ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117
Focal plane cameras for ESA optical astronomy missions D.H. Lumb, N. Rando, A. Peacock, F. Favata, M.A.C. Perryman* Science Department, European Space Agency, ESTEC, Postbus 299, 2200AG Noordwijk, The Netherlands
Abstract Charge-coupled devices (CCDs) have played an important role in imaging and spectroscopic measurements for visible wavelength astronomy. Their introduction has been largely responsible for driving the deployment of 8–10 m class telescopes, and in consequence have revolutionized experimental cosmology. Other scientific initiatives such as planet-finding are rapidly increasing in importance for the astronomical community, and we review two future European Space Agency missions which emphasize such new directions. In each case, we address some detailed design issues for their focal planes, showing how new CCD development activities may satisfy the mission requirements. r 2003 Elsevier B.V. All rights reserved. Keywords: Detectors; Astronomy; Charge-coupled devices
1. Introduction From the appearance of the charge-coupled device (CCD) as an optical image sensor [1] to its first deployment in ground-based astronomy [2] was the comparatively short interval of 7 years. Despite the relatively small size of arrays, the overwhelming advantages of quantum efficiency, linearity, dynamic range and stability, quickly made them the detector of choice in the 1980s. Much of the rapid early development in the US community benefited from the significant investment from the NASA Space Telescope programme [3], which for the first Wide Field and Planetary Camera resulted in the additional key technologies of back-illumination and 2-side buttability, as well
*Corresponding author. Tel.: +31-71-5654446; fax: +31-715654690.. E-mail address:
[email protected] (D.H. Lumb).
as many ground-breaking performance features of read out noise and charge transfer efficiency. Notwithstanding the protracted delays in the Hubble Telescope flight opportunity, the CCD had in the meantime become the pre-eminent technology in ground-based astronomy. For imaging applications the above-mentioned characteristics, for example, led to sensitivities allowing detection of cosmologically distant targets with typical R magnitudes B26 [4], which compares with B21 magnitudes with large photographic plates. At the same time, the readout of dispersed spectra revolutionized quasar red-shift identifications. It was this improvement that was the catalyst for the development of the new generation of 8–10 m telescopes: the existing 4 m class telescopes had finally become photon starved. The era of ‘‘precision’’ experimental cosmology is now being facilitated with the development of large mosaic arrays of CCDs that at last approach the dimensions of photographic plates
0168-9002/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2003.08.013
ARTICLE IN PRESS D.H. Lumb et al. / Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117
(e.g. Megacam [5] with 40 CCDs offering a total of 350 Mpixels). However these scientific interests are challenging silicon as the detector material of choice. Red-shifts of 4–5 place peak emission of galaxies in the IR, while signatures of early star formation and obscured accretion processes also peak in the IR, driving the choice of new detector technologies for the Next Generation Space Telescope. At the same time there are significant paradigm shifts in the most popular astronomy topics, for example the enormous interest in future initiatives to discover Earth-like planets in other stellar systems. Such new initiatives are addressed in the ESA New Horizons programme. Two ESA missions focus on the rather different techniques of astrometry and asteroseismology.
2. Eddington Eddington [6] was proposed in 2000 in response to the call for flexi-missions, and subsequently selected as a reserve mission. At the Science Program Committee meetings of 2002, it was suggested that this mission could be deployed as part of the Herschel–Planck project line to leverage use of the spacecraft. In this context, an instrument must be ready for launch in 2007, and therefore no development activity can be accommodated, and the challenging requirements must be met with existing technology. The mission has two complementary scientific aims: to produce data on stellar oscillations, necessary for the understanding the interior structure and evolution of stars, and to detect and characterize habitable planets around other stars via measurement of transits. 2.1. Eddington science requirements The stellar oscillations must be measured by long duration photometry on thousands of stars. The required frequency range demands measurements on timescales of B10’s seconds for 30 days. The photon shot noise requirements drive the photon collection requirements, while the field of view and magnitude range ensure that sufficient star samples can be taken in one field. These
113
Table 1 Eddington scientific requirements for asteroseismology Parameter
Requirement
Goal
Freq. precision measurement Frequency range Magnitude range Field of view Fourier noise amplitude level Flux scaling
0.3 mHz
0.1 mHz
1 mHz–10 MHz mv=5–15 2.5 diam. o1.5 ppm in 30 days 6 105 elect/s for mv=11 G0 sophoton noise/3 for mv=11 30 days o 30 s
1 mHz–100 MHz mv=3–15 6 diam.
