The Improved Limb Atmospheric Spectrometer (ILAS) was a satellite-based solar occultation sensor that was developed by the Environment Agency of Japan (EA) to monitor and study the stratospheric ozone layer. This paper describes the characteristics of the ILAS instrument and its performance in orbit. ILAS measured the vertical distribution of ozone, nitric acid, nitrogen dioxide, nitrous oxide, methane, water vapor, temperature, pressure, and aerosol extinction coefficients at 1.6-km vertical resolution. ILAS was equipped with two spectrometers: an infrared (IR) spectrometer with an uncooled pyroelectric linear array detector to sense between 6.21 and 11.76 μm and a visible spectrometer to monitor 753–784 nm. In addition, a Sun-edge sensor (SES) assigned the tangent height of the instantaneous field-of-view (IFOV). A two-axis gimbals control system on ILAS used two Sun position sensors to track the center of brightness of the Sun during occultation measurements. Before launch onboard the Advanced Earth Observing Satellite (ADEOS), the performance of ILAS was checked on the ground using several methods, including gas-cell measurements, time response measurements, Sun-tracking tests, and hollow-cathode lamp measurements. After the launch of ADEOS on 17 August 1996, ILAS functioned successfully for 8 months of routine operation, from 30 October 1996 to 30 June 1997, collecting more than 6700 solar occultation measurements, after which time the satellite failed due to a failure in a solar paddle. The time delay response of the IR channel was characterized using stepwise IR input. Instrument functions of the ILAS IR and visible spectrometers were determined by combining theoretical optical calculations, experimental measurements using a gas-cell before launch, and in-orbit data. The signal-to-noise ratio (SNR) of each element in the IR channel was estimated to be 400–1200. In the visible channel, it was 1600–1800 for a 100% direct Sun signal. At sunset occultation, ILAS was able to track the Sun below a tangent height of 10 km in some cases. The method of determining the solar edges from the SES data worked correctly, giving adequate tangent height information for observations. Output signal levels of the SES, visible channel, and IR channel showed slight degradation during the period that ILAS was operational, which is attributed to space-borne contaminants. However, changes in absolute signal levels do not affect data retrieval, because the solar occultation technique was self-calibrating. Overall, ILAS worked as designed during its operation in orbit and gathered valuable data for ozone layer studies.
 The Improved Limb Atmospheric Spectrometer (ILAS) was a solar occultation satellite sensor that was developed by the Environment Agency of Japan (EA) to monitor the stratospheric ozone layer [Sasano et al., 1999a, 1999b]. ILAS was onboard the Advanced Earth Observing Satellite (ADEOS), which was launched on 17 August 1996, from the Tanegashima Space Center of the National Space Development Agency of Japan (NASDA).
 ADEOS, the first genuine Earth observing satellite developed in Japan, carried eight instruments to observe the Earth's surface and atmosphere. The subrecurrent orbit of ADEOS was Sun synchronous at approximately 800-km altitude, 98.6 degrees inclination, 101 min period, and 1040 local mean Sun time at the descending node. The recurrent period was 41 days [Shimoda, 1999]. ILAS made two solar occultation measurements (one sunrise, one sunset) during an orbit. Due to the measurement principle and the orbit of the satellite, the latitudinal coverage of ILAS measurements was from 57 to 72 degrees in the Northern Hemisphere and from 64 to 89 degrees in the Southern Hemisphere. The latitudinal coverage was affected by seasonal changes in the configuration between the orbital plane of ADEOS relative to the rotation axis of the Earth [Sasano et al., 1999a]. Ozone-related chemical species at high latitudes can be measured continuously throughout the year. ILAS development began in 1989, with an Engineering Model (EM) and Proto-Flight Model (PFM) developed through March 1992 and September 1994, respectively [Suzuki et al., 1995]. The PFM was launched into orbit.
