Global‐Scale Observations of the Limb and Disk Mission

Home , Kourou

RESEARCH ARTICLE Global‐Scale Observations of the Limb and Disk 10.1029/2020JA027797 Mission Implementation: 1. Instrument Design Special Section: and Early Flight Performance Early results from the Global‐scale Observations of William E. McClintock1 , Richard W. Eastes1 , Alan C. Hoskins1, Oswald H. W. Siegmund2 , the Limb and Disk (GOLD) 2 3 4 4 mission Jason B. McPhate , Andrey Krywonos , Stanley C. Solomon , and Alan G. Burns 1Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, CO, USA, 2Space Sciences Laboratory, 3 This article is a companion to University of California, Berkeley, CA, USA, Florida Space Institute, University of Central Florida, Orlando, FL, USA, McClintock et al. (2020), https://doi. 4National Center for Atmospheric Research, High Altitude Observatory, Boulder, CO, USA org/10.1029/2020JA027809.

The Global‐scale Observations of the Limb and Disk (GOLD) is a National Aeronautics and Key Points: Abstract • GOLD makes thermospheric images Space Administration mission of opportunity designed to study how the Earth's ionosphere‐thermosphere of OI and N2 LBH emissions, system responds to geomagnetic storms, solar radiation, and upward propagating atmospheric tides and ionospheric images of OI emission waves. GOLD employs an instrument with two identical ultraviolet spectrographs that make observations of and observes O2 absorption on the limb the Earth's thermosphere and ionosphere from a commercial communications satellite owned and • An overview of the instrument operated by Société Européenne des Satellites (SES) and located in geostationary orbit at 47.5° west design and performance based on longitude (near the mouth of the Amazon River). They make images of atomic oxygen 135.6 nm and N laboratory characterization is 2 provided Lyman‐Birge‐Hopfield (LBH) 137–162 nm radiances of the entire disk that is observable from geostationary • Imaging and spectroscopic orbit and on the near‐equatorial limb. They also observe occultations of stars to measure molecular oxygen fi performance con rm laboratory column densities on the limb. Here, we provide an overview of the instrument and compare its prelaunch results. Radiometric sensitivity using stars is ~20% less than ground and early flight measurement performance. Direct comparison of LBH spectra of an electron lamp taken measurement before launch with spectra on orbit provides evidence that both cascade and direct excitation are important sources of thermospheric LBH emission. Plain Language Summary The Global‐scale Observations of the Limb and Disk (GOLD) is a Correspondence to: W. E. McClintock, National Aeronautics and Space Administration mission of opportunity designed to study how the [email protected] Earth's ionosphere‐thermosphere system responds to geomagnetic storms, solar radiation, and upward propagating tides on time scales as short as 30 min. GOLD employs two identical ultraviolet spectrographs

Citation: that make observations of the Earth's thermosphere and ionosphere from a commercial communications McClintock, W. E., Eastes, R. W., satellite owned and operated by SES and located in geostationary orbit at 47.5° west longitude (near the Hoskins, A. C., Siegmund, O. H. W., mouth of the Amazon River). They make images of atomic oxygen 135.6 nm and N2 LBH radiances of the McPhate, J. B., Krywonos, A., et al. ‐ (2020). Global‐scale observations of the entire disk that is observable from geostationary orbit and on the near equatorial limb. They also observe limb and disk Mission implementation: occultations of stars to measure molecular oxygen column densities on the limb. Here we describe the GOLD 1. Instrument design and early ﬂight instrument including its optical system and detector. Its performance was characterized in the lab before performance. Journal of Geophysical Research: Space Physics, 125, launch. We compare measurements of laboratory sources made then to observations of the thermosphere e2020JA027797. https://doi.org/ after launch and ﬁnd good agreement. 10.1029/2020JA027797

Received 11 JAN 2020 1. Introduction Accepted 13 APR 2020 ‐ Accepted article online 15 APR 2020 The Global scale Observations of the Limb and Disk (GOLD) is a National Aeronautics and Space Administration mission of opportunity designed to study how the Earth's ionosphere‐thermosphere system responds to geomagnetic storms, solar radiation, and upward propagating tides (Eastes et al., 2017). GOLD employs two identical ultraviolet spectrographs that make images of the Earth's thermosphere and ionosphere from a commercial communications satellite owned and operated by SES and located in geostationary equatorial orbit at 47.5° west longitude (near the mouth of the Amazon River). From this vantage

point, the instrument measures atomic oxygen 135.6 nm and N2 LBH emission intensities and performs stellar occultation observations. These support the speciﬁc measurement requirements of the GOLD Mission: Images on the disk:

©2020. American Geophysical Union. 1. Images of the disk of the Earth, over a latitude range of ±60° and a longitude range of ±70° relative to All Rights Reserved. spacecraft nadir, with a cadence of 30 min

MCCLINTOCK ET AL. 1of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

2. Measurements of emission from atomic oxygen (O) at 135.6 nm and from the Lyman‐Birge‐Hopﬁeld

(LBH) bands of molecular nitrogen (N2) with a cadence of 30 min and a spatial resolution of 250 × 250 km (at nadir) and with a spectral resolution of 0.2 nm full width half maximum 3. Measurements of the atomic oxygen (O) 135.6 nm emission from the nighttime equatorial arcs (northern and southern) at a 1‐hr cadence and a spatial resolution of 2° in latitude and a longitudinal resolution of 100‐km (at nadir) Proﬁles on the limb:

1. Measurements of N2 LBH emission up to 350 km above the surface at the equatorial limbs with an altitude resolution of 50 km 2. Stellar occultation measurements in the 135–155 nm wavelength range and 140–200 km altitude range. In each channel, a single‐mirror telescope images the Earth onto the 11.2° tall entrance slit of an imaging spectrograph. The spectrograph disperses the light and images it onto a two‐dimensional detector that provides an independent spectrum for every position along the slit resulting in a single one‐dimensional spatial by one‐dimensional spectral image. Only a narrow slice of the image made by the telescope enters the spectrograph at a given time and the second spatial dimension of a three‐dimensional image cube is built up “scanning” the Earth's image across the slit in small steps from east to west (Figure 5). This is accomplished within the GOLD instrument by two scan mirrors, one located in front of each telescope. The GOLD instrument has two operating states: science and safe. During science, it operates from 03:00 spacecraft local time (SLT) until 21:30 SLT each day. From 03:00 SLT until 20:00 SLT, it acquires complete spatial‐spectral image cubes of the disk and sunlit limb with a 30‐min cadence. Beginning at 17:00 SLT and continuing until 21:30 SLT, it also scans the nightside disk beginning at longitudes approximately 15° east of the terminator. Limb scans and occultations are interleaved with the disk scans to maintain a 30‐min cadence as described by McClintock et al. (2020). In order to protect itself from incursion by the Sun, GOLD enters its safe state at 21:30 SLT and remains there until 03:00 SLT the following day.

2. Instrument Design The GOLD instrument is a hosted payload aboard the SES‐14 spacecraft, which is based on the Eurostar E3000EOR bus manufactured by Airbus Defense and Space (ADS). GOLD was integrated to the spacecraft at the ADS facility in Toulouse France and is operated by SES in Luxemburg through a contract to SES Government Solutions. GOLD was launched aboard Ariane ﬂight VA241 from Ariane Launch Complex No. 3 in Kourou, French Guiana on 25 January 2018. After orbit insertion and spacecraft commissioning, GOLD instrument checkout started in early September. Nominal operations and observing began 9 October 2018.

