Direct Detectors for FIR/submm Light: Overview of Existing & Emerging Concepts Albrecht Poglitsch FIR Detectors 1
Outline Direct FIR/submm Detectors in an Astronomical Context Photon Detectors Basics Semiconductor Bulk Photoconductors Photodiodes BIB Detectors STJ SQPC Readout electronics / multiplexers Bolometric Detectors Basics Thermometers (semiconducting, superconducting) Absorber coupled bolometers Antenna coupled bolometers Filled arrays vs. single mode horns Bolometers without thermal noise? Readout / multiplexers FIR Detectors 2
FIR Detectors for Present/Future Observatories Observatories Ground based mm/submm telescopes SOFIA Future FIR space observatories (SAFIR, FIR Interferometers) Instruments Imaging broadband photometers/polarimeters Medium resolution, wideband and/or imaging spectrometers Boundary conditions for detectors Very large span in power/pixel between photometry with warm, ground based telescope and spectroscopy with cryogenic space telescope BLIP taken for granted Speed and stability requirements depend on application Large format arrays (observing efficiency) multiplexing One device unlikely to be optimum for all needs FIR Detectors 3
Limiting Sensitivities and Detector Needs Noise = 10 photons (at 200 µm) after ½ s integration FIR Detectors 4
Side Remark: Coherent vs. Direct Detection FIR Detectors 5
Photon Detector Basics Ideal direct detection of light: Convert each photon reaching the detector into an electrical signal (before energy gets thermalized) and count Noise determined by fundamental limit of photon statistics Today s approximation in FIR detectors: Convert a fraction of the photons (quantum efficiency) into charge carriers and integrate charge Detector may add extra noise Analog readout electronics adds noise / has limited dynamic range and resolution FIR Detectors 6
Semiconductor Bulk Photoconductor hν>e g E c hν>e i E c E d hν>e i D D + D D D A A A A A E c E a E v E v E v intrinsic n-type extrinsic p-type Photon absorption generates mobile charge carriers Intrinsic PC: electron hole pair Extrinsic PC: electron or hole (+ ionized donor/acceptor) External electric field causes drift of carriers Measured as photocurrent in contact Photoconductive gain strongly bias-dependent impact on readout Recombination after mean life time τ / drift length l Recombination is reverse process of generation same statistics Fundamental noise limit 2 x photon noise (g-r noise) FIR Detectors 7
FIR Photoconductor Characteristics Extrinsic photoconductors only practical devices so far demonstrated NEP of ~ 10-18 W/ Hz at low background sufficiently low dark current at non-demanding T (1.8 K) low doping concentration essential low absorption arrays assembled from individual, ~mm 3 size Ge detectors, limiting maximum array size to ~10 3 pixels beyond 120 µm complex mechanical stress mechanism necessary, maximum wavelength ~205 µm performance degradation from cosmic rays problematic transient response at low background potential of bulk devices has probably been used up Intrinsic, e.g. Hg 1-x Cd x Te, photoconductors are difficult to control for the small bandgaps needed Superlattice materials, where layer thickness rather than alloy composition determines bandgap, may be promising FIR Detectors 8
Ge:Ga Photoconductor Arrays MPIS 32 x 32 unstressed Ge:Ga array FIFI LS 16 x 25 stressed Ge:Ga array FIR Detectors 9
Photoconductor Transient Response and Memory Effects PC has two time constants (simplest model), light-level dependent τ g-r Cosmic rays can affect deeper states than FIR light and lead to responsivity change Calibration Curing schemes FIR Detectors 10
"Bandgap-Engineered" Semiconductors Bandgap of Hg1- x Cd x Te alloys vs. temperature for various Cd fractions x A superlattice (SL) is a composite semiconductor consisting of a large number of alternating well and barrier layers hence multiple quantum wells FIR Detectors 11
FIR Photodiodes? internal field (space charge in p-n transition) external field (reverse bias) Diode: space charge of p-n transition prevents transport of majority carriers Electron-hole pairs generated in or near transition by photon absorption separated by E-field bias independent, G=1 no g-r noise If semiconductor with sufficiently small bandgap (alloy or superlattice) becomes practical, photodiodes will be possible FIR Detectors 12
Blocked Impurity Band Detectors Advantages of BIB device: Small volume reduced cosmic ray susceptibility No g-r noise (diode-like) 2 gain in sensitivity over bulk device Monolithic, planar arrays large formats feasible Challenge: doping concentration profile (Ge:Ga BIB still no competitive QE) IR active layer 10 µm 4 µm metal contact implanted contact blocking layer Position (microns) FIR Detectors 13 Electric Field (V/cm) 140 120 100 80 60 40 20 0-20 -40-60 5 x 10 12 cm -3 3 x 10 13 cm -3 1 x 10 14 cm -3 E y VB x U conduction band impurity conduction band 0 2 4 6 8 10 12 14 16