Noise Monitoring time Time sampling
o5 s
requirements are summarized in Table 1. A design has been developed, that meets these goals within a compact spacecraft physical envelope, and comprises four co-aligned telescopes, each with a dedicated focal plane camera. The planet-finding phase of the mission requires the detection of sufficient transits to determine the statistical significance as to the proportion of earth-like planets in our galaxy. The transits should be detectable on time scales of B10 min, but to access the required number of late-type dwarfs implies a working magnitude range of mv B11–17 and a noise level B6 10 5 in 1 h, and hence a certain photon shot noise and throughput. The requirements are summarized in Table 2. In order to attain high photometric accuracy, a CCD detector must have minimal response nonuniformity, and this can be enhanced by defocusing a stellar spot, on scales large compared with pixel-to-pixel variations. The signal handling capacity of CCDs scales with pixel size, and furthermore four-phase electrode architecture offers an additional benefit over conventional threephase designs, due to a combination of the relatively higher fraction of pixel which can be kept inside the confining potential well during transfer, and the effects of fringing fields at the potential well edges. A back-illuminated topology minimizes intra- and inter-pixel sensitivity variations. The requirement for high dynamic range is
ARTICLE IN PRESS 114
D.H. Lumb et al. / Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117
Table 2 Eddington scientific requirements for planet- finding Parameter
Requirement
Light curve noise level Late-type dwarfs to monitor Time sampling Magnitude range Flux scaling
10
Noise Monitoring time
5
Goal
in 39 h
20,000 at above S:N o600 s 11–17 7 104 elect/s for mv at the limit sophoton noise/3 for mv at limit >3 years
100,000 at lower S:N o30 s .
especially taxing, because a large field of view implies a long readout duration and thus a large accumulated signal is to be sensed at the output node while at the same time measuring with low noise. 2.2. CCD selection In principle, a custom CCD design could be implemented that would best match these objectives. However given a compressed program requiring delivery of a flight detector in ca. 2004 it was decided to see if an off-the-shelf design could satisfy the requirements. As a baseline we considered the E2 V (aka Marconi Applied Technologies) CCD42 family of devices, which has been widely used for astronomical applications, including some space-based instruments [7]. While not necessarily the only CCD solution available, this selection allows us to size other aspects of the instrument design realistically. The key features are a pixel size 13.5 mm square, back illumination, three-side buttability, image section full well of B1.5 105 electrons, but with serial register and amplifier designs that accommodate at least four image pixels of charge. The width of the array is 2048 pixels, and height can be chosen according to different photolithographic mask components. We consider an array that is the maximum height that can be accommodated in a 5 in. diameter silicon wafer and still split in identical store and image registers. The spacecraft envelope limits the telescope focal length to B1 m,
Fig. 1. Eddington field of view offered with a 3 2 array of CCD42-C0 devices. Image area (white) is 4.7o squared. Darkened rectangles are identically sized store sections.
and this drives a trade-off for focal plane scale and defocusing. Based on a B10 arcsec spot size, we therefore need to define readout mode timings that accommodate the maximum dynamic range, and compare the performance with the stated requirements. Combining four rows of image pixels into the output register, and reading at a maximum pixel rate of B2–3 MHz would allow a frame interval of less than 1 s. A defocusing of the spot to 50 pixels in area thus allows a signal handling of Bmv=8 and down to a noise limit less than mvB15. To achieve the required dynamic range requires restricting the accumulation time by an order of magnitude, but results in losing a comparable livetime fraction in this mode. However such brightest stars may be limited within the focal plane and different readout options applied from CCD to CCD (Fig. 1 and Table 3). This example demonstrates the highly adaptable nature of CCDs to satisfy operational requirements, relying only on the flexibility of readout modes to access a large parameter space. The next application in contrast is so specialized that it can only be satisfied by custom developments of new detectors.
