Final Report Summary - : E-SQUID (Development of SQUID-based multiplexers for large Infrared-to-X-ray imaging detector arrays in astronomical research from space)
Executive Summary:
One of the main questions of science is the early phases of the evolution of our Universe and the evolutionary scenarios after the Big Bang. The signature of this evolution is imprinted in the electromagnetic radiation from the deep space. A detailed visibility to the first billion years in the evolution requires instruments capable of imaging and spectroscopy with an extremely high sensitivity, since the science requires ultimately precise measurements of radiation ranging from X-rays originated in the vicinity of the super-massive black holes in the nuclei of the first galaxies to infrared radiation by the cool gas in the interstellar space up to distances of more than 10 billion light-years.
In that context, E-SQUID project drives progress in the development of versatile multiplexed SQUID (Superconducting QUantum Interference Device) readout of Cryogenic Detectors with the main exploitation focus in space instruments, with also significant potential for ground-based applications. SQUIDs can be designed for optimized readout of X-ray micro-calorimeter arrays, as well as similar detectors for longer wavelengths extending to far infrared/sub-mm wavelengths. The most competitive technology in this area is presently available from non-European sources. A core objective of E-SQUID is to upgrade the performance of European solutions to be competitive with the best achievements in the world.
The practical task of the E-SQUID project was to develop prototypes for next generation multiplexed SQUID readout schemes of X-ray and far-infrared Cryogenic Detector arrays, modifications of which may then be used for optical wavelengths. The improvements were performed in several steps, ensuring the validity of each improvement by tests and performance characterization, and moving to next phase using improved guidelines provided by the test results. The goal is to find an optimal scalable multiplexing method for each wavelength that enables future design and developing of large detector arrays with minimum noise characteristics.
Imaging detectors that are based on Transition Edge Sensors (TES) form the core of future space-based telescopes for X-ray and infrared astronomy. SQUID technology is essential to obtain detector noise limited readout performance of these TES based detectors. Mass, complexity and power consumption constraints in space-based instruments form the main driving force behind the development of efficient multiplexing schemes.
In the E-SQUID project new SQUIDs have been developed for use to read individual detectors, as well as new improved SQUID multiplexers for reading multi-pixel detector arrays. From the possible alternative multiplexing schemes, frequency domain multiplying (FDM) turned out the most promising concept for space applications. The usefulness of code domain multiplexing (CDM) could not be demonstrated in E-SQUID project, but the results indicate that further work on this may lead to success as well.
It was found that the developed SQUID multiplexers are well suited for the required dimensions of the TES detector arrays for two future space telescopes: the XIFU instrument on ATHENA X-ray telescope, and the SAFARI instrument on SPICA infrared telescope. The use for future instruments of ALMA is also possible.
The technology developed in E-SQUID can also be utilized for the development of a number of important ground-based applications, such as passive imaging of concealed weapons, and measurements of fragile biological samples and processes, such as the interaction of biological or chemical agents with living cells.
Project Context and Objectives:
A major task for modern astronomy is the examination of the early Universe and the evolutionary scenarios after the Big Bang. The signature of this evolution is imprinted in the electromagnetic radiation from the deep space, including the ubiquitous cosmic microwave background and the faint glow from stars and galaxies. A detailed visibility of this multi-billion-year tale requires instruments capable of imaging and spectroscopy with an extreme level of sensitivity, since science requires ultimately precise measurements of radiation ranging from X-rays originated in the vicinity of the super-massive black holes in the nuclei of the first galaxies to infrared radiation by the cool gas in the interstellar space up to distances of more than 10 billion light-years.
In that context, E-SQUID project drives progress in the development of versatile multiplexed SQUID (Superconducting QUantum Interference Device) readout of Cryogenic Detectors with the main exploitation focus in space instruments, with also significant potential for ground-based applications. SQUIDs can be designed for optimized readout of X-ray micro-calorimeter arrays, as well as similar detectors for longer wavelengths extending to far infrared/sub-mm wavelengths. The most competitive technology in this area is presently available from non-European sources.
The high level objective of E-SQUID is to upgrade the performance of European solutions to be competitive with the best achievements in the world.
This technology is already used in modern ground-based astronomical instruments such as SCUBA-2 at James Clerk Maxwell Telescope (Hawaii). The performance and array sizes of presently available solutions are, however, not sufficient for the scientific aims of future instruments like the X-ray micro-calorimeter for ATHENA (L2 mission of the ESA), the infrared space mission SPICA (JAXA/ESA), and next generation ground-based far infrared experiments for ALMA (Atacama Large Millimeter Array). Hence, the development of improved Cryogenic Detector systems with larger arrays and smaller noise are very important to the future of European space science. In addition to space science, the technology developed for E-SQUID can be utilized for the development of a number of important ground-based applications, such as passive imaging of concealed weapons, and measurements of fragile biological samples and processes, such as the interaction of biological or chemical agents with living cells.
For setting up the scientific requirements for the photon detectors of each energy range, we have selected the following reference instruments. They are APEX telescope (Chajnantor/Chile) at far infrared/Submm, S-Cam 3 (La Palma/Canary Islands) at optical, and the ATHENA (space) at X-ray range. With the aimed performance, the reference instruments can provide observations of the cold interstellar dust enabling to trace the history of star formation (infrared), multicolour observation capability of the exoplanet transits across the parent stars (optical), and black hole accretion disk observations providing information on space-time structure of the black hole event horizon (X-rays). In parallel, the requirements serve also the more general aim, which is to develop larger ultra-sensitive Cryogenic Detector arrays with smaller noise for sensors that can be used for space astronomy and ground-based applications.
The practical task of the E-SQUID project is to develop prototypes for next generation multiplexed SQUID readout schemes of X-ray and far-infrared Cryogenic Detector arrays, modifications of which may then be used for optical wavelengths. The improvements are performed in several steps, ensuring the validity of each improvement by tests and performance characterization, and moving to next phase using improved guidelines provided by the test results. The goal is to find an optimal scalable multiplexing method for each wavelength that enables future design and developing of large detector arrays with minimum noise characteristics. The signal multiplexing means a reduced amount of signal lines between the Cryogenic Detectors and room temperature main electronics, which enables, in practice, the use of considerably bigger detector arrays, and also reduces electrical disturbances and heat load.
The power dissipation of cryogenic detectors is typically extremely small. An example is the 6 pW / pixel power dissipation for the breadboard TESs for the International X-ray Observatory (Athena) . The biggest power input, and heat leakage, to the lowest temperature part of the detector is via the electrical harness to drive and read-out the detectors. Multiplexing techniques are used to minimize the required number of wires for reading large arrays, and thus their development is a crucial task.
The only read-out amplifier that has so far been proven to be suitable for the above approach is a SQUID-based amplifier next to the detector.
Therefore, our first objective is to use our technological skills to develop cryogenic SQUID amplifiers whose figure-of-merit is improved to bring their performance closer to the thermodynamic limit.
Moreover, to realize an effective multiplexing scheme, a sizable number of these first stage amplifiers need to be integrated into a single output channel, using an appropriate addressing method i.e. time (TDM), digital (CDM) or frequency (FDM) encoding.
Therefore, the second objective of E-SQUID project is to select the multiplexing method which is most suitable for the chosen approach of SQUID readout of cryogenic detectors. The selection will be made on the basis of considerations regarding achievable multiplexing level (e.g. number of merged channels), technology readiness level (TRL) and complexity of adjacent electronics.
Depending on that selection, a full set of specifications regarding the setup of various cryogenic detector arrays can be derived. It seems possible to build up a universal architecture based on these specifications, using a sophisticated SQUID layout and adapted room temperature electronics.
Therefore, the third objective of E-SQUID projectl is to develop both cryogenic and room-temperature sections of a complete multiplexing electronics unit, which we aim to standardize across the two major European superconductive foundries, IPHT Jena and VTT.
Project Results:
Concepts and Readout Methods of detectors
Space X-Ray and IR missions
Introduction
Imaging detectors that are based on Transition Edge Sensors (TES) form the core of future space-based telescopes for X-ray and infrared astronomy. SQUID technology is essential to obtain detector noise limited readout performance of these TES based detectors. Mass, complexity and power consumption constraints in space-based instruments form the main driving force behind the development of efficient multiplexing schemes.
As the prototypical space missions we have taken as the project guidelines, are the X-IFU instrument for the ATHENA+ mission and the SAFARI instrument for the SPICA mission. The planned ATHENA+ mission is an orbiting X-ray telescope, whose imaging capability comes from silicon pore x-ray optics with ~2 m2 collecting area. The X-IFU instrument will perform energy dispersive single photon counting, i.e. capture each incident x-ray photon in the 0.3 – 12 keV range with 70% quantum efficiency at order-of 100 cps count rate, and resolve energy of each photon with order-of 2.5 eV resolution. The planned SPICA mission is an orbiting far-infrared telescope, with a 3.5-metre antenna dish, actively cooled to 5 K temperature. The SAFARI instrument is effectively a ‘crystal set’ radio receiver for 34 – 210 m wavelengths, equipped with Fourier Transform Spectrometer and imaging capability. The SAFARI receiver sensitivity is going to be limited by the photon noise of the 2.7 K blackbody background, which implies order-of 2×10-19 W/Hz noise-equivalent power (NEP).
Although missions seem very different, the same detector technology can be used in both, namely absorption of the radiation and measuring the temperature change in the absorber. The sensitivity requirements dictate operation of detectors at a very low temperature, order-of 50 mK, and a very sensitive thermometer, for which a superconducting phase transition is used. In the SAFARI instrument the detectors are operated in bolometric mode, meaning that they react to the average incident radiative power, whereas in the X-IFU instrument they are operated in calorimetric mode, i.e. each incident photon causes a temperature excursion. As a consequence, bolometers of the SAFARI instrument always operate close to thermal equilibrium, implying that the work well with Frequency Domain Multiplexing (FDM, see below), whereas calorimeters of the X-IFU instrument spend a large fraction of their time out-of equilibrium. This difference has consequences on the multiplexing method.
Operating principle TES-based detectors
Imaging TES-based detector arrays consist of a matrix of pixels. A pixel consists of a thermally isolated island equipped with an absorber, and a thermometer. When the optical signal from the source gets absorbed in the absorber, its energy is transferred into heat. The transition edge sensor senses the resulting increase in temperature, which is a superconductor in its superconducting phase transition. Electro thermal feedback stabilizes the detector in an operating point in the transition.
During operation, the detector is in a dynamic equilibrium, i.e. its temperature is stabilized above the thermal bath temperature, which is set by the critical temperature Tc of the TES by dissipation in its resistance. As a result, a constant heating power is needed to keep the pixel in its operating point. When an optical signal gets absorbed, less heating current is needed to maintain the temperature. The latter change in heating current forms the signal.
In the microcalorimeter operating mode, the detector is dimensioned such that it responds to individual photons, which makes the detector energy dispersive. In the bolometer operation mode, the detector is too slow to deconvolve individual photon absorption events. As a result, in this mode the detector is not energy dispersive, but only measures the average power of the applied light.
Multiplexed readout of TES-based detectors
Multiplexed readout as opposed to direct readout of TES-based cryogenic microcalorimeters and bolometers is mainly driven by the goal of minimizing the required cooling capacity and complexity at the cold stages. The main sources of heat load in the readout system are SQUID current amplifiers, and the heat leak through the bias and readout wiring.
Without any multiplexing, each pixel would require a single wire pair for its voltage bias, and three wire pairs for the SQUID trans-impedance amplifier. Multiplexed readout implies that multiple pixels are readout by a single SQUID amplifier, and multiple TES bias voltages are fed through one wire pair.
Rationale for SQUID readout
TES-based detector readout can be seen as monitoring the resistance of a temperature dependent resistor, which value is a function of the applied signal power. In order to benefit from electro-thermal feedback, the monitoring is done by applying a constant voltage, and reading the temperature-dependent current.
To comply with the required voltage bias condition, a current amplifier is needed with an effective input resistance that is significantly smaller that the set point resistance (typically R0=40mΩ) of the TES. In addition to that, electro-thermal stability requires that the electrical inertia of the bias circuit should be approximately a factor of 6 smaller than the thermal inertia. This limits the maximum input reactance of the current amplifier at the bolometer bandwidth to 6* R0 .
Because of the stringent science requirements, it has been chosen that the readout circuit is allowed to consume ~10% of the NEP budget. This implies that the required effective noise temperature of the current amplifier should be at least one third of the effective noise temperature of the bolometer, taking into account that the detector noise and the amplifier noise are uncorrelated. The latter noise temperature is approximately equal to the operating temperature, i.e. ~100mK. Hence, the effective noise temperature of the current amplifier has to be smaller than ~30mK. This requirement, combined with the input impedance requirement, rules out the use of semiconductor amplifiers, and leave SQUID current amplifier readout as the only viable option.
Multiplexing of TES-based bolometers
Multiplexing of TES-based bolometers implies using the wire pairs and SQUID amplifiers of a single bolometer for the independent readout a larger number of bolometers. Consequently, the bolometer signals need to be added on a single wire pair, on which then several bolometer signals are transmitted simultaneously. Note that simple addition of the signals is not acceptable, as the signals overlap in time, frequency and phase space, and would therefore become indistinguishable.
Because of the stringent science requirements, there is virtually no headroom in the NEP budget to assign to information loss as a result of multiplexing. As a consequence of that, the multiplexing scheme has essentially to be lossless, and cross talk-less. In text books it can readily be found that any lossless multiplexing scheme, i.e. a scheme in which the constituent signals need to remain fully recoverable, requires two fundamental steps before the signals can be added while remaining (mathematically) independent. The first step is to confine the signal bandwidth, and the second step is to transform the signals (“fingerprinting”) in such a way that they become mathematically independent. The latter mathematical transformation implies the multiplication of the time-dependent signal functions of the pixels with an independent carrier function per pixel, in such a way that they become orthogonal and therefore mathematical independent and fully distinguishable.
As a result of the mathematical transformation (i.e. the “multiplexing”), N multiplexed signals require a bandwidth which is at least N times the bandwidth of a single signal, depending on the chosen set of carrier functions. As will become clear below, SQUID amplifiers can provide sufficient power gain over bandwidths of several MHz On the other hand, TES-based detector pixels require a bandwidth of the order of 1 kHz per pixel. Therefore, from bandwidth point of view, SQUID amplifiers are in principle capable of accommodating the signals of at least 102 to 103 TES-based pixels.
Code domain multiplexing in TES-based detector arrays
The low operating temperature implies that extremely small power dissipation can be tolerated. Every nanowatt dissipated in the 50 mK stage of the refrigerator must be lifted to higher temperature where the radiator of the spacecraft reside. This means that every dissipated nanowatt must be multiplied at least by Carnot efficiency, in practice by the much poorer actual efficiency of the refrigerator, to obtain the heat generation at the S/C radiator. This is a particularly stringent constraint in space, where cooling water is not available, but the heat must be radiated away.
Two additional problems – Shannon capacity and need for multiplexing - arise from the fact that both SAFARI and X-IFU are imaging instruments. SAFARI will contain 61 x 61 detector pixels in 34-60 μm band, 34 x 34 pixels in 60-110 μm band, and 18 x 18 pixels in 110-210 μm band, whereas X-IFU is planned to contain 3840 X-ray detector pixels.
Shannon channel capacity indicates the amount of information flowing through the amplification and data acquisition chain, the capacity equalling signal bandwidth times the dynamic range, and being expressed as bits per second. If one X-IFU pixel generates 500 kbit/s information flow, the full 3840-pixel instrument will generate 2.6 Gbit/s, assuming the flow cannot be compressed. Although it is not known whether the connection is fundamental – in analogy with the Landauer’s principle – high Shannon capacity seems in practice to go hand-in-hand with increased power dissipation in amplifiers and with increased heat leakage through cabling.
Multiplexing implies that signals from several pixels are encoded and summed into a single cable. It would be impractical eg. in X-IFU to have 3840 dedicated cables from the 50 mK refrigerator stage to room temperature, and a dedicated digitizer for each pixel. Multiplexing entails an orthogonal set of encoding functions, with each signal being multiplied by one member function, before the signal are summed together. Three basis sets, sinusoids for Frequency Domain Multiplexing (FDM), boxcar functions for Time Domain Multiplexing (TDM) and Hadamard codes for Code Domain Multiplexing (CDM), offer certain technical advantages and have been pursued in the past. One then needs a cryogenic multiplying element, capable of multiplying the detector signal with a basis function (Figure 4).
In FDM, the detectors can be biased by the sinusoids forming the basis set, so that Ω’s law performs the effective multiplication of the bias voltage with the detector conductance. This works well in SAFARI-like bolometers, which remain internally close to thermal equilibrium and perform the multiplication smoothly. X-IFU –like calorimeters, in the other hand, had been observed to suffer from degradation of the energy resolution before the years 2009 when the E-SQUID project was planned. The degradation was suspected to be of fundamental origin. Hence, the CDM was originally targeted in the E-SQUID project, with the emphasis on implementation of the Hadamard-encoding elements.
As Hadamard functions are two-level, a multiplication by +1 and -1 is sufficient in CDM. This can be implemented by a commutating switch. We have studied such switches implemented with Josephson junctions, in which case the periodic switch response due to quantum interference brings in an additional advantage: replication of so-called Sylvester’s construction of Hadamard matrices in hardware (Figure 5). This opens a way to address N pixels with log2N address lines.
SQUID-like flux-controlled two-junction current steering switches (CSSw’s) require a very clean addressing current, because fluctuations in address current couple as noise to the detector signal. We have instead studied CSSw’s based on three-junction Zappe interferometers (Figure 6). They have almost square-shaped flux response, as opposed to sine-like response of two-junction SQUIDs, hence the address current fluctuations have a much lower effect.
Resistive CSSw’s require the low-value resistor Rq (Figure 6 b) which are (i) hard to implement and (ii) suspected to cause problems with SRON calorimeters. Tens of nanovolts of voltage drop can be tolerated across the switch, which implies Josephson oscillation at tens of MHz, which is disturbingly close to the information band of the calorimeters. Therefore, we have studied inductive CSSw’s which always remain in the superconductive state (Figure 6 c,d).
Dynamic range of current-steering switches easily becomes the dominating bottleneck. There are several effects, one of which is: a large current handling capacity implies a small loop inductance, which makes loop dimensions to approach the Josephson junction diameter, which subsequently creates a Fraunhofer envelope (Figure 8), in analogy with the optical double-slit interferometric effect. When periodicity is lost, the log2N addressing is no longer feasible.
There is the alternative CDM method, the so-called flux-summing technique (Figure 7). This is a moderately interesting technique, however having two disadvantages: (i) only sqrt(N) –proportional reduction in the number of cables can be reached easily, and (ii) each readout SQUID must handle an N-fold dynamic range, because signals from all N pixels are summed at SQUID inputs.
Due to these observations, and encouraged the experimental and theoretical results from SRON in late 2013 that FDM may still be feasible with calorimeters without degraded energy resolution, project emphasis was shifted towards FDM.
