Journal of Low Temperature Physics manuscript No.(will be inserted by the editor)

Vincent Reveret · Philippe Andre · Michel Talvard ·Louis R. Rodriguez · Olivier Boulade · Eric Doumayrou ·Pascal Gallais · Benoit Horeau · Jean Le Pennec · MichelLortholary · Jerome Martignac · Patrick Agnese

HERSCHEL - PACS bolometer arrays for submillimeterground-based telescopes

Received July 23, 2007, Accepted September 15, 2007

Abstract The Herschel Space Observatory will carry onboard a new kind of bolometric architecture for thePACS1 submillimeter photometer. These new generation CCD-like multiplexed bolometer arrays are buttableand enable the conception of large fully sampled focal planes either for space or for ground-based telescopes.We present here some development for ground-based applications in the context of the ARTEMIS2 project.We have developed an electro-thermal numerical model that simulates the performances of these semicon-ducting bolometers under specific ground-based conditions (different wavelengths and background powersfor example). This model permits to determine the optimal parameters for each condition and shows thatthe bolometers can be background limited in each atmospheric transmission window between 200 and 450microns. We also describe the optical system that provides a high optical efficiency in each submillimeter at-mospheric window. Astronomical observations made with a prototype on the APEX telescope are presented.

Keywords Large bolometer array · ground-based telescopes · submillimeter astronomy · simulations ·submillimetric absorption · thin dielectric layers

1 CEA Bolometers - From Space to Ground

1.1 The PACS photometer

The Herschel Space Observatory which will be launched in the second half of 2008 [1] is the third cornerstoneof the European Space Agency. This submillimeter space observatory will carry onboard three instruments:HIFI (high resolution heterodyne spectrometer), SPIRE and PACS (both spectro-imagers) for operation be-tween 60 and 670 µm.

In 1997, two CEA laboratories (LETI/LIR and DAPNIA/SAp) have started the development of new tech-nology bolometers to meet the requirements of the PACS instrument [2] : how to build a wide-field submilli-metric camera with fast mapping capability and a very good sensitivity in a relatively high background powerenvironment? The approach was to use different known technologies, many of them developed during the

V. ReveretEuropean Southern Observatory, Casilla 19001, Santiago 19, ChileE-mail: vreveret@eso.orgV. Reveret · P. Andre · M. Talvard · L. R. Rodriguez · O. Boulade · E. Doumayrou · P. Gallais · B. Horeau · J. Le Pennec · M.Lortholary · J. MartignacCEA/DSM/DAPNIA Service dAstrophysique, Saclay, 91191 Gif sur Yvette Cedex, France

P. AgneseLETI/CEA Grenoble, 17 avenue des Martyrs, 38054 Grenoble Cedex 9, France1 PACS stands for Photodetector Array Camera and Spectrometer2 ARTEMIS stands for ARchitecture de bolometres pour des TElescopes sub-MIllimetriques au Sol

2

!!"

Indium Bump

Thermal Link

Metallic pattern

Implanted thermometer

Si:P:B

Reector

Heat Sink 300mK

PACS Focal plane with 2048 pixels 1 Pixel

Fig. 1 (Color online) The picture on the left shows a pre-version of the 100 µm focal plane of the HERSCHEL PACS pho-tometer. It is made of 8 subarrays each containing 256 bolometers. The sketch on the right side shows one of these pixels. A firstelectronic stage (impedance reduction) is located below the quarter wave cavity reflector, at 300 mK. CMOS transistors at 2Klocated a few cm below the pixels perform the multiplexing.

ISOCAM project at CEA [3], and put them together to build a new kind of bolometer array. Such technolo-gies include ionic implantation (for resistive thermometers), silicon micro-machining, flipchip technologyusing indium bumps to bond the different elements and low temperature CMOS amplification. It was alsodecided to avoid Winston cones, and instead, to use a ”resonant metallic absorption” system.

