10th International Workshop On Low Temperatures Electronics

PROCEEDINGS

October 14-17, 2013 Paris, France

Organizing Committee

C. FERRARI, Univ. Paris Diderot/AIM M. PIAT, Univ. Paris Diderot/APC D. PRELE, Univ. Paris Diderot/APC A. ANTHORE, Univ. Paris Diderot/LPN A. CAVANNA, CNRS/LPN U. GENNSER, CNRS/LPN Y. JIN, CNRS/LPN X. DE LA BROISE, CEA/IRFU C. PIGOT, CEA/IRFU-AIM V. REVERET, CEA/IRFU-AIM L. RODRIGUEZ, CEA/IRFU-AIM J.-L. SAUVAGEOT, CEA/IRFU-AIM F. ANIEL, Univ. Paris Sud/IEF

PAPER LIST

Part 1 Low frequency electronics

- GeFRO: a new readout circuit for cryogenic ionization detectors with low radioactive background contribution - Lorenzo Cassina, Carla Cattadori, Andrea Giachero, Claudio Gotti, Matteo Maino, Gianluigi Pessina p. 4

- Development of cryogenic readout circuit for far-infrared image sensors with fully- depleted silicon-on-insulator (FD-SOI) CMOS process - Koichi Nagase, Takehiko Wada, Hirokazu Ikeda, Yasuo Arai, Morifumi Ohno p. 9

- On the MOSFET-based Temperature Sensitive Element for Bolometer application - Fuxa Etienne, Yon Jean-Jacques, Jomaah Jalal p.14

Part 2 High frequency electronics

- Modeling and Experimental Investigation of HF noise of SiGe:C HBTs at Cryogenic Temperature - Luis Diaz-Albarran, Eloy Ramirez-Garcia, Martha Galaz-Larios, Mauro Enciso- Aguilar, Nicolas Zerounian , Frederic Aniel p. 20

- Developments of the Cryogenic Integrated Circuits with GaN/AlGaN HEMTs and the New High Sensitivity Terahertz Detectors - Yasunori Hibi, Jiandong Sun, Hua Qin, Hiroshi Matsuo, Lin Kang, Jian Chen, Peiheng Wu p. 25

- SiPM cryogenic operation down to 77K– Damien Prele, D. Franco, D. Ginhac, K.Jrad, F. Lebrun, S. Perasso, D. Pellion, A. Tonazzo, F. Voisin p. 30

- Zero Maintenance X Band Cryogenic Low Noise Amplifier - Remi Rayet, T. Bonhoure, B. Fauroux p. 35

- Design and evaluation of low-frequency RFID transponder for cryogenic applications - Frank Ihmig, Heiko Zimmermann p. 40

Part 3 Tests and cryogenics setup

- Experimental Set-up to test low temperature electronics for X-ray micro-calorimeters with high impedance sensors - Marco Barbera, Alfonso Collura, Xavier de La Broise, Giuseppe Lo Cicero, Ugo Lo Cicero, Claude Pigot, , Jean-Luc Sauvageot, Salvatore Varisco p. 45

Part 4 Detectors

- Ballistic Phonon Transport in Low-Temperature Detectors - Djelal Osman, Stafford Withington, David Goldie, Dorota Glowacka p. 50

- Optimisation of kinetic inductance detectors for millimeter and

submillimeter wave detection and the study of titanium nitride thin films for future applications - Amar Adane, Grégoire Coiffard, Arnaud Barbier, Catherine Boucher, Martino Calvo, Johannes Goupy, Alessandro Monfardini, Karl-Friedrich Schuster p. 55

- Superconducting microresonator detectors for neutrino mass

measurements in Milano - Andrea Giachero, Marco Faverzani, Peter Day, Elena Ferri, Claudia Giordano, Benno Marghesin, Riccardo Nizzolo, Angelo Nucciotti p. 65

- Development of Transition Edge Sensors based on Al/Ti Bilayer

Thin Films - Qingya Zhang, Genfang He, Wenhui Dong, Tianshun Wang, Junkang Chen, Jianshe Liu, Tiefu Li, Xingxiang Zhou, Wei Chen p. 70

Part 5 Quantum information processing

- A low-temperature device architecture for the statistical study of electrical characterisation of 256 split-gate devices - Haider Al-Taie, Luke W. Smith, Boruo Xu, Patrick See, Jon P. Griffiths, Harvey Beere, Geb A. C. Jones, David A. Ritchie, Michael J. Kelly, Charles G. Smith p. 75

- Multi-channel high frequency signalling at ultra-low

temperatures for quantum computing applications - Morteza Erfani, Michael Cuthbert, Steve Chappell, Anthony Matthews, Graham Batey, Robin Brzakalik (Oxford Instru.) p. 81

GeFRO: a new readout circuit for cryogenic

ionization detectors with low radioactive background

contribution

L. Cassina, C. Cattadori, A. Giachero, G. Gotti, M. Maino, G.Pessina

INFN e Universita di Milano Bicocca , Piazza della Scienza 2, 20126 Milano

E-mail: lorenzo.cassina@mib.infn.it

Abstract. We present GeFRO (Germanium FROntend) a novel approach to the readout ofcryogenic ionization detectors. The circuit consists of a very front-end placed close to thedetector and made of bare dies: an input common source JFET transistor and a resistor whichcloses the feedback path for DC restoration of the input node. Due to such a small number ofcomponents near the detector, GeFRO shall ensure a minimal impact added on the radioactivebackground in those experiments where very low signal rates are expected, such as GERDA andMAJORANA. A remote second stage, at room temperature and designed to be either AC andDC coupled, allows the amplification with a rise time of about 20 ns. In case of AC coupling,the second stage is provided of a Baseline Restorer. The two stages can be connected with morethan 10 meters of a terminated line whose series resistance very marginally affects the signalamplitude as would happen with a traditional approach. Only two signal cables are necessaryfor biasing and readout, so that power consumption, crosstalk and mass present in the detectorarea are minimized. We present the GeFRO performances using a BEGe detector; the resultsare obtained in the laboratory of the University of Milano Bicocca.

1. IntroductionSemiconductor detectors are widespread sensors able to collect the charge produced by ionizingradiation interacting inside their active area. Among these, germanium detectors are usuallyemployed in cryogenic environment for gamma spectroscopy measurements because they ensurea full absorption of incoming radiation and high energy resolution. Particularly, experimentssuch as GERDA or MAJORANA, whose primary goal consists of searching the neutrinolessdouble beta decay (0νββ) in 76Ge, use ultrapure germanium semiconductor detectors isotopicallyenriched in 76Ge, so they even fulfil the task of radiation source, maximizing the detectionefficiency. This kind of detector is usually read by a charge sensitive preamplifier circuitwhich produces an output voltage proportional to the collected input charge. One of the mostsignificant figure of merit of the 0νββ experiments is the sensitivity S0ν , given by the followingequation [1]:

S0ν = ln2 · ε · i.a.A

√M · t

∆E ·B(1)

In equation 1, ε is the detection efficiency, i.a. the abundance of 76Ge isotope, A thegermanium molar mass, M the overall source mass, t the observing time, ∆E the energyresolution and B the background events per unit of energy, time and mass. In order to maximizeS0ν , equation 1 suggests to minimize the background contribution and to design a tailoredradiopure read-out circuit which could ensure stable gain and high energy resolution. Our read-out circuit, designed for the phase II of GERDA experiment, is GeFRO: its main features arehigh gain stability and low electrical noise together with a minimal number of components madeof radiopure material near the detectors. Hence, GeFRO is responsible for a negligible impacton the overall background counts. Furthermore, the power consumption inside the cryostatand crosstalk between neighboring channels are minimized. In addition, referring to GERDAexperiment, an off-line pulse shape analysis is planned to be adopted both for pile-up rejectionand for the discrimination of γ muti-site events. This active technique is potentially able toreduce by about one order magnitude the spurious γ events in the energy range of interest,halving the whole background amount [2]. Hence, a wide bandwidth read-out circuit is supposedto be used in order to preserve the shape of the signal. For this reason, GeFRO was designedwith a “fast” output with a bandwidth larger than 20 MHz which ensures a timing resolutionof the order of 20 ns.In the first part the main features of GeFRO will be presented. Then, using a Broad EnergyGermanium detector (BEGe), the GeFRO performance observed in the laboratories of MilanoBicocca University will be shown.

2. The GeFRO read-out circuit

-

+

RP

CP

RS

RPul

Room temperature

Cryogenic temperature

II stage

Front-end

> 10 m

RT

VPul

CPul

RF

> 10 mJin

Detector

RB

VCC

RQ

VRef

AF

CC VFastG

HV

CF

VSlow

Figure 1. Simplified GeFRO schematic circuit in AC coupled version.

As shown in the simplified schematic of figure 1, the GeFRO circuit consists of two stages.The very front-end, inside the cryostat, is composed of only the feedback elements (RF and CF )and the input common source JFET (Jin), all of them placed as close as possible to the detectorin order to minimize the stray capacitance at the input node and chosen in order to minimizetheir noise contribution at cryogenic temperature. Another constraint to be considered is theradiopurity of the components: transistors and resistors in pure Silicon die were found to beadequate. At room temperature and several meters away from the very front-end, a secondstage ensures an adequate amplification and closes the slow feedback loop which discharges theinput node after every signal. The second stage is connected to the front-end part by means of

two 50 Ω coaxial cables 10 and more meters long and may be AC or DC coupled as it will beshown next. In the AC coupled configuration (figure 1) the DC component at the drain node isfiltered by the coupling capacitance CC and only the AC components of the signal are amplified.Resistor RT = 50 Ω terminates the connection line to avoid reflections. To minimize shifting,CC has to be very large (several hundreds of µF). Otherwise, in case the acquisition rates werelarger than a few kHz, a base line restorer circuit should be used. As alternative to the ACsolution, the DC coupled version is shown in figure 2. Resistors RP1, RP2 and RP3 have tobe adjusted in order to suppress the offset value of the VFast node. We have that Vref = VDwhen VSlow is close to VRef . In this case DC component of VFast is then suppressed by settingRP2 = RT and RP3 = R1. Furthermore, RP1, RFF and RB can be chosen for the biasing settingof Jin, voltage and current, respectively. In our solution, VD = 2 V and IDS = 10 mA (RP1

is 5.1 kΩ, RT = RP2 = 50 Ω, R1 = RP3 = RB = 1 kΩ). In the actual situation VSlow is thegate voltage of Jin, VGin, (detector leakage current is negligible) and VFast node has an offset

proportional to VGinR1RT

RQRP

, that is negligible when RP RQ. Eventually, an offset currentcould be injected at the inverting node of the feedback operational amplifier with a resistor tothe power supply. In the DC configuration no overshoot is expected so the base line restorercircuit should be unnecessary.

Det

HV

Jin

RFCF

RS

RP

CP

-

+

VCC

RFF RB

CFF

CFF

RT

A

-

+

RP2

RP1

CFF

RQ

C1

R1

VFast

VSlow

Cryogenictemperature

Front-end

RP3

VCC

VRef

VD

II StageRoom temperature

Figure 2. Simplified GeFRO schematic circuit in DC coupled version.

Let us now describe the GeFRO operating principle referring to figure 1. A more rigorousdescription as well as first and second order calculations might be found in [3]. At DC, supposingRP RQ, the drain of Jin is held at Vref by the feedback loop. As for signal amplification, thefast detector pulse is integrated across the total capacitance Cin present at the gate node of Jin.The signal gate voltage Vgs = Q/Cin is converted to the signal current I = −gmVgs by Jin thatdrives the terminated coaxial line. After the amplification stage, the “fast” signal amplitude isVFast = −G · gmRT · Q/Cin. After this transient, the slow feedback path discharges the inputnode, so VFast is described by equation 2:

VFast(t) ' −G · gmRT ·Q

Cin· e−

tτL−τD (2)

where τL 'RQRP

CinRFgmRT

is determined by the feedback loop and τD represents the delay of

the connection lines. The stability of the loop gain is ensured for τL > 2τD [3]. Providedthis last condition is met, we have a flexibility to change the value of the discharging timeconstant changing some resistor values. It could be observed that, neglecting the small gate-drain capacitance of Jin, the output amplitude does not depend on the series resistance of thetransmission line since Jin converts the input voltage signal into a current signal. So, evensmall section coaxial cables could be used to minimize the heat injection in the cryostat andnegligible effects should be observed due to their higher series resistance. Furthermore, since Jinoperates in open loop region on the rising edge of the signal, the rise-time of the “fast” outputis limited only by the bandwidth of the operational amplifier G (20 MHz in our prototype).Finally, according to equation 2, VFast ∝ C−1

in . Hence, the GeFRO fast gain depends on theinput capacitance Cin. If a non-ideal behavior occurs, such as a detector biasing instability,the input capacitance would fluctuate and gain instability could be observed. Hence, GeFRO isequipped of a “slow” output VSlow described by equation 3 [3]:

VSlow(t) ' − Q

CF· τFτF − τL + τD

(e− tτF − e−

tτL−τD ) (3)

As in a classic charge sensitive preamplifier circuit, VSlow only depends on the feedbackcapacitance and gain instability is prevented. The “slow” signal fall-time is regulated byτF = RFCF . Finally, since the biasing current to Jin is received through a terminated signaltransmission line, signal reflections which give rise crosstalk across the supply lines are prevented.In addition, GeFRO ensures a low noise contribution: expected values and detailed calculationscan be found in [4].

3. GeFRO performance with a BEGe detectorThe performance of the GeFRO read-out circuit was tested in Milano Bicocca Universityand INFN laboratories. GeFRO performance was tested using a CANBERRA Broad EnergyGermanium detector (BEGe), held in vacuum, which represents the baseline detector chosenin GERDA because it allows not only to discriminate between single-site and multi-site eventswith pulse shape off-line analysis but also to obtain high energy resolution thanks to its verysmall capacitance [1]-[5]. The front-end was connected to the AC coupled second stage bymeans of RG174, 50 Ω coaxial cables, diameter 2.7 mm, 10 m long. Both the “fast” and“slow” signals were filtered by a gaussian 10 µs shaper and finally acquired by a 14 bits Ortec919 multichannel analyzer. Every hour a 22Na and 228Th spectrum was acquired; the averageoverall events rate was about 1300 counts per second. Good performance at high rate is notstrictly requested in our experimental context, although a rate of about 600 Hz is foreseen forthe periodical channel calibration. The results concerning gain stability and energy resolutionare shown in figure 3. As it can be observed, in the absence of input capacitance fluctuationboth “fast” and “slow” signals show a good energy resolution in the energy range of interest:∆EFWHM ' 2.8 keV @2614 keV (208Tl full-energy peak) and ∆EFWHM ' 2.1 keV @1275 keV(22Na full-energy peak) were measured. Furthermore, pulse shape analysis [2] was performedconsidering the “fast” output and 90% of the multi-site γ events were rejected, in agreementwith the expectations. As for the “fast” signal, a small drift (∼ 20 ppm per hour) and a smallresolution fluctuation could be observed compared to the “slow” output. Moreover, for the “fast”signal only, the resolution does not increase proportionally to the square root of energy, but anunexpected extra-contribution could be observed. Because this effect decreases at lower eventrate, we attribute the reason to the pile-up at the input node of Jin which induces a small gate-source capacitance, drain-source resistance and transconductance variation which might affectthe overall gain and deteriorate resolution, almost proportionally to the signal energy. Anyway,referring to the already mentioned setup, both signal satisfy the expectations. However, in ourtests the BEGe detector was operated in vacuum at liquid nitrogen temperature (TLN2 = 77K),

(a) “Fast” signal relative variation of the peakposition.

(b) “Fast” signal full width energy resolution.

(c) “Slow” signal relative variation of the peakposition.

(d) “Slow” signal full width energy resolution.

Figure 3. Gain stability and energy resolution performance of GeFRO coupled with BEGedetector. Signal was obtained using 22Na and 228Th γ-sources and the average acquisition ratewas about 1300 events per second.

although in GERDA it will be operated naked in liquid argon (TLAr = 90 K) which behavesalso as shielding material against external radiation background. Moreover, the signal pin ofthe BEGe detector will be connected to the front-end by means of a bonding gold wire. Hence,fluctuation of the total input capacitance might occur in case of bonding wire vibration due toa turbulence in the cryogenic liquid. That condition might cause a gain instability fluctuationof a few tens of ppm on the “fast” output, but it is expected to weakly affect the pulse shapeanalysis efficiency which does not depend on the gain value. So, in the final GERDA setup, the“slow” signal should be used for a stable and well-resolved energy spectra acquisition, while the“fast” waveform could be analyzed for multi-site γ events rejection.

References[1] Abt I et al. 2004 Proposal to the LNGS P38[2] de Orduna R G, Hultand M, Andreotti E, Marissens G, Budjas D and Misiaszek M 2010 JRC Scientific and

Technical Reports[3] Cassina L, Cattadori C, Giachero A, Gotti C, Maino M and Pessina G 2013 Submitted to Transactions on

Nuclear Science (TNS) ArXiv:1307.5233 [physics.ins-det][4] Cattadori C, Gallese B, Giachero A, Gotti C, Maino M and Pessina G 2011 Journal of Instrumentation

(JINST) 6 P05006[5] Ackermann K H et al. 2013 EPJC 73 P2330

Development of cryogenic readout circuit for

far-infrared image sensors with Fully Depleted

Silicon on Insulator CMOS process

Koichi Nagase1, Takehiko Wada1, Hirokazu Ikeda1, Yasuo Arai2 andMorifumi Ohno3

1 Institute of Space and Astronautical Science (ISAS) Japan Aerospace Exploration Agency(JAXA)2High Energy Accelerator Research Organization3National Institute of Advanced Industrial Science and Technology

E-mail: nagase@ir.isas.jaxa.jp

Abstract.We are developing far-infrared image sensors for astronomical observations. In far-infrared

observations, detectors and readout circuits must be cooled under liquid helium temperature (4.2K) to reduce thermal background. We designed and evaluated the readout integrated circuitsthat have good characteristics under 4.2 K using fully-depleted silicon-on-insulator (FD-SOI)CMOS. We demonstrated proper operations of the CMOS operational amplifier, analog switchand elementary logic circuits at 4.2 K.

1. IntroductionLarge format image sensors are very important for astronomical observations because they makeit possible for us to have both wide field of view and high spatial resolution, simultaneously.Unfortunately, there are small format image sensors for far-infrared (FIR) astronomy. Our aimis to develop large format FIR image sensors.

The FIR image sensors consist of germanium photoconductor, operational amplifier(OPAMP), reset switch and X-Y selectors for multiplexing. The structure and the schematicdiagram of the image sensors are shown in figure 1 and 2, respectively.

FIR image sensors must be operated at cryogenic temperature in order to reduce the thermaldark current. As a result, the readout integrated circuit (ROIC) must be operated at cryogenictemperature with low power consumption. Especially, it is essential to reduce the powerconsumption per a pixel in order to realize large format FIR image sensors. The previousROIC has high power consumption (10 µW/pixel) because they are based on PMOS circuit[1][2][3]. In order to realize low power consumption (1 µW/pixel), CMOS architecture must beused. However, conventional CMOS ROIC does not work well because of unstable operation ofNMOS FET below 30 K [4][5].

Our approach is to use FD-SOI CMOS technology in which both NMOS and PMOS workproperly at cryogenic temperature. We have already shown that both P-channel and N-channelFET which is fabricated by a commercial foundry (LAPIS Semiconductor Co., Ltd.) work

properly at cryogenic temperature [6]. We have also developed CMOS OPAMPs and logics [7][8].In this paper, we review our previous works on the cryogenic FD-SOI CMOS and report therecent results of the development of CMOS analog switches which are necessary for multiplexing.

Figure 1. Structure of the image sensorsFigure 2. Schematic diagram of the imagesensors

2. Cryogenic characteristics of FD-SOI MOSFETNMOSs fabricated by conventional bulk-CMOS process suffer from kink and hysteresis becauseof impact-ionization and carrier freeze-out [4][5]. Kink phenomenon is caused by the followingreasons. The mobility of electrons is faster at low temperature than that at room temperature.Accelerated electrons make additional holes and electrons by impact-ionization. The excesscurrent by the holes and the electrons make kink in I-V curve. Hysteresis phenomenon atcryogenic temperature is caused by following reason. At low temperature, impact-ionized carrierscan not drain into the substrate because of carrier freeze-out. Then, the voltage of the substratebecome unstable. At room temperature, there is no carrier freeze-out and they can drain intothe substrate.

FD-SOI MOSFETs are not affected by kink and hysteresis because their body is very thin andis fully depleted. Impact-ionized carriers go rapidly to drain or source and do not accumulate inthe body. The figure 3 and 4 show the I-V curve of FD-SOI MOSFETs with W/L=0.63 µm/5.0µm at 4.2 K. There are no kink effect both NMOS and PMOS in the drain-source voltage rangebetween 0 and 1 V.

3. The cryogenic circuits using FD-SOI CMOSWe successfully developed the cryogenic OPAMP for FIR image sensors [7][8]. Figure 5 showsthe schematic. Our OPAMP has PMOS FET inputs (M2 and M3). The gate length and widthare 5 µm and 5 µm, respectively. Two-stage cascodes (M6/M8 and M7/M9) consist of NMOSFETs. The high resistance load obtained by the NMOS cascodes give us a high gain. OurOPAMP has a open-loop gain which is over 7000. Output node is a source follower (M11).Table 1 shows the performances of the OPAMP at 4.2 K.

4. CMOS analog switchWe use analog switches to multiplex the signal from each pixels. Our CMOS analog switchconsists of W/L=0.63/5.0 µm body-tied MOSFETs (Fig.6). We successfully demonstrated theproper operation of the CMOS analog switch at 4.2 K (Fig.7).

We also measured ON resistance of the CMOS analog switch at 4.2 K (Fig.8). ON resistancewas estimated by R=V/I (V: Voltage between input and output, I: Output current). ONresistance of our CMOS analog switch is sufficiently low for multiplexing.

Figure 3. I-V curves of the FD-SOINMOS with gate-source voltage of 0.900 V,0.950 V, 1.000 V, 1.050 V and 1.100 V

Figure 4. I-V curves of the FD-SOIPMOS with gate-source voltage of -1.250V, -1.375 V, -1.500 V, -1.575 V, -1.650 V

Figure 5. Circuit diagram of ouroperational amplifier

Vdd

Inverter

Vdd

M1

M2

Input Output

Control

Figure 6. Circuit diagram of the CMOSanalog switch

Table 1. Performances of the FD-SOI CMOS amplifier at 4.2 K [7][8]

design measurement

Open loop gain ＞ 1000 ＞ 7000Power consumption 1.1 µW 1.3 µWOutput Voltage swing ＞ 1V 1.3 V

Input referred noise at 1 Hz 14-20 µV/√Hz 19 µV/

√Hz

Input offset voltage 0 mV 2mVVariation of input offset voltage 0 mV 4.2 mV(1σ)Leak current of reset switch 0.1 fA 0.1 fA

5. Future planWe successfully developed CMOS OPAMPs, logics and analog switch. We have already designedand fabricated 2×2 pixels readout array (Fig.9). We will evaluate the 2×2 pixels array in nextyear. We will also develop 32×32 pixels array in next two years. We are planning to make

Figure 7. Signal of the CMOS analogswitch at 4.2 K measured by an oscilloscopewith 1 MΩ input impedance. Top is theinput signal (+0.5 V for clearance). Centeris the output. Bottom is the control signal(×1/10− 0.5 V)

Figure 8. ON-state resistance of theCMOS analog switch at 4.2 K as a functionof input voltage. VDD and VSS are +3 Vand -3 V. The voltage of control signal is3.0 V.

astronomical observation with balloon telescopes using our FIR image sensors in next five years.

Figure 9. Array configuration

AcknowledgmentThis work was supported by JSPS KAKENHI Grant Numbers 23340053 and 25109005.

References[1] H.Nagata, et al., IEEE Transactions on electron devices, VOL. 51, No 2, 270, (2004)[2] T. Hirao et al., Advances in Space Research, vol.30, pp.2117-2122, (2002)

[3] M. Kawada et al., PASJ, Publ. Astron. Soc. Japan 59, S389-S400 (2007)[4] Robert M. Glidden et al., SPIE, Infrared readout electronics, Vol.1684, (1992)[5] B.Dierickx, L.Warmerdam, E.Simoen, J.Vermeiren and C.Clays, IEEE Transactions on electron devices, VOL.

35, NO. 7, (1988)[6] H. Nagata, T.Wada, H.Ikeda, Y.Arai and M. Ohno, AIP Conf. Proc. 1185, 267, (2009)[7] H. Nagata, T.Wada, H.Ikeda, Y.Arai, M.Ohno and K.Nagase, IEICE Trans. Commun. Vol.E94-B, 2952, (2011)[8] T. Wada, H. Nagata, H. Ikeda, Y. Arai, M. Ohno and K. Nagase, J Low Temp Phys 167:602-608 ( 2012)

On the MOSFET-based Temperature Sensitive Element for

Bolometer Application

E.Fuxa (1), J.J.Yon (1), J.Jomaah (2)

(1)CEA LETI, Grenoble, France

(2)IMEP, Grenoble, France

etienne.fuxa@cea.fr

Abstract. This paper focuses on the study of thermal performances of MOS transistors for

bolometer applications. Series of measurements were conducted to obtain TCC (Temperature

Coefficient of Current) versus gate voltage and temperature curves. The measurements were

confronted to atlas simulations, and showed that in the subthreshold region the TCC ranges

from 4%/K all the way to 9%/K. It was also determined that gate length does not have an

influence on the TCC until short channel effects factor in.

