Observing the universe in the submillimeter spectral region

Cooling 2,560 bolometers
One challenge remained, however: Ensuring that the sensitivity of the new detectors would be limited solely by the fluctuations in radiation originating in the telescope’s mirror (albeit cooled to close to –200 °C). Previously, bare detectors involving a surface area not even equal to 1 mm2 were able to detect a source equivalent to a 100-watt light bulb positioned 300 kilometers away. The Herschel observatory can detect a similar source at a distance comparable to the distance from Earth to the moon. Achieving such a performance entails cooling every bolometer, standing as it does as a veritable thermal detector of radiation, to a far lower temperature than that of the instrument itself, i.e. to 0.3 Kelvin. To achieve such a temperature in space, researchers turned to closed-circuit helium (3He) evaporation cryocoolers, developed by CEA’s Low Temperatures Service (SBT).

The chief issue arising with this type of system relates to the total power “budget” available across the focal plane, to wit, 10 mW – this power being indispensable for biasing purposes, readout, and the detectors’ electrical and mechanical connections. Development of an ingenious multiplexing setup enabled researchers to deploy 2,560 operational bolometers in the PACS camera, within that power budget. The more conventional SPIRE instrument is only able to control 300 bolometers, even though it draws on the same cryocoolers.

The detectors are distributed across two focal planes to cater to the 60–130 mm, and 130–210 mm ranges, respectively. Within the camera, their layout makes it possible to map the same region in the sky at different wavelengths. At the shorter wavelengths, focal plane complexity stands at 64x32 pixels, making it possible to maximize the advantage of the resolution afforded by Herschel’s large mirror; on the other hand, a complexity of 32x16 pixels proves adequate for the longer wavelengths.

Hybridization ensures functionality
Each focal plane is built up from 16x16-pixel modules, which may be joined end-to-end on three sides. And, as the detectors exhibit a wide spectral-absorption band, modules may be mounted indifferently onto either of the optical channels. It is solely the optical filters positioned in front of each focal plane that determine the wavelength band detected. Every pixel in the module corresponds to a silicon bolometer, which has undergone prior hybridization, by way of an indium bump, to a readout and multiplexing circuit based on CMOS technology. The hybridization operation ensures three functionalities:

• the usual electrical interconnection of bolometer, and readout circuit
• the setting up of an optical cavity, ensuring the efficient trapping of incident light
• bolometer mechanical strength, and thermalization

The size of the hybridization bumps was calculated to ensure that the resonance cavity would be tuned to the wavelength to be absorbed. This cavity comprises the bolometer array, and a reflector, positioned over the CMOS circuit: this promotes maximum absorption of the incident wave, close to unity. Absorption occurs on metal deposited onto the array, exhibiting a surface impedance matching that of the vacuum.

Array’s twofold advantage
The silicon bolometer array is suspended by way of very thin (≈ 2 mm), low-thermal-conductivity beams. This setup allows the tenuous absorbed radiation to induce a measurable rise in temperature. Finally, a doped-silicon thermometer, positioned at the center of the array, effects the measurement, making use of an exponential law, relating resistance to temperature. This exhibits a temperature coefficient close to 3,000%/K. Compared to a “solid” surface, the array affords a twofold advantage: first, a lower heat capacity is exhibited, ensuring a swifter thermal response rate; second, it is less susceptible to the ionizing cosmic particles in the space environment. Obviously, the use of an array to absorb light may raise questions: would not such light “get through” the gaps in the array? No, because light does not “detect” details smaller than its own wavelength. All that is required is to fabricate an array with a pitch smaller than the wavelength to be detected.

As regards the 16x16-pixel modules, every one of them was individually evaluated, several years before launch of the mission and based on their performance, they were integrated into the focal planes, and calibrated a first time. Once the complete camera had been integrated into the PACS instrument – including its cryocooler, and flight electronics – final calibration was carried out. Early in 2008, the PACS instrument in turn was mounted on board the Herschel satellite, alongside SPIRE, and HIFI (Heterodyne Instrument for the Far Infrared). After the final adjustments, e.g. replacing faulty connectors and rerouting cables to preclude interference from the solar panels, the satellite was determined to be “ready for service” in December 2008.