IEEE International Workshop on Advances in Sensors ...

2 downloads 14077 Views 22MB Size Report
were calculated for standard 1 M Na+ using the program ...... accelerometers”, ASME Journal of Applied Mechanics,. 42, pp. ...... The car's mechanics itself,.
Proceedings

2nd IEEE International Workshop on Advances in Sensors and Interfaces

2007

26-27 June 2007 Bari, Italy

Copyright ©2007 by the Institute of Electrical and Electronics Engineers Inc All rights reserved. Copyright and Reprint Permission: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy beyond the limit of U.S. copyright law for private use of patrons those articles in this volume that carry a code at the bottom of the first page, provided the per-copy fee indicated in the code is paid through Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923. For other copying, reprint or republication permission, write to IEEE Copyrights Manager, IEEE Operations Center, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855-1331. All rights reserved. Copyright ©2007 by the Institute of Electrical and Electronics Engineers Inc

IEEE Catalog Number: 07EX1794 ISBN: 1-4244-1245-5 Library of Congress: 2007925367 © 2007 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

Contents Foreword Organizing Committee Session IA: Advances in Sensors and Interfaces C. Van Hoof, R. F. Yazicioglu, T. Torfs, P. Merken: “Ultra-low power Biopotential Interfaces and their application in wearable and implantable systems” R. Thewes: “CMOS Chips for Bio Molecule Sensing Purposes” P. D. Franzon: “Molecular Electronic Circuits”

Session IB: Bio-sensing for health care L. Rothberg, H. Li: “Detection of Specific Nucleotide Sequences using Electrostatic Interactions of DNA with Gold Nanoparticles” M. Lanzoni, B. Riccò, G. De Cesare: “Smart Sensors for Fast Biological Analysis” G.M. Farinola, L. Torsi, F. Naso, P.G. Zambonin, L. Valli, M.C. Tanese, O. Hassan Omar, F. Babudri, F. Palmisano: “Chemical design, synthesis and thin film supramolecular architecture for advanced performance chemo- and bio-sensing organic field effect transistors” G. Costantini, M. Carota, G. Maccioni, D. Giansanti: “A New Integrated Kinematic Sensor for the Classification of Sit-to-Stand Locomotion Task” E. Ieva, K. Buchholt, L. Colaianni, N. Cioffi, I. D. van der Werf, A. Lloyd Spetz, P.O. Käll, L. Torsi: “Gold Nanoparticle Sensors For Environmental Pollutant Monitoring” A. Fort, C. Lotti, M. Mugnaini, R. Palmerini, R.Palombari, S. Rocchi, L.Tondi, V.Vignoli: “A Two Electrode C - NiO Nafion® Amperometric Sensor for NO2 Detection” D. Caputo, G. de Cesare, C. Manetti, A. Nascetti, R. Scipinotti: “Chromatographic System Based on Amorphous Silicon Photodiodes” P. Delizia, S. D’Amico, A. Baschirotto: “A Readout Circuit in 0.35 μm CMOS Technology for Lab-ona-Chip Applications” V.M.N. Passaro, B. Casamassima, F. De Leonardis, F. Dell’Olio, F. Magno: “Modeling and Design of a Microdisk Photonic Sensor for Biological Applications” L. Abbati, P. Placidi, A. Scorzoni, M. Lanzoni: “A Configurable Architecture for the Detection of DNA Sequences based on a E2PROM device” M. Barbaro, A. Caboni, D. Loi: “A CMOS Integrated DNA-chip for hybridization detection with digital output” D. De Venuto, G. Indiveri, A. Valentini: “Design of an Integrated Low-Noise Read-Out System for DNA Capacitive Sensors”

Session IC: MEMS and Sensor Networks E.M. Yeatman: “Energy Scavenging for Wireless Sensor Nodes” A. Flammini, P. Ferrari, D. Marioli, E. Sisinni, A. Taroni: “Sensor networks for industrial applications” G. Costantini, M. Todisco, M. Carota, G. Maccioni, D. Giansanti: “A New Adaptive Sensor Interface for Composing and Performing Music in Real Time” D. Dondi, D. Brunelli, L. Benini, P. Pavan, A. Bertacchini, L. Larcher: “Photovoltaic Cell Modeling for Solar Energy Powered Sensor Networks” P. Bruschi, M. Piotto, N. Bacci: “Postprocessing Technologies, Interface Circuits and Packaging Strategies for CMOS Compatible Gas Flow Sensors” L. Rossi, G. Breglio, A. Cusano, A. Irace, V. Pascazio, A. Tutolo: “Multiplexing of Fiber Bragg Grating Sensors: Time Windowed Improved C-PFM Reading Technique. An Experimental Validation” R. Sharma R. Mary Lourde: “Crosstalk Reduction in Balise and Infill Loops in Automatic Train Control” A. Barberis, L. Barboni, M. Valle: “Assessment of the MAC Layer Behavior of Wireless Sensor Networks Simulators Using Experimental Testbeds” G. Melone, P. Miodushevsky, L. Rizzi, L. Vasanelli: “Miniaturized Thin Film Temperature Sensor For Wide Range of Measurements”