 ILAS had two grating spectrometers. One was an infrared (IR) spectrometer with an uncooled pyroelectric linear array detector to monitor 6.21–11.76 μm; the other was a visible spectrometer covering 753–784 nm. Using the IR spectrometer, ILAS measured vertical profiles of ozone, nitric acid, nitrogen dioxide, nitrous oxide, methane, and water vapor [Yokota et al., 2002]. With the visible spectrometer, ILAS measured vertical profiles of temperature, pressure [Sugita et al., 2001], and the aerosol extinction coefficient at 780 nm [Hayashida et al., 2000]. The entrance slit of the spectrometer, when projected out to the atmosphere at the tangent height for the occultation geometry, provided an instantaneous field-of-view (IFOV) of a 1.6-km-high window. The effective vertical resolution, accounting for the size of the IFOV and the effects of atmospheric refraction, varied from 1.9 to 3.5 km at a tangent height range between 15 and 55 km [Yokota et al., 2002]. The characteristics of the ILAS instrument compared with other solar occultation satellite instruments (ATMOS, HALOE, SAGE II, and POAM II) are described in Table 1. A low spectral resolution IR grating spectrometer was chosen because a previous feasibility study showed that it gave sufficiently accurate mixing ratios for several gas species related to ozone chemistry.
Table 1. Comparison of the Characteristics of Solar Occultation Satellite Instruments
Fourier transform spectrometer
Grating/filter Sun photometer
Filter Sun photometer
O3, HNO3, NO2, N2O, CH4, H2O, aerosols
O3, HCl, HF, CH4, H2O, NO, NO2, aerosols
Aerosols, O3, H2O, NO2
Aerosols, O3, H2O, NO2
IR-ch.: 6.21–11.76 μm (850–1610 cm−1)
Fil. 1 (625–1100 cm−1)
O3: 9.852 μm
Ch. 1: 1.020 μm
Ch. 1: 352.3 nm
Vis-ch.: 753–784 nm
Fil. 2 (1100–2000 cm−1)
HCl: 3.401 μm
Ch. 2: 0.940 μm
Ch. 2: 441.6 nm
Fil. 3 (1550–3700 cm−1)
HF: 2.452 μm
Ch. 3: 0.600 μm
Ch. 3: 448.1 nm
Fil. 4 (3750–5100 cm−1)
CH4: 3.459 μm
Ch. 4: 0.525 μm
Ch. 4: 601.4 nm
Fil. 9 (625–2450 cm−1)
H2O: 6.605 μm
Ch. 5: 0.453 μm
Ch. 5: 761.2 nm
Fil. 12 (625–1450 cm−1)
NO: 5.263 μm
Ch. 6: 0.448 μm
Ch. 6: 781.0 nm
NO2: 6.254 μm
Ch. 7: 0.385 μm
Ch. 7: 921.0 nm
CO2: 2.799 μm
Ch. 8: 936.4 nm
Ch. 9: 1060.3 nm
Channel (pixel; filter) number
IR-ch.: 44 pixels
Vis-ch.: 1024 pixels
IR-ch.: 0.129 μm
0.01 cm−1 (unapodized)
Depends on channels
Vis-ch.: 0.03 nm
Coarse Sun position sensor (CSPS) + Fine Sun position sensor (FSPS)
 After the launch and initial checkout period of ADEOS, ILAS started normal operation on 1 November 1996. For 8 months, ILAS functioned successfully and gathered data for approximately 6700 occultation measurements, before ADEOS stopped working on 30 June 1997 [Sasano et al., 1999a].
 This paper describes the characteristics and performance of the ILAS instrument, including the optical, thermal, and mechanical characteristics of ILAS in orbit as revealed by detailed postlaunch analysis.
2. The ILAS Instrument
 ILAS consisted of a Sun-tracking mechanism, telescope and relay optics, IR and visible spectrometers, IR and visible detectors, a Sun-edge sensor (SES), a signal processor, an electric power supply, and the structure body. Figure 1 shows the layout of each component of the ILAS instrument, and Table 2 summarizes the characteristics of the ILAS instrument [Araki et al., 1993; Suzuki et al., 1995].
 ILAS had two grating spectrometers, one for the IR region to measure O3 and other chemical species, and a second for the near visible region to measure pressure and temperature (using the molecular oxygen A absorption band). Hereafter, we call the latter the visible spectrometer. A similar spectrometer was on the LAS (Limb Atmospheric Infrared Spectrometer) onboard the EXOS-C satellite [Matsuzaki et al., 1985], and a similar visible spectrometer was used on a rocket experiment carried out at the Institute for Space and Astronautical Science (ISAS) in the 1980s [Matsuzaki et al., 1984]. ILAS also had a SES for measuring the IFOV direction. Figure 2 shows a block diagram of ILAS.