Frequency Domain Multiplexing in TES-base detector arrays
Basic operating principles of FDM
As already mentioned above, multiplexing involves mounting a bandwidth-limited signal on a carrier function by means of multiplication (modulation), followed by summing of the modulated signals for transportation through a shared channel. In the SRON FDM implementation, the orthogonal set of carrier functions consist of sinusoidal TES bias voltages of different frequencies. The signal-dependent TES resistance amplitude-modulates the resulting bias current. The required bandwidth limitation, and separation in frequency space is provided by LC band pass filters, which consist of inductors (L) and capacitors (C). A schematic diagram is shown in Figure 9. The filtered signals are summed in a current summing point, with a finite but low impedance Zc, which schematically represents the SQUID current amplifiers input impedance. The SQUID amplifier chain is not indicated for simplicity.
For cross talk minimization reasons, the summing point (or “common”) impedance needs to be low with respect to the TES resistance as it is a common electrical circuit element between the bias circuits of the pixels, and for that reason provides a route for cross talk. Other sources of cross talk in FDM are finite out-of-band damping of the band pass filters, nonlinearity in the amplifier chain, and inductive coupling between the LC filters, which also induces currents in neighboring bias circuits like common impedance does.
Figure 10 shows the practical implementation of FDM in more detail. The readout chain consists of three main building blocks, i.e. the cold electronics, the front-end electronics, and the digital electronics. The cold electronics provides the TES-based detectors with LC filters and the first stage amplification by means of SQUIDs. The front-end electronics consists of the cable connection to the room temperature electronics, and the analog signal and buffer amplifiers, as well as bias sources. The digital electronics provides the demodulation functionality, the sinusoidal (AC) carrier generation, and dynamic range enhancing feedback signals for the SQUID amplifiers by means of baseband feedback (BBFB).
There are many aspects of this design, which require more detailed design motivations and discussions. However, these issues are outside the scope of this document. Instead, in the following subsections the design concepts for the building blocks will be briefly touched, to summarize the context.
Multiplexing factor and frequency space assignment
The multiplexing space is determined by the available bandwidth and the minimum allowable separation in frequency space. There are several practical factors influencing these constraints, which are listed below.
1. Component size. The physical dimensions of the lithographically produced LC filters scale inversely proportional with their center frequency, i.e. fcenter α 1/(Area)2. Taking into account the dimensional constraints of the focal plane arrays (FPA), in combination with the production constraints, this gives a practical lower limit of 1 MHz.
2. SQUID bandwidth. The noise temperature of a SQUID is proportional to the bandwidth, which for the bolometer requirements results in an upper limit to the available bandwidth of 3 - 7 MHz, depending on the SQUID dimensions.
3. Cable bandwidth. An ideal (lossless) transmission line has an infinite bandwidth. Practical (lossy) transmission lines attenuate the signals. Thermal considerations dictate a lower limit to the cable resistances, which limit the usable bandwidth of a cable. Practical cables of these dimensions show a cable pole around 5 – 10 MHz, depending on the available cooling power around 4K, and the length of the cable. Note that below 4K superconducting wires can be used, which are virtually lossless for the signals, but isolating for thermal loads.
4. Dissipation in the digital electronics. The dissipation in the digital electronics scales proportional with the operating frequency. The highest required operating frequency is proportional to the highest frequency in the FDM band. The available power budgets for the digital electronics currently yield the upper limit (typically 3 – 8 MHz).
5. Baseband feedback stability. The baseband feedback gain-bandwidth-product (GBP) is typical set at 2 – 3 times the electrical bandwidth of a pixel, as a result of dynamic range and linearity considerations. The GBP sets the minimum band separation at approximately 10 times the GPB kHz to guarantee stable operation.
6. Cross talk. Finite damping in LC band pass filters leads to finite leakage of power neighboring pixels, inverse proportional to their distance in frequency space. Given the cross talk requirements, this sets the ratio between minimum distance between pixels and the electrical bandwidth per pixel at >6.
7. Dynamic range. The higher the multiplexing factor, the higher the dynamic range requirements for the SQUID amplifier chain, and the carrier generation DAC's and the feedback DAC's. The availability of space qualified digital-to-analog converters limit the number of channels to approximately 160 for bolometer applications, and 40 for micro calorimeter applications. Note that this limit could be circumvented by using adding the signals of a larger number of DACs. However, the latter is constraint by power limitations.
Combining all factors has for the bolometer system design led to the choice of a multiplexing factor of 160 pixels/SQUID channel, with a band separation of 12.5 kHz, and a lower and upper operating frequency of 1 and 3 MHz, respectively. For the micro calorimeter case the multiplexing factor has been set at 40 pixels/SQUID channel, with a separation of 100 kHz between pixels, and a frequency range between 1 and 5 MHz.
Construction of the SQUID amplifier chain
The function of the SQUID amplifier chain is to amplify and transport the TES detector signals to the room temperature electronics over the cable harness. As the cable harness is exposed to electromagnetic interference (EMI), the signal power levels need to be above the EMI thresholds, to ensure signal integrity. Simultaneously, there is a minimum energy resolution requirement at the input of the readout chain, as set by the noise temperature of the detectors, in combination with cross talk requirements.
It turns out that these requirements can be met best by a two-stage SQUID system, where the first stage is sitting close to the detector, and is optimized for an optimal energy resolution in combination with maximum power consumption. The second stage is sitting at an elevated temperature (~2K) where more cooling power is available, so that more powerful SQUIDs can be used to drive the cables to room temperature.
Ground Based FIR telescopes
The new generation of large FIR/Submm telescopes is designed from the start to have a large field of view (FOV), in contrast to most older telescopes. The speed for mapping large areas of the sky is proportional to the number of detectors. Therefore in order to take advantage of the large field of view, one would like to fill it with detectors, eventually. Even if this might not yet be feasible today, detector arrays with thousands of pixels will be required to cover at least a substantial part of the telescope field of view. As continuum detectors in this spectral range need to operate well below 1 K for optimum sensitivity, reading out a very large number of such detectors is only feasible via cryogenic and multiplexed readout techniques.
APEX, the Atacama Pathfinder Experiment, is one of these new generation telescopes. The site on the Llano de Chajnantor at 5100 m altitude in Chile has excellent atmospheric transmission. The surface quality of the reflector is very high and there is a large field of view at the Cassegrain focus (Figure 11). In the following, instruments for APEX are taken as reference, because of our close association with APEX and our detailed knowledge about its instruments. However, the conclusions apply to any FIR/Submm telescope with a large field of view, that is situated on a good site.
The integration time for mapping a given area to a specified depth is inversely proportional to the number of pixels. On the other hand, if integration time and mapping area are fixed, the minimum detectable flux is inversely proportional to the square root of the number of detectors. Therefore care must be taken not to degrade the number advantage by lower pixel sensitivity or higher readout noise, as compared to a direct readout of fewer pixels.
APEX is a freestanding telescope, not protected by a dome. Even so, the surface quality of the main dish is very high (15 microns rms). This and the quality of the site make it uniquely efficient for observations at wavelengths as short as 350 microns.
The FOV at the Cassegrain focus of APEX is limited by the central hole in the primary to 0.5 degrees (30 arcmin) in diameter. LABOCA, the large bolometer camera for APEX is an example for a FIR/submm astronomical camera. It operates at a wavelength of 870 microns and covers a field of view of 0.2 degrees (12 arcmin) with 300 pixels. The existing LABOCA has semiconductor bolometers and operates in a liquid Helium cryostat. In the following, this camera will be referred to as LABOCA-1, to distinguish it from the superconducting version LABOCA-2. In both LABOCAs all pixels are antenna coupled and fully efficient at a spacing of 2 f l, where f is the focal ratio at the final focal plane and l is the wavelength. Some further gain could be obtained by changing to small Nyquist sampling pixels (at 0.5 f l), but only at the cost of increasing the pixel number by a factor of 16! At the present state of the art, extending the field of view with fully efficient pixels seems a more effective use of resources. Even in this concept there is plenty of room for growth in pixel numbers.
The spectral band-pass is defined by means of cold composite filters, a quarter wavelength back short and a short (2 x diameter) piece of cylindrical waveguide at the exit of the conical horn. The conical horns couple the radiation to a micro machined array of TES detectors on a hexagonal grid with a grid constant of 4 mm. Each detector has lossy crossed dipoles as absorbers on a structured silicon nitride membrane and a TES consisting of a bilayer of Mo/Au-Pd. The whole detector array is fabricated on a single 4-inch wafer with on-array wiring to the out-of-plane multiplexers. The multiplexers are set up as multi-chip modules with superconducting bonds to the array. The focal plane is cooled to about 300 mK by a He-3 sorption cooler.
FIR/Submm detectors on ground based telescopes have to cope with a substantial and variable background power load, which restricts the achievable detector sensitivity because of photon noise. This implies a relaxation of the cooling effort since adequate detector performance can be achieved at relatively high operating temperatures, like 300mK.
The ultimate goal for the multiplexer architecture is to integrate as many pixels as possible into one readout chain. Although the single-pixel noise requirements are relatively relaxed because of the background limited detector performance, the implementation of a large number of multiplexed channels (>> 10), while avoiding multiplexer losses, constitutes a remarkable technical challenge.
Conventional TDM’s have the fundamental problem of increasing the current noise of N multiplexer SQUIDs by Sqrt[N] because of aliasing. Since the SQUID noise is typically one order of magnitude lower than the intrinsic detector noise, lossless TDM could be achieved by low pass filtering the detector noise at half the MUX sampling rate. However such a filter between the MUX chip and the TES will be difficult to implement because of the impedances involved. For N about 30 this would still work, while fore larger multiplexing levels the SQUID noise becomes an issue. Known concepts for overcoming this problem are coded (CDM) or FDM. FDM of far-infrared TES architectures is not trivial because the required modulation of DC detectors with an AC bias in a narrow frequency band constitutes a technical challenge which cannot be easily solved. Especially the need for cryogenic high frequency filters raises a purely technological problem of manufacturing the required large capacitances.
In CDM, the detectors that are connected to one multiplexing chip are not read out sequentially, like in TDM. Instead all detectors are sampled continuously, while the polarity of the signal from each detector is cycled according to a sequence of orthonormal unit step functions (Hadamard- or Walsh-functions). Consequently, coded multiplexing seems the method of choice, where the SQUID noise aliasing disappears due to the use of the full time interval for all channels instead of only single time slots in the TDM case. A scalable CDM approach was therefore chosen for the E-SQUID project with respect to FIR/Submm applications on the ground.
The preferred astronomical mapping scheme is to use the telescope as scanner, that means moving continuously over the area of interest, while taking data. Multiple passes will then result in a Nyquist sampled map. This mapping scheme, puts several demands on the pixel performance:
1. A single detector should achieve an electrical NEP below the estimated photon noise of background. This performance must not be compromised by the multiplexer, especially in the data bandwidth of the astronomical signals between 0.1 and 10Hz.
2. Crosstalk between pixels should be less than 1%
3. Extended emission in the mapped area will be transformed to low signal frequencies by the scanning motion of the telescope. Therefore detectors should have almost no 1/f noise above 0.1Hz.
Since the 2 f l spacing between neighboring pixels allows for extended wiring between pixels, a multi-chip approach with multiplexers out-of-plane is a straightforward and reasonable approach. Conceptually, multiplexer chips will be mounted alongside the detector chip, connected electrically with superconducting bond wires. To ease the implementation, this multi-chip approach implies the setup of multiplexers at the same operation temperature as the detector. This has consequences for the multiplexer design (critical SQUID parameters) and the dissipation balance. The power load of the multiplexer itself is almost negligible.
Second stage amplifiers are needed, typically SQUID arrays for pure amplification or Fourier SQUID arrays (SQIF) for single valued response. These are installed at a higher operation temperature around 1K on the next cooling stage. In that way, no notable direct thermal load on both cooling stages is expected, as long as the wiring between MUX and amplifier is designed properly.
Optical
In optical wavelength areas, the basic idea was to utilise the single photon counting SQUID technique similar to X-Ray region. A decision was made to replace Metallic Magnetic Calorimeters (MMC) with Superconducting Tunnel Junction (STJ) detectors. However, the SQUID readout for multiplexing of the STJ would require chips with dimensions that are too bulky to be housed in the cryostat at the University of Leicester (ULEIC) and would also be in general impractical for large number of pixels. Therefore the SQUID readout for multiplexing the STJ has been abandoned. A readout based on silicon germanium transistors and working at cryogenic temperature was considered as a solution. Amplifiers with performance that would fit the requirement for optical photon detection have been developed by VTT, and were available for a test of a single pixel read-out. However, synergy could not be maintained anymore with the sub-mm and X-ray mission approaches.
Design considerations
Vital goal of the E-SQUID project is the achievement of fabrication compatibility between the two major European foundries for superconducting electronic circuits: VTT/Finland and IPHT/Germany. Thus, in view of the potential dissemination of E-SQUID devices for future space missions, both foundries can act as supplier to European space programs with the crucial advantage of an available back-up in case of unforeseen technological or organizational problems.
At the start of the project, such compatibility was not pre-existing. Although general aspects of superconducting technology for SQUID-based devices were used at both sites, application demands from antecedent projects have driven the foundries to different implementations. Moreover, since the intended realization of code-division multiplexers would potentially benefit from an on-chip digital control, IPHT has advocated the consideration of RSFQ as a promising technology, further increasing the requirements on a common technology.
In the design phase, mainly the two partners concerned (VTT and IPHT) have worked together on a common set of design rules. The work based on existing foundry processes at VTT (MRI-MEG) and IPHT (RSFQ 1F). The following basic requirements have been agreed as starting point:
• three superconducting metallization layers, allowing for moderately complex designs
•
• projection lithography, allowing for high-density designs
•
• resistors compatible for operation at temperatures below 1 Kelvin
•
• width of metallic lines smaller than 7 micron, facilitating spontaneous flux expulsion (prerequisite for large coherent SQUID arrays)
•
At the same time, the partners agreed to maintain differences in their processes, which otherwise would imply too far-reaching changes with unpredictable risks for the respective foundry:
1. IPHT uses an electro-chemical oxidation procedure (anodisation) in order to passivate niobium structures before the deposition of insulation layers (SiO2), whereas VTT uses only SiO2 films. At the design level this implies the requirement of a network of auxiliary connection between all niobium structures to be anodized for the IPHT version of a common process, which are unnecessary for the VTT implementation.
2. For the formation of Josephson junctions, IPHT deposits a specific trilayer package to the junction regions, whereas VTT etches junctions out of a trilayer which covers the whole processing wafer. The thorough solution to handle that difference was to define a common circuit primitive (Josephson junction definition and contact windows) as exchangeable design cell, which can be expressed differently for the respective fabrication processes.
Another crucial difference in the processes turned out to be intolerable: In order to handle the problem of insufficient edge coverage in wire-crossing steps, VTT is using a chemical-mechanical planarization to level the subjacent wire and the isolation film, whereas IPHT relies on solely using material-additive techniques, i.e. larger thickness of the overlying wire. The resulting thickness incompatibility could have been acceptable in purely-SQUID designs, but for digital logic circuitry, the accurate sheet inductance mainly defined by the distance between two wires becomes crucial. In simulations at IPHT it was determined that design margins for digital logic would become prohibitive if the insulator thickness variations persist between the VTT and IPHT processes.
In consequence, according process development was initiated at VTT. A new proprietary technique of selective insulator deposition at the steps was developed, which resolves the step-coverage problem without chemical-mechanical planarization (see Figure 12). With this novel technique, the thicknesses of the insulation between the wiring layers became comparable between VTT and IPHT process, allowing for the use of similar RSFQ designs.
The described design variances have been incorporated into a script for layer operation, which automatically translates a design from the originating foundry into the respective version of the other. With this script, designs both from VTT and IPHT have been successfully transferred to the respective builds within the frame of the E-SQUID project.
Fabrication
The original work plan scheduled a regular sequence of 4 alternating fabrication rounds (so called “builds”) at the foundries (VTT: Build 1&3, IPHT: Build: 2&4). In the early project phase, technical difficulties caused a considerable delay: whereas at VTT a major etching tool was under repair, at IPHT the resumption of regular operation after a recent full reconstruction of the clean room facility was still in progress. In view of potential risks and in order to minimize the delay, a change of the sequence was decided. The first build started at IPHT late 2011; however, because of the still unstable clean-room operation that build failed. A successful replacement build was fabricated between February and April 2012. That build mainly contained technology tests and building blocks of code division multiplexers and RSFQ electronics. Amongst these, also ambitious experimental design have been included anticipating the intended complexity of forthcoming devices.
After testing and characterization of Build 1, an extensive design phase was executed. Because testing of building blocks (especially current steering switches) has revealed partially unexpected behavior, the start of build 2 was intentionally delayed until the design issues have been resolved. In consequence, the start of Build 2 was expected only at the beginning of 2014, which would have seriously challenged the project schedule. Therefore, the common decision was made to perform Build 2 and 3 in parallel at VTT and IPHT; abandoning the possibility of design iteration between Build 2 and 3 in favor of a successful demonstration of the compatibility between the foundry processes. Both builds have been fabricated successfully; however, only the VTT Build 3 yielded high quality parameters whereas the IPHT Build 2 suffered from an insufficient insulation quality . The scheduled Build 4 at VTT was used to test the new ESQUID technology on topic related devices, namely FDM’s for simultaneous ESA projects as SAFARI. The intention of this build is the demonstration of the versatility of the novel foundry process for superconducting electronics in a wider context.
Despite the occured technological problems, still, the successful accomplishment of fabrication builds at IPHT (Build 1) and VTT (Build 3) allows for a direct comparison of the technologies. At the example of functional 8 channel multiplexer devices and RSFQ logic circuits, the compatibility of both foundries was clearly demonstrated. For the first time in the international superconductivity community, the direct design transfer between two independent manufactures was shown. That achievement enables future joint work in the field of superconducting electronics and strengthens the European research especially in comparison to competitors in the United States.
Test environment
Code Domain Multiplexing for Space Applications
VTT equipped its BlueFors LD400 dilution refrigerator with a retractable wideband wiring set, set of shielded boxes at 4K plate, and an experimental Cu box at mixing chamber. Because wiring plays a crucial role in wideband readout systems, it was designed to be rapidly modifiable, with spring-loaded thermalization blocks and homemade vacuum-tight feedthroughs at removable KF flanges. A modular electronics cage is installed at the top flange, where rapidly prototyped, standard-sized electronics cards can be installed.
VTT also constructed a ‘broomstick’ probe for quick tests in liquid helium, where the same KF-flanged wiring can be installed (Figure 14). A modular electronics cage is located on top of the flange. The ‘broomstick’ makes it possible to test the same electronics/wiring combinations with the same parasitic effects in liquid helium, before installing them into dilution refrigerator.