1. Introduction

Bolometers are uncooled infrared detectors that can be broken down in three main parts: an

absorption layer that will absorb the incoming infrared radiation, a sensitive element that is thermally

connected to the absorber and act as a thermometer, and a ROIC (Read-Out Integrated Circuit). The

most common devices used as the temperature sensitive elements are thermistor materials, mainly

vanadium oxides (VOx) and amorphous silicon (a-Si). Their TCR (Temperature Coefficient of

Resistance) is usually between -2%/K and -4%/K [1-2]. It was shown that quantum wells can be used

as thermistors with TCR of 3%/K and with values of over 4%/K considered attainable after

optimization of the structure [3]. Diodes can also be used as sensitive element with even higher

temperature dependency, up to 7%/K [4]. Finally, SOI-CMOS transistors are also being used and have

been shown to exhibit TCC close to 6%/K close to the operating point [5].

2. Measured devices and atlas simulations

The measured devices were 180nm technology node NMOSFETs with 3.3V operating voltage, and

were available with 20µm and 0.340µm channel length. Atlas simulations were done on a simplified

(ideal) MOS model with adjustable geometry, as show Fig. 1 hereafter:

Figure 1: Atlas mosfet used for simulations

It was drawn deep enough to ensure a bulk behaviour as the silicon would never be partially or

completely depleted.

3. Experimental method

The drain current was measured with a gate voltage sweep for two different values of the source-

drain potential difference: 50mV and 3V so that we could compare linear and saturation regimes while

the bulk is connected to the ground. Since the bolometers are uncooled devices, the measurements

were conducted from 297 to 339 kelvins with steps of 2 kelvins. Measurements were done with two

transistors that only differed in gate length: 20 microns or 340 nanometers to check short channel

effects influence on the TCC.

4. Experimental results

First we show figures of drain current and transconductance versus gate voltage with linear and

logarithmic scale for the current, for measurement and models. Lines are Atlas simulated curves while

marks are measured data.

Figure 2: Ids versus gate voltage – logarithmic scale

Figure 3: Ids versus gate voltage – linear scale

Figure 4: gm versus gate voltage – logarithmic scale

Figure 5: gm versus gate voltage – linear scale

These show that the model is a close fit for the measurements in the subthreshold region and for

high gate voltages even though precise parameters for the Lombardi model used for the mobility in

Atlas were not extracted, default values of the material were used instead. That is due to the fact that,

as will be shown, the region of interest is the subthreshold region and thus mobility model, specifically

its behavior with longitudinal and transverse fields, is not as critical.

Below is a figure showing the TCC versus gate voltage for the long gate (20 microns) transistor.

Figure 6: TCC versus gate voltage

We can see that in both the simulation and the measurement, the TCC is much higher in the

subthreshold region and is very sensitive with gate voltage. This is something of importance as it will

have a large influence for defining the way the voltages will be controlled and so in the definition of

the ROIC.

Next page are the results of Atlas simulations, that allowed to plot the TCC of several transistors

that only differed in gate length. We used varied lengths ranging from 50 nanometers to 20

micrometers here.

Figure 7: TCC versus gate length for large lengths

Figure 8: TCC versus gate length for short lengths

We can see on figure 7 that for gate lengths from 1 to 20 microns, the TCC values exhibit no

difference: the difference between the 10 microns and 20 microns simulations is always less than one

percent. However, for a gate length of 340 nanometers or less the TCC is lower as is showed in figure

8 (the 20µm channel simulation is shown here again for comparison). This can be attributed to the

short channel effects that factor in for such small lengths.

For long channel bulk MOSFETs, the approximate formulas for the drain current in the three mode

of operation are as follows [5]:

(

)

( )

( ) (

) (1)

( )

(2)

[( )

] (3)

Where

and are the depletion, fast surface state and oxide capacitances

and is the threshold voltage. Considering that at close to room temperature,

and that we

can neglect the variations of the n parameter with temperature as it was extracted to show a linear

behavior with a 10-4 slope with temperature [5] which would imply a contribution of approximately

0.01% to the TCC, we can obtain approximate formulas for the TCC:

( )

(1)

(2)

(3)

In the subthreshold region, the influence of the drain to source voltage on the TCC would be

negligible for larger than 100mV values because of the exponential factor in eq. 4. For very small

values the TCC would slightly decrease with lower voltages. The first two terms are both positive and

of the same order of magnitude and show that in this region the TCC would decrease with rising

temperatures.

By definition, the current does not depend on drain to source voltage in the saturation region so it is

obvious that neither will the TCC. In ohmic mode however, higher drain to source voltages would

hamper the TCC as long as the second term is higher than the first one that is due to the mobility

dependency on temperature. For higher gate voltages the latter becomes predominant and the TCC

changes sign, eventually diminishing the drain to source voltage’s influence.

It can be noted that in accordance with the statements above, both simulations and measurements in

the figure below show that the TCC grows with lower temperatures in the region of interest for

bolometers applications. It is important for this application as it can prevent thermal runaway.

Figure 9: TCC versus temperature

Finally simulations were conducted to confirm the theoretical behaviour of the TCC expected with

drain to source voltage variations.

Figure 10: Influence of the drain to source voltage on the TCC

The figure 10 above shows TCC versus gate voltage curves for three different drain to source

voltages. On the right are close-up views of the subthreshold (top) and saturation/linear (bottom)

regions. As expected the TCC shows little dependency on drain to source voltage in the subthreshold

region even though the 50mV curve is very slightly below the other two, and no dependency at all in

saturation mode, but its influence shows in the ohmic region: for 1V (resp. 50mV) drain to source

voltage, the mode switches from saturation to ohmic when the gate voltage nears 1,6V (resp. 0.7V)

and we can see the curves diving a little, confirming theoretical analysis.

5. Conclusions

We have shown that bulk MOS transistors in the subthreshold regime exhibit TCC much higher

than state of the art bolometers TCRs, and that the TCC is independent of gate length for large

transistors. More precisely, gate length has no influence as long as short channel effects are absent;

when they occur they then tend to have a negative influence on the TCC. We have also seen than in

the subthreshold area the TCC is quite stable with drain-source voltage variations.

Acknowledgment

We would like to thank the team at IMEP that helped to obtain the experimental data and Jean

Louis Ouvrier Buffet for his Atlas expertise.

References

[1] Moreno M, Torres A, Ambrosio R, Kosarev A, "Un-cooled microbolometers with amorphous

germanium-silicon (a-GexSiy:H) thermo-sensing films", 2012

[2] Vedel C, Martin J.L, Ouvrier Buffet J.L, Tissot J.L, Vilain M, Yon J.J, "Amorphous silicon

based uncooled microbolometer IRFPA, 1999

[3] Lapadatu A, Kittisland G, Elfving A, Hohler E, Kvisterøy T, Bakke T, Ericsson P, "High-

performance long wave infrared bolometer fabricated by wafer bonding", 2010

[4] Kropelnicki P, Vogt H, "A new DC-temperature model for a diode bolometer based on SOI-pin-

diode test structures", 2010

[5] Socher E, Beer S.M, Nemirovsky Y, "Temperature sensitivity of SOI-CMOS transistors for use

in uncooled thermal sensing", 2005

Modeling and Experimental Investigation of HF noise of SiGe:C HBTs at Cryogenic Temperature

L M Diaz-Alabarran1, E Ramirez-Garcia1 1,2, M C Galaz-Larios1, F Aniel2, N Zerounian2 and M A Enciso-Aguilar1 1 Instituto Politécnico Nacional, UPALM, Edif. Z-4 3er piso, C.P. 07738, México D.F, México, 2 Institut d’Electronique Fondamentale, Univ. Paris-Sud, CNRS UMR 8622, 91405 Orsay, France. Abstract. SiGe heterojunction bipolar transistors (HBT) have demonstrated unity gain transition frequencies higher than 0.5 THz at low temperatures for the current gain (fT), and at room temperature for the maximum oscillation frequencies (fMAX ). Most of the existing high-frequency (HF) noise models neglect the thermal noise contribution produced by electrons at the base emitter (BE) junction leading to unreliable noise performances for the development of THz applications. This noise source becomes non negligible when operating frequencies (f) are close to fT. An experimental and modeling study of HF noise at room and at low temperatures (40 K) is presented using the equivalent circuit in Π-configuration. The modeling is based on a proper extraction of the parameters of the small-signal equivalent circuit. For a 0.13 µm emitter wide HBT with a germanium profile of 5-30%, at optimum dynamic performances the noise resistance is underestimated by 20% when BE thermal noise is neglected. Minimum noise factor is impacted with a difference of 17.5%.

1. Introduction SiGe heterojunction transistors (HBT) are credible contenders to develop applications in the terahertz (THz) domain. Recently, different technologies of SiGe HBTs have already exhibited unity gain transition frequencies higher than 0.5 THz at low temperatures for the current gain (fT) [1], and already at room temperature for the maximum oscillation frequencies (fMAX ) [2]. Moreover, to develop THz applications a reliable knowing of the HF noise performances is crucial. In [3] two different technological variations of SiGe HBT were studied to investigate their fT/fMAX and HF noise performances at cryogenic temperatures. A good agreement is found between model and measurements of HF noise at 300 and 77 K and for operating frequencies (f) up to 40 GHz (f << fT, with fT,max greater than 250 GHz). However, as discussed in [4] most of the HF noise models are limited because they neglect the thermal noise contribution produced by electrons at the base emitter (BE) junction (TNe), which is not negligible when f ≈ fT [4]. Hence, it is necessary to consider TNe to correctly assess noise performances (at room and at low temperatures) for operation frequencies f near fT. We present an experimental study and the modeling of the HF noise performances of the HBT, at room and at low temperatures (40 K), using the equivalent circuit in Π-configuration. The modeling is based on a proper extraction of the parameters of the small-signal equivalent circuit. 1ramirezg@ipn.mx

In section two, we introduce the device specifics, the measurement set-ups and the electrical circuit. In section three, results of the experimental and modeling HF noise are discussed, and finally we draw some conclusions. 2. Device Characteristics, Measurement Set-ups and Noise Modeling The npn SiGe:C HBTs used for the study, hereafter referred as HBT are developed by STMicroelectronics. These transistors have a fully self-aligned architecture, using a selective implanted collector and a selective epitaxial growth of the intrinsic base layer through the emitter window. This architecture helps to improve the maximum oscillation frequency (fMAX ) in comparison to a quasi-self-aligned architecture and to non-self-aligned architectures. Due to the very narrow design of the emitter junction width, and of the emitter to extrinsic base layer distance, the base resistance and junction capacitances are reduced. More details about the device fabrication can be found in [5]. Table 1 introduces some technological parameters of HBTs used for the study with: the base thickness (WSiGe), the germanium content of the base layer (%Ge) at the emitter side of the base (the lower value) and at the collector side, the maximum boron doping level in the base (NA,pk). Further, the emitter area of HBT is SE = 0.13×5.74 µm2.

Table 1. Some technological parameters of the HBTs, and measured maximum fT at 300 and 40 K.

Device WSiGe (nm)

Ge (%)

NA,pk (×1019 cm-3)

fT,max (GHz) 300 K/40 K

B1 24 5–20 2.9 216/295

B2 24 5–25 2.6 238/326

B3 24 5–30 2.5 267/376 We performed DC, S-parameters up to 45 GHz at room and low temperature, and high-frequency noise (HFN) measurements between 0.5 GHz and 18 GHz with an in-house noise measurement setup, only at room temperature. The noise measurement principle is described in [6]. A SOLT calibration is performed up to the coplanar probes, and an open noise de-embedding technique [7] is used to subtract the contribution in noise and in S-parameters of the access lines and pads. Table 1 gives maximum values of fT reached at optimum collector current density (JCopt of about 13-15 mA/µm2), at 300 and 40 K. The small-signal equivalent circuit in Π-configuration is represented in figure 1 (a), with a thermal noise source associated to each resistor and two correlated shot noise sources for the base-emitter (BE) and the base-collector (BC) junctions. The noisy electrical circuit and the noise analysis were performed using Advance Design System (ADS) from Agilent. We use an analytical-based method to extract a part of the small-signal parameters at 300 K and at 40 K. The emitter resistance (RE) is obtained from the extrapolation of Re(Z12) at infinite emitter current. The apparent base resistance (RB) can be obtained from Re(Z11–Z12) in the low frequency region of the measurement and the base-collector capacitance (CBC) is extracted from Im(Z22–Z21). Remaining parameters such as the BE dynamic resistance (RΠ), the BE capacitance (CΠ), the transconductance (gm), and its high-frequency in-excess delay (τΠ) are determined by error minimization between measured and simulated S-parameters. At low temperature, for the division of the total base resistance RB into each different contributions, such as RB ≈ RBx + X RBi [8], where X is the base-collector distribution factor, RBx and RBi are the extrinsic and intrinsic base resistances, respectively, we use the approach described in [9] for room temperature, and we consider X300 K = X40 K.

With the equivalent circuit extracted from S-parameters measured at 40 K, we can model the noise performance at this low temperature. The noise model used in this work was described in [4]. Noise sources take into account the self-heating, respecting a thermal resistance (RTH) for HBTs. At room temperature, values of RTH are taken from [10] and values are deduced at 40 K with RTH,40 = RTH,room/10, due to the increase of the thermal conductivity of Si at low temperature. For TNe produced at the BE junction, this noise source results from the diffusive nature of electrons as already described in [4] and references therein. TNe is taken into account within the two correlated noise sources i1 and i2 (see relations (4), (8), (9) and (10) in [4] for a complete description). Another important parameter to properly model the noise performances is the noise transport time, τ equals to τB + τC, where τB and τC are the base and the collector transit times, respectively. For more explanation, see [4] and references therein. τB and τC are extracted using the approach presented in [11]. This technique is based on hydrodynamic modeling of the HBT to compute how the electron transit time is divided into each subsequent region from emitter to collector terminals. The total transit time can be extracted from S-parameter measurements. For HBTs investigated here, the value of τ is given in table 2 (other parameters of table 2 are used in section 3). For the collector charging time RCCBC, it is about 0.062 ps at 300 K [11]. At 40K, CBC and RC decrease of about 13% and RCCBC,40 K is estimated to be equal to 0.054 ps. The decrease of the value of τ between 300 and 40 K is due to the increase of the part of the electrons in a non-stationary transport regime, enhanced by the higher Ge gradient between B1 and B3 that induces a faster transit of electrons through the base layer. Figure 1 (b) shows a satisfactory agreement between the simulated (continuous grey lines) and measured (circles) S-parameters of B1 at the two temperatures. For the three devices, the agreement stands for all bias levels.

X CBC

E

B C

RE

RBi

X CBC(1 )−

VBER Π CΠ

gm ⋅e VBE-j ω τ

22i

21i

RBx

2

BxRv

2

ERv

2

BiRv RC2

CRv

Π

21S

12S

11S22

S

18 GHz

1 GHz

/ 10Meas.Model

21S

12S

11S 22

S

40 GHz1 GHz

/15

(a) (b)

Figure 1. (a) Noisy electrical circuit in Π configuration (b) Modeled and measured S-parameters at room temperature (at 40 K in inset), for the device B1, biased at JC = 8 mA/µm2 (7 mA/µm2),

VCE = 1.28 V (1.43 V) at 300 K (40 K).

Table 2. Extracted noise transport time (τ = τB + τC), and simulated NFmin and Rn/Z0 (= 50 Ω), biased at JC ≈ JCopt. Superscripts: 1at f = 1GHz, 2at f ≈ fT,max with TNe, and 3at f ≈ fT,max without TNe.

Device ττττ (ps)

300 K/40 K

1NFmin (dB) 300 K/40 K

2NFmin (dB) 300 K/40 K

3NFmin (dB) 300 K/40 K

2Rn/Z0 300 K/40 K

3Rn/Z0 300 K/40 K

B1 0.48/0.38 2.3/0.79 5.0/3.5 4.7/3.1 0.96/0.14 0.91/0.13

B2 0.43/0.35 2.0/0.47 5.2/5.1 4.8/4.5 0.83/0.24 0.80/0.20

B3 0.38/0.30 1.85/0.13 5.3/7.0 5.0/6.3 1.26/0.48 1.17/0.4

3. High Frequency Noise Analysis For the device B1, figure 2 (a) and (b) give a comparison between the model (continuous lines) and measurements (symbols) for the four noise parameters (4NP), i.e. the minimum noise figure (NFmin), the normalized equivalent noise resistance (Rn/Z0, where Z0 = 50 Ω), and the optimum reflection coefficient Γopt (magnitude and phase), at room temperature. Continuous lines in figure 2 (a) underline the influence of TNe included in noise sources in black lines and when it is not included in grey lines for frequency close to fT. It is shown that the HF noise model (black and grey lines) properly reproduces the measurement done between 1 GHz and 18 GHz. A similar agreement is found for the other two HBTs at all bias levels. These results along with the results of the previous section demonstrate that all the elements of the Π-model are correctly extracted. Table 2 draws the simulated values of NFmin and Rn at 300 and 40 K at f ≈ fT,max, with and without the influence of TNe, with the electrical model biased at optimum current density. Now we focus on modeled NFmin and Rn/Z0. Figure 2 (a) shows no difference between the noise model when TNe is considered (black lines) or not (grey lines) at operation frequencies around 100 GHz, i.e. f ≈ fT/2. At fT, the minimum noise factor Fmin (NFmin = 10⋅log(Fmin)) is 7% higher when TNe is considered. Regarding Rn/Z0 the results highlighted in figure 2 (a) indicate that for operation frequencies lying below fT/2 the difference is less than 1% with or without TNe. At fT, Rn/Z0 is increased by 6% when TNe is taken into account.

0

1

2

3

4

5

6

0

1

1 10 100

NF

min (

dB

)

Rn /Z

0

f (GHz)

0

0.5

1

0

1

2

1 10 100

|Γ op

t|

Ph

ase Γ opt (rad)

f (GHz) (a) (b)

Figure 2. Measured –symbols– and modeled –black (with TNe) and grey lines (without TNe)– four noise parameters in function of the frequency at 300 K,

device B1 biased at JC ≈ 8.1 mA/µm2 and VCE = 1.28 V.

Modeling results are shown at 40 K in figure 3 for the device B3, with NFmin and Rn/Z0 as a function of the frequency f. Black lines represent the modeling results for the case when TNe is considered and in grey lines when TNe is neglected. As for results at room temperature, there is not any difference in NFmin and Rn when TNe is considered or not for frequencies less than 150 GHz ( ≈ fT/2). For the upper frequency range (f > fT/2) if TNe is neglected the model underestimates Fmin and Rn by 17.5% and 20%, respectively. Such a misestimating is worse at low temperature than at room temperature. This difference arises from the correlation factor: as described in [4] when noise transport time tends to zero, so does the correlation factor, i1 and i2 become non-correlated and NFmin increases with the operation frequency. The results demonstrate that the contribution of TNe plays a growing role when the HBT dynamic performances are higher. This is the case at low temperature where dynamic performances are

enhanced from those at room temperature. Hence this noise source must be included in HF noise modeling.

0

5

0 100 200 300 400

NF

min (

dB

)

f (GHz)

0

0.5

1

0 100 200 300 400

Rn/Z

0 f (GHz)

(a) (b)

Figure 3. Modeled (a) NFmin and (b) Rn at 40 K in function of frequency, for the device B3 biased at JC ≈ JCopt, in black lines with TNe and in grey lines without TNe.

4. Conclusion We presented an experimental investigation and modeling results of HF noise at room temperature and at 40 K. The model used to carry out this study considers all significant contributions to noise including the thermal noise produced by electrons at the BE junction (NTe). We found that at both temperatures if this term is neglected minimum noise factor and equivalent noise resistance are underestimated by more than 7% for operating frequencies closed to fT. For the device with the best dynamic performances among the three devices presented, B3 with a germanium graduality of Ge = 5-30%, the situation is worse if NTe is neglected. Fmin and Rn/Z0 may be underestimated by more than 17%. This implies that NTe contribution should be included in HF noise models for applications requiring very high operating frequency reaching the maximum capability of the HBT. Acknowledgements Authors would like to thank the Instituto Politécnico Nacional (Mexico) for the financial support under the contract number ICyTDF/325/11. We thank also STMicroelectronics for providing the devices. References [1] Zerounian N, Aniel F, Chevalier P, Barbalat B and Chantre A 2007, Electron. Lett. 43, 774. [2] Heinemann B, et al. 2010 Proc. of the Int. Electron Devices Meeting 688. [3] Waldhoff N, et al. 2009 Proc. of the Eur. Solid-State Device Research Conf. 121-124. [4] Ramirez-Garcia E, et al. 2012. Sem. Sc. and Tech. 27 065016. [5] Chevalier P, et al. 2005 J. of Solid-State Circ. 40 2025-2034. [6] Danelon V, et al. 1998 Electron. Lett. 34 1612–1613. [7] Aufinger K and Böck J 1996 Proc. of the Eur. Solid-State Device Research Conf. 957–960. [8] Zerounian N, et al. 2009 Solid-State Electr. 53 483–489. [9] Ramirez-Garcia E, et al. 2012 Sem. Sc. and Tech. 27 085005. [10] Barbalat B, et al. 2005 Proc. of the Eur. Solid-State Device Research Conf. 223–6. [11] Ramirez-Garcia E, et al. 2011 Solid-State Electr. 61 58–64.

Development of the Cryogenic Integrated Circuits withGaN/AlGaN HEMTs and the New High Sensitivity TerahertzDetectors

Yasunori Hibi1*, Jiandong Sun2, Hua Qin2, Hiroshi Matsuo3, Lin Kang1, JianChen1**, and Peiheng Wu1

1Research Institute of Superconductor Electronics (RISE), Nanjing University,Nanjing 210093, China2Suzhou Institute of Nano-Tech and Nano-Binics (SINANO), Chiniese Academy ofScience (CAS), Suzhou 215123, China3National Astronomical Observatory of Japan (NAOJ), Mitaka, Tokyo 181-8588, Japan

E-mail: *yasu0318@nju.edu.cn, **chenj63@nju.edu.cn

Abstract. GaN/AlGaN High Electron Mobility Transistors (HEMTs), which are originallyused for the purpose of high-speed operation at room temperature, they also have goodperformances at cryogenic temperature. We have developed cryogenic integrated circuits (ICs)with the HEMTs and examined applications using these ICs. We have investigated thecryogenic properties of the single HEMTs and we are designing some basic ICs, presently.We see that the GaN/AlGaN HEMTs and superconductor devices can be fabricated andintegrated on same substrate. On chip integration allows for compact cryogenic instrumentrealization. This also allows for wire number reduction. Hence, the reliability of instrument isnot only improved, the stray capacitance is but also eliminated. Therefore, it becomes possibleto make apparatus with improved speed and lower power consumption.That both devices can fabricate on the same substrate means that development of newequipment is attained. We have started to evaluate two types of photon detectors. One is thenear infrared photon counter which unified the multi-channel serial connected SNSPDs and theserially connected cryogenic readout circuit. The other is a sensitive terahertz photon detectorby employing photon to charge conversion in an integrated circuit including a HEMT withnormal-metal gate and superconducting terahertz photon absorber.

1. IntroductionTo show the best performance of cryogenic detectors, signal readout system should be set beside andoperated in cryogenic temperature. Almost all of the Si semiconductor devices cannot work incryogenic temperature on the low power consumption conditions because of carrier freeze out. Thoughthere are some developments which make the cryogenic operation possible [1-5], those investigationsused high doped substrate [1], took special circuit design [2] and used special channel structure [3,4].On the other hand, III-V semiconductor devices work in cryogenic temperature although there aresome performance indexes variations [5].

HEMT (High Electron Mobility Transistor) is a well-known circuit element with high speed andlow noise. Since LNA (Low Noise Amplifier) for microwave amplification is always constructed bysome HEMTs and so on, HEMT is workable in cryogenic temperature at very high frequency. On the

other hand, Dong [6] and Liang [7] showed that AlGaAs/GaAs HEMTs have good static propertiesand very low voltage noise in cryogenic temperature on the low power consumption conditions. Thesefacts indicate that HEMT is a good candidate for cryogenic circuit element for low frequency usage.

In this paper, we show some cryogenic low frequency performances (static properties and lowfrequency noise) of GaN/AlGaN HEMT on low power consumption. We pay attention about HEMTsand superconductor devices can be fabricated onto same substrate and propose some plans of newcryogenic detectors.

2. HEMT and HEMT Integrated Circuit (IC)Our evaluation targets are GaN/AlGaN HEMTs fabricated in SINANO (Suzhou Institute of Nano-Tech and Nano-Bionic). These HEMTs are fabricated onto sapphire and silicon substrate. Weinvestigated cryogenic performances of single HEMT. Both of them have good cryogenic propertieson low power consumption conditions. In this paper, we show the cryogenic performances of HEMTfabricated onto silicon substrate and its gate size is 13 m width and 2 m length. In figure 1, roomand cryogenic temperature static characteristics are shown. By cooling down, threshold voltagechanges about 200 mV. Measured gate leak current was about 50 pA when Ids = 1 A. This valueshould be upper limit of the measurement system.