Read-out and Multiplexing Large-format detector arrays require cryogenic multiplexed readout electronics Fabrication of semiconductors devices for operation below 10... 20 K (carrier freeze-out) difficult Basic design depends on detector type Detectors with weak bias dependency (photodiodes) can use passive integration (e.g. into FET input capacitance) and source follower / switching multiplexer scheme Detectors with strong bias dependency (photoconductors) have to use active integrator (CTIA) circuit for each pixel + multiplexer Photoconductor readout is more complex than photodiode readout and more likely to introduce readout noise FIR Detectors 14
Passive Integration Multiplexer For bias-independent detectors, the passive integration scheme is simple affected by input voltage noise only after integration (read noise) Cryogenic MOS source follower standard Input capacitance used as integration capacitor Non-destructive, multiple read-out up the ramp V bias MUX Out Scheme used for NIR arrays with read noise ~ few electrons FIR Detectors 15
C f CTIA: Readout Scheme for PCs MUX Out In C AC -A C S/H -V bias CTIA architecture Capacitively fed-back highgain amplifier keeps input node at constant potential constant detector bias Photocurrent is integrated in feedback capacitor Amplifier input noise (both voltage & current noise density) affects integrated signal SBRC 190 FIR Detectors 16
Superconducting Tunnel Junction Single Quasiparticle Photon Counter (a) SQPC: Absorption of radiation, coupled in by an niobium antenna, breaks Cooper pairs in an aluminum strip and gives a current pulse through a tunnel junction connected to a RF-SET. (b) RF-SET: an input gate signal changes both SET output impedance and amount of rf carrier power reflected by a resonant LC-circuit. Reflected power is amplified by a HEMT for detection at room temperature. Multiple RF-SETs with different resonant frequencies can share the same HEMT. FIR Detectors 17
STJ SQPC (cont.) Predicted NEP @ 300 GHz: 10-19 W/ Hz (limited by noise of bias resistor) Dynamic range ~ R = 1... 1000 on cryogenic space telescope FIR Detectors 18
NEP (W/ Hz) IR Group Science Retreat Ringberg, Nov 2002 Bolometer Basics Convert photon energy into heat that 10-15 NEP at optimal bias vs optical power Red band //210 µm Si beam raises the temperature of the sensing element Some form of thermometer required to measure this temperature rise To maintain operating temperature, sensing element is connected through 10-16 (weak) link to thermal photon bath thermal (phonon) noise, MOS requiring readout very low operating temperaturejohnson Current in (resistive) thermometer introduces (T-dependent) phonon load electrothermal feedback, affecting 10 speed/responsivity. Lower limit for -17 optical power (W) x 10 current by thermistor Johnson noise -12 1 2 3 4 5 6 7 8 9 10 T=T 0 + T FIR Detectors 19 C P G R(T) T 0 U, I Responsivity: S= U/ P=IαR/G with α=1/r dr/dt Speed: τ=c/g Noise mechanisms: conventional bolometer photon background noise phonon noise: NEP= 4kT 2 G Johnson noise: NEP= 4kTR /S readout noise
New Variation on Theme: Integrating Bolometer Superconducting heat switch makes extremely weak link to bath. If brought to normal by magnetic field, conduction goes up by large factor (10 4 for Al at 100 mk) During integration very little phonon noise, then rapid reset. Correlated double sampling eliminates switching noise In combination with suitable thermometer, NEP < 10-20 W/ Hz not precluded FIR Detectors 20
Bolometer Basics (cont.) Large variety of bolometer implementations comes from combination of: Two main technologies for thermometers semiconductive thermistor α<0, current biased superconductive oresistive: TES α>0, voltage biased; α very large o non-dissipative: kinetic inductance, magnetic penetration Two schemes of coupling to and dissipation of electromagnetic radiation Combined impedance matching and dissipation: absorber coupled bolometers this easily allows filled arrays Separate coupling and dissipation: antenna coupled bolometers antenna couples radiation into waveguide where it propagates to absorber (resistor) for dissipation, this allows beam shaping / straylight suppression + filtering FIR Detectors 21
Semiconductive thermistor Thermometers α<0, current biased measured quantity is voltage Ge (neutron transmutation doped) o Advantages: high α, very low 1/f knee (~0.02 Hz) odisadvantage:discrete bulk element, to be mounted/contacted on bolometer structure Si (ion implantation, diffusion, mesa,...) Ge (NTD) o Advantages: can be integrated in lithographic production process, allow very high impedances compatible with MOS noise o Disadvantages: not fully competitive in terms of 1/f onset and α FIR Detectors 22