VTT has designed and constructed two types of cryogenic semiconductor amplifiers and one room-temperature LNA. These are to be used after SQUIDs in the amplification chain. The first cryogenic type has voltage input, order-of 100 MHz bandwidth, and 0.5 K noise temperature into a few k input load. The second type (Figure 15) is designed into 50 transmission line –based systems, has >1 GHz bandwidth, and 5 – 10 K noise temperature into 50 input load depending on frequency. The room-temperature LNA is based on paralleled SiGe bipolar transistors, it has lower-than 0.5 nV/sqrt(Hz) voltage noise, and provides optional low-noise active-feedback 50 input termination. The 1 GHz amplifier is targeted towards sin/cos flux counting readout, and possibly towards reactive microwave SQUIDs.
Because large-volume cryogenic testing of SQUID devices is foreseen in the pathway to the certified flight hardware, we have taken the first steps to alleviate the mass testing. We shall attempt spring loaded contacting for tests, to avoid the labour-intensive wire bonding step. The wafer is first diced into reticles (Figure 16), which are tested in liquid helium. The reticles are the further diced into individual steps. The development efforts towards mass testing are beyond the scope of E-SQUID project, but preparatory steps have been taken, by including test pads and the internal test wiring in the Build 4 reticle design.
In the measurements performed at UH, the same measurement geometry and test system was used, that is described under VTT’s test environment section. The cryogenic test system was build inside a Bluefors SD250 dilution fridge situating in EMI shielded room. The measurement PC was operated from outside the shielded room to minimize its interference with the test system. The transmission between the laboratory PC and measurement system electronics was arranged by optical lines and feed-troughs provided in the wall of the EMI shielded room.
UH test facilities situate at Kumpula Campus Acceleration Laboratory premises, in a grotto below the main building about 20 meters underground. The experiment in UH test is presented in Figure 17. The cryostat’s pulsetube, which has potential to add interference in measurement signal, is installed outside the EMI shielded room. Altogether, the low ambient EMI background was anticipated to provide a good base for measurement work. However, the thermalization of the wiring turned out to be problematic in UH’s dilution fridge as a consequence of low cooling power as compared to the VTT’s dilution fridge, in which the wirings were initially build and tested.
According to the original plan, the CDM multiplexing test were planned to be performed at UH, but because renovation of the laboratory premises and delays in dilution fridge installation, the test were transferred to VTT. Accordingly only one measurement session was performed at UH premises.
FDM for Space Applications (SRON)
As the development work at SRON over the last years has been aimed at developing readout for bolometers for the SPICA/SAFARI project, a new cryogenic setup had to be designed to be able to demonstrate the combination of the SQUIDs and X-ray micro calorimeters. The other components besides the SQUIDs, i.e. the LC band pass filters, the micro calorimeters and the room temperature electronics have been made available outside the scope of this contract.
We have separated the demonstrations in an experiment on the SQUID performance, on the single pixel performance, and on a multiplexing demonstration with micro calorimeters. The results are described in the following sections.
Cryogenic experimental FDM setup
The function of the cryogenic experimental FDM setup is to provide a temperature and magnetic field controlled environment for the TES detector chip, the LC filter chip, and the SQUIDs. A radioactive Fe55 source is used to illuminate the X-ray detectors with photons of 5.9keV with a count rate of a few counts per second.
Two versions of the setup are available, one with the two SQUID stages at two temperatures, to demonstrate compatibility of the SQUIDs with the intended configuration in a satellite, and one with both SQUID stages at base temperature (less than 50mK), which is experimentally an easier configuration. The bracket with TES detector chip, the LC filter chip, and the SQUIDs is the same for both cases; only the location of the 2nd stage (or upper) SQUID is different. Both setups are described below.
The 2-stage 2-temperature set-up consists of a 2nd stage SQUID in a cryoperm tube, that is thermally linked to the still level, and an FDM bracket that is thermally linked to the mixing chamber plate of a helium-free dilution refrigerator from Leiden Cryogenics. The thermal link of the 2nd stage SQUID is dimensioned such that its temperature can be controlled between 0.6 and 2 K, without disturbing the operation of the dilution refrigerator. Dimensioned with the same boundary conditions, the temperature of the FDM bracket with the 1st stage SQUID can be regulated between 17 and 200 mK without altering significantly the base temperature of the mixing chamber.
The FDM bracket assembly that holds the FDM multiplexing circuit, consists of a low magnetic impurity copper bracket, providing mounting space for the front-end SQUID, the LC filters and the TES array chips and a printed circuit board for the electrical connections. A ruthenium oxide thermometer and a heater are glued at the bottom of the copper bracket for the readout and stabilization of the temperature.
The FDM bracket can be used with both a low NEP TES bolometer array, and with a highenergy resolving power x- ray micro calorimeter array. The former configuration requires very low parasitic optical loading (less than 1aW), which is achieved by means of light blocking filters in the signal loom feed through and a light-tight assembly.
The X-ray micro calorimeters are sensitive to magnetic fields and their performance is optimal at static field lower than 1µTesla). Special care has been taken to design the magnetic shielding and to improve the uniformity of the applied magnetic field across the array. The TES arrays chip fits into a superconducting Helmholtz coil fixed at one end of the bracket. The coil is used to generate a uniform perpendicular magnetic field over the whole pixels array. The shielding of the external magnetic field was achieved by fitting the bracket into a Nb can wrapped by few layers of metallic glass tape. The interface of the Nb can with the lid of the bracket was designed such that it forms a labyrinth, filled with carbon-loaded epoxy on the copper lid side. In this configuration the Nb can provides both the required magnetic shielding, and the required stray light shielding.
The circuitry PC-board currently allows to read-out a maximum of 18 pixels under ac bias in a FDM configuration. The SQUID chip is fixed to the copper bracket high thermal conductivity glue (Ge Varnish). Two photographs of the setup are shown in Figure 18.
Optical Applications (ULEIC)
A read-out based on SiGe amplifier has been considered as an alternative to SQUID read out for STJ detectors. A cryogenic Voltage–Voltage amplifier developed by VTT has been modified has a current to voltage transimpedance amplifier.
The amplifier is seated at the 4 K stage of the cryostat (Figure 19a) while the detector operates at 300 mK (Figure 19b) The signal from the detector is carried to the input of the amplifier through an NbTi twisted pair. The amplifier is encapsulated in an aluminium die cast box. The Amplifier control lines, bias line and signal line from 4K To the room temperature amplifier are made of 8 constantan twisted pairs.
Far-infrared applications for ground based observatory (MPG)
While fully functional CDM chips from the E-SQUID project were not yet available, useful development work for the project was still possible. MPIfR and IPHT are engaged in building an array of three hundred TES sensors (LABOCA-2) for the APEX telescope in Chile. LABOCA-2 is meant to replace LABOCA-1, which has semiconductor bolometers and operates in a liquid helium cryostat. The new camera is required to match sensitivity of LABOCA-1. Both LABOCA versions operate at a wavelength of 870 micron.
LABOCA-2 operates with TDM readout, using a ten channel SQUID multiplexer chip manufactured by IPHT. Thirty multiplexer chips are arranged around the edge of a 4” wafer with 300 TES in a hexagonal grid (Figure 21). The signals from each MUX are further amplified by a SQIF before entering the analog and digital readout electronics at room temperature.
LABOCA-2 is cooled by a three stage He-4/He-3 sorption cooler on a pulse tube (Figure 22). Operating temperatures are regulated at 230mK for the array with the MUXs and at 800mK for the SQIFs. The temperature of the array of conical horns which couple the radiation to the TES is around 280mK.The LABOCA cryostat is a very convenient test platform as it features:
1. Fast automatic cool down: 300K to 230mK in 25h.
2. Fast forced warmup: 14h
3. Hold time > 48h
4. Cold automatic recycling: 6h
The disadvantages are:
1. Tests can only be done on complete 4” LABOCA wafers
2. Tests, with the pulse tube stopped, are possible for only about 2 min.
Three subarrays of LABOCA-2, each with 100 TES, are routed to three FLL-electronics boxes on top of the cryostat. The basic MUX sampling frequency is 15 kHz. Three operating modes are available:
1. In the normal TDM-mode the signals are demultiplexed and for large signals, there is an option to correct for flux jumps of the SQUIDs. The frequency limit in this mode is 750 Hz.
2. In the “step-mode” the MUX is switched off and one particular TES can be selected for readout at 15kHz, resulting in a frequency limit of 7.5 kHz.
3. The three FLL boxes are identical and can be interchanged. One of the boxes is fitted with a connector to the output of the SQIF. In step-mode, it allows to measure the high frequency noise with an FFT spectrum analyzer. Our model SRS 780 has a high frequency limit of 102.4 kHz.
The warm electronics are virtually identical to that delivered to the E-SQUID project by Supracon, except for the much higher number of channels. Changing from TDM to CDM would mainly require changes in the software.
In radio astronomy, the response of a receiver to radiation is calibrated by taking hot/cold data. That means that absorbers at 300K and 77K are put in front of the receiver and the change in output is measured. 300K and 77K are typical background temperatures in the FIR as well, therefore the same method can be applied. If a complete I/V characteristic is taken at both temperatures, then the response for all bias settings is obtained at once.
Noise data were taken in any of the three alternative ways: MUX-, step- or spectrum analyzer mode. NEP values can be derived by combining the noise with the hot/cold data. Other sensitivity criteria are the direct comparison with LABOCA-1 or the amount of 300K photon background noise.
Results
Code Domain Multiplexing for Space Applications
After Build 1, VTT used the Jena-fabricated superconductive switch chip from Build 1 to demonstrate 8-pixel TDM of X-ray calorimeters at 50 mK temperature (Figure 24). A 5x5 –pixel calorimeter array from SRON was used as the detectors, which was irradiated from a radioactive 55-Fe source. The resulting time traces are shown in Figure 25. The switch chip was primarily designed for measuring dc characteristics of multipixel bolometer/calorimeter arrays (Figure 26), but the chip adapted into TDM role, too. The work was reported in the EUCAS-2013 conference.
After Build 2/3, a VTT-fabricated version of the switch chip, with the same layout as the Jena version from Build 1, was tested (Figure 27) at 4.2K without calorimeters, and found to perform essentially the same way as the IPHT-fabricated version (Figure 28).
An amplifier SQUID with a large mutual inductance (Figure 29), order-of 0.8 _mu A / Phi_0 sensitivity, was tested at 4.2 K, and found to perform well-enough that testing with detectors is warranted, either with SPICA bolometers or X-IFU calorimeters. The SQUID is designed for the so-called ‘high-Z summing point’ circuit topology, in which the lower 50 mK SQUID is replaced by a passive (non-dissipating) transformer, and the first SQUID resides in the 2 K refrigerator stage.
General functionality of the so-called SQUELCH chip was verified at 4.2 K. The idea behind the SQUELCH is to utilize the compressibility of the information flow from a calorimeter array, namely the fact that co-incident counts from many adjacent pixels are not likely. Two cascaded SQUELCH chips are designed to generate a 5-channel binary-coded output signal from 32 calorimeter pixels. The inactive channels not registering an x-ray count, reside in superconducting state and hence do not contribute to noise. This is the motivation behind the name ‘SQUELCH’. The chip is depicted in Figure 30.
After Build 4, the amplifier chip pair intended for FDM readout was successfully tested at 4.2K (Figure 31). The lower amplifier is a single SQUID intended for 50 mK operation, while the upper amplifier is a 4-parallel 184-series SQUID array, to be used as the second amplifier stage at 2.5 K temperature . The 4x184 array shows flux noise of 7E-8 Phi0 / sqrt(Hz), which implies a larger native dynamic range than any other SQUID device in world.
Also tested from Build 4 is a high-sensitivity SQUID intended for high-Z summing point, which is equipped with local feedback circuitry for response linearization. The SQUID is depicted in Figure 32.
Outside of the E-SQUID fabrication rounds, a system-level experiment on open-loop sin/cos flux counting was performed, by using pre-existing SQUID devices from other projects. The flux-counting approach is seen as a way to increase the dynamic range without increasing heat dissipation. The work was reported in DOI: 10.1088/0953-2048/27/7/075005. A chip design was included in Build 3 which combines flux-summing CDM circuit (i.e. signal summing coils with Hadamard-coded polarities) and the sin/cos flux counting readout SQUIDs. Tests on the CDM chip are underway.
FDM for X-Ray and Far-IR Space Applications (SRON)
To verify the SQUID behavior of a G1/F5 combination strictly independent of the TES or multiplex behaviour, the SQUIDs have been characterized during an independent cool down, without any load, such as LC filters and TES based detectors, connected to the input coil.
Both the upper-SQUID (F5) and the lower-SQUID (G1) are mounted on a PCBoard in a magnetically shielded SQUID test module (not shown). In Figure 34a shows the voltage-to-flux characteristics of the upper-SQUID measured as a function of the bias current, and with the lowerSQUID in the superconducting state.
Figure 34b shows the voltage-to-flux characteristics of the two-stage SQUIDs measured as a function of the bias current of the lower-SQUID array (G1). The two-stage SQUID array shows a clean V-ϕ curves with a maximum voltage modulation of 2.4mV at a lower-SQUID bias current of 50 µA. Note that the lower-SQUID is effectively voltage biased at 50 µV since the bias current is applied to a 1Ω shunt resistor connecting the lower-SQUID output to the upper-SQUID input coil.
We measured the current noise of the two-stage SQUID at three different temperatures, respectively 3K, 100mK and 20mK. Two calibration tones at 12kHz and 2MHz have been applied via the feedback coil to monitor the SQUID gain as a function of temperature. The magnitude of the calibration tone corresponds to an applied flux to the lower SQUID of 0.01Φ0. In Figure 35a the spectral flux and current density taken at temperature of 3K and 25mK is shown. The current noise has been calculate using the nominal input sensitivity of the G1 SQUID array equal to M−1 = 7µA/ Φ0 In Figure 35b a zoom-in of the SQUID noise spectra around the 2 MHz calibration tone is shown. We measured a flux noise equal to 0.35 μΦ 0/√Hz and 0.18 μΦ0/√Hz at 3K and 20mK respectively. This corresponds to an energy resolution of ε = SΦ/2Lsq ≃ 9ħ, where an Lsq = 70 pH is assumed, and is limited by the room temperature amplifier and the data acquisition system.
A relatively straightforward optimisation of the room temperature set-up should lead to a further improvement of the SQUID sensitivity. The flux sensitivity for the G1 SQUID array is √SΦ,QL = 6 • 10−8 Φ0/√Hz, which should be achievable if the temperature of the SQUID shunt resistors can be cooled at temperature below 150mK. The measured current noise at the SQUID input is 1.2 pA/√Hz at 20mK, which corresponds to a coupled energy resolution of ε c = ½LinSI ∼ 20ħ, where Lin ∼ 3nH is the input inductance of the SQUID.
We have observed no difference in the noise when the temperature of the mixing chamber was raised to 100mK indicating that the actual SQUID shunt resistors are thermalized at higher temperature and that the noise is limited by the room temperature electronics.
Single pixel performance under AC bias
TES based detectors need a voltage bias source for stable operation. The bias source can be either a constant voltage (DC) source, or an alternating voltage (AC) source. Because of the physical processes in the detector it cannot be assumed automatically that the performance under AC and DC bias is equal. The purpose of this section is to give an overview of the status of the performance measurements under AC vs DC bias.
Biasing the TES with a sinusoidal alternating current (AC bias) is an essential ingredient for FDM. Whether the performance of the detector under AC bias and DC bias is the same or similar is an important difference for applications of TES detectors. For bolometer pixels it has been shown consistently that the dark NEP measurements under both AC and DC bias give the same results. For microcalorimeters, however, this demonstration has not been given consistently until now.
Figure 36 (left) shows the latest results of NEP measurements on an X-ray microcalorimeter which has been characterized under both DC and AC bias. The AC results are shown for two different settings of the anti-alias filter in the electronics. The corresponding integrated NEP, which is equivalent to the energy resolution of the pixel, is shown in the legend of the plot. The results under DC bias show the lowest NEP and therefore the best integrated NEP or energy resolution (ΔE=2.3eV FWHM), while the best integrated NEP under AC bias currently equals 2.8eV FWHM.
From the plot it becomes clear that, by comparing the configuration of the anti-alias filter with the red curve and the green curve, either the low or high frequency part of the curves overlap. This implies that we have reasons to assume that the NEP under AC and DC bias are equal, once this issue in the room temperature electronics has been solved.
Figure 36 (right) shows the measured integrated NEP as function of bias point resistance in the transition, normalized on the normal state resistance, for two different magnetic fields. The plot shows that the best results can be obtained for a range of bias points and magnetic fields.
The measured X-ray resolution under AC bias is currently 3.7eV while 2.3eV has been observed under DC bias for the same pixel. We cannot exclude at this moment that the discrepancy is caused by drifts in the responsivity of the detector, as a result of insufficient temperature stability in both the refridgerator and the room temperature electronics. Ongoing experimental work is aimed at improving stability of the setup on the time scale level of 1 hour, to make representative measurements under AC bias possible.
Multiplexing demonstration
FDM multiplexing with X-ray calorimeters has been demonstrated earlier at SRON in a different setup and with different SQUIDs. With these results in mind, the main purpose of this demonstration is to show that in the combination with 5 pixels over an increased frequency range the noise level has improved with more than an order of magnitude.
An overview plot is shown in Figure 37, where the current noise spectral density at the output of the SQUID amplifier chain is shown as function of frequency. It is clear that within the measured range between 1 and 6 MHz, the SQUID noise level is independent of frequency at a low level of 2.2 pA/√Hz. The peaks in the spectra which belong to the microcalorimeter pixels are labeled px 1-5. The narrow peaks in between the pixels originate from imperfections in the room temperature electronics.
The same noise levels are observed after demodulation at the output of the digital baseband feedback electronics, as is shown in Figure 38. Here we see the noise of three resonators in the superconducting state. The high noise level of approximately 20 pA/√Hz corresponds to the Johnson noise of the stray resistances in the bias circuit. The spread in Q-factor of the LC resonators causes the different level and corner frequency for the 3.6 MHz resonator. Above the corner frequency the noise level rolls of to the SQUID current noise level.
One possible form of pixel interactions when multiple pixels are switched on is bias power leakage between pixels. The latter effect depends mainly on the separation in frequency space between pixels, normalized on the bandwidth of an individual pixel. In this experiment the spacing has been chosen such that no observable power leakage is to be expected. The measurement of a current-voltage (IV) characteristic of a pixel with and without neighboring pixels biased, is shown in Figure 39. It is clear that at the level of less than 1% the curves overlap, which confirms the expected lack of power leakage.
The last experiment to be discussed, consists of measuring the time response of 3 pixels simultaneously. A time domain plot of 3 biased pixels is shown in Figure 40. The time range (1 s) of the plot is chosen such that a small number of photon absorption can be seen. The response time of the detector to the photon hit is barely distinghuishable in this plot. The single sided signals are photon reponses, while the double sides signals are caused by electrical cross talk between the pixels. Whether the level of the cross talk is consistent with the expectations has not been investigated yet. It is, however, easy to conclude that such cross talk events are easy to reject, as absorbed photon responses are single sided only, while the cross talk events are clearly double sided.