During the day, GOLD measures O atomic (135.6 nm) and N2 LBH (137–162 nm) thermospheric radiances above the disk and on the limb. It also observes stellar occultations to measure O2 densities by absorption on the limb. At night, it measures 135.6 nm radiances emitted by O+ as it recombines in the ionosphere. Data collected by the instrument are packetized and routed to the spacecraft for immediate downlink to an SES ground station in Woodbine Maryland where they are recorded and then transmitted to the GOLD Science Operations Center (SOC) at the Laboratory for Atmospheric and Space Physics (LASP) at the University of Colorado. There they are formatted before transmission to the GOLD Science Data Center (SDC) at the University of Central Florida. The SOC also prepares weekly instrument command loads that are transmitted to the SES satellite operations center for upload to the spacecraft where they are routed to the instrument over a 1553 communications bus.

2.1. Instrument Description The instrument, illustrated in Figure 1 and summarized in Table 1, is mounted to the spacecraft nadir‐facing deck and views the entire ~18° observable disk and limb of the earth (~±80° in longitude and latitude about the subspacecraft point). It consists of two identical, independent telescope‐imaging spectrograph channels, referred to as Channels A and B. These are mounted on either side of the GOLD Electronics Box (GOLD E‐box) that controls the channels and provides the instrument electrical interface to the spacecraft. The channels are equipped with large sunshades and protective covers that were opened after the spacecraft

MCCLINTOCK ET AL. 2of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 1. (a) The GOLD instrument consists of two independent telescope‐spectrograph channels, referred to as A and B, and an electronics box located between the two channels. It is mounted to the spacecraft nadir deck using four kinematic struts. The top surfaces of A and B face north and south, respectively. (b) Section view of Channel A viewing nadir in the east‐west direction. Channel B is identical to A and is rotated 180° about nadir for packaging convenience.

reached its ﬁnal orbit. In order to avoid mechanical interference between the sunshades, the channels are oriented so that the top of A faces north and the top of B faces south.

2.2. Optical Design Figure 1b shows a section view for a single channel. A sunshade with a 31° east‐west opening angle defines the instrument field of regard (FOR), which is larger than the Earth disk. This accommodates any spacecraft pointing adjustments and accounts for possible instrument‐spacecraft alignment errors. This aperture is followed by a plane mirror, mounted in a rotation mechanism that can control its angle with − a precision and accuracy of 1.373 × 10 3 degrees. The mirror projects light from the Earth through the system aperture stop and onto a telescope consisting of a single spherical mirror with a 150 mm focal length. The telescope images the scene onto one of three interchangeable entrance slits of an imaging spectrograph. These are designated high‐spectral‐resolution (GOLD‐HR: 0.2 × 29.4 mm), low‐spectral‐resolution (GOLD‐LR: 0.4 × 29.4 mm), and stellar occultation (GOLD‐OCC: 2.6 × 27.3 mm). The spectrograph is a modified Wadsworth design (Wadsworth, 1896) in which the parabolic collimating mirror and spherical grating are replaced by a collimator and grating each with a toroidal surface (Krywonos et al., 2006). This collimator‐grating combination (Table 1) forms a 0.71 × 0.96 (spectral × spatial) demagnified image of the slit at the spectrograph focal plane. GOLD gratings, which were fabricated holo- graphically by Jobin Yvon in Palaiseau France, have a ruling density of 3000 grooves per mm that produce a mean dispersion of 1.17 nm/mm. GOLD mirrors and gratings were fabricated from low‐expansion Ohara Clearceram™‐Z and Corning® 7980 fused silica, respectively and coated with aluminum‐magnesium fluor- ide (Quijada et al., 2014). The measured average spectral resolutions across the focal plane, which include optical aberration and detector point spread function (PSF) are 0.19 and 0.35 nm for HR and LR, respectively. There is a small, but nonnegligible focal plane distortion: the spectral images of the straight entrance slit are curved, and

MCCLINTOCK ET AL. 3of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Table 1 the amount of curvature is a function of wavelength so that the disper- GOLD Instrument Summary sion varies from 1.17 nm/mm at slit center to 1.174 nm/mm at slit top Scan mirror‐telescope and bottom. This variation is removed in GOLD data products by Entrance pupil 30.0 × 30.0 mm (square) applying an optical distortion correction, discussed below, which enables the use of a single wavelength vector for all the rows of a Focal length 150 mm spectral‐spatial image. The grating is mounted on a precision bidirec- Field of view HR 0.076° × 11.2° (0.2 × 29.4 mm) tional mechanism, the grating yaw mechanism (GYM), that can be LR 0.153° × 11.2° (0.4 × 29.4 mm) rotated by an uplinked command to adjust the wavelength coverage Occultation 1.00° × 10.4° (2.6 × 27.3 mm) over a range of ~±10 nm in ~0.05 nm steps. This is used to periodically Field of regard 30°E‐W × 21.2°N‐S reposition the spectrum on the detector in order account for gain fati- Spectrograph gue that occurs at the location of the bright 135.6 nm oxygen emission Effective focal length 285.2 mm Collimating mirror Toroidal (section 2.3, section 4.2). a Radii of curvature RDisp = 735.0 mm RX‐Disp = 585.0 mm Each channel is equipped with a shutter, located directly in front of the Grating Toroidal‐holographic a scan mirrors. These are closed between 21:30 SLT and 03:00 SLT to pre- Radii of curvature RDisp = 494.0 mm RX‐Disp = 570.4 mm Ruling density 3000 grooves/mm vent sunlight from entering the instrument. Dispersion 1.17 nm/mm Wavelength range 135–165 nm Slit magnification 0.71 × 0.96 (Spectral × Spatial) 2.3. Detector Spectral resolution HR slit 0.21 nm Spectra are recorded using a microchannel plate (MCP) detectors LR slit 0.35 nm equipped with cesium iodide photocathodes and cross‐delay line (XDL) Detector anodes (Siegmund et al., 1993; Siegmund et al., 2016). The detectors, Anode Cross delay line which are contained within vacuum housings with doors that can be Active area 27.0 × 32.0 mm (Spectral × Spatial) opened and closed by command, have a 27.0 × 32.0 mm (x, y) active area Format 1,600 × 1,800 (Spectral × Spatial) Pixel dimension where x is the dispersion (across slit) direction and y is the imaging (along Channel A 0.0172 × 0.0185 mm (Spectral × Spatial) slit) direction. They operate in a single‐event mode. Within each detector, Channel B 0.0174 × 0.0190 mm (Spectral × Spatial) photons arriving at the top surface of the input MCP initiate photo‐events Instrument by liberating electrons from the photocathode. Individual events are Mass 36.6 kg amplified in a stack of three MCPs arranged in a z‐configuration. Average power 75 W Dimensions 61.7 × 54.1 × 23.1 cm3 Nominal operating voltages across the plates, which depend on the stack a resistance, are 3,200v and 3,600v for Channels A and B, respectively. Disp is the dispersion plane X‐Disp is the cross‐dispersion plane. − These voltages result in a charge packets with a modal 107 e (1.6 pico‐coulombs) size. The packets impinge upon the anode where they divide between separate x and y collection electrodes. The fast (5 ns wide) signals on each axis travel in opposite directions to two pairs of amplifiers and associated processing electronics that locate them via time difference with 12‐bit precision in anode space and measure the amplitude of the pulse with 8‐bit precision, one photon at a time. Each event is reported by the electronics as two 12‐bit position data numbers (DN) relative to the anode coordinate system and one 8‐bit pulse amplitude DN. In addition to photon event locations, the detector also encodes the location of two fiducial pulses produced by the electronics at a 10 Hz rate. These are used to correct thermal stretch and shift introduced into the detector (x,y) locations by temperature changes in the amplifiers. The detector sensitivity to far ultraviolet light is significantly enhanced by a quantum efficiency grid (QEG) located parallel‐to and 6 mm above the MCP input surface. It consists of a square mesh with 0.02 mm diameter conducting wires with 0.34 mm spacing that is rotated 45° with respect to the x‐y axes. Negative high voltage is connected to this grid and reaches the MCP stack through a resistor that is in series with the grid and the stack. The differential voltage between the QEG and the top MCP, which is ~ −650 volts, repels photoelectrons that are created on the web between microchannels and directs them toward the channels. These would be lost in the absence of the grid. Because the GOLD detectors process one event at a time, there is an upper limit to the allowable instantaneous count rate across the entire 2‐D array, which is set by the electronics processing time or “global dead − − time” of ~τ =10 6 s. Events that arrive in times less than 10 6 s after an initial event are simply ignored (Siegmund et al., 2016). This results in a correctable nonlinear response in which the “true” count rate is related to the observed count rate by

