Table 1:
Selected Spectral Lines Available to SPIFI
Current Status:
Galactic Center
NGC 253 & M82
The Future
Array Status
Wavelength Coverage
Resolving Power
Array Status
Wavelength Coverage
Resolving Power
Sample Calculation
of Integration Time for JCMT Observations
SPIFI Design
Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6

NOTE: To properly view this page the MS font "Symbol" is needed.

The South Pole Imaging Fabry-Perot Interferometer (SPIFI)

PI: Gordon J. Stacey, Department of Astronomy, Cornell University

C.M. Bradford1, G.J. Stacey1, M.R. Swain1, J.A. Davidson2, M. Savage2, J.M. Jackson3, and A.D. Bolatto3

1Department of Astronomy, Cornell University, 2SETI Institute, NASA/Ames Research Center 3Department of Astronomy, Boston University


We have designed and constructed and commissioned the South Pole Imaging Fabry-Perot Interferometer (SPIFI). SPIFI is a direct detection imaging spectrometer designed for use in the far-IR and submm (200, 350, 450 mm) windows available to the 1.7 m Antarctic Submillimeter Telescope and Remote Observatory (AST/RO) at the South Pole, and in the submm (350 and 450 mm) windows available to the James Clerk Maxwell Telescope (JCMT) on Mauna Kea in Hawaii. SPIFI's detector is a 5 5 element array of monolithic silicon bolometers cooled to 60 mK in an adiabatic demagnetization refrigerator. SPIFI uses free standing metal mesh Fabry-Perot interferometers to deliver spectroscopic images at resolving powers, R Dl/l as large as R = 10,000 (or a velocity resolution, Dv as small Dv = 30 km s-1) over the entire array. The resolving power can be set to any value between R = 1000 and 10,000 (Dv = 300 to 30 km s-1) in a few minutes time at the telescope. Higher resolving powers (better than 20,000, or Dv < 15 km s-1) are possible for the inner 9 pixels. The beam sizes at 370 mm are ~ 7" and 55" on the 15 m JCMT and the 1.7 m AST/RO telescopes respectively. The array is a laid out on a square grid and delivers 35" 35" and 300 300" fields of view respectively on the JCMT or AST/RO telescopes.

At present, SPIFI can tune to any frequency in the 350 micron window, and continuously scan 13 spectral resolution elements at any given wavelength. It is also possible to set up in the 450 micron window. In the near future, we expect to be able to easily switch between the two windows while the instrument is cold. Important astrophysical lines in the 350 micron window include the 371 micron [CI] fine structure line, and the rotational transitions of CO (J = 7 6) (372 mm), HCN (J = 10 9, and 9 8) (338 and 376 mm), and HCO+ (J = 10 9, and 9 8) (336 and 374 mm). For the first observing runs, we will focus on the [CI] and CO (7 6) lines to explore the physical conditions of the gas and gas dynamics in the Galactic Center, Galactic starformation regions and external galaxies. The brighter lines that are available to SPIFI are listed in Table 1 below.

SPIFI was developed at Cornell University under NASA grants NAGW-3925 and NAG5-4503. SPIFI is a collaborative venture between individuals at Cornell, Boston University and the SETI institute. Our group welcomes scientific collaborations with other JCMT users. Please contact Prof. G. J. Stacey at Cornell University (stacey@astrosun.tn.cornell.edu) to arrange collaborative efforts.

SPIFI is more fully described in: (1) Stacey, G.J., Bradford, C.M., Swain, M.R., Jackson, J.M., Bolatto, A.D., Davidson, J.A., & Savage, M. 1996, ESA SP-388, and (2) Swain, M.R., Bradford, C.M., Stacey, G.J., Bolatto, A.D., Jackson, J.M., Savage, M. & Davidson, J.A. 1998, Infrared Detectors and. Instrumentation, SPIE Proceedings, 0000, 000 (1998).

Table 1: Selected Spectral Lines Available to SPIFI
A (s-1)

1Excitation potential, i.e. energy of upper level above the ground state.
2Molecules: Collision partner H2 (100 K). Atoms: [CI] (H atoms), [NII] (electrons).