We therefore conclude that the combination of the 2-stage SQUID tandem works well in combination with the microcalorimeter readout sofar. More experiments outside the scope of this work package will be needed to make more quantitative estimates of inter-pixel behavior possible, and to study the impact of a closer packing density of pixels in frequency space.
Far-infrared applications for ground based observatory (MPG)
Before the E-SQUID project, IPHT and MPG/MPIfR had been developing TES technology for astronomical and terrestrial applications. For that purpose, TES arrays with 7 elements were manufactured at IPHT and tested at both institutions in small He-3 cryostats with direct SQUID readout. As shown in Figure 41, photon background limited noise was achieved in a 300K environment. To achieve a similar kind of performance, but via a multiplexer of some kind, is the aim of E-SQUID project as applied for the FIR.
It was expected that the first LABOCA wafer fabricated according to the best 7-element test array would be good enough to go to the APEX telescope. Unfortunately this was not the case. For any given bias, which by design is common to 50 TES, some TES would show good sensitivity but others not. No common bias with good performance could be found. At the time, the problem was believed to be in the fabrication of that particular wafer, and it was decided to make more wafers.
Early in the year 2011, soon after the fabrication of the first LABOCA-2 wafer, the cleanroom of IPHT was shut down for a major expansion and installation of new modern production equipment. Unfortunately, all LABOCA wafers (11 in total), produced after the reopening of the cleanroom on new equipment had some kind of defect, like wrong Tc, poor uniformity or multiple transitions. By the end of the year 2012, the decision was made to go back to the old machine for production of LABOCA wafers. This lead to considerable improvement in the quality.
Back on the old machine, four wafers in a new layout have been produced in that period at IPHT, two with thermistors of 50 x 100 microns and two with thermistors of 100 x 200 microns.
The new LABOCA-2 wafers with 100 x 200 micron thermistors have almost ideal I-V characteristics, in the sense that the dissipated power in the transition range is almost constant and the transition to normal conduction is abrupt (Figure 42).
The instability at the superconducting transition is much closer to the normal transition for the wafers with 50 x 100 micron thermistors, making them less useful.
A most surprising result was the appearance of high frequency noise in these TES with quasi ideal I/V characteristics. If any broad maximum in the noise appears at one bias setting, it can be maximized by very small bias steps.
Tests were run by connecting old spare 7-element test chips to the new LABOCA-2 data acquisition. Two test chips were glued to one of the new LABOCA-2 wafers with 50 x 100 micron thermistors and bonded to two of the existing MUX, replacing the TES on the wafer. One MUX was left bonded to the TES on the wafer. The latter TES show high frequency noise (Figure 43).
On each test chip, several different TES structure had been implemented. Some of them therefore went into normal conduction at different bias voltages. In Figure 44, noise data were combined that were taken at about 0.2 V bias below the normal transition of each TES. Surprisingly, the noise is quite well behaved at high frequencies!
One particular TES on the test chip is very similar to the LABOCA design. Its noise spectrum is shown in more detail in Figure 45, with the pulse tube stopped. The suppressed Johnson noise at low frequencies in the range of the thermal time constant is evident! Measurements with input radiation were not possible in this improvised configuration. Three test chips from two wafers have been tested to date. All have low high-frequency noise, some slope in the power characteristic and reduced Johnson noise at low frequencies.
All the TES on the test chips have very small series resistors (about 8 mΩ) implemented on one side of the thermistor. At the time, these series resistors were only meant to provide a way to measure the I/V characteristic in both directions of the bias voltage. Also, all of the TES on the test chips have a Gold ring around the absorbing area. The TES on the wafer have neither the series resistor nor the Gold ring. At the time of writing this report it is not clear whether the different layouts have anything to do with the good performance of the test chips or if that is due to some lucky “accident” in the fabrication. More tests are necessary. We can exclude however, that the high frequency noise is caused by internal or external RF interference.
TDM could still be a valid MUX scheme if an electrical anti-aliasing filter were incorporated on the MUX chip. Space on the wafer for the larger MUX chip was foreseen in the new layout. The first MUX chips with coils at the input were produced at IPHT and are tested on a LABOCA-2 wafer. This was more with the idea of blocking the high frequency noise. To serve as anti-aliasing filter, TES with much lower normal resistance would have to be developed.
Another alternative is to use the thermal time constant of the TES for anti-aliasing purposes. It is expected that this will happen to some extent anyway at high background levels.
Optical Applications
A read-out based on SiGe amplifier has been considered as an alternative to SQUID read out for STJ detectors. A cryogenic Voltage–Voltage amplifier developed by VTT has been modified has a current to voltage transimpedance amplifier.
The stability of the STJ bias is crucial. A bias between 100 to 200 µV that seats within the superconducting gap is needed for photon detection. Moreover, a diagnosis of the Junction performance and states is required prior of photon detection. The diagnosis consists of an IV curve covering the entire superconducting gap. A typical IV curves obtained in our set up with a room temperature IV tracer is shown in Figure 46.
An in-built biasing system was used. However, the actual gain of the biasing voltage divider was a priori unknown and also could depend on the impedance of the load. This gain was experimentally determined using three different known load resistors (Figure 47). On the other hand, the current to voltage amplifier gain was known and was 105 V/A.
Top left, the voltage applied at the input of the voltage divider is plotted versus the measured current for 3 different resistance load. Top right, graphical determination of the two equivalent resistors of the voltage divider. Bottom left, voltage divider gain versus the load resistance deduced from the resistors value.
Figure 48 shows IV curves of STJ obtained at 300 mK when a tuned magnetic field is applied (red line) and without magnetic field (blue line). The Red curve is characteristic of a high impedance device. However, flux trapping or excess noise due to EMI degrade the IV curves (cf. Figure 46). Even though, the red curve is the best curve obtained so far with the cryogenic amplifier, attempts to measure X–Ray pulses within these conditions were unsuccessful. Therefore, no optical photon detection was considered.
To illustrate the multispectral imaging capability of STJs based camera, measurements of a human fibroblast cell using a triple bandpass emission filter were recorded. The cell was stained with 3 different dyes. The filter allows simultaneous transmission of fluorescence from multiple fluorochromes that emit in the bandpass regions. A typical spectrum integrated over the whole camera during a 1 s acquisition is shown in Figure 49. There are three well defined peaks that reflect the three emission bands of the filter. The peaks are broadened by the Gaussian response of the STJ. In order to estimate the contribution of each dye, their tabulated emission spectra are superposed with the measured spectrum. Even though, emission bands are not necessary centered at the maximum emission wavelength of the fluorochromes, each band has its signal clearly dominated by a single dye contribution. Therefore, while all dye’s contributions are measured simultaneously, data post processing allows selecting signal from a specific band (i.e. specific fluorochrome). Figure 50 shows images that have been obtained by counting event that occurred in the three specific bands indicated Figure 49. In the 420-450 nm band, the signal is located in the middle right of the image and shows the nucleus. In the two other bands, the entire cell is visible. The corresponding images using usual CCD cameras are also shown. In this case, each image corresponds to a different measurement using a different filter set.
High counts rate associated with high sensitivity of STJs gives the possibility of reducing the acquisition time with the benefit of, for example, reducing the risk of photo bleaching. Figure 51 shows an image of the nucleus of a human fibroblast cell for different times of acquisition. After only few milliseconds, the DAPI signal, concentrated at the nucleus of the fibroblast cell, is clearly distinguishable from the background, while conventional epi-fluorescence imaging requires generally acquisition time beyond the second.
Conclusion
Cryogenic imaging detectors based on TES arrays with SQUID MUX readout will have an important role in instrumentation of future space-based telescopes. The capability of cryogenic imaging detector arrays promise unrepresented observational capabilities in many fields of modern astronomy. The technology is applicable in large band of electromagnetic spectrum, from FIR to X-ray. Accordingly, several instruments based on this technology are proposed, such as Athena's X-IFU and SAFARI's SPICA. There are also ground based applications that will greatly benefit on maturation of this technology, such as the LABOCA camera on APEX.
Three out of the four fabrication rounds were successful (Builds 1, 3 and 4), from which Build 4 was devoted to demonstrate the selected FDM technique.
The partners VTT and IPHT defined the common design rules for the cross-compatibility between the VTT and IPHT foundries. The almost similar fabrication rounds demonstrated the cross-compatibility of the IPHT (Build 2) and VTT (Build 3) foundries with the Zappe-TDM chip as the example device.
The SQUID tandem which has been developed for FDM has been tested separately, and in combination with a small array of microcalorimeters. The measurements show that the energy resolving power of the SQUIDs is nearly quantum limited at 20ħ, which shows that low current noise spectral density can be combined with a low summing point inductance in an FDM system. This creates a multiplexing space of approximately 160 for the bolometer TES detectors, and approximately 40 for microcalorimeters.
In optical wavelengths, the SQUID readout for multiplexing of the STJ does not work due to magnetic field. Instead, cryogenic SiGe semiconductor amplifier with 1 GHz bandwidth and ~5 K noise temperature was designed, built and demonstrated, but the synergy with SQUID devices was loosed.
Potential Impact:
There are significant socio-economic impacts arising from the E-SQUID project.
Firstly, it will improve the Technological Readiness Level (TRL) in Europe in the area of cryogenic multiplexers, which is considered as the breakthrough solution that fulfils the extremely strict science requirements of future space observatories especially in the X-ray and the Sub-mm range. E-SQUID will also enhance technological spin in and collaboration between the top European developers of SQUID-based solutions, and the company partners will benefit from the results of the project as improved competitiveness in the international market. Furthermore, the results also support the needs of European security and defense sectors, as well as the medical sector.
As the highest level socio-economic impacts, E-SQUID will improve the European non-dependence in critical (space) technologies, and thus the competitiveness of European space industry in the global market. Consequently, the improved competitiveness will have a positive effect in Europe’s economy and thus employment. The future space science applications from EQUID project will bring new high quality data for the use of European science community, thus also raising the impact of European science worldwide.
The quantification of the socio-economic impact in the economy is difficult in terms of economic impact and their social implications to e.g. employment, since the lead time of the possible commercial applications is very long, i.e. of the order of a decade, and the various final commercial products and the extent of business that will exploit the commercialized technologies are not yet known.
However, European non-dependence in this critical technology is a target that can be considered as achieved in E-SQUID project, considering the needs of the European scientific community. This means that the capability of European foundries to develop and procure scalable SQUID multiplexers with sufficient noise characteristics *is demonstrated* at level, which justifies relying on purely European foundries for the use of future European space missions, but the *final quantitative requirements were not achieved* in E-SQUID project. It can be stated with high confidence that the quantitative requirements will be reached within sufficiently short time frame considering the needs of future European space science missions.
With the ESQUID project, substantial progress on that field promises to pave the way to more effective, user-friendly and cost-effective devices. First ground-based scientific instruments based on European technology are already in routine operation by observation staff, and future European space missions might benefit from the ESQUID technology. With that achievement, for the first time in cryogenic detector research, it seems promising to look for their application also beyond science.
The potential applications arising from the E-SQUID project include concealed weapon detection through clothing at airports, and detection of Near-Earth Objects (NEO’s) threatening the security of the whole human civilization. Also infrared camera applications by which a number of human activities can be detected from space including forest fires and missile launches are possible. The technology based on E-SQUID can also be used for Earth-based applications in ultra-sensitive THz cameras for medical applications.
In the following the impacts and exploitation of E-SQUID project results are explained in more detail.
Among possible fields, the imaging of human bodies for security checks has spearheaded because of its high social relevance by being an alternative to the much disputed conventional body scanners. Terror attacks of the kind which have taken place recently in public places such as airports or train stations would have been prevented if a bomb concealed underneath the clothing of the attacker had been detected in a timely manner. Therefore, a ‘body scanner’ which can generate an electromagnetic visualization of hidden threats is a very attractive concept for future security screening.
However, although the desirability of such devices is clear, their current realization as active X-ray or millimeter-wave scanners raises certain health and privacy concerns. One considerably more attractive alternative are cameras which could generate passive images more rapidly without the need for the targets to stop what they are doing and co-operate with the screening process. This is the principle of Smart Security (IATA’s checkpoint of the future) and its vision to guarantee “...an uninterrupted journey from curb to aircraft door, where passengers proceed through the security checkpoint with minimal need to divest, where security resources are allocated based on risk, and where airport amenities can be maximized…” This ambition can be fulfilled, with current knowledge on possible technologies, only by using cryogenic detectors, as it has been demonstrated by two European groups, both also members of the ESQUID consortium: A. Luukanen et al. from VTT and T. May et al. from IPHT. Their devices will strongly benefit from the ESQUID technology, thus improving the performance at the same time with reducing the manufacturing costs. The social impact of such cameras is obvious: as they will improve security whilst making journeys more comfortable and respecting the privacy of the traveler, the public acceptance of such unavoidable measures will be increased substantially.
The dissemination activities in that field include the presentation of ESQUID results in frame of conferences related to security and Terahertz research (see list), the introduction to industry partners (e.g. Rohde & Schwarz as system vendor for security devices), and in semi-popular articles for EU-related audience and decision makers (see list).
Another very promising direction for future dissemination of ESQUID technologies is the field of life science, where imaging applications might benefit from the unprecedented performance of cryogenic optical detectors. One nearby application is fluorescence microscopy. In modern medical analysis, fluorescence microscopy has become an indispensable tool in imaging of biological tissue on a cellular level. A large variety of life processes as e.g. chromosome dynamics or gene expression can be visualized, benefiting medical diagnosis as well as basic research. In case of using diverse types of probes which fluoresce at different wavelength, this technique is able to separately label different organelles of a cell, thus studying their interaction at the same view.
With an energy-resolving single photon detector array (e.g. a TES or a STJ) it would be possible to simultaneously record images at different wavelength without any signal losses due to today’s scanning or tunable filter techniques. With that, the study of fragile biological samples at cautiously faint illumination levels becomes possible; preventing fluorophor photobleaching and sample damage to the best possible degree. That becomes especially important in medical research, where life processes need to be studied with minimal influence on the sample. Although fluorescence microscopy was shown to achieve unprecedented performance in terms of spatial resolution and signal quality, the inevitable impact of the fluorophor excitation on the life process to be observed still hinders an un-spoiled observation.
Utilizing cryogenic detectors, the required excitation could be reduced to a fundamental minimum, simply because the sophisticated detector is able to record literally every single photon emitted by the probe or sample, respectively. This promises to pave the way to a new performance category of photon detectors which could be exploited as focal plane arrays for multispectral microscopes. However, despite the obvious advantages, a substantial risk for a successful commercialization arises from the need to achieve a proper balance between application benefit and technical complexity. Again, ESQUID technology can act as an enabler for such focal-plane instruments with a high spatial resolution.
In the future, such instruments could primarily benefit medical research in the context of analyzing the interaction of biological or chemical agents with living cells. Because such interactions potentially can be studied with minimal interference by the observation technique, best possible significance can be achieved. That becomes especially advantageous for the development of tailored medicinal products which treat diseases at a cellular level. With such focus, the technique will be used rather by a few medical research centers than by doctors at standard hospitals. Although that limits the potential market, the social impact, nevertheless, is hardly less significant. In an ageing society, the treatment of diseases becomes more and more important. With each individual advancement of therapeutic procedures, ethical and economic advantages arise from improved well-being and smaller health care costs, respectively. That motivates the research into such high-end technologies.
University of Leicester and IPHT already have agreed to collaborate in such direction and have propsed a research project in frame of a bilateral UK/German project (Biophotonics Plus) which plans exploiting and enhancing parts of the ESQUID technology.
There are also niche applications in microcalorimetry for material analysis based on X-ray spectroscopy, and nuclear isotope analysis, based on gamma-ray spectroscopy. The results of this project can be used to develop detectors dedicated for such applications.
To assess the broader impact of the project in scientific context, one must include the time scale, to which one is referring to. E-SQUID is intended for space applications, which have a rather long lead time. The devices that have been produced within this project are intended as demonstrators for instruments with a launch date in the 2020’s or even 2030‘s. It is clear that the space telescopes in which the devices are intended to fly, will open new windows for astronomy, and therefore produce scientific papers and newspaper articles.
This technology is already used in modern ground-based astronomical instruments such as SCUBA-2 at James Clerk Maxwell Telescope (Hawaii). The present state-of-the-art does not, however, fulfill the requirements of the next generation of astronomical instruments.
The E-SQUID project results will address several future needs in astronomical research. Application of the results of the E-SQUID project is in the plan for the X-ray micro-calorimeter, X-IFU, of ATHENA (L2 mission of the ESA, current launch schedule 2028), is foreseen applicable for the SAFARI instrument for the infrared space mission SPICA (JAXA mission with ESA contribution, current launch schedule 2025), and is also possible for the next generation ground-based far infrared experiments of ALMA (ESO, in the 2020’s). (Figures 1, 2, and 3).
The X-IFU (X-ray Integral Field Unit) will be an advanced, actively shielded X-ray microcalorimeter spectrometer for high-spectral resolution imaging. The detector will be a large array of absorbers read out by transition edge sensors (TES). The TES microcalorimeter senses the heat pulses generated by X-ray photons when they are absorbed. The minute temperature increase occurring with the incident photon energy is measured by the sharp change in the electrical resistance of the TES. This must be cooled to a temperature of less than 100 mK and biased into its transition between superconducting and normal states.
SAFARI will cover the FIR window that extends from ~30 μm (the upper cut-off of the MIR instruments) to ~210 μm (just long ward of the [NII] 206 fine structure line) with a field-of-view of 2'x2'. Assuming diffraction limited performance, SAFARI will provide angular resolutions from ~2" to 15" (20 to 150 AU at 10 pc) at wavelengths not covered by JWST and at more than 2 orders of magnitude higher sensitivity than Herschel/PACS. This huge increase in sensitivity could potentially open EP research to wavelengths completely blocked by the Earth's atmosphere, but representing the emission peak of many cool bodies (gas-giant planets, asteroids and so on). SAFARI will have its major strength in measuring excess radiation from dusty proto-planetary disks in hundreds of stars at almost all galactic distances. It will also perform medium spectral resolution observations (R~2,000) over a spectral range also rich in dust features, water vapour rotational lines (temperatures below ~500 K), atomic oxygen fine structure lines at ~63 μm and the solid state water-ice features at ~44 and ~62 μm.
The baseline optical configuration of this European instrument is a Mach-Zehnder imaging Fourier Transform Spectrometer (FTS). The principal advantages of this type of spectrometer for ESI are the high mapping speed of the FTS due to spatial multiplexing, the ability to incorporate straightforwardly a photometric imaging mode and the operational flexibility to tailor the spectral resolution of the science programme. The detector technology candidates of the insturment are Si:Sb and Ge:Ga photoconductors, transition edge sensor (TES) bolometers and kinetic inductance detectors (KIDs).
List of Websites:
http://fusion.gfl.helsinki.fi/esquid/
Coordinator: Dr Juhani Huovelin, Department of Physics, P.O.Box 48, FI-00014 University of Helsinki, Finland.