MCCLINTOCK ET AL. 4of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

CG−obs CG−true ¼ : (1) 1 − τ·CG−obs

There is also a local count rate limitation that is due to the ﬁnite time required for individual pores in the MCP to recover after a photon event and is proportional to MCP stack resistance. Local dead time was characterized during ground calibration. The best‐ﬁt quadratic function is

CL−true ¼ CL−obs·1ðÞþ 0:098*CL−obs ; (2)

−1 which is valid for CL–obs <5s . The two count rate corrections, which are applied during ground processing (McClintock et al., 2020), primarily affect stellar observations because global and local count rates for disk and limb observations are typically less than 1.4 × 104 Hz and 0.1 Hz, respectively. Global corrections can be significant for occultations when the 1°‐wide occultation slit intersects the sunlit disk. In those cases, total event rates can reach 1–2×105 Hz for stars of average brightness. Local corrections can be important for the three brightest stars in the GOLD observing catalog (Gamma Orionis, Epsilon Orionis and Zeta Orionis), which are not observed because they exceed the 5 Hz limit in some wavelength intervals. Local count rates for all other GOLD stars are <0.5 Hz. Anode space is distorted relative to physical space, and two corrections are applied during data processing in order to convert locations from the first, distorted space (DNs in anode space) to locations within discrete bins (pixels) in the second undistorted space. The first is a temperature‐dependent stretch‐shift factor, which slightly expands anode space for increasing temperature and shifts it (Wilkinson et al., 2003). GOLD instrument diurnal temperatures vary between 15 °C and 40 °C leading to a maximum stretch‐factors of 1.003 in x and 1.005 in y and maximum shift values of ~8 DN in each axis. Stretch and shift are determined from the locations of the fiducial pulses. The second, which is independent of temperature, is nonlinear and is an artifact of the timing circuit that determines pulse‐centroid location (Wilkinson et al., 2003). It was characterized during assembly and test by placing a mask, which had a grid of 0.010 mm diameter holes separated by 1.000 mm, in contact with the input surface of the top MCP. This was illuminated with an ultraviolet lamp, providing a pattern of equally spaced “point” sources. Figures 2a and 2b show an image of the mask taken with the Channel A detector. Blue symbols indicate the pinhole locations reported by the output of the anode electronics. These are irregularly spaced. Red symbols are the positions of a regularly spaced grid in physical space that matches the average spacings of the blue symbols. Its grid spacing was determined from the average best‐fit

linear polynomials that map anode 12‐bit detector DNs (xDN(i,j), yDN(i,j)) to the physical locations of the pinholes (xM(i,j),yM(i,j)), measured in mm (xM(i,j) = i and yM(i,j) = j), where i,j is the ith column and jth row of the mask pinhole matrix.

i ¼ Δxj·xDN ðÞþi; j Cxj; (3) ¼ Δ ðÞþ; ; j yi·yDN i j Cyi (4)

Δxj and Δyi were determined for each row and each column of pinholes by least squared ﬁtting using equa- tions 3 and 4. These were averaged to yield the spacings Δx = 0.0172 mm per anode‐DN (58.14/mm) and Δy = 0.0185 mm (54.05/mm) per anode‐DN for Channel A and Δx = 0.0174 mm per anode‐DN (57.53/

mm) and Δy = 0.0190 mm (52.63/mm) per anode‐DN for Channel B. Cxj and Cyi were averaged to provide the offsets for the grid for each detector. Although the anode, which is larger than the MCP active area, can encode 4,096 × 4,096, the data processing software requires only a 1,600 × 1,800‐pixel array format to capture an entire 27.0 mm x 32.0 mm spectral x spatial focal plane image. Once the grid parameters are established, the distortion correction for each mask location is deﬁned as

i ¼ Δx·xDN ðÞþi; j Cx þ δxDN ðÞi; j ; (5) ¼ Δ ðÞþ; þ δ ðÞ; : j y·yDN i j Cy yDN i j (6)

Arrows in Figures 2a and 2b pointing from blue to red show the magnitudes of the correction that must be applied to anode space in order to specify the pinhole locations in physical space.

MCCLINTOCK ET AL. 5of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 2. (a) Blue asterisks indicate the locations reported by the anode electronics for a grid of pinholes with 1.000 mm spacing. Red circles are the locations determined by ﬁtting single linear polynomials in x and y to the anode locations. A box inscribes a region of the detector shown in an expanded view in (b) arrows indicate corrections that must be applied to map anode space to physical space. (c) and (d) Images of the displacements that must be added to the x,y data numbers reported by the detector electronics in order to transform then to locations in physical space.