2.1. IR Spectrometer
 Before entering the ILAS IR spectrometer, sunlight was split by a dichroic mirror and focused on the entrance slit using a single Ge relay lens optics (F/4.0, F = 480 mm). The IFOV of the IR spectrometer was designed to be 13′50″ wide and 2′08″ high, which corresponds to 13.2 and 2.0 km, respectively, at a tangent height of 20 km seen from ADEOS at an orbital altitude of 800 km. However, the center of the entrance slit of the IR spectrometer had a small (28″) offset compared to the telescope exit slit, as shown in Figure 3. This offset resulted from adjustments to the optics after the prelaunch environmental test. As a result, the actual IFOV height was 1′40″, which corresponds to a 1.6-km tangent height, which is the vertical resolution of the ILAS IR spectrometer.
 An optical chopper was mounted in front of the entrance slit. The IR spectrometer used a Czerny–Turner type grating spectrometer design (F/4.0, F = 100 mm) with a plane grating of 30 grooves/mm. ILAS used a 44-element PbTiO3 pyroelectric array detector (developed by Matsushita Electric Industrial Co., Ltd.) [Ishigaki, 1997] for IR channel signal detection between 6.21 and 11.76 μm, with 0.129-μm spectral resolution. An advantage of this pyroelectric array detector was that it did not require cooling devices. Furthermore, it had a sufficient signal-to-noise ratio (SNR) (noise equivalent power of 0.7 nW/Hz1/2) for solar occultation measurements in the IR region. We define the vertical direction of the array detector as the dispersion direction of the grating. The vertical element size of the IR detector (0.38 mm high and 0.02 mm gap) was nearly equal to that of the slit image (0.4 mm high) and the vertical IFOV. In addition, the horizontal element size (2.0 mm wide) corresponded to the horizontal slit image and the horizontal IFOV. There was some cross-talk between adjacent elements, which was caused by optical, thermal and electrical interference. The cross-talk between each element was estimated experimentally to be about 2.25% for each element [Yokota et al., 2002].
2.2. Visible Spectrometer
 Visible sunlight was reflected (F/4.0) from the beam splitter and focused (F/8.0 perpendicular, F/16 spectral direction) on the entrance slit of the visible spectrometer using four-lens relay optics. The visible spectrometer used only part of the 12-cm telescope (3 × 1.5 cm in effective area), because there was too much solar energy in this spectral region. The IFOV of the visible spectrometer was designed to be 2′04″ wide and 2′08″ high, both of which correspond to 2.0 km at a tangent height of 20 km. The center of the entrance slit of the visible spectrometer had an offset of 28″ compared to the telescope exit slit, as shown in Figure 3. As a result, the actual IFOV height was 1′40″, which corresponds to 1.6 km in tangent height. This number gives the vertical resolution of the ILAS visible spectrometer. Also note that the IFOV of the visible and IR spectrometers had a 28″ offset, which corresponds to about 480 m in tangent height. This difference was corrected during the course of data retrieval [Yokota et al., 2002].
 The visible spectrometer used a concave holographic grating (F/8.0, F = 400 mm, 1800 grooves/mm) with a 1024-element metal oxide semiconductor (MOS) photodiode array detector to cover the 753–784 nm region with a 0.15 nm full-width at half-maximum (FWHM) resolution. Due to this high spectral resolution, the visible spectrometer could be accurately calibrated for wavelength using observational data of the solar Fraunhofer lines. The detailed procedure for calibrating the wavelength is described in section 3.3.
2.3. Sun-Edge Sensor
 Accurate information on the absolute direction of the IFOV during Sun tracking is very important for solar occultation measurements. Therefore, ILAS was equipped with a SES with 747 ± 12.5 nm optical filter to determine the angular distance of the IFOV from the top edge of the Sun. The SES used a 1024-element MOS photodiode array with a telescope (f = 630 mm) with 8″ resolution, which corresponds to about 130 m in tangent height. The tangent height for ILAS measurements has an accuracy of about 150 m in absolute altitude when the refraction effect is considered correctly. The tangent height accuracy computations consider the accuracy of the satellite clock (12 ms of 1σ uncertainty) and orbit determination (150 m of 3σ in XYZ direction). A detailed description of the tangent height determination is given elsewhere [Nakajima et al., 2002].
