Version 1 (modified by trac, 16 months ago) ( diff )


Instrument upgrades / New instruments

Fiber feed from 3.5m to SDSS/APOGEE



We propose to initiate a project to implement a fiber optic network at Apache Point Observatory such that the existing (and future) spectrographs on site can be supplied with target light from any of four telescopes. The first phase of this implementation will be to develop the site infrastructure to connect all of the telescopes through the addition of a new utility pathway running from the ARC 3.5m past both of the small aperture telescopes and terminating at the SDSS telescope. This will be accomplished through a combination of below-ground conduits, above-ground cable trays, and existing cable routing paths available at each telescope. This infrastructure will be used to implement a fiber run between the SDSS/APOGEE spectrograph, the ARC 3.5 m telescope, the ARCSAT 0.5 m telescope, and the NMSU 1 m telescope. We propose to route 270 of the 300 available APOGEE fibers to the TR1 bent Naysmyth port of the 3.5 m telescope. Of these 270 fibers, 217 will be grouped together to form a integral filed unit ~32 arcsec size on sky. The remaining 53 fibers will be placed outbound of the IFU to sample sky. The combination of the 3.5m aperture and the APOGEE spectrograph used as an integral field unit will open exciting new science opportunities at a relatively low cost for the capability, leveraging the existing APOGEE instrument. The remaining 30 APOGEE fibers will be routed to the two small aperture telescopes, with 15 fibers going to the focal plane of the ARCSAT 0.5 m telescope for a future instrument application and the remaining 15 will be routed to the focal plane of the NMSU 1 m telescope where they will replace and improve the existing fiber run.

The full paper can be downloaded here.

Assumed Parameters for APOGEE:

  • R=22,500 from 1.51 - 1.70 um, 300 2" dia fibers in SDSS 2.5m f/5 focal plane
  • ~15% throughput --> S/N = 100/pix for H > 12.2 in t = 3 hrs
  • radial velocity to 100 m/sec

Possible 3.5m Feed Configurations

  • Single fiber: 0.7" at F/10, 1.4" at f/5 (i.e., with focal reduction)
  • IFU 1.4" spaxels, 217 elements, ~39" across, lenslet array in front for ~100% fill factor and f/5 conversion
  • MOS ~ 200-300 1.4" fibers (with lenslets) in 8' FOV (pretty tightly packed!) or ~30 Randomly targeted objects (e.g in 7-fiber bundles) in 8' FOV.

Science Cases Overview

John Bally, Kevin Bundy, Jon Holtzman, Don York, Jennifer Sobeck


  • Nearby Galactic Star and star clusters [Bally ~ See attached]
  • Nearby HII regions and post-main-sequence objects [Bally ~ See attached]
  • Young Massive Clusters (YMCs) and Super Star Clusters (SSCs) [Bally ~ see attached]
  • Massive Stellar Transient Nearby Galaxies [Bally ~ See attached]
  • Individual abundances in globular cluster stars from near-IR, perhaps from an IFU [Holtzman]
  • Integrated light of globular clusters
  • Dwarf spheroidals [Holtzman, Sobeck]
  • Dwarf galaxies [Bundy, Holtzman]
  • Extragalactic (Emission lines, Stellar populations, and Stellar Dynamics) [Bundy ~ See below]

Single object science

  • Abundances of Hipparcos sub-giants [Holtzman]
  • Radial velocity monitoring of late-type stars
  • supernova followup in H band
  • Survey of B[e] stars [Chojnowski]


  • The nature of star and clusters in the Central Molecular Zone (CMZ) and nearby starburst galaxies [Bally ~ See attached]
  • Blind Emission-Line Searches of Deep Extra-Galactic Fields [Bally ~ See attached]

Future Fiber Feeds

  • A Visual-Wave fiber bundle for the APO 3.5 meter [Bally ~ See attached]
  • Visual wave band IFU observations of nebula [York ~ See below]

Detailed Science Cases

A number of science cases have been put forward by John Bally in an attached document "Science Case for a~300 fiber connection from the APO 3.5m to the SDSS spectrograph" for details please see the attachments of this page.