That led to the conception of a non traditionnal bolometric detection system which is more similar tothe CCD concepts used in other domains of astrophysics. 256 pixels compose the basic array, each pixelhaving a 0.5F! field of view to fully sample the image in the focal plane. Pixels contain two semiconductingthermometers (silicon made with phosphorus implantation and boron compensation) which work at 300 mKin the ”hopping conduction” regime. The submillimeter light absorption is done via a quarter wave cavity.The bottom of the cavity is a silicon layer coated with a thin gold layer that acts as a reflector. Indiumbumps support the upper absorbing layer made of silicon which is metal coated (TiN, superconductive at300 mK but absorbing for submillimeter frequencies). This layer has a grid-like geometry in order to reducethe heat capacity of the bolometer. The signal is read out at the middle point of the resistor bridge and isprocessed by CMOS transistors. Because of the intrinsic noise of CMOS followers, the impedances of thebolometer resistors have to be very high (! G" ) in order to provide a very high response. A cold (2K)CMOS multiplexing system is located below the detection stage. These bolometer arrays can be butted end toend to form very large focal planes. CEA has built two bolometric focal planes for the PACS photometer onHERSCHEL [4]. Flight models are now being tested and calibrated. The detectors fulfill all the requirementsin particular in terms of sensitivity (BLIP3 conditions in the two bands), bandwidth (electrical and thermal)and 1/ f noise suppression.

1.2 Development for ground-based telescopes.

Since 2005, in the context of the ARTEMIS project, we have started to work on the ground-based applicationsfor these bolometers [5]. The wavelength range between 200 µm and 1 mm is of primary importance in today’sastronomy. Topics such as the study of the first stages of star formation, or distant submillimeter galaxies canstrongly benefit from the large diameters (and hence the good angular resolutions) of ground-based telescopes.The technology developed for the PACS instrument seems to be very well suited for ground-based conditionslike the high and variable background power incoming on the focal plane. This technology is also clearly agood solution to cover large fields of view.

Figure 2 shows the typical transparency of the atmosphere in the submillimeter regime, for different valuesof water vapor contents. Only a few sites on Earth are suitable for submillimeter astronomy between 200 and865 µm. The Chajnantor plateau in the Atacama desert provides very good conditions, even at the shortestwavelengths.

The ARTEMIS project aims at developing a submillimeter bolometer camera with 4096 pixels (64"64)operating in the three bands : 200, 350 and 450 µm (and possibly later the 865 µmband). We have investigated3 BLIP : Background Limited Performances

3

!"" #"" $"" %"" &"""

'()*+*,-./01,023

"

"4!

"4#

"4$

"4%

&

5.3678/*91:0;9(,7317716,

<'=0>0"4!033

<'=0>0"4?033

<'=0>0&033

Fig. 2 (Color online) The different atmospheric transmission windows in the submillimeter regime at the Chajnantor site in theAtacama desert. The curves correspond to different values of precipitable water vapour contents (adapted from [6]). Only highaltitude dry sites provide the best conditions for astronomical use of the 200, 350 and 450 µm bands.

the performances of our bolometers in these atmospheric windows considering a 12m diameter telescope ona site like Chajnantor.

1.3 Electro-Thermal considerations

In order to estimate the best performances for our bolometers in the different atmospheric bands, we have de-veloped a numerical and dynamical model which permits us to obtain optimum values of different parameters(such as bias, or thermal conductance for example). The model is based on the general differential equationthat manages thermal exchanges in the bolometer [7] :

Pph+PJ =CdTbdt

+Gth(Tb#T0) (1)

with Pph = total photonic power incident on one pixel, PJ = electrical power coming from the bias of theresistive element in the bolometer,C= total heat capacity of the pixel, Tb = bolometer temperature, T0 = heatsink temperature, constant at 300 mK, Gth = average value of the thermal conductance between Tb and T0,and t = time.