2.4. Sun Tracker
 ILAS tracked the center of brightness of the Sun. Sun tracking used a digital feedback system with an 80-ms feedback period, using a 2 × 2 quad Si-photo detector (fine Sun position sensor, FSPS) that had an IFOV of about ±0.5 degrees. The FSPS used a four-lens telescope with a 25.0 mm diameter and 72.0 mm focal length. The actual tracking ability of ILAS was better than 24″ peak-to-peak in line-of-sight for the azimuth and elevation directions. A coarse Sun position sensor (CSPS), which used a two-dimensional position-sensing detector with an IFOV of about ±5 degrees, was used to lock onto the Sun before tracking by the FSPS. The CSPS used a refractive telescope with an 8.5 mm diameter and a 21.7 mm focal length. The spectral response of both FSPS and CSPS had a peak at 750 nm with 140 nm FWHM. Initial azimuth and elevation positions were digitally relayed from the ground.
2.5. ILAS Observation Sequence
 The sampling frequencies of the SES and the IR and visible spectrometers were all 12 Hz. The speed of the change in viewing tangent height due to satellite motion was about 2.1 km/s at a tangent height of 20 km, (about 3.1 km/s in outer space). Thus, the sampling interval of these sensors corresponded to about 0.18 km at a tangent height of 20 km, or about 0.25 km in outer space.
Figure 4 shows the observation sequence of ILAS operation modes. ILAS was put into standby mode (2) 100 min after the launch of ADEOS. In this standby mode, only heater power was needed for ILAS survival. For the sunrise occultation measurement, ILAS first pointed at a predetermined position inside the instrument box for checkout (3), then pointed toward outer space for 0% emission calibration (4), then pointed at the expected sunrise position calculated from orbital ephemeris of ADEOS and searched for the Sun with a Sun tracker. After it captured the Sun and started tracking, ILAS made a sunrise observation (5). When the Sun moved out of the Earth's atmosphere, ILAS made a 100% solar emission calibration (6). Then, it again pointed toward outer space for 0% emission calibration (4), pointed at a predetermined position inside the instrument box (3), and then was put into standby mode (2).
 At the time of the sunset occultation measurement, ILAS performed steps (3) and (4), then pointed toward the Sun to make a 100% solar emission calibration (7). Then, when the Sun moved into the Earth's atmosphere, it made a sunset observation (8), and then resumed steps (4), (3), and (2).
 When a light load mode (LLM) command was issued from the ADEOS bus, ILAS was automatically put into standby mode (2) from any of modes (3), (4), (5), (6), (7), and (8). In an emergency, when the satellite could not supply even heater power to ILAS, ILAS was put into All-OFF mode (1).
2.6. ILAS Data Retrieval Procedure
 ILAS used an onion-peeling retrieval method, in which the entire spectrum was least squares fitted in order to derive simultaneously the volume mixing ratios of the gas species for each layer. In the radiative transfer calculations used to obtain theoretical ILAS signals, a look-up table was used for rapid calculation of IR cross sections as a function of air pressure and temperature. The look-up table was calculated in advance using a line-by-line method with the HIRAN 96 line parameter database [Rothman et al., 1998]. Errors (internal error) in the retrieved gas profiles were estimated from residuals of the fitted spectrum and the Jacobi values for each gas molecule. Possible errors (external error) caused by other factors, such as temperature uncertainties and nongaseous correction errors, were modeled and evaluated in advance. The error bar, which was provided as an ILAS product, is the root-sum-square of these two components. No constraint was applied in the retrieval procedure, and the effects of initial gas profiles on the retrieved results were negligibly small in general. Details of ILAS retrieval and error estimation procedures are given by Yokota et al. .
3. Performance of Each Component
 The characteristics and performance of each component of ILAS hardware were determined by prelaunch ground tests and by postlaunch analysis of in-orbit ILAS mission data. Ground tests and experiments were executed, not only to confirm that the instruments worked properly, but also to derive the instrument parameters that are indispensable for actual operational data processing of ILAS data. The ground tests consisted of the following:
Measurements of output signals of the IR spectrometer, visible spectrometer, and SES by tracking the Sun as a light source.