Session IIA: High Energy Physics Detector W. Snoeys: “Electronic system trends and challenges in present day particle experiments” L. Musa: “Highly Integrated System-On-Chip Circuits for the Readout of High-Energy Physics Detectors” F. Corsi, M. Foresta, C. Marzocca, G. Matarrese, A. Del Guerra: “Current-Mode Front-End Electronics for Silicon Photo-Multiplier Detectors” M.C. Rossi, G. Conte, V. Ralchenko: “Polycrystalline Diamond X-ray Sensors: Intensity and Field Dependent Response” A. Baschirotto, S. D’Amico, M. De Matteis, F. Grancagnolo, M. Panareo, R. Perrino, G. Chiodini, G. Tassielli: “A CMOS high-speed front-end for cluster counting techniques in ionization detectors” M. Traversa, P. Prete, I. Farella, P. Paiano, F. Marzo, A. Cola, N. Lovergine, A.M. Mancini: “A MOVPE Technology for Fabrication of CdTe-based Homoepitaxial p-i-n Diode Structures as Nuclear Radiation Detectors”

Session IIB: Automotive and Industrial Sensors and Sensors Interfaces E. Pajot: “Olfactory nanobiosensors based on sniffing yeasts” M.J. Ohletz, F. Schulze: “Requirements for Design, Qualification and Production of Integrated Sensor Interface Circuits for High Quality Automotive Applications” L. Milor, C. Jia: “BIST for Testing of Delay”

P. Lopalco, S. Lobasso, A. Corcelli, M. Dibattista, R. Araneda, Z. Peterlin, S. Firestein: “Do Olfactory Receptors Respond to Explosives?” A. De Marcellis, G. Ferri, M. Patrizi, V. Stornelli, A. D’Amico, C. Di Natale, E. Martinelli, A. Alimelli, R. Paolesse: “An Integrated Analog Lock-In Amplifier for Low-Voltage Low-Frequency Sensor Interface” M. Melani, F. D’Ascoli, L. Fanucci, F. Iozzi, A. Gianbastiani, A. Rocchi: “Inertial sensors rapid prototyping for automotive application” F. Dell’Olio, V.M.N. Passaro, F. De Leonardis: “Sensitivity Analysis of Rib Waveguides for Integrated Optical Sensors” F. Placentino, F. Alimenti, A. Battistini, W. Bernardini, P Mezzanotte, V. Palazzari, S. Leone, A. Scarponi, N. Porzi, M. Comez, L. Roselli: “Measurements Of Length And Velocity Of Vehicles With A Low Cost Sensor Radar Doppler Operating At 24GHz” C. Bonserio, M. Giannini , A.M. Losacco, G. Pappalettera: “Realization of a physical mask by laser micro-cutting process”

Session IIC: Imaging Systems for security G. Sansoni, M. Trebeschi, F. Docchio: “Application of three-dimensional optical acquisition to the documentation and the analysis of crime scenes” G.P. Suranna: “Ter-anthrylene-ethynylenes: new anthracene based structures for solution deposited OTFT with potential sensing properties” G. Costantini, D. Casali, R. Perfetti, M. Carota: “A Binocular Sensor Interface for Moving Objects Detection” F. Lopez, M. Mennuni, M. Giustini, M. Giomini, M. Dezi, G. Venturoli, A. Mallardi, G. Palazzo: “Photosynthetic Reaction Centers Embedded in Polyelectrolyte Multilayer as a Tool in the Determination of PSII Herbicides” S. Maggi, N. De Leo, M. Fretto, V. Lacquaniti: “Superconducting Tunnel Junction X-Ray Detectors with Ultra-Low Subgap Current” R. Loiacono, F. Dell’Olio, V.M.N. Passaro: “Hollow Core Waveguides for Optical Chemical Sensing” A. Guerriero, R. Matarrese, A. Morea, C. Pasquale, F. Ragni, K. Tijani: “A Grid Portal to Improve SST Maps” M. Lucci, V. Sessa, S.Orlanducci, E. Tamburri, F. Toschi, M.L. Terranova, A. Reale, A. Fiorello, C. Falessi: “A carbon nanotube-based quartz crystal nanobalance for NH3 detection : toward the assembling of a sensing platform” V. Lacquaniti, D. Andreone, M. Fretto, N. De Leo, S.Maggi, F. Francone, R. Rocci, D. Serazio, A. Sosso: “Overdamped Josephson Junctions for Applications to precision measurement”