One of the main questions of science is the early phases of the evolution of our Universe and the evolutionary scenarios after the Big Bang. The signature of this evolution is imprinted in the electromagnetic radiation from the deep space. A detailed visibility to the first billion years in the evolution requires instruments capable of imaging and spectroscopy with an extremely high sensitivity, since the science requires ultimately precise measurements of radiation ranging from X-rays originated in the vicinity of the super-massive black holes in the nuclei of the first galaxies to infrared radiation by the cool gas in the interstellar space up to distances of more than 10 billion light-years.
In that context, E-SQUID project drives progress in the development of versatile multiplexed SQUID (Superconducting QUantum Interference Device) readout of Cryogenic Detectors with the main exploitation focus in space instruments, with also significant potential for ground-based applications. SQUIDs can be designed for optimized readout of X-ray micro-calorimeter arrays, as well as similar detectors for longer wavelengths extending to far infrared/sub-mm wavelengths. The most competitive technology in this area is presently available from non-European sources. A core objective of E-SQUID is to upgrade the performance of European solutions to be competitive with the best achievements in the world.
The practical task of the E-SQUID project was to develop prototypes for next generation multiplexed SQUID readout schemes of X-ray and far-infrared Cryogenic Detector arrays, modifications of which may then be used for optical wavelengths. The improvements were performed in several steps, ensuring the validity of each improvement by tests and performance characterization, and moving to next phase using improved guidelines provided by the test results. The goal is to find an optimal scalable multiplexing method for each wavelength that enables future design and developing of large detector arrays with minimum noise characteristics.
Imaging detectors that are based on Transition Edge Sensors (TES) form the core of future space-based telescopes for X-ray and infrared astronomy. SQUID technology is essential to obtain detector noise limited readout performance of these TES based detectors. Mass, complexity and power consumption constraints in space-based instruments form the main driving force behind the development of efficient multiplexing schemes.
In the E-SQUID project new SQUIDs have been developed for use to read individual detectors, as well as new improved SQUID multiplexers for reading multi-pixel detector arrays. From the possible alternative multiplexing schemes, frequency domain multiplying (FDM) turned out the most promising concept for space applications. The usefulness of code domain multiplexing (CDM) could not be demonstrated in E-SQUID project, but the results indicate that further work on this may lead to success as well.
It was found that the developed SQUID multiplexers are well suited for the required dimensions of the TES detector arrays for two future space telescopes: the XIFU instrument on ATHENA X-ray telescope, and the SAFARI instrument on SPICA infrared telescope. The use for future instruments of ALMA is also possible.
The technology developed in E-SQUID can also be utilized for the development of a number of important ground-based applications, such as passive imaging of concealed weapons, and measurements of fragile biological samples and processes, such as the interaction of biological or chemical agents with living cells.
Project Context and Objectives:
A major task for modern astronomy is the examination of the early Universe and the evolutionary scenarios after the Big Bang. The signature of this evolution is imprinted in the electromagnetic radiation from the deep space, including the ubiquitous cosmic microwave background and the faint glow from stars and galaxies. A detailed visibility of this multi-billion-year tale requires instruments capable of imaging and spectroscopy with an extreme level of sensitivity, since science requires ultimately precise measurements of radiation ranging from X-rays originated in the vicinity of the super-massive black holes in the nuclei of the first galaxies to infrared radiation by the cool gas in the interstellar space up to distances of more than 10 billion light-years.
In that context, E-SQUID project drives progress in the development of versatile multiplexed SQUID (Superconducting QUantum Interference Device) readout of Cryogenic Detectors with the main exploitation focus in space instruments, with also significant potential for ground-based applications. SQUIDs can be designed for optimized readout of X-ray micro-calorimeter arrays, as well as similar detectors for longer wavelengths extending to far infrared/sub-mm wavelengths. The most competitive technology in this area is presently available from non-European sources.
The high level objective of E-SQUID is to upgrade the performance of European solutions to be competitive with the best achievements in the world.
This technology is already used in modern ground-based astronomical instruments such as SCUBA-2 at James Clerk Maxwell Telescope (Hawaii). The performance and array sizes of presently available solutions are, however, not sufficient for the scientific aims of future instruments like the X-ray micro-calorimeter for ATHENA (L2 mission of the ESA), the infrared space mission SPICA (JAXA/ESA), and next generation ground-based far infrared experiments for ALMA (Atacama Large Millimeter Array). Hence, the development of improved Cryogenic Detector systems with larger arrays and smaller noise are very important to the future of European space science. In addition to space science, the technology developed for E-SQUID can be utilized for the development of a number of important ground-based applications, such as passive imaging of concealed weapons, and measurements of fragile biological samples and processes, such as the interaction of biological or chemical agents with living cells.
For setting up the scientific requirements for the photon detectors of each energy range, we have selected the following reference instruments. They are APEX telescope (Chajnantor/Chile) at far infrared/Submm, S-Cam 3 (La Palma/Canary Islands) at optical, and the ATHENA (space) at X-ray range. With the aimed performance, the reference instruments can provide observations of the cold interstellar dust enabling to trace the history of star formation (infrared), multicolour observation capability of the exoplanet transits across the parent stars (optical), and black hole accretion disk observations providing information on space-time structure of the black hole event horizon (X-rays). In parallel, the requirements serve also the more general aim, which is to develop larger ultra-sensitive Cryogenic Detector arrays with smaller noise for sensors that can be used for space astronomy and ground-based applications.
The practical task of the E-SQUID project is to develop prototypes for next generation multiplexed SQUID readout schemes of X-ray and far-infrared Cryogenic Detector arrays, modifications of which may then be used for optical wavelengths. The improvements are performed in several steps, ensuring the validity of each improvement by tests and performance characterization, and moving to next phase using improved guidelines provided by the test results. The goal is to find an optimal scalable multiplexing method for each wavelength that enables future design and developing of large detector arrays with minimum noise characteristics. The signal multiplexing means a reduced amount of signal lines between the Cryogenic Detectors and room temperature main electronics, which enables, in practice, the use of considerably bigger detector arrays, and also reduces electrical disturbances and heat load.
The power dissipation of cryogenic detectors is typically extremely small. An example is the 6 pW / pixel power dissipation for the breadboard TESs for the International X-ray Observatory (Athena) . The biggest power input, and heat leakage, to the lowest temperature part of the detector is via the electrical harness to drive and read-out the detectors. Multiplexing techniques are used to minimize the required number of wires for reading large arrays, and thus their development is a crucial task.
The only read-out amplifier that has so far been proven to be suitable for the above approach is a SQUID-based amplifier next to the detector.
Therefore, our first objective is to use our technological skills to develop cryogenic SQUID amplifiers whose figure-of-merit is improved to bring their performance closer to the thermodynamic limit.
Moreover, to realize an effective multiplexing scheme, a sizable number of these first stage amplifiers need to be integrated into a single output channel, using an appropriate addressing method i.e. time (TDM), digital (CDM) or frequency (FDM) encoding.
Therefore, the second objective of E-SQUID project is to select the multiplexing method which is most suitable for the chosen approach of SQUID readout of cryogenic detectors. The selection will be made on the basis of considerations regarding achievable multiplexing level (e.g. number of merged channels), technology readiness level (TRL) and complexity of adjacent electronics.
Depending on that selection, a full set of specifications regarding the setup of various cryogenic detector arrays can be derived. It seems possible to build up a universal architecture based on these specifications, using a sophisticated SQUID layout and adapted room temperature electronics.
Therefore, the third objective of E-SQUID projectl is to develop both cryogenic and room-temperature sections of a complete multiplexing electronics unit, which we aim to standardize across the two major European superconductive foundries, IPHT Jena and VTT.
Project Results:
Concepts and Readout Methods of detectors
Space X-Ray and IR missions
Introduction
Imaging detectors that are based on Transition Edge Sensors (TES) form the core of future space-based telescopes for X-ray and infrared astronomy. SQUID technology is essential to obtain detector noise limited readout performance of these TES based detectors. Mass, complexity and power consumption constraints in space-based instruments form the main driving force behind the development of efficient multiplexing schemes.
As the prototypical space missions we have taken as the project guidelines, are the X-IFU instrument for the ATHENA+ mission and the SAFARI instrument for the SPICA mission. The planned ATHENA+ mission is an orbiting X-ray telescope, whose imaging capability comes from silicon pore x-ray optics with ~2 m2 collecting area. The X-IFU instrument will perform energy dispersive single photon counting, i.e. capture each incident x-ray photon in the 0.3 – 12 keV range with 70% quantum efficiency at order-of 100 cps count rate, and resolve energy of each photon with order-of 2.5 eV resolution. The planned SPICA mission is an orbiting far-infrared telescope, with a 3.5-metre antenna dish, actively cooled to 5 K temperature. The SAFARI instrument is effectively a ‘crystal set’ radio receiver for 34 – 210 m wavelengths, equipped with Fourier Transform Spectrometer and imaging capability. The SAFARI receiver sensitivity is going to be limited by the photon noise of the 2.7 K blackbody background, which implies order-of 2×10-19 W/Hz noise-equivalent power (NEP).
Although missions seem very different, the same detector technology can be used in both, namely absorption of the radiation and measuring the temperature change in the absorber. The sensitivity requirements dictate operation of detectors at a very low temperature, order-of 50 mK, and a very sensitive thermometer, for which a superconducting phase transition is used. In the SAFARI instrument the detectors are operated in bolometric mode, meaning that they react to the average incident radiative power, whereas in the X-IFU instrument they are operated in calorimetric mode, i.e. each incident photon causes a temperature excursion. As a consequence, bolometers of the SAFARI instrument always operate close to thermal equilibrium, implying that the work well with Frequency Domain Multiplexing (FDM, see below), whereas calorimeters of the X-IFU instrument spend a large fraction of their time out-of equilibrium. This difference has consequences on the multiplexing method.
Operating principle TES-based detectors
Imaging TES-based detector arrays consist of a matrix of pixels. A pixel consists of a thermally isolated island equipped with an absorber, and a thermometer. When the optical signal from the source gets absorbed in the absorber, its energy is transferred into heat. The transition edge sensor senses the resulting increase in temperature, which is a superconductor in its superconducting phase transition. Electro thermal feedback stabilizes the detector in an operating point in the transition.
During operation, the detector is in a dynamic equilibrium, i.e. its temperature is stabilized above the thermal bath temperature, which is set by the critical temperature Tc of the TES by dissipation in its resistance. As a result, a constant heating power is needed to keep the pixel in its operating point. When an optical signal gets absorbed, less heating current is needed to maintain the temperature. The latter change in heating current forms the signal.
In the microcalorimeter operating mode, the detector is dimensioned such that it responds to individual photons, which makes the detector energy dispersive. In the bolometer operation mode, the detector is too slow to deconvolve individual photon absorption events. As a result, in this mode the detector is not energy dispersive, but only measures the average power of the applied light.
Multiplexed readout of TES-based detectors
Multiplexed readout as opposed to direct readout of TES-based cryogenic microcalorimeters and bolometers is mainly driven by the goal of minimizing the required cooling capacity and complexity at the cold stages. The main sources of heat load in the readout system are SQUID current amplifiers, and the heat leak through the bias and readout wiring.
Without any multiplexing, each pixel would require a single wire pair for its voltage bias, and three wire pairs for the SQUID trans-impedance amplifier. Multiplexed readout implies that multiple pixels are readout by a single SQUID amplifier, and multiple TES bias voltages are fed through one wire pair.
Rationale for SQUID readout
TES-based detector readout can be seen as monitoring the resistance of a temperature dependent resistor, which value is a function of the applied signal power. In order to benefit from electro-thermal feedback, the monitoring is done by applying a constant voltage, and reading the temperature-dependent current.
To comply with the required voltage bias condition, a current amplifier is needed with an effective input resistance that is significantly smaller that the set point resistance (typically R0=40mΩ) of the TES. In addition to that, electro-thermal stability requires that the electrical inertia of the bias circuit should be approximately a factor of 6 smaller than the thermal inertia. This limits the maximum input reactance of the current amplifier at the bolometer bandwidth to 6* R0 .
Because of the stringent science requirements, it has been chosen that the readout circuit is allowed to consume ~10% of the NEP budget. This implies that the required effective noise temperature of the current amplifier should be at least one third of the effective noise temperature of the bolometer, taking into account that the detector noise and the amplifier noise are uncorrelated. The latter noise temperature is approximately equal to the operating temperature, i.e. ~100mK. Hence, the effective noise temperature of the current amplifier has to be smaller than ~30mK. This requirement, combined with the input impedance requirement, rules out the use of semiconductor amplifiers, and leave SQUID current amplifier readout as the only viable option.
Multiplexing of TES-based bolometers
Multiplexing of TES-based bolometers implies using the wire pairs and SQUID amplifiers of a single bolometer for the independent readout a larger number of bolometers. Consequently, the bolometer signals need to be added on a single wire pair, on which then several bolometer signals are transmitted simultaneously. Note that simple addition of the signals is not acceptable, as the signals overlap in time, frequency and phase space, and would therefore become indistinguishable.
Because of the stringent science requirements, there is virtually no headroom in the NEP budget to assign to information loss as a result of multiplexing. As a consequence of that, the multiplexing scheme has essentially to be lossless, and cross talk-less. In text books it can readily be found that any lossless multiplexing scheme, i.e. a scheme in which the constituent signals need to remain fully recoverable, requires two fundamental steps before the signals can be added while remaining (mathematically) independent. The first step is to confine the signal bandwidth, and the second step is to transform the signals (“fingerprinting”) in such a way that they become mathematically independent. The latter mathematical transformation implies the multiplication of the time-dependent signal functions of the pixels with an independent carrier function per pixel, in such a way that they become orthogonal and therefore mathematical independent and fully distinguishable.
As a result of the mathematical transformation (i.e. the “multiplexing”), N multiplexed signals require a bandwidth which is at least N times the bandwidth of a single signal, depending on the chosen set of carrier functions. As will become clear below, SQUID amplifiers can provide sufficient power gain over bandwidths of several MHz On the other hand, TES-based detector pixels require a bandwidth of the order of 1 kHz per pixel. Therefore, from bandwidth point of view, SQUID amplifiers are in principle capable of accommodating the signals of at least 102 to 103 TES-based pixels.
Code domain multiplexing in TES-based detector arrays
The low operating temperature implies that extremely small power dissipation can be tolerated. Every nanowatt dissipated in the 50 mK stage of the refrigerator must be lifted to higher temperature where the radiator of the spacecraft reside. This means that every dissipated nanowatt must be multiplied at least by Carnot efficiency, in practice by the much poorer actual efficiency of the refrigerator, to obtain the heat generation at the S/C radiator. This is a particularly stringent constraint in space, where cooling water is not available, but the heat must be radiated away.
Two additional problems – Shannon capacity and need for multiplexing - arise from the fact that both SAFARI and X-IFU are imaging instruments. SAFARI will contain 61 x 61 detector pixels in 34-60 μm band, 34 x 34 pixels in 60-110 μm band, and 18 x 18 pixels in 110-210 μm band, whereas X-IFU is planned to contain 3840 X-ray detector pixels.
Shannon channel capacity indicates the amount of information flowing through the amplification and data acquisition chain, the capacity equalling signal bandwidth times the dynamic range, and being expressed as bits per second. If one X-IFU pixel generates 500 kbit/s information flow, the full 3840-pixel instrument will generate 2.6 Gbit/s, assuming the flow cannot be compressed. Although it is not known whether the connection is fundamental – in analogy with the Landauer’s principle – high Shannon capacity seems in practice to go hand-in-hand with increased power dissipation in amplifiers and with increased heat leakage through cabling.
Multiplexing implies that signals from several pixels are encoded and summed into a single cable. It would be impractical eg. in X-IFU to have 3840 dedicated cables from the 50 mK refrigerator stage to room temperature, and a dedicated digitizer for each pixel. Multiplexing entails an orthogonal set of encoding functions, with each signal being multiplied by one member function, before the signal are summed together. Three basis sets, sinusoids for Frequency Domain Multiplexing (FDM), boxcar functions for Time Domain Multiplexing (TDM) and Hadamard codes for Code Domain Multiplexing (CDM), offer certain technical advantages and have been pursued in the past. One then needs a cryogenic multiplying element, capable of multiplying the detector signal with a basis function (Figure 4).
In FDM, the detectors can be biased by the sinusoids forming the basis set, so that Ω’s law performs the effective multiplication of the bias voltage with the detector conductance. This works well in SAFARI-like bolometers, which remain internally close to thermal equilibrium and perform the multiplication smoothly. X-IFU –like calorimeters, in the other hand, had been observed to suffer from degradation of the energy resolution before the years 2009 when the E-SQUID project was planned. The degradation was suspected to be of fundamental origin. Hence, the CDM was originally targeted in the E-SQUID project, with the emphasis on implementation of the Hadamard-encoding elements.
As Hadamard functions are two-level, a multiplication by +1 and -1 is sufficient in CDM. This can be implemented by a commutating switch. We have studied such switches implemented with Josephson junctions, in which case the periodic switch response due to quantum interference brings in an additional advantage: replication of so-called Sylvester’s construction of Hadamard matrices in hardware (Figure 5). This opens a way to address N pixels with log2N address lines.
SQUID-like flux-controlled two-junction current steering switches (CSSw’s) require a very clean addressing current, because fluctuations in address current couple as noise to the detector signal. We have instead studied CSSw’s based on three-junction Zappe interferometers (Figure 6). They have almost square-shaped flux response, as opposed to sine-like response of two-junction SQUIDs, hence the address current fluctuations have a much lower effect.
Resistive CSSw’s require the low-value resistor Rq (Figure 6 b) which are (i) hard to implement and (ii) suspected to cause problems with SRON calorimeters. Tens of nanovolts of voltage drop can be tolerated across the switch, which implies Josephson oscillation at tens of MHz, which is disturbingly close to the information band of the calorimeters. Therefore, we have studied inductive CSSw’s which always remain in the superconductive state (Figure 6 c,d).
Dynamic range of current-steering switches easily becomes the dominating bottleneck. There are several effects, one of which is: a large current handling capacity implies a small loop inductance, which makes loop dimensions to approach the Josephson junction diameter, which subsequently creates a Fraunhofer envelope (Figure 8), in analogy with the optical double-slit interferometric effect. When periodicity is lost, the log2N addressing is no longer feasible.
There is the alternative CDM method, the so-called flux-summing technique (Figure 7). This is a moderately interesting technique, however having two disadvantages: (i) only sqrt(N) –proportional reduction in the number of cables can be reached easily, and (ii) each readout SQUID must handle an N-fold dynamic range, because signals from all N pixels are summed at SQUID inputs.
Due to these observations, and encouraged the experimental and theoretical results from SRON in late 2013 that FDM may still be feasible with calorimeters without degraded energy resolution, project emphasis was shifted towards FDM.