The values of δxDN and δxDN for each detector xDN,yDN location were determined by interpolating the valueδxDN(i,j) and δyDN(i,j) at the detector locations xDN(i,j), yDN(i,j) of the mask. Figures 2b and 2c show the displacements that must be added to the detector DN values, xDN and yDN, respectively, in order to convert anode space to physical space. Detector spatial resolution (PSF), which is determined by the signal‐to‐noise in the timing ampliﬁers and timing jitter in the electronics, was determined at each pinhole location by ﬁtting Gaussians to the distributions of x,y values around mean values for the ~850 pinhole images obtained with the mask. The resulting mean full width half maximum (FWHM) values for both detectors were ~0.050 mm in both x and y with a standard deviation ~0.015 mm. In addition to location, the detector processing electronics generates an 8‐bit word proportional to the electron packet size (pulse height) produced by the MCP stack for the detected photon. Thus, each detected photon is encoded into a 12‐bit address +12‐bit address +8‐bit pulse‐height word. The instrument E‐box processor assembles these into a science data packet, which includes a header and up to 15,358 events, total for both detectors, for immediate downlink at a rate of 10 packets per second (equivalent to the 6.0 Mbit/s spacecraft data rate allocation). This approach enables all data binning to be done on the ground, allowing

MCCLINTOCK ET AL. 6of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

spatial‐temporal resolution analysis to be performed on an event‐by‐event basis. The maximum instantaneous data downlink rate required to support GOLD's disk and limb observations significantly is less than 153,580 events per second (4.9 Mbit/s). Detector events for daytime occultations can reach 2 × 105 Hz because the 1° slit subtends the illuminated atmosphere. This is reduced to <0.9 × 105 Hz telemetering only the 97 rows of the detector centered on the star's spectral‐spatial image. In order to maintain a constant 6.0 Mbit/s data rate, the processor appends fill words to packets that contain fewer than 15,358 events. The high‐gain MCP detectors in the GOLD instrument have a finite life- time due to MCP charge depletion, or gain fatigue, which lowers detector quantum efficiency as a function of total accumulated counts (Sahnow et al., 2011; Siegmund et al., 2016) This effect is most significant for OI—135.6 nm radiation, which produces local count rates that are an order of magnitude higher than those from LBH (Figure 6). fi Sensitivity changes are tracked and corrected during ground processing Figure 3. Blue asterisks are the instrument detection ef ciencies measured ‐ “fl fi ” for Channel A at five discrete wavelengths. Error bars represent the using images of a miniature Pen Ray® cold cathode at eld lamp systematic uncertainty at each wavelength, which is ±20%. The blue curve located within each channel. These are operated for 20 min prior to is the relative sensitivity derived by comparing a laboratory spectrum the nominal observing sequence that begins at 03:00 SLT. and a model LBH spectrum. Its shape has a relative uncertainty of ±5%. The Additionally, the GYM is used periodically to shift 135.6 to undepleated fi lled black circles and black line are the corresponding results for Channel regions of the detector. B. 3. Prelaunch Performance 3.1. Radiometric Sensitivity The aggregate instrument detection efficiency is defined as the product of all mirror reflectances, grating reflectance and diffraction efficiency, and the detector quantum efficiency. It was measured for both channels before launch at five wavelengths in a LASP vacuum‐calibration facility equipped with a monochromator that feeds an off‐axis parabolic mirror. At each wavelength, each channel was independently illuminated by a collimated, nearly monochromatic beam of light from the monochromator‐mirror assembly, and the count rate from its detector was recorded. Next, the rate that photons entered the aperture of the channel under test was measured by raster‐scanning the input beam with a calibrated photomultiplier tube and comparing the detector count rate to the raster‐sum photomultiplier count rate. The absolute detection efficiency of the photomultiplier was measured in a separate facility by comparing it to a National Institutes for Standards and Technology (NIST) calibrated photodiode (Canfield et al., 1973). Details of the technique are described by McClintock et al. (2015). The measurements, which have a wavelength‐to‐wavelength uncertainty of ±10% and an aggregate systematic uncertainty of ±20%, are shown in Figure 3 for Channels A and B as blue asterisks and black circles, respectively. The solid lines are the relative calibrations obtained from observations of an electron lamp discussed below. Their relative wavelength‐to‐wavelength uncertainties are estimated to be ±5% (Ajello et al., 1988). After instrument commissioning the detection efficiency values in Figure 3 were multiplied by 0.7 when a preliminary comparison of observations of the star HD 30836 (π4 Ori), which were made by allowing it to drift across the entrance slits of the spectrometers during commissioning, with published irradiance values from International Ultraviolet Explorer (IUE; Boggess et al., 1978; Bohlin, 1996; Bohlin & Bianchi, 2018) suggested that the ground calibration had overestimated instrument responsivity. This factor has been revised to 0.78 and 0.85 for Channel A and Channel B, respectively, based on above‐the‐atmosphere occultation observations of multiple stars (see section 4.7). However, data released through 22 December 2019 are calibrated with the initial value of 0.7 for each channel. These will be updated in the next data release.

3.2. Spectral Performance GOLD spectral imaging performance was characterized during ground test using an electron impact lamp

operating at 30 electron volts to produce an N2 LBH emission source with a spherical emitting volume ~ 1 mm in diameter (Ajello et al., 1988). The instrument was mounted on a manipulation platform and

MCCLINTOCK ET AL. 7of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 4. Laboratory LBH spectrum corrected for detector and optical distortion (black) compared to a simulated spectrum calculated using the model from Estes (2000). The lamp spectrum includes emission from NI multiplets located at 131.07, 131.27, 131.95, and 149.3 nm, which are not included in the model. Other NI multiplets at 132.7 and 141.2 nm, also not in the model, are obscured by LBH emission. Locations of ν′–ν″ bands are identiﬁed for ν′ =0–5.

placed in a large vacuum chamber with the lamp approximately 3 m in front of it resulting in an out‐of‐focus image of the source on the spectrograph entrance slits that was ~2 mm in diameter and ﬁlled ~100 detector rows. Entire spectral‐spatial images were built up by rotating the manipulator in order to slew the lamp image along the entire length of the slit mapping the observing angle of emitting region onto the detector row‐by‐row using the equation:

ψ ¼ Δ ðÞ− ; FT ·tan MY · Y· rDet rDet0 (7)

where FT is the telescope focal length (150 mm), ψ is the incoming angle of the central ray from the lamp relative to the telescope optic axis, MY is the spectrograph optical magniﬁcation (0.96), rDet is the detector row number, and rDet0 is the detector row number for ψ =0. Comparison of the row‐to‐row spectra during ground calibration indicated that the single detection efﬁciency vector for each channel (Figure 3) is adequate to describe the spectral response of the instru-

ment. If QT(λj, ψk) represents the detection efficiency for wavelength λj and angular slit position ψk, then QT (λj, ψk)=QT (λj). Spectra extracted from lamp images were compared to an LBH emission model in order to determine instrument optical distortion and spectral resolution and to derive its relative radiometric sensitivity as a function of wavelength. The black line in Figure 4 is a spectral plot made from the sum of the central 1,500 rows of a Channel A image after transforming detector output from anode space to physical space, applying an optical distortion correction to remove slit‐image curvature, and dividing by the relative sensitivity curve shown in Figure 3. Shown in red is a simulated spectrum generated by convolving an LBH line‐by‐line model (Eastes, 2000) calculated for room temperature (T = 293°K) with a Gaussian function that has a 0.19 nm FWHM. It is normalized to the measurements using the total photons in each spectrum, excluding the 149–150 nm and 130–132 nm wavelength ranges. In addition to in‐band LBH, atomic nitrogen and OI—135.6 nm emissions focused onto the detector, GOLD spectra include background signals arising from two sources associated with the optical system. The first is HI—121.56 nm and OI—130.4 nm emissions that are scattered by the diffraction grating into the primary 132–162 nm wavelength range. The second is light that is reflected from the top surface of the input MCP, returning to the grating. Here, it is collimated by the grating and reflected back toward the detector where

MCCLINTOCK ET AL. 8of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 5. Slit projections for Channel A point east and images of the slit are both tilted and slightly curved (imperceptible in this ﬁgure). These two features are introduced by the tilted scan mirrors and the imaging properties of the telescopes. Because Channel B is rotated by 180° about nadir relative to Channel A, its slit projections for point west and the greatest north‐south slit projection overlap occurs on the east limb.

it illuminates a fraction of the active area, which is wavelength dependent. Whereas 130.4 nm exhibits distinct contributions from both grating scatter and detector reflection, the geometry for 121.56 nm is such that there is no obvious detector‐reflected component. Nominal HI—121.56 nm and OI—130.4 nm radiances are ~60 kiloRayleighs (kR) and 10 kR, respectively (Meier et al., 2015), where 1 Rayleigh = 1010 − − − − − − photons m 2 s 1 nm 1 emitted into 4π steradians = 7.96 × 104 photons cm 2 steradian 1 s 1 (Hunten et al., 1956). These are significantly larger than radiances from OI—135.6 nm and LHB (typically less than 2 kR), and their contributions dominate the scattered and detector‐reflected signals in the flight data (Figure 6).