Foreword from the General IWASI07 Chair Today the application of silicon micro-sensors and micro-actuators has been increasing at high rate. Micro-sensors appear in great numbers, notably in automobiles, process controls, biomedical applications, and scientific instrumentation. A shift of emphasis in Micro-ElectroMechanical Systems (MEMS) research is emerging, i.e. from core micro-fabrication technologies to application-specific micro-sensors and micro-actuators. More than ever the work on micro-devices is focused on application-driven selection of fabrication technologies (or integration of several technologies), suitable materials, system-level integration, and packaging. MEMS technology is always in competition with non-lithography based manufacturing and often is the preferred solution in applications. Due to lower costs, superior performance and monolithic integration with processors and actuators, micro-sensors are widely employed on the factory floor and in home devices. Recent advances in semiconductor fabrication technology have enabled the development and production of sensors of a new generation, referred to as intelligent sensors. Those are characterized by having significant data processing, storing and analyzing power. These intelligent sensors can be used as autonomous systems or deployed in large numbers to form powerful sensor networks. The sensor networks may depend on multiple embedded processors to simultaneously gather and process information from many sources. They are often flexible, self-organizing and fault tolerant, thus making them well suited for mission critical applications. The Second IEEE International Workshop on Advances in Sensors and Interfaces (IWASI 2007), is held June 26-27, 2007, in Bari, Italy. It is aimed at bringing the gap between sensor devices and their integration with the electronics by using newly developed technologies. Besides this, electronic design tools and manufacturing technologies are key to achieve high design quality and to meet the tight time-to-market requirements. IWASI provides a forum for the exchange of ideas and results. It spans a range from sensor applications, over biological and chemical sensors for high-energy physics sensors to sensor interfaces and networks. All papers submitted have been peer-reviewed and a collection of papers is published in the proceedings, thus becoming an archive for practitioners and researchers. A session on advances in sensors and interfaces presents new trends for electronics for molecular electronics and the applications in the area of bio-molecule, portable and implantable sensors. The session on MEMS-based sensors and sensor networks represents recent developments in sensor integration. Another session deals with high-energy physics sensors discussing novel techniques in radiated environments. Automotive and Industrial sensors are a specific focus as in this area high volume sensor applications are installed and require specific design and production techniques. The applications of sensor in imaging systems for map detection and safety are described. The variety of sensors covered in this workshop demonstrates the diverse applications of which sensors have managed to become an essential part of. As time progresses we will see sensors become even more ubiquitous in any fields of applications.

Prof. Dr. Daniela De Venuto General Chair IWASI 2007

Organizing Committee Workshop Chair: D. De Venuto (Politecnico di Bari and INFN Bari-Italy) Steering Committee: B. Courtois (CMP-TIMA Grenoble-France) M. Declercq (EPFL Lausanne-Switzerland) B. Riccò (Univ. di Bologna-Italy) C. Van Hoof (IMEC Leuven-Belgium) Local Committee: E.Nappi (INFN Bari - Italy) D. De Venuto (Politecnico di Bari and INFN Bari-Italy) E. Cantatore (Politecnico di Bari-Italy) Technical Program Committee: L. Benini (Univ. di Bologna-Italy) E. Cantatore (Eindhoven Univ. of Technology.-NL) H. Casier (AMIS Bruxelles-Belgium) K. Chakrabarty (Duke University-USA) B. Courtois (CMP-TIMA Grenoble-France) G. De Cesare (Univ. La Sapienza Roma-Italy) D. De Venuto (Politecnico di Bari and INFN Bari-Italy) G. Gielen (Univ. Leuven-Belgium) F. Grancagnolo (INFN Lecce-Italy) M. Kayal (EPFL Lausanne-Switzerland) S. Mir (TIMA Grenoble-France) E. Nappi (INFN Bari -Italy) A. Taroni (Univ. di Brescia - Italy) L. Torsi (Univ. di Bari-Italy) M. J. Ohletz (ZMD AG-Germany) M. Savino (Politecnico di Bari-Italy) B. Vigna (ST Microelectronics-Cornadero-Italy)

Session IA Advances in Sensors and Interfaces

Ultra-low power Biopotential Interfaces and their application in wearable and implantable systems

Chris Van Hoof, Refet Firat Yazicioglu, Tom Torfs, Patrick Merken Imec, kapeldreef 75, 3001 Leuven, Belgium Abstract: With the advent of ultra-low power sensor interfaces, long-term ambulatory monitoring using wearable devices and more energy-autonomous implants are becoming a reality. This paper wil present suitable architectures and circuits, and will present several application case examples.