Frequency Domain Multiplexing in TES-base detector arrays
Basic operating principles of FDM
As already mentioned above, multiplexing involves mounting a bandwidth-limited signal on a carrier function by means of multiplication (modulation), followed by summing of the modulated signals for transportation through a shared channel. In the SRON FDM implementation, the orthogonal set of carrier functions consist of sinusoidal TES bias voltages of different frequencies. The signal-dependent TES resistance amplitude-modulates the resulting bias current. The required bandwidth limitation, and separation in frequency space is provided by LC band pass filters, which consist of inductors (L) and capacitors (C). A schematic diagram is shown in Figure 9. The filtered signals are summed in a current summing point, with a finite but low impedance Zc, which schematically represents the SQUID current amplifiers input impedance. The SQUID amplifier chain is not indicated for simplicity.
For cross talk minimization reasons, the summing point (or “common”) impedance needs to be low with respect to the TES resistance as it is a common electrical circuit element between the bias circuits of the pixels, and for that reason provides a route for cross talk. Other sources of cross talk in FDM are finite out-of-band damping of the band pass filters, nonlinearity in the amplifier chain, and inductive coupling between the LC filters, which also induces currents in neighboring bias circuits like common impedance does.
Figure 10 shows the practical implementation of FDM in more detail. The readout chain consists of three main building blocks, i.e. the cold electronics, the front-end electronics, and the digital electronics. The cold electronics provides the TES-based detectors with LC filters and the first stage amplification by means of SQUIDs. The front-end electronics consists of the cable connection to the room temperature electronics, and the analog signal and buffer amplifiers, as well as bias sources. The digital electronics provides the demodulation functionality, the sinusoidal (AC) carrier generation, and dynamic range enhancing feedback signals for the SQUID amplifiers by means of baseband feedback (BBFB).
There are many aspects of this design, which require more detailed design motivations and discussions. However, these issues are outside the scope of this document. Instead, in the following subsections the design concepts for the building blocks will be briefly touched, to summarize the context.
Multiplexing factor and frequency space assignment
The multiplexing space is determined by the available bandwidth and the minimum allowable separation in frequency space. There are several practical factors influencing these constraints, which are listed below.
1. Component size. The physical dimensions of the lithographically produced LC filters scale inversely proportional with their center frequency, i.e. fcenter α 1/(Area)2. Taking into account the dimensional constraints of the focal plane arrays (FPA), in combination with the production constraints, this gives a practical lower limit of 1 MHz.
2. SQUID bandwidth. The noise temperature of a SQUID is proportional to the bandwidth, which for the bolometer requirements results in an upper limit to the available bandwidth of 3 - 7 MHz, depending on the SQUID dimensions.
3. Cable bandwidth. An ideal (lossless) transmission line has an infinite bandwidth. Practical (lossy) transmission lines attenuate the signals. Thermal considerations dictate a lower limit to the cable resistances, which limit the usable bandwidth of a cable. Practical cables of these dimensions show a cable pole around 5 – 10 MHz, depending on the available cooling power around 4K, and the length of the cable. Note that below 4K superconducting wires can be used, which are virtually lossless for the signals, but isolating for thermal loads.
4. Dissipation in the digital electronics. The dissipation in the digital electronics scales proportional with the operating frequency. The highest required operating frequency is proportional to the highest frequency in the FDM band. The available power budgets for the digital electronics currently yield the upper limit (typically 3 – 8 MHz).
5. Baseband feedback stability. The baseband feedback gain-bandwidth-product (GBP) is typical set at 2 – 3 times the electrical bandwidth of a pixel, as a result of dynamic range and linearity considerations. The GBP sets the minimum band separation at approximately 10 times the GPB kHz to guarantee stable operation.
6. Cross talk. Finite damping in LC band pass filters leads to finite leakage of power neighboring pixels, inverse proportional to their distance in frequency space. Given the cross talk requirements, this sets the ratio between minimum distance between pixels and the electrical bandwidth per pixel at >6.
7. Dynamic range. The higher the multiplexing factor, the higher the dynamic range requirements for the SQUID amplifier chain, and the carrier generation DAC's and the feedback DAC's. The availability of space qualified digital-to-analog converters limit the number of channels to approximately 160 for bolometer applications, and 40 for micro calorimeter applications. Note that this limit could be circumvented by using adding the signals of a larger number of DACs. However, the latter is constraint by power limitations.
Combining all factors has for the bolometer system design led to the choice of a multiplexing factor of 160 pixels/SQUID channel, with a band separation of 12.5 kHz, and a lower and upper operating frequency of 1 and 3 MHz, respectively. For the micro calorimeter case the multiplexing factor has been set at 40 pixels/SQUID channel, with a separation of 100 kHz between pixels, and a frequency range between 1 and 5 MHz.
Construction of the SQUID amplifier chain
The function of the SQUID amplifier chain is to amplify and transport the TES detector signals to the room temperature electronics over the cable harness. As the cable harness is exposed to electromagnetic interference (EMI), the signal power levels need to be above the EMI thresholds, to ensure signal integrity. Simultaneously, there is a minimum energy resolution requirement at the input of the readout chain, as set by the noise temperature of the detectors, in combination with cross talk requirements.
It turns out that these requirements can be met best by a two-stage SQUID system, where the first stage is sitting close to the detector, and is optimized for an optimal energy resolution in combination with maximum power consumption. The second stage is sitting at an elevated temperature (~2K) where more cooling power is available, so that more powerful SQUIDs can be used to drive the cables to room temperature.
Ground Based FIR telescopes
The new generation of large FIR/Submm telescopes is designed from the start to have a large field of view (FOV), in contrast to most older telescopes. The speed for mapping large areas of the sky is proportional to the number of detectors. Therefore in order to take advantage of the large field of view, one would like to fill it with detectors, eventually. Even if this might not yet be feasible today, detector arrays with thousands of pixels will be required to cover at least a substantial part of the telescope field of view. As continuum detectors in this spectral range need to operate well below 1 K for optimum sensitivity, reading out a very large number of such detectors is only feasible via cryogenic and multiplexed readout techniques.
APEX, the Atacama Pathfinder Experiment, is one of these new generation telescopes. The site on the Llano de Chajnantor at 5100 m altitude in Chile has excellent atmospheric transmission. The surface quality of the reflector is very high and there is a large field of view at the Cassegrain focus (Figure 11). In the following, instruments for APEX are taken as reference, because of our close association with APEX and our detailed knowledge about its instruments. However, the conclusions apply to any FIR/Submm telescope with a large field of view, that is situated on a good site.
The integration time for mapping a given area to a specified depth is inversely proportional to the number of pixels. On the other hand, if integration time and mapping area are fixed, the minimum detectable flux is inversely proportional to the square root of the number of detectors. Therefore care must be taken not to degrade the number advantage by lower pixel sensitivity or higher readout noise, as compared to a direct readout of fewer pixels.
APEX is a freestanding telescope, not protected by a dome. Even so, the surface quality of the main dish is very high (15 microns rms). This and the quality of the site make it uniquely efficient for observations at wavelengths as short as 350 microns.
The FOV at the Cassegrain focus of APEX is limited by the central hole in the primary to 0.5 degrees (30 arcmin) in diameter. LABOCA, the large bolometer camera for APEX is an example for a FIR/submm astronomical camera. It operates at a wavelength of 870 microns and covers a field of view of 0.2 degrees (12 arcmin) with 300 pixels. The existing LABOCA has semiconductor bolometers and operates in a liquid Helium cryostat. In the following, this camera will be referred to as LABOCA-1, to distinguish it from the superconducting version LABOCA-2. In both LABOCAs all pixels are antenna coupled and fully efficient at a spacing of 2 f l, where f is the focal ratio at the final focal plane and l is the wavelength. Some further gain could be obtained by changing to small Nyquist sampling pixels (at 0.5 f l), but only at the cost of increasing the pixel number by a factor of 16! At the present state of the art, extending the field of view with fully efficient pixels seems a more effective use of resources. Even in this concept there is plenty of room for growth in pixel numbers.
The spectral band-pass is defined by means of cold composite filters, a quarter wavelength back short and a short (2 x diameter) piece of cylindrical waveguide at the exit of the conical horn. The conical horns couple the radiation to a micro machined array of TES detectors on a hexagonal grid with a grid constant of 4 mm. Each detector has lossy crossed dipoles as absorbers on a structured silicon nitride membrane and a TES consisting of a bilayer of Mo/Au-Pd. The whole detector array is fabricated on a single 4-inch wafer with on-array wiring to the out-of-plane multiplexers. The multiplexers are set up as multi-chip modules with superconducting bonds to the array. The focal plane is cooled to about 300 mK by a He-3 sorption cooler.
FIR/Submm detectors on ground based telescopes have to cope with a substantial and variable background power load, which restricts the achievable detector sensitivity because of photon noise. This implies a relaxation of the cooling effort since adequate detector performance can be achieved at relatively high operating temperatures, like 300mK.
The ultimate goal for the multiplexer architecture is to integrate as many pixels as possible into one readout chain. Although the single-pixel noise requirements are relatively relaxed because of the background limited detector performance, the implementation of a large number of multiplexed channels (>> 10), while avoiding multiplexer losses, constitutes a remarkable technical challenge.
Conventional TDM’s have the fundamental problem of increasing the current noise of N multiplexer SQUIDs by Sqrt[N] because of aliasing. Since the SQUID noise is typically one order of magnitude lower than the intrinsic detector noise, lossless TDM could be achieved by low pass filtering the detector noise at half the MUX sampling rate. However such a filter between the MUX chip and the TES will be difficult to implement because of the impedances involved. For N about 30 this would still work, while fore larger multiplexing levels the SQUID noise becomes an issue. Known concepts for overcoming this problem are coded (CDM) or FDM. FDM of far-infrared TES architectures is not trivial because the required modulation of DC detectors with an AC bias in a narrow frequency band constitutes a technical challenge which cannot be easily solved. Especially the need for cryogenic high frequency filters raises a purely technological problem of manufacturing the required large capacitances.
In CDM, the detectors that are connected to one multiplexing chip are not read out sequentially, like in TDM. Instead all detectors are sampled continuously, while the polarity of the signal from each detector is cycled according to a sequence of orthonormal unit step functions (Hadamard- or Walsh-functions). Consequently, coded multiplexing seems the method of choice, where the SQUID noise aliasing disappears due to the use of the full time interval for all channels instead of only single time slots in the TDM case. A scalable CDM approach was therefore chosen for the E-SQUID project with respect to FIR/Submm applications on the ground.
The preferred astronomical mapping scheme is to use the telescope as scanner, that means moving continuously over the area of interest, while taking data. Multiple passes will then result in a Nyquist sampled map. This mapping scheme, puts several demands on the pixel performance:
1. A single detector should achieve an electrical NEP below the estimated photon noise of background. This performance must not be compromised by the multiplexer, especially in the data bandwidth of the astronomical signals between 0.1 and 10Hz.
2. Crosstalk between pixels should be less than 1%
3. Extended emission in the mapped area will be transformed to low signal frequencies by the scanning motion of the telescope. Therefore detectors should have almost no 1/f noise above 0.1Hz.
Since the 2 f l spacing between neighboring pixels allows for extended wiring between pixels, a multi-chip approach with multiplexers out-of-plane is a straightforward and reasonable approach. Conceptually, multiplexer chips will be mounted alongside the detector chip, connected electrically with superconducting bond wires. To ease the implementation, this multi-chip approach implies the setup of multiplexers at the same operation temperature as the detector. This has consequences for the multiplexer design (critical SQUID parameters) and the dissipation balance. The power load of the multiplexer itself is almost negligible.
Second stage amplifiers are needed, typically SQUID arrays for pure amplification or Fourier SQUID arrays (SQIF) for single valued response. These are installed at a higher operation temperature around 1K on the next cooling stage. In that way, no notable direct thermal load on both cooling stages is expected, as long as the wiring between MUX and amplifier is designed properly.
Optical
In optical wavelength areas, the basic idea was to utilise the single photon counting SQUID technique similar to X-Ray region. A decision was made to replace Metallic Magnetic Calorimeters (MMC) with Superconducting Tunnel Junction (STJ) detectors. However, the SQUID readout for multiplexing of the STJ would require chips with dimensions that are too bulky to be housed in the cryostat at the University of Leicester (ULEIC) and would also be in general impractical for large number of pixels. Therefore the SQUID readout for multiplexing the STJ has been abandoned. A readout based on silicon germanium transistors and working at cryogenic temperature was considered as a solution. Amplifiers with performance that would fit the requirement for optical photon detection have been developed by VTT, and were available for a test of a single pixel read-out. However, synergy could not be maintained anymore with the sub-mm and X-ray mission approaches.
Design considerations
Vital goal of the E-SQUID project is the achievement of fabrication compatibility between the two major European foundries for superconducting electronic circuits: VTT/Finland and IPHT/Germany. Thus, in view of the potential dissemination of E-SQUID devices for future space missions, both foundries can act as supplier to European space programs with the crucial advantage of an available back-up in case of unforeseen technological or organizational problems.
At the start of the project, such compatibility was not pre-existing. Although general aspects of superconducting technology for SQUID-based devices were used at both sites, application demands from antecedent projects have driven the foundries to different implementations. Moreover, since the intended realization of code-division multiplexers would potentially benefit from an on-chip digital control, IPHT has advocated the consideration of RSFQ as a promising technology, further increasing the requirements on a common technology.
In the design phase, mainly the two partners concerned (VTT and IPHT) have worked together on a common set of design rules. The work based on existing foundry processes at VTT (MRI-MEG) and IPHT (RSFQ 1F). The following basic requirements have been agreed as starting point:
• three superconducting metallization layers, allowing for moderately complex designs
•
• projection lithography, allowing for high-density designs
•
• resistors compatible for operation at temperatures below 1 Kelvin
•
• width of metallic lines smaller than 7 micron, facilitating spontaneous flux expulsion (prerequisite for large coherent SQUID arrays)
•
At the same time, the partners agreed to maintain differences in their processes, which otherwise would imply too far-reaching changes with unpredictable risks for the respective foundry:
1. IPHT uses an electro-chemical oxidation procedure (anodisation) in order to passivate niobium structures before the deposition of insulation layers (SiO2), whereas VTT uses only SiO2 films. At the design level this implies the requirement of a network of auxiliary connection between all niobium structures to be anodized for the IPHT version of a common process, which are unnecessary for the VTT implementation.
2. For the formation of Josephson junctions, IPHT deposits a specific trilayer package to the junction regions, whereas VTT etches junctions out of a trilayer which covers the whole processing wafer. The thorough solution to handle that difference was to define a common circuit primitive (Josephson junction definition and contact windows) as exchangeable design cell, which can be expressed differently for the respective fabrication processes.
Another crucial difference in the processes turned out to be intolerable: In order to handle the problem of insufficient edge coverage in wire-crossing steps, VTT is using a chemical-mechanical planarization to level the subjacent wire and the isolation film, whereas IPHT relies on solely using material-additive techniques, i.e. larger thickness of the overlying wire. The resulting thickness incompatibility could have been acceptable in purely-SQUID designs, but for digital logic circuitry, the accurate sheet inductance mainly defined by the distance between two wires becomes crucial. In simulations at IPHT it was determined that design margins for digital logic would become prohibitive if the insulator thickness variations persist between the VTT and IPHT processes.
In consequence, according process development was initiated at VTT. A new proprietary technique of selective insulator deposition at the steps was developed, which resolves the step-coverage problem without chemical-mechanical planarization (see Figure 12). With this novel technique, the thicknesses of the insulation between the wiring layers became comparable between VTT and IPHT process, allowing for the use of similar RSFQ designs.
The described design variances have been incorporated into a script for layer operation, which automatically translates a design from the originating foundry into the respective version of the other. With this script, designs both from VTT and IPHT have been successfully transferred to the respective builds within the frame of the E-SQUID project.
Fabrication
The original work plan scheduled a regular sequence of 4 alternating fabrication rounds (so called “builds”) at the foundries (VTT: Build 1&3, IPHT: Build: 2&4). In the early project phase, technical difficulties caused a considerable delay: whereas at VTT a major etching tool was under repair, at IPHT the resumption of regular operation after a recent full reconstruction of the clean room facility was still in progress. In view of potential risks and in order to minimize the delay, a change of the sequence was decided. The first build started at IPHT late 2011; however, because of the still unstable clean-room operation that build failed. A successful replacement build was fabricated between February and April 2012. That build mainly contained technology tests and building blocks of code division multiplexers and RSFQ electronics. Amongst these, also ambitious experimental design have been included anticipating the intended complexity of forthcoming devices.
After testing and characterization of Build 1, an extensive design phase was executed. Because testing of building blocks (especially current steering switches) has revealed partially unexpected behavior, the start of build 2 was intentionally delayed until the design issues have been resolved. In consequence, the start of Build 2 was expected only at the beginning of 2014, which would have seriously challenged the project schedule. Therefore, the common decision was made to perform Build 2 and 3 in parallel at VTT and IPHT; abandoning the possibility of design iteration between Build 2 and 3 in favor of a successful demonstration of the compatibility between the foundry processes. Both builds have been fabricated successfully; however, only the VTT Build 3 yielded high quality parameters whereas the IPHT Build 2 suffered from an insufficient insulation quality . The scheduled Build 4 at VTT was used to test the new ESQUID technology on topic related devices, namely FDM’s for simultaneous ESA projects as SAFARI. The intention of this build is the demonstration of the versatility of the novel foundry process for superconducting electronics in a wider context.
Despite the occured technological problems, still, the successful accomplishment of fabrication builds at IPHT (Build 1) and VTT (Build 3) allows for a direct comparison of the technologies. At the example of functional 8 channel multiplexer devices and RSFQ logic circuits, the compatibility of both foundries was clearly demonstrated. For the first time in the international superconductivity community, the direct design transfer between two independent manufactures was shown. That achievement enables future joint work in the field of superconducting electronics and strengthens the European research especially in comparison to competitors in the United States.
Test environment
Code Domain Multiplexing for Space Applications
VTT equipped its BlueFors LD400 dilution refrigerator with a retractable wideband wiring set, set of shielded boxes at 4K plate, and an experimental Cu box at mixing chamber. Because wiring plays a crucial role in wideband readout systems, it was designed to be rapidly modifiable, with spring-loaded thermalization blocks and homemade vacuum-tight feedthroughs at removable KF flanges. A modular electronics cage is installed at the top flange, where rapidly prototyped, standard-sized electronics cards can be installed.
VTT also constructed a ‘broomstick’ probe for quick tests in liquid helium, where the same KF-flanged wiring can be installed (Figure 14). A modular electronics cage is located on top of the flange. The ‘broomstick’ makes it possible to test the same electronics/wiring combinations with the same parasitic effects in liquid helium, before installing them into dilution refrigerator.