Current Status: First Light at JCMT in April 99

First light with SPIFI was obtained on the JCMT in late April, 1999. Figure - shows some photos of the run. We had planned a series of CO (76) (371 mm) and [CI] 3P23P1 (370 mm) observations of the Galactic Center and external galaxies. Being conservative on our first run, we focused initially on the CO (76) line since it is easier to set up with the instrument. The first three shifts were spent as engineering and checkout - measuring telescope efficiencies and flux calibration, (on the Moon, Mars, and Jupiter), obtaining beam shapes and array orientation (on Mars), verifying the conversion between CSO t225GHz and the 370 mm opacity (via telescope tips), and estimating system sensitivity. We measured a main beam efficiency hMB ~ 30%, a beam size ~ 7", and a forward coupling of ~ 65% in agreement with standard JCMT values. Telluric transmission was marginal (tsky(zenith) ~ 6 to 17%) at best. In this weather, we mapped the Galactic Center and the cores of a few galaxies in the CO (76) line --- these results are briefly described below. Unfortunately, the evening that we switched over to observe in [CI], the weather became very poor: t225 GHz >0.1 (hsky(370 mm) < 3.8%). The weather deteriorated for the remainder of the run, so that we only obtained CO(76) line observations.

Our system sensitivity was very close to what we expected from laboratory measurements. For example, the best pixels in our CO (76) observations of NGC 253 (see below), taken with hsky ~ 8.5% have an rms error in the main beam brightness temperature TMB (rms) ~ 0.40 - 0.5 K in 132 seconds of integration time per point, so that:

TMB (rms) = 2 Tsys k /{hMB (DnDt) 1/2} Tsys ~ 10,000 - 12,500 K

where Dn is the bandwidth of the spectrometer, Dt is the integration time, and k is the backend degradation factor (= 1.15 for a 2-bit digital correlator). Scaling these values to a good night, hsky ~ 40%, we expect Tsys ~ 2000 K (see Figure -). For comparison with heterodyne receivers, we calculate an equivalent double side-band receiver temperature, Trec (DSB) :

Tsys = 2[Trec (DSB) + htelTsky (1-hsky) + (1- htel)(Ttel]/( hskyhtel)

Taking htel = 65%, Tsky = 270 K, Ttel = 265 K, and hsky = 8.5%, we have:

Trec (DSB) < 100 K

Results: Galactic Center

For the CO (76) observations presented here, we used a velocity resolution ~ 70 km s-1 (R = l/Dl ~ 4300), and scanned 8.6 resolution elements (600 km s-1), sampling at 50 km s-1 intervals. At present the array has a 4 4 pixel footprint with one pixel missing, and three not operating properly. We plan to replace the nonfunctional pixels and have the rest of the 5 5 array installed at Goddard this summer.

Despite the mediocre weather, SPIFI's good sensitivity and imaging capabilities enabled us to map the entire circumnuclear ring (CNR, roughly 200 spectra) in the CO (76) line. The map consists of 13 pointings of the array - these "footprints" are delineated in
Figure -. Typical integration times were ~ 12 minutes, or about 60 seconds per spectral sample. Since the data were obtained very recently, they are not yet properly reduced, and only a single array footprint is presented (Figure -). This was obtained at the south-westernmost regions of the ring - a region where both the [OI] 63 mm line and HCN line emission peak (cf. J1) - and known to be bright in CO(76) from previous large beam measurements (H1). Strong ring emission is apparent over the entire region mapped. It is anticipated that these high spatial resolution maps will reveal shocks within the clumpy ring, and at the regions where the "Northern Streamer" crossed the ring in the north-west. We are eager to observe the same regions (and more) in the [CI] 370 mm line thereby tracing the mass and extent of the ring, and when combined with the CO map, providing a better delineation of clumps and shocks. As pointed out by Serabyn et al. 1995 (S3), both the 610 and 370 mm [CI] lines are good tracers of the molecular ring: both lines are optically thin in the ring and largely so in the foreground gas along the line of sight as well. Foreground absorption greatly hampers studies of the CNR in the lower J CO lines. Since the 370 mm line occurs between excited states of Co, it is especially immune to problems with foreground absorption.