4. Flight Measurement Performance 4.1. Telescope Imaging and Sunshade Performance Each GOLD channel acquires spectral‐spatial image cubes by rotating its scan mirror to slew the image of the Earth across a selected spectrograph slit from east to west. The combination of telescope focal length and 29.4 mm slit length results in an 11.2° north‐south instantaneous ﬁeld of view (FOV). In order to image the entire ~18° tall Earth, the scan mirror is double‐sided. One side is tilted north by 2.5°, and the other south by 2.5°. This enables GOLD to image the disk in two swaths, one covering −0.6° to 10.6° and the other covering −10.6° to 0.6°. In the east‐west direction, GOLD's nadir spatial resolution is ~50 and ~100 km for HR and

MCCLINTOCK ET AL. 9of18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 6. Typical first‐light spectrum obtained before nominal operations began. The various background components are shown as dashed lines. Detector reflect is 130.4 nm light that is reflected from the top surface of the input MCP and returns to the grating. There it is collimated and reflected back toward the detector where it uniformly illuminates a fraction of the active area. Positions of OI and N1 atomic emissions as well as N2 LBH molecular bands are marked by short vertical lines.

LR, respectively. On the limb, these increase to 56 and 112 km. In the north‐south direction spatial resolution at nadir, including telescope and spectrograph optical aberration and detector PSF, is ~30 km. Because the scan mirrors are tilted in their mounts, which have rotation axes that are perpendicular to the telescope optic axes, the projections of the spectrograph entrance slits onto the disk are not parallel to the rotation axes. Figure 5 illustrates the effect for the Channel A slit where slit projections are shown in ~1.5° increments of its scan mirror angle. As the Channel A scan mirror rotates counter clockwise, increasing the opening angle (Figure 1), the slit image moves west, its angle with respect to north‐south (twist) increases, and the projection of slit center moves toward the equator. Channel B slit projection patterns are rotated 180° about nadir. Its scan mirror opening angle decreases as the slit image moves west, its angle with respect to north‐south decreases, and the projection of slit center moves away from the equator. In addition to tilt, the slit projections are also slightly curved as a result of the single‐element telescope imaging performance. The magnitude of the curvature is imperceptible in Figure 5. Each channel is equipped with a shutter located directly in front of the scan mirrors. These are closed between 21:30 and 03:00 SLT to prevent sunlight from entering the instrument. The instrument sunshades enable observations during the remainder of the day. Their performance was characterized during instrument commissioning by rotating the scan mirrors to their maximum viewing angles (±15.5°) while the sun drifted between 45° and 60° of the boresights. This experiment was performed in both the early morning and late afternoon. No solar spectral signature was observed above the nominal detector background. Further, no evidence for a solar contribution to the signal has been evident during observations.

4.2. Thermal Environment The instrument is covered with thermal blankets, and the sunshades are thermally isolated from it. Nonetheless, the incursion of sunlight into the sunshades results in a signiﬁcant diurnal internal temperature variation that can be as large as 20 °C at the solstices. This has a relatively small and easily correctable effect on imaging performance and wavelength registration (see section 4.3). It does, however, complicate detector operation because MCP resistance is a function of temperature and varies between 100 and 150 MΩ during the day. The resistor that controls the QEG voltage is 30 MΩ. Its resistance is approximately 20% that of the MCP stack when the temperature is ~20 °C but increases to approximately 30% when the temperature is 40 °C. This causes a variation in the voltage applied to the MCP stack that changes the

MCCLINTOCK ET AL. 10 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

amplitude of photo‐event electron pulses (commonly referred to as “gain”) at a rate of ~0.04 pico‐coulombs per volt. Changing gain results in small changes in detector sensitivity. This is managed for the GOLD detectors by adjusting the high voltage on the QEG resistor every 30 min during observations in order to maintain the MCP voltage change to <±25 volts throughout the day. This is adequate to reduce gain‐induced sensitivity changes to ~1% during nominal operation. However, gain‐induced sensitivity changes can become significant when the MCP begins to experience gain fatigue through charge depletion (see section 2.3). This primarily affects the sensitivity of the instrument to the 135.56 nm component of the GOLD oxygen doublet, which is ~3.5 times brighter than the 135.85 nm component and ~10 times brighter than individual LBH bands (Figure 6). This is mitigated by using the GYM to reposition the spectrum on the detector. To date, this has occurred once for Channel B, on 14 March 2019. Since that time, the observations with Channel B have been restricted to primarily nighttime viewing. This has reduced additional burn in rate to ~10% per year. The spectrum has been moved three times for Channel A; 26 April 2019, 10 October 2019, and 21 March 2020. Dates for repositioning are determined by monitoring peaks in photon‐event pulse heights throughout the day to determine each detector's gain‐high voltage relationship. The effects of gain sag on GOLD data products and methods used to correct them are described in McClintock et al., 2020. Diurnal temperature variation also give rise to expansion‐contraction and distortion of spectrograph cases leading to displacements of the spectral‐spatial images on the detectors. Stretch and shift in the wavelength scale is removed during Level 1C processing (McClintock et al., 2020) using the positions of the atomic oxygen and nitrogen lines. Displacements of the image parallel to the spectrograph entrance slit have been measured over the course of a day. These can be as large as ±0.17 mm corresponding to ±40 km at nadir. They are not corrected in data released through 22 December 2019. Additionally, thermally induced changes in telescope pointing, which could be on the scale of ± 20 km, have not been characterized.

4.3. First Light Spectra and Background Components A spectral plot from Channel A observations, acquired on 8 September 2018 after initial instrument checkout but before routine observations began and using the HR slit, is shown in Figure 6. It was extracted from the central 400 rows of a spectral‐spatial image made by pointing the spectrograph slit to nadir and integrat- ing for 15 min. The spectrally resolved OI 135.56, 135.85 nm doublet is the dominant emission, which is approximately an order of magnitude brighter than the most prominent LBH bands (e.g., (2,0) near 138.4 nm). In addition, the nitrogen multiplet near 149.3 nm is also present. Spectral resolution and dispersion were determined by fitting both the OI and NI lines using profiles consisting of two Gaussian components each. The results indicate that a Gaussian profile with FWHM ~0.20 nm for the HR slit is an accurate representation of the spectrograph line spread function. Fits to spectra with the LR slit result in profiles with FWHM ~0.35 nm. Both FWHM show 10% diurnal variation as instrument bulk temperature vary throughout the day by ~25 °C while the dispersion changes by ~0.2%. Channel B results are nearly identical. Dashed lines in Figure 6 indicate the backgrounds, which include grating scatter, detector‐reflected light, and particle‐induced dark counts. No attempt is made to separate the individual components during ground processing. The sum can be adequately removed from the data by fitting a piecewise linear curve to regions

of the spectrum where emission from O1 135.6 nm and N2 LBH can be neglected. These are marked by ﬁlled circles in the Figure 6.