Measurements of several minor gases using a gas-cell and a standard light source to determine the instrument functions of the IR spectrometer [Suzuki et al., 1995; Suzuki, 1996].
Measurement of the stepwise input response of the IR channel to determine the time delay response of the IR signal processing circuit.
Confirmation of solar tracking using a pseudosolar source.
 Several other ground tests were used to confirm the performance of each component of ILAS, which worked properly during these tests. Ground tests (2) and (3) are now described in detail. The results of these tests were used to determine important instrument parameters for the operational data retrievals of ILAS data.
3.1. Time Delay Response of the IR Signal Processing System
 The time delay response of the output signal of the ILAS/PFM IR channel to a stepwise IR input achieved by opening a shutter against a 1673 K blackbody source was measured for each of the 44 elements. The time delay response of the IR signal processing system of ILAS/PFM was approximated by a third-order delay system model as:
T1, T2, and T3 were determined for each element. The average values of T1, T2, and T3 for each element found by experiment were 189 ± 20, 110 ± 17, and 93 ± 12 ms with 1σ uncertainty, respectively.
 To determine the delay time t0 for all elements, we used fluctuations in Sun tracking of in-orbit data. When there are fluctuations in Sun tracking during the solar occultation, both the intensity of the IR signal and the position of the upper and lower edges of the Sun should fluctuate in phase. Therefore, when we calculated a autocorrelation function between IR signal intensity and the position of the Sun edge as a function of time difference, we could determine the t0 value from the time difference that maximized the autocorrelation function. The determined t0 value was 65 ± 22 ms with 1σ uncertainty.
3.2. Gas-Cell Measurements
 The nominal instrument function of each detector element of the ILAS IR spectrometer was determined with a theoretical optical calculation. This nominal instrument function had to be experimentally calibrated. First, the absolute wavelength of each element of ILAS was measured using a monochrometer and a gas-cell. From this method, the center wavelength of element 43, an element dedicated to wavelength calibration, was 11.7634 μm, within 0.01% of the designed value (11.7647 μm). In addition, the dispersion of each element was 0.12928 μm, within 0.1% of the designed value (0.12915 μm).
 Second, the absorption spectra of several sample gases measured by the ILAS IR spectrometer were calibrated using a commercial Fourier transform IR spectrometer (Bomem Inc., MB-100, 1 cm−1 resolution) [Suzuki et al., 1995; Suzuki, 1996]. The gas-cell measurements were made with a 1673 K blackbody source. Measured sample gases were O3, CH4, N2O, NO2, SO2, C2H6, CFC-11, CFC-12, HCFC-21, HCFC-22, and CCl3CCF3. Sample gases were diluted with N2 gas to achieve 0.1–1000 torr sample pressure. The output transmittance of ILAS and that of FTIR were compared to calibrate the ILAS IR instrument function. The alignment error of the ILAS IR spectrometer was negligibly small, so that the designed wavelength value for each element of the ILAS IR spectrometer could be used.
 As an example, Figure 5 compares the absorption spectrum of 0.125 torr HCFC-22 in a 100.39 torr total pressure gas-cell with an optical path length of 1 m, measured by ILAS (diamonds), with that measured by FTIR and then convolved with the instrument function of the ILAS IR spectrometer. These calculations also correct for the aberration of the grating spectrometer. The different symbols in the figure for the FTIR spectra represent calculations using different cross-talk values (triangles: 0%, stars: 2.25%) between neighboring elements. The convolution result of the FTIR spectrum with 2.25% cross-talk gave minimal residual compared with the ILAS spectrum, as shown in Figure 5. This instrument function was used for the actual data processing of the ILAS IR spectrometer data for the Version 5.20 data product.
3.3. Determination of ILAS Visible Channel Instrument Function
 ILAS visible channel instrument function was determined primarily by theoretical optical calculations for the slit image on the detector plane of the visible spectrometer. In addition, we used the line spectrum of a K hollow-cathode lamp to correct the absolute wavelength of the instrument function. We found that the FWHM of the instrument function of the ILAS visible spectrometer was wider than designed, and it could not distinguish two absorption lines of the P-branch of the O2A band that differed in spin quantum energy.