Single object possibilities

A single-object 3.5m feed is desirable for objects that are sufficiently spaced in the sky such that observing with the 2.5m does not offer benefits from wide field and for specific single objects of interest. The 3.5m would also potentially allow observations of fainter objects than can be done with the 2.5m. Note that the throughput gain comes from telescope area (3.5/2.5)2 = 1.96 (0.73 mag fainter) and also from the fact that the corrector at the 2.5m is estimated to absorb ~30% of the light in the H band (another 0.3 mag). On the other hand, there might be some additional light losses from a longer fiber run, especially if it were to involve multiple connectors. Also to avoid seeing losses that would be larger with the 1.4" fibers even at F/5 from the 3.5m, one would want to consider a lenslet array in from of a fiber bundle, such that light was collected over multiple fibers without any loss between fibers. This might lead to additional throughput gains. But ballpark, a 3.5m feed would allow observations 1 mag fainter than at the 2.5m in the same time, e.g. S/N=100/pixel at H=13.2 in 3 hours, or S/N=100/pixel at H=11 in 1 hour.

One particular application might be observations of stars identified as subgiants once GAIA parallaxes become available. The great advantage of subgiants is that, with known distances, accurate ages can be determined. Ages in conjunction with the chemical abundances provided by APOGEE spectra provide a powerful tool for studying Galactic evolution. Subgiants are intrinsically fainter and bluer than the giants that are the targets of the main APOGEE survey. At M(H)~2, the 3.5m feed would be able to get good abundances to a distance of ~2 kpc. Of course, to build up a sample of a reasonable number of stars, one at a time, would require a significant amount of observing time.

Another application might be a survey of B[e] stars, which are B-type emission line stars that differ from classical Be stars due to the presence of forbidden emission lines and strong IR excesses. These features are attributed to a circumstellar dust component not present in the case of classical Be stars. The ~100 or so known B[e] stars are a heterogeneous group often found to be supergiants (sgB[e]), pre-main sequence (HAeB[e]), or compact planetary nebulae (cPNB[e]). However, the fact that ~50% of them remain unclassified (unclB[e]) is a testament to the complexity and richness of the emission line spectra, and also to the difficulty of conducting a uniform survey given the isolated nature of the stars. On average, B[e] stars are almost 4 magnitudes brighter in the H-band than in V-band, making them ideal targets for NIR spectroscopy. For more information, see Lamers et al. 1998, and Miroshnichenko et al. 2007 A + B.


Prepared by Kevin Bundy


A 3.5m APOGEE feed coupled to large fiber-IFU could be very exciting in the context of extragalactic science. The overall sensitivity is a key challenge, but is worth more thought. Several science cases could be pursued further, but the instrument’s power for galaxy science could be enhanced if some changes in the spectrograph setup (wavelength range and resolution) were possible, namely a somewhat lower resolution and coverage in the J-band.

Emission Lines:

The current APOGEE wavelength range misses many strong emission lines, including Pachen Beta (1.282 um), H2, and several [FeII] lines in the J-band. This is too bad, because mapping narrow emission lines at high velocity resolution would be compelling for two reasons: 1) Less sensitivity to dust in the near-IR allows deeper probes of gas in dusty systems (e.g., LIRGS and ULIRGS) and 2) The [FeII] lines are diagnostics of electron temperature, density, and ionization field that arise in shocks and high-ionization regions (e.g., Thompson+95). They are therefore useful probes of AGN, winds, and shocks, phenomena that are often difficult to study again because of dust obscuration at bluer wavelengths (Iserlohe+13).

Stellar Populations:

Obviously, there are many spectral features however that provide chemical diagnostics of integrated stellar populations, including potential IMF-sensitive features that vary on the few percent level with IMF shape (requiring S/N~100, see Conroy & van Dokkum 2012). The spectral resolution required is about 6000-8000, so higher S/N could be achieved by spectral binning. The near-IR wavelengths are relatively unexplored territory for galaxy stellar population work, and APOGEE on the 2.5m provides a valuable “library” for interpreting it.