3. GAIA GAIA is conceived as one of the major future missions in ESA’s portfolio [8], scheduled for launch in the 2010-1012 time frame. It will rely on
ARTICLE IN PRESS D.H. Lumb et al. / Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117 Table 3 Summary parameters for the Eddington camera when implemented with CCD42-C0 CCDs Parameter
Value
Field of view Pixel format Pixel rate Readout frame Normal Reduced Binned pixel Peak signal Max brightness Wavelength range
4.8 4.8 3072 2048 2.5 MHz 0.75 s 0.05 1 4=13.5 54 mm 6 105 electrons/pix/frame mv B7.7 (B4 in reduced frame) 400–800 nm
the proven principles of ESA’s Hipparcos [9] mission to solve one of the most difficult yet fundamental challenges in modern astronomy: to create an extraordinarily precise three-dimensional map of more than one billion stars throughout our Galaxy and beyond. In the process, it will map their motions, which encode the origin and subsequent evolution of the stars. Through a comprehensive photometric classification, it will provide detailed physical properties of each star observed: characterizing their luminosity, temperature, gravity, and elemental composition. This massive stellar census will provide the basic observational data to tackle an enormous range of important problems related to the origin, structure, and evolutionary history of our Galaxy. GAIA will achieve this by repeatedly measuring the positions of all objects down to mv=20. Final accuracies of 10 marcsec at mv=15, will provide distances accurate to 10% as far as the Galactic Centre, 30,000 light years away. 3.1. GAIA capability *
*
*
Catalogue: of the order of 1 billion stars; 0.34 106 to mv=10; 26 106 to mv=15; 250 106 to mv =18; 1000 106 to mv=20, completeness to about 20 mag. Sky density: mean density approximately 25,000 stars deg 2, maximum density: about 3 106 stars deg 2. Accuracies: median parallaxes of 4 mas at mv =10, 10 mas at mv=15, 160 mas at mv=20.
*
* *
115
Distance accuracies: from Galaxy models: 21 million better than 1%, 46 million better than 2%, 116 million better than 5%, 220 million better than 10%. Radial Velocities: 1–10 km s 1 to mv=16–17. Photometry: to mv=20 in four broad and 11 medium bands.
3.2. GAIA measurement techniques Measuring the trigonometric parallax for a star yields the only fully reliable way of measuring distances in the local Universe. The first spacebased astrometry mission, ESA’s Hipparcos satellite, demonstrated that milliarcsecond accuracy was achievable by means of a continuously scanning satellite that observed two directions simultaneously. In addition, the global nature of the measurements, the fact that the positions and changes in positions caused by proper motion and parallax are determined in a reference system consistently defined over the whole sky, leads to the determination of absolute parallax measurements. (In ground-based parallax measurements the transformation of relative parallaxes to absolute distances is a non-trivial problem.) Global astrometry has many intrinsic advantages over pointed observations: a global instrument calibration can be performed in parallel with the observations; and the interconnection of observations over the celestial sphere provides the rigidity and reference system needed for the kinematical interpretation of the observations themselves. The core science case for GAIA requires measurement of luminosity, effective temperature, mass, age, and composition for the stellar populations in our own Milky Way Galaxy and in its nearest galaxy neighbours. These quantities can be derived from the spectral energy distribution of the stars, through multi-band photometry. A GAIA photometric system has been derived following study of the scientific and technical requirements of the GAIA payload. This resulted in the current baseline system of four broad bands that will be in the main astrometric instrument focal plane and 11 photometric bands covering the spectral range 280–920 nm, that will
ARTICLE IN PRESS 116
D.H. Lumb et al. / Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117
occupy the focal plane of a dedicated spectrometric instrument. Stellar radial velocities are needed if we wish to know the space velocity of stars with respect to the Sun. They are thus essential to our understanding of the kinematics of our Galaxy. The astrometric measurements supply only two components of the space motion of the target stars (vr is needed for proper kinematical or dynamical studies). To maximize the radial velocity signal even for metal-poor stars, strong saturated lines are desirable. Broad lines also allow the radial velocity to be accurately derived from even moderate resolution spectra. The Ca II triplet around 860 nm will be used. 3.3. GAIA astrometric CCD The scanning motion of the sky is focused onto the focal plane and the resulting photoelectron signal charge pattern is clocked synchronously via. a Time Delay & Integration readout through the CCD. By completely oversampling the Point Spread Function (PSF) by some six pixels, the centroid of individual star PSFs can be determined with sub-pixel accuracies. The required accuracies demand a Signal-to-Noise ratio that can only be compiled over an interval that is much longer than could be sustained with a single CCD (scan rate B60 arcsec/s). The concept that has been developed is to pass the star pattern repeatedly over a column of identical CCDs, and recognize the passage of selected stars using a set of mapping CCDs located at the start of the scanning array. Thus only the windows for selected stars are read out of each astrometric CCD. The challenge of this centroiding technique is to match the compact optics design with small (9 mm) CCD pixels, yet obtain a relatively long CCD (58 mm). At the same time such a CCD requires the usual enhancements of lowest noise and high detection efficiency delivered with back illumination technology. In addition the maximization of the stars’ signal must be achieved by packing the whole focal plane as densely as possible. The concept is shown in Fig. 2. Ten columns of 11 CCDs comprise the astrometric CCDs, while two rows of CCDs perform the
Fig. 2. GAIA astrometric focal plane, comprising 110 astrometric CCDs sandwiched between rows of sky mappers and photometers. Stars scan through the field from top to bottom and are extracted from each CCD in turn.
