Terahertz focal plane arrays for astrophysics and remote sensing Christopher Groppi Arizona State University School of Earth and Space Exploration
Emission at 115 GHz from the CO molecule was first detected in 1970. Astronomers have been imaging THz light ever since.
Full Moon This is what we can do today: a 120 square degree image of 13 CO(1-0) at 110 GHz. Beamsize is ~1 arcminute in this image. Every pixel is an integrated high resolution spectrum.
IEEE mm-wave and THz Workshop 350 GHz continuum SWIR image 1.2, 1.6, 2.1 microns We can do the same trick looking at the thermal emission from small, solid particles ( dust ). Dust emits a blackbody spectrum, peaking at THz frequencies at typical molecular cloud temperatures.
Dust is 1% by mass of a molecular cloud, but is easy to detect and provides information about the cloud temperature and structure. IEEE mm-wave and THz Workshop Blackbody emission from dust
How do molecules emit light? Molecular clouds are 99% molecular gas, mostly H 2. We can t see H 2 (no dipole moment) so we look at other, less abundant molecules instead (like carbon monoxide). Blue / Higher Frequency Red / Lower Frequency
How do we make images? Until recently, virtually all telescopes were equipped with single pixel THz detectors. Images are made using the On The Fly Mapping technique. Antenna is raster scanned across the source at a fixed angular rate. Receiver is read out rapidly (several Hz). Lots of short integrations at closely spaced intervals are convolved with a Gaussian kernel the size of the beam on a Nyquist sampled grid. Much faster than point by point mapping, since multiple spatial pixels share the same reference.
Tradeoff between spatial resolution and mapping speed. Large antennas have smaller beams, resulting in better spatial resolution in the final image. This also results in more pixels in an image of a given angular size. Bigger telescopes map a given area slower. If you want your cake and eat it too (wide areas AND high spatial resolution) you need multiple spatial pixels.
Full Moon This is what we can do today: a 120 square degree image of 13CO(1-0) at 110 GHz. Beamsize is ~1 arcminute in this image. Every pixel is an integrated high resolution spectrum.
Coherent array receivers SEQUOIA was built for the 15m FCRAO antenna in Massachusetts. Operates from 85-115 GHz. 32 cryogenic HEMT amplifier based pixels (16 in each linear polarization).
PoleStar array for AST/RO 810 GHz 4 Pixel SIS superconducting mixers Solid state local oscillator source (~0.3 mw) L-band IF 1-2 GHz Trec~550-650 K 4 channel array AOS backend spectrometer
Supercam System 2 8 channel downconverter modules LO System with 8 way power divider LO Optics LO Beamsplitter & dewar window CTI 350 cooler Sumitomo 4K cooler Omnisys Spectrometer 64x250 MHz complete system Prototype 8 channel bias system (1 6U card with power supplies) Spectrometer and bias control computer
1x8 Mixer Module Bias DC connector Gilbert GPPO blind mate IF connectors Magnet DC connector Electromagnets Horn Extension Block
A Closer Look Ground beamlead SOI SIS Chip Beamlead alignment tabs IF Beamlead Magnet probes Input matching network WBA13 MMIC chip Bias circuitry Output coax
Low Noise Amplifiers 32 db Gain, 5 K Noise at 8mW power dissipation N. Wadefalk, J. Kooi, H. Mani & S. Weinreb, Caltech
Supercam IF Processing 1x8 Downconverter Module (Caltech: G. Jones and J. Bardin) Total power metering 250 MHz and 500 MHz bandwidth modes (1 GHz with filter change) Digital attenuators Low cost surface mount components
Supercam Spectrometer (Omnisys) Real-Time FFT system Virtex 4 SX55 FPGA 4x 500 MHz or 2x 1 GHz per board 1024 channels power consumption 25W per board Ethernet interface SuperCam spectrometer initially uses 8 identical boards for 64 x 250 MHz operation Allan time (with IF processor) ~250 s IEEE mm-wave and THz Workshop
LO Multiplexing 64-Way Waveguide Power Divider
CNC Metal Micromachining 350 GHz TWT 650 GHz Sideband Separating Mixer
Technological Challenges and Proposed Solutions for Even Larger Coherent Arrays 1. Mechanical and Electrical Complexity Solution: Use 2D integration to simplify design 2. Detector yield and focal plane assembly Solution: Use simple, robust SIS device design with self aligning SOI chips 3. Heat load from LNAs becomes dominant with large pixel count Solution: Develop integrated ultra-low power dissipation LNAs 4. Economical and fast WG and feed fabrication Solution: use drillable feeds (e.g. Leech et al, this conference), CNC micromachining for WG. 5. RF and DC interconnects and wire count Solution: Develop multi-conductor cryo-ribbon interconnects 6. Magnetic field for SIS devices Solution: Use engineered permanent magnets to replace individually adjustable electromagnets
KAPPA FPU Concept 6 mm Waveguide SiGe MMIC Bonding Pad Via to surface mount G3PO connector IF Microstrip
KAPPA SiGe LNA 500 um 800 um SiGe MMIC Chip 0.5-4.5 GHz 16 db Gain 7K noise temperature (predicted) 9K (measured) 2 mw power dissipation S11 <-11 db On-chip bias tee Small chip size 2 stage version also fabricated (32 db gain, 4-5 mw power dissipation)
LNA Integration
IF Flex Circuit 16 channel stripline flex circuit with microstrip terminations. 1m circuit has ~4 db (simulated) loss at 4.5 GHz. Heat load with 9 micron thick copper conductors: ~5 mw (heat sunk at 15K, 150mm length. Heat load could be further reduced with cryogenically compatible metal replacing copper (i.e. phosphor-bronze). Rogers Ultralam 3000 flexible LCP dielectric: 200um total thickness 10mm wide 18mm wide Stripline to microstrip transition