The main challenge is sensitivity, as shown in estimates below. Following up bright nearby galaxies from the ATLAS3D survey is worth further consideration however. The typical H-band surface brightness of these sources at 1 Re is about 20 mag per sq.arcsec, compared with APOGEE’s target of muH = 13.4 mag per sq.arcsec (H=12.2) at a S/N~140. Stellar population modeling (without IMF sensitivity) combined with fiber stacking and longer integrations could make achieving S/N~20 feasible in annular bins, however.

Stellar Dynamics:

The high spectral resolution naturally draws one to studying dynamically cold systems, where, for example, R~10,000 is sufficient for measuring out-of-plane galaxy disk velocity dispersions (5-10 km/s). This is the basis of exciting work by the DiskMass Survey (Bershady+10) to “weigh” the mass of galactic disks. Again, surface brightness is a challenge. A fairly bright galaxy in their sample reaches 18.0 Hmag/sq.arcsec at R~10’’. Their CaII Triplet focused observations aimed to achieve S/N~10. But, absorption lines at APOGEE wavelengths are much weaker. However, the advantage would be the ability to potentially trace kinematically a different stellar population and, again, better penetrate the dust in the near-IR.


Here are some down-and-dirty (and possibly suspect!) scaling arguments on the expected S/N. I’d be very curious if there are further gains with #5, and the benefits of a J-band option (#6) that would open access to more emission lines. With reductions in S/N targets, longer exposures, and by smoothing in wavelength, we start to approach the surface brightness limits argued for above.

Scaling S/N arguments, background limited:

  1. Start with S/N = 140/pix for H > 12.2 (Vega, from 2MASS) in t = 3 hrs (on the 2.5m)
  2. Larger 3.5m mirror (just D3.5/D2.5): S/N=100 at H=12.6 (muH = 13.8)
  3. Reduce S/N to 20: ΔH=2.1 (H=14.7) Reduce S/N to 10: ΔH=2.9
  4. Smooth in wavelength by a factor of 10 (Resolution now ~2000). Goes as sqrt(smoothing): ΔH=1.25 (H=16, muH = 17.2)
  5. Gains from less moon illumination (dark time), no MW background: ??
  6. Move to J-Band?
  7. Large (much larger?) fiber apertures: ?

More about APOGEE applicability for extragalactic studies

Prepared by Dmitry Bizyaev

Absorption spectra:

We dedicated a few APOGEE fibers in order to investigate what we can get for extragalactic studies in the frames of an APOGEE ancillary program in 2011. We have observed centers of M31, M32, and M110 in single APOGEE fibers (2" diameter) with the 2.5m telescope. Below is a table of the signal-to-noise obtained with one "visit" (67 min exposure time, typically with a plenty of Moon light). In addition to the center-of-galaxy exposures, a few fibers were allocated at "SDSS plate collision distance" (typically 80 arcsec) from the center of M32 along the minor and major axes. Figures below show combined spectra of the centers of M31 and M32, unbinned in the spectral direction, some 10 visit combined. The most prominent lines are Fe, Mg, Si, Al, and some molecular bands (CO).

No image "m31_2.png" attached to Instruments No image "m32_2.png" attached to Instruments

Table: Signal-to-noise per pixel from extragalactic objects with APOGEE on 2.5m 

Object     SB(H),mag/"**2  SNR/pix, 1 visit (67 min)
M32 center   11.42         350
M31 center   11.76         178
M110 center  15.12           7

Off-center SB(H),mag/"**2  SNR/pix,7 combined visits
M32         ~18.2            4

The central surface brightness in the H-band is taken from the 2MASS Atlas of Large Galaxies. The cases of M110 and off-centered fibers in M32 show practical limitation of the spectrograph's applications for extended object studies. Thus, according to the same Atlas, the central part of a nearby face-on spiral galaxy NGC 628 that is brighter than 18.2 mag/sq.arcsec in the H spans 27 arcsec, which well matches the proposed size of 300-fiber IFU. Here we can study hidden bars, nuclear rings, nuclear spirals, and other kinematically distinctive structures.