1-4244-1245-5/07/$25.00 ©2007 IEEE

CMOS Chips for Bio Molecule Sensing Purposes Roland Thewes [email protected]

1-4244-1245-5/07/$25.00 ©2007 IEEE

Adenine -Thymine (A -T): 2 Hydrogen bridges Guanine -Cytosine (G -C): 3 Hydrogen bridges

• Negatively charged • Mass of 1 nucleotide pair ! 1.1 ! 10-21 g • Lengths: - salamander:

~ 9 ! 1010 base pairs

- corn: - humans: - tomato:

~ 1.5 ! 10 10 base pairs ~ 3 ! 109 base pairs ~ 7 ! 108 base pairs

- E coli: - simple viruses: - plasmids:

~ 5 ! 106 base pairs > several 10 k 1...250 k

2 nm

0.34 nm base pair

Fig. 1: Important physical and chemical DNA properties.

3.4 nm

In recent years, CMOS-based DNA and protein micoarrays have attracted much attention. They are believed to provide a huge potential in the area of medical diagnosis, environment monitoring, and further applications in life sciences and biotechnology. The desoxyribonucleine acid (DNA) contains the full genetic information of organisms, and conserves and transfers the genetic information by replication. The ribonucleic acid (RNA) carries a short term copy of the DNA information and expresses the genetic information. Proteins are essential for cell operation, provide significant contributions to entire cell metabolism, and have various specific functions. Among theses bio molecules DNA is by far the most stable and easiest to handle. Moreover, techniques exist to re-write information about gene expression and thus protein generation into complimentary DNA (cDNA). For these reasons DNA related investigations play a predominant role in the field of microarray based techniques. A few important basic DNA properties are summarized in Fig. 1. The purpose of DNA microarrays [1, 2] is to highly parallel investigate a given sample concerning the presence / absence or quantitative amount of specific (predefined) DNA sequences. Most important applications are genome research and drug development; in particular the field of medical diagnosis is under development. Depending on the specific application different requirements arise concerning array density, sensitivity, dynamic range, and specificity. The basic setup of a DNA microarray is shown in Fig. 2: A slide (“chip”) of the order mm2 ... cm2, usually made of glass, polymer material, or Si, provides a number of test sites. At these sites single-stranded DNA probe molecules are immobilized with different sequences of typically 20 – 40 bases. In Figs. 2b) and c), two different sites are considered after the immobilization phase (for simplicity, probe molecules with only five bases are depicted here). In the next step (Figs. 2d) and e)), the sample containing the target molecules is applied to the whole chip. The target molecules can be up to two orders of magnitude longer compared to the probe molecules. Hybridization between probe and target molecules is obtained in case they “match”, i.e. if they have complementary sequences. If they mismatch, (Fig. 2e)), chemical binding does not occur. Finally, after a washing step, double-stranded DNA is obtained at the match

• Complementary (matching) base pairs:

pentose

1. Introduction

• Usually organized as double helix

phosphor acid rest

Abstract - A topical review is given concerning CMOSbased DNA micoarrays. Considering the entire application chain, functionalization techniques, required CMOS process extensions and consequences, and a number of readout techniques are discussed. Related circuit design issues and CMOS implementations are highlighted as well.

positions (Fig. 2f)), and single-stranded DNA remains at the mismatch sites (Fig. 2g)). Thus, provision of the information which amount of double-stranded DNA 36is° found at the respective sites (cf. chapter 4) reveals the composition of the sample. Before discussing further technical details we briefly consider the entire application and manufacturing chain of microarrays (Fig. 3): Starting with a naked chip or substrate in a first step it has to be functionalized, i.e. the required bio-molecules have to be brought to and immobilized at their target positions. Then, the chip must be assembled in a suitable package which must provide microfluidic functionality as well. In case of electronic chips this means that the package must provide both an electrical and a fluidic interface. Note, that depending on the manufacturing procedure, the order of packaging and functionalization may also be changed. After functionalization and packaging the chip must be stored under suitable conditions which are more sophisticated as compared to standard CMOS chips for standard applications (temperature, humidity, sterility, protection of fluidic interface, …). To finally operate the chip, usually first the sample (e.g. blood) has to be prepared and specifically processed. As schematically depicted in the figure, both branches meet in the reader unit providing the experimental results which then have to be further interpreted so that disciplines as bio-informatics come into play at this point.