Results: NGC 253 and M82

We observed several galaxies in the CO(76) line, and despite poor transmission, had some success.
Figure - is a preliminary reduction of the line emission seen in a single footprint centered on the nucleus of NGC 253. These data were obtained when telluric transmission towards the source was ~ 10%. The total integration time was ~ 25 minutes, or 132 seconds per spectral sample. NGC 253 is the best example of a nearby (D ~ 2.5 Mpc) spiral galaxy with a starburst nucleus. It is a very dusty, highly inclined (i ~78 system containing a massive molecular bar within which a massive starburst has recently occurred.

Harris et al. (1991) previously detected bright (TMB ~ 5 K) CO(65) line emission from the nuclear regions. More recently, Israel et al. (1994) mapped the CO(43) line emission over a broader region. Our data indicates the high J emission is also widespread - the line is clearly detected over most of the array footprint. The brightness of the CO (76) line relative to the lower J lines is a sensitive indicator of the gas density and UV field strength (Kaufman et al. 1999). The bright CO (76) line indicates the emitting gas is remarkably warm (Tgas ~ 100 K), and dense, n ~ 4 104 cm-3 - highly excited by starburst. Starbursts may be self-limiting, in that the effect of the large UV fields is to heat and fragment the ambient molecular ISM thereby inhibiting further star formation. Figure - shows our initial reduction of a pointing toward the other prototypical starburst galaxy, M82. The CO (76) line is substantially weaker than in NGC 253, with TMB peaking around 1 K, consistent with the lower-J lines observed in this source by Wild et al. (1992).

We plan to add substantially to these maps in the near future. We will map the inner 35" 105" regions of the two sources in both the CO (76) and [CI] 370 mm lines. The [CI] 370 mm observations can ascertain the origins of the enhanced C0 / CO abundance observed in these starburst nuclei (H3): PDR origins would mean strong 370 mm line emission since the gas in PDRs is warm. Cosmic ray enhancement of abundances would not lead to enhanced 370 mm line emission since cosmic rays are much less effective at heating the gas. Furthermore, if the [CI] emission arises from PDRs, we expect very good correspondence between the CO (76) line (tracing warm dense PDRs) and the [CI] 370 mm line.

The Future

There are several improvements will be made to SPIFI in the near future: 1) Fill out the 5 5 footprint. This is top priority, and should be done this summer. 2) Improve the transmission of the JCMT relay optics. A 30% gain in performance can be realized with an AR coating of these lenses, and better optical alignment. 3) A 20% improvement may result by eliminating our 2nd long-pass filter. 4) Stray light in the dewar is equivalent to the background at high resolving powers. Eliminating this light should result in a factor 2 improvement in sensitivity. 5) Enable observations in the 200 and 450 mm windows 6) Large format SQUID multiplexed superconducting bolometer arrays now exist. These arrays are planned to have formats as large as 32 32 pixels. These arrays would greatly increase the utility and mapping efficiency of SPIFI: -- We can subsample (l/2D pixels) the SPIFI focal plane yielding instantaneous fully sampled images. The overall field of view would be a factor of 2 larger. -- Smaller pixel size (1 1mm) changes the plate scale. The smaller f/# would permit the array to be placed within the main SPIFI cryostat, resulting in: -- Near unit filling factor array is much more efficient for extended sources than our Winston cones. -- Much easier optical alignment since all optical elements are on one bench. -- A milli-K pupil where an aperture stop could be installed - stray light within the dewar becomes much less of an issue. A small filter wheel could be installed at this pupil making for easier changes of submm window. -- The ADR salt pill and 3He thermal guard are moved into the main SPIFI cryostat, likely reducing the 4He requirements for SPIFI by a factor of two.