4.4. Laboratory and Flight Performance Comparison

A typical GOLD spectrum of LBH obtained using an electron lamp and N2 (red) is compared to a nadir‐viewing flight spectrum obtained during instrument commissioning in Figure 7a. Both spectra were converted from detector count rate per pixel to radiance expressed in Rayleighs per nanometer using the algorithms described by McClintock et al. (2020) and the instrument detection efficiency derived from observations of π4 Ori made between 26 October 2018 and 18 November 2018 (see section 4.7). The laboratory spectrum was then scaled to flight using the total spectral counts excluding values at wavelengths 135–136 nm and 149–150 nm. Radiances in the flight spectrum for 135–136 nm have been multiplied by 0.25 in order to scale them to the plot. This region contains both oxygen 135.56, 135.85 emissions and the 135.4 nm LBH ν′–ν″ (3,0) band. Both spectra contain emission from the nitrogen 149.3 multiplet. Short vertical lines mark the locations of the brightest, well‐isolated band progressions for ν′ =0–5 upper levels.

MCCLINTOCK ET AL. 11 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 7. (a) Comparison of a typical flight spectrum with the O 135.6 nm and 135.4 nm LBH ν′–ν″ (3,0) band emission strengths is reduced by a factor of 4 in order to scale them to the plot, taken at ~15:00 Coordinated Universal Time (UTC) and disk center (black), and an LBH spectrum from an electron lamp operated at 30 eV energy and taken during ground calibration (red). Both spectra were calibrated using instrument detection efficiency derived from observations of π4 Ori made between 26 October 2018 and 18 November 2018 (see section 4.7). Vertical lines mark the positions of isolated, bright LBH bands for ν′ =0–5 and ν″ =0–6. The laboratory spectrum was scaled to flight using the integrated LBH radiance. With this scaling, emissions from bands with ν′ = 0 show evidence for enhancement in the flight spectrum relative to those in the lab. (b) A comparison of the flight spectrum and a model with T = 600 K and using the same scaling also show evidence for increased ν′ = 0 emission.

The general shape of the two spectra is in good agreement suggesting that the laboratory measurements provide a good representation of instrument relative responsivity. However, there are also important differences. First, the long wavelength tails of the LBH emission bands in the ﬂight spectrum are extended toward longer wavelength than those from the laboratory because the column‐density weighted

MCCLINTOCK ET AL. 12 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

temperature of the thermosphere is greater than that of the laboratory vacuum chamber, which was operated near T = 293 K. As a result of the normalization approach, the peaks of the bands in the lab spectrum are expected to exceed those in flight, but the integrated radiances from band to band should be nearly equal for the two (e.g., (4,0) at 132.5 nm). Second, whereas most of the bands with ν′ ≥ 1 appear to have comparable band‐integrated radiances, those with ν′ = 0 appear brighter in the flight spectrum (e.g., (0,2) at 155.5 nm). This suggests that the flight spectrum includes contributions from cascade as well as from direct excitation (Eastes, 2000). Direct excitation is the dominant source of emission from the ~1 mm emitting region of the lamp where cascade effects would be negligible (Ajello et al., 2020). Cascade preferentially populates the lower ν′ = 0,1 states relative to those with higher ν' (Eastes, 2000). To quantify the apparent ν′ = 0 excess in the Figure 8. Comparison of integrated radiances from the relatively well fl fl fi ight data, integrated radiances were computed for the ight and lamp isolated bands identi ed in Figure 7. The black solid curve, which is fi offset by +1.5, displays the ratio of emission from each ν′ level summed spectra using the relatively well isolated bands identi ed in Figure 7a. over all ν″ levels indicated in Figure 7 (e.g., for ν′ = 0 the ratio of sums of For each ν′ level the ratio of flight‐to‐lamp emission was calculated for emission from (0,1), (0,2), and (0,3)) for the flight spectrum and lamp the sum of all available ν″ levels (e.g., for ν′ = 0, the ratio of sums of emis- spectrum, respectively. Vertical lines represent the standard deviation of sion from (0,1), (0,2), and (0,3) are included for both spectra) using inte- the individual ratios. The black dashed line corresponds to ratio = 1. Red, gration intervals that captured the red tails of the flight bands. The blue, and green curves present the ratios computed by comparing the flight spectrum with a model using T = 600 K, the lamp spectrum with a model results are displayed as the top curve (black) in Figure 8, which has been using T = 293 K, and a model using T = 600 K with a model using offset by +1.5 for clarity. Vertical lines represent the standard deviation of T = 293 K. These have been offset by +1.0, +0.5, and 0, respectively. The the individual band ratios included in each ν′ sum. The black dashed line ratios for the flight data indicate excess ν′ = 0 emission not seen in the lamp corresponds to ratio = 1. This plot provides evidence for enhanced emis- or model ratios. sion from the ν′ = 0 level in the flight spectrum and also suggests that the ν′ = 1 level may also have a cascade component. In order to test for bias in determining the ratios of individual ν′–ν″ measurements caused by the differences in temperature‐dependent band shapes, the flight spectrum in Figure 7a was also compared to the model described in section 3.2 using T = 600 K. These spectra, normalized to the total LBH emission, are shown in Figure 7b. The ratios from that comparison, which are plotted as the red curve in Figure 8 and offset by +1.0 also indicate the presence of enhanced emission from the ν′ = 0 level in the flight spectrum. Ratios computed for models with T = 650 and 550 K showed essentially identical results. As further confirmation of cascade in the flight data the band emission ratios were also calculated for the lamp spectrum shown in Figure 4 and the model with T = 293 K and for the model with T = 600 K and the model with T = 293 K. Those results are shown in Figure 8 as blue and green curves, offset by +0.5 and 0, respectively. The last two show no evidence for ν′ = 0 enhancement, indicating that the bias in the band calculations resulting from temperature differences in the flight‐lab comparison is not the cause of the ν′ = 0 enhancement. This additionally supports the conclusion that there is a cascade contribution to the ν′ = 0 level in the flight data.

4.5. Particle Induced Detector Dark Counts Because GOLD is located in geostationary orbit, it is subjected to large fluxes of high energy particles, which not only consist primarily of up to 10 Mev electrons but also include protons. The detectors are shielded by the minimum 6 mm‐thick aluminum walls of the instrument case and by 2 mm‐thick tantalum plates that surround the sides and backs of their housings. While these structures efficiently block the primary electrons, some of the bremsstrahlung X‐rays produced by their deceleration pass through the tantalum or enter the fronts of the detectors causing background counts as they interact with the MCPs. Images of these are measured at the top of every limb scan and the rates per square cm averaged over each 8.64 cm2 detector are shown in Figure 9 for the first 400 days of the mission. The background counts vary by a factor of ~200 over the first year of the mission and can also vary significantly on hourly time scales.