We measured input voltage noise and current noise. Typical voltage noise Power Spectrum Density(PSD) and the voltage noise dependence to drain to source current (Ids) are shown in figure 2. In SipMOSFET or GaAs JFET case, deep cryogenic temperature (<4 K) voltage noise is always higherthan room temperature noise. In this case, cryogenic noise is about 10 times smaller than roomtemperature noise. When drain to source voltage (Vds) was + 1.0 V and Ids was 1.0 A, the inputvoltage noise was 70 nVrms/Hz0.5 at 1 kHz. This value seems to be the minimum value at thismeasurement interval, voltage noise increases even if Ids is increased or decreased. When 1/f0.5

dependence is assumed to the form of noise PSD, the input voltage noise at 1 Hz should be about 2Vrms/Hz0.5. This value is almost equal to that of Si pMOSFET [3] or GaAs JFET [5]. When Ids was100 A and Vds was +0.5 V at cryogenic temperature, the voltage noise is higher than the other Ids

cases. This is because the HEMT is operated in the transition region between linear and saturate region.About current noise, in case of Vds was + 1.0 V and Ids was 10 A, it was 0.25 pArms/Hz0.5 at 1 kHz.Since we used commercial JFETs (2SK30, 2SK246) in place of the HEMT and we got 0.15pArms/Hz0.5 at 1 kHz current noise, this HEMT current noise may be treated as upper limit.

To realize large scale cryogenic HEMT ICs, now we are designing and fabricating some smallscale ICs. Concrete contents are differential amplifiers, high speed amplifiers and sample-and-hold.We are also fabricating resistances, capacitances and diodes in same fabrication process.

Figure 1. Static characteristics of the GaN/AlGaN HEMT at room and cryogenic temperature. a) is Ids

- Vds at room temperature. Vgs were -1.80 V, -1.75 V, -1.70 V, -1.65 V, -1.60 V and -1.55 V and their

-2 10-6

0

2 10-6

4 10-6

6 10-6

8 10-6

1 10-5

0 0.5 1 1.5 2 2.5 3

Ids-V

ds@300K

Vgs=-1.80VVgs=-1.75VVgs=-1.70VVgs=-1.65VVgs=-1.60VVgs=-1.55V

Vds(V)

-2 10-6

0

2 10-6

4 10-6

6 10-6

8 10-6

1 10-5

0 0.5 1 1.5 2 2.5 3

Ids-V

ds@4K

Vgs=-1.55VVgs=-1.50VVgs=-1.45VVgs=-1.40VVgs=-1.35VVgs=-1.30V

Vds(V)

-1 10-5

0

1 10-5

2 10-5

3 10-5

4 10-5

5 10-5

-2 -1.8 -1.6 -1.4 -1.2 -1

Ids-V

gs(V

ds=+3.00 V)

@300K@4K

Vgs(V)

a) b) c)

symbols are open circle, filled circle, open square, filled square, open diamond and filled diamond,respectively. b) is Ids - Vds at cryogenic temperature. Vgs were -1.55 V, -1.50 V, -1.45 V, -1.40 V, -1.35V and -1.30 V and their symbols are open circle, filled circle, open square, filled square, opendiamond and filled diamond, respectively. c) is Ids - Vgs at room and cryogenic temperature, so calledthreshold curves. Vds was + 3.00 V. Open circle and filled circle are indicated at room and cryogenictemperature, respectively.

Figure 2. Measured results of the GaN/AlGaN HEMT voltage noise. a) is typical output powerspectrum density (PSD). This PSD was taken at Ids = 1 A, Vds = + 1.0 V and 4 K. Frequency rangeis between 10 Hz and 3.2 kHz and voltage range is between 1 Vrms/Hz0.5 and 1 Vrms/Hz0.5. To getinput voltage noise, those values should be divided by 1,100. Though there are many harmonic noise,we think they are derived from power source of the laboratory. b) is the relation between Ids and inputvoltage noise at 1 kHz. Open symbols and filled symbols are noise at room temperature and noise at 4K, respectively. Vds of circle symbols are + 0.5 V, square symbols are + 1.0 V and diamond symbolsare + 3.0 V, respectively.

3. Integration of Superconductor and Semiconductor DevicesIf superconductor and semiconductor devices can be fabricated onto same substrate, the number of I/Olines and the volume of instruments could be reduced. As mentioned above, GaN/AlGaN HEMTshave enough cryogenic properties for low frequency circuit elements and the ICs constructed by theseHEMTs are able to work in cryogenic temperature. Since both superconductor devices andGaN/AlGaN HEMTs are fabricated onto sapphire or silicon substrate, monolithic hybrid cryogenicdevices can be realized. We have started to investigate this integration technology.

There are some merits of integration superconductor and semiconductor devices except forreduction of the volume of instrument. First is that this technology makes electric connectionreliability higher. Second is that this makes stray capacitance smaller. Reduction of stray capacitancebecomes possible to make apparatus with improved speed and lower power consumption.

4. Two New Detector Plans Making Use of Integration TechnologyWe propose two detector plans realized by this integration technology.

As first, although the concept of this detector proposed by Jahanmirinejad and Fiore [8], it is a kindof photon number resolving detector which is constructed by many serial connected superconductornanowire single photon detectors (SNSPDs) and signal readout HEMT. Jahanmirinejad and Fiore

required to the signal readout device to have high input impedance (>1 M), low input capacitance(<0.1 pF) and high speed (> GHz). Because gate capacitance of GaN/AlGaN HEMT with its gate size13 m by 2 m is about 80 fF, our integration technology could satisfy these requirements especiallyfor such low input capacitance. If SNSPDs and the cryogenic readout device are made separately,large stray capacitance is unavoidable arisen by electric connections of bonding wires and bondingpads. This concept is shown in figure 3. We already investigated good SNSPD photon counter [9,10],next step is to co-fabricate SNSPDs and GaN/AlGaN HEMT readout circuits onto same substrate. Inrealization of this detector, it is important that how cut heat dissipation by the HEMT formsuperconductor detectors.

Secondly, it is sensitive photon detector which is constructed by superconductor terahertz photonabsorber and HEMTs charge sensing circuit. The concept of this detector is shown in figure 4. Thisconcept is basically same as charge sensitive infrared phototransistor (CSIP) [11]. The differencepoints of this detector from CSIP are that our detector uses superconductor absorber for photonabsorption and HEMT gate for charge sensing. By using Nb/NbN photon absorber, detectivefrequency can be extended to more than 1 THz. Terahertz photon is absorbed at NbN bow tie antennaand generates quasi-particles in Nb line. These generated quasi-particles are shut up in the Nb line andtransferred to normal metal wire. Transferred electrons are integrated at the HEMT gate capacitanceand sensed as Ids variation. The NEP of this type detector is limited by charge loss at charge sensingpart. Since present upper limit of the gate leak current is about 50 pA, the detector NEP can beestimated by using equation (22) of Komiyama 2011 [11],

NEP = ൫(2Γ)ଵ ଶ/ ℎߥ൯ />ߟ 1.2 × 10ଵ /ߟ [W/Hz0.5] (1)where is dark count rate, h is Planck constant, is frequency of observation target and is totaloptical efficiency, respectively. We set = 50x10-12/2e and = 1 THz, where e is unit of charge. Toimprove sensitivity, ultimately low gate leak current is required. We will confirm real gate leak currentof the HEMT by using the fabricating sample-and-hold circuit mentioned in section 2. If we getsmaller than 1x10-18 A gate leak current just like the cryogenic GaAs JFET [5], the detector NEP couldreach to the order of 10-22 W/Hz0.5.

Figure 3. Schematic diagram of the serialconnected multi-channel superconductornanowire single photon detectors (SNSPDs)with cryogenic readout amplifier. Enclosed partby dotted line is fabricated onto same substrate.

Figure 4. Schematic diagram of the chargesensitive HEMT amplifier with superconductorterahertz photon absorber.

SNSPD

No

rmal

Me

tal

RoomTemperatureAmplifierHEMT

Amplifier

NbN

NbHEMT(ChargeIntegration)

HEMT(Charge Reset)

Read Out

Vdd

Vss

RL

Vb

5. SummaryWe show that the GaN/AlGaN HEMT have good static characteristics, input voltage noise and currentnoise in cryogenic temperature. These mean that the cryogenic integrated circuits for low frequencyare able to be realized with the GaN/AlGaN HEMTs. We pay attention that the GaN/AlGaN HEMTsare fabricated onto sapphire or silicon substrate, we have investigated a technology whichsuperconductor and semiconductor devices are fabricated onto same substrate. By using thistechnology, reduction of volume, higher reliance and higher speed with limited power consumption ofsuperconductor instruments are expectable. We also proposed new cryogenic detectors which can berealized by using the integration technology superconductor and semiconductor devices onto samesubstrate.

AcknowledgmentsThis work was supported by the National High Technology Research Program of China (Grant No.2011AA010204), the National Basic Research Program of China (Grant Nos. 2011CBA001 and2014CB339800), the National Natural Science Foundation of China (Grant No. 11173015).

References[1] E. T. Young, J. T. Davis, C. L. Thompson, G. H. Rieke, G. Rivlis, R. Schnurr, J. Cadien and L.

Davidson 1998 Proc. SPIE 3354 57-65[2] P. Merken, T. Souverijns, J. Putzeys, Y. Creten and C. van Hoof 2006 Proc. SPIE 6275 627516[3] T. Hirao, Y. Hibi, M. Kawada, H. Nagata, H. Shibai, T. Watabe, M. Noda and T. Nakagawa

2002 Advances in Space Research 30 2117-2122[4] H. Nagata, T. Wada, H. Ikeda, Y. Arai and M. Ohno 2009 AIP Conf. Proc. 1185 286-289[5] M. Fujiwara and M. Sasaki 2004 IEEE Trans. Electron. Devices 51 2042-2047[6] Q. Dong, Y. X. Liang, U. Gennser, A. Cavanna and Y. Jin 2012 J. Low Temp. Phys. 167 626-

631[7] Y. X. Liang, Q. Liang, U. Gennser, A. Cavanna and Y. Jin 2012 J. Low Temp. Phys. 167 632-

637[8] S. Jahanmirinejad and A. Fiore 2012 Optics Express 20 5017-5028[9] L. Zhang, L. Kang, J. Chen, Y. Zhong, Q. Zhao, T. Jia, C. Cao, B. Jin, W. Xu, G. Sun and P.Wu

2011 Appl. Phys. B, Lasers Opt. 102 867-871[10] Q. Zhao, L. Zhang, T. Jia, L. Kang, W. Xu, J.Chen and P.Wu 2011 Appl. Phys. B, Lasers Opt.

104 673-678[11] S. Komiyama 2011 IEEE J. Selected Topics in Quantum Electronics 17 54-66

SiPM cryogenic operation down to 77 K

D. Prele1, D. Franco1, D. Ginhac2, K. Jradi2, F. Lebrun1, S. Perasso1,D. Pellion2, A. Tonazzo1, F. Voisin1,1APC, Univ. Paris Diderot, CNRS/IN2P3, CEA/Irfu, Obs. de Paris, Sorbonne Paris Cite,Paris, France2Le2i, Univ. de Bourgogne, CNRS/INSIS-INS2I, Dijon, France

E-mail: damien.prele@apc.univ-paris7.fr

Abstract. Silicon PhotoMultiplier (SiPM) is composed of extremely sensitive photosensorsbased on the Geiger Mode Avalanche PhotoDiode (GM-APD), which operate as a digital pixelsensitive to single photons. SiPMs are being considered for applications in low temperatureenvironments, such as noble-liquid detectors for dark matter searches or neutrino physics andGM-APD is promising technology for space Compton telescopes. While it is well known thatthe dark count rate, one of the main limitations of SiPM, is reduced at low temperature, adetailed study of the behavior of the device in cryogenic environment is necessary to assessits performances. In this paper, we present measurements of static parameters as breakdownvoltage and quenching resistance of a commercial SiPM (Hamamatsu MPPC S10362-11-100C).Evolution of these parameters as well as junction capacitance between room temperature and77 K is discussed.

1. IntroductionSilicon PhotoMultipliers (SiPMs) are being considered as possible photosensors for detectorsoperating at low temperatures, such as noble liquid (argon or xenon) time projection chambersfor direct dark matter searches and neutrino detectors, or for Compton telescopes in space. Adetailed study of their parameters in cryogenic environment are necessary to assess the expectedperformances.

A SiPM is composed of Geiger Mode Avalanche PhotoDiode (GM-APD). Geiger modeprovides internal gain as large as 106 carriers per incident photon. By means of a self-sustainingavalanche process: successive impact ionizations of carriers are accelerated as long as the electricfield is sufficient, above a voltage breakdown threshold (VBD). The GM-APD current rises tothe order of a milli-ampere in a few nano-secondes (rise time). This current passes througha quenching resistor [1] (RQ in Fig. 1), placed in series with the voltage biasing, which hasthe effect of lowering the electric field in the GM-APD below the breakdown threshold. Thiscauses the avalanche process to be turned out in few 10 ns (fall time). Without self-sustainingavalanche process and as long as there is no new incoming photon, the current drops below thepico-ampere level (leakages), and the GM-APD electric field is restored above the breakdownthreshold. After this current pulse, the GM-APD is again ready to detect a new photon.

The main parameters of a GM-APD, and thus of a SiPM, are: the quenching resistance RQ,influencing the fall time; the junction capacitance CJ , which limits the rise and fall times and thebreakdown voltage VBD. Rise and fall times are represented in the Figure 2. The differential

resistance of the junction Rdj during the avalanche process is also an interesting parameterallowing to determine the rise time.

HVbias

RQ . . . RQ

Vout

Figure 1. Electrical scheme of a SiPMshowing the individual GM-APD and thequenching resistors RQ. A small resistor(typically 50 Ω) in series with the SiPM andthe high voltage bias (≈ 70 V in the case ofthe MPPC) converts the current to voltage.

t

rise time∝ CJRdj

fall time∝ CJRQ

Figure 2. Typical current waveform crossinga GM-APD cell when photon is detected. AGM-APD pulse consists of two parts, rise andfall. The rise time depends to the value of thedifferential resistance Rdj of the diode duringthe avalanche process and of the junctioncapacitor CJ , while the fall time depends toRQ and CJ .

We determined the parameters VBD, RQ and CJ as a function of temperatures for acommercial SiPM, the Hamamatsu MPPC S10362-11-100C [2]. Most of parameters wereextracted from the Current-Voltage characteristic (I(V), see Fig. 3) from 300 K to 77 K. Only,junction capacitors is obtain from dynamic measurements. This allows us to know what kind ofcryogenic operation we can expect from the photosensor in cooled environements.

2. Quenching resistance extraction from the forward bias characteristicUsing the forward bias characteristic of the SiPM, it is possible to extract the value of thequenching resistor RQ. We assume that N cells under forward bias are similar to a single diodein series with a RQ/N resistor 1.

Following this method to extract RQ, we have measured the forward characteristic of theMPPC S10362-11-100C from 300 K to 77 K. Figure 4 shows the I(V) forward characteristics

1 SiPM ≡ Diode + RQ/N. Indeed, N diodes in parallel are equivalent to one diode with an area N times larger,where the small-signal equivalent resistance is always rd = kT

qIwhatever N.

Figure 3. Typical SiPM I(V) characteristic at 300 K with and without incident photons. Azoom around the breakdown voltage is also shown using a log scale of the current absolute value.

from room temperature to liquid nitrogen temperature. The slope of the curve is inverselyproportional to the quenching resistance, so we clearly see that quenching resistance increasesat low temperatures. Figure 5 shows the quenching resistance deduced from I(V) curves. It isnoticeable that the quenching resistance increases more quickly as the temperature decreasesand finally is 10 times larger at 77 K than at 300 K.

Figure 4. I(V) curves show thatthe quenching resistor increases at lowtemperature. Indeed, in an I(V) curve,the resistance is inversely proportional tothe slope of the curve.

Figure 5. Quenching resistance RQ as afunction of the temperature. RQ is extract fromthe figure 4 and by multiplying RQ/N by thenumber of SiPM cells (N=100).

The quenching resistance (per cell) goes from 100 kΩ at room temperature to 1 MΩ at liquidnitrogen temperature. As a consequence, for a constant junction capacitance, the fall time willincreases by an order of magnitude. This RQ(T) negative temperature coefficient could inhibitthe quenching process. Indeed, during the fall time, the Joule heating could reduce the RQ

resistance and would lead to stay in the avalanche process. This is probably the effect that willbe shown in the following section with the reverse bias characteristics (dashed line of Figure 6).

3. Breakdown voltage measurement from the reverse bias characteristicReverse bias characteristic is a common way to characterize SiPM and extract the breakdownvoltage VBD. As shown in figure 3, the measurement in the dark also allows us to know themaximum voltage biasing (VZ

2) before SiPM self-triggering.As it is shown in figure 6, at low temperature the photo-current is too small to be measured

by our equipment (Agilent B1500A source measure unit with a noise floor ≈ 100 pA). Moreover,below 250 K, a ”full SiPM latching process” trigs the overall cells unexpectedly above thebreakdown voltage even without photon flux (”Dark”). The arbitrary voltage where this latchingoccurs suggests that it is a different phenomenon than the ”Zener” threshold VZ . Moreover, thisphenomenon could be compatible with a positive electro-thermal feedback due to the negativetemperature coefficient of the RQ.

Nevertheless, using a light source to trigger the SiPM, it is possible to clearly see thebreakdown voltage, as it shown in figure 6 - solid line. The breakdown voltage is measuredand plotted as a function of the temperature in Figure 7. It appears that its evolution is closeto a linear function ; with a temperature coefficient equal to -50 mV/K. VBD decreases by morethan 10 V (in absolute value) from room temperature to liquid nitrogen temperature.

2 The Zener voltage VZ corresponds to the breakdown voltage in the dark.

Figure 6. Reversed biased I(V) curves of theSiPM. To see small current phenomena a logscale is used, so in absolute value.

Figure 7. Breakdown voltage as a functionof temperature. A linear fit gives a fluctuationof the breakdown voltage of −50 mV/K.

4. Dynamic characteristicsSiPMs are used to detect fast pulses, therefore a dynamic characterization is mandatory todetermine the SiPM behavior. A small signal analysis allows one to extract the capacitance ofthe junction reverse biased.

Figure 8 shows the equivalent AC electrical model of the SiPM. Rdj is the differentialresistance of each junction and is only considered in presence of an avalanche process ie biasedbelow VBD (in absolute value). CJ is the junction capacitance, which is due to the depleted regionof the APD. The thickness of this depleted layer is proportional to the reverse biasing, thereforethe junction capacitor decreases as the reverse bias increases in absolute value. However, as isshown in figure 9, this value remains constant for voltage biases larger than 30 V. Moreover,above VBD, the avalanche process prohibits the measurement of the junction capacitance dueto the drop of the junction differential resistance.

HVbias+ac

N in //

RQ/N

I+i

≡

ac

Rdj/N

RQ/N

i

N×CJ

Figure 8. Small signal electrical schematicof a SiPM for ac measurement. Rdjcorresponds to the differential resistance ofthe GM-APD junction (without RQ). CJ

is the junction capacitor.

Figure 9. Evolution of the junction capacitanceas a function of the voltage biasing and fordifferent temperatures from room temperatureto liquid nitrogen temperature.

The junction capacitance as a function of temperature, measured with a bias voltage of 60 V,is finally reported in figure 10. We clearly notice a decrease of the junction capacitance at lowtemperature. However, this decrease is less than a factor of 2 between 300 K and 77 K.

Figure 10. Junction capacitance as a function of the temperature. The measurement has beenmade with a 60 V biasing.

5. ConclusionThe breakdown voltage VBD, the quenching resistance RQ and the junction capacitance CJ

of a commercial SiPM MPPC S10362-11-100C [2] have been measured at temperatures from300 K to 77 K. A quasi-linear decreasing of the breakdown voltage with temperature is found,with a rate of -50 mV/K. The quenching resistance increases by a factor of 10 between 300 Kto 77 K. Finally, the junction capacitance decreases by less than a factor of 2 in the sametemperature range. These experimental results give a good idea of the voltage biasing requiredfor MPPC cryogenic operation. Furthermore, they allow us to predict a fall time increaseby a factor of 5 at liquid nitrogen temperature. Moreover, these tests clearly point out aphenomenon of ”full triggering” of the tested SiPM at low temperature and for biasing justabove the breakdown voltage. Therefore, the biasing voltage range seems to be strongly reducedat cryogenic temperatures.

Most of these results has been obtained by using ”simple” I(V) measurements and are fullycompatible with similar measurements using time response [3].

AcknowledgmentsMeasurements presented in this paper has been made with the help of B. Manceur during itsmaster degree internship at APC. Authors are also grateful to M. Piat, who provided part ofthe cryogenic equipment needed for this study.

References

[1] Cova S. et al 1996 Avalanche photodiodes and quenching circuits for single-photon detection (Applied Optics,Vol. 35 No 12)

[2] HAMAMATSU datasheet, MPPC (multi-pixel photon counter) - S10362-11 series[3] Otono H. et al 2007 Study of MPPC at Liquid Nitrogen Temperature (International Workshop on neaw

Photon-Detectors, Kobe, Japan)

ZERO MAINTENANCE X-BAND CRYOGENIC LOW NOISE AMPLIFIER

T. Bonhoure1, B. Fauroux1, R. Rayet1 and S. Rawson1

1Callisto, 12 avenue de Borde Blanche, 31290 Villefranche de Lauragais, France. Email : thomas.bonhoure@callisto-space.com

I. INTRODUCTION/ABSTRACT

Callisto has over 18 years experience producing cryogenic Low Noise Amplifiers for satellite ground stations.The goal of cooling the LNA is to improve the sensitivity of the downlink chain by improving the G/T by atleast 2dB compared to a COTS ambient temperature LNA. By using such a cryogenic LNA the antenna buildcost can be reduced by 20% or more and the data transfer rates can be improve up to 60%.The drawback of using standard cryogenic cooling techniques is the maintenance of cryo-cooling system whichin the past had to be done between 10 000 and 25 000hrs of operation depending on the cryo-systems used.Moreover, standard cryostats need a vacuum isolation to be able to cool the amplifier down to cryogenictemperature and this requires the use of a vacuum pump for each cooldown session. During these activities thesystem is not operational and this can cause down time of several hours. Callisto has developed a prototype X-band compact LNA using a maintenance free cryocooler. This compactLNA has two X Band channels achieving a Noise Temperature of less than 20K. This level of NoiseTemperature is achieved with a compact and simple system with a MTBM better than 90 000hrs (10 years). In2013 Callisto started the production of these X-Band Cyro Compact LNAs and the first production units aredestined for the CNES Cormoran project. It is also planned to produce K-band and Ka-Band units based on thesame design.

II. CRYOGENIC LOW NOISE AMPLIFIER

The Low Noise Amplifier (LNA) is one of the most critical elements of the reception chain of an antenna forspace telecommunications (Fig. 1). High performance LNAs increase the sensitivity of the ground station and/orcan reduce the antenna size and then build cost. To achieve the highest performance for the LNA it is necessaryto cool the LNA to very low temperature. The most performing LNA are the cryogenically cooled LNA, thistype of LNA is used for satellite missions where link margins are very critical. In 2006, Callisto has developed afirst generation of Compact X-Band cryo LNA under an ESA contract, 3 units have been delivered to ESA(Kourou 15m) and 2 units to CNES (Aussaguel 9m). Theses cryo LNAs allow the increase of the G/T ofexisting antenna but required preventive maintenance [1].

Ground Station

Signal

LNA Data

LNA Purpose

Fig. 1. Basic schematic of the LNA purpose

A. Advantages of cooling a LNA The factor of merit of the reception chain of ground station (G/T) is linked to the gain of the antenna which isdriven by the size of the dish (diameter) and the antenna noise which is driven by the LNA performance. Inorder to improve the G/T of a ground station, the antenna size can be increased but this solution is costly.Another approach is to improve the LNA performance; the most efficient way is to cryo-cool the LNA. Thebenefits of using a cryogenic LNA is that the Noise Temperature (NT) of the LNA can be typically halvedcompared to a system with a conventional un-cooled LNA. This enables the use of a smaller diameter antenna toachieve the required G/T performance. The Cryo-LNA benefits can be:

· Reduced antenna builds costs by 20% ormore.

· Reduced windage issues and hence reducedtracking equipment costs.

· Improved antenna pointing operation(smaller antenna = wider beam).

· Improved data transfer rates, of up to60%.

· Reduced satellite emission power.Additional margin allows a decrease of thesatellite transmit power, extending missionlife.

Example for an X-Band antenna with G/T of 40dB/K the diameter of the dish will be 10m with a conventionalLNA (0.8dB Noise Figure) and will be reduced to 8m with a cryogenic LNA (0.3dB NF).

