The Cosmic Infrared Background Experiment

(CIBER): the Narrow Band Spectrometer

P. M. Korngut1,2,13, T. Renbarger3, T. Arai4,5, J. Battle2, J. Bock2,1,S. W. Brown6, A. Cooray7, V. Hristov2, B. Keating3, M. G. Kim8, A. Lanz2,D. H. Lee9, L. R. Levenson2, K. R. Lykke6, P. Mason2, T. Matsumoto4,8,12,

S. Matsuura4, U. W. Nam9, B. Shultz10, A. W. Smith6, I. Sullivan11,K. Tsumura4, T. Wada4, and M. Zemcov2,1

ABSTRACT

We have developed a near-infrared spectrometer designed to measure theabsolute intensity of the Solar 854.2 nm Ca II Fraunhofer line, scattered byinterplanetary dust, in the Zodiacal light spectrum. Based on the known equiva-lent line width in the Solar spectrum, this measurement can derive the Zodiacalbrightness, testing models of the Zodiacal light based on morphology that areused to determine the extragalactic background light in absolute photometrymeasurements. The spectrometer is based on a simple high-resolution tipped fil-ter placed in front of a compact camera with wide-field refractive optics to providethe large optical throughput and high sensitivity required for rocket-borne obser-vations. We discuss the instrument requirements for an accurate measurement ofthe absolute Zodiacal light brightness, the measured laboratory characterization,and the instrument performance in flight.

Subject headings: infrared: diffuse background — instrumentation: spectrograph —methods: laboratory — space vehicles: instruments — techniques: spectroscopic —zodiacal dust

1Jet Propulsion Laboratory (JPL), NationalAeronautics and Space Administration (NASA),Pasadena, CA 91109, USA

2Department of Physics, California Institute ofTechnology, Pasadena, CA 91125, USA

3Department of Physics, University of Califor-nia, San Diego, San Diego, CA 92093, USA

4Department of Infrared Astrophysics, Insti-

tute of Space and Astronautical Science (ISAS),Japan Aerospace Exploration Agency (JAXA),Sagamihara, Kanagawa 252-5210, Japan

5Department of Physics, Graduate School ofScience, The University of Tokyo, Tokyo 113-0033,Japan

6Sensor Science Division, National Institute ofStandards and Technology (NIST), Gaithersburg,

1

arX

iv:1

307.

0016

v1 [

astr

o-ph

.CO

] 2

8 Ju

n 20

13

1. Introduction

The intensity of the infrared Extragalac-tic Background Light (EBL) is a corner-stone of cosmological observations, a mea-sure of the total radiation produced bystellar nucleosynthesis and gravitationalaccretion over cosmic history. Accuratemeasurements can determine if the EBLis consistent with the calculated surfacebrightness from galaxy counts, or if an ad-ditional background component is present,e.g., unaccounted flux from known galax-ies or missing galaxy populations. Fur-thermore, light from the epoch of reion-ization must contribute to the total EBL,with Lyman cutoff and Lyman-α featuresoriginating at z > 6 redshifted into thenear-infrared (NIR) wavelength band.

Current direct measurements of the NIREBL intensity are mutually inconsistent.The dominant factor causing this discrep-ancy arises not from instrumental effects,but from different approaches to astro-physical foreground removal. In the NIR,all astronomical observations taken 1 AU

MD 20899, USA7Center for Cosmology, University of Califor-

nia, Irvine, Irvine, CA 92697, USA8Department of Physics and Astronomy, Seoul

National University, Seoul 151-742, Korea9Korea Astronomy and Space Science Institute

(KASI), Daejeon 305-348, Korea10Materion Barr Precision Optics & Thin Film

Coatings, Westford, MA, 01886, USA11Department of Physics, The University of

Washington, Seattle, WA 98195, USA12Institute of Astronomy and Astrophysics,

Academia Sinica, Room 1314, Astronomy-Mathematics Building, National Taiwan Univer-sityNo.1, Roosevelt Rd, Sec. 4 Taipei 10617, Tai-wan, R.O.C.

13Contact author, pkorngut@caltech.edu

from the Sun are completely dominated byZodiacal Light (ZL). ZL originates frominterplanetary dust scattering sunlight atoptical and near-infrared wavelengths, andemitting thermal radiation at mid-infraredto far-infrared wavelengths. In the early1990’s, the Diffuse Infrared BackgroundExperiment (DIRBE) produced absolutemeasurements of the astrophysical skybrightness in multiple bands over a widerange of Solar elongation angles (Hauseret al. 1998). A model of ZL based onmorphology derived from annual modu-lation of the signal measured by DIRBEwas developed for subtracting the ZL fore-ground in order to assess the EBL (Kel-sall et al. 1998). Applying this modelhas led to high EBL estimates (Cambresyet al. 2001; Dwek & Arendt 1998; Mat-sumoto et al. 2005). Zodiacal dust mod-els based on morphology alone are notunique, and a separate ZL model was sub-sequently developed based on the assump-tion that the mid-infrared sky brightnessobserved by DIRBE is entirely due to ZL(Wright 2001). Implementing this fore-ground treatment led to a significant de-crease in estimated EBL levels (Levensonet al. 2007). Dwek et al. (2005) make theimportant observation that a 23% underes-timate in ZL intensity in the Kelsall et al.(1998) model would account for the entireobserved NIR EBL excess, as the residualEBL spectrum displays the same color asZL within the statistical uncertainties. Anindependent test of the ZL contribution tototal sky brightness is needed.

Solar Fraunhofer lines scattered by in-terplanetary dust provide an alternativemethod of estimating ZL brightness. Asthe equivalent widths of Fraunhofer lines

2

are accurately known in the Solar spec-trum, a measurement of the line in scat-tered light is directly related to the ZLcontinuum brightness. Empirically, themeasured equivalent width of Fraunhoferlines in ZL matches the Solar spectrum to< 2% (Beggs et al. 1964). Early measure-ments by Dube et al. (1977) were the firstto estimate ZL from its Fraunhofer signa-ture. Bernstein et al. (2002a) attemptedto measure the optical EBL using a com-bination of absolute sky brightness mea-surements with the Hubble Space Tele-scope and Fraunhofer line measurementswith a ground-based telescope. This pi-oneering measurement was complicated byatmospheric airglow emission, but also bysystematic uncertainties from atmosphericscattering, ground scattering, and straylight (Mattila 2003). Mattila et al. (2011)have developed a technique to remove ZLby measuring the difference in brightnessbetween a high-latitude dark nebula andits surrounding area, using Fraunhofer linemeasurements to remove diffuse Galacticlight from starlight scattered by interstel-lar dust. They have applied this techniqueto estimate the EBL at 400 nm and pro-vide an upper limit at 520 nm. A recent re-analysis of Pioneer 10 and 11 optical dataat 440 nm and 640 nm from outside the Zo-diacal cloud (Matsuoka et al. 2011) also de-rive an EBL that is consistent with Mattilaet al. (2011). These measurements, donewith a model-independent ZL foregroundestimation, suggest a total EBL consistentwith the integrated light from galaxies inthe optical.

In this paper, we describe the de-sign and implementation of a NarrowBand Spectrometer (NBS) aboard the

Cosmic Infrared Background Experiment(CIBER) sounding rocket payload (seeZemcov et al. 2011). The instrument isoptimized to measure ZL intensity by ob-serving a single Fraunhofer line at 854.2nm caused by Solar Ca II absorption, andwork in synergy with a low-resolution spec-trometer (LRS) (see Tsumura et al. 2011)to estimate the mean intensity of the EBLin the NIR. The LRS will measure thespectrum of the ZL (e.g. Tsumura et al.2010) from 0.75 to 2.1µm so that a deter-mination of the ZL brightness at 854.2 nmcan be directly compared with predictionsfrom ZL models based on the DIRBE 1.25and 2.2µm bands, where the EBL discrep-ancies are the largest. The CIBER payloadalso contains two wide field imaging cam-era (Bock et al. 2012), designed to probethe EBL through its spatial fluctuationcharacteristics.