VTT has designed and constructed two types of cryogenic semiconductor amplifiers and one room-temperature LNA. These are to be used after SQUIDs in the amplification chain. The first cryogenic type has voltage input, order-of 100 MHz bandwidth, and 0.5 K noise temperature into a few k input load. The second type (Figure 15) is designed into 50 transmission line –based systems, has >1 GHz bandwidth, and 5 – 10 K noise temperature into 50 input load depending on frequency. The room-temperature LNA is based on paralleled SiGe bipolar transistors, it has lower-than 0.5 nV/sqrt(Hz) voltage noise, and provides optional low-noise active-feedback 50 input termination. The 1 GHz amplifier is targeted towards sin/cos flux counting readout, and possibly towards reactive microwave SQUIDs.
Because large-volume cryogenic testing of SQUID devices is foreseen in the pathway to the certified flight hardware, we have taken the first steps to alleviate the mass testing. We shall attempt spring loaded contacting for tests, to avoid the labour-intensive wire bonding step. The wafer is first diced into reticles (Figure 16), which are tested in liquid helium. The reticles are the further diced into individual steps. The development efforts towards mass testing are beyond the scope of E-SQUID project, but preparatory steps have been taken, by including test pads and the internal test wiring in the Build 4 reticle design.
In the measurements performed at UH, the same measurement geometry and test system was used, that is described under VTT’s test environment section. The cryogenic test system was build inside a Bluefors SD250 dilution fridge situating in EMI shielded room. The measurement PC was operated from outside the shielded room to minimize its interference with the test system. The transmission between the laboratory PC and measurement system electronics was arranged by optical lines and feed-troughs provided in the wall of the EMI shielded room.
UH test facilities situate at Kumpula Campus Acceleration Laboratory premises, in a grotto below the main building about 20 meters underground. The experiment in UH test is presented in Figure 17. The cryostat’s pulsetube, which has potential to add interference in measurement signal, is installed outside the EMI shielded room. Altogether, the low ambient EMI background was anticipated to provide a good base for measurement work. However, the thermalization of the wiring turned out to be problematic in UH’s dilution fridge as a consequence of low cooling power as compared to the VTT’s dilution fridge, in which the wirings were initially build and tested.
According to the original plan, the CDM multiplexing test were planned to be performed at UH, but because renovation of the laboratory premises and delays in dilution fridge installation, the test were transferred to VTT. Accordingly only one measurement session was performed at UH premises.
FDM for Space Applications (SRON)
As the development work at SRON over the last years has been aimed at developing readout for bolometers for the SPICA/SAFARI project, a new cryogenic setup had to be designed to be able to demonstrate the combination of the SQUIDs and X-ray micro calorimeters. The other components besides the SQUIDs, i.e. the LC band pass filters, the micro calorimeters and the room temperature electronics have been made available outside the scope of this contract.
We have separated the demonstrations in an experiment on the SQUID performance, on the single pixel performance, and on a multiplexing demonstration with micro calorimeters. The results are described in the following sections.
Cryogenic experimental FDM setup
The function of the cryogenic experimental FDM setup is to provide a temperature and magnetic field controlled environment for the TES detector chip, the LC filter chip, and the SQUIDs. A radioactive Fe55 source is used to illuminate the X-ray detectors with photons of 5.9keV with a count rate of a few counts per second.
Two versions of the setup are available, one with the two SQUID stages at two temperatures, to demonstrate compatibility of the SQUIDs with the intended configuration in a satellite, and one with both SQUID stages at base temperature (less than 50mK), which is experimentally an easier configuration. The bracket with TES detector chip, the LC filter chip, and the SQUIDs is the same for both cases; only the location of the 2nd stage (or upper) SQUID is different. Both setups are described below.
The 2-stage 2-temperature set-up consists of a 2nd stage SQUID in a cryoperm tube, that is thermally linked to the still level, and an FDM bracket that is thermally linked to the mixing chamber plate of a helium-free dilution refrigerator from Leiden Cryogenics. The thermal link of the 2nd stage SQUID is dimensioned such that its temperature can be controlled between 0.6 and 2 K, without disturbing the operation of the dilution refrigerator. Dimensioned with the same boundary conditions, the temperature of the FDM bracket with the 1st stage SQUID can be regulated between 17 and 200 mK without altering significantly the base temperature of the mixing chamber.
The FDM bracket assembly that holds the FDM multiplexing circuit, consists of a low magnetic impurity copper bracket, providing mounting space for the front-end SQUID, the LC filters and the TES array chips and a printed circuit board for the electrical connections. A ruthenium oxide thermometer and a heater are glued at the bottom of the copper bracket for the readout and stabilization of the temperature.
The FDM bracket can be used with both a low NEP TES bolometer array, and with a highenergy resolving power x- ray micro calorimeter array. The former configuration requires very low parasitic optical loading (less than 1aW), which is achieved by means of light blocking filters in the signal loom feed through and a light-tight assembly.
The X-ray micro calorimeters are sensitive to magnetic fields and their performance is optimal at static field lower than 1µTesla). Special care has been taken to design the magnetic shielding and to improve the uniformity of the applied magnetic field across the array. The TES arrays chip fits into a superconducting Helmholtz coil fixed at one end of the bracket. The coil is used to generate a uniform perpendicular magnetic field over the whole pixels array. The shielding of the external magnetic field was achieved by fitting the bracket into a Nb can wrapped by few layers of metallic glass tape. The interface of the Nb can with the lid of the bracket was designed such that it forms a labyrinth, filled with carbon-loaded epoxy on the copper lid side. In this configuration the Nb can provides both the required magnetic shielding, and the required stray light shielding.
The circuitry PC-board currently allows to read-out a maximum of 18 pixels under ac bias in a FDM configuration. The SQUID chip is fixed to the copper bracket high thermal conductivity glue (Ge Varnish). Two photographs of the setup are shown in Figure 18.
Optical Applications (ULEIC)
A read-out based on SiGe amplifier has been considered as an alternative to SQUID read out for STJ detectors. A cryogenic Voltage–Voltage amplifier developed by VTT has been modified has a current to voltage transimpedance amplifier.
The amplifier is seated at the 4 K stage of the cryostat (Figure 19a) while the detector operates at 300 mK (Figure 19b) The signal from the detector is carried to the input of the amplifier through an NbTi twisted pair. The amplifier is encapsulated in an aluminium die cast box. The Amplifier control lines, bias line and signal line from 4K To the room temperature amplifier are made of 8 constantan twisted pairs.
Far-infrared applications for ground based observatory (MPG)
While fully functional CDM chips from the E-SQUID project were not yet available, useful development work for the project was still possible. MPIfR and IPHT are engaged in building an array of three hundred TES sensors (LABOCA-2) for the APEX telescope in Chile. LABOCA-2 is meant to replace LABOCA-1, which has semiconductor bolometers and operates in a liquid helium cryostat. The new camera is required to match sensitivity of LABOCA-1. Both LABOCA versions operate at a wavelength of 870 micron.
LABOCA-2 operates with TDM readout, using a ten channel SQUID multiplexer chip manufactured by IPHT. Thirty multiplexer chips are arranged around the edge of a 4” wafer with 300 TES in a hexagonal grid (Figure 21). The signals from each MUX are further amplified by a SQIF before entering the analog and digital readout electronics at room temperature.
LABOCA-2 is cooled by a three stage He-4/He-3 sorption cooler on a pulse tube (Figure 22). Operating temperatures are regulated at 230mK for the array with the MUXs and at 800mK for the SQIFs. The temperature of the array of conical horns which couple the radiation to the TES is around 280mK.The LABOCA cryostat is a very convenient test platform as it features:
1. Fast automatic cool down: 300K to 230mK in 25h.
2. Fast forced warmup: 14h
3. Hold time > 48h
4. Cold automatic recycling: 6h
The disadvantages are:
1. Tests can only be done on complete 4” LABOCA wafers
2. Tests, with the pulse tube stopped, are possible for only about 2 min.
Three subarrays of LABOCA-2, each with 100 TES, are routed to three FLL-electronics boxes on top of the cryostat. The basic MUX sampling frequency is 15 kHz. Three operating modes are available:
1. In the normal TDM-mode the signals are demultiplexed and for large signals, there is an option to correct for flux jumps of the SQUIDs. The frequency limit in this mode is 750 Hz.
2. In the “step-mode” the MUX is switched off and one particular TES can be selected for readout at 15kHz, resulting in a frequency limit of 7.5 kHz.
3. The three FLL boxes are identical and can be interchanged. One of the boxes is fitted with a connector to the output of the SQIF. In step-mode, it allows to measure the high frequency noise with an FFT spectrum analyzer. Our model SRS 780 has a high frequency limit of 102.4 kHz.
The warm electronics are virtually identical to that delivered to the E-SQUID project by Supracon, except for the much higher number of channels. Changing from TDM to CDM would mainly require changes in the software.
In radio astronomy, the response of a receiver to radiation is calibrated by taking hot/cold data. That means that absorbers at 300K and 77K are put in front of the receiver and the change in output is measured. 300K and 77K are typical background temperatures in the FIR as well, therefore the same method can be applied. If a complete I/V characteristic is taken at both temperatures, then the response for all bias settings is obtained at once.
Noise data were taken in any of the three alternative ways: MUX-, step- or spectrum analyzer mode. NEP values can be derived by combining the noise with the hot/cold data. Other sensitivity criteria are the direct comparison with LABOCA-1 or the amount of 300K photon background noise.
Results
Code Domain Multiplexing for Space Applications
After Build 1, VTT used the Jena-fabricated superconductive switch chip from Build 1 to demonstrate 8-pixel TDM of X-ray calorimeters at 50 mK temperature (Figure 24). A 5x5 –pixel calorimeter array from SRON was used as the detectors, which was irradiated from a radioactive 55-Fe source. The resulting time traces are shown in Figure 25. The switch chip was primarily designed for measuring dc characteristics of multipixel bolometer/calorimeter arrays (Figure 26), but the chip adapted into TDM role, too. The work was reported in the EUCAS-2013 conference.
After Build 2/3, a VTT-fabricated version of the switch chip, with the same layout as the Jena version from Build 1, was tested (Figure 27) at 4.2K without calorimeters, and found to perform essentially the same way as the IPHT-fabricated version (Figure 28).
An amplifier SQUID with a large mutual inductance (Figure 29), order-of 0.8 _mu A / Phi_0 sensitivity, was tested at 4.2 K, and found to perform well-enough that testing with detectors is warranted, either with SPICA bolometers or X-IFU calorimeters. The SQUID is designed for the so-called ‘high-Z summing point’ circuit topology, in which the lower 50 mK SQUID is replaced by a passive (non-dissipating) transformer, and the first SQUID resides in the 2 K refrigerator stage.
General functionality of the so-called SQUELCH chip was verified at 4.2 K. The idea behind the SQUELCH is to utilize the compressibility of the information flow from a calorimeter array, namely the fact that co-incident counts from many adjacent pixels are not likely. Two cascaded SQUELCH chips are designed to generate a 5-channel binary-coded output signal from 32 calorimeter pixels. The inactive channels not registering an x-ray count, reside in superconducting state and hence do not contribute to noise. This is the motivation behind the name ‘SQUELCH’. The chip is depicted in Figure 30.
After Build 4, the amplifier chip pair intended for FDM readout was successfully tested at 4.2K (Figure 31). The lower amplifier is a single SQUID intended for 50 mK operation, while the upper amplifier is a 4-parallel 184-series SQUID array, to be used as the second amplifier stage at 2.5 K temperature . The 4x184 array shows flux noise of 7E-8 Phi0 / sqrt(Hz), which implies a larger native dynamic range than any other SQUID device in world.
Also tested from Build 4 is a high-sensitivity SQUID intended for high-Z summing point, which is equipped with local feedback circuitry for response linearization. The SQUID is depicted in Figure 32.
Outside of the E-SQUID fabrication rounds, a system-level experiment on open-loop sin/cos flux counting was performed, by using pre-existing SQUID devices from other projects. The flux-counting approach is seen as a way to increase the dynamic range without increasing heat dissipation. The work was reported in DOI: 10.1088/0953-2048/27/7/075005. A chip design was included in Build 3 which combines flux-summing CDM circuit (i.e. signal summing coils with Hadamard-coded polarities) and the sin/cos flux counting readout SQUIDs. Tests on the CDM chip are underway.
FDM for X-Ray and Far-IR Space Applications (SRON)
To verify the SQUID behavior of a G1/F5 combination strictly independent of the TES or multiplex behaviour, the SQUIDs have been characterized during an independent cool down, without any load, such as LC filters and TES based detectors, connected to the input coil.
Both the upper-SQUID (F5) and the lower-SQUID (G1) are mounted on a PCBoard in a magnetically shielded SQUID test module (not shown). In Figure 34a shows the voltage-to-flux characteristics of the upper-SQUID measured as a function of the bias current, and with the lowerSQUID in the superconducting state.
Figure 34b shows the voltage-to-flux characteristics of the two-stage SQUIDs measured as a function of the bias current of the lower-SQUID array (G1). The two-stage SQUID array shows a clean V-ϕ curves with a maximum voltage modulation of 2.4mV at a lower-SQUID bias current of 50 µA. Note that the lower-SQUID is effectively voltage biased at 50 µV since the bias current is applied to a 1Ω shunt resistor connecting the lower-SQUID output to the upper-SQUID input coil.
We measured the current noise of the two-stage SQUID at three different temperatures, respectively 3K, 100mK and 20mK. Two calibration tones at 12kHz and 2MHz have been applied via the feedback coil to monitor the SQUID gain as a function of temperature. The magnitude of the calibration tone corresponds to an applied flux to the lower SQUID of 0.01Φ0. In Figure 35a the spectral flux and current density taken at temperature of 3K and 25mK is shown. The current noise has been calculate using the nominal input sensitivity of the G1 SQUID array equal to M−1 = 7µA/ Φ0 In Figure 35b a zoom-in of the SQUID noise spectra around the 2 MHz calibration tone is shown. We measured a flux noise equal to 0.35 μΦ 0/√Hz and 0.18 μΦ0/√Hz at 3K and 20mK respectively. This corresponds to an energy resolution of ε = SΦ/2Lsq ≃ 9ħ, where an Lsq = 70 pH is assumed, and is limited by the room temperature amplifier and the data acquisition system.
A relatively straightforward optimisation of the room temperature set-up should lead to a further improvement of the SQUID sensitivity. The flux sensitivity for the G1 SQUID array is √SΦ,QL = 6 • 10−8 Φ0/√Hz, which should be achievable if the temperature of the SQUID shunt resistors can be cooled at temperature below 150mK. The measured current noise at the SQUID input is 1.2 pA/√Hz at 20mK, which corresponds to a coupled energy resolution of ε c = ½LinSI ∼ 20ħ, where Lin ∼ 3nH is the input inductance of the SQUID.
We have observed no difference in the noise when the temperature of the mixing chamber was raised to 100mK indicating that the actual SQUID shunt resistors are thermalized at higher temperature and that the noise is limited by the room temperature electronics.
Single pixel performance under AC bias
TES based detectors need a voltage bias source for stable operation. The bias source can be either a constant voltage (DC) source, or an alternating voltage (AC) source. Because of the physical processes in the detector it cannot be assumed automatically that the performance under AC and DC bias is equal. The purpose of this section is to give an overview of the status of the performance measurements under AC vs DC bias.
Biasing the TES with a sinusoidal alternating current (AC bias) is an essential ingredient for FDM. Whether the performance of the detector under AC bias and DC bias is the same or similar is an important difference for applications of TES detectors. For bolometer pixels it has been shown consistently that the dark NEP measurements under both AC and DC bias give the same results. For microcalorimeters, however, this demonstration has not been given consistently until now.
Figure 36 (left) shows the latest results of NEP measurements on an X-ray microcalorimeter which has been characterized under both DC and AC bias. The AC results are shown for two different settings of the anti-alias filter in the electronics. The corresponding integrated NEP, which is equivalent to the energy resolution of the pixel, is shown in the legend of the plot. The results under DC bias show the lowest NEP and therefore the best integrated NEP or energy resolution (ΔE=2.3eV FWHM), while the best integrated NEP under AC bias currently equals 2.8eV FWHM.
From the plot it becomes clear that, by comparing the configuration of the anti-alias filter with the red curve and the green curve, either the low or high frequency part of the curves overlap. This implies that we have reasons to assume that the NEP under AC and DC bias are equal, once this issue in the room temperature electronics has been solved.
Figure 36 (right) shows the measured integrated NEP as function of bias point resistance in the transition, normalized on the normal state resistance, for two different magnetic fields. The plot shows that the best results can be obtained for a range of bias points and magnetic fields.
The measured X-ray resolution under AC bias is currently 3.7eV while 2.3eV has been observed under DC bias for the same pixel. We cannot exclude at this moment that the discrepancy is caused by drifts in the responsivity of the detector, as a result of insufficient temperature stability in both the refridgerator and the room temperature electronics. Ongoing experimental work is aimed at improving stability of the setup on the time scale level of 1 hour, to make representative measurements under AC bias possible.
Multiplexing demonstration
FDM multiplexing with X-ray calorimeters has been demonstrated earlier at SRON in a different setup and with different SQUIDs. With these results in mind, the main purpose of this demonstration is to show that in the combination with 5 pixels over an increased frequency range the noise level has improved with more than an order of magnitude.
An overview plot is shown in Figure 37, where the current noise spectral density at the output of the SQUID amplifier chain is shown as function of frequency. It is clear that within the measured range between 1 and 6 MHz, the SQUID noise level is independent of frequency at a low level of 2.2 pA/√Hz. The peaks in the spectra which belong to the microcalorimeter pixels are labeled px 1-5. The narrow peaks in between the pixels originate from imperfections in the room temperature electronics.
The same noise levels are observed after demodulation at the output of the digital baseband feedback electronics, as is shown in Figure 38. Here we see the noise of three resonators in the superconducting state. The high noise level of approximately 20 pA/√Hz corresponds to the Johnson noise of the stray resistances in the bias circuit. The spread in Q-factor of the LC resonators causes the different level and corner frequency for the 3.6 MHz resonator. Above the corner frequency the noise level rolls of to the SQUID current noise level.
One possible form of pixel interactions when multiple pixels are switched on is bias power leakage between pixels. The latter effect depends mainly on the separation in frequency space between pixels, normalized on the bandwidth of an individual pixel. In this experiment the spacing has been chosen such that no observable power leakage is to be expected. The measurement of a current-voltage (IV) characteristic of a pixel with and without neighboring pixels biased, is shown in Figure 39. It is clear that at the level of less than 1% the curves overlap, which confirms the expected lack of power leakage.
The last experiment to be discussed, consists of measuring the time response of 3 pixels simultaneously. A time domain plot of 3 biased pixels is shown in Figure 40. The time range (1 s) of the plot is chosen such that a small number of photon absorption can be seen. The response time of the detector to the photon hit is barely distinghuishable in this plot. The single sided signals are photon reponses, while the double sides signals are caused by electrical cross talk between the pixels. Whether the level of the cross talk is consistent with the expectations has not been investigated yet. It is, however, easy to conclude that such cross talk events are easy to reject, as absorbed photon responses are single sided only, while the cross talk events are clearly double sided.