B. Drawbacks of standard cooling techniques The typical cryogenic concepts are considered as not reliable, complicate to operate and require regularmaintenance. The high performance X-Band cryo LNA developed by Callisto for ESA Deep Space Antenna useGM cryo-coolers. The bearings and seals of this type of cryo cooler need to be changed regularly, so the coolerneeds to be serviced every 1 to 3 years. The maintenance for this cooler at the end of this service life requiresthe cooler to be replaced and this cannot be performed on site. This maintenance process needs around twomonths (including shipping time to and from factory) and is quite expensive. In terms of operation, a vacuum pump has to be connected during the cooldown in order to reach cryogenictemperature. The purchase of such equipment is expensive. Furthermore, the cryo LNA must be installed closeto the antenna feed and the connection of the vacuum pump to start the system can be difficult due to the lack ofroom.

C. Zero Maintenance Concept In order to eliminate the maintenance issues, the new generation (MKII) of X-Band Compact LNA developedby Callisto integrates world leading cryogenic technology, and it uses the advantages of a light-weight, highlyefficient and long-life cryogenic cooler. Working at 80K, a Stirling-cycle cryo cooler is used, that integrates thecold finger and compressor into one compact unit that can work in any orientation. The cooler has beendeveloped to serve various applications including mobile telephone base station filter cooling, delivered in veryhigh volumes and for which extremely long MTBF performance has been demonstrated. Contactless bearingsand seals are designed for ultra-long operating life. The Compact LNA MKII also includes a new type of thermal insulation based on a mix of Aerogel and drygas. This insulation concept has also been used in recent developments for high performance cryo LNAs forDeep Space Antenna under ESA contract and is patented by Callisto [3]. The use of Aerogel/dry gas insulationremoves the need for a vacuum pump in operation and extends the life time of the system up to 10 years.

III. DESIGN OVERVIEW

The X-Band Compact MKII has been developed with the following main goals:

· MTBF > 10 years· Dual channel for dual polarization use· Noise Temperature < 20K

The cryo LNA is shown in Fig. 2, it consists of a waveguide window with two apertures for bothpolarizations, a sealed enclosure which includes the cryo MIMIC LNA, a post amplifier box and the cryocooler. The overall system is 475x172x165mm, weight 8.5Kg and can operate in any orientation, on a movableantenna structure. The cryo Compact LNA is monitored and controlled by a specific power supply unit. Thisunit is housed in an industrial standard 19” rack mounting drawer.

Sealed Enclosure

CryoLNA1

PostAmp1

DC/DC Bias Board

Post Box

Cryo Cooler

Power Supply Unit

PostAmp2Cryo

LNA2

Main Power230V/50Hz, 110V/60Hz

WG to coax transition

WG to coax transition

WR112 UBR84 RF Input2

WR112 UBR84 RF Input1

N-Type (F)RF Output2

N-Type (F)RF Output1

Cooler Power

M&C

Remote Interface (TCP/IP)

Fig. 2. X-Band Compact MKII overview & block diagram

The cryo cooler of the Compact MKII needs an air flow around the cooling fins in order to regulate its skintemperature as shown in Fig. 2.

IV. THERMAL DESIGN

A. Cryocooler long term reliability To reach the targeted MTBF of 10 years, “Sapphire” Stirling cryo-cooler manufactured by STI has beenselected; this cooler is a unique long-life Stirling design [2]. Specifically developed for mass-production andhigh reliability, STI advertises a MTBF of >1million hours based on the track record of installed units. But toensure this MTBF the cooler has to operate under specific conditions:

· Low input power· Low heat rejector temperature

The Fig. 3 presents the cryocooler heat lift at 77K versus the input power for various heat reject temperatures:

Fig. 3. STI Sapphire cryocooler heat lift at 77K

As the heat reject temperature increases, the input power increases and the long term reliability is degraded.The heat load is driven by the configuration inside the sealed housing. An important design objective was tominimize the heat load hence keep the heat reject temperature in the lower range of the specified envelop.

B. Detailed thermal design The thermal design is based on a complex trade-off between the physical temperature of the MMIC LNA, theNoise Temperature and the long term reliability. To decrease the NT it is necessary to decrease the physicaltemperature of the MMIC LNA, these two parameters are directly linked. So to decrease the NT, more powerhas to be injected to the cooler. The increase of the power impacts the reject temperature and hence the longterm reliability is degraded. Another approach to lower the physical temperature of the MMIC withoutdegrading the lifetime of the cooler is to keep a nominal power input and minimize the heat load. The thermaldesign is based on this approach.

The main contributors to the heat load are the solid state insulation (Aerogel+dry gas) and the waveguide tocoaxial transition (Fig. 2). The efforts to reduce the heat load have been focused on the configuration and thematerial choice of the waveguide to coaxial transition.

V. RF DESIGN

The RF design is mainly driven by the thermal constraints and the targeted NT of 20K. In order to achieve theNT, the waveguide to coaxial transition is included in the vacuum enclosure and directly connected to the cryoMMIC LNA. A post amplifier is added after the cryo MMIC LNA in a post box to increase the overall gain ofthe system. The two RF chains are independent and identical. The components used are the same for both chains (Fig. 2).The RF circuit inside the enclosure comprises the following elements: a WR112 waveguide vacuum window, acustom long waveguide to coaxial transition, a state of the art MMIC Low Noise Amplifier developed byChalmers University for ESA and an output coaxial cable. The RF circuit inside the post box comprises a postamplifier and an output isolator to match the gain and output return loss specification.

VI. PERFORMANCE AND TEST RESULTS

Fig 4. View of the X-Band Compact MKII

A. Thermal performance The Fig. 5 shows a cooldown of the X-Band Compact MKII. At the beginning the temperature required hasbeen set to 90K, the system reaches the set point in 70 min; the input power needed to maintain this basetemperature is 73W. Then the temperature required has been set to 95K and the input power stabilizes at 63W.63W of input power is half of the maximum power capability for the cryo cooler. This value is reasonable toensure the 10 years lifetime of the cryo cooler.

0

10

20

30

40

50

60

70

80

90

100

110

120

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160

Pow

er(W

)

Tcry

o(K

)

Time (min)

Cooldown X-Band Compact MKII Cryo Temp Cooler Input Power

Fig. 5. Cooldown on the X-Band Compact MKII

B. RF performance The RF parameters has been measured at 95K of base temperature, the results are shown in Fig. 6 and Table 1.

50

55

60

65

70

5

10

15

20

25

7.85 7.9 7.95 8 8.05 8.1 8.15 8.2 8.25 8.3 8.35 8.4 8.45 8.5

Gain

(dB)

NT

(K)

Freq (GHz)

Gain & NT Measurement X-Band Compact MKII NT Gain

-35

-30

-25

-20

-15

-10

-5

0

7.85 7.9 7.95 8 8.05 8.1 8.15 8.2 8.25 8.3 8.35 8.4 8.45 8.5

Retu

rnLo

ss(d

B)

Freq (GHz)

Return Loss Measurement X-Band Compact MKII S11 S22

Fig. 6: Gain, NT and Return Loss measurement.

Table 1. Measurements SummaryParameter Specification Measurement

Frequency Range 8.0-8.5 GHz 8.0-8.5 GHz

Input type Waveguide WR112 Waveguide WR112

Output type N-Type female N-Type female

Dimension 500x172x165mm 475x172x165mm

Weight < 10Kg 8.5Kg

Noise Temperature < 20K 19.4Kmax

Gain @cryo 54-59dB 59.8-60.3dB

Gain flatness @cryo ±2 dBpp across band 0.5dBpp

Pout 1dB @cryo > +10dBm +14.4dBm

Input RT Loss @cryo > 16dB 15.3dBmin

Output RT Loss @cryo > 20dB 24.9dBmin

VII. CONCLUSION

One prototype X-Band Cryo Compact LNA MKII has been manufactured and tested. One production unit hasbeen delivered to Zodiac Data System for the CNES CORMORAN project, others units will be manufacturedfor this project but also to other ZDS customers, and ESA is planning to order one unit to upgrade its KRU 15mantenna. This LNA is the first zero maintenance cryogenic LNA. The tests results show a Noise Temperaturebelow 20K and a gain around 59dB. This product can be installed in medium size antenna due to its simplicityand its compactness compare to other cryogenic product. The concept described in this paper can be adapted toother frequency band and Callisto plan to use this design for a Ka-Band cryo Compact LNA (25.5-27GHz).

VIII. REFERENCES

[1] B. Fauroux, “X-Band Compact Ultra-Low Noise Amplifier for the CNES STEREO Mission Support,”Callisto Application Note, http://www.callisto-space.com/CallistoCryo-LNA-CNE_Aussaguel_Application_Note.pdf , 2012.

[2] R. Rayet, “MTBF & Reliability of the Stirling cryo-cooler for The Comapct Ultra-Low Noise Amplifier”Callisto Technology Note, 2012.

[3] S. Halté, “Second Generation 15K X-Band Cryogenic Low Noise Amplifier for the DSA Network” , TTC2010.

Design and evaluation of low-frequency RFID transponder for

cryogenic applications

F R Ihmig1*, S G Shirley1, and H Zimmermann1,2

1Fraunhofer–Institut fuer Biomedizinische Technik (IBMT), Main Department for

Biophysics and Cryotechnology, Ensheimer Straße 48, 66386 Sankt Ingbert, Germany 2Chair of Molecular and Cellular Biotechnology/Nanotechnology, Saarland

University, Saarbruecken, Germany

*E-mail: frank.ihmig@ibmt.fraunhofer.de

Abstract. Electronic identification of cryobiological specimens is an emerging

technology in the field of advanced biological specimen banking. In this paper, we

focus on the design and evaluation of a radio-frequency identification transponder

(RFID) using a low-frequency carrier signal for wireless specimen identification

between room temperature and liquid-nitrogen temperatures. The transponder design

is based on a commercial-off-the-shelf CMOS integrated circuit. The chip contains

1 kbit of EEPROM for user-defined data storage. We describe the design technique

and results for the component and transponder characteristics at room temperature and

77 K. We have assembled a batch of prototype devices and tested them down to

liquid-nitrogen temperature to verify the transponder design and present test results for

read and write operations. We suggest further development steps including

miniaturization, packaging and extended functionality.

1. Introduction

Electronic circuits and modules for cryogenic applications, such as tagging cryopreserved biological

material, must operate between room temperature (RT) and liquid-nitrogen temperature (LNT).

Cryopreservation involves the storage of cryoprotected living cells at ultra-low temperatures down to

77 K (–196 °C). As an example, the preservation of living stem cells allows basic research into

cellular reproduction and differentiation with significant potential for future biological and clinical

applications. The reception, storage, processing and distribution of biological specimens take place in

a biorepository or biobank.

Over ten years Fraunhofer IBMT has developed advanced technology for cryogenic biobanking

based on cryoengineering and cryoelectronics [1-4]. The aim is to overcome typical shortcomings of

conventional cryostorages, namely specimen mis-identification, ice formation, instability of the cold

chain, and difficulties in specimen exchange. Cryoelectronics can intimately couple data with the

specimen to which it relates and access this data through interfaces in the cold storage environment.

Since we first introduced it, the electronic identification of cryobiological specimens has been

recognized as emerging technology in the field of advanced biological specimen banking [5, 6].

Consequently, there is a great demand for its integration into various specimen containers such as

bags, micro-well plates, vials and straws.

Here, we focus on the design and evaluation of passive radio-frequency identification (RFID)

transponders using a low-frequency carrier signal for wireless specimen identification down to LNT.

RFID systems are automatic identification systems used in a wide range of applications. Systems can

be classified as active or passive (no battery power), and as programmable or read-only. Passive

transponders receive power through inductive coupling with the reader and this prompts them to

communicate. The distance at which a passive transponder can receive sufficient power defines its

range. Passive transponders are the most widely used type of RFID transponders. Low-frequency

(50 kHz - 150 kHz) systems are generally low cost, work at close range, do not require licensing and

are tolerant of metal, liquids and electrical noise.

An RFID system consists of two main elements, the RFID transponder, where data is encoded, and

the RFID reader, which is used for extracting the encoded data from the transponder. Figure 1 is a

block diagram of a typical RFID system. The reader transmits the RF signal and, when an RFID

transponder tagged item enters the interrogation zone, it modulates the signal according to the data

stored in the transponder. The reader decodes the ID of the tagged item from the modulated signal and

updates it in a shared database for further processing.

The parallel assembly of the electrical coil and the capacitor Ct forms a transponder’s resonant

circuit, which should be tuned to the carrier frequency of the reader. The induced voltage in the

transponder coil is maximized by resonance, rectified and supplies the chip with power. Very low-

power chips are required because the coupling between the two coils is usually very weak.

Figure 1. Schematic diagram of a typical RFID system.

2. Transponder design and evaluation method

Cryopreservation usually involves small specimen volumes and low volume containers (2 ml and

below) such as cryovials (typ. o.d. 12.5 mm) and straws (typ. o.d. 4 mm). Antenna design is crucial

with small form factors. We require wide-range temperature operation and a read range of a few

centimeters with small transponder geometry.

2.1. Design Considerations

For prototyping, we select and evaluate components for use at cryogenic temperatures. The

transponder is based on a commercial-off-the-shelf CMOS IC. For confidentiality, the manufacturer of

the selected chip is not given. The operating temperature range is specified by the manufacturer as –40

to +85 °C and the operating frequency range is 100 to 150 kHz. The chip contains 1 kbit of EEPROM

for user data storage including a unique serial number and device identification. The memory can be

protected with a 32-bit password for write and read. The chip has a voltage rectifier and limiter and a

resonance capacitor (170 pF ±2%) in a CID package (6 x 4 mm). The coil is the only external

component required. Since small antenna geometry implies low inductance, the chip requires an

additional tuning capacitor.

The read range of the transponder is set by the power consumption of the chip. Good reading

distances require proper tuning of the transponder’s resonant circuit. Parameter variation with

temperature in the components must be considered because of the wide-temperature range. The

resistance of a copper coil decreases with decreasing temperature so a quality factor rise is expected.

Generally, a high quality factor, Q, is desirable to ensure signal gain. However a high Q gives a

narrow bandwidth and requires close tolerance, low temperature-variation components.

Reader Transponder

Chip

Magnetic field

CtCr

Ri

Reader Transponder

Chip

Magnetic field

CtCr

Ri

A complicating factor in the use of transponders in cryogenic storage systems is that the

transponders are required to operate in two different modes: (1) during preparation and processing in

the laboratory, specimens can be well separated and a conventional, fairly long range, reader is

appropriate; (2) in the cryotank, specimens are densely packed (tens or hundreds of thousands of

specimens per tank) and, to avoid collision problems, a large number of reader coils (or screening

coils) are needed in a highly multiplexed system. The necessary readers are therefore of very short

range. Locating a specimen precisely within the tank is an important function also requiring very short

range readers. The transponder itself must be capable of functioning in both environments.

We designed and manufactured two small coil types: an air-cored coil from copper wire of 0.1 mm

diameter and a ferrite-cored coil from 0.056 mm copper wire. Table 1 gives the dimensions and

parameters. Core materials can aid antenna miniaturization. Ferromagnetic materials such as ferrites

are commonly used to increase inductance and/or reduce coil cross section. But losses in the core

materials degrade performance and can cause detuning. This degradation becomes worse at lower

temperatures [7]. The selected core is a medium permeability MnZn ferrite (3B1 by Ferroxcube).

Table 1. Coil dimensions and parameters.

Core Out. dia.

[mm]

Inn. dia.

[mm]

Height

[mm]

No. of turns Permeability Inductance

(meas.) [mH]

Air 6.18 4.00 6.25 500 1 0.62

Ferrite 2.3 0.74 7.5 600 900 1.57

The external resonant capacitors are C0G. Our preliminary work (unpublished data) showed that

C0G capacitors behave well at low temperatures. Generally, capacitors with a simple dielectric show

less temperature drift than high dielectric constant types [8]. The SIL-1500 long-range RFID reader

and SAT-A4-LR-P-125kHz antenna of 180 mm diameter (both Scemtec, Germany) were used. This

system operated at room temperature with a carrier frequency of 125 kHz.

2.2. Experimental procedure

For inductance and capacitance measurement, we used a high-end digital LCR meter (GW Instek

LCR-821) set to slow measurement speed for best accuracy. Measurements of each coil and capacitor

at RT and 77 K were obtained in the frequency range from 20 kHz to 200 kHz. For transponder

characterization, we used the RFID reader, a waveform generator (TTi TGA1230) and a digital

phosphor oscilloscope (Tektronix TDS5052). We measured the induced voltage variation as a function

of frequency and as function of distance at RT and LNT for both transponder types. 16 prototype

transponders with air-cored coils were evaluated for variations in read range and for bit errors.

3. Results and discussion

Two transponder types were assembled: one with air-cored coils and a nominal parallel capacitance of

2,350 pF; the other with ferrite-cored coils and a nominal parallel capacitance of 930 pF. Leaded

solder alloy (Sn60/Pb39/Cu1) was used for prototyping. Due to the conductivity change of copper the

DC resistance of the coils decreased by a factor of about 6 between RT and 77 K. This will affect the

Q of the circuit. However, should this create a problem, the Q is sufficiently high that the coil could be

‘padded’ with a temperature insensitive resistor to reduce the variation in quality factor.

Table 2 shows the parameters of the components. C0G capacitor details are those used with the air-

cored coil. Values measured at 100 kHz at RT and 77 K are given in the table. This is the closest

obtainable frequency of the LCR meter to the resonance frequency of the transponder. All components

show little variation at 77 K except the ferrite-cored coil. The inductance of the air-cored coil was

virtually the same at RT and 77 K. This is to be expected as the effects that could change the

inductance are very small (contraction of the coil ~0.3% and the permeability difference between air

and nitrogen ~0.004%). The coil Q improves at low temperatures. The inductance of the ferrite-cored

coil changes significantly with temperature (about 12% decrease), mainly due to the drop in magnetic

permeability of the ferrite core. The effect is large enough to have a significant impact on the

resonance frequency of a circuit which makes it more difficult to trade off the read range at both

temperatures. The C0G capacitor value change is below 0.5% and the chip capacitor value decrease is

below 2%, still within its tolerance.

Table 2. Component types and exemplary parameters at RT and 77 K.

Component Manufacturer Type Nom. value RT value

@ 100 kHz

77 K value

@ 100 kHz

Variation

Inductance IBMT Air core 0.62 mH 0.6198 mH 0.6156 mH -0.0042

Inductance IBMT Ferrite core 1.57 mH 1.5721 mH 1.3770 mH -0.1951

Chip Cap. - - 170 pF 172.27 pF 169.49 pF -2.78

Capacitor Kemet C0G 1206 2.2 nF 2.2737 nF 2.2832 nF +0.0095

Capacitor Kemet C0G 0603 150 pF 148.65 pF 148.97 pF +0.32

To separate the various factors affecting read range we measured 1) the induced voltage in the coil

with no chip connected; 2) the induced voltage with a chip connected and 3) the induced voltage in a

cold coil connected to a warm chip. Figure 2a only shows the 77 K curves for clarity. For the non-

loaded coil there is virtually no difference in the voltage-frequency curves between RT and 77 K.

Also, a warm chip connected to the cold coil causes little change to the RT result. The biggest

variation observed with a loaded coil depends on whether the chip is cold or warm.

The measured voltage-distance characteristic of the transponder with air-cored coil gives the shape

of the theoretical variation of magnetic field with distance (Figure 2b). Read operation was possible up

to a distance of about 7.5 cm at RT and up to about 6.5 cm at 77 K. For the same distance, voltage

values at 77 K are slightly less than those at RT. Also, the chip exhibits an increase in threshold

voltages at 77 K. This increase, coupled with the steep curve, explains the reduced reading distance in

the cold.

The operation of 16 prototype transponders with air-cored coil gave a mean read range of about

7.5 cm (±1.1) at RT and about 6.0 cm (±1.2) at 77 K. The continuous write and read operation with

100 cycles within the complete user memory area at 77 K showed no bit errors in the random test

pattern. Read operation of the prototype transponder with ferrite-cored coil was possible up to a

distance of about 3.1 cm at RT and up to about 3.4 cm at 77 K. This seems to be a good trade off in the

transponder’s resonant circuit.

Figure 2. (a) Voltage-frequency characteristic of the air-cored transponder with and without chip. (b)

Voltage-distance characteristic at RT and 77 K. The chip detection by the reader is marked. Inset

shows photograph of prototype transponder assembly with air-cored coil.

4 5 6 7 8 9 10 11 12 13 14 15 16

0

1

2

3

4

5

6

7

8

9

10

11

RT, air core

77 K, air core

Ind

uce

d v

oltag

e, V

pp [V

]

Distance [cm]100 120 140 160 180 200

0

10

20

30

40

50

60

70

77 K, w/o chip

77 K, with chip

77 K, chip RT

Ind

uce

d v

oltag

e, V

pp [m

V]

Frequency [kHz]

(a) (b)

Chip

detection

Variation

with chip

cold and

warm

4 5 6 7 8 9 10 11 12 13 14 15 16

0

1

2

3

4

5

6

7

8

9

10

11

RT, air core

77 K, air core

Ind

uce

d v

oltag

e, V

pp [V

]

Distance [cm]100 120 140 160 180 200

0

10

20

30

40

50

60

70

77 K, w/o chip

77 K, with chip

77 K, chip RT

Ind

uce

d v

oltag

e, V

pp [m

V]

Frequency [kHz]

(a) (b)

Chip

detection

Variation

with chip

cold and

warm

As simulation tool we used a finite element solver for 2D and axisymmetric magnetic problems

(FEMM 4.2, freeware, http://www.femm.info). This showed the ferrite-cored coil had higher induced

voltage at very short range but a faster fall-off with distance than the air-cored coil. The simulated

distances at which sufficient power would be transferred were 3.5 cm (ferrite-cored coil) and 7.5 cm

(air-cored coil) respectively. These are in good agreement with the observed read distances at RT.

4. Conclusion and outlook

We have shown that the design of miniaturized RFID transponders for cryogenic applications is

feasible. Cryo-compatible building blocks for an RFID system are available. The RFID chip itself will

function at LNT and power can be transferred via transformer coupling. Important design factors are

component selection and resonant circuit tuning. The evaluated batch of transponder prototypes shows

proper operation and acceptable read range for the intended application. The selected chip has the full

range of state-of-the-art RFID functionality including password protection and encryption algorithms

that address privacy concerns.

A practical design must consider the whole RFID system and its environment. We need the

smallest transponder coil that provides acceptable read range. This problem may be eased because the

read range for LNT can be very small and, for RT, minimizing the reader power is not a primary

concern. An air-cored coil gives an inductance that is stable with temperature. The inductance might

be increased using core materials with stable permeability at low temperature [9], which would allow

the use of smaller capacitors. The change in coil resistance, and hence Q, with temperature can be

masked with a small series resistor because a very high Q is not necessary.

Further development includes miniaturization, packaging and extended functionality. Low-

temperature coefficient magnetic materials such as magnetic powder cores [10] may help, as well as

coil-on-chip technology. Packaging possibilities include various form factors using, e.g. borosilicate

glass, ceramic or epoxy molding, and evaluation must consider thermo-mechanical behavior [11]. It

should be possible to extend functionality using low-power sensors or other memory technologies

such as FRAM. The next major step is to develop a cryo-compatible short range, multi position reader.

Acknowledgments

This work is partially supported by the Ministry of Economic Affairs of Saarland (Germany). We

thank Vitalij Mast for the simulation work and Stephane Djoumbou for excellent experimental work

and transponder assembly. We appreciate the proof-reading by Randall K. Kirschman.

References

[1] Ihmig F R, Shirley S G and Zimmermann H 2003 Proc. 2nd VDE World Microtechnologies

Congress (Munich) pp 643-648

[2] Ihmig F R, Shirley S G and Zimmermann H 2004 Proc. 6th European Workshop on Low

Temperature Electronics (Noordwijk) WPP-227 (ESA Publication) pp 153-160

[3] Shirley S G, Durst C H P, Fuchs C C, Zimmermann H and Ihmig F R 2009 Cryogenics 49 638-

642

[4] Ihmig F R, Shirley S G, Kirschman R K and Zimmermann H 2013 IEEE Pulse 4 35-43

[5] Bettendorf E, Malenfant C and Chabannon C 2005 Cell Preserv. Technol. 3 112-114

[6] Zarabzadeh A, Hayati F, Watson R W G, Bradley G and Grimson J 2011 Proc. IEEE Int. Conf.

RFID-Technology and Applications (Sitges) pp 228-235

[7] Mukhopadhyay P K, Barat P, Kar S K, Bandyopadhyay S K and Sen P 1994 Cryogenics 34

241-243

[8] Pan M 2005 Cryogenics 45 463-467

[9] Geithner R, Heinert D, Neubert R, Vodel W and Seidel P 2013 Cryogenics 54 16-19

[10] Gerber S S, Elbuluk M E, Hammoud A and Patterson R L 2002 Proc. 37th Intersociety Energy

Conversion Engineering Conference (Washington) pp 249-254

[11] Cauchois R., Yin M S, Gouantes A and Boddaert X 2013 Microelectr. Rel. 53 885-891

Experimental set-up to test low temperature electronics for X-ray micro-calorimeters with high impedance sensors

M Barbera1,2, A Collura2, X de La Broise3, G Lo Cicero1, U Lo Cicero2, C Pigot3, S Varisco2 and J L Sauvageot3

1 Università degli Studi di Palermo, Dipartimento di Fisica e Chimica, Palermo, IT 2 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Palermo, Palermo, IT 3 CEA-Saclay, DSM-IRFU-SEDI, Gif-sur-Yvette CEDEX, FR

Abstract. We have installed our single stage ADR cryostat inside an EMI shielded room built in-house, and have set-up it to test the performance of low temperature read-out electronics of X-ray micro-calorimeters with high impedance sensors. The present configuration allows to test four independent channels, although an extension to a larger number of channels with the use of a multiplexing approach is foreseen. We describe the adopted solutions to wire the cryostat minimizing EMI, and present the configuration used to test micro-calorimeters with front-end electronics based on Si J-FETs or HEMT, and with multiplexing.