APOGEE has unnecessary high spectral resolution (~25 km/s FWHM, 13 km/s per pixel), binning on the spectral direction will allow to increase SNR. Given enough signal in IFU fibers for kinematics and absorption spectra abundance analysis, the 1.5 arcsec fiber has size

Distance  1.5" size
D=10 Mpc  0.07 kpc
D=30 Mpc  0.22 kpc

Conclusion 1: APOGEE+3.5m NIR IFU should be able to study kinematics and abundances of stellar population in the central parts of nearby galaxies. The very centers of elliptical galaxies will be seen with superior SNR, while the spiral galaxies at ~ 10 Mpc will get moderate SNR if binning on the spectral direction is applied.

Emission lines:

APOGEE was designed for stellar studies, and its original working wavelength range is set to the region with minimum gas emission contamination. Unfortunately, the principal NIR gas emission lines avoid APOGEE spectral range (1.5-1.7 microns). Bright hydrogen emission lines can be seen in redshifted galaxies (see "z" range and corresponding 1.5" size in the table below):

Line     A   z_min  z_max 1.5" size,kpc

 beta  12820  0.18  0.31  5.7  9.7
 gamma 10940  0.38  0.54 12.0 16.7
 delta 10050  0.51  0.67 15.8 20.9

 break 14580  0.04  0.15  1.2  4.7

Some higher level Brackett lines fall into the APOGEE range, and can be seen in the absorption from the integral stellar population spectra.

One interesting case of [FeII] emission falling into APOGEE range is mentioned by Raimondo et al. (2013). They studied the very central part of a Seyfert 1 galaxy (MCG–6-30-15) and noticed decoupled kinematics of gas (e.g. from [FeII] 1.644, 1.677 mu) and stars.

Conclusion 2: APOGEE+3.5m NIR IFU can be used for AGN studies in relatively nearby galaxies. Benefits from observing z = 0.2-0.3 galaxies are doubteful because of poor spatial resolution (a few kpc per spaxel), which means using only a few spaxels out of 300.

Optical IFU for Nebula Work

Prepared by Don York

There are some applications of an IFU for projects on gas that require a feed to a high resolution spectrograph R>8000. The bundle(s) would be fed from the 2.5 meter to a 3.5m spectrograph. I list two science cases and do not consider technical challenges.

1) My main interest is in obtaining the highest resolution spectra possible of reflection nebulae. Two to three arcsec diameter fibers would be fine, as the main issue here is to get the best spectrum of the integrated nebula in the shortest time possible (by stacking the output of all fibers). The surface brightness of the nebulae is around 18 mag/sq. arcsec. Typical sizes are one arcminute.

The spectrum of a reflection nebulae is the same (almost) as the spectrum of the illuminating star. In many cases, the ISM on the line of sight is BEHIND the star. The back shining light hits dust and the mirrored light comes back through the same gas, to Earth. The dust in such nebulae is typically heated, by the photoelectic effect, to 50-100 degrees K and the ISM seen in absorption in the spectrum will experience excitation from a black body radiation higher than that of most of the ISM, which is affected mainly by the CMBR at 3 degrees K

In particular, more than one level of CN will be populated and produce absorption lines. There is a class of small molecules like CH and CH+ with upper levels not populated by the CMBR that would be excited by the thermal radiation from the reflection nebula black body. Additionally, and this is my main motivation, a candidate for the class of molecules that produce the Diffuse Interstellar Bands (DIBs) that have 5-7 atoms per molecule, would have very closely spaced rotational levels, and would produce a "band" under excitation by the CMBR. This would explain the minimum width of the unidentified DIBs that is seen for the full known set of over 600 DIBs (30-35 km/sec, as opposed to <1 km/sec for the three diatomic molecules mentioned above, which have very wide separations of the excited level from the ground state such that the excited lines have typical widths of interstellar absorption lines.

Dalhstrom et al (2013 ApJ, 773, 41) and Oka et al (2013 ApJ, 773 42) present the case for enhancement of DIB widths in one line of sight (Herschel 36, in NGC6530) which uniquely shows excited CH+, excited CH, as well as a doubling of the widths of several DIBs (representing the population of many more rotational levels in the ground vibrational state). The star happens to have an IR star only 400AU away in projection, so a diluted IR black body is available to pump the rotational levels and explain the extra widths for the DIBs.