2. Instrument Description

We have chosen to measure the 854.2nm line, the strongest of the Ca II tripletlines at 849.8, 854.2 and 866.2 nm. We se-lected a long-wavelength Fraunhofer linefor comparison with the DIRBE/Kelsallmodel at 1.25 and 2.2µm, and the 854.2 nmline is the brightest non-hydrogenic Fraun-hofer line, with an equivalent width (EW)of 0.37 nm (Allen 1976), at wavelengthslonger than 400 nm. We avoided hydro-gen lines, which have a contribution fromrecombination line emission from the in-terstellar warm ionized medium (Mar-tin 1988). The narrow-band spectrome-ter (NBS) uses a tipped filter spectrome-ter with high-throughput refractive optics(Eather & Reasoner 1969). Our goal is tomake a percent-level determination of the

3

ZL intensity, in comparison to the current23 % error needed to explain the discrep-ancy between the excess EBL and galaxycounts (Dwek et al. 2005).

2.1. ZL Detection

The NBS measures the absolute skybrightness on and off a known Fraunhoferline, producing a photocurrent

i(λ) =ηAΩ

hc

∫λ′Iλ′T (λ, λ′)dλ′ [e−/s],

(1)where h and c are Planck’s constant andthe speed of light, respectively, η is thepeak efficiency, AΩ is the detector through-put, Iλ′ is the sky surface brightness, andT (λ, λ′) is the normalized instrument re-sponse function centered at wavelength λ.High instrument sensitivity is needed in or-der to make precise measurements duringthe short observation time available in asounding rocket flight. In the simplifiedcase where the sky consists of two bright-ness components, the ZL and the EBL,and the Fraunhofer line is unresolved, thephoto currents measured on and off the lineare

ion 'ηAΩ

hc(λIλ,ZL∆λ−λIλ,ZLW+λIλ,EBL∆λ),

(2)

ioff 'ηAΩ

hc(λIλ,ZL∆λ+ λIλ,EBL∆λ), and

(3)

ioff − ion =ηAΩ

hcλIλ,ZLW, (4)

where λIλ,ZL and λIλ,EBL are the ZL andEBL continuum brightness off the line, Wis the equivalent width of the Fraunhoferline, and the spectral resolution is

∆λ =

∫T (λ, λ′)dλ′. (5)

Thus the estimated ZL local continuumbrightness is simply

λIλ,ZL =hc

WηAΩ(ioff − ion), (6)

determined with an uncertainty given by

σλIλ,ZL =hc

WηAΩ(σi2off + σi2on)1/2, (7)

where σioff and σion are the instrumentaluncertainties in the on and off line detectorcurrents. If the noise on detector currentscales as the number of pixels N on and offthe line, the ZL sensitivity is given by

σλIλ,ZL =hc

WηAΩσi(2/N)1/2, (8)

where σi is the per pixel current noise.Thus to optimize surface brightness sen-sitivity, we need to maximize the pixelthroughput AΩ, and incorporate as manypixels on and off the line as possible.

In addition to the measurement errors,W technically contains uncertainty itself,dominated by the stability of the Fraun-hofer line profile over time. To quan-tify this contribution, we obtained pub-licly available high resolution spectra fromthe SOLIS1 instrument which sampled theCa II line daily over the course of sev-eral years (2008 to 2012). The equiva-lent width of the line measured between853.55 nm and 854.80 nm was observed tovary at the 0.5% level over this time pe-riod, and is therefore taken to be a highlysub-dominant source of error in ZL mea-surement using this technique.

In addition to raw sensitivity, we requiresufficient spatial resolution to remove the

1http://solis.nso.edu/iss/

4

foreground contribution from stars. As theCa II Fraunhofer line is expected to have acomparable depth in the interstellar radia-tion field as in the Solar spectrum (Mattila1980b), residual starlight will add to theinferred ZL intensity in Eq. 5. We havechosen a pixel resolution of 2×2 arcmin inorder to mask stars down to S < 14 Vegamagnitude using positions and fluxes froman external catalog.

2.2. Optical Design

The NBS instrument uses a narrow-band interference filter placed at the aper-ture of an imaging camera with high-throughput refractive optics. This ‘tilted-filter photometer’ provides an intrinsicsensitivity advantage by having higherthroughput than spectrometers using aprism or grating (Jacquinot 1954, Eather& Reasoner 1969). The design of thenarrow-band spectrometer optics is shownin Figure 2.2, and detailed specificationsare summarized in Table 1.

The narrow-band interference filter wasfabricated by Materion Barr Precision Op-tics & Thin Film Coatings2 to customizedspecifications from the science team. Thefilter is tipped by 2 from normal in or-der to optimize sensitivity, based on thenumber of detectors on and off the line.The narrow-band filter is combined withtwo filters that provide out-of-band spec-tral blocking.

The wide-field camera is a refractivedesign with six lenses developed and man-ufactured by the Genesia Corporation,Japan3 based on the cryogenic refrac-

2\protecthttp://materion.com/3\protecthttp://www.genesia.co.jp/

tion measurements discussed in Yamamuroet al. (2006). The lenses are manufacturedfrom fused silica and are spherical exceptfor the back surface of the entrance lensand the front surface of the last lens, whichare 6th order aspheres. The lenses are anti-reflection coated for < 0.2% reflectance persurface at 854.2 nm.

A 256 × 256 PICNIC HgCdTe detectorarray manufactured by Teledyne Scientific& Imaging4 is placed at the f/0.92 focusof the camera, yielding a 2′× 2′ pixel scaleand an 8.5× 8.5 field of view on the sky.A mechanical shutter and calibration lampare also installed in the NBS instrument(see Zemcov et al. (2011) for detailed spec-ifications of these subsystems). The cryo-genic shutter is placed between the opticsand the detector array in order to moni-tor array dark current. A calibration lampassembly feeds light from a light emittingdiode into the optical chain with an opticalfiber between the third and fourth lenses.The lamp is used as a transfer standardto track any changes in response in flightcompared with the laboratory calibration.

2.3. Filter Design

The spectrometer disperses radiation byshifting the effective filter bandpass by thechange in incident angle over the field ofview (FOV) according to

λ = λ0 cos(θi), (9)

where λ0 is the response at normal inci-dence and the internal angle θi is relatedto the incident external angle θe by Snell’slaw sin(θi) = sin(θe)/n. Figure 2 showsthe calculated peak wavelength response of

4\protecthttp://www.teledyne-si.com/

5

Fig. 1.— Ray trace and photo of the NBS optics. The spectrometer uses a fixed interferencefilter combined with 2 blocking filters placed at the entrance aperture of a wide-field camera(see Table 1 for specifications). The entire optical assembly and focal plane assembly iscooled to 77 K in operation. The ray trace shows an extended “pop up” baffle tube usedafter the first flight that is not presented in the photograph.

6

Table 1: NBS Component Parameters.Optics wide field refractive camera with 6 lensesFilters 2 blocking filters and 1 tipped narrow-band filter.Detector 256× 256 MBE HgCdTe PICNICAperture 75 mmf/# 0.92Pixel size 2× 2 arcminField of view 8.5 × 8.5

Filter central wavelength at 295 K 855.57 nmFilter central wavelength at 77 K 854.55 nmFree spectral range at 77 K 852.0− 854.5 nmFilter resolution λ/∆λa 1120Filter tip angle 2.0

Lens efficiency 0.98ED 541b narrow-band filter efficiency 0.84ED 536 blocking filter efficiency 0.66ED 686 blocking filter efficiency 0.83Detector QEc 0.83Total QE 0.37Correlated double sample noise 28 e−

Calculated off-line responsivityd 529 nW m−2 sr−1 / e− s−1

a ∆λ is the integral width ∆λ =∫T (λ′)dλ′.

b ED NNN is a manufacturer’s part number.c Measured at 2.2µm.d Responsivity given by η∆λAΩ/hc in Eq. 3.

7

pixels across the array using the parame-ters listed in Table 1, and exhibits the char-acteristic shift given by Equation 9. The8.5× 8.5 field of view combined with the2 filter tip modulates the wavelength re-sponse from λ − λ0 = 0 at normal filterincidence to λ − λ0 = 0.003λ0 at the ex-treme corners. The divergence-limited res-olution (e.g. Bock et al. 1994) varies fromλ/∆λ = 3.6 × 107 at the array center to2.8 × 104, at the minimum wavelength atthe array corner. The spectral range ofλ/∆λ = 1/0.003 = 330 is sufficient to sam-ple the ZL spectrum both on and off theCa II 854.2 nm Fraunhofer line, which it-self has a width of λ/∆λFWHM ≈ 4000.