1. Introduction We have built an Adiabatic Demagnetization Refrigerator (ADR)[1] at the X-Ray Astronomy Calibration and Testing (XACT) facility[2] of INAF-OAPA to test cryogenic X-ray detectors for laboratory and astrophysical applications[3][4][5][6][7]. The ADR cools detectors down to 40 mK and can keep them at 60 mK for more than 20 hours with an heat-load < 1 μW. A gold plated copper box, thermally coupled to the cold finger, hosts four micro-calorimeters and associated feedback resistors of the front end read-out circuit. A second box, thermally coupled to the liquid He stage, hosts the J-FET transistors operating in the temperature range 100÷140 K. Figure 1 shows a schematic of the read-out circuit, while Figure 2 shows the two boxes mounted inside the ADR. A few neutron transmutation doping (NTD) germanium micro-calorimeters have been successfully tested allowing us to study the thermalization properties of absorber materials[8]. The electromagnetic interferences (EMI) have prevented us, up to now, to measure the intrinsic energy resolution of the tested detectors. Within the activities of the research project CESAR - Cryogenic Electronics for Space Applications and Research (FP7-SPACE-2010-1)[9], we will test X-ray micro-calorimeters with semiconductor thermistors operated with a new low temperature read-out electronics developed at CEA-Saclay. In the following sections we describe the modifications that have been designed and implemented in our cryogenic laboratory to conduct such tests.

2. EMI shielded room Different sources of noise have been affecting our measurements in the past. Low frequency noises include: mains frequency and harmonics, high speed motors (e.g. vacuum pumps), digital electronic equipment, and mechanical vibrations. High frequency noises include: broadcast radio, cellular phones, wireless systems, and digital electronic equipment. 1 Send correspondence to Marco Barbera; e-mail: marco.barbera@unipa.it

Figure 1. Schematic of the wired circuit inside the ADR to read-out four NTD Germanium X-ray micro-calorimeters.

Figure 2. J-FET box (top), and Detector box (bottom).

Such noise sources affect both the temperature stability of the X-ray micro-calorimeters (our goal is ΔT < 2 µK while we usually get 10 µK at 60 mK), and the signal along the photon pulse detection chain (our goal is to keep the electronic noise contribution to the energy resolution below 1 eV while we usually measure ΔE ≈ 50 eV). To improve the working conditions we have built a 20 m2 EMI shielded room, we have rewired the system with particular care to avoid ground loops, and have settled the ADR on active pneumatic isolation posts. Figure 3 shows the layout of the new laboratory.

Figure 3. Layout of the cryogenic laboratory. Only the thermometer resistance bridges and micro-calorimeter pre-amplifier electronics are located inside the shielded room. All the other electronic equipment is located in the control room, e.g. magnet power supply, resistance bridge interfaces, J-FET box heater power supply, A/D boards and data acquisition and control computers.

The walls and roof of the shielded room have been realized with galvanized steel panels while the floor has been realized with a copper foil on bitumen covered with a stiff walking layer. All joints between walls, roof, and floor are made with copper adhesive tape protected by galvanized steel profiles. The glass windows of the door have been covered with an aluminum mesh made of a 0.2 mm Ø wires with 1.5 mm x 1.9 mm rectangular pitch. The electrical continuity between the door and walls and floor is guaranteed by the use of spring-line beryllium-copper gaskets. The holes opened in the walls for gas pipes, and for ventilation windows are shielded with two layers of aluminum mesh.

3. Grounding scheme

In order to minimize noise, a proper grounding scheme and rewiring of all the lines connecting the ADR to the measurement and control instrumentation have been implemented. Figure 4 shows a schematic drawing of the adopted configuration.

Figure 4. Schematic drawing of the ADR wiring and grounding scheme.

The ground to the ADR is provided only by one line which also provides ground to the resistance bridges and rack. The mains earth is disconnected on the power line of the resistance bridges. The I- wires for the four-wire measurement of the thermometer resistances need to be grounded on both sides, for this reason these wires run together with the main ground going from the ADR to the bridges to avoid pick-up loops. All the wires are shielded and are grounded only on one side.

4. Line filtering

The mains powering the electric utilities inside the shielded room are provided through isolation transformers with the central point of the output winding grounded in order to split the voltage into balanced power lines of 115 V + 115 V. This configuration provides a galvanic isolation from the noise present in the general mains of the building, and reduces the intensity of the mains electromagnetic fields inside the shielded room, thus lowering the intensity of capacitive coupling onto the instrumentation and the detected signals. The power lines from the isolation transformers enter the shielded room through filters in L-C-L configuration (#1 in Figure 5). Each couple of inductors are winded in the same magnetic core and in the same direction. In this way the load current flows in the two wires in opposite direction, canceling the effect of magnetization of the core. This solution permits the use of very high impedance inductors avoiding saturation of the magnetic core. The noise signals, usually in common mode on the two wires, see the full inductance and are attenuated by the filter. Possible (unusual) differential mode noise signals are bypassed by the capacitors. The capacitors of the filter on the J-FET Box heater line (#2 in Figure 5) are on the cryostat-side of the filter. The ground of the capacitors is connected to the shield of the cable coming from the filters panel and isolated from the ground of the cryostat. The thermometers are read in 4-wire configuration with the I- line connected to ground and the capacitors on the cryostat-side of the filter (#3 in Figure 5). In

the present configuration four thermometers, are fully wired. The power supply line of the magnet is filtered by three filters. The first one is on the filters panel, the second on the cryostat base and the third on the cryostat (#4, #5, and #6, respectively, in Figure 5). In order to avoid saturation of the inductors a balanced configuration has been used similarly to the mains filters. The resistance bridge interface with the computer uses a low speed bus, in order to limit EMI. The communication protocol is converted, outside the shielded room, to a standard GPIB interface. The filters used in the different lines of the bus are # 7, #8, and #9 in Figure 5. The capacitors in the data lines filter are through hole coaxial capacitors mounted on the wall of the shielded room.

Figure 5. Adopted filters: 1. Power lines, 2. J-FET Box Heater, 3. Thermometers, 4. Magnet filter #1, 5. Magnet filter #2, 6. Magnet filter #3, 7. Resistance bridge picobus power supply, 8. Resistance bridge picobus logic ground line, 9. Resistance bridge picobus data line.

5. New front-end circuit wiring inside the ADR In order to operate X-ray micro-calorimeters with high impedance sensors adopting the read-out electronics designed at CEA-Saclay we had to implement a modification of the ADR wiring (Figure 6), as well as building a new electronics box. Figure 7 shows the schematic of the new electronic board mounted inside the ADR in the new box that will replace the J-FET box shown in Figure 2.

Figure 6. Schematic of the new circuit inside the ADR designed to read-out four NTD Germanium X-ray micro-calorimeters with the new cold electronics.

Figure 7. Mounting of the new electronic board, operating at 2K, inside the ADR.

6. Summary and conclusions

We have upgraded our low-temperature testing facility, based on an ADR cryostat, to minimize the EM and mechanical interferences, in particular: we built a large size shielded room, implemented a new ground scheme, rewired and filtered the connections between the ADR and measurement and control instrumentation, designed a new configuration inside the ADR cold stage to test the low temperature front end electronics developed at CEA-Saclay within the research activities of the CESAR project (FP7-SPACE-2010-1). Test results will be discussed in a forthcoming paper.

7. References [1] Barbera M, Artale M A, Candia R, Collura A, Di Cicca G, Perinati E, Serio M, Serio S,

Varisco S, Silver E, Bandler S 2004 Proc. SPIE 5501 366. [2] Barbera M, Candia R, Collura A, Di Cicca G, Pelliciari C, Sciortino S, Varisco S 2006 Proc.

SPIE 6266 3F-1. [3] E. Silver, H. Schnopper, S. Bandler, N. Brinckhouse, S. Murray, M. Barbera, E. Takacs, J.

Gillaspy, J. Porto, I. Kink, M. Laming, N. Madden, D. Landis, J. Beeman, and E. Haller 2000 Astrophys J 541 495.

[4] E. Silver, S. Bandler, H. Schnopper, S. Murray, M. Barbera, N. Madden, D. Landis, J. Beeman, and E. Haller 2000 Proc. SPIE 4140 397.

[5] Chianetta G, Arnone C, Barbera M, Beeman J, Collura A, Lullo G, Perinati E, Silver E 2008 J Low Temp Phys 151 387.

[6] Lo Cicero U, Arnone C, Barbera M, Collura A, Lullo G 2012 J Low Temp Phys 167 535. [7] Lo Cicero U, Arnone C, Barbera M, Collura A, Lullo G 2012 J Low Temp Phys 167 541. [8] Perinati E, Barbera M, Varisco S, Silver E, Beeman J, Pigot C 2008 Rev Sci Instr. 79

053905-1. [9] V. Reveret, Y. Jin, C. Pigot, J. Putzeys, L. Rodriguez CESAR: Cryogenic Electronics for

Space Applications and Research 2013 J Low Temp Phys in press.

Acknowledgements

We acknowledge partial support by the European Union, project CESAR - Cryogenic Electronics for Space Applications and Research (FP7-SPACE-2010-1, project N. 263455).

Ballistic Phonon Transport in Low-Temperature Detectors

Djelal Osman, Stafford Withington, David Goldie and Dorota GlowackaUniversity of Cambridge, Cavendish Laboratory, JJ Thomson Ave, Cambridge CB3 0HE

E-mail: d.osman@mrao.cam.ac.uk

Abstract. We have successfully manufactured Transition Edge Sensors (TESs) whose thermalcharacteristics are completely characterised by ballistic phonon transport. In order to explore the regime forwhich ballistic transport dominates, these TESs have extremely small SiN support legs: 0.2 mm thick, 0.7to 1.0 mm wide, and 1.0 to 4.0 mm long. We show how, using classical acoustic wave theory, one requiresonly the geometry and elastic properties of the TES support legs to precisely predict the saturation power,thermal conductance and thermal fluctuation noise of these devices. For our TESs, which have transitiontemperatures of approximately 107mK, the majority of heat flux is transported by fewer than 6 acousticmodes per leg, having wavelengths ranging from 0.3 to 4.0 mm. We conclude that, for the first time, TESshave been created whose thermal response is fully characterised by a classical acoustic wave model; in thiscase, by the physics of few-mode ballistic transport.

1. Introduction

The aim of our work is to understand the underlying thermal transport mechanisms in amorphous dielectricmicrobridges, so as to be able to incorporate phononic filters into the support legs of Transition Edge Sensors(TESs). We have successfully manufactured TESs that not only operate in the regime for which ballisticphonon transport dominates, but for which saturation powers, thermal conductances and noise values arefully accounted for by classical acoustic wave theory. These results complement previous work, in whicha variety of elastic and inelastic phonon scattering mechanisms were used to simulate the low-temperaturethermal behaviour of patterned dielectric microbridges[1][2]. In order to operate in the ballistic regime,thermal power must be transmitted over distances that are short with respect to the acoustic attenuationlength of the material. In this paper, we show experimental verification that our ballistic acoustic modelprecisely predicts the thermal behaviour of TESs with short support leg structures. There are four thermaltransport mechanisms in amorphous dielectrics: ballistic, elastic diffusive, inelastic diffusive and localisedtransport. Being able to predict which of these mechanisms will dominate for a dielectric microbridge ofany geometry, and by also understanding how these mechanisms affect the operating thermal parameters ofa TES (such as the saturation power, thermal conductance and noise equivalent power (NEP)), we aim tocreate predictable and reproducible methods for the design and creation of ultra low-noise TESs, which willno longer require the production of ever longer support leg structures to achieve lower NEPs.

2. Experiment

To probe the dynamics of a TES in the regime where thermal transport is dominated by ballistic phonons, onemust first ensure that the dielectric thermal support structures between the TES and the heat bath are shortrelative to the acoustic attenuation length of the material. Additionally, narrow support legs act to restrictthe number of power-carrying acoustic modes, which reduces phonon noise. As such, a series of devices weredesigned and manufactured with leg widths varying from 0.7 to 1.25 mm, and leg lengths varying from 1 to4 mm, so as to investigate the geometric regime in which ballistic transport dominates in amorphous SiN.

For the measurement of TES current, we used a two-stage SQUID designed and manufactured byPhysikalisch-Technische Bundesanstalt (PTB), having a low noise current contribution of approximately 8pA/

pHz. The TESs and the SQUIDs were placed inside an Adiabatic Demagnetisation Refrigerator (ADR).

The coldest temperature achieved by our refrigerator at the sample stage was 60 mK. By regulating thecurrent in the ADR magnet, we were able to control the sample stage temperature.

3. Acoustic Wave Theory

To predict the thermal power flux along a microbridge as a function of bath temperature, we first calculatedthe acoustic phonon modes that are available to carry power between the TES and the heat bath througha dielectric support leg of arbitrary dimensions and of rectangular cross-section. To do this, we solved thestress-strain tensor wave equation:

!2ui + Cijkl@2uk

@xj@xl= 0, (1)

where ui is the strain for a given cartesian direction, i, Cijkl is the fourth-rank stiffness tensor, is themass density of the material and ! is the angular frequency of the traveling wave. We solved this relationnumerically by decomposing the displacement vector into a set of Cartesian power-law basis functions, whichspan the entire volume of the microbridge; as was done by Nishiguchi (1997)[3]. Equation (1) gives usthe wavenumber-dependent eigenfrequencies of propagating acoustic modes for any material, and for anyrectangular cross-section.

Figure 1 shows a plot of calculated dispersion profiles for phonon frequencies up to 10GHz in anamorphous SiN microbridge of cross-section 700nm x 200nm. The only parameters that were requiredto calculate these dispersion relationships were the cross-sectional dimensions of the microbridge and theYoung’s modulus, Poisson ratio and mass density of the material. For amorphous SiN, we used a Young’smodulus of 280 GPa, Poisson ratio of 0.28 and a mass density of 3.14 g/cm3. In Figure 1, we also plotanalytic approximations to the lowest four dispersion relationships, and find that these fit very well in thelow-wavenumber limit.

0 0.2 0.4 0.6 0.8 10

2

4

6

8

10

q [µm!1]

v [G

Hz]

Figure 1. Dispersion relationships calculated for a rectangular cross-section SiN microbridge of dimension700nm x 200nm. Axes are frequency, , and wavenumber, q. The black lines are the modeled acousticdispersion profiles. The dashed coloured lines show the analytic approximations for the lowest four acousticmodes, which fit the model very well at low wavenumbers.

4. Calculating Power Flow

With the acoustic dispersion relationships, we are able to calculate the net power transmission betweenthe two ends of the microbridge. In the case of our devices, one end of the support legs connects to theTES island at the superconducting transition temperature, whilst the other end connects to the rest of thewafer at the bath temperature. Given that the TES island and the wafer are so much larger than the low-dimensional bridging structure, we make the reasonable assumption that the TES and the heat bath behaveas infinite reservoirs of infinite degrees of freedom. Additionally, we assume that the density of states of theseinfinite reservoirs is determined by Bose-Einstein statistics. Thus, the occupation number of phonons at thetermination boundaries of a microbridge is taken as dependent only upon the temperature of the heat loadat each end.

Mathematically, there are an infinite number of independent acoustic modes that are capable oftransmitting power. However, the power that is transmitted through these modes has a blackbody spectraldistribution, and thus rolls-off to zero above some temperature-dependent frequency. Hence, the total powertransferred through the microbridge is calculated by summing over the total power transmitted by each of theindependent acoustic modes, in which each acoustic mode transfers some portion of the blackbody emissionspectrum determined by its cut-off frequency:

P (T, Tb) =X

i

2

41

i

h

exp( hkT ) 1

d 1

i

h

exp( hkTb

) 1d

3

5 , (2)

where T is the temperature of the TES island, Tb is the temperature of the bath (i.e. the temperature of theADR), is the non-angular wave frequency, i is the cut-off frequency for acoustic mode i, h is the Planckconstant and k is the Boltzmann constant.

0240

5

10

15

20

25

P/d! [yW/Hz]

! [G

Hz]

Th (104.5mK)Tc (63.0mK)

0240

5

10

15

20

25

P/d! [yW/Hz]

! [G

Hz]

0 0.5 1 1.50

5

10

15

20

25

1/" [1/µm]

! [G

Hz]

Figure 2. The plot on the left shows the PSD emitted from the hot end of the microbridge at a temperatureTh=104.5mK (red line), and the PSD emitted in the opposing direction from the cold end, Tc=63mK (blueline). The central plot shows the net PSD transmitted across the microbridge, which is the PSD at the hotend subtracted by the PSD at the cold end. The plot on the far right shows the dispersion profiles of theacoustic modes that are carrying this power. These plots are for a SiN microbridge with a cross-section of830nm x 200nm.

To illustrate this concept, we can plot the power spectral density (PSD) of the total power passingthrough the acoustic channels expressed in equation (2). Figure 2 shows how the availability of propagatingacoustic waves above their respective cut-off frequencies introduces step-wise increases in the net powertransfer between a TES and the colder heat bath as a function of phonon frequency. By comparing thesePSD plots with their corresponding dispersion profiles, we find that all thermal power is transmitted byphonon wavelengths between 300 nm and 4 mm at the operating temperatures of our TESs.

5. Thermal Power Transmission

Knowing the net PSD for a SiN microbridge with specific hot and cold termination temperatures allows usto plot the total net power transmitted from the TES to the heat bath. Experimentally measuring this netthermal power transmission is equivalent to measuring the saturation power of the TES. If we say that thehot termination temperature, Th, is equal to the TES superconducting transition temperature, and that thecold termination temperature, Tc, is equal to the temperature of the ADR, then we can integrate the netPSD to find the temperature dependent thermal power transmission of a TES for which the cross-sectionaldimensions of its support legs are known. For this purpose, precise measurements of the dimensions of ourTES support legs were made using Scanning Electron Microscopy (SEM). Thus, calculating the expected netpower flow, we have found that our model matches our measured powers extremely well. The results fromone of our six devices is shown in Figure 3. The other five devices show an equally excellent match betweenthe model and measured data.

0 20 40 60 80 1000

50

100

150

Tc [mK]

P [fW

]

Figure 3. Thermal power transmission, P, as a function of bath temperature, Tc, for one of the sixTESs measured. The square data points are the experimental measurements. The red line is the resultof the acoustic model using the known dimensions of the TES support legs, and the measured transitiontemperature for this device. The model and data agree very closely.

6. Thermal Fluctuation Noise

Using a similar method to that outlined previously for the calculation of TES saturation power, we can alsomodel the phonon contributions to the total noise in the device as a function of bath temperature. However,in place of equation (2), we use the following expression for the r.m.s. fluctuations in thermal power flow:

4P (T )2 =X

n

2B

ˆn

"h

exp[( hkT ) 1]

#2

d X

n

2B

ˆn

[h]2

exp[( hkT ) 1]

d, (3)

where B is the readout bandwidth. The first term on the R.H.S. corresponds to classical bunched noise, whilstthe second term corresponds to shot noise. Thus, calculating 4P (T ) for T = Th and T = Tc, and summingthe two values in quadrature, one arrives at the total phonon NEP for a given temperature difference acrossthe microbridge. Figure 4 shows the measured frequency-dependent total NEP of one of our TESs (blackdata points) and the modeled phonon NEP (red line). As well as phonon noise, the measured NEP in Figure4 includes 1/f noise, electronic readout noise, and Johnson noise contributions. However, these additionalnoise contributions are negligible in the operating frequency range of the TES (between 10 - 50Hz). Hence,this figure shows that not only are our measured devices phonon noise limited, but our acoustic model cancalculate the phonon contribution to the NEP accurately.

100

101

102

103

0

1

2

3

4

5

! [Hz]

NE

P [aW

Hz

!1/2

]

Figure 4. Readout frequency, , against NEP. The black data points are the measured noise spectrum fromthe same TES measured in Figure 3, which has a thermal conductance of 2.85 pW/K. The red line showsthe expected phonon NEP contribution calculated from our acoustic model, which is 1.2 aW/

pHz.

7. Summary

In summary, we have successfully manufactured TESs with extremely small support leg structures (<0.7 mmwide, <1 mm long and 0.2 mm thick). We have also experimentally verified that, using classical acoustic wavetheory, one can predict the power transmission, thermal conductance, and thermal fluctuation noise of thesedevices. Emphasis should be placed on the fact that the only parameters one requires to model the thermalcharacteristics of these TESs are the cross-sectional dimensions of the supporting legs, the Young’s modulus,Poisson ratio and mass density of SiN, and the superconducting transition temperature; all of which areknown values. Thus, remarkably, our model does not fit for any free parameters. We have shown that, forthe first time, TESs have been manufactured whose thermal response is completely characterised by few-modeballistic phonon transport. The number of effective power modes available in each of the support legs of ourTESs can be calculated by normalising the total net power transmission by the total power transmitted fora single acoustic mode across the entire blackbody frequency spectrum. By doing so, we find that our TESsoperate in the regime for which all the power is transmitted by approximately the lowest 6 acoustic modesper leg, given a bath temperature of 60mK. We have also found that the phonon noise in our devices isof the order of 1.2 aW/

pHz. By analysing the frequency-dependence of the integrands in equation (3), it

can be shown that ballistic phonon noise is dominated by shot noise. For our future work, we aim to use ouracoustic wave model to incorporate phononic filters into the support legs of TESs, so as to decrease thermalconductance and noise without lengthening the microbridges.

References

[1] Withington S and Goldie D. J. 2011 Phys. Rev. B 83 195418[2] Withington S and Goldie D. J. 2013 Phys. Rev. B 87 205442[3] Nishiguchi N, Ando Y and Wybourne M. N. 1997 Phys.: Condens. Matter 9 5751

Optimisation of kinetic inductance detectors for millimetre and submillimetre wave detection and the study of titanium nitride thin films for future applications

Grégoire Coiffard1, Karl-Friedrich Schuster1, Alessandro Monfardini2, Amar Adane1, Arnaud Barbier 1, Catherine Boucher1, Martino Calvo2, Johannes Goupy2, Samuel Leclercq1 and Stéphane Pignard3

1 Insitut de Radioastronomie Millimétrique, 300 rue de la Piscine, 38406 Saint Martin d'Hères, France 2 Institut Néel, CNRS et Université de Grenoble, 25 rue des Marytrs, 38042 Grenoble, France 3 Laboratoire des Matériaux et du Génie Physique, Grenoble-INP Minatec 3 parvis Louis Néel, 38016 Grenoble, France

E-mail: coiffard@iram.fr

Abstract. We present some of the recent updates that have been carried out to improve the sensitivity of lumped element kinetic inductance detectors (LEKIDs) for the NIKA experiment [1]. The optical absorption of such detectors can be improved by adding an anti-reflection coating on the back-side of the silicon substrate. Here, we apply deep silicon etching to the substrate in order to decrease the effective dielectric constant of the silicon which allows a broader frequency response of the detector. Another problem, that disturbs the detector response, is the presence of slot-modes in the frequency multiplexing coplanar feed/readout line. To eliminate this, we developed superconducting bridges over the feed line. Our current KID detectors are made of very thin aluminium films which present some limitations concerning the design layout and sensitivity. We present the development of titanium nitride (TiN) at IRAM as an alternative material, and recent progresses that have been made in the deposition process of TiN thin films. Because of its high kinetic inductance, non-stoichiometric TiN seems to be a promising material for KIDs. However deposition of such TiN is complex and we have to find a reproducible process. The deposition process is examined and the different techniques used to characterise our films are presented.

1. Introduction Millimetre and submillimetre radioastronomy is of great interest to study the early stages of stars or galaxy formation. Historically, bolometer arrays have been used for millimetre wave detection but the use of larger arrays to increase the mapping speed of extended sources is needed, a challenge which requires complex readout systems for most bolometer based solution. As an alternative kinetic inductance detectors have seen a rapid development for 10 years [2]. These detectors are easily multiplexed in the frequency-domain which is a great advantage for building very large arrays. The kinetic inductance detectors developed for the IRAM 30-meter telescope at Pico Veleta (Spain) have

rapidly reached sensitivities comparable to the state-of-the-art bolometers and present now a real alternative for ground based millimetre and sub-millimetre wave detection.

2. Kinetic inductance detectors (KIDs) Kinetic inductance detectors are superconducting devices that use photon-assisted pair breaking to detect incoming waves.