We therefore conclude that the combination of the 2-stage SQUID tandem works well in combination with the microcalorimeter readout sofar. More experiments outside the scope of this work package will be needed to make more quantitative estimates of inter-pixel behavior possible, and to study the impact of a closer packing density of pixels in frequency space.
Far-infrared applications for ground based observatory (MPG)
Before the E-SQUID project, IPHT and MPG/MPIfR had been developing TES technology for astronomical and terrestrial applications. For that purpose, TES arrays with 7 elements were manufactured at IPHT and tested at both institutions in small He-3 cryostats with direct SQUID readout. As shown in Figure 41, photon background limited noise was achieved in a 300K environment. To achieve a similar kind of performance, but via a multiplexer of some kind, is the aim of E-SQUID project as applied for the FIR.
It was expected that the first LABOCA wafer fabricated according to the best 7-element test array would be good enough to go to the APEX telescope. Unfortunately this was not the case. For any given bias, which by design is common to 50 TES, some TES would show good sensitivity but others not. No common bias with good performance could be found. At the time, the problem was believed to be in the fabrication of that particular wafer, and it was decided to make more wafers.
Early in the year 2011, soon after the fabrication of the first LABOCA-2 wafer, the cleanroom of IPHT was shut down for a major expansion and installation of new modern production equipment. Unfortunately, all LABOCA wafers (11 in total), produced after the reopening of the cleanroom on new equipment had some kind of defect, like wrong Tc, poor uniformity or multiple transitions. By the end of the year 2012, the decision was made to go back to the old machine for production of LABOCA wafers. This lead to considerable improvement in the quality.
Back on the old machine, four wafers in a new layout have been produced in that period at IPHT, two with thermistors of 50 x 100 microns and two with thermistors of 100 x 200 microns.
The new LABOCA-2 wafers with 100 x 200 micron thermistors have almost ideal I-V characteristics, in the sense that the dissipated power in the transition range is almost constant and the transition to normal conduction is abrupt (Figure 42).
The instability at the superconducting transition is much closer to the normal transition for the wafers with 50 x 100 micron thermistors, making them less useful.
A most surprising result was the appearance of high frequency noise in these TES with quasi ideal I/V characteristics. If any broad maximum in the noise appears at one bias setting, it can be maximized by very small bias steps.
Tests were run by connecting old spare 7-element test chips to the new LABOCA-2 data acquisition. Two test chips were glued to one of the new LABOCA-2 wafers with 50 x 100 micron thermistors and bonded to two of the existing MUX, replacing the TES on the wafer. One MUX was left bonded to the TES on the wafer. The latter TES show high frequency noise (Figure 43).
On each test chip, several different TES structure had been implemented. Some of them therefore went into normal conduction at different bias voltages. In Figure 44, noise data were combined that were taken at about 0.2 V bias below the normal transition of each TES. Surprisingly, the noise is quite well behaved at high frequencies!
One particular TES on the test chip is very similar to the LABOCA design. Its noise spectrum is shown in more detail in Figure 45, with the pulse tube stopped. The suppressed Johnson noise at low frequencies in the range of the thermal time constant is evident! Measurements with input radiation were not possible in this improvised configuration. Three test chips from two wafers have been tested to date. All have low high-frequency noise, some slope in the power characteristic and reduced Johnson noise at low frequencies.
All the TES on the test chips have very small series resistors (about 8 mΩ) implemented on one side of the thermistor. At the time, these series resistors were only meant to provide a way to measure the I/V characteristic in both directions of the bias voltage. Also, all of the TES on the test chips have a Gold ring around the absorbing area. The TES on the wafer have neither the series resistor nor the Gold ring. At the time of writing this report it is not clear whether the different layouts have anything to do with the good performance of the test chips or if that is due to some lucky “accident” in the fabrication. More tests are necessary. We can exclude however, that the high frequency noise is caused by internal or external RF interference.
TDM could still be a valid MUX scheme if an electrical anti-aliasing filter were incorporated on the MUX chip. Space on the wafer for the larger MUX chip was foreseen in the new layout. The first MUX chips with coils at the input were produced at IPHT and are tested on a LABOCA-2 wafer. This was more with the idea of blocking the high frequency noise. To serve as anti-aliasing filter, TES with much lower normal resistance would have to be developed.
Another alternative is to use the thermal time constant of the TES for anti-aliasing purposes. It is expected that this will happen to some extent anyway at high background levels.
Optical Applications
A read-out based on SiGe amplifier has been considered as an alternative to SQUID read out for STJ detectors. A cryogenic Voltage–Voltage amplifier developed by VTT has been modified has a current to voltage transimpedance amplifier.
The stability of the STJ bias is crucial. A bias between 100 to 200 µV that seats within the superconducting gap is needed for photon detection. Moreover, a diagnosis of the Junction performance and states is required prior of photon detection. The diagnosis consists of an IV curve covering the entire superconducting gap. A typical IV curves obtained in our set up with a room temperature IV tracer is shown in Figure 46.
An in-built biasing system was used. However, the actual gain of the biasing voltage divider was a priori unknown and also could depend on the impedance of the load. This gain was experimentally determined using three different known load resistors (Figure 47). On the other hand, the current to voltage amplifier gain was known and was 105 V/A.
Top left, the voltage applied at the input of the voltage divider is plotted versus the measured current for 3 different resistance load. Top right, graphical determination of the two equivalent resistors of the voltage divider. Bottom left, voltage divider gain versus the load resistance deduced from the resistors value.
Figure 48 shows IV curves of STJ obtained at 300 mK when a tuned magnetic field is applied (red line) and without magnetic field (blue line). The Red curve is characteristic of a high impedance device. However, flux trapping or excess noise due to EMI degrade the IV curves (cf. Figure 46). Even though, the red curve is the best curve obtained so far with the cryogenic amplifier, attempts to measure X–Ray pulses within these conditions were unsuccessful. Therefore, no optical photon detection was considered.
To illustrate the multispectral imaging capability of STJs based camera, measurements of a human fibroblast cell using a triple bandpass emission filter were recorded. The cell was stained with 3 different dyes. The filter allows simultaneous transmission of fluorescence from multiple fluorochromes that emit in the bandpass regions. A typical spectrum integrated over the whole camera during a 1 s acquisition is shown in Figure 49. There are three well defined peaks that reflect the three emission bands of the filter. The peaks are broadened by the Gaussian response of the STJ. In order to estimate the contribution of each dye, their tabulated emission spectra are superposed with the measured spectrum. Even though, emission bands are not necessary centered at the maximum emission wavelength of the fluorochromes, each band has its signal clearly dominated by a single dye contribution. Therefore, while all dye’s contributions are measured simultaneously, data post processing allows selecting signal from a specific band (i.e. specific fluorochrome). Figure 50 shows images that have been obtained by counting event that occurred in the three specific bands indicated Figure 49. In the 420-450 nm band, the signal is located in the middle right of the image and shows the nucleus. In the two other bands, the entire cell is visible. The corresponding images using usual CCD cameras are also shown. In this case, each image corresponds to a different measurement using a different filter set.
High counts rate associated with high sensitivity of STJs gives the possibility of reducing the acquisition time with the benefit of, for example, reducing the risk of photo bleaching. Figure 51 shows an image of the nucleus of a human fibroblast cell for different times of acquisition. After only few milliseconds, the DAPI signal, concentrated at the nucleus of the fibroblast cell, is clearly distinguishable from the background, while conventional epi-fluorescence imaging requires generally acquisition time beyond the second.
Conclusion
Cryogenic imaging detectors based on TES arrays with SQUID MUX readout will have an important role in instrumentation of future space-based telescopes. The capability of cryogenic imaging detector arrays promise unrepresented observational capabilities in many fields of modern astronomy. The technology is applicable in large band of electromagnetic spectrum, from FIR to X-ray. Accordingly, several instruments based on this technology are proposed, such as Athena's X-IFU and SAFARI's SPICA. There are also ground based applications that will greatly benefit on maturation of this technology, such as the LABOCA camera on APEX.
Three out of the four fabrication rounds were successful (Builds 1, 3 and 4), from which Build 4 was devoted to demonstrate the selected FDM technique.
The partners VTT and IPHT defined the common design rules for the cross-compatibility between the VTT and IPHT foundries. The almost similar fabrication rounds demonstrated the cross-compatibility of the IPHT (Build 2) and VTT (Build 3) foundries with the Zappe-TDM chip as the example device.
The SQUID tandem which has been developed for FDM has been tested separately, and in combination with a small array of microcalorimeters. The measurements show that the energy resolving power of the SQUIDs is nearly quantum limited at 20ħ, which shows that low current noise spectral density can be combined with a low summing point inductance in an FDM system. This creates a multiplexing space of approximately 160 for the bolometer TES detectors, and approximately 40 for microcalorimeters.
In optical wavelengths, the SQUID readout for multiplexing of the STJ does not work due to magnetic field. Instead, cryogenic SiGe semiconductor amplifier with 1 GHz bandwidth and ~5 K noise temperature was designed, built and demonstrated, but the synergy with SQUID devices was loosed.
Potential Impact:
There are significant socio-economic impacts arising from the E-SQUID project.
Firstly, it will improve the Technological Readiness Level (TRL) in Europe in the area of cryogenic multiplexers, which is considered as the breakthrough solution that fulfils the extremely strict science requirements of future space observatories especially in the X-ray and the Sub-mm range. E-SQUID will also enhance technological spin in and collaboration between the top European developers of SQUID-based solutions, and the company partners will benefit from the results of the project as improved competitiveness in the international market. Furthermore, the results also support the needs of European security and defense sectors, as well as the medical sector.
As the highest level socio-economic impacts, E-SQUID will improve the European non-dependence in critical (space) technologies, and thus the competitiveness of European space industry in the global market. Consequently, the improved competitiveness will have a positive effect in Europe’s economy and thus employment. The future space science applications from EQUID project will bring new high quality data for the use of European science community, thus also raising the impact of European science worldwide.
The quantification of the socio-economic impact in the economy is difficult in terms of economic impact and their social implications to e.g. employment, since the lead time of the possible commercial applications is very long, i.e. of the order of a decade, and the various final commercial products and the extent of business that will exploit the commercialized technologies are not yet known.
However, European non-dependence in this critical technology is a target that can be considered as achieved in E-SQUID project, considering the needs of the European scientific community. This means that the capability of European foundries to develop and procure scalable SQUID multiplexers with sufficient noise characteristics *is demonstrated* at level, which justifies relying on purely European foundries for the use of future European space missions, but the *final quantitative requirements were not achieved* in E-SQUID project. It can be stated with high confidence that the quantitative requirements will be reached within sufficiently short time frame considering the needs of future European space science missions.
With the ESQUID project, substantial progress on that field promises to pave the way to more effective, user-friendly and cost-effective devices. First ground-based scientific instruments based on European technology are already in routine operation by observation staff, and future European space missions might benefit from the ESQUID technology. With that achievement, for the first time in cryogenic detector research, it seems promising to look for their application also beyond science.
The potential applications arising from the E-SQUID project include concealed weapon detection through clothing at airports, and detection of Near-Earth Objects (NEO’s) threatening the security of the whole human civilization. Also infrared camera applications by which a number of human activities can be detected from space including forest fires and missile launches are possible. The technology based on E-SQUID can also be used for Earth-based applications in ultra-sensitive THz cameras for medical applications.
In the following the impacts and exploitation of E-SQUID project results are explained in more detail.
Among possible fields, the imaging of human bodies for security checks has spearheaded because of its high social relevance by being an alternative to the much disputed conventional body scanners. Terror attacks of the kind which have taken place recently in public places such as airports or train stations would have been prevented if a bomb concealed underneath the clothing of the attacker had been detected in a timely manner. Therefore, a ‘body scanner’ which can generate an electromagnetic visualization of hidden threats is a very attractive concept for future security screening.
However, although the desirability of such devices is clear, their current realization as active X-ray or millimeter-wave scanners raises certain health and privacy concerns. One considerably more attractive alternative are cameras which could generate passive images more rapidly without the need for the targets to stop what they are doing and co-operate with the screening process. This is the principle of Smart Security (IATA’s checkpoint of the future) and its vision to guarantee “...an uninterrupted journey from curb to aircraft door, where passengers proceed through the security checkpoint with minimal need to divest, where security resources are allocated based on risk, and where airport amenities can be maximized…” This ambition can be fulfilled, with current knowledge on possible technologies, only by using cryogenic detectors, as it has been demonstrated by two European groups, both also members of the ESQUID consortium: A. Luukanen et al. from VTT and T. May et al. from IPHT. Their devices will strongly benefit from the ESQUID technology, thus improving the performance at the same time with reducing the manufacturing costs. The social impact of such cameras is obvious: as they will improve security whilst making journeys more comfortable and respecting the privacy of the traveler, the public acceptance of such unavoidable measures will be increased substantially.
The dissemination activities in that field include the presentation of ESQUID results in frame of conferences related to security and Terahertz research (see list), the introduction to industry partners (e.g. Rohde & Schwarz as system vendor for security devices), and in semi-popular articles for EU-related audience and decision makers (see list).
Another very promising direction for future dissemination of ESQUID technologies is the field of life science, where imaging applications might benefit from the unprecedented performance of cryogenic optical detectors. One nearby application is fluorescence microscopy. In modern medical analysis, fluorescence microscopy has become an indispensable tool in imaging of biological tissue on a cellular level. A large variety of life processes as e.g. chromosome dynamics or gene expression can be visualized, benefiting medical diagnosis as well as basic research. In case of using diverse types of probes which fluoresce at different wavelength, this technique is able to separately label different organelles of a cell, thus studying their interaction at the same view.
With an energy-resolving single photon detector array (e.g. a TES or a STJ) it would be possible to simultaneously record images at different wavelength without any signal losses due to today’s scanning or tunable filter techniques. With that, the study of fragile biological samples at cautiously faint illumination levels becomes possible; preventing fluorophor photobleaching and sample damage to the best possible degree. That becomes especially important in medical research, where life processes need to be studied with minimal influence on the sample. Although fluorescence microscopy was shown to achieve unprecedented performance in terms of spatial resolution and signal quality, the inevitable impact of the fluorophor excitation on the life process to be observed still hinders an un-spoiled observation.
Utilizing cryogenic detectors, the required excitation could be reduced to a fundamental minimum, simply because the sophisticated detector is able to record literally every single photon emitted by the probe or sample, respectively. This promises to pave the way to a new performance category of photon detectors which could be exploited as focal plane arrays for multispectral microscopes. However, despite the obvious advantages, a substantial risk for a successful commercialization arises from the need to achieve a proper balance between application benefit and technical complexity. Again, ESQUID technology can act as an enabler for such focal-plane instruments with a high spatial resolution.
In the future, such instruments could primarily benefit medical research in the context of analyzing the interaction of biological or chemical agents with living cells. Because such interactions potentially can be studied with minimal interference by the observation technique, best possible significance can be achieved. That becomes especially advantageous for the development of tailored medicinal products which treat diseases at a cellular level. With such focus, the technique will be used rather by a few medical research centers than by doctors at standard hospitals. Although that limits the potential market, the social impact, nevertheless, is hardly less significant. In an ageing society, the treatment of diseases becomes more and more important. With each individual advancement of therapeutic procedures, ethical and economic advantages arise from improved well-being and smaller health care costs, respectively. That motivates the research into such high-end technologies.
University of Leicester and IPHT already have agreed to collaborate in such direction and have propsed a research project in frame of a bilateral UK/German project (Biophotonics Plus) which plans exploiting and enhancing parts of the ESQUID technology.
There are also niche applications in microcalorimetry for material analysis based on X-ray spectroscopy, and nuclear isotope analysis, based on gamma-ray spectroscopy. The results of this project can be used to develop detectors dedicated for such applications.
To assess the broader impact of the project in scientific context, one must include the time scale, to which one is referring to. E-SQUID is intended for space applications, which have a rather long lead time. The devices that have been produced within this project are intended as demonstrators for instruments with a launch date in the 2020’s or even 2030‘s. It is clear that the space telescopes in which the devices are intended to fly, will open new windows for astronomy, and therefore produce scientific papers and newspaper articles.
This technology is already used in modern ground-based astronomical instruments such as SCUBA-2 at James Clerk Maxwell Telescope (Hawaii). The present state-of-the-art does not, however, fulfill the requirements of the next generation of astronomical instruments.
The E-SQUID project results will address several future needs in astronomical research. Application of the results of the E-SQUID project is in the plan for the X-ray micro-calorimeter, X-IFU, of ATHENA (L2 mission of the ESA, current launch schedule 2028), is foreseen applicable for the SAFARI instrument for the infrared space mission SPICA (JAXA mission with ESA contribution, current launch schedule 2025), and is also possible for the next generation ground-based far infrared experiments of ALMA (ESO, in the 2020’s). (Figures 1, 2, and 3).
The X-IFU (X-ray Integral Field Unit) will be an advanced, actively shielded X-ray microcalorimeter spectrometer for high-spectral resolution imaging. The detector will be a large array of absorbers read out by transition edge sensors (TES). The TES microcalorimeter senses the heat pulses generated by X-ray photons when they are absorbed. The minute temperature increase occurring with the incident photon energy is measured by the sharp change in the electrical resistance of the TES. This must be cooled to a temperature of less than 100 mK and biased into its transition between superconducting and normal states.
SAFARI will cover the FIR window that extends from ~30 μm (the upper cut-off of the MIR instruments) to ~210 μm (just long ward of the [NII] 206 fine structure line) with a field-of-view of 2'x2'. Assuming diffraction limited performance, SAFARI will provide angular resolutions from ~2" to 15" (20 to 150 AU at 10 pc) at wavelengths not covered by JWST and at more than 2 orders of magnitude higher sensitivity than Herschel/PACS. This huge increase in sensitivity could potentially open EP research to wavelengths completely blocked by the Earth's atmosphere, but representing the emission peak of many cool bodies (gas-giant planets, asteroids and so on). SAFARI will have its major strength in measuring excess radiation from dusty proto-planetary disks in hundreds of stars at almost all galactic distances. It will also perform medium spectral resolution observations (R~2,000) over a spectral range also rich in dust features, water vapour rotational lines (temperatures below ~500 K), atomic oxygen fine structure lines at ~63 μm and the solid state water-ice features at ~44 and ~62 μm.
The baseline optical configuration of this European instrument is a Mach-Zehnder imaging Fourier Transform Spectrometer (FTS). The principal advantages of this type of spectrometer for ESI are the high mapping speed of the FTS due to spatial multiplexing, the ability to incorporate straightforwardly a photometric imaging mode and the operational flexibility to tailor the spectral resolution of the science programme. The detector technology candidates of the insturment are Si:Sb and Ge:Ga photoconductors, transition edge sensor (TES) bolometers and kinetic inductance detectors (KIDs).
List of Websites:
http://fusion.gfl.helsinki.fi/esquid/
Coordinator: Dr Juhani Huovelin, Department of Physics, P.O.Box 48, FI-00014 University of Helsinki, Finland.