Array Status: For the past year, we operated with 9 pixels (3 3) installed in the array. The array is now down at GSFC having the additional 16 pixels installed.

Sensitivity: In the lab, we have achieved system noise equivalent powers, NEPs ~ 9 10 -16W Hz-1/2 @ 370 mm and R = 6000 (Dv = 50 km s-1). The NEP defined here is the noise power in a 1 Hz electrical bandwidth ( second integration time) and includes the factor of 2 loss due to chopping. For a point source, the noise equivalent flux (NEF) is given by the NEP/(Ahahsky), where A is the surface area of the telescope, hsky is the transmission of the sky along the line of sight to the source and ha is the aperture efficiency. For the 15m JCMT, ha = 20% at 370 mm, so that assuming hsky = 40% we have: NEF ~ 6.3 10-17W m--2Hz-1/2. At present, SPIFI is not background limited except at low resolving powers, so that the NEP and NEF are roughly independent of R It is difficult to compare SPIFI to heterodyne receivers since the figure of merit for heterodyne receivers is the double side band receiver temperature, Trec(DSB), which translates in very subtle ways to our NEP (see below for more information). The major complication is that our quoted NEP includes contributions due to the thermal background, but usually receiver temperatures quoted in the literature do not. Therefore, at the telescope the relevant noise for heterodyne receivers is Trec + Tbac, while for a background limited system such as SPIFI at low resolution, Trec 0, so that the relevant noise is Tbac.

Wavelength Coverage: At present, SPIFI optimized to run in the 350 mm window, and our bandpass filter (courtesy Dr. Peter Ade, Queen Mary and Westfield College, London) enables observations between 345 and 389 mm with good sensitivity. Notice, however, that telluric transmission gets poor beyond 380 mm, and there is a strong telluric feature near 359 mm.

Resolving Power: In its current configuration, SPIFI can achieve resolving powers between ~ 1000 and ~ 10,000 (velocity resolutions between 300 and 30 km s-1) across the entire array. About 13 resolution elements may be scanned continuously centered on any wavelength in the 345 to 389 mm band. Longer spectra (up to the width of the bandpass filter (44 mm) can be obtained in a straightforward manner by piecing together 13 resolution element scans.

SPIFI's Features: Goals

Array: The entire 5 5 array should be installed by December 1, 1998. In the more distant future, it is possible to install a significantly larger array to enable subsampling of the beam and imaging a larger field of view.

Sensitivity: The best NEP we expect to achieve at the South Pole is ~ 4.2 10-16 sqrt(1000/R) W-Hz-1/2. For the JCMT, the limit is about a factor of 1.7 higher. These are the background limited system NEPs based on the known system transmission and quantum efficiency.

Wavelength Coverage: In the near future, we plan to enable observations with SPIFI in either the 350 or 450 mm windows without warming up the system. This is a goal for the planned spring 1999 run at JCMT. For polar operations, we plan to be able to switch quickly between the 200 and 350 mm windows to make best use of the instrument. A more challenging task on the horizon is to enable switching between the 200, 350, and 450 mm windows while cold.

Resolving Power: We have designed and are constructing a next generation Fabry-Perot that will be able to continuously scan many (> 30) resolution elements without resetting the scan center. It is hoped that this new Fabry-Perot will be available for the planned spring 1999 run at the JCMT. We are also investigating modifications that will enable changing the resolving power of the instrument from 1000 to 20,000 while cold (velocity resolutions between 300 and 30 km s-1). Note, however, that only the inner 9 pixels are usable at the highest resolutions.