mapping function. At the end of the array are 5 rows of Broad Band Photometer CCDs. This challenging design surpasses any planned mosaic of ground-based CCD cameras in area of silicon deployed. Each CCD is B27 cm2, and with 180 CCDs in the main focal plane this represents B0.5 m2 of close-packed detectors. The challenge is not only for the CCD design itself, but the ability to package this array efficiently and integrate the drive and readout electronics as close to the focal plane as possible. At the same time complex (and different) windowed readout schemes in each CCD have to be maintained without penalizing noise performance through cross-talk.
4. Space environment Mention should be made of special requirements of operating in space. In particular, the radiation environment has important implications for these and other applications. The prompt response to cosmic rays manifests as a signal of typically 2000– 3000 electrons spread over a handful of pixels. While this represents a minor contamination to the signal in high dynamic range applications, where the background at the lowest noise levels must simultaneously be measured, some form of median filtering must be employed. Such algorithms must
ARTICLE IN PRESS D.H. Lumb et al. / Nuclear Instruments and Methods in Physics Research A 513 (2003) 112–117
account for the much higher cosmic ray rate than ground applications (7 cm 2 s 1 in interplanetary space). The effects of prolonged exposure to space radiation, such as solar flares, is well known to cause a degradation in charge transfer efficiency. Trap sites at dislocation defects remove signal from charge packets, in a manner that varies strongly with signal density, device architecture and clocking rates. Especially for the large CCDs described here, the transfer laterally through many cm of damaged silicon, presents a tremendous challenge for device performance. The characterization of effects on accurate photometry and centroiding will be a significant calibration effort.
5. Future instruments CCDs are likely to remain state of the art detectors for space-borne astronomy missions for at least another decade. However there are clear trends towards future requirements that cannot be met by traditional CCD designs. Astrophysical phenomena whose emission peaks in the IR range are important studies that will need detector materials with narrower bandgap than Si. Current advances in the USA for multiplexed pixel arrays of InSb and HgCdTe have been tremendous, producing Megapixel class arrays with readout noise approaching that of CCDs.
117
Silicon may remain an important sensor material through the medium of CMOS active pixel sensors. Integrating the readout functions onto each pixel will mitigate against radiation damage concerns, and allow extremely rapid access to pixels of interest. Integrating intelligence into the front-end readout control could optimize performance in the context of reading out restricted windows of interest, and would have been ideally suited to the Eddington camera. More generally such sensors will facilitate highly integrated, lower power, low-mass instruments.
References [1] W.S. Boyle, G.E. Smith, Bell Syst. Technol. J. 49 (1970) 587. [2] R.B. Wattson, S.A. Rappaport, E.E. Frederick, Icarus 27 (1976) 417. [3] F.P. Landauer, J.R. Janesick, S.L. Knapp, et al., Proceedings of the 1978 Government Micro Applications Conference P394, Monterey, CA, November, 1978. [4] J.E. Gunn, J.A. Westphal, Proc. SPIE 290 (1981) 16. [5] O. Boulade, X. Charlot, P. Abbon, et al., Proc. SPIE 4008 (2000) 657. [6] Eddington Assessment Study, F. Favata, I. Roxburgh, J. Christensen-Dalsgaard (Ed.), ESA-SCI(2000)-8 pub, European Space Agency, Noordwijk, The Netherlands, 2000. [7] W.W. Weiss, A. Baglin, Proc. SPIE 4013 (2000) 450. [8] GAIA Assessment Study, K. De Boer, G. Gilmore, E. Hog, et al., (Eds.), ESA-SCI(2000)-4 pub, European Space Agency, Noordwijk, The Netherlands, 2000. [9] M.A.C. Perryman, E. Hog, J. Kovalevsky, et al., Astron. Astrophys. 258 (1992) 1.