Finding such an extraordinary coincidence is unlikely. Using the IR field of the reflection nebula is an alternative way to confirm the excitation that explains the anomalous DIBs, and to confirm the argument (based on molecular spectroscopy) that the DIBs are from small molecules (5-7 atoms), and idea that goes against much of the grain of the argument these days, but on weak spectroscopic and theoretical arguments.

A second case involves classic H II regions and extended star formation regions, also, nebular environments. Higher resolution is necessary to go beyond standard abundance analysis, to detect weak lines needed for evaluating densities and temperatures, and to measure fluctionations over small scales in nebulae (implied by differences in T from different diagnostics). For galaxies with separated, resolved H II regions, such as M101, higher resolution is necessary to evaluate tradiditional methods of determining physical parmeters of the gas, which differs from very high to very low densities in the same galaxy: an IFU is the instrument of choice and high resolution is needed to do critical work.

Meeting Notes

Meeting Notes January 14th 2015

APOGEE fiber Feed cost reduction Options

At the request of the fiber feed study group I have developed the following options for reducing project cost for running fibers from the APOGEE instrument to the 3.5 m focal plane. There are certainly many more combinations that could be considered. I’ve tried to present options that include the largest effect on total project cost. Many permutations on these 6 options could be considered but this should give us a bases for talking through the benefits and costs of a reduced scope project.

Reduced Cost Options for the APOGEE Fiber feed to the 3.5 m Telescope

Project DescriptionCost ReductionTotal Project cost
Original Proposal$0$351,820
Elimination of the ARC-SAT Run$17,964$333,856
Reduce IFU fiber count to 91$100,472$251,348
Reduce IFU fiber count to 37$128,628$223,192
Eliminate Lenslet Coupling$75,928$275,892
Bare Bones (30 fibers to Echelle focal Plane)$276,650$75,170

Elimination of the ARC-SAT Run

Almost $18k was allotted in the budget to run 15 fibers to the dome of the ARC-SAT telescope. Removing this cost from the project has no effect on the 3.5m run or the replacement of the 1m run.

Reduce IFU fiber count to 91

This reduces the number of elements in the IFU by a little over half. For a simple hexagonal packed IFU with a 1.4 arc-sec fiber size the long axis of the IFU would be reduced from 37 to 23 arc-seconds. This still assumes 47 sky fibers terminated in mini bundles around the central IFU.

Reduce IFU fiber count to 37

This reduces the number of elements in the IFU by almost a factor of 6. For a simple hexagonal packed IFU with a 1.4 arc-sec fiber size the long axis of the IFU would be reduced from 37 to 15 arc-seconds. This still assumes 47 sky fibers terminated in mini bundles around the central IFU.

Eliminate Lenslet Coupling

The primary effect of eliminating the lenslet coupling to the IFU is a reduction in the fill factor. Coupled IFUs can reach fill factors near unity (~90%). By switching to a ‘bare’ fiber system the fill factor of the IFU would be reduced to ~38%. This assumes the f-ratio of the 3.5m telescope is corrected to f/5 which will maintain the 1.4 arc-second spaxal size. MaNGA has shown that accurate dither patterns will fill in the gaps of bare fiber IFUs so the primary cost is a ~62% reduction in observing efficiency. There may also be some implications to the data regularity when the dither sets are combined.

Bare Bones Option

This is an option proposed by Bruce to use as much existing infrastructure as possible to join APOGEE to the 3.5 m. It would involve using existing conduits and would integrate into the Echelle focal plane utilizing the Echelle guider. I am assuming we would still correct the f-ratio to f/5, which is one of the remaining high cost items in this option. The two obvious advantages of this system are that it is far cheaper than any other option proposed, and it does not require forest service approval as no modifications to the site is required. The primary affects are as follows. 1) There is no field rotator on NA1. So the system would only be available for single object point source work no MOS or IFU mode would be available. 2) Throughput would be reduced due to the longer fiber run an extra ~12% (~10% to ~22%).

1/20/15 - I've updated the budget for the 'bare bones' option based on input from Mark and Bill - NKM

Note: See TracWiki for help on using the wiki.