Solving this equation permits us to get the temperature of the bolometer as a function of time. All theother parameters depend on this temperature :

– Bolometer impedance (empirically adapted from Efros Law [8] :

R= R0 exp

(√T0T

)exp

(#qL(T )EkT

)(2)

where R0 and T0 are constants caracterising the type of resistor we are using (Silicon type with implantedPhosphorus and Boron compensation), q is the electric charge of the electron, L(T ) is the hopping lengthand E is the electric field inside the resistor.

4

– Thermal conductance :

G(T ) = #1T +$1T 3 (3)where the T term characterizes the heat conduction in metals (from Drude theory) and the T 3 term corre-sponds to a dielectric conduction (Casimir model).#1 and $1 are constants.

– Heat capacity :

C(T ) = #2T +$2T 3+ %1 exp(&1TCT

)(4)

where the T term corresponds to the doped silicon resistor, the T 3 term corresponds to the dielectric(silicon grid) and the exponential term is the contribution of the superconducting absorbing metal (TiN,with a transition temperature of TC). #2, $2, %1 and &1 are constants.

All the constant terms and empiric laws for these parameters have been determined experimentally andend up as expected from previous literature. After having tested the validity of this model by comparingit to experimental results, we used this simulation to determine what would be the best performances of ourbolometers under ground-based conditions. For each band, we have estimated the background power incomingon one pixel (see table 1). It is possible to determine the optimum bias that leads to the best performance (interms of NEP4). Note that, we also tried to optimize the value of the thermal conductance and found thatthe best performances are obtained for a small change of the value compared to the one used for PACS (Gthincreased by a factor of 1.2).

Table 1 gives the estimated NEP and NEFD for each considered atmospheric band. The NEFD (NoiseEquivalent Flux Density) is an estimation of the instrument sensitivity during an astronomical observation. Itis equivalent to the flux of an object that is detected with a Signal / Noise = 1 in 1s of integration and is relatedto the telescope diameter, the optics transmission and the atmospheric conditions. The results show that ourbolometers can be background limited in the first 3 submillimeter bands.

Table 1 Estimated performances of the bolometers in the different atmospheric windows.

Wavelength in µm 200 350 450 850Average Window Transmission 50% 65% 65% 85%

Beam FWHM in arcsec 4 7.5 9.5 18Typical Background per pixel in pW 70 35 40 8Field of View (for a 64"64 array) 1.8’"1.8’ 3.2’"3.2’ 4.2’"4.2’ 7.8’"7.8’

NEP Photon in W/Hz1/2 5.5"10#16 2.9"10#16 2.8"10#16 0.7"10#16NEP Detector in W/Hz1/2 2.2"10#16 1.6"10#16 1.7"10#16 0.7"10#16NEP Total in W/Hz1/2 6.0"10#16 3.5"10#16 3.3"10#16 1.0"10#16NEFD in mJy (1' , 1s) 350 160 150 40

1.4 Absorption of the submillimeter light

Our bolometers use a resonant cavity to absorb submillimeter waves [9]. When an absorbing material5 isplaced at a distance !/4 over a reflecting layer, it is in theory possible to absorb 100% of the energy at thisparticular wavelength ! . To extend the absorption profile, we use specific metallic patterns like crosses orloops (horizontal resonance, see figure 3a).

In our design, we use indium bumps to tune the cavity height to the desired wavelength we want todetect. These bumps are also used as electrical links between the silicon grid (containing the resistors) and thereflector layer where we find the first stage of signal processing. The current limit in size for these bumps isaround 80 µm. A 60 µm cavity would give very high absorption in the three submillimeter bands (see figure4a). For the longer wavelengths (865 µm), an antireflective layer (silicon,112 µm thickness) can be used onto give an absorption better than 95% (see [10] for more details about this method).4 NEP : Noise Equivalent Power, the minimum power that the system can detect, inW/

$Hz.