MCCLINTOCK ET AL. 13 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 9. Background rates for Channel A (black) and Channel B (blue) resulting from energetic particles for the ﬁrst 400 days of the mission.

Background frequency distributions, shown in Figure 10a, indicate that Channel B rates are signiﬁcantly higher in the 25–75 Hz range than those of Channel A. Figure 10b shows the cumulative distributions, which have rates of 68 and 105 counts/cm2/s for Channels A and B, respectively. During instrument design, the Geant4 transfer code (Agostinelli et al., 2003) and an electron energy spectrum from the AE8 radiation model (Vette, 1991) were used to design the detector shields and estimate backgrounds. Those calculations, which included only shielding by instrument structures, predicted median rates of 55 counts/cm2/s and cumulative rates <300 counts/cm2/s at the 95th percentile. On orbit median rates are larger by a factor of 1.2 and 1.9 for Channels A and B, respectively. The difference between the two channels may arise from the different view factors between each detector and spacecraft structures that provide additional shielding. Figure 11 compares the pulse height distributions (PHDs), normalized to unity, that arise from photon events with a modal gain of 1.5 picocoulombs and particle backgrounds. These are plotted as a function of pulse height, measured in picocoulombs, where the conversion from pulse height DN to charge (Q), measured in picocoulombs, is

Q ¼ 0:049 þ 0:0215·DN: (8)

Approximately 8% of background events produce extremely high gain (DN ≥ 255), which are most likely

Figure 10. (a) Frequency histograms of particle background rates for the two channels. (b) Comparison of observed integral distributions with a pre‐launch predicted distribution calculated using Geant4 and the AE8 radiation model (red line).

MCCLINTOCK ET AL. 14 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

associated with charged particles interacting tangentially in the MCPs. These events overwhelm the XDL readout electronics in the Y channel, imposing bright, isolated horizontal features on the background images as illustrated in Figure 12a for Channel B backgrounds, which were taken with the scan mirror stowed so that no light entered the spectrograph. Most of these are identified and removed during L1A‐L1B data processing (McClintock et al., 2020), as illustrated in Figure 12b where the bright horizontal strips and additional arced structures are absent after filtering the data to exclude pulse height values 0, 1, and >200. The dark vertical bands in both images are the result of detector “burn in” of atmospheric emissions (e.g., Figure 6). These are corrected for both channels during L1B–L1C data processing using images of an onboard “flat field” lamp as described by McClintock et al., 2020.

4.6. Flat Field Corrections fl fi Figure 11. Pulse height distributions for photon events (black) with a Figure 13 shows examples of at eld correction images used to remove modal gain = 1.5 picocoulomb (~68 DN) and particle backgrounds detector “burn in” artifacts during L1B–L1C data processing (blue). Eight percent of the particle events have pulse heights with (McClintock et al., 2020). Flat field images are acquired each day during DN ≥ 255 (5.53 picocoulombs). a20‐min period before nominal disk scans begin. These are averaged over a7‐day window to increase the signal‐to‐noise ratio. Correction images are then constructed from the ratio of a current 7‐day average to that of the 7‐day image acquired just after commissioning. Bright vertical lines on the left of each image result from oxygen 135.56 nm emission, which requires significant correction factor as time progresses. Once the factor reaches ~2, the grating yaw mechanism (see section 2.2) is actuated to move it to the right, toward longer wavelengths. Channel B exhibited substantial initial burn in and its GYM was actuated once on 14 March 2019. After that time, Channel B has been used primarily for early morning disk scans and night disk scans when emissions are weak and for special brief events (e.g., the 2 July 2019 solar eclipse). Channel A continues full disk, limb, and occultation viewing. Its GYM has been actuated three times (27 April 2019, 10 October 2019 and 21 March 2020). This has produced the double image seen in Figure 12a. Whereas corrections for 135.6 nm are significant, those for LBH emissions are modest and are typically less than 10%.

Figure 12. (a) Horizontal features in the Channel B particle background image result from extremely energetic events. These are largely identified and removed by excluding detector events with pulse height measurement values of 0, 1, and >200 resulting in the image shown in (b). Dark vertical bands in the images are depressions in the detector flat field that arise from “burn in” on the detector by atmospheric emissions (McClintock et al., 2020). Channel A background images exhibit nearly identical behavior events with pulse heights = 255.

MCCLINTOCK ET AL. 15 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Figure 13. Detector ﬂat ﬁeld correction images for 15 August 2019. These were derived from an average of seven daily observations of lamp images made before nominal disk scans begin.

4.7. Flight Radiometric Performance Measurements of channel detection efﬁciency, which were made by allowing the star π4 Ori (HD 30836) to drift across the spectrographs low‐resolution entrance slits during instrument commissioning (section 3.1), indicated that the values determined during ground calibration should be multiplied by a factor of 0.7 in order to agree with published values from IUE. Repeated observations of above‐the‐atmosphere measurements of multiple stars taken during routine occultations when the star is present unocculted by the atmosphere for more than 30 s indicate that more appropriate multipliers are 0.78 and 0.85 for Channel A and Channel B, respectively. The black curve (Channel A) and blue curve (Channel B) in Figure 14 show π4

Figure 14. Black (Channel A) and blue (Channel B) curves compare the irradiance from HD 30836 (π4 Orionis) derived from GOLD above‐the‐atmosphere occultation observations to the irradiance reported by IUE (red curve) using the detection efﬁciencies determined during ground calibration (Figure 3) multiplied by 0.78 and 0.85 for Channel A and Channel B, respectively.

MCCLINTOCK ET AL. 16 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Ori irradiance derived from GOLD observations made between 26 October 2018 and 18 November 2018 using the ground‐measured detection efficiency values (Figure 3) multiplied by these updated factors. These observations were taken during routine stellar occultations when the star tangent altitude was >250 km and have spectral sampling of 0.12 nm. The red curve shows the IUE results which have slightly lower spectral sampling and resolution. Comparisons of three additional stars, π5 Ori, η Ori and υ Ori give nearly identical results. Repeated observations of π4 Ori made with Channel A between October 2018 and March 2020 show no evidence for changes in its radiometric sensitivity during that time. Routine occultations measurements with Channel B have been suspended since December 2018. Data released through 22 December 2019 have been calibrated using the detection efficiency measured during commissioning; thus, radiance and irradiance values reported in publicly released data underestimate the atmospheric radiances by ~10% for Channel A and ~20% for Channel B. These errors will be retroactively corrected in future releases. 4.8. Data Products and Science Data Processing The algorithms used to convert detector counts to data expressed in geophysical units, which are defined as the GOLD Mission Level 1C data products (radiance for emissions and irradiance for occultations), and the data processing approach used implement them are described in a companion paper (McClintock et al., 2020). That paper describes the status of the L1C data and discusses the artifacts present in the current implementation. These are also described in data release notes available on the GOLD Science Data web site (http://gold.cs.ucf.edu/search/). McClintock et al. (2020) also describe ongoing work directed at mitigating those artifacts.