In a superconducting thin film, photons with energy greater than the superconducting gap ∆>= 2νhE break apart Cooper pairs. This implies an increase of the quasi-particles number which

changes the surface impedance SZ of the film ( kinSS LjRZ ω+= where SR is the surface resistance

and kinL the kinetic inductance). This effect is interesting when a resonator is patterned in the

superconducting thin film. An increase of the number of quasi-particles implies an increase of the kinetic inductance which lowers the resonance frequency of the resonator. By coupling the resonator to a readout transmission line, this shift can be measured.

Figure 1. The increase of Lkin due to the absorption of millimetre radiation shifts the microwave-resonance to a lower frequency (black curve) and the resonance becomes less deep due to a higher RS

Resonators with slightly different resonant frequencies are designed and coupled to one readout

transmission line. This frequency multiplexing allows to design and fabricate very large arrays of several hundred resonators with only one readout line.

For the moment, KIDs arrays are made of aluminium. To achieve a high kinetic inductance which leads to a better sensitivity of the detector, the films have to be very thin (20 nm). The resonator consists of an inductive meander and an interdigital capacitor, the length and the number of the fingers of the interdigital capacitor determines the resonance frequency of each resonator (Figure 2 (b)). For NIKA, the resonant frequencies are between 1 and 2 GHz with a spacing of 2 MHz. The resonators are coupled to the transmission line via the inductive meander, the distance between the line and the resonator is fixed to 14 µm to achieved the same coupling factor for each pixels of the array (Figure 2 (c)).

Figure 2. (a) Layout of a dual-polarization LEKID [3] (b) Zoom on the fingers of the interdigital capacitor (c) Coupling area of the LEKID to the coplanar waveguide transmission line

3. Optimization of the optical coupling In the millimetre and submillimetre region of the electromagnetic spectrum only a few bands are available for detection at ground based telescopes. Here we focus on the 1 mm atmospheric window which lies between 200 and 280 GHz.

3.1. Classic array One of the characteristics of the KIDs is the direct absorption of the incoming power into the substrate. Thus, it is very important to maximise this absorption in order to optimise the detector response and sensitivity.

Figure 3. Illustration of the room temperature set-up used to measure the absorption of KIDs array. (a) Harmonic mixer (b) Frequency multiplier (c) coupler (d) Corrugated feed-horn + lens (e) Silicon substrate (f) KIDs (g) Backshort

We have carried out simulations with FEM simulator (CST Microwave Studio) in order to determine the optimum values for the substrate thickness and the backshort-detector distance. The results of these simulations can be compared to room temperature quasi-optical reflectivity measurements which are less time consuming than a cryogenic experiment. The room temperature measurement set-up consists (Figure 3) of a vector network analyzer (VNA) that generates a low frequency band. This band is then translated into a millimetre band (200-270 GHz for our experiment), then a feed-horn focuses the wave on the sample. Finally, the reflected wave returns the way back to the VNA which measures the reflection coefficient S11 in dB. The absorption of the structure is then

determined as 2

111 S− . We use the resistance residual ratio (RRR) of aluminium to adapt the room

temperature resistivity of the film to the cold one. For our aluminium films, 2=RRR so the films for room temperature measurements have to be two time thicker than the one for cryogenic measurements if we want to compare the results. Low temperature measurements are done with a Martin-Puplett interferometer and show comparable results for optical absorption.

(a) (b)

Figure 4. (a) Structure of the detector with the silicon substrate, the KID and the backshort (b) Comparison of CST simulation and room temperature measurement of the absorption for an electric field perpendicular to the interdigital capacitor

In Figure 4 we see that the absorption determined with the room temperature set-up is in good

agreement with the simulation calculation. The slight differences can be due to the environment of the RF measurement set-up. We notice that the absorption reaches a maximum around 240 GHz which matches the centre of the 1 mm band but the absorption decreases too sharply and we observe that the 1 mm band is not fully exploited with this kind of KID structure.

3.2. Anti-reflection coating In electronics, impedance matching is often useful to minimise reflections in a microwave circuit. Here we use the same principle to effectively match the millimetre wave to the detector. An anti-reflection layer is added between the air and the substrate in order to minimise reflections on the silicon. The anti-reflection theory is well known [4], and to reduce reflections between vacuum and

silicon we have to insert a material with a dielectric constant equal to 42.30 == SiAR εεε .

Moreover, the thickness of this layer should be 1704

0 ==AR

ARLε

λµm.

A material with such a dielectric constant is difficult to find and moreover it will be a real challenge to attach it on the silicon substrate without any defects or air gap. An elegant solution is to use directly micro-machined silicon as anti-reflection layer. Silicon can be etched in order to decrease its dielectric constant from 11.7 to 3.42 [5]. Another advantage of this technique is that the substrate and the anti-reflection coating have the same thermal expansion coefficient which is interesting when the detector is cooled down. We choose to etch hexagonal structures for symmetry and robustness reasons. The dimensions of the hexagons are determined to achieve a filling factor of 35 %, giving an effective dielectric constant equal to 3.42, that corresponds to a side of 170 µm and a spacing between hexagon of 70 µm. Simulations have been carried out in order to confirm a better absorption of detectors with anti-reflection in the 1 mm band (Figure 5).

Substrates of 300 µm are used and the backside is etched with deep reactive ion etching (DRIE). The depth of the hexagon is around 150 µm and he thickness of the pure silicon is therefore 150 µm too. The KIDs array is patterned on the other side of this etched substrate. Finally, the absorption with the anti-reflection is determined at room temperature (Figure 5).

(a) (b)

Figure 5. (a) Scanning electron microscopy images of the anti-reflection etched silicon (b) Comparison of room temperature measurements and simulations of a normal detector and a detector with anti-reflection coating

First measurements of arrays with anti-reflection coating are promising. The filling of the 1 mm

band is much better and the absorption is always greater than 50 %. Nevertheless we need to perform more experiments in order to really determine the sensitivity of these news KIDs arrays.

4. Improvement of the microwave readout line Millimetre wave coupling is not the only parameter that can be improved. The microwave frequency multiplexing feed line can also be upgraded in order to gain even more sensitivity for the array. The coplanar waveguide (CPW) transmission line is made with a centre strip separated from two ground planes by a gap. This line is well adapted for our application as we can directly match it to 50 Ω with reasonable dimensions for the centre strip (20 µm) and the gaps (12 µm) and it can be easily connected with the coaxial RF connector. Unfortunately, due to turns and ground planes asymmetry, the potential of the two ground planes can differ which in turn generates undesired slot-modes. These modes not only can lead to undesired parasitic in-band resonances and standing waves along the feed line but also lead to parasitic pixel-to-line and pixel-to-pixel coupling.

To prevent the propagation of these parasitic modes, the ground planes have to be set to the same potential. We attempt this by connecting ground planes with bond wires before and after each turns of the CPW and every three pixels. Such a technique suppresses parasitics resonances in the feed line and allows to optimise the distribution of the coupling quality factor (QC) of resonators (Figure 6), which determines how much power each resonator receives from the transmission line. The QC of a resonator defines the shape of its resonance curve (depth and width), therefore it is important to achieve the best QC distribution over an array in order to obtain the best readout signal for all resonators.

51047.056.0 ×±=CQ 51008.023.0 ×±=CQ

(a) (b)

Figure 6. (a) Histogram of the coupling quality factor for all the resonators of an array (b) Histogram of the QC for an array with bond wires between the ground planes of the CPW

Encouraged by these results we have developed two different techniques to connect the ground planes with bridges to avoid manual bonding which is time consuming. The first one uses a SiO2 insulating layer on the centre strip of the CPW and small superconducting bridges can be fabricate over it. The other one consists of micro-fabrication of aluminium air-bridges, this technique is appealing but required special care in order to adapt the process to KIDs array. Recently, we carried out the first measurement of array with bridges and we need to characterise and investigate the different effects of these techniques with more arrays.

5. Development of titanium Nitride (TiN) thin films Despite the various optimisations presented above, the performance of our KIDs remains limited by the material used currently (Al). To increase the sensitivity of the detector, a very efficient technique is to increase the kinetic inductance in the inductive meander by lowering its volume. We are currently working with aluminium films of 20 nm thickness, giving a kinetic inductance Lkin = 2 pH/square, but the sensitivity is still not optimal. However thinner films and a narrower inductive meander are difficult to be produced with good homogeneity. A solution is to use a material such as the non-stoichiometric titanium nitride (Lkin, TiN = 20 pH/square) which has a high normal conducting resistivity. Moreover, TiN has a very high internal quality factor that allows to build very sensitive KIDs array [6]. However deposition of non stoichiometric TiN is complex, with significant challenges in material physics but also reproducibility.

5.1. Sputtering deposition of titanium nitride In our investigation titanium nitride films are sputter deposited on 2'' silicon wafer from a 4'' Ti target

at room temperature and with a base pressure of 8102 −×=baseP mbar. The inlets provide Ar and N2

atmosphere for the reactive sputtering of TiN. TiN is very sensitive to species such as O2 or H2O and it is important to eliminate them as much as possible from the reactive chamber. Before the TiN deposition, the reactive chamber is thus pumped down during many hours and then pure Ti is sputtered and evaporated in order to have a getter effect that traps the most polluting species for TiN. During this cleaning process, the relative quantities of different species present in the chamber are visualised with a mass spectrometer. When the levels of oxygen and water stabilise, the deposition process of TiN begins.

Firstly, the silicon wafer is cleaned in a pure Ar RF plasma in order to eliminate the native oxide layer. Then the Ti target is sputtered in a pure Ar atmosphere to eliminate contamination. This pre-sputtering is made at a pressure of 0.2 Pa with an Ar flow of 30 sccm and a power of 450 W during 5 minutes. The N2 valve is opened and the flow is controlled to have the Ar/N2 ratio that gives the desired TiN stoichiometry. When the plasma is well established the shutter is opened and the sputtering of TiN begins on the silicon substrate.

5.2. Characterization of the TiN films To break Cooper pairs a photon energy has to be CBTkhE 2

32 =∆>= ν , where ∆ is the

superconducting gap, TC the superconducting transition temperature and kb the Boltzmann constant. For millimetre wave detection this implies a TC = 1 K. Stoichiometric TiN has a TC of 4.5 K; by lowering the nitrogen content of the material we can reach TC around 1 K. Figure 7 presents the variation of TC of our TiN1-x films versus the N2 flux during the deposition process. The variation of TC is relatively steep in particular around the target value of 1 K; moreover a relatively low N2 flux is needed to obtain non-stoichiometric TiN1-x films which makes the process control difficult.

Figure 7. Evolution of the critical temperature TC of TiN thin films depending on the variation of the N2 flow rate in the plasma mixture

As the determination and the control of TC with classical low temperature measurements are relatively time consuming, we are using spectroscopic ellipsometry to correlate room temperature spectra with the value of TC [7].

The spectroscopic ellipsometry measures the change of the polarization of the light (combination of

p and s-waves) after reflection on a sample. The change is written as ∆Ψ= ie)tan(ρ where Ψ is the

ratio of the amplitude diminutions of the p-waves over the s-waves and ∆ is the phase difference induced by the reflection. Here Ψ and ∆ are acquired as a function of the wavelength (range 400-850 nm) and we choose to use the complex dielectric function of the sample which can be expressed as

ΦΦ

+−+Ψ=+ 22

2

221 sintan

1

1sin

ρρεε i where o70=Φ is the angle of incidence of the beam.

The complex dielectric function is interesting because the energy at which 01 =ε corresponds to

the unscreened plasma energy

2/1

0

2

*

=

m

NeE p ε

h where N is the density of conduction electron, e

is the electron charge, 0ε is the permittivity of free space and *m is the electron effective mass. Our

results show that the less nitrogen in the film, the more free carriers are available. In other words, pE

increases when TC decreases through the correlation with the nitrogen stoichiometry (Figure 8). However we do not have access to the absolute stoichiometry of the film a quantity which can be evaluated with more powerful characterisation technique such as Rutherford backscattering spectrometry (RBS) that we plan to use in the future.

Figure 8. TC plotted vs. the photon energy where ε1=0

6. Conclusion We have optimised lumped element kinetic inductance detector arrays in order to maximise the absorption of millimetre wave in the 1 mm band available at the telescope. DRIE of the silicon substrate gives encouraging results, the absorption is greater than 50 % within the whole 1 mm band. More measurements regarding the sensitivity of these arrays have to be carried out to confirm an absolute increase in sensitivity.

Concerning the microwave readout line, simple wire bondings suppressed the parasitic slot line modes and now that alternative and more advanced fabrication processes for bridges are well established, characterisations measurements are scheduled.

Finally, the study of the deposition process of titanium nitride is now reproducible and we have found an efficient way to correlate the critical temperature of the film to spectroscopic ellipsometry spectra. Now we plan to investigate the spatial homogeneity in nitrogen content across the substrates with chemical analysis. Moreover, we hope to improve this homogeneity with a new 6'' target and modified gas inlet for the reactive chamber.

Acknowledgment This work has been partially funded by the Foundation Nanoscience Grenoble, the ANR under the contracts "MKIDS" and "NIKA". This work has been partially supported by the LabEx FOCUS ANR-11-LABX-0013.

This work has benefited from the support of the European Research Council Advanced Grant ORISTARS under the European Union's Seventh Framework Programme (Grant Agreement no. 291294).

References [1] A. Monfardini et al., NIKA: A Millimeter-Wave Kinetic Inductance Camera, Astronomy and

Astrophysics 521, A29, 2010 [2] P. K. Day et al., A broadband superconducting detector suitable for use in large arrays, Nature,

vol. 425, p. 817-821, 2003 [3] M. Roesch et al., Dual polarization Lumped Element Kinetic Inductance Detectors (LEKIDs)

for 1.25 and 2.05 mm, 23rd ISSTT, 2012 [4] M. Born & E. Wolf, Principles of Optics, Pergamon Press, 1980 [5] K.F. Schuster et al., Micro-machined quasi-optical Elements for THz Applications, 16th ISSTT,

2005 [6] H.G. Leduc et al., Titanium nitride films for ultrasensitive microresonators detectors, Appl.

Phys. Lett. 97, 102509, 2010 [7] M. R. Vissers et al., characterization and in situ monitoring of sub stoichiometric adjustable TC

titanium nitride growth, arXiv:1209.4626v1, 2012

Superconducting microresonator detectors for

neutrino mass measurements in Milano

A. Giachero1, P. K. Day2, P. Falferi3,4 M. Faverzani1, E. Ferri1, C.Giordano3, H. G. LeDuc2 B. Marghesin3, F. Mattedi3, R. Mezzena5,R. Nizzolo1, A. Nucciotti1,1University and INFN of Milano-Bicocca, Department of Physics, Milan, Italy2Jet Propulsion Laboratory and California Institute of Technology, Pasadena, CA, USA3Bruno Kessler Foundation, Trento, Italy4Istituto di Fotonica e Nanotecnologie, CNR, Trento, Italy5University of Trento, Department of Physics, Trento, Italy

E-mail: Andrea.Giachero@mib.infn.it

Abstract. Microwave Kinetic Inductance Detectors (MKIDs) are superconducting detectorssuitable for large-scale multiplexed frequency domain readout and characterized by good energyresolution and fast response. Our goal is to develop arrays of MKIDs applicable to the neutrinomass estimation through the calorimetric measurement of the energy spectra of the 163Hoelectron capture (EC) decay. In order to achieve this goal, a study aimed to the selectionof the best design and material for the detectors is in progress.

These superconductive resonators are low temperature detectors excited by a probe signal,generated by a microwave synthesizer, and readout by a cryogenic HEMT amplifier at 4 K and,subsequently, by room-temperature amplifier. Homodyne mixing with the original excitationsignal is accomplished on the amplified signal employing an IQ-mixer. The resulting I and Qsignals are acquired by a commercial DAQ and then projected offline into the frequency anddissipation quadrature signals. In this contribution we present the developed readout setupand some characterization results obtained from different materials (stoichiometric TiN, sub-stoichiometric TiN and Ti/TiN multilayer) using two different X-ray sources in the range of theenergy of interest (Al X-ray, 1.5 keV, and 55Fe, 6 keV).

1. IntroductionThe determination of the neutrino masses is one of the most challenging open issue in particlephysics. Experiments, based on kinematic analysis of electrons emitted in single β-decay, arethe only ones dedicated to effective electron-neutrino mass determination. The method consistsin searching for a tiny deformation caused by a non-zero neutrino mass to the spectrum nearits end point. The most stringent results come from electrostatic spectrometers on tritiumdecay (E0 = 18.6 keV). The Troitsk experiment has set an upper limit on neutrino mass ofme ≤ 2.5 eV [1], while the Mainz collaboration has reached mµ ≤ 2.2 eV [2]. KATRIN, anext generation experiment, is designed to reach a sensitivity of 0.2 eV/c2 in five years [3, 4].Alternative approaches to the spectrometry are focused on using microwave antennas to detectcoherent cyclotron radiation emitted by individual decay electrons in a magnetic field [5] andon using microcalorimeters where beta source is embedded in the detector so that all the energyemitted in the decay is measured, except the one taken away by the neutrino [6, 7].

The measurement of the end point of nuclear beta or electron capture (EC) decays spectrais the only theory-unrelated method usable to sort out the problem of the neutrino mass.In order to achieve enough statistics avoiding artifacts, such as pile up, a large number ofdetectors (' 104), characterized by good energy resolution (∆E ' eV @ keV) and fast response(' µs), is required. Among the family of nuclear beta or electron capture decays, an interestingisotope suitable for the neutrino mass experiment could be the 163Ho, which decays via EC withQ ' (2÷ 3) keV, as proposed by De Rujula and Lusignoli in 1982 [8].

Low temperature detectors (LTD) have proven their potential for applications requiring veryhigh sensitivity over the last 50 years. Currently, a major focus for advancing this technologyis toward producing large arrays optimizing the sensitive sensors like Transition Edge Sensors(TES), Metallic Magnetic Calorimeters (MMC) or Kinetic Inductance Detectors (MKID). Inany case the only way to improve the overall sensitivity of a measurement is to increase thenumber of detectors.

Our aim in this challenge is to develop arrays of microresonator detectors [9] (MKIDs)applicable to the calorimetric measurement of the energy spectra of 163Ho. Currently, a studywith the purpose to select the best material and the best design for the detectors is in progress.In this contribution the current progress on the design and on the performed tests is presented.

2. Detector productionThe detector array chips we designed are composed by 16 individual microresonators withfrequencies distributed in the range from 3 to 6 GHz and with a variety of coupling qualityfactors. The resonator geometry (lumped-element resonator) consists of two interdigitalcapacitors (IDC) connected with a coplanar strip (CPS) transmission line that works as inductor.Each resonator is capacitively coupled on one side to a coplanar waveguide (CPW) line that isused for the readout. The strength of the coupling, which sets the coupling quality factor Qc ,is determined by the width of the gap between the CPW line and the resonator. The spacingbetween the IDC finger is 10 or 30 µm. This large spacing is intended to reduce the two-levelsystem (TLS) noise associated with amorphous dielectric layers at surfaces. For more detailssee [10].

The goal is to implement this detector geometry in superconductive films with a criticaltemperature Tc in the range (1 ÷ 1.5) K in order to have a low energy gap and long quasi-particle recombination time (i.e. low generation-recombination noise). Titanium Nitride (TiN)has been recently investigated as superconducting material it is a well suited for the productionof MKIDs. TiN resonators show very high quality factors and large ratios of kinetic to magneticinductance. Furthermore, the amount of Nitrogen added to the Titanium film allows to producesub-stoichiometric TiNx (where x is the concentration of N); in this way it is theoreticallypossible to tune the transition temperature in the range (0÷ 4.5) K [11]. Nevertheless we foundthat the targeted Tc has a strong dependence on x and control this resulted very difficult. Forthese reasons we considered a different approach using a multilayer of pure Ti and stoichiometricTiN [12, 13]. Exploiting the proximity effect the Tc of a superconducting material can be reducedby the superposition of a normal metal or a metal with a superconductive transition at lowertemperature. In this configuration the Cooper pairs leak into the normal metal, consequentiallythe pair density in the superconducting metal is reduced as well as the Tc. The magnitude ofthis effect depends on the thickness of the two layers. In the case of Ti/TiN multilayers, byadjusting these thickness one is able to tune the Tc between the transition temperature of Ti(0.4 K) and TiN (4.5 K).

We produced Ti/TiN multilayer by the superposition of layers Ti/TiN with Ti at the bottominterface and TiN on the top. The maximum number of stacked layers is 8. The films have beensputtered on high resistance Silicon wafers, which had been cleaned and etched with hydrofluoricacid (HF) to remove the native oxide. Results prove that the superconductive transition can

3.0300f [GHz]

3.0275 3.03253.0250 3.0350

0

5

-5

-10

-15

-20

Forw

ard

Tran

smitt

ion

- S 2

1 [dB

]

50

150

250

350

450

T [m

K]

100 200 300 400 5000T [mK]

Q-1

. 10-4

0

2

4

6

8

datafit

Figure 1. Example of energy gap ∆ measurement for a multilayer film. It is easy to see howthe resonant frequency f0 and the merit factor Q of one single resonator decrease with thetemperature increasing.

Film TC [K] ∆ [meV] QStoichiometric TiN (250nm) 4.6 0.8 105

Sub-stoichiometric TiN (160nm) 2.5 0.39 106

Ti(10nm)/TiN(15nm) 8 layers 1.6 0.26 105

Ti(10nm)/TiN(12nm) 8 layers 1.2 0.17 5 · 105

Table 1. Critical temperature, gap parameter and internal quality factor obtained for theproduced films. The sub-stoichiometric was provided by Jet Propulsion Laboratory (Pasadena,CA, USA) while the stoichiometric and multilayer films by Bruno Kessler Foundation (Trento).

be tuned in the (0.5 ÷ 4.6) K temperature range by properly choosing the Ti thickness withinthe (0 ÷ 15) nm range, and the TiN thickness within the (7 ÷ 100) nm range. The Tc can belowered by reducing the TiN thickness with fixed thickness of Ti or by keeping fixed the TiNthickness and increasing that of Ti. In this latter case the Tc is more sensitive to layer thicknessvariations in the (5÷10) nm range and it is nearly stable for Ti thickness < 5 nm and > 10 nm.The produced Ti/TiN multilayers have shown a high uniformity and high quality factor [12].

3. Device characterizationThe produced detectors were placed in a dilution fridge (Oxford MX 40, Cryogenics Laboratoryof the University of Milano-Bicocca) and cooled down to temperature as low as 100 mK. Foreach resonator, we estimated the critical temperature (Tc), the resonance frequency (f0) andthe quality factor (Q) by measuring and fitting the forward transmission S21 as a function ofthe temperature (figure 1, left), with a Vector Network Analyzer (HP 8753E, 30 kHz ÷ 6 GHz)remotely controlled. For each temperature value we extrapolate the internal quality factor Qiand fitting its behavior versus the temperature it was possible to estimate the energy gap ∆ inaccording with the Mattis-Bardeen theory [14]. The measured values for four different films arereported in the table 1.

Triggered PulseAverage Pulse

0.05 0.1 0.15 0.20Time [ms]

1

2

0

3

Dis

sipa

tion

[a.u

.]

0

0.5

1

-0.5

1.5

Freq

uenc

ySh

ift [a

.u.] <tf> ≃ 100 µs

<tr> ≃ 5 µs

0 2 4 6 8 10Frequency [a.u.]

1

10

100

1000

Cou

nts/

bin

1.5 keV6.0 keV

Ti/TiN multilayer (200nm) @ 105mK

Figure 2. Left plot: observed pulser response for Ti/TiN multilayers from an Aluminum x-ray fluorescence source. Data are digitized at 2.5 MS/s and projected into the frequency anddissipation quadrature. Right plot: energy spectra obtained with two different sources: 55Fe(6.0 keV, blue line) and Aluminum X-ray fluorescence (1.5 keV, red line). No peaks are observedin the energy spectra.

4. Measurements with X-ray sourcesTo test the detectors response to the release of energy the devices were illuminated with twocalibration sources: 55Fe (' 6 keV) and fluorescence of Al (' 1.5 keV). In order to performcoincidences study between adjacent resonators, a two-channels readout was implemented thatworks as follow. Two different probe signals LO1 and LO2, one for each microresonator, weregenerated by two microwave synthesizers and then merged together by a power coupler. Theresulting signal was used to excite the resonators. The forward transmitted waveform wasamplified with a cryogenic HEMT amplifier, mounted on the 4 K stage and, subsequently, witha room-temperature amplifier. The amplified signal was then split by a power divider in twocopies, RF1 and RF2. Homodyne mixing between each copy (RF1 and RF2) and the originalexcitation signals (LO1 and LO2) is accomplished employing two IQ-mixers. The resulting I1, Q1

and I2, Q2 signals were acquired by a commercial 14bit-simultaneous-sampling data acquisitionboard with a sample rate fs = 2.5 MHz, and projected offline into the frequency and dissipationquadrature signals. Pulses due to X-ray absorption events were easily identified in the detectoroutput timestream, as it is shown in figure 2 (left) for a multilayer composed by 8 layers ofTi/TiN (176 nm of thickness, Ti/TiN=10 nm/12 nm).

The observed pulses showed several different behaviors: slower pulses, with a singleexponential decay constant, very fast pulses decaying faster than exponential, and pulseswith two exponential decay times. We interpret the slow component as the quasi-particlerecombination time and the fast component as a result of self-recombination of quasi-particlesdue to a local excess of quasi-particles. Unfortunately, the detectors do not resolve amonochromatic energy and in the energy spectra acquired no peaks were observed (figure 2,right).