5 We use a TiN alloy as the absorber because it is compatible with the manufacture processes and superconducting at 300mKbut absorbing for submillimeter waves.

5

Simple Silicon layer

Vertical Resonance

Horizontal Resonance

Pacs type array

Minor technological development

(a) (b) (c)

Absorbing grid

Silicon Layer

Cavity A

Cavity B

Fig. 3 (Color online) (a) Schematic representation of the ”double resonance” absorption system of one pixel. (b) A possiblesolution to tune the absorption of the bolometer to long wavelengths. (c) In practice, this solution consists in adding a thindielectric layer (for example silicon) on the top of an existing PACS type array.

200 400 600 800 1000

Wavelength in µm

0

0.2

0.4

0.6

0.8

1

Ab

so

rptio

n

200 400 600 800 1000

Wavelength in µm

0

0.2

0.4

0.6

0.8

1

Ab

so

rptio

n

Cavity 80 µm

Cavity 80 µm + 112 µm Si layer

Fig. 4 (Color online) (a - Left) Simulated absorption curve for a 60 µm cavity. At least 90% absorption can be obtained in eachband. (b - Right) When using a 112 µm silicon antireflective layer, an array with 80 µm cavities can reach 96% absorption inthe 850 µm band (continuous line). The dash line shows the theoretical absorption for a simple 80 µm cavity (72% absorption).

NGC6334

!"#$%&'(')*+,'

-./ -0/

Fig. 5 (Color online) (a) The APEX 12m submillimeter antenna located at 5100m in the Atacama desert. Thanks to its uniquelocation, it can make observation in all the submillimeter windows between 200 µm and 1.2 mm (see [6]). (b) The star formingregion NGC6334 as seen by the P-Artemis instrument in March 2007. It is a well known object, but some details in the Southpart of the map have never been observed before.

2 On-Sky tests - Perspectives

In March 2007, we have tested a prototype of the bolometer camera (called P-Artemis) on the APEX submil-limeter antenna in Chile. The instrument has a 16"16 multiplexed array, working at 450 µm. The bolometersthemselves were not optimised for APEX, as we used a spare detector which was not selected for PACS. This

6

first run on a telescope was mainly done to test the different concepts of the ARTEMIS project. In particular,we successfully tested the antireflective layer system.

We were able to detect several submillimeter sources (see figure 5b) and we obtained a sensitivity at 450µm of! 3 Jy.s1/2 which is quite high compared to our estimates. This high value can be explained by differentfactors :

– bad transmission and self-emission of the three HDPE6 lenses used in the optical system. A system usingmirrors could improve the sensitivity by as much as a factor !5.

– bad IR filtering inside the cryostat. A factor!2 in performances can be expected with an improved design.– non optimised detector. Bolometers specifically built for APEX could give performances improved by afactor 1.5 to 2 compared to the prototype.

New tests with an improved prototype (optics and filters) are scheduled for the near future in order to preparethe full ARTEMIS instrument that should become available by the end of 2008.

References

1. G. L. Pilbratt, AAS, 207-37, p. 1219, (2005).2. A. Poglitsch et al, SPIE Proceedings, 6265, p. 62650B, (2006).3. C. Cesarsky, A. Abergel, and P. Agnese, A&A, 315, p. L32, (1996).4. N. Billot et al, SPIE Proceedings, 6265, p.62650D, (2006).5. M. Talvard et al, SPIE Proceedings, 6275, p. 627503, (2006).6. L.-A. Nyman, P. Schilke, and R. S. Booth, The Messenger, 109, p.18, (2002).7. P. Richards, J. Appl. Phys., 76, p.1, (1994).8. C. Buzzi, Thesis of the University Joseph Fourier, Grenoble 1, (1999).9. L. Hadley and D. Dennison J. Opt. Soc. Am., 37, p. 451, (1947).10. V. Reveret et al, SPIE Proceedings, 6275, p. 627502 (2006).

6 HDPE stands for High Density PolyEthylene.