Thermometers (cont.) Superconductor transition edge sensor (TES) Resistive, α > 0, voltage biased measured quantity is current, requiring special readout (in practice, SQUID) Extremely large α possible with proper design (homogeneous material + T, achieved by S/N sandwich) strong electrothermal feedback can decrease thermal τ by factor >100 or improve NEP phonon by factor >10 R TES Electrothermal feedback: TES can be thought of as a thermostat: best operating range (empirical) For constant U bias, R will adjust itself to keep the TES at T c by keeping P photon + U I = const I is directly proportional to P photon T C T FIR Detectors 23
Non-dissipative: kinetic inductance Thermometers (cont.) Surface impedance of superconductor is Figure (almost) entirely inductive (no DC resistance, but inertia of cooper pairs which have to be accelerated by AC field) Presence of quasiparticles (broken Cooper pairs) blocks some states for Cooper pairs and changes both parts of impedance Quasiparticle density grows exponentially with T (thermometer), but Cooper pairs can also be broken by photons with hν>2 photon detector FIR Detectors 24
Kinetic inductance Differential reactance and resistance follow quasiparticle density Quality factor Q=X/R grows exponentially with decreasing T T-measurement with microwave resonator: T-change leads to phase change in transmitted signal Different pixels can have resonators with different frequency (multiplexing) Thermometers (cont.) FIR Detectors 25
Thermometers (cont.) Non-dissipative: magnetic penetration Meissner-Ochsenfeld effect: no magnetic field in (bulk) superconductor Thin film: B(x) = B 0 exp(-x/λ), λ=f(t) Constant B-field, e.g. stored in superconducting drive coil on one side of thin film Pick-up coil on other side of film has to maintain magnetic flux by adjusting current accordingly Current in pick-up coil measured with SQUID Lossless, wireless thermometer! Superconducting thin film FIR Detectors 26 N S Magnetometer (SQUID) B 0 B(T)
Absorber Bolometer: Resistive Film Thin film of e.g. Bi, matching free-space impedance, deposited on micromachined membrane, can be combined with λ/4 cavity (backreflector) for better efficiency FIR optical properties of 400 Ω/sq Bi/SiO bilayer on 1 µm SiN membrane FIR Detectors 27
Absorber Bolometer: Resistive Film (cont.) Filled array achieved by pop-up design which moves readout into 3rd dimension, linear modules densly stacked 1 µ PUDs Thick silicon frame mushroom bolometer: bolometer/readout sandwich Fabricated mainly by photolithography Load resistors mounted here Mushroom bolometer.065 K heat sink Works with either integrated Si thermistor or TES sensor Silicon Bridges 1.3 K heat sink Electrical pins Pop-up Device FIR Detectors 28
Absorber Bolometer: Resonant Grid Filled array Periodic absorber pattern + λ/4 cavity (backreflector) Use of compatible metal absorbers (superconductive, to reduce the heat capacity) Completely fabricated by photolithography Readout electronics indium-bump bonded to back of bolometer array to allow large size, butted arrays CMOS buffer/multiplexer (for semiconductive thermistors) Concept can be modified for TES Variation on theme: spiderweb bolometer same idea, but optimized for feed horn coupling FIR Detectors 29
Interconnection circuit CEA/LETI Design 0.3K >50 µm 400 µm Detection Layer thermometers silicon grid + absorber reflector Indium bumps indium bumps achieve, at the same time, the λ/4 cavity depth and electrical link between sensors and readout electronics FIR Detectors 30
CEA/LETI Design (cont.) Reference thermometer Pixel Pixel thermometer r Absorbing grid Silicon rod Thermistor impedance ~ 10 10 Ω double SOI wafers 750 µm -2) MESA diffused thermometers. P:B FIR Detectors 31
FIR Detectors 32
Antenna Coupled Bolometers Single-moded horn antenna Heritage from coherent techniques Feedhorns where adjacent horn is used FWHM to couple beams one on the in the focal (Gaussian) plane mode into sky waveguide don t overlap Used in conjunction with e.g. spiderweb-absorber type bolometers in cavity for increased absorption efficiency and clean beam definition (no side lobes, no straylight from warmer parts of instrument) Good coupling requires ~2fλ diam. in focal plane which causes 2 FWHM beam spacing on sky Jiggling/special scanning for full beam sampling needed Beam FWHM λ/d Beam separation 2λ/D 16 pointings needed for fully-sampled image SPIRE PLW Array FIR Detectors 33