The instrument is insensitive to thechoice of spectral resolution to first or-der (see Equation 7). However, the reso-lution does vary the number of pixels onand off the line. Higher resolution makesNon < Noff , slowly reducing sensitivity butalso providing for a larger spectral contrastto ZL according to

ioff−ion

ioff

=W

∆λ. (10)

The actual response to ZL over theFOV, given the finite resolution and thefinite width of the Fraunhofer line, is theconvolution of the instrument responsewith the Solar spectrum (see Equation1). The spectral contrast on and off theCa II line is 26 %, giving the characteristicpattern shown in the noiseless instrumentsimulation in Figure 3. Using a simulatorwe investigated a range of resolution andfilter tips, balancing sensitivity and con-trast, and determined that λ/∆λ ≈ 1000was a good choice. Since the wavelengthresponse of the pixels varies with radiusfrom the maximum wavelength, data from

Fig. 2.— Instrument wavelength map cal-culated from Equation 9 using the parame-ters listed in Table 1. The colorscale showsthe peak wavelength response of each pixelin the 256 × 256 detector array, with thecontours at intervals of ∆λ = 0.2 nm.

the NBS measures the convolved imageshown in Figure 3, from which the abso-lute ZL brightness can be determined. Thespectral response convolved with the Solarspectrum is shown in Figure 4.

2.4. Detectors and Electronics

The NBS uses a 256× 256 HgCdTe de-tector array employing a PICNIC readout integrated circuit (ROIC). A molec-ular beam epitaxy (MBE) process is usedto provide short wavelength response com-pared with HgCdTe arrays fabricated onsapphire substrates (M. Farris, privatecommunication). The detectors have a40µm pixel pitch, giving a large area opti-mal surface brightness sensitivity. A cus-tom electronics chain is used to addresspixels on the multiplexer and read out anddigitize the detector signals, described indetail in Zemcov et al. (2011). The de-tector bias and reset voltages are trimmedas in Lee et al. (2010). We do not use

8

Fig. 3.— The convolved Solar spectrumwith the calculated spectral response overthe field of view of the NBS using the in-strument parameters given in Table 1. Weplot the ZL intensity normalized to unityat the ZL continuum level. The contoursshow annuli of constant brightness and arespaced at intervals of 0.03. The drop insignal around the maximum wavelength re-gion is due to the 854.2 nm Ca II line. An-nular averages centered on the filter re-sponse pattern give a known profile fromthe Solar spectrum that is directly propor-tional to the ZL intensity.

Fig. 4.— Comparison of the Solar spec-trum (Delbouille et al. 1990) near the Ca II

854.22 nm line, the spectral response of theNBS filter, and the convolution of the filterwith the Solar spectrum. The NBS mea-sures the convolved spectrum from 852.0to 854.5 nm. The NBS filter has been nor-malized to the peak transmittance listed inTable 1.

the built-in buffer amplifier circuitry onthe ROIC, which emits photons and in-creases the detector dark current, in favorof external junction field-effect transistor(JFET) source followers (Hodapp et al.1996). Since the same detectors are usedin the both the NBS and LRS, we referthe reader to Tsumura et al. (2011) fora detailed discussion of the electrical andresponsivity properties of the CIBER PIC-NIC arrays.

3. Instrument Assembly and Labo-ratory Testing

We tested the instrument performancein the laboratory prior to flight. The pay-load section is designed to accommodate

9

several test configurations (Zemcov et al.2011):

1. Using a vacuum lid with fused quartzwindows the NBS can view exter-nal sources to measure focus, wave-length response, and absolute cali-bration using the SIRCUS (SpectralIrradiance and radiance ResponsivityCalibrations using Uniform Sources)facility (Brown et al. 2006).

2. Using a vacuum box the NBS viewsan integrating sphere without inter-vening windows to measure the flatfield.

3. Using the flight shutter door equippedwith a radiatively cooled inner shieldwe can measure the noise propertiesand dark current under dark condi-tions.

3.1. Focus testing

We measure the optical focus with theCIBER cryostat outfitted with two sets ofwindows: warm vacuum windows installedat the bulkhead, and cold windows oper-ating at ≈ 150 K to intercept thermal ra-diation from the warm window and bulk-head. The NBS vacuum window is anti-reflection (AR) coated to eliminate reflec-tions between the window and the NBSfilters, though the cold window was not.The transmittance of the warm and coldwindows at 850 nm are 0.94 and 0.92, re-spectively.

We introduced a collimated beam usinga 254 mm diameter f/3.6 Newtonian tele-scope with white light shining through a50µm diameter pinhole located at the tele-scope focus. The best focus of the Newto-

nian telescope was first determined usingan auto-collimation technique with a flatmirror. We then illuminated the NBS en-trance aperture with the collimated tele-scope beam, and the pinhole was steppedin 2.5 mm increments over a 50 mm rangeabout the telescope focus. The best fo-cus of the NBS was measured and the dis-placement from the array position, ∆x2,was calculated using

f 21

∆x1

=f 2

2

∆x2

, (11)

where f1 and f2 are the focal lengths and∆x1 and ∆x2 are the displacement frombest focus of the collimating telescope andthe NBS, respectively. The range of mea-surement allows a 300µm through focusto be probed in increments of 15µm fromNBS focus. We measured focus near thecenter of the array, at each of the four cor-ners of the array, and then repeated thefocus test at the array center for redun-dancy.

Once a focus test is completed, the po-sition of the focal plane was determinedby plotting the full-width half-max spotsize as a function of pinhole position. Aquadratic polynomial was fit to the spotsize as a function of focus position, andthe minimum of this fit was taken to bethe NBS array position relative to ideal fo-cus. This program allowed us to focus theNBS to within 10−20µm, well within theflp ≈ 40µm depth of focus of the NBS,where f = 0.92 is the f/# of the NBS andlp = 40 µm is the pixel size.

3.2. Wavelength Calibration

We developed a system of filters man-ufactured by Barr Associates consisting of

10

a narrow-band filter and two blocking fil-ters. Due to the precise tuning require-ments, Barr measured the change in re-sponse (∆λ = −1.0 nm) on cooling to77 K on a witness filter, and calibratedthe shift in response versus incident angleat room temperature. The change in re-sponse is well fit to a material index of 1.75,as shown in Figure 5. The narrow-bandfilter gives a peak response of 855.57 nmwith an integral width of 0.76 nm at roomtemperature, which changes to 854.55 nmwhen cooled to 77 K. Barr measured thetransmittance in 7 positions from the cen-ter to the edge at room temperature, andfound an average peak transmittance of84% and that the peak wavelength variesby 0.34 nm, maximum to minimum, at dif-ferent positions across the filter.

Fig. 5.— Measured peak wavelength withangle of incidence at room temperature, fitto an index of 1.75 and λ0 = 852.57 nm.The theory curve follows Equation 9. Thewavelength response at normal incidenceis a function of temperature so that λ0 at77 K is 854.55 nm.

We calibrated the NBS as a functionof wavelength using the SIRCUS lasersand an absolutely calibrated reference ra-diometer. Because the laser SIRCUS Ti-Sapphire line-width is essentially zero, thenarrow band pass of the NBS is fully re-solved. Etalons placed in the laser cavityprovided wavelength stability and fine tun-ing. We measured wavelength response toan uncertainty of approximately 0.001 nmreferenced to a calibrated wavemeter in-corporating a stabilized HeNe laser refer-ence. The absolute radiance of the inte-grating sphere, viewed by the NBS, was in-ferred using a monitor detector separatelycalibrated to an absolutely calibrated ra-diometer. This two-step bootstrapping ap-proach is necessary due to the dynamicrange mismatch between the radiometerand the NBS. In order to achieve sufficientsignal to noise on the monitor detector, themonitor was located on a smaller fiber-fedsphere injecting light into the larger sphereviewed by the NBS. The relatively spec-trally flat ratio of the large sphere radi-ance to the monitor signal was measuredat a few wavelengths and interpolated tothe NBS wavelengths. The sphere radiancewas stabilized using the monitor signal anda liquid crystal laser intensity modulator.The monitor signal was recorded through-out the NBS calibration measurements.

