend Blend Blend Blend

807

Element Hg ......

Comment

Na ......

1060 530 525 200 23 15

Blend Very broad and reversed blend Very broad and reversed blend Blend Blend and OH Blend and OH

145

Blend

(Osterbrock et al. 1992). The weak He line at 7065 A has a ÿux of about 0.02 Ha in NGC 7000 but is considerably larger at 0.10 in Orion. Figure 7 is the spectrum of a dawn twilight clear sky, recorded in a 600 s exposure with the Sun 7¡ below the .5 horizon at midexposure. The image from which this spectrum was derived is remarkable in the presence of a bright narrow Na D emission. As the calibrated spectrum shows, the ÿux in this line is of the order of 2000 R above the nearby continuum. This "" sodium ÿash îî (Roach & Gordon 1973 ; Chamberlain 1961, p. 345) occurs when solar radiation is optimal to excite resonance ÿuorescence from the layer of mesospheric atmospheric Na at an altitude of about 95 km. A laser guide star produced by ÿuorescence in this layer is a useful reference for adaptive optics imaging (Lloyd-Hart et al. 1998). Spectral images recorded in twilight skies have a background similar to this, although generally much weaker and lacking the sodium ÿash. A spectral image of comet C/1995 O1 (Hale-Bopp) in Figure 8 shows 1 of a processed full frame. The slit was 4 oriented north-south, 0¡ 8 antisunward from the nucleus. .4 Night sky lines run the full width of the image. Emission from the prominent Na tail falls in the reversal dip of the broad urban scattered light proïle. The scattered Fraunhofer continuum from the dust tail runs down through the center of the frame, and emission lines of CO` ions are strong. This image shows the ability of WISPI to distinguish spatially localized spectral features. A full spectrum of the ion tail of Hale-Bopp is in Figure 9, where several CO` bands and weaker lines of H O` and 2 [O I] are labeled. Subtraction of the o-tail background in the wide-ïeld image allows detection of features of the order of 5 R in 600 s exposures such as this one. 4. CONCLUDING COMMENTS The wide-ïeld spectroscopic imager described here has a useful coverage from 4000 to 11000 A over a 3¡ ïeld of .2 view with its 400 mm focal length objective. WISPI is best suited for measuring extended emission line sources, but it has the versatility to work well on bright stars. Stellar imaging allows the derivation of an instrumental response function which is then used to transform the raw signal from an extended source to a ÿux in rayleighs. A calibration on Vega shows that at Ha the instrumental signal is about 7 ADU (compared to a readout noise of 1 ADU) for an incident exposure of 1 photon cm~2 A~1 above the atmo sphere. Given that the ÿux in this spectral region from a V \ 0 star is about 670 photons cm~2 A~1 s~1, an 1800 s exposure would produce a signal 10] readout noise for a V \ 14.8 star. A 13 R monochromatic object also produces a signal of about 10 ADU in an 1800 s exposure. Objects at the 1 R level are detectable by dierencing on- and otarget regions of the same image, with quantitative mea-

ment. The reversed center of the line has a weak secondary central peak which is about 50 R high and to which the airglow Na emission makes a variable contribution, as described below. The D-line wing creates a continuum with signiïcant contributions to the night sky background from about 5500 to 6300 A. We have not seen other lines due to the buer gases in these lamps, but emission from Ne, Ar, and possibly Kr also should be present in late twilight as outdoor lighting ïrst turns on. In spite of the interference from airglow and urban light pollution, very weak features can be extracted by taking the dierence of two parts of a single wide-ïeld image. As an example, Figure 6 shows the result of subtracting a region in NGC 7000 in which there is less nebular emission from one where the nebular spectrum is stronger. With careful dierencing, the Na continuum disappears, leaving a dierential readout noise of the order of 1 R. The usual H II region emission lines are indicated. The observed ratios for the [N II]: Ha :[N II] lines are 0.078 : 1.00 : 0.24, about the same as 0.071 : 1.00 : 0.21 reported for the Orion Nebula

2000 PASP, 112 : 801õ808


808

KIELKOPF & GRAHAM and his advice on design goals and to Jeremy Raper for his help during the ïrst ïeld tests. This work was funded by grants from the Air Force Office of Scientiïc Research and from NASA through the Kentucky Space Grant Consortium.

surements of fainter emission requiring exposures of several hours. We are grateful to Dr. Frank O. Clark of the Air Force Research Laboratory for discussions that led to this project

REFERENCES
Allen, C. W. 1976, Astrophysical Quantities (London : Athlone) Broadfoot, A. L., & Kendall, K. R. 1968, J. Geophys. Res., 73, 426 Chamberlain, J. W. 1961, Physics of the Aurora and Airglow (New York : Academic) Crosswhite, H. M. 1972, in American Institute of Physics Handbook (New York : McGraw-Hill), 7-29 Gaustad, J. E., McCullough, P. R., & Van Buren, D. 1996, PASP, 108, 351 Gaustad, J., Rosing, W., Chen, G., McCullough, P., & Van Buren, D. 1997, BAAS, 190, 30.05 Hentges, P. J., & McCullough, P. R. 1998, BAAS, 192, 40.05 Kielkopf, J. F., & Hinkle, S. 1987, PASP, 99, 442 Lloyd-Hart, J., et al. 1998, ApJ, 493, 950 McCullough, P. R. 1997, AJ, 113, 2186 Meaburn, J. 1976, Detection and Spectrometry of Faint Light (Dordrecht : Reidel) Osterbrock, D. E., Fulbright, J. P., Martel, A. R., Keane, M. J., & Trager, S. C. 1996, PASP, 108, 277 Osterbrock, D. E., Tran, H. D., & Veilleux, S. 1992, AJ, 389, 305 Reynolds, R. J., Hausen, N. R., Tufte, S. L., & Haner, L. M. 1998, ApJ, 494, L99 Roach, F. E., & Gordon, J. L. 1973, The Light of the Night Sky (Dordrecht : Reidel) Sandage, A. 1976, AJ, 81, 954 Scherb, F. 1981, ApJ, 243, 644 Striganov, A. R., & Sventitskii, N. S. 1968, Tables of Spectral Lines of Neutral and Ionized Atoms (New York : Plenum) Szomoru, A., & Guhathakurta, P. 1998, ApJ, 494, L93 Tufte, S. L., Reynolds, R. J., & Haner, L. M. 1998, ApJ, 504, 773 Tug, H., White, N. M., & Lockwood, G. W. 1977, A&A, 61, 679

2000 PASP, 112 : 801õ808