 As is typical for a high spectral dispersion grating spectrometer, environmental temperature affected the wavelength position of the ILAS visible spectrometer instrument function. The environmental temperature of the ILAS visible spectrometer changed by about 4°C over half a year, due to the change in the distance between the satellite and the Sun, as shown in Figure 6. This figure shows the long-term change in the temperature of the optical bench of ILAS (hereafter, denoted as “ILAS temperature”) after launch. After the initial checkout period of ADEOS, the power for all the sensors was turned on, resulting in a 7°C increase in ILAS temperature, on 30 October 1996. After that, ILAS temperature showed a gradual sinusoidal pattern, with an amplitude of about 4°C, peaking at the end of January. This change was probably due to the change in the distance between the Earth and the Sun. This temperature change resulted in about a 0.5-element change of the wavelength (0.5 × 0.03 nm = 0.015 nm) at the focal plane of the visible spectrometer. Thermal expansion of the optical base caused a change in the location of the focal plane, due to the obliqueness of the incident beam of the spectrometer.
 To reduce the effect of this error, we determined the wavelength axis of the instrument function of the visible spectrometer each week for both sunrise and sunset observations. For this purpose, Fraunhofer lines seen in the 100% signal of the Sun were used to minimize the root-sum-square differences between element numbers of calculated and observed spectra at 13 selected wave number regions that included strong, distinct solar lines outside the O2A band, using the downhill simplex method. At that time, the instrument function of the ILAS visible spectrometer was convolved with a high-resolution solar spectrum measured by ground-based FTIR at Kitt Peak [Wallance et al., 1993]. The error in the wavelength center was estimated to be ±0.012 nm (0.4 pixel) from their maximum residuals.
 The width of the visible instrument function was also evaluated by comparing the transmittance spectrum of the atmosphere with the simulated one, to minimize the residual. In total, 72 instrument functions (36 weeks × 2 occultation modes) were assigned for each 1024 pixels, and used for events that occurred within the respective weeks.
3.4. Output Signal of the IR Channel
Figure 7a shows raw output signals of the ILAS IR channel at several tangent heights. All 44 elements were divided into 6 regions with different amplification factors of the following amplifier as shown in the figure, to take the difference of solar irradiance in those wavelength regions into account. Figure 7b shows an example of the transmittance spectrum of the ILAS IR channel at a 30-km tangent height. The spectrum was derived from the raw output signal through the atmosphere divided by a direct Sun (100%) signal. A strong ozone absorption band around element 27 (9.6 μm) is evident. The spectrum fitted with the line-by-line nonlinear least squares fitting method [Yokota et al., 2002] is also shown in this figure. The residual of the fit shown in Figure 7c is mostly small (less than 1%), except for the central part of the ozone 9.6-μm absorption band, which may be caused by slight deficiencies in instrument function of each element (difference in wavelength) around this band.
Figure 8 shows examples of the SNR of each element of the IR channel for a direct Sun (100%) signal. The SNR is defined as the ratio of the absolute mean count number of the signal to the root mean square (RMS) of signal fluctuation. As shown in the figure, the SNR of a small element number has higher values (more than 1,000) than that of a larger element number (about 400 for element 43). This is because the solar emission energy has a peak at around 530 nm, and it decreases as the wavelength increases in the IR.
3.5. Output Signal of the Visible Channel
Figure 9a shows raw output signals of the ILAS visible channel at several tangent heights. The R-branch of the molecular oxygen A band absorption features is present between elements 200 and 300, while that of the P-branch is between elements 300 and 600. Several solar Fraunhofer lines are also in the spectra.
Figure 9b shows an example of the transmittance spectrum of the ILAS visible channel at 25-km tangent height. This spectrum was derived from the raw output signal divided by a direct Sun (100%) signal. There is strong absorption due to R- and P-branches of the molecular oxygen A band between 759 and 771 nm. Note that the maximum transmittance around this region is about 88%, not 100%. The attenuation was caused by Rayleigh scattering of air molecules, Mie scattering by aerosols, and absorption due to the ozone Wulf band. These attenuating effects should be removed when the aerosol extinction is retrieved from the visible channel [Hayashida et al., 2000; Yokota et al., 2002], or when temperature and pressure are retrieved from the ILAS visible channel [Sugita et al., 2001].