For each pixel in the array, a Gaus-sian spectral response function was fit tothe laser data, yielding measurements ofthe central wavelength and integral widthover the array. These data and the sub-sequent fits are shown in Figure 6 for asubset of individual pixels across the ar-ray. The central wavelength response overthe NBS field is shown in Figure 7. The

11

average measured resolution over all pix-els is λ/∆λ = 1004 ± 12 , which averagesover any spatial variation in wavelength re-sponse over the narrow-band filter. Themeasured system wavelength response isclose to the values listed in Table 1 basedon the measured narrow-band filter, with amaximum wavelength of 854.51± 0.02 nm.

Fig. 6.— The wavelength response forindividual pixels sampled at four discreteangles from the maximum wavelength re-sponse of the NBS. The points show themeasurements for four pixels, starting atthe peak wavelength of 854.51 nm andmoving shortward. The measurements usethe SIRCUS facility, taken in wavelengthsteps of ∆λ = 0.1 nm. The curves show thefit of a Gaussian, yielding a central wave-length for each pixel.

In addition to the SIRCUS facility, wecan verify the reproducibility of the NBSwavelength response using an integrat-ing sphere coupled to either a R ∼ 600monochromator, a stabilized 852 nm diodelaser and wavemeter, or a Ne lamp withan emission line at 854.5 nm, close to the

Fig. 7.— Measured spectral response overthe instrument field of view. The spec-tral response was measured using a high-resolution tunable laser injected into an in-tegrating sphere with an exit port viewedby the NBS. The minimum wavelength isshifted from the center of the field of viewby tipping the filter by ≈ 2 in order tooptimize the instrument sensitivity. Con-tours and color scale are identical to thosein Figure 2.

NBS bandpass. These were used for quicktests to verify the wavelength calibrationdid not change over multiple thermal cy-cles and after payload vibration tests.

3.3. Anomalous Thermal Response

With the original filter set, consisting ofthe narrow-band filter and a single block-ing filter (referred to as ED536), installedin the NBS instrument, we observed asignificant background signal of 310 e−/swhen looking out into a dark laboratorythrough the fused quartz vacuum win-dow. By modulating the temperature ofthe room-temperature window and holder,we determined this radiation was thermalemission incident on the detectors. Thesignal greatly exceeded the expected levelbased on the measured on-axis filter block-

12

ing, though the source of this emission isoff-axis and well outside the field of view.To reduce this radiation we added a sec-ond blocking filter, ED686 as shown in Fig-ure 8, to improve the long-wavelength at-tenuation. However under the same condi-tions we still observed a large backgroundlevel of 110 e−/s with the new filter inplace, certainly an improvement but notthe factor of > 104 reduction expectedfrom the additional long-wavelength block-ing.

The CIBER instrument was flown in itsfirst flight with the NBS using this filterconfiguration. During the flight the LRSobserved a large long-wavelength signal as-sociated thermal emission from the skinof the sounding rocket, heated by air fric-tion during ascent (Tsumura et al. 2011).The skin section had a direct view of theinside surface of the cold baffle tube andfirst optical element of all the imagers andspectrometers, and thus thermal radiationcould scatter to the detectors. Unfortu-nately the NBS saw an elevated photo-current of ∼ 150 e/s− as well, related tothis thermal emission. After discussionswith Barr, we determined that the filterblocking degrades at large incident angles,as shown by the model curves in Figure 8.In order to improve the large-angle perfor-mance, Barr designed and fabricated anabsorbing dielectric coating to the long-wavelength ED686 blocking filter, namedED686-R in Figures 8 and 9.

According to calculations, by adding theabsorbing coating, the large-angle block-ing should improve by 102 to 104, over therange of 0 to 80 incidence, with little lossof in-band transmission, from 89.0 % to87.5 %, as measured at 77 K. As shown in

the bottom panel of Figure 8, the measuredlarge angle performance at 60 to 80 an-gle of incidence in the 2000−2500 nm rangeis ∼ 102 better than the calculated per-formance without the coating, but some-what worse than the calculated perfor-mance with the coating. The re-coated fil-ter was installed in the instrument, and wemeasured a background level that was sig-nificantly reduced looking through the vac-uum window, 1.4 e−/s, consistent with themeasured improvement of ∼ 80 at 60 to80.

In addition to improving the filter de-sign, we added a cold extending/retractingbaffle to extend the baffle tube beyond theheated rocket skin, eliminating all lines ofsight from the skin to the optics and the in-side surface of the baffle, described in de-tail in Zemcov et al. (2011). In the sec-ond flight the measured thermal responsein the LRS was at least 100 times smallerthan in the first flight, and we did not ob-serve a measurable response in the NBS,indicating that the baffles and large-angleblocking filters performed as expected.

4. Instrumental Systematic ErrorControl

4.1. Requirements

The instrument design controls system-atic instrumental errors by incorporating1) a cold shutter to measure array darkcurrent; 2) low scatter optics to eliminatestray light from off-axis sources; 3) a filter-ing scheme with excellent spectral blockingto control out-of-band radiation; 4) a spe-cialized aperture-filling laboratory sourceto measure array flat field response; and 5)a calibration program based on laboratory

13

Fig. 8.— Optical Density as a function ofincident angle for the two blocking filters.(Optical Density = − log10(T )). The long-wavelength blocker ED686 was recoatedwith an absorbing dielectric to improveits large-angle blocking. The panels showfrom top to bottom 1) calculated large-angle transmittance of the ED536 filter;calculated performance of the ED686 filter2) before and 3) after applying the absorb-ing dielectric layer; and 4) the measuredlarge-angle transmittance of the ED686 fil-ter with the absorbing coating. The mea-sured transmittance in the 2000 - 2500range for 60 to 80 incident angles ap-pears to not show the full calculated im-provement and is consistent with the ob-served reduction by factor of ∼ 80 to ther-mal emission measured in the laboratory.

measurements.

The laboratory characterization pro-gram addresses the systematic error re-quirements. In addition to the tests de-scribed in Section 3, these measurementsinclude sensitivity, filter blocking, darkcurrent, temperature stability, flat fieldand calibration determination.

4.2. Dark Current

The generation of charge carriers withinthe depletion region of HgCdTe infrareddetectors leads to a small (∼0.1 e−/s) darkcurrent even in the absence of incidentphotons. In order to perform absolutemeasurements of the ZL, this dark cur-rent must be understood and removed fromflight data. In flight, the dark current ismonitored by closing the instrument’s coldshutter prior to launch, during data collec-tion in the middle of the flight, and againfollowing the vacuum shutter door closeevent during reentry.

To determine dark current behavior be-fore flight, we measure the baseline darkcurrent with the instrument in flight con-figuration, with the shielded vacuum door,providing a dark environment in the exper-iment section (Zemcov et al. 2011). TheNBS cold shutter is closed during the mea-surement to ensure the darkest conditionspossible at the detector array. To deter-mine the dark current, the array is readout for ∼ 1 hour in this configuration. Be-cause resetting the array causes a transientresponse, the first several frames followinga reset are discarded from the analysis. Wefit a slope and offset over each integration,obtaining an array median dark current of0.5± 0.1 e−/s, where the uncertainty is es-timated from the variance of twelve 50 s in-

14

tegrations. The pattern of the dark currentis brightest at the array corners, suggest-ing residual emission from the ROIC eventhough we are not using the final bufferstage. In the case of the NBS, the dom-inant dark current characteristic for sys-tematic control is not the mean level, butthe stability of spatial morphology acrossthe array. Gradients or features spuriouslyaligned with the wavelength response canproduce artificial spectral structures whichcould mimic the Fraunhofer line. We quan-tified this effect through simulations as de-scribed in Section 4.8.

To verify that the dark current is sta-ble over long periods, we periodically ob-tained a standard calibration set includ-ing a dark current measurement over thecourse of several years. Though typicallytaken with the light bulkhead in place ofthe rocket door, these measurements showthat the dark current is stable over longperiods and many cryogenic cycles. Theselong-term dark levels are consistent withthe images obtained on the launch gantryjust prior to flight, and the dark measure-ments obtained during observations withthe shutter closed.