Filled Absorber vs. Horn Arrays Which gives better observing efficiency? It depends! Assumption: perfectly baffled system, only proper light from telescope aperture reaches focal plane (easier for horn!) Observation of point source with known position (+ no pointing errors): Gaussian horn receives one mode (signal + background) and couples ~80% of source power to detector. Filled array (fully sampling, with θaccordingly lower background) has to apply spatial filtering to signals from pixels around source position for best (S/N) power measurement. Gaussian horn wins by factor 1.2 in S/N Observation of point source with position/pointing accurate within region (λ/d) 2 : Gaussian horn needs jiggle (e.g. 7 positions around nominal). Filled array has to apply same spatial filtering to data, but with sliding center position. Filled array wins by factor 1.2 in S/N FIR Detectors 34
Filled Absorber vs. Horn Arrays (cont.) Mapping of extended area for point source extraction: (Further assumption: same focal plane size in both cases) Gaussian horn array, hexagonally packed, spaced by ~2 f# λ in focal plane effective focal plane ~0.3 For otherwise ~ comparable observing efficiency (see above), filled array gains factor of 3.4 in observing time! FIR Detectors 35
Antenna Coupled Bolometers (cont.) Planar antenna structures Commonly used in coherent receivers Multiple slot antennae + microstripline seems favoured concept for bolometers Superconducting structures set upper limit in frequency Dual slot microstrip antenna, which is cut into a superconducting niobium ground plane, coupled to a TES. The electric field from the microstrip antenna is propagated along planar transmission lines. These microstrip lines consist of a niobium superconducting lead separated from the ground plane by 2000Ǻ of SiO dielectric. The electric fields from each antenna are passively added and terminated at a thin film Au resistor located on a thermally isolated silicon nitride island. On the same island, an Al/Ti/Au TES film responds to the temperature rise of the resistor. FIR Detectors 36
Antenna Coupled Bolometers (cont.) Planar antenna structures Stripline allows integrated filter structures for e.g. multiband photometers, realized in one focal plane array FIR Detectors 37
Antenna Coupled Bolometers (cont.) Planar antenna structures Phased arrays (passively) of slot antennae allow beam shaping and potentially a good compromise between absorber and classical horn 4x3 phased array pixel (dual polarization) conceptual drawing Beam map and one row of 15x16 slot antenna array FIR Detectors 38
Readout and Multiplexing Current-biased bolometers need voltage readout Thermistors usually reasonably to very high impedance ontd Ge ~ 10MΩ, low enough for readout within ~10 cm. JFETs, heated to ~100K, have very low voltage noise and very low 1/f knee; used in most sensitive, really existing Low Frequency Noise Stability 10 bolometers. -5 No multiplexing at K or sub-k level possible, + hefty dissipation/ conduction from 100 K readout to cryogenic environment limits number of bolometers severely. FIR Detectors 39 10-6 10-7 10-8 20 nv Hz -1/2 20 nv Hz -1/2 20 nv Hz -1/2 0.001 0.01 0.1 1 10 Freq [Hz]
Readout and Multiplexing (cont.) Current-biased bolometers need voltage readout Thermistors usually reasonably to very high impedance o Si thermistors can be made with very high impedance (10 GΩ) for large voltage responsivity to become compatible with much higher voltage noise of cryogenic MOS FETs. This requires but also allows operation of readout within ~1 mm from sensor to keep RC low enough and prevent pickup. MOS technology allows complex circuits multiplexing. Also allows schemes for suppression of intrinsically too high 1/f knee. FIR Detectors 40
Differential Readout Scheme 300 mk VH RC RB VL VH VCH RC Rch RB VL 300 mk 300 mk VRL VDL VGL1 SEL2 SEL1 active pixels blind pixels SEL1 SEL2 VGL1 VDL VRL 300 mk SEL3 SEL3 Thermistor bridge for 2 K VDL_BU VDDeach pixel VGG1 VGL_BU1 Electronic chopping S1 between bridge and Buiffer Unit reference voltage to eliminate buffer 1/f noise Use of blind pixels to Buiffer reduce Unit influence of thermal drifts and EMI on signal lines S1 VGL_BU1 VGG1 VDD Achieved performance: 2 K VDL_BU NEP ~10-16 W/ Hz @ 7 Hz bandwidth FIR Detectors 41
Readout and Multiplexing (cont.) Voltage-biased bolometers need current readout TES sensor has very low impedance; current needs to be measured at 0 voltage drop Presently, SQUID only realistic approach: TES current flows through s.c. coil, magnetic field measured with SQUID TES = I(T) B(T) I B V(T) V B SQUID SQUID is periodic device and requires somewhat complex control/readout electronics for operation FIR Detectors 42
High SQUID noise margin allows timedomain multiplexing Activation of individual pixel SQUIDS by bias through row select Amplification by serial SQUID array in each column Rigorous magnetic shielding mandatory! Readout and Multiplexing (cont.) FIR Detectors 43