To explain this we hypothesized 3 different mechanisms. Given the thickness of the film andthe energy of the X-ray, only a small fraction of the X-rays (about 15%) are absorbed in thesuperconductor, while the remaining interact in the substrate. The events of the latter casecould give a non-zero response in the detector because of the phonons produced in the substratethat could interact in the superconductor. Since the efficiency in collecting the phonons isproportional to the solid angle under which they see the inductor, only a small fraction of

energy is detected. Besides, the energy seen by the detector has a dependence on the positionof the interaction. On the other hand, if the X-ray is absorbed in the film, two phenomenacould happened. In one hypothesis the total energy could be degraded, because of the phononsescaping from the superconductor to the substrate, while in the second hypothesis, because ofthe slow diffusion of quasiparticles in TiN, a local excess of quasiparticles in the superconductorcould happened, bringing to a local transition of the superconductor to normal metal state.

In order to study this behavior we are considering 3 different future upgrades: 1) resonatorswith the inductor line suspended from the substrate to prevent the exchange of phonons betweenthe film and the substrate; 2) resonators with the inductor coupled to a higher-gap absorber witha faster diffusion to prevent the recombination of the quasi-particles before they diffuse; 3) andultimately, thermal devices, with low Tc (around 0.5 K), where the non-equilibrium quasiparticlepopulation from photon absorption is generated by temperature change [15].

5. ConclusionTi/TiN multilayer films suitable for a next generation of experiments in the neutrino physicsfield are produced and tested with encouraging results. The characteristic parameters, such asthe Tc , the gap parameter ∆ and the quasiparticle lifetime τqp of different TiN superconductingfilms have been measured. Test with calibration sources have been performed but the energyspectra acquired have not yet shown well defined peaks. We believe this to be due to variouseffects (contact between the inductor and the substrate, slow diffusion of the quasiparticles inthe film). To understand and avoid these problems new solutions are under investigation.

AcknowledgmentsThis work is supported by Fondazione Cariplo through the project Development ofMicroresonator Detectors for Neutrino Physics (grant International Recruitment Call 2010, ref.2010-2351).

References[1] Lobashev V et al. 2001 Nucl. Phys. B. (Proc. Suppl.) 91 280–286 doi:10.1016/S0920-5632(00)00952-X[2] Kraus C et al. 2005 Eur. Phys. J. C 40 447–468 doi:10.1140/epjc/s2005-02139-7[3] Angrik J et al. 2004 Tech. Rep., Forschungszentrum, Karlsruhe, Germany 447–468 NPI ASCR Rez, EXP-

01/2005, FZKA Scientific Report 7090, MS-KP-0501[4] Drexlin G et al. 2013 Advances in High Energy Physics 1–39 doi:10.1155/2013/293986[5] Monreal B and Formaggio J A 2009 Phys. Rev. D 80(5) 051301 doi:10.1103/PhysRevD.80.05130[6] Ferri E et al. 2012 J. Low Temp. Phys. 167 1035–1040 doi:10.1007/s10909-011-0421-6[7] Nucciotti A 2012 Nucl. Phys. B. (Proc. Suppl.) 229232 155–159 doi:j.nuclphysbps.2012.09.025[8] Rujula A D and Lusignoli M 1982 Physics Letters B 118 429–434 doi:10.1016/0370-2693(82)90218-0[9] Day P K et al. 2001 Nature 425 817–821 doi:10.1038/nature02037

[10] Faverzani M, Day P K et al. 2012 J. of Low Temp. Phys. 167 1041–1047 doi:10.1007/s10909-012-0538-2[11] Faverzani M, Day P K et al. 2013 NIMA 718 492 – 494 doi:10.1016/j.nima.2012.11.060[12] Giachero A, Day P K et al. 2013 submitted to J. of Low Temp. Phys. arXiv:1307.3781 [cond-mat.supr-con][13] Vissers M R et al. 2013 Appl. Phys. Lett 102 232603 doi:10.1063/1.4804286[14] Mattis D C and Bardeen J 1958 Physical Review 111(2) 412–417 doi:10.1103/PhysRev.111.412[15] Golwala S et al. 2008 J. of Low Temp. Phys. 151 550–556 doi:10.1007/s10909-007-9687-0

Development of transition edge sensors based on Al/Ti bilayer

thin films

Qing-Ya Zhang1,2,3

,Gen-Fang He1,2,3

,Wen-Hui Dong1,2,3

,Tian-Shun Wang4,5

,Jun-

Kang Chen4,5

,Jian-She Liu1,2,3

,Tie-Fu Li1,2,3

,Xingxiang Zhou4,5

and Wei Chen1,2,3

1 Tsinghua National Laboratory for Information Science and Technology, Tsinghua

University, Beijing 100084, China 2 Department of Micro/Nanoelectronics, Tsinghua University, Beijing 100084, China

3 Institute of Microelectronics, Tsinghua University, Beijing 100084, China

4 Department of Optics and Optical Engineering, University of Science and

Technology of China, Hefei 230026, Anhui, China 5 CAS Key Laboratory of Quantum Information, University of Science and

Technology of China, Hefei 230026, Anhui, China

E-mail: weichen@tsinghua.edu.cn, xizhou@ustc.edu.cn

Abstract. We report our progress on the development of Al/Ti bilayer transition edge sensors

(TESs) for applications in millimeter and submillimeter wave detection. The bilayers were

deposited by sputtering at room temperature on silicon wafers coated with 200nm-thick low-

stress SiN, and patterned into TESs with different geometries using standard photolithography

and etching. We also etched away the silicon substrate underneath the TESs to release

suspended SiN membranes, and applied black wax coating to support and protect the structures

on the front side. Testing results show that the Al/Ti bilayers have sharp superconducting to

normal state transition with a transition width of 3-6 mK under a bias current of 10 A. The critical temperature of the bilayers can be tuned by changing the thickness of the Ti layer, and

the fabricated TESs were found to exhibit long-term stability. The dependence of the critical

temperature on bias current was studied based on Landau-Ginsburg theory and the relevant

device parameters are obtained.

1. Introduction Transition edge sensors (TESs) are superconducting thin-films biased in the transition region from the

superconducting to the normal state. The thin films work as sensitive thermometers and can measure

the temperature change after photons or particles absorption. Functioning by the principles of electro-

thermal feedback, TESs have lower noise, better linearity and faster response than semiconductor thermal detectors [1]. Thus, bolometers and microcalorimeters based on TESs have been developed for

radiation detection in a wide spectrum from millimeter waves to gamma-rays.

Presently available bolometric detectors for sub-millimeter and millimeter waves are background limited, and further improvement in experiment sensitivity can be achievable only by significantly

increasing the number of detectors in a focal plane.TES bolometers can be monolithically fabricated in

large arrays and can be readout by SQUID multiplexing circuits. It has become an attractive detection technology in sub-millimeter and millimeter waves astrophysics in the last decade. TES bolometer

arrays with hundreds to thousands of pixels

have been successfully deployed in

experiments like APEX-SZ and

POLARBEAR [2]. They are also being used or proposed to be used in other experiments,

such as BICEP2 [3] and SPTpol [4].

We are now developing TES bolometers based on Al/Ti bilayers with potential

applications in sub-millimeter and

millimeter wave astronomy. In the TES

bolometer, the metal film absorber and Al/Ti bilayer thermometer are placed on a thin

suspended SiN membrane in order to

decrease the thermal coupling to the silicon substrate, which acts as a cold bath. The SiN

membrane is supported by several SiN

beams. Because of the narrow beams, the thermal conductance is decreased and the

device sensitivity is increased. The target

critical temperature (Tc) for our TESs is 550

mK. The substrate temperature is set at 300 mK. The desired Tc can be achieved by

selecting appropriate values for the

thickness of Ti thin layers due to the proximity effect [5]. We have fabricated

Al/Ti TESs with different geometries to

study how the device performance is

affected by its geometry. The suspended SiN membranes of the devices have three

different dimensions, 0.93×0.93 mm2,

1.43×1.43 mm2 and 1.93×1.93 mm

2. We

will study and compare their thermal

properties. Figure 1 shows Al/Ti TESs with

different geometries on a 1.93×1.93 mm2

SiN membrane. In this paper, we report the

details of the fabrication process for Al/Ti-

based TESs and discuss the superconducting

properties of the devices.

2. Fabrication details

A shown in Figure 2(a), the fabrication

process begins with the growth of 0.2 µm of low-stress SiN on both sides of a 4 inch

silicon wafer that is polished on both sides.

The growth of SiN is done by low-pressure chemical vapor deposition (LPCVD). Note

that the SiN layer should have a low tensile

stress at around 300 MPa to ensure a high

yield of intact membranes. The silicon wafer has a starting thickness of 470 µm. In

the next step of the fabrication, the backside

Figure 2. Process flow diagram

Figure 1. Al/Ti TESs on a SiN membrane

SiN windows for defining the size of SiN membranes are opened using reactive ion etch (RIE), as

shown in Figure 2(b). When performing this backside etching, the SiN on the front side of the wafer is

protected by a photoresist (PR).

Before bilayer deposition, wafers with the SiN coating are cleaned in a 5:1 mixture solution of H2SO4 and H2O2 for 10 minutes and then rinsed by de-ionized water to remove surface contamination.

In a multi-target sputtering system, the Al/Ti bilayers are deposited by magnetron sputtering at room

temperature with a base pressure of 8.0×10-8 Torr (Figure 2(c)). The system has an isolated load lock

and the process chamber can be maintained at high-vacuum level for a long period of time. In the

chamber five targets are placed on the same plane around a central axis. The wafers are rotated around

the central axis over the target with a speed of 20 rotations per minute to improve uniformity during

the deposition. Baffles inside the chamber are used to prevent cross contamination. The purity of Al target is as high as 99.999% and the purity of Ti target is 99.995%. A 46 nm thick Al layer is

deposited first at a rate of 0.08 nm/s. The sputtering power and Argon gas pressure are 400 watts and 2

mTorr, respectively. Under these conditions, the deposited Al films have a Tc around 1.35 K and show low tensile stress below 100 MPa. The Ti layer is deposited after 2 minutes at a rate of 0.09 nm/s. The

Ti film thickness is controlled by the sputtering time and power.

The front side Al/Ti bilayer is patterned using double sided photolithographic techniques aligning to the backside alignments. We first selectively remove the Ti layer by sulfur hexafluorid (SF6) based

RIE, leaving it only in the area of the TES. The Al layer is etched using an Al etch

(H3PO4:CH3COOH:HNO3 100:10:1 by volume) (Figure 2(d)). When we pattern the bilayers using lift-

off, it is found that the fabricated bilayers often have bad transition characteristics. Also the superconducting properties of the bilayers degrade significantly when the bias current is relatively

large. The probable cause is the PR contamination during the lift-off process. Therefore, we pattern the

bilayers using etching methods and the bilayer show good and reproducible superconducting performance.

The niobium (Nb) leads and bonding pads are made by lift-off. This is done by patterning a PR

mask (Figure 2(e)), depositing a 120 nm thick Nb layer (Figure 2(f)) and lifting off (Figure 2(g)).

Before the Nb deposition, argon plasma cleaning is performed on the Al/Ti bilayers in order to remove the thin oxide films on the surface, making good electrical contacts.

The wafers are then etched in an KOH solution at 75°C to form suspended SiN membranes (Figure

2(h)). Ultrasonic agitation is introduced in our etching process to improve the etching uniformity and reduce the surface roughness [6].The front side structures are protected by a commercial black wax

during the wet etch. While SiN membranes can easily be damaged by ultrasonic agitation, the black

wax mechanically supports the released membrane till the end of the process and

keeps the membranes intact. The black wax

is easy to apply and it can persist long

enough in the KOH etchant. It is an ideal choice as a protective coating.

For our future TES bolometers fabrication,

perforations will be made in the suspended SiN membrane, leaving only several

suspended SiN beams to support the central

SiN island. This process step involves RIE etching SiN and removing PR in an oxygen

plasa. The geometry of SiN beams determine

the thermal conductance between

thermometers and cold bath. By designing the geometry of SiN beams, TES bolometers

with different thermal conductance and

different saturation power can be fabricated.

Figure 3. Superconducting transitions for three

different dAl/dTi: 46/5, 46/15 and 46/55 (nm)

3. Results and discussions

As expected, a decrease of Tc is observed by depositing Ti thin layers on Al thin films, due to the

proximity effect. We have fabricated Al/Ti TESs with Tc values ranging from 470 mK to 800 mK by

varying the Ti layer thicknesses. Figure 3 displays the transition curves of three selected bilayers with equal Al layer thickness (dAl) and different Ti layer thickness (dTi). The results show that the Tc and the

normal resistance (Rn) of bilayers decrease with increasing dTi.

The fabricated Al/Ti bilayers show sharp transitions with a ΔTc of 3-6 mK under a bias current of 10 µA. The bilayers grown on SiN membranes and on bulk SiN/Si on the same chip have almost equal

Tc values.We have not observed an appreciable difference between the transition curves of bilayers on

the membrane and on the bulk on the same chip. The membrane does not alter the superconducting

properties of the bilayers. Long term stability of TESs is essential for application of TESs as bolometric detectors. In

particular, the device parameters Tc, ΔTc and Rn should remain stable during the detector’s lifecycle.

Our TESs, consisting of 46 nm/60 nm Al/Ti bilayers have shown no significant change in ΔTc and Rn after several cooling and warming cycles over a period of 3 months, though a slight decrease of Tc by

2 mK from the original value of 500 mK has been observed. For TES bilayers, it has been reported

that the film oxidation can affect device parameters [7]. The probable reason for the stability of Al/Ti bilayers is that Ti layer film can form highly stable protective oxides layers on surface.

We have studied the dependence of transition curves on bias current in a wide range between 10

µA and 400 A. We measured two different TES devices on the same chip, TES A and TES B. The dimensions of Al/Ti bilayers on TES A and TES B are 100 µm × 200 µm and 50 µm × 100 µm,

respectively. The bilayers consist of 46 nm Al and 62 nm Ti. The transition curves of TES A at five different bias currents are displayed in the Figure 4(a). It shows that larger current shifts transition

curve to lower critical temperature. The transition width are 5.4 mK,4.8 mK,5.0 mK,7.3 mK and 9.2

mK, showing a slight increase as the bias current increases. We obtained similar results for TES B. The relations between critical temperature and bias current for TES A and TES B are shown in

Figure 4(b). Applying Landau-Ginsburg (GL) theory to a thin superconducting film, we can derive an

expression for the critical temperature dependence of the bias current: 2 3

0 0( ) (1 ( ) )c cT I T I I

where Tc(I) is the critical temperature with bias current I, Tc0 is the zero-current critical temperature,

and I0 is the zero-temperature critical current. The measured data of TES A and TES B fit quite well

Figure 4. (a) The dependence of transition curves on bias current for TES A, (b) The relation

between critical temperature and bias current for TES A and TES B (including the measured data

and fitting curves).

with this equation using a least square fit. We obtained Tc0 = 499.9 mK, I0 = 12.1 mA for TES A and Tc0

= 499.3 mK, I0 = 5.3 mA for TES B.

4. Conclusions

We have developed a reliable process to manufacture Al/Ti-based TES detectors and used it to fabricate devices with different geometries. Al/Ti proximity bilayers have shown long-term stability

and sharp transitions with ΔTc of 3-6 mK. The Tc of bilayers can be tuned from 470 mK to 800 mK for

different applications. The dependence of Tc on bias current is studied based on GL theory and the relevant device parameters are obtained.

Acknowledgments

This work is supported partially by the State Key Program for Basic Research of China (Grant No.

2011CBA00304), the Chinese national science foundation (Grant No.60836001, Grant No. 11273023), the Tsinghua National Laboratory for Information Science and Technology (TNLIST) Cross-discipline

Foundation and the Central Government University Fundamental Research Fund.

References [1] Irwin K D and Hilton G C 2005 Top.Appl.Phys. 99 63-78

[2] Arnold K S 2010 ph.D. Dissertation (Berkeley:University of California,Berkeley) pp 56-75

[3] Crites A T et al. 2011 IEEE Trans.Appl.Supercond. 21 184 [4] Orlando A et al. 2010 Antenna-coupled TES bolometer arrays for BICEP2/Keck and SPIDER:

Proc. Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for

Astronomy V (San Diego, California, United States, June 29-July 2, 2010) (SPIE Proceedings

vol 7741) ed Wayne S. Holland and Jonas Zmuidzinas pp77410H1-8 [5] Gennes P D 1964 Rev.Mod.Phys. 36 225-37

[6] Chen J, Liu L, Li Z, Tan Z, Jiang Q, Fang H, Xu Y and Liu Y 2002 Sens.Actuators A 96 152-6

[7] Zieger G, Anders S, Bone H, Dellith J, Born Detlef, May T and Meyer H-G 2012 Supercond.Sci.Technol. 25 125005

A low-temperature device architecture for the

statistical study of electrical characteristics of 256

quantum devices

H Al-Taie1,2, L W Smith2, B Xu2, P See3, J P Griffiths2, H E Beere2,G A C Jones2, D A Ritchie2, M J Kelly1,2 and C G Smith2

1 Centre for Advanced Photonics and Electronics, Electrical Engineering Division,Department of Engineering, 9 J. J. Thomson Avenue, University of Cambridge, CambridgeCB3 0FA, United Kingdom2 Cavendish Laboratory, Department of Physics, University of Cambridge, J. J. ThomsonAvenue, Cambridge CB3 0HE, United Kingdom3 National Physical Laboratory, Hampton Road, Teddington, Middlesex TW11 0LW, UnitedKingdom

E-mail: ha322@cam.ac.uk, lws22@cam.ac.uk

Abstract. Research in the field of low-temperature electronics is limited by the small numberof electrical contacts available on cryogenic set ups. This not only restricts the number ofdevices that can be fabricated, but also the device and circuit complexity. We present an on-chipmultiplexing technique which significantly increases the number of devices locally measurableon a single chip, without the modification of existing fabrication or experimental set-ups. Wedemonstrate the operation of the multiplexer by performing electrical measurements of 256quantum wires formed by split-gate devices using only 19 electrical contacts on a cryogenicset-up. The multiplexer allows the measurement of many devices and enables us to performstatistical analyses of various electrical features which exist in quantum wires. We use thisarchitecture to investigate spatial variations of electrical characteristics, and reproducibility ontwo separate cooldowns. These statistical analyses are necessary to study device yield andmanufacturability, in order for such devices to form the building blocks for the realisation ofquantum integrated circuits. The multiplexer provides a scalable architecture which makes awhole series of further investigations into more complex devices possible.

Cryogenic experimental set ups limit the complexity of low-temperature electronics by thesmall number of electrical wires available on a cryostat. This also restricts the number of devicesthat can be fabricated on a single chip, and multiple cooldowns of many samples are necessaryfor electrical characterisation of devices [1]. Not only is this extremely time consuming andcostly, but variations between cooldowns may lead to misleading characterisation of electricalfeatures due to changes in the electron density.

For devices to be considered for applications in spintronics [2] or quantum informationprocessing [3], one must consider reliability, yield and manufacturability/reproducibility. Onedevice which has been identified as having great potential in these applications [4, 5] is thesplit-gate transistor [6]. The split gate is an ideal component to investigate because it can beseen as a building block for more complex devices (such as laterally-defined quantum dots). Itis also the simplest device to exhibit quantum phenomena; the quantisation of conductance in

Figure 1. (a) Schematic diagram of a two-way quantum multiplexer. The mesa (blue) branchesfrom input contact V to eight separate output paths, labelled 1 - 8. Addressing gates are shownin yellow (off) and white (on), and insulating regions are shown in grey. Negative voltagesapplied to addressing gates G2, G4 and G6 route the input voltage along the path indicatedby the arrow, to output 1. (b) Optical micrograph of the device, consisting of two multiplexersaddressing an array of 256 split gates. The contacts for the source, drain and input-voltage arelabelled S, D and V, respectively. (c) Schematic diagram of a three-way quantum multiplexer.(d) Conductance against Vsg for an example split gate, in which Vp is indicated by the arrow.

units of 2e2/h as electrons become confined in a one-dimensional (1D) channel [7, 8].We presented the first study of a large number of individually-addressable, gate-defined

quantum devices on a GaAs/AlGaAs heterostructure [9]. Separate measurement of each devicewas enabled using an on-chip increase in number of electrical contacts: A quantum multiplexer.The multiplexer addresses an array of 256 geometrically-identical split gates. Since manydevices can be measured during a single cooldown in a cryostat, a study of the fabrication andquantum yield can be performed [9]. Additionally, statistical variations in quantum phenomenacan be investigated [10]. The focus of this present study is the reproducibility of electricalcharacteristics, both from device-to-device and after thermal cycling.

Devices were fabricated on a modulation-doped GaAs/AlGaAs heterostructure, in which thetwo-dimensional electron gas (2DEG) forms 90 nm below the surface of the wafer. Data arepresented before and after illumination with a light emitting diode. The carrier density (n) andmobility (µ) were measured to be 1.7×1011 cm−2 and 0.94×106 cm2V−1s−1, respectively, beforeillumination. After illumination, n = 2.9 × 1011 cm−2 and µ = 2.2 × 106 cm2V−1s−1. Standardoptical lithography was used to define the surface gates, and the split gates (0.4 µm long and 0.4µm wide), were patterned using electron-beam lithography. Two-terminal lock-in measurementswere performed at 1.4 K using excitation voltages between 25 and 100 µV at 77 Hz.

Figure 1(a) illustrates the operation of the multiplexer. A raised ‘mesa’ which confines the2DEG branches from the contact labelled V , to which an input voltage (Vi) is applied. Theoutput path through the multiplexer is determined by 6 ‘addressing gates’ (G1 to G6). Thesegates cover multiple mesa-defined branches of the multiplexer; in Fig. 1(a) gate G4 crosses

Figure 2. Colour scale showing the spatial distribution of Vp normalised to Vp(µ) (a) beforeillumination and (b) after illumination, where each box represents a split gate in the array.Clear boxes indicate split gates which did not define a 1D channel, therefore Vp could notbe determined. (c) Histogram of Vp before illumination, for a bin size of 10 mV. The dataare categorised according to whether disorder effects were or were not evident, represented byshaded and solid bars, respectively.

the mesa at points A, B, and C. A photo-definable insulator (polyimide HD4104), is locatedunderneath the addressing gates at selected locations, for example at B. Therefore, a negativevoltage applied to G4 is sufficient to deplete the 2DEG at A and C, but not B. To direct Vialong path 1 (as illustrated by the arrow), only gates G2, G4 and G6 are ‘on’, i.e. a negativevoltage is applied. The Vi can be directed to any of the other output paths depending on whichaddressing gates are ‘on’ or ‘off’. A voltage offset (Vo) must be maintained between Vi and theaddressing gates that are ‘on’, where Vo is the voltage which must be applied to each addressinggate to deplete electrons in the 2DEG when Vi = 0. Addressing gates which are ‘off’ are heldat the same voltage as Vi.

Figure 1(b) shows an optical micrograph of the chip, in which the source, drain, and contactto which the split-gate voltage (Vsg) is applied are labelled S, D, and V, respectively. The splitgates are arranged in a rectangular array, and each device is measured individually, using twomultiplexers (the left multiplexer selects the row, and the top multiplexer selects the column).The operation of the chip is described more fully in Ref. [9].

As a proof-of-concept, Fig. 1(c) shows a schematic diagram of a three-way multiplexer, whichgives a 50% increase in the number of output paths while maintaining the same number ofelectrical contacts (i.e. 6 addressing gates and 1 input contact). The operation is identical tothe two-way multiplexer except that if the desired path is through either outer mesa channelsD or F , the voltage applied to addressing gates G2 or G1 must be large enough to deplete the2DEG beneath the insulator at E. However, if the required path is through the central channel(E), a voltage is applied to both G1 and G2 such that the 2DEG is depleted at D and F ,but not at E. This provides a scalable architecture that can be extended to address an evenlarger number of devices on chip. The number of output paths from the two-way (N2mux) andthree-way (N3mux) multiplexers are given by N2mux = 2(n−1)/2 and N3mux = 2(n−1)/2 +2(n−3)/2,respectively, where n is the number of electrical contacts (including the addressing gates andinput contact). In either design, adding two additional addressing gates doubles the number ofoutput paths.

The conductance (G) through each split gate was measured individually, as a function ofVsg. Due to damage that occurred during fabrication, fifteen split gates did not define a 1Dchannel, leaving 241 usable devices. This gives a yield of 94%, which may be improved withcareful manufacture. Figure 1(d) shows conductance G as a function of Vsg for an example split

Figure 3. (a) Scatter plot of Vp from each split gate on two separate cooldowns; Vp1 and Vp2.Categories A, B and C refer to whether the devices showed evidence of disorder on both, neither,or only one cooldown. As a guide, the dotted line shows Vp1 = Vp2. (b), (c) Conductance asa function of Vsg for two example split gates. In each panel the conductance traces from bothcooldowns are shown. In (c) the arrow marks a weakened plateau, evidence of disorder on thesecond cooldown.

gate (before illumination). The pinch-off voltage (Vp) is defined as the voltage at which G = 0,marked by the arrow.