We compared dark current measure-ments taken with a cold lid placed at77 K over the optics and the cold shut-ter open and closed, showing a DC offsetof 0.33 e/s. This indicates an additionalphotocurrent is present, induced by emis-sion from the ROIC reflecting off the NBSoptics. While the amplitude of this sig-nal is large compared to the mean darkcurrent, it has a distinctive stable spatialmorphology along the array. This consistsof a spatially flat component and a com-pact region of excess photocurrent. The

reflected light can be subtracted, or thepixels in the compact region can be sim-ply discarded in the analysis of flight data.A ZL spectrum was fit to this spatial ar-ray response to quantify the systematiccontribution of this effect, if no furthercorrections are applied. The best fit spec-trum had an amplitude of -1.88 nW m−2

sr−1, corresponding to a ∼ −0.4% effectfor nominal ZL intensities, assuming thecalibration factor determined in Section4.7.

4.3. Temperature stability

Because the output voltage of the ROICvaries with temperature, typically∼ 1000 e/K,we have implemented a multi-stage tem-perature control system. As described inZemcov et al. (2011), the pressure of thecryogenic bath is regulated in flight by anabsolute pressure valve. In addition, thefocal plane temperature is controlled in2 thermal stages. An intermediate stageis actively regulated. The focal plane isthermally staged from the controlled stageto passively filter high-frequency tempera-ture variations. The temperature of bothstages are precisely monitored using ahigh-precision temperature bridge.

To verify that the regulation systemworks as designed, we cooled the NBSfocal plane to its quiescent temperature(typically ∼ 80 K) at which point we acti-vated the temperature control unit. Ther-mometry data are collected during thetime the focal plane unit takes to reacha steady operating temperature, which isset in the firmware of the thermal controlunit. Though the thermal set point is ad-justable, it must be set to a value largerthan the highest base temperature so that

15

it can be controlled in all instrument flightand testing configurations. At our chosenset point of 82.23 K, it typically takes 1.5 hfor the heaters to raise the temperature ofthe focal plane to its set point ∼ 2 K abovethe base. When stabilized, the active con-trol proportional-integral-derivative (PID)loop controls the intermediate stage aboutthe set point.

To quantify the systematic error on ZLintensity measurement induced by thermalinstabilities, dark current was measuredperiodically during the stabilization pro-cess. Images were produced from darkdata with thermal drifts ranging up to1.2 mK/s. This drift produces a responseover the array of typically ∼ 1 e/s. Be-cause this response is largely spatially uni-form, the effect of a thermal drift on ZL sig-nals is minimal. After subtracting a tem-plate of dark current produced by aver-aging many integrations taken previouslywith the temperature fully controlled, aFraunhofer line was fit to the thermally un-stable data. The inferred ZL brightnesswas observed to depend very steeply onthermal instability, with the amplitude ofthe induced ZL signal falling below the sta-tistical uncertainty (described in Section4.8) at a drift level of 0.45 mK/s, equiv-alent to ∆ZL = 21.1 nW m−2 sr−1in Ta-ble 2. When fully controlling, the aver-age drift of the focal plane temperature is< 10µK/s over a 50 s integration, makingthe effect of thermally varying dark currenta negligible systematic.

The NBS control system is powered onseveral hours before flight to ensure thatthe focal plane temperature has stabilizedat launch. During flight, the temperatureof both thermal bridges are telemetered to

the ground. The flight data show no signif-icant deviation from the ground behaviorduring astronomical observations in flight.Figure 10 of Zemcov et al. (2011) showsthat the thermal drift of the NBS focalplane, at its worst during the first hundredseconds of flight, is 5.3µK/s in the first 50 sof astronomical data acquisition, which ex-ceeds the specification. However, after thisperiod of time the drift is < 0.02µK/s andthermally-induced drift is negligible.

4.4. Out of Band Blocking

For precise measurements of the Fraun-hofer line brightness, we require a high de-gree of out-of-band blocking of the ZL con-tinuum. The HgCdTe detectors are onlyresponsive over a 2100 nm range span-ning 400−2500 nm, which defines the outerwavelength limit for potential contamina-tion. The fact that the ZL continuumpeaks at optical wavelengths, and that anyspectral leak is unlikely to be modulatedover the field of view as the Ca II line,further reduces the systematic error fromleaks. Barr originally designed a singleblocking filter (ED536) to meet the OD =− log10(TOB) > 6 blocking requirement.This filter (ED536) provides an in-bandtransmittance of 66 %. The blocking per-formance in Fig. 9 is excellent, though lim-ited by measurement accuracy to OD= 5.

The on-axis blocking performance of theNBS was characterized using the tunablelong wavelength laser system coupled tothe integrating sphere as described in Zem-cov et al. (2011). The laser wavelengthwas changed to discrete values separatedby ∼ 100 nm over the range 1100 ≤ λ ≤2300 nm. At each wavelength step, wedetermined the response of the NBS by

16

Fig. 9.— Out-of-band blocking of compo-nent filters measured at normal incidence.Note the straight lines indicate instrumen-tal limits on the out-of-band blocking, notactual transmittance. Of the two blockingfilters, ED686 was measured before and af-ter an absorbing coating was applied to re-duce large-angle transmission.

summing the photocurrent over the ar-ray. To remove the photocurrent arisingfrom background light in the apparatus,we recorded ambient frames when the laserwas shuttered. We differenced the laser onand shuttered data to measure the pho-tocurrent arising from out-of-band light.

We simultaneously monitored the ra-diance incident on the NBS using a cal-ibrated detector viewing the integratingsphere. Because the laser is capable ofproducing a large radiance in a narrowband, the signal to noise ratios of the mea-surements are large, even though the ab-solute response to out of band light isnot. The average photocurrent arisingfrom 1100 ≤ λ ≤ 2300 light is 2 × 10−5

e−/s / nW m−2 sr−1, which is consistentwith the OD > 5 limit derived from di-rect filter testing. We produce an out-of-band spectral response function by inter-polating the measurements over the range850 ≤ λ ≤ 2500 nm.

To determine if the measured out ofband blocking is sufficient, we estimatethe photocurrent expected in flight fromZL by assuming a Solar ZL spectrum for850 ≤ λ ≤ 2500. The expected signalat the NBS detector is then calculated bynumerically integrating the product of theZL spectrum and our out-of-band responsefunction which yields iOB ≤ 3 me−/s at theecliptic pole. This results in a calculatedlevel of iOB/iIB ≤ 0.2% and is therefore isdeemed negligible.

4.5. Flat Field Determination

To measure the Fraunhofer line inten-sity we must first normalize the relative re-sponsivity of the individual detector arraypixels to a common value, typically termed

17

the “flat field” correction. Because theflat field of the NBS detector array cannotbe determined independently of the astro-physical signal in flight, we must measurethe flat field response in the laboratory.The internal instrument calibration lampserves as a transfer standard to verify thatthe responsivity has not changed betweenthe laboratory testing and flight.

The flat field measurement consists ofa quartz halogen lamp which is fiber cou-pled into an integrating sphere inside avacuum chamber. We measured the out-put spectrum of the integrating sphere andadded a filter at the lamp to synthesize aSolar spectrum over the wavelength rangeof interest, 800 nm≤ λ ≤ 2500 nm. TheCIBER instrument shares a common vac-uum space with the test chamber so thereare no optical surfaces between the inte-grating sphere port and the NBS aperture.Since the illumination pattern of the out-put of the sphere has been measured tobe uniform to < 0.3% at angles of inci-dence < 10 the illumination measured atthe NBS focal plane is a good tracer of thearray flat field.

The flat field response is measuredby turning on the quartz halogen lamp,recording an image, and then subtractingimages before and after the lamp is offto remove any background signal. Finallythese measurements are bracketed by darkframes to ensure that the array baselinebehavior is stable. The flat field is mea-sured at a variety of input flux levels from1000 to 10 e−/s to ensure the flat fieldmeasurement does not vary significantlywith source brightness due to detector non-linearity.