Figure 10 shows an example of the SNR of each element of the visible channel for a direct Sun (100%) signal, which was defined in a similar way as for the IR channel. Once again, the tendency of the SNR to decrease for large element numbers is seen (from a SNR of about 1800 for element 0 to about 1600 for element 1023). This can also be attributed to the change in the solar emission energy in wavelength.
3.6. Output Signal of the SES
 ILAS carried a SES to determine the tangent height of the IFOV geometrically, using the instantaneous positions of the ADEOS satellite, the Sun, and the Earth [Nakajima et al., 2002]. Figure 11 shows raw output signals of the SES at several tangent heights. A smaller element number in the horizontal axis corresponds to a higher tangent height. The dotted rectangle at the center indicates the location and opening height of the telescope's entrance slit. From these intensity curves of the output signals, the upper and lower edges are derived as the inflection points of each curve. As tangent height decreases, the intensity of the SES signal decreases due to scattering by air molecules and aerosols. In addition, the width between the upper and lower edges decreases because the image of the Sun becomes elliptical due to refraction of the solar ray path. Note that the shapes of curves at 30- and 20-km tangent heights show apparent asymmetry because there is stronger extinction of the solar ray at lower altitudes due to aerosols.
 As shown in Figure 11, the solar profiles are not hummock shaped, as shown in the 747 nm curve in Figure 12, and as was expected theoretically and from the EM test. The cause of this discrepancy was investigated using the backup optics of ILAS/PFM and shown to be the result of light other than in the expected wavelength region (center = 747 nm, HWHM = 12.5 nm) transmitted through the optics due to imperfections in the optical filter. Figure 12 shows that the intrusion of 464–636 nm wavelength light, which peaks around 590 nm, has a broader skirt due to the chromatic aberration of the optics, resulting in the blurred Sun profile seen in the “Total” curve in Figure 12. In spite of this effect, determining the upper and lower edges of the Sun using the inflection points of each curve gives a sufficiently accurate value for tangent heights [Nakajima et al., 2002]. Detailed error estimates of tangent height determination using the SES are described by Nakajima et al. .
3.7. Sun Tracking With Two-Axis Gimbals in Space
 Sun tracking with the two-axis gimbals of ILAS was essential for occultation measurements. There were three fundamental requirements for Sun tracking. The first was to track the Sun to as low a tangent height as possible, which became the lowest tangent height for gas retrieval. The second requirement was to minimize the fluctuation in the tracking to reduce the retrieval error. The last requirement was to avoid random and unknown offsets to the tracked radiometric center of the Sun. These requirements minimized the systematic error of retrieval, which may arise from differences between measured and theoretical transmittance.
Figure 13 shows the distributions of the lowest tangent height of Sun tracking for both sunrise and sunset measurements. The distribution of lowest tangent heights for sunset measurements has a peak at 9–10 km, while that for sunrise measurements has a peak at 12–13 km. This difference arises from two effects. One is the difficulty in capturing the brightness center of the Sun within a short time period just after sunrise, when the shape of the Sun is distorted due to the refraction effect. The other is the discrepancy in tropopause height between sunrise and sunset occultation measurements. The orbital characteristics of ADEOS caused sunrise occultation in the Northern Hemisphere to occur mostly at lower latitudes than sunset occultation in the Southern Hemisphere [Sasano et al., 1999a]. The higher tropopause height in sunrise occultation in the Northern Hemisphere resulted in the higher lowest tangent height of the measurements compared with sunset occultation, because of the frequent existence of cirrus clouds at the tropopause.