5. Summary The GOLD mission is framed by a well‐deﬁned set of measurement objectives that include imaging oxygen

135.6 nm and N2 LBH emissions with a spatial resolution of 250 × 250 km (at nadir) and a spectral resolution of 0.2 nm, measuring N2 LBH emission and observeing stellar occulations on the limb during the day, and measuring 135.6 nm radiances emitted by O+ as it recombines in the ionosphere during the night. These are met by an instrument consisting of a pair of telescope‐imaging spectrograph channels equipped with scan mirrors. Here, we describe the GOLD instrument including its optical system and detector. Its performance was characterized in the lab before launch. We compare measurements of laboratory sources made then to observations of the thermosphere after launch and ﬁnd good agreement. Direct comparison of LBH spectra of an electron lamp taken before launch with spectra on orbit provide evidence that both cascade and direct excitation are important sources of thermospheric LBH emission.

Acknowledgments References This research was supported by NASA — Contract 80GSFC18C0061 to the Agostinelli, S., Allison, J., Amako, K., Apostolakis, J., Araujo, H., Arce, P., et al. (2003). Geant4 A simulation toolkit. Nuclear Instruments – University of Colorado, Boulder. The and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 506(3), 250 303. https://doi. ‐ ‐ flight data used here are available at the org/10.1016/S0168 9002(03)01368 8 GOLD Science Data Center (http:// Ajello, J. M., Evans, J. S., Veibell, V., Malone, C. P., Holsclaw, G. M., Hoskins, A. C., et al. (2020). The UV spectrum of the ‐ ‐ fi gold.cs.ucf.edu/search/) and at NASA's Lyman Birge Hop eld band system of N2 induced by cascading from electron impact. Journal of Geophysical Research: Space Physics, Space Physics Data Facility (https:// 125, e2019JA027546. https://doi.org/10.1029/2019JA027546 spdf.gsfc.nasa.gov). We thank the Ajello, J. M., Shemansky, D. E., Franklin, B., Watkins, J., Srivastava, S., Simms, W. T., et al. (1988). A simple ultraviolet calibration source – reviewers who provided detailed with reference spectra and application to the Galileo ultraviolet spectrometer. Applied Optics, 27(5), 890 914. https://doi.org/10.1364/ comments and suggestions that greatly AO.27.000890 improved the accuracy and quality of Boggess, A., Carr, F. A., Evans, D. C., Fischel, D., Freeman, H. R., Fuechsel, C. F., et al. (1978). The IUE spacecraft and instrumentation. – this work. Nature, 275(5679), 372 377. https://doi.org/10.1038/275372a0 Bohlin, R. C. (1996). Spectrophotometric standards from the far‐uv to the near‐ir on the white dwarf flux scale. The Astrophysical Journal, 111, 1743–1747. https://doi.org/10.1086/117914 Bohlin, R. C., & Bianchi, L. (2018). A correction for IUE UV flus distributions from comparisons with CALSPEC. Astronomical Journal, 155(4), 162–170. https://doi.org/10.3847/1538‐3881/aaaec1 Canfield, R. L., Johnston, R. G., & Madden, R. P. (1973). NBS detector standards for the far ultraviolet. Applied Optics, 12(7), 1611–1617. https://doi.org/10.1364/AO.12.001611 Eastes, R. W. (2000). Modeling the N2 Lyman‐Birge‐Hopfield bands in the dayglow: Including radiative and collisional cascading between the singlet states. Journal of Geophysical Research, 105(A8), 18,557–18,573. https://doi.org/10.1029/1999JA000378 Eastes, R. W., McClintock, W. E., Burns, A. G., Anderson, D. N., Andersson, L., Codrescu, M., et al. (2017). The global‐scale observations of the limb and disk (GOLD) mission. Space Science Reviews, 212(1‐2), 383–408. https://doi.org/10.1007/s11214‐017‐0392‐2 Hunten, D. M., Roach, F. E., & Chamberlain, J. W. (1956). A photometric unit for the airglow and aurora. Journal of Atmospheric and Terrestrial Physics, 8(6), 345–346. https://doi.org/10.1016/0021‐9169(56)90111‐8

MCCLINTOCK ET AL. 17 of 18 Journal of Geophysical Research: Space Physics 10.1029/2020JA027797

Krywonos, A., Harvey, J. E., Daniell, R., Parent, N., & Eastes, R. (2006). Scanless ultraviolet remote sensor for limb profile measurements from low earth orbit. Optical Engineering, 45(10). https://doi.org/10.1117/1.2360200 McClintock, W. E., Eastes, R. W., Beland, S., Bryant, K., Burns, A. G., Correira, J., et al. (2020). Global‐scale measurements of the limb and disk (GOLD) mission implementation: 2. observations, data pipeline and level 1 data products. Journal of Geophysical Research:Space Physics, 125,e2020JA027809 https://doi.org/10.1029/2020JA027809 McClintock, W. E., Schneider, N. M., Holsclaw, G. M., Clarke, J. T., Hoskins, A. C., Stewart, I., et al. (2015). The imaging ultraviolet spectrograph (IUVS) for the MAVEN mission. Space Science Reviews, 195(1‐4), 75–124. https://doi.org/10.1007/s11214‐014‐0098‐7 Meier, R. R., Picone, J. M., Drob, D., Bishop, J., Emmert, J. T., Lean, J. L., et al. (2015). Remote sensing of Earth's limb by TIMED/GUVI: Retrieval of thermospheric composition and temperature. Earth and Space Science, 2,1–37. https://doi.org/10.1002/2014EA000035 Quijada, M. A., Hoyo, J., & Rice, S. (2014). Enhanced far‐ultraviolet reflectance of MgF2 and LiF over‐coated Al mirrors. Proceedings of SPIE, 9144, 91444G. https://doi.org/10.1117/12.2057438 Sahnow, D. J., Oliveira, C., Aloisi, A., Hodge, P. E., Massa, D., Osten, R., Proffitt, C., et al. (2011). Gain sag in the FUV detector of the cosmic origins spectrograph. Proceedings of SPIE, 8145, 81450Q. https://doi.org/10.1117/12.893989 Siegmund, O. H., Gummin, M. A., Stock, J. M., Marsh, D. R., Raffanti, R., & Hull, J. S. (1993). High‐resolution monolithic delay‐line readout techniques for two‐dimensional microchannel plate detectors. Proceedings of SPIE, 2006. https://doi.org/10.1117/12.162851 Siegmund, O. H. W., McPhate, J., Curtis, T., Jelinsky, S., Vallerga, J. V., Hull, J., & Tedesco, J. (2016). Ultraviolet imaging detectors for the GOLD mission. Proc. SPIE, 9905, 99050D1‐D10. https://doi.org/10.1117/12.2232219 Vette, J. I. (1991). The AE‐8 Trapped Electron Model Environment. NASA‐TM‐107820, NAS 1.15:107820, NSSDC/WDC‐A‐RS‐91‐24. Greenbelt, MD. Wadsworth, F. L. O. (1896). The modern spectroscope XV. On the use and mounting of the concave grating as an analyzing or direct comparison spectroscope. Astrophysical Journal, 3,47–62. https://doi.org/10.1086/140176 Wilkinson, E., Béland, S., Penton, S. V., Vallerga, J. V., McPhate, J. B., & Sahnow, D. J. (2003). Algorithms for correcting geometric dis- tortions in delay‐line anodes. Review of Scientific Instruments, 74(1), 38–46. https://doi.org/10.1063/1.1524715

MCCLINTOCK ET AL. 18 of 18