To quantify the stability of the flat field

and its effect on the instrument’s abilityto accurately measure the depth of theFraunhofer line, the measurement was re-peated three independent times. Eachmeasurement was conducted in a separatecooldown of the cryostat. To quantify sta-bility, the amplitude of the Fraunhofer linewas fit in a direct measurement of the So-lar spectrum (described in detail in Section5) using each of the three independent flatfield measurements. The measured ampli-tude was shown to vary at the 1.4% levelbetween each of the three measurements.

4.6. Stray Light Performance

The NBS optics must also reject sourcesof off-axis emission. In the case of CIBER,this is dominated by stray light from theEarth. The response to extended off-axisemission is given by Bock et al. (1995) tobe

Istray =1

4π

∫g(θ)Is(θ, φ)dΩ, (12)

where Istray is the stray light signal ob-served surface brightness on the sky, Isis the surface brightness of the emittingoff-axis source, and g(θ) = 4π

ΩFOVG(θ) is

the telescope gain function, where G(θ) isthe telescope response to an off-axis pointsource normalized to unity on axis, andΩFOV is the telescope field of view.

To measure G(θ) in the laboratory,we followed the procedure described inTsumura et al. (2011). This consisted ofreplacing the PICNIC array with a warmsilicon photo-diode5 and mounting the

5S10043 manufactured by Hamamatsu photonics(http://www.hamamatsu.com)

18

NBS to a custom optical bench capable ofrotation to 90 from on-axis. A collimatedhalogen source was chopped at 20 Hz asthe angle to the NBS boresight was var-ied. The amplitude of the modulated sig-nal was measured with a lock-in amplifierreferenced to the signal from the chopper.The dynamic range of this measurementwas limited due to the small bandwidth ofthe NBS such that the largest angle withsufficient signal to noise was 25 from cen-ter. At this angle, G(25) = 5 × 10−5 andthe function displayed a sharp slope, sug-gesting it continues to fall at higher angles.

The night-side brightness of the earthin the NBS band is dominated by airglowemission and is approximately 103 nW m−2

sr−1 (Leinert et al. 1998) at 854 nm. As theearth limb is > 55 off-axis during observa-tions, the maximum projected solid angleis Ω⊕ = 0.2 sr. Even in the overly con-servative case in which G(θ) is assumedto remain flat at 5 × 10−5 to higher an-gles than were measured in the lab, theintegral in Equation 12 results in a signalIstray < 1 nW m−2 sr−1. We therefore con-clude that off-axis emission from the earthcontributes a negligible systematic error tothe measurement.

4.7. Calibration

The NBS must be accurately calibratedin order to compare our ZL measurementswith ZL models based on DIRBE obser-vations. Specifically, we need to trans-late the NBS ZL measurements to theDIRBE 1.25µm band with an accuracyof . 5 % to distinguish between models.To perform this comparison, we must ab-solutely calibrate the NBS, then translatethe ZL brightness measured at 854 nm to

1.25µm using the CIBER-LRS ZL spectralcolor. To reduce the contribution of theuncertainty in the NBS calibration to theoverall uncertainty in this chain, the NBSmust be accurately calibrated relative toDIRBE. For comparison, the absolute cali-bration uncertainty in standard calibrationstars like Vega is typically ∼ 1% and the1.25 µm DIRBE channel absolute calibra-tion uncertainty is 3% (Burdick & Mur-dock 1997) obtained on Sirius. In addi-tion to the laboratory calibration we de-scribe below, the NBS observes Vega inevery flight for an independent calibrationcheck.

A schematic representation of the NBScalibration set up is shown in Figure 15of Zemcov et al. (2011), where light from astable source is coupled into an integratingsphere which is viewed by the NBS anda reference detector absolutely calibratedby NIST. The radiance of the light sourceis varied and the response of the NBS isreferenced to the detector.

Our primary photometric calibrationmeasurements are derived from the monochro-matic data described in detail in Sec-tion 3.2 which use the SIRCUS laser fa-cility as the stabilized light source (Brownet al. 2006). The reference detectors whichmeasure absolute radiance are constructedof single element large area Si photodi-odes and baffles that limit the angular ex-tent light reaching the detector. A smallersphere feeding the larger sphere, which isviewed by the NBS, is constantly moni-tored with Si photodiodes to ensure thelight levels remain known during the NBScalibration given the ratio of large sphereradiance to monitor signal previously de-termined with the reference detectors. Un-

19

certainties in the deduced narrow band ra-diance are approximately 0.3%, dominatedby contributions from the sphere unifor-mity and the effects of out-of-field light.

Photocurrent measurements using thelaser source determine the NBS absoluteresponsivity, at a discrete set of wave-lengths. We calculate an array-wide map

of the calibration factor (∫Iλ′dλ

′

i) for a spec-

trally flat source from the ratio of the mea-sured NBS photocurrent and absolute radi-ance. The integral is evaluated from scansof the SIRCUS laser across a 100 nm widerange. After applying the flat field cor-rection, described in Section 4.5, the arraywide mean response factor is (602±10 nWm−2 sr−1) / (e− s−1) after correcting forthe transmittance of the additional windowused in the laboratory. The 1σ uncertaintyin the quoted absolute calibration factor isdetermined by the variance across all pix-els in the array, added in quadrature withthe 0.3% systematic uncertainty from theabsolute calibration quoted by NIST.

This calibration factor was checked bycoupling a supercontiuum laser source tothe integrating sphere referenced to a de-tector system capable of R ∼ 500 spectralmeasurements, calibrated to measure radi-ance traceable to fixed-point blackbodies.The supercontinuum laser produces a rel-atively flat featureless spectrum across theNBS band. Measurements of the photocur-rent induced by the broad band sourcewere taken at four light levels rangingfrom 500 to 2000 e−/s while simultane-ously monitoring the absolute radiance ofthe sphere. Dark signals were referencedto the ambient light background of the lab-oratory. Each exposure is corrected usingthe flat field template described in Sec-

tion 4.5. The absolute calibration factor istaken to be the slope of the linear corre-lation between the NIST measured abso-lute radiance and the mean photocurrentof each exposure. The resulting value is(603 ± 12 nW m−2 sr−1) / (e− s−1), aftercorrecting for the transmittance of the ad-ditional window used in the laboratory, inexcellent agreement with the measurementfrom the narrow band source. The quoteduncertainty is dominated by instrumentnoise in the NIST spectrographs and lin-earity during the boot-strapping measure-ments which pushes the dynamic range ofthe reference spectrographs.

The SIRCUS data set provides abso-lute calibration and a full spectral responsemeasurement for every pixel. This allowsfor a more sophisticated treatment of flightdata. The response over the array may becalculated for any source spectrum usingEquation 1. Thus we can make templateresponse maps over the array to ZL andEBL that fully include flat field gain andvariation in the spectral response over thearray.

4.8. Sensitivity

We estimate the complete instrumentsensitivity by simulating a series of 50 sobservations using measured dark data asinput noise. The data used in this cal-culation were taken in full flight configu-ration on the launch gantry one day be-fore the second CIBER flight with the coldshutter closed. There were 30 independent50 s integrations obtained. The variancebetween instances contains fluctuations inboth dark current spatial morphology andcorrelated read noise. We added a modelZL spectrum convolved with the instru-

20

ment response (e.g. Figure 7) to each in-dependent integration. The amplitude ofthe Fraunhofer line in the simulated obser-vation was then fit and compared to theinput value. The measured 50 s integra-tion sensitivity is taken to be the stan-dard deviation of the distribution of resid-uals (input sky brightness - measured skybrightness), which includes all sources ofcorrelated and uncorrelated noise over thearray. The sensitivity is determined to be∆ZL = 21.1 nW m−2 sr−1, shown in Table2, converted to ZL units assuming the cal-ibration factor from Section 4.7. Since theflight photocurrent varies between 0.5 and1.5 e−/s, the contribution of photon noiseis small. Given the typical brightness of500 nW m−2 sr−1 for high ecliptic latitudeZL brightness (Tsumura et al. 2010), thisleads to a ∆ZL/ZL . 4.4 % accuracy onthe ZL brightness in a 50 s observation.