Figure 14 shows SES element numbers of the upper and lower edges of the Sun, as well as the middle point of those edges for one sunset occultation scene. Minimum data unit of ILAS measurement corresponds to 83.139 ms in sampling time (12.028 Hz sampling). The upper and lower edges were determined as inflection points of the SES data [Nakajima et al., 2002]. Small fluctuations in the Sun tracking by a few elements were seen in both edges. The maximum tracking fluctuation of the Sun tracker was about 24″ peak-to-peak, but this does not affect the accuracy of tangent height registration because it only requires the angular distance between the top edge of the Sun and the IFOV direction [Nakajima et al., 2002]. The sharp spikes at around 45 s (relative time) and near the end of the occultation are errors due to imperfect edge determination, which should be removed at the time of data retrieval [Yokota et al., 2002]. When the Sun moved into the Earth's atmosphere, the diameter of the Sun, defined as the difference between the upper and lower edges, decreased, due to the refraction effect, at around 50 s. Nevertheless, the element numbers of the middle point of the Sun remained constant. This verifies that the Sun tracking worked normally during the course of the occultation event, and the edge-detection algorithm also worked properly. Details can be found in the work of Nakajima et al. .
 Optical offsets between the radiometric center of the tracking optics and that of the entrance slit were measured before launch. They were 6″ in the vertical, and 28″ in the horizontal direction, which are negligible compared to the radius of the Sun (15′ 32.58″).
3.8. Long-Term Changes in Instrument Temperature and Output Signal
 As described in section 3.3, the long-term change in ILAS temperature shown in Figure 6 could have affected the visible spectrometer's instrument function. In addition, this temperature change could have affected the determination of tangent height using the SES. The procedure for correcting tangent height determination due to a change in instrument temperature is described by Nakajima et al.  in detail. For the IR spectrometer, the effect of instrument temperature change was negligible because the spectral resolution of the IR spectrometer was low.
 The absolute signal levels of satellite sensors sometimes deteriorate gradually in the harsh environment of space. In the case of ILAS, the 100% solar signal showed a gradual decrease after launch, as shown in Figure 15. Sensors that detected shorter wavelengths (visible channel, SES, and FSPS) had larger degradation slopes than the IR channel or CSPS (not shown). The degradation may have been partly due to contamination. However, changes in the absolute signal levels do not affect data retrieval, as long as the degradation in the wavelength of each spectrometer is sufficiently smooth, because we measured the 100% signal as a reference. In other words, measurements are self-calibrated for each occultation scene. Moreover, the decrease in absolute count numbers can be adjusted by applying different gains of the amplifiers in order to maximize the SNR.
 ILAS was designed and developed by the EA to monitor the stratospheric ozone layer, and the PFM of ILAS that was launched was manufactured in September 1994. After ILAS was launched on board ADEOS on 17 August 1996, it made about 6700 solar occultation measurements in both the Northern and Southern Hemispheres between October 1996 and June 1997. The performance of each component of ILAS was characterized using prelaunch test data and in-orbit performance data. Before the launch, gas-cell measurements, time response measurements, Sun-tracking tests, hollow-cathode lamp measurements, etc. were conducted. To check in-orbit performance, Sun-tracking performance, the raw signal of IR and visible spectrometers, SNR of IR and visible spectrometers, and long-term changes in the output signals of each sensor were investigated. The time delay response of the IR channel was characterized using stepwise IR input. Instrument functions of ILAS IR and visible spectrometers were determined by combining theoretical optical calculations, experimental measurements using a gas-cell before launch, and in-orbit data. The temperature of the optical bench of ILAS increased by about 4°C during the normal operation period; this change was attributed to the change in the distance between the Earth and the Sun. The absolute wavelength of the visible spectrometer shifted slightly with this temperature change. To compensate for this effect, instrument functions of the visible spectrometer were determined each week for each sunrise and sunset, using the solar Fraunhofer lines seen in the direct solar data. The method used to determine the solar edges from the SES data worked correctly, giving adequate tangent height information for the observations. ILAS worked as designed during its operation in orbit and gathered valuable data for ozone layer studies.
 The authors wish to thank the National Space Development Agency of Japan (NASDA) for their cooperation with the ILAS project and their successful launch and operation of the ADEOS satellite. They thank N. Takeuchi for his contribution in the design stage of the ILAS instrument. They also appreciate Y. Itou, M. Kaji, E. Iwama, and their coworkers at Fujitsu F.I.P. Co. for their contribution in processing the ILAS data at the ILAS Data Handling Facility (ILAS-DHF) of the National Institute for Environmental Studies. The ILAS project is funded by the Environment Agency of Japan (reorganized as the Ministry of the Environment in January 2001).