SPIFI Sensitivity

The natural units for a direct detection spectrometer such as SPIFI are the NEP and NEF as defined above. Currently SPIFI is only background limited at modest resolving powers (R < 3000), so that the NEP is only weakly dependent on R: NEP = 7.48 10-16(sqrt(1.28 + 1000/R)) W-Hz-1/2. In the near future, through improvements in the cold electronics and system transmission, we expect to become background limited at all resolving powers, so that NEP R-1/2. For point sources, the minimum (1s) detectable line flux (MDLF) then given by:

MDLF = sqrt(N)NEP/{(AhtelhMBhsky)sqrt(2tint)}

Where N is the number of spectral resolution elements in the scan, htel is the receptive efficiency of the telescope (= 65%, JCMT web page), and hMB is the fraction of the forward beam that couples into the main beam (= 30%, JCMT web page). Note that ha = htelhMB. The transmission of the atmosphere along the line of sight, hsky is given by: hsky = exp(-tzenithsec(q)), where tzenith is the zenith opacity, and q is the angle from the zenith to the source.

It is instructive to compare our sensitivity to current heterodyne receivers. It is difficult to compare the two directly since the figure of merit for heterodyne receivers is Trec(DSB), which translates in very subtle ways to our NEF listed above. {The primary complication is that the NEF includes the shot noise contribution of the background Tbac, while Trec does not. Therefore, at the telescope the relevant noise for heterodyne receivers is Trec + Tbac, while for a background limited direct detection system, like SPIFI at small R, Trec 0, so that the relevant noise is Tbac}. A more straightforward method, is to convert SPIFI's NEF to a system temperature on the telescope by comparing the detected power to a main beam rms noise, TMB(rms), then converting TMB(rms) back to Tsys. From the JCMT web page:

NEF [W-m-2Hz-1/2 ] = 2kTMB(rms)/l2{DnWbeam}
TMB(rms) = 2Tsysk/(Dnt)1/2/hMB

where Wbeam is the beam solid angle (= 1.28 10-9 sr at 370 mm), Dn is the resolution bandwidth (Dn = p/2n/R = 1.29 109 Hz, for R = 1000), k is the backend degradation factor (= 1.15 for a 2-bit digital correlator), and the integration time, t, is second. For hsky= 40%, hMB = 30%, and htel = 65%, SPIFI has an equivalent Tsys = 800 K for R = 1000 at 370 mm. If SPIFI were background limited at all resolutions, Tsys would be independent of R (since TMB (Dn) R1/2 for that case). We are working towards, but have not yet reached this goal. Figure 1 below plots the expected Tsys for SPIFI on JCMT as function of R for both today's system and the limit we expect to reach (the "goal" system). For the "goal" system, SPIFI's sensitivity will improve by a factor of 1.5 at the lowest resolutions, and up to a factor of 3.7 at R ~ 10,000. This Tsys gives the minimum detectable line flux per resolution element if we do not spectrally scan. This is only appropriate for very deep line searches where the line to continuum ratio is expected to be large. Normally we will need to spectrally scan, typically 5 resolution elements to form a spectrum, so we also plot the effective Tsys for this case as well.

Recall that Tsys is related to TA*(rms), TR*(rms), and TMB*(rms) by (cf. JCMT web page):

TA* = 2Tsysk/(Dntint)1/2;
TR*(rms) = TA*(rms)/hfss;
TMB(rms) = TA*/hMB.

The "antenna temperature", TA*(rms) is that actually measured at the telescope after corrections for telescope and antenna losses, while TR*(rms) is the internationally accepted scale obtained by correcting the forward spillover and scattering efficiency, hfss (~ 0.60, JCMT web page).

Sample Calculation of Integration Time for JCMT Observations

Suppose I wished to map a region in a source in the 370 mm [CI] line. The source is at 1.2 air masses and the zenith transmission of 40%, or tzenith = 0.92. If the source is best modeled as a distribution of point sources, then the relevant value is TMB(rms) (or MDLF), if it is more extended, then the relevant value is TR*(or MDLF/(hfiss/hMB). I wish to run at a resolving power of 2,000 (Dv = 150 km s-1), scan 5 resolution elements, and spend one hour at each pointing. Plugging in relevant values I have: NEP = 1.00 10-15 W Hz-1/2, and hsky = 0.33, so that MDLF = 2.3 10-18W m-2 (1s, 1 hour, 5 resolution element scan).