5. Laboratory Measurement of theSolar Spectrum

As the NBS is designed to measure ab-sorption in Solar light scattered off the in-terplanetary dust cloud, a high signal tonoise measurement of sunlight is a usefultool for testing end to end system func-tionality. To obtain these data, we pointeda long optical fiber at the Sun and cou-pled the other end to the integrating sphereto uniformly illuminate the NBS aperture.The coupling efficiency to the sphere wasoptimized to produce an array photocur-rent near 200 e/s, where the signal to noiseis very high in each pixel without nearingsaturation. We measured the Solar spec-trum presented from a 200 s exposure andthe flat-field correction described in Sec-tion 4.5.

For comparison, we calculated a pre-dicted model spectrum by convolving thehigh resolution Solar measurements of Del-bouille et al. (1990) with the instrumentspectral response function obtained fromdiscrete-wavelength SIRCUS laser data.The model was fit to the data with a sim-ple χ2 minimization using two free param-eters: a normalization factor and an addi-tive offset to account for stray light fromthe laboratory. The measured Solar spec-trum is shown in Figure 10 in both oneand two dimensions alongside the best fitmodel. With only two free parameters, themaximum deviation of the residual of datafrom model is 0.24%.

6. Second Flight Performance

The CIBER instrument was flown on11 July 2010 from White Sands MissileRange. During this flight the NBS per-formed well, with the shutter, calibrationlamp, extendable/retractable baffle andarray all operating normally. All flightevents occurred according to the plannedsequence, and the attitude control systemgave stable pointing on all of the astro-nomical fields. The NBS obtained imagesof the sky in flight consistent with focusdata from laboratory measurements. TheCIBER instrument was recovered for post-flight testing and future flights.

The NBS photocurrents in the secondflight range between 2 e−/s at an eclipticlatitude of 10 to 1 e−/s at an ecliptic lat-itude of 90. Based on the laboratory cal-ibration, these levels correspond to a ZLbrightness ranging from 500 to 1000 nWm−2 sr−1, which matches a simple extrap-olation of the DIRBE sky brightness toNBS wavelengths. There is no measurable

21

Table 2: Measured System-Level Instrument Characteristics.Parameter Measured ValueMean Dark Current Stability 0.5± 0.1 e−/sFocal Plane Temperature Stability < 5.3µK/sSpectral Resolution λ/∆λ 1004 ± 12Line Contrast 26 %Lab Responsivity (602± 10 nW m−2 sr−1) / (e− s−1)Out of band spectral response TOB < 10−6 for 0.4 ≤ λ ≤ 2.5µmFlat field response ∆FF/FF gives ∆ZL/ZL = 1.4 %ZL sensitivity in 50 s (1σ) ∆ZL = 21.1 nW m−2 sr−1

Fig. 10.— Left: Measured and best fit 1DSolar spectra obtained from coupling sun-light to the integrating sphere in the labo-ratory. The calculated Solar response wasgenerated from the high resolution mea-surements of Delbouille et al. (1990) con-volved with the measured NBS spectral re-sponse. Error bars are 1 σ statistical. Topright: Flat field corrected 2D measured So-lar spectrum. Bottom right: CalculatedSolar response with fitted amplitude andoffset. The color scales are matched in the2D images.

thermal response in the NBS second flightdata. The in-flight dark level obtainedwith the shutter closed was 0.43 e−/s, con-sistent with the pre-flight dark images ob-tained on the launch gantry. The arraytemperature was stable in flight, giving amaximum drift of 5.3 µ K / s over a 50s integration during astronomical observa-tions.

7. Astrophysical Systematics

In addition to instrumental effects, esti-mating the ZL intensity from a Fraunhoferline is subject to several astrophysical sys-tematic errors.

7.1. Dynamical Effects

The Fraunhofer line profile can be al-tered by dynamical effects within the in-terplanetary dust cloud. Reynolds et al.(2004) have done an extensive study ofthese effects using the Mg I Fraunhofer lineat 581.4 nm along various lines of sight.The Doppler inferred velocity of the linewas found to depend strongly on the Solarelongation angle, ε, with peak magnitudesof ∼15 km/s around ε = 90. At the spec-tral resolution of the NBS (shown in Ta-ble 2), this effect is barely detectable (∼

22

10× smaller than the instrument FWHM).To quantify uncertainty caused by veloc-ity variation, we simulated shifting the ZLFraunhofer line across a conservative rangespanning ± 20 km/s from rest and fittingthe resulting spectrum assuming zero. Thebest fit ZL amplitude was altered by atmost 0.3%.

Reynolds et al. (2004) also find a kine-matic broadening of the Fraunhofer linecorresponding to a velocity dispersion of∼ 15 km/s. The Ca II line has an intrinsicwidth of 70 km/s and the FWHM of theNBS bandpass is 245 km/s. Since the lineis not fully resolved by the NBS and the ef-fects of broadening are small compared tothe intrinsic width, the dynamical effectshave negligible impact on total ZL bright-ness measurements.

7.2. Galactic Fraunhofer Line Fore-grounds

NIR EBL measurements are subject toGalactic foregrounds consisting of Inte-grated Stellar Light (ISL) from unmaskedstars and Diffuse Galactic Light (DGL),caused by starlight scattered by interstel-lar dust. Both of these signals will containFraunhofer lines at some level. Detectedstars are masked and removed from anal-ysis, but the aggregate signal from fainterstars is present and must be modeled.

Bernstein et al. (2002b) have generatedISL spectral templates at 0.4 nm resolu-tion in the optical which contain the Ca II

H (396.85 nm) and K (393.37 nm) lines.Both lines appear shallower in the ISLthan in the solar spectrum, but by differ-ent amounts. For this reason, a generalizedconclusion about the depth of Fraunhoferlines in the ISL relative to Solar would be

inaccurate. Unfortunately, these high reso-lution spectra do not cover the Ca II tripletand are therefore not directly applicable tothis study. Instead, for this estimate weapply the lower resolution ISL template ofMattila (1980a,b), which spans the UV toNIR at 5 nm resolution. This resolution isinsufficient to individually resolve all of theCa II triplets in the band. By comparinghigh-resolution measurements of the So-lar spectrum (Delbouille et al. 1990) con-volved with a 5 nm Gaussian responsefunction, we estimate that the Ca II tripletappears at near Solar EW in the ISL.

Fraunhofer lines in DGL have also beendetected in the optical (Mattila et al.2011). As DGL arises from stellar lightscattered by diffuse dust clouds, we assumethe EW of the Fraunhofer line in the DGLis the same as it is in the ISL. The totalcontribution from DGL can be modeled byscaling the measured DGL to 100 µm dustemission (Matsuoka et al. 2011) as well asneutral hydrogen column density (Mattilaet al. 2011).

To estimate the expected level of theGalactic Fraunhofer signal, we take the ex-ample of a field in the direction of Bootesobserved by CIBER (ecliptic longitude of200.57 and latitude of 46.633) and as-sume that both DGL and ISL contributeCa II absorption features with Solar EW.The level of the ISL is determined from thestar counts predicted by the Trilegal modelcode (Vanhollebeke et al. 2009) below acutoff magnitude of 14, comparable to theexpected completeness in an NBS image.The DGL intensity is estimated using thecorrelation with neutral hydrogen columndensity (nH) proposed by Mattila et al.(2011), with nH = 1.03×1020 cm−2 (Dickey

23

& Lockman 1990). The ZL intensity istaken from Kelsall et al. (1998), scaledto NBS wavelengths assuming the spec-tral template generated from the CIBER-LRS (Tsumura et al. 2010). With theaforementioned assumptions, we estimatea contribution at 854 nm of 15 nW m−2

sr−1from ISL, 10 nW m−2 sr−1from DGLand 386 nW m−2 sr−1from ZL, or a 6.5%contribution from the Galactic Fraunhoferline signal to the total sky brightness. ThisGalactic contribution can be assessed andremoved with an improved ISL model spec-trum.

7.3. Raman Scattering

If significant Raman scattering effectsare present in the ZL, Fraunhofer lines willbe filled, causing the relationship betweenline depth and continuum brightness to besystematically biased. Dube et al. (1979)compare this ratio for a Fraunhofer line insunlight reflected off the moon to the ZL,and find them to agree within 1.5%, lim-ited by instrumental error. This suggeststhat if present at all, Raman scatteringis a small effect, consistent with the no-tion that interplanetary dust particles havelarge dielectric constants with few free elec-trons.