Figures 2(a) and (b) show colour-scale plots of Vp as a function of spatial distribution beforeand after illumination, respectively. The Vp is normalised to the mean pinch-off voltage [Vp(µ)],and the range of the colour bars in (a) and (b) are identical, to allow direct comparison. Eachbox represents a split gate in the array, and clear boxes indicate the 15 split gates where Vp couldnot be determined. From Fig. 2(b) it can be seen that the normalised variation of Vp reducesafter illumination. Prior to illumination Vp(µ) = −0.76 V, and standard deviation σ = 97 mV.Following illumination, Vp(µ) = −2.86 V (due to the increased n), and the absolute value of σincreased to 144 mV. However, there is a reduction in the coefficient of variation (defined as adimensionless ratio of the standard deviation to the absolute of the mean), from 0.127 to 0.054,such that a higher n appears to lead to greater uniformity of Vp.

Quantum phenomena in mesoscopic devices can be affected by disorder. Two main sourcesof disorder in GaAs HEMTs are scattering from ionised donors and ionised impurities. In 1Dmeasurements, this can lead to deviations in the quantised conductance from 2e2/h, weakenedor missing plateaux, and evidence of Coulomb blockade [14].

Figure 2(c) shows a histogram of Vp (before illumination), for a bin size of 10 mV. Thedata are categorised according to whether disorder effects were observed at any point on theconductance trace. Before illumination, 70 (171) devices did not (did) show evidence of disorder(in this measurement T = 1.4 K, at a lower T more disorder effects may become apparent). Therange of Vp is smaller where disorder was not observed (≈ 0.3 V), and the standard deviationσ = 67.3 mV compared to 104.7 mV. The total distribution in Fig. 2(c) is perhaps bi-modal, forwhich an explanation has yet to be determined.

Certain disorder effects (such as missing or weakened plateau), were much less evidentafter illumination, since the higher electron density results in better screening of impurities.However, after illumination almost all devices showed occasional length resonant-like features inthe conductance, perhaps due to the stronger 1D confining potential [15].

The sample was thermally cycled and measured on a second cooldown. Figure 3(a) showsa scatter plot of Vp2 against Vp1, where Vp1 (Vp2) is Vp from the first (second) cooldown. Anadditional split gate did not define a 1D channel on the second cooldown, leaving a total of240 devices. The data are categorised according to whether disorder was observed on both

cooldowns (67.9%), on neither cooldown (24.2%), or only on one cooldown (7.9%). Therefore,whether disorder effects were or were not observed did not change for 92.1% for devices. ThePearson-product moment correlation coefficient r quantifies the degree of correlation betweentwo variables. For Fig. 3(a) r = 0.89 (including all data), indicating a very strong correlation.The mean change in pinch-off voltage δVp(µ) = 39.2 mV, which may be related to a change inn2D on successive cooldowns. A large δVp (> 0.15 V), occurred for only 7 devices, irrespectiveof whether disorder effects were observed. Since Vp is dependent on electron density, a large δVpmay reflect local variations in n due to a redistribution of ionised donors in the dopant layer.

By comparing conductance data from successive cooldowns it may be possible to distinguishbetween likely sources of disorder. Figure 3(b) shows G as a function of Vsg for a device in whichthe conductance characteristics are highly disordered. Despite the disorder, the conductance isremarkably similar between cooldowns. The strength of the effect, and the reproducibility onsuccessive cooldowns suggests that an ionised impurity exists near the 1D channel.

Figure 3(c) shows G against Vsg for another device, where the arrow indicates a weakenedplateau on the second cooldown. Such cooldown-dependent effects may be related to remoteionised donors, since the distribution of donors that are ionised varies on thermal cycling.Additionally, the effect is relatively weak, consistent with the origin of the disorder being furtheraway from the 1D channel (i.e. separated from the 2DEG by a 40 nm wide spacer layer).

In summary: the reproducibility of electrical characteristics in 1D devices has beeninvestigated, as a function of spatial location and after thermal cycling. This was made possibleusing a scalable on-chip multiplexer, which allowed 256 split gates to be individually measuredduring a single cooldown. This approach can be used to determine limiting factors on devicereproducibility, and identify where to direct efforts for improvement. In the present study,disorder, variations in the density, and the re-distribution of dopant ions on the second cooldownwere all limiting factors. Improvements can be made by fabricating devices on a higher qualityGaAs/AlGaAs heterostructure (quantified by a greater µ), which will likely result in smallervariations of Vp between devices, and fewer observable disorder effects. Additionally, dopantscan be removed entirely by fabricating devices on undoped heterostructures [16]. It may alsobe instructive to investigate how variations in density can be reduced by controlling the coolingprocess, and by illuminating the sample during the cooldown. This is an initial study, whichhas presented a framework for considering reproducibility in quantum devices. It can be usedto test the suitability of devices for applications in quantum computing and spintronics.

AcknowledgmentsThis work was supported by the Engineering and Physical Sciences Research Council Grant No.EP/I014268/1. The authors would like to thank C. J. B. Ford, I. Farrer and F. Sfigakis forinvaluable discussions, and R. D. Hall for electron-beam exposure.

References[1] Q.-Z. Yang, M. J. Kelly, I. Farrer, H. E. Beere, and G. A. C. Jones, Appl. Phys. Lett. 94, 033502 (2009).[2] D. D. Awschalom, and M. E. Flatte, Nat. Phys. 3, 153 (2007).[3] D. Loss, and D. P. DiVincenzo, Phys. Rev. A 57, 120 (1998).[4] T.-M. Chen, M. Pepper, I. Farrer, G. A. C. Jones, and D. A. Ritchie, Phys. Rev. Lett. 109, 177202 (2012).[5] P. Debray, S. M. S. Rahman, J. Wan, R. S. Newrock, M. Cahay, A. T. Ngo, S. E. Ulloa, S. T. Herbert, M.

Muhammad, and M. Johnson, Nat. Nanotechnol. 4, 759 (2009).[6] T. J. Thornton, M. Pepper, H. Ahmed, D. Andrews, and G. J. Davies, Phys. Rev. Lett. 56, 1198 (1986).[7] B. J. van Wees, H. van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwenhoven, D. van der Marel,

and C. T. Foxon, Phys. Rev. Lett. 60, 848 (1988).[8] D. A. Wharam, T. J. Thornton, R. Newbury, M. Pepper, J. Ahmed, J. E. F. Frost, D. G. Hasko, D. A. Ritchie,

and G. A. C. Jones, J. Phys. C 21, L209 (1988).[9] H. Al-Taie, L. W. Smith, B. Xu, P. See, J. P. Griffiths, H. E. Beere, G. A. C. Jones, D. A. Ritchie, M. J.

Kelly, and C. G. Smith, Appl. Phys. Lett. 102, 243102 (2013).

[10] L. W. Smith, H. Al-Taie, B. Xu, F. Sfigakis, P. See, J. P. Griffiths, H. E. Beere, G. A. C. Jones, D. A. Ritchie,M. J. Kelly, and C. G. Smith, submitted to PRB.

[11] D. V. Lang, R. A. Logan, and M. Jaros, Phys. Rev. B 19, 1015 (1979).[12] J. G. Williamson, C. E. Timmering, C. J. P. M. Harmans, J. J. Harris, and C. T. Foxon, Phys. Rev. B 42,

7675 (1990).[13] K. J. Thomas, J. T. Nicholls, M. Y. Simmons, M. Pepper, D. R. Mace, and D. A. Ritchie, Phys. Rev. Lett.

77, 135 (1996).[14] C.-T. Liang, I. M. Castleton, J. E. F. Frost, C. H. W. Barnes, C. G. Smith, C. J. B. Ford, D. A. Ritchie, and

M. Pepper, Phys. Rev. B 55, 6723 (1997).[15] G. Kirczenow, Phys. Rev. B 39, 10452 (1989).[16] B. E. Kane, L. N. Pfeiffer, K. W. West, and C. K. Harnett, Appl. Phys. Lett. 63, 2132 (1993).

Multi-channel high frequency signalling at ultra-low

temperatures for quantum computing applications

Morteza Erfani, Anthony Matthews, Graham Batey, Steve Chappelland Michael Cuthbert

Oxford Instruments Omicron NanoScience, Tubney Woods, Abingdon, Oxon OX13 5QX, UK

E-mail: morteza.erfani@oxinst.com

Abstract. We report on the development of an ultra-low temperature platform forexperimentation on multiple solid-state qubit devices. As the number of qubits in a solid-statestructure scales up, the cryogenic platform for the device operation faces serious technologicalchallenges. A critical demand is the need for a large number of signals to enable qubit operation.This calls for a high cooling power in the sample area as well as efficient thermal anchoring toavoid thermal fluctuations induced by the room temperature electronics. We have developedsuch a platform in a cryogen-free dilution refrigerator integrated with high density wiring,including DC and high frequency RF lines. These lines are thermally anchored at all coolingstages of the fridge to ensure minimum heat transfer from room temperature to the mixingchamber. We have also incorporated a rapid sample exchange mechanism to allow samplecooling from room temperature to ∼10 mK within 8 hours without warming up the entirecryostat. This is of particular advantage for systems integrated with large superconductingmagnets as the long thermal cycle of the large mass of the magnet is avoided, saving timeand reducing the running cost. The loading system also contains wiring for grounding, sampleconditioning and protection against static discharge.

1. IntroductionSince being proposed in the 1980s, the field of quantum information processing (QIP) has grownat a fast pace and has developed from a theoretical concept into the practical realisation ofcomputation with a handful of qubits. Achieving a suitable environment for a given physicalsystem in order to reveal quantum behaviour entails significant technical challenges. Suchenvironments are typically realised under extreme conditions such as the extremely small, theextremely cold and those subjected to extreme magnetic fields. In recent years, the integratedcryogen-free dilution refrigerator and magnet platform has developed into a workhorse forquantum computation as it provides such extreme conditions with excellent usability. Asopposed to the liquid helium bath in a traditional ”wet” fridge, a dry system utilises a two-stage mechanical cooler, known as a pulse tube refrigerator, to provide a cold environment ofaround 4 K. Cryogen-free fridges have several advantages compared to their wet counterparts,including more experimental access, simplified user operation and safety as there is no liquidhelium in the system to be handled. This paper discusses a cryogen-free platform by OxfordInstruments [1], the Triton dilution refrigerator (Fig.1), which, despite being a general purposeplatform, is employed in a variety of QIP projects ranging from highly respected academicprogrammes to the commercial realisation of quantum computation. Here we outline specific

Figure 1. Photograph of a Triton cryogen-free dilution refrigerator showing the top plate(room temperature), 70 K plate, 4 K plate,still plate, 100 mK plate and the mixingchamber plate.

Figure 2. Photograph of a Triton refriger-ator showing high density RF wiring and in-line attenuators between the still and 100 mKplates useful for QIP applications.

technology enhancements to the Triton mainly targeted at QIP applications. These include thehigh density RF wiring (Fig.2) and its integration into sample exchange methods known as thetop loading (TL) and bottom loading (BL) load-locks (Fig.3). Having access to high frequencysignals at the sample area with a minimal heat load is vital for QIP instrumentation in orderto enable qubit manipulation and read-out. The high density of the RF lines is particularlyattractive for multi-qubit operations as it complies with the DiVincenzo criteria for scaling upthe number of qubits for general purpose quantum computing [2].

2. Qubit scale-up at ultra-low temperatureThe integration of high frequency RF lines coupled with the quick TL and BL sample turnaroundmethods are solutions that make the Triton dilution refrigerator a customised platform for QIP.The novelty of the BL/TL design is apparent from its method of cooling a so-called samplepuck from room temperature to the mK temperature range in under 8 hours. This is achievedby directly loading the puck into an operating fridge running at sub-Kelvin temperatures. Themass of the puck is designed to be at an optimum so that it is small enough to deliver a fastcool down time and, at the same time, large enough to provide ample space for the sample, DCand RF wiring.

Depending on the TL or BL option, a gate valve is mounted on the top or bottom of thefridge along with a vacuum lock that allows a sample holder to be loaded onto the cold fridgevia a probe without disrupting the integrity of the insulating vacuum. The system is designedso that the sample puck can be thermally anchored to a platform that is at the centre of themagnetic field. This platform, known as the docking station, is in intimate thermal contactwith the mixing chamber. When the loading operation is complete, the probe can be retracted,and rotatable flaps close off the central access to protect the system from thermal radiation.Owing to this novel flap design, the probe retraction is of particular advantage in the system as

Figure 3. Schematic views of atop loading (left) and a bottomloading (right) magnet-integratedTriton refrigerator.

it eliminates any additional heat load from room temperature through the probe. As the probeis retracted the sample puck is left in the cryostat, thermally anchored to the mixing chamberand electrically connected to the experimental wiring installed on the large plates of the dilutionrefrigerator. This sample handling method is in contrast with traditional top or bottom loadingmechanisms where the loading probe must remain attached to the cryostat, which affects thecool-down time and the base temperature as well as the quantity and thermal performance ofany wiring payload. In such systems installing the cold electronics for QIP applications is veryinefficient as the experimental wiring has to be fitted down the small diameter probe, not downthe cryostat.

The sample puck and the docking station have counterpart interfaces, both mechanically andelectrically, which allow the puck to plug in and out of a cold fridge. Alignment pins on thesample puck interface ensure that it mates reliably with the docking station (its counterpartinside the fridge). M4 threads are tightened to 5 Nm via drive rods on the loading probe, andtherefore clamp the sample puck inside the fridge before the loading probe is retracted. There isfinite thermal resistance at any demountable screw-fastened joint. As reported by Okamoto etal. [3], increasing the effective contact area and removing the oxide layer on the copper surfacewill lower this thermal resistance. This can effectively be achieved by gold plating the contactsurface. Therefore the thermal contact, mating surface to the dilution refrigerator, is through agold plated copper to copper connection.

The mass of the sample puck is kept to a minimum to minimise the cool down time. Aradiation shield cooled to the temperature of the mixing chamber minimises the radiationimpinging on the sample. This shield has a dual purpose as it protects the sample when thesample holder is loaded into the fridge and also provides a thermal shield to minimise theelectron temperature for experiments after the sample holder is loaded. Electrically accessiblefrom room temperature, the DC and RF lines both terminate on the docking station wherethe puck mates and provides contact with the sample. The integration of large numbers of RFinterconnects operating at mK temperatures in high magnetic fields within physical space, heatload and electrical noise limits is particularly demanding. Compatibly with the BL/TL samplehandling, we have developed a modular platform using blind-mate connection methods for bothRF and DC signals. In particular, the blind-mate RF interconnects allow for a certain level ofmisalignment in both axial and radial directions without compromising the RF performance,which makes these connectors particularly well suited for the blind-mating BL/TL method ofpuck loading.

3. Experimental advantages3.1. Cool down timeDespite significant advantages, using a mechanical cooler for the first stage cooling in a cryogen-free system has a known drawback: limited cooling power. As a result, the initial cool downtime is long, especially if the system is integrated with a large magnet. Therefore, in a basicfridge, the sample exchange process can take over 72 hours before a new sample runs at basetemperature. It also demands manpower to disassemble the tail sets, change samples, rewire andthen reassemble the tails, and each of these processes provides an opportunity to introduce a faultinto the system. The TL/BL method resolves this issue by enabling sample loading into a coldfridge in an easy-to-use process with fast turnaround. This is particularly attractive for the QIPexperimentation, as it enables faster iterations of the multi-qubit device design-fabrication-testcycle.

3.2. High density RF signallingHigh density RF signalling is a requirement for scaling up the number of qubits. The RFperformance is evaluated by measuring the scattering parameter of individual lines using aVector Network Analyser. This gives us an attenuation expression of a transmission coefficient,the S12 parameter. As a purpose-built platform for QIP applications, such a measurementcharacterises the system by providing a quantitative description of the sample area as anenvironment. Combining this with the TL/BL feature makes the system even more attractiveas a ready-for-measurement instrument with all the DC and RF lines pre-wired and terminatedat the docking station.

3.3. Device safetyWhether for QIP applications or solely for characterisation purposes, the devices measured atmK range are generally susceptible to static charge. The TL/BL loading probe is fitted withDC connectors that mate with the sample puck. This allows us to bias and ground the sampleduring the loading process. This feature eliminates the risk of static discharge damaging thesample as it is connected to a reference potential.

4. ConclusionIn 1995 DiVincenzo showed the universality of double-qubit gate operations [4]. This is believedto be a breakthrough towards the physical realisation of a quantum computer with a largescale of operating qubits. However, providing a large number of control signals in extremeQIP environments remains an ongoing challenge. The magnet integrated cryogenic platformdiscussed in this paper is designed to address this experimental demand as a scalable environmentfor accommodating multi-qubit modules. This is clearly an ongoing development to tackle theincreasing demand of the community to provide even larger numbers of control signals in asmaller footprint. Due to scalability and the simplicity in operation we believe that the cryogen-free dilution refrigerator discussed here can provide a suitable environment for a large scale andoperational quantum computer.

References[1] Batey G, Buehler M, Cuthbert M, Foster T, Matthews A, Teleberg G and Twin A 2009 Cryogenics 49 727 –

734[2] DiVincenzo D P 2000 Fortschritte der Physik 48 771–783[3] Okamoto T, Fukuyama H, Ishimoto H and Ogawa S 1990 Review of Scientific Instruments 61 1332–1334[4] DiVincenzo D P 1995 Phys. Rev. A 51(2) 1015–1022

LIST OF PARTICIPANTS

Nom Prénom Organisme Pays Email

1 AL-TAIE Haider University of Cambridge UK ha322@cam.ac.uk

2 ANIEL Frédéric Université Paris Sud, Orsay France frederic.aniel@u-psud.fr

3 ANTHORE Anne CNRS-LPN France anne.anthore@lpn.cnrs.fr

4 ARCAMBAL Christian CEA-Orme des Merisiers France christian.arcambal@cea.fr

5 BALESTRA Francis IMEP-LAHC, Grenoble INP-CNRS France balestra@minatec.inpg.fr

6 BARBERA Marco Università degli Studi di Palermo - Dipartimento di Fisica

Italy marco.barbera@unipa.it

7 BOUCHIER Daniel CNRS/IEF France daniel.bouchier@u-psud.fr

8 BOULADE Olivier CEA/Service d’Astrophysique France olivier.boulade@cea.fr

9 BOUNAB Ayoub CEA Saclay France bounab31@gmail.com

10 CASSINA Lorenzo University and INFN of Milano-Bicocca Italy loricass88@gmail.com

11 CATALFAMO Andrea Oxford Instruments Omicron Nanoscience UK andrea.catalfamo@oxinst.com

12 CATARINO Isabel FCT/UNL Portugal isacat@fct.unl.pt

13 CAVANNA Antonella CNRS-LPN France antonella.cavanna@lpn.cnrs.fr

14 CHIODI Francesca CNRS/IEF France Francesca.chiodi@u-psud.fr

15 COIFFARD Grégoire Institut de radioastronomie millimétrique France coiffard@iram.fr

16 COURAUD Laurent CNRS/LPN France laurent.couraud@lpn.cnrs.fr

17 DARSON David ENS Paris (LPA) France david.darson@lpa.ens.fr

18 DE LA BROÏSE Xavier CEA/DSM-IRFU France labroise@cea.fr

19 DE MARCILLAC Pierre CNRS/IAS France marcillac@ias.u-psud.fr

20 DÉBARRE Dominique CNRS/IEF France Dominique.debarre@u-psud.fr

21 DECOURCELLE Tanguy Laboratoire Astroparticule et Cosmologie France tanguy.decourcelle@apc.univ-paris7.fr

22 DELISLE Cyrille CEA Gif Sur Yvette France cyrille.delisle@cea.fr

23 DENIS Anne Laboratoire P.Aigrain France anne.denis@lpa.ens.fr

24 DONG Quan CNRS-LPN France quan.dong@lpn.cnrs.fr

25 EDWARDS Megan University of Cambridge UK me321@cam.ac.uk

26 ERFANI Morteza Oxford Instruments UK morteza.erfani@oxinst.com

27 ESMAIL Adam University of Cambridge UK ae287@cam.ac.uk

28 FEVE Gwendal LPA/ENS Paris France gwendal.feve@lpa.ens.fr

29 FUXA Étienne Commissariat à l'énergie atomique et aux énergies alterna

France etienne.fuxa@cea.fr

30 GANDOLFO Nicolas Institut de Physique Nucléaire France gandolfo@ipno.in2p3.fr

31 GENNSER Ulf CNRS-LPN France ulf.gennser@lpn.cnrs.fr

32 GHRIBI Adnan Laboratoire AstroParticule et Cosmologie France ghribi@in2p3.fr

33 GIACHERO Andrea University and INFN of Milano-Bicocca Italy Andrea.Giachero@mib.infn.it

34 HIBI Yasunori NanJing University, Research Institute of Superconductor Electronics

China yasu0318@nju.edu.cn

35 HOREAU Benoît CEA Saclay France bhoreau@cea.fr

36 IHMIG Frank Fraunhofer IBMT Germany frank.ihmig@ibmt.fraunhofer.de

37 JIN Yong CNRS/LPN France yong.jin@lpn.cnrs.fr

38 JUILLARD Alexandre Institut de Physique Nucléaire de Lyon France alexandre.juillard@ipnl.in2p3.fr

39 KALTENBACHER Thomas CERN Norway thomas.kaltenbacher@uib.no

40 KIRALY Sandor RCAES HAS (Konkoly Observatory) Hungary skiraly@konkoly.hu

41 KUZMIN Leonid Chalmers University Sweden kuzmin@chalmers.se

42 LAMBERT Nicholas University of Cambridge UK nl249@cam.ac.uk

43 LE COGUIE Alain CEA/DSM-IRFU France lecoguie@cea.fr

44 LE PENNEC Jean CEA Saclay France Jean.le-pennec@cea.fr

45 LEONE Bruno European Space Agency UK bruno.leone@esa.int

46 LO CICERO Ugo Istituto Nazionale di Astrofisica Italia locicero@astropa.inaf.it

47 LUGIEZ Francis CEA/IRFU France francis.lugiez@cea.fr

48 MALNOU Maxime CNRS ESPCI LPEM France maxime.malnou@espci.fr

49 MANEVAL Jean-Paul Laboratoire Pierre Aigrain France maneval@lpa.ens.fr

50 MARTIGNAC Jérome CEA/DSM-IRFU France jerome.martignac@cea.fr

51 MATSUO Hiroshi National Astronomical Observatory of Japan Japan h.matsuo@nao.ac.jp

52 MOREAU Vincent CEA Saclay - Service d’Astrophysique France vincent.moreau@cea.fr

53 NAGASE Koichi Japan Aerospace Exploration Agency Japan nagase@ir.isas.jaxa.jp

54 NAVICK Xavier CEA/IRFU France navick@cea.fr

55 NONES Claudia IRFU - CEA Saclay France claudia.nones@cea.fr

56 OSMAN Djelal University of Cambridge UK d.osman@mrao.cam.ac.uk

57 PERBOST Camille APC - Université Paris Diderot France camille.perbost@apc.univ-paris7.fr

58 PIAT Michel APC - Université Paris Diderot France piat@apc.univ-paris7.fr

59 PIGOT Claude CEA/DSM/Irfu/Service d'Astrophysique France claude.pigot@cea.fr

60 PRELE Damien Laboratoire APC France damien.prele@apc.univ-paris7.fr

61 RAYET Rémi CALLISTO France remi.rayet@callisto-space.com

62 REVERET Vincent CEA Service d’Astrophysique France vincent.reveret@cea.fr

63 RODRIGUES Matias CEA/ LNHB France matias.rodrigues@cea.fr

64 RODRIGUEZ Louis CEA Saclay France louis.rodriguez@cea.fr

65 ROSTICHER Michael Laboratoire Pierre Aigrain France michael.rosticher@ens.fr

66 SADOULET Bernard University of California USA sadoulet@berkeley.edu

67 SAUVAGEOT Jean-Luc CEA/DSM/Irfu/Service d'Astrophysique France jean-luc.sauvageot@cea.fr

68 SCHOUTEN Raymond Delft University The Netherlands

r.n.schouten@tudelft.nl

69 SMITH Luke Cavendish Laboratory, Department of Phyiscs, University of Cambridge

UK lws22@cam.ac.uk

70 STAHL Stefan Stahl-Electronics Germany develop_group@web.de

71 TARTARI Andrea Laboratoire APC - PCCP France tartari@apc.univ-paris7.fr

72 TORRES Lidia CEA Service d’Astrophysique France lidia.torres@cea.fr

73 TREUTTEL Jeanne Observatoire de Paris France jeanne.treuttel@obspm.fr

74 VISTICOT Francois CEA Orme des Merisiers France francois.visticot@cea.fr

75 WADA Takehiko Japan Aerospace Exploration Agency Japan wada@ir.isas.jaxa.jp

76 YANICHE Jean-françois Institut de Physique Nucléaire France yaniche@ipno.in2p3.fr

77 ZEROUNIAN Nicolas Université Paris-Sud - IEF France nicolas.zerounian@u-psud.fr

78 ZHANG Qingya Tsinghua University China qy-zhang08@mails.tsinghua.edu.cn