In terms of temperatures, since R = 2000, I find Tsys= 1020 K for no spectral scanning at the zenith for a zenith transmission of 40%, or tzenith = 0.92. Making the correction to 1.2 airmasses, Tsys = 1225. I can correct for scanning by realizing that scanning 5 resolution elements means that the effective integration time per element is 1/5 the total integration time (or 12 minutes). Making this correction, I obtain (since Dn = p/2n/R = 6.36 108Hz), TA*(rms) = 4.2 mK, TR*(rms) = 6.9 mK, and TMB(rms) = 13.9 mK. These values are 1s, per resolution element. The scan width is 5 resolution elements, and the total integration time is 1 hour.

Each of these values include chopping losses, but do not include overheads due to acquiring the source, nod delays etc, which will likely entail an additional factor of 1.5 or so. So, given these parameters, you should expect the noise values in the last paragraph in a total wall clock time of roughly 60 1.5 = 90 minutes (Figure 1).

SPIFI on JCMT Compared with SPIFI on AST/RO

For any given instrument sensitivity, SPIFI on JCMT will be ~ 12 to 16 times more sensitive to point sources than SPIFI on AST/RO. (The JCMT has 80 times the surface area as AST/RO, but ~ 25 to 30% of the aperture efficiency for point sources, and the atmosphere is ~ 1.5 times more transmissive at the South Pole.) However, the SPIFI - AST/RO system is ~ 3 times more sensitive to extended sources than the SPIFI-JCMT system since the AST/RO telescope has ~ twice the aperture efficiency for extended sources, and the atmosphere is ~ 1.5 times more transmissive at the South Pole.


Nearly all submm spectroscopy to date has been done with heterodyne receivers. These receivers will remain the spectrometers of choice for very high velocity resolutions: Dv < 5 10 km s-1. However, recent advances in bolometer technology permit background limited direct detection to similar resolutions. It can be shown that direct detection is in principle more sensitive than phase sensitive schemes (cf. Harris, A.I. 1990, in "From Ground-Based to Space-Borne Sub-mm Astronomy", ESA SP-314, 165).

Monochrometers such as grating spectrometers and Fabry-Perot interferometers (FPIs) are the instrument of choice for background limited systems. However, for spectral resolutions > 10,000 at 370 mm, a single pass grating needs to be > 1.8 m long! Clearly a multiple pass instrument such as a FPI is preferred.

SPIFI Design

SPIFI was designed to mount on the azimuth arm pedestal of AST/RO accessing an f/8.3 focus (Figure 2). With a polyethylene relay lens, SPIFI accesses the f/15 Naysmith focus of the JCMT (Figure 3). Figure 4 illustrates the SPIFI optical design. The beam transits a calibration module containing a gas cell and blackbody for spectral calibration and flat fielding, reaches a focus just outside the SPIFI, then enters the main dewar via a polyethylene window. The beam is collimated at 9 cm by M1, sent through a scatter filter, the high order FPI (HOFPI), and the Lyot stop, then decollimated by M4 and sent through the low order FPI (LOFPI) and the filter wheel (containing fixed 1st order FPI's) before entering the NASA Ames adiabatic demagnetization refrigerator (ADR). The final f12.6 beam passes through low and high pass filters at 300 mK before imaging on the Winston cones leading to the bolometer array.

SPIFI attains very high spectral resolution with good spectral purity by using three FPI in series. Each has a free standing metal mesh etalon, with finesse ~ 30 to 60, and transmissions ~ 80 to 60% . The LOFPI and HOFPI are fully tunable, and use flex vane translation stages with PZT pushers for spectral scanning. The HOFPI also employs a coarse translation stage enabling gross changes in spectral resolution (e.g. R ~ 1500 to 10,000) while SPIFI is cold. A capacitive bridge is employed for measuring and stabilizing the plate spacing.

The detectors are Goddard monolithic silicon bolometers kept at ~ 60 mK with the AMES ADR. The bolometers have detector noise equivalent powers < 5 10-18 W Hz-1/2, a few times smaller than the noise from the thermal background radiation referred to the detector.