8. Summary

We describe the design and performanceof a specialized wide-field, narrow-bandspectrometer for measuring the absoluteZL brightness using the EW of the 854.2nm Ca II Fraunhofer line, viewed in sun-light scattered by interplanetary dust. Thesystematics limiting the uncertainty, bothinstrumental and astrophysical, are dis-cussed and quantified through an exten-

sive suite of laboratory characterizationand simulation. The aggregate instrumen-tal uncertainty is dominated by read noisefluctuations and flat field error.

The instrument was flown on a secondsounding rocket flight in July 2010 anddemonstrated expected performance, withnominal photo currents, temperature sta-bility, calibration lamp response, and darkcurrents. Preliminary reductions of theflight data show clear significant detectionsof the Fraunhofer line. A third flight wascarried out successfully in March 2012. In-depth analysis of the science data is cur-rently underway and results will be re-leased in the near future.

Acknowledgements

This work was supported by NASAAPRA research grants NNX07AI54G, NNG05WC18G,NNX07AG43G, NNX07AJ24G, and NNX10AE12G.Initial support was provided by an awardto J.B. from the Jet Propulsion Labo-ratory’s Director’s Research and Devel-opment Fund. Japanese participationin CIBER was supported by KAKENHI(20·34, 18204018, 19540250, 21340047 and21111004) from Japan Society for the Pro-motion of Science (JSPS) and the Ministryof Education, Culture, Sports, Science andTechnology (MEXT). Korean participa-tion in CIBER was supported by the Pi-oneer Project from Korea Astronomy andSpace science Institute (KASI).

Certain commercial equipment, instru-ments, or materials are identified in thispaper to foster understanding. Such iden-tification does not imply recommendationor endorsement by the National Instituteof Standards and Technology, nor doesit imply that the materials or equipment

24

identified are necessarily the best availablefor the purpose.

This work utilizes SOLIS data obtainedby the NSO Integrated Synoptic Program(NISP), managed by the National SolarObservatory, which is operated by the As-sociation of Universities for Research inAstronomy (AURA), Inc. under a cooper-ative agreement with the National ScienceFoundation.

We would like to acknowledge the dedi-cated efforts of the sounding rocket staff atthe NASA Wallops Flight Facility and theWhite Sands Missile Range. We also ac-knowledge the work of the Genesia Corpo-ration for technical support of the CIBERoptics. A.C. acknowledges support from anNSF CAREER award, B.K. acknowledgessupport from a UCSD Hellman FacultyFellowship, K.T. acknowledges supportfrom the JSPS Research Fellowship forYoung Scientists, and M.Z. and P.M.K ac-knowledge support from NASA Postdoc-toral Fellowship.

REFERENCES

Aharonian, F., et al. 2006, Nature, 440,1018

Allen, C. W. 1976, Astrophysical Quanti-ties, ed. Allen, C. W.

Beggs, D. W., Blackwell, D. E., Dewhirst,D. W., & Wolstencroft, R. D. 1964, MN-RAS, 127, 329

Bernstein, R. A., Freedman, W. L., &Madore, B. F. 2002a, ApJ, 571, 85

—. 2002b, ApJ, 571, 85

—. 2002c, ApJ, 571, 107

—. 2005, ApJ, 632, 713

Berta, S., et al. 2010, A&A, 518, L30+

Bock, J., et al. 2012, ArXiv e-prints

Bock, J. J., Lange, A. E., Matsumoto,T., Eisenhardt, P. B., Hacking, P. B.,& Schember, H. R. 1994, ExperimentalAstronomy, 3, 119

Bock, J. J., Lange, A. E., Onaka, T., Mat-suhara, H., Matsumoto, T., & Sato, S.1995, Appl. Opt., 34, 2268

Brown, S. W., Eppeldauer, G. P., & Lykke,K. R. 2006, Appl. Opt., 45, 8218

Burdick, S. V., & Murdock, T. L. 1997,COBE Final Report: DIRBE CelestialCalibration, Tech. rep.

Cambresy, L., Reach, W. T., Beichman,C. A., & Jarrett, T. H. 2001, ApJ, 555,563

Delbouille, L., Roland, G., & Neven, L.1990, Atlas photometrique DU spec-tre solaire de [lambda] 3000 a [lambda]10000, ed. Delbouille, L., Roland, G., &Neven, L.

Dickey, J. M., & Lockman, F. J. 1990,ARA&A, 28, 215

Dube, R. R., Wickes, W. C., & Wilkinson,D. T. 1977, ApJ, 215, L51

—. 1979, ApJ, 232, 333

Dwek, E., & Arendt, R. G. 1998, ApJ, 508,L9

Dwek, E., Arendt, R. G., & Krennrich, F.2005, ApJ, 635, 784

25

Eather, R. H., & Reasoner, D. L. 1969,Appl. Opt., 8, 227

Fixsen, D. J., Dwek, E., Mather, J. C.,Bennett, C. L., & Shafer, R. A. 1998,ApJ, 508, 123

Hauser, M. G., et al. 1998, ApJ, 508, 25

Hodapp, K., et al. 1996, New Astronomy,1, 177

Jacquinot, P. 1954, Journal of the OpticalSociety of America (1917-1983), 44, 761

Juvela, M., Mattila, K., Lemke, D., Klaas,U., Leinert, C., & Kiss, C. 2009, A&A,500, 763

Keenan, R. C., Barger, A. J., Cowie, L. L.,& Wang, W.-H. 2010, ApJ, 723, 40

Kelsall, T., et al. 1998, ApJ, 508, 44

Kutyrev, A. S., Arendt, R., Dwek, E.,Moseley, S. H., Rapchun, D., & Silver-berg, R. F. 2008, in Presented at theSociety of Photo-Optical Instrumenta-tion Engineers (SPIE) Conference, Vol.7014, Society of Photo-Optical Instru-mentation Engineers (SPIE) ConferenceSeries

Lee, D. H., et al. 2010, Journal of Astron-omy and Space Sciences, 27, 401

Leinert, C., et al. 1998, A&AS, 127, 1

Levenson, L. R., Wright, E. L., & Johnson,B. D. 2007, ApJ, 666, 34

Madau, P., & Silk, J. 2005, MNRAS, 359,L37

Marsden, G., et al. 2009, ApJ, 707, 1729

Martin, P. G. 1988, ApJS, 66, 125

Matsumoto, T., et al. 2005, ApJ, 626, 31

Matsuoka, Y., Ienaka, N., Kawara, K., &Oyabu, S. 2011, ApJ, 736, 119

Matsuura, S., et al. 2011, ApJ, 737, 2

Mattila, K. 1980a, A&AS, 39, 53

—. 1980b, A&A, 82, 373

—. 1980c, A&A, 82, 373

—. 2003, ApJ, 591, 119

Mattila, K., Lehtinen, K., Vaisanen, P.,von Appen-Schnur, G., & Leinert, C.2011, ArXiv e-prints

Misconi, N. Y. 1976, A&A, 51, 357

Penin, A., et al. 2011, ArXiv e-prints

Reynolds, R. J., Madsen, G. J., & Moseley,S. H. 2004, ApJ, 612, 1206

Schroedter, M., et al. 2005, ApJ, 634, 947

Tsumura, K., et al. 2010, ApJ, 719, 394

—. 2011, ApJS (Submitted)

Vanhollebeke, E., Groenewegen, M. A. T.,& Girardi, L. 2009, A&A, 498, 95

Wright, E. L. 1998, ApJ, 496, 1

—. 2001, ApJ, 553, 538

Yamamuro, T., Sato, S., Zenno, T.,Takeyama, N., Matsuhara, H., Maeda,I., & Matsueda, Y. 2006, Optical Engi-neering, 45, 083401

Zemcov, M., et al. 2011, ArXiv e-prints1112.1424

26

Zemcov, M., Blain, A., Halpern, M., &Levenson, L. 2010, ApJ, 721, 424

This 2-column preprint was prepared with theAAS LATEX macros v5.2.

27