2. Functionalization Fig. 4 shows an overview about different functionalization techniques. For low and medium density arrays deposition of off-chip synthesized probe molecules by microspotters [3, 4] is widely applied. Today, such devices are able to handle volumes in the sub-nl range at pitches of order 100 µm or even below. An alternative approach to direct off-chip synthesized molecules to their on-chip target position makes use of the negative charge of DNA: electrophoretic forces are applied via noble metal electrodes (Fig. 5). To protect the assembled DNA from electrochemical reactions the electrodes are covered by a permeation layer [5, 6].

species 1 (probe molecules) species 2 species 3

species N (probe molecules)

a)

Provision of sample (with single stranded DNA target molecules) to the whole chip and start of hybridization

b)

d)

Detection of hybridization after washing

f)

sensor site

sensor site sensor site sensor sensor area

c)

ELECTROLYTE

e)

A G C T T G

A G C T T G

mismatch matchsite sensor

electrophoretic force G C C T A G

application of positive voltage permeation layer noble metal electrode

+++

g) already functionalized

sensor site

G C C T A G

Immobilization of different single -stranded DNA sequences (probe molecules) at different sites

G C C T A G

microarray chip

under functionalization

to be funct. in a forthcoming step

sensor site

Fig. 5: Schematic plot demonstrating DNA microarray functionalization using electrophoretic forces.

Fig. 2: Basic operation principle of DNA microarrays. sensor area

Sample

Optionally

Optional use of

Silicon

CMOS functionality Sample preparation, PCR, ...

Chip (processed solid state material)

Functionali zation

Packaging

Storage

Readout

sensor area

Interpretation (i.e. make use of the result)

or vice versa!

on -chip

Optical control of in -situ growth Electronic control of in -situ growth

off -chip

Application area

DNA synthesis

Fig. 3: Entire application/manufacturing chain of microarry chips.

10

Electrophoresis driven placement Spotting 0

10

1

10

2

10

3

104

105

106

Test sites/ chip

Diagnostics Drug research

Fig. 4: DNA microarray functionalization techniques and related application areas.

Whereas these two techniques are applicable for low and medium density chips (usually aiming for diagnostic applications) on-chip in-situ growth techniques must be applied for high density arrays. The basic idea using an optical technique is schematically sketched is Fig. 6: We consider two sites starting at a point where molecules already have been grown with five bases. The upper end of the DNA is terminated by a protection group which can be released by illumination. Thus, provision of light to selected sites and subsequent washing of the chip leads to the situation shown in the middle column of the figure. After deactivation of the light source a new base terminated with a protection group is applied to all sites in parallel but can only bind to the unprotected DNA strands. This procedure finally leads to the situation depicted in the rightmost column of the figure. For that purpose lithography-based mask techniques are used similar to that known from the semiconductor world [7, 8]. Alternative optical techniques use digital projector based systems instead [9].

Fig. 6: Schematic plot demonstrating optically driven in-situ

DNA synthesis. Electrical approaches for on-chip in-situ synthesis replace the light triggered mechanisms by electrical signals applied to noble metal electrodes controlling the required chemical processes [10, 11]. Due to the high-density focus, the latter technique has to use CMOS as a platform. Use of CMOS has also been demonstrated for the electrophoresis based approach (Fig. 5). From an engineering standpoint, a processing challenge arises to provide noble metal electrodes on CMOS, concerning design issues we find that only relatively large signals have to be handled at relaxed frequencies. As will be discussed in forthcoming chapters, CMOS performance requirements are increased if electrical readout is requested as well.

3. CMOS Integration The interaction of CMOS chips with the wet world of DNA microarrays requires extra processing steps to provide biocompatible transducer materials. Fig. 7 provides a schematic overview of related options. The need to introduce noble metal electrodes (e.g. Au, Pt) has already been addressed before and will be stressed in chapter 4 again. Moreover, an option using electrodes covered by a dielectric is also considered in the figure (cf. chapter 4.3). Since CMOS production lines are usually not compatible with such extra processes due to contamination problems, the concept of CMOS postprocessing is frequently applied. Also there, however,

standard CMOS passivation

last CMOS additional metal (Al) passivation (optionally)

noble metal (Pt, Au)

noble metal

permeation layer

passivation

dielectric with relatively high k / low thickness thin metal electrode

40

standard CMOS

30 35

no anneal VDD

350°C, 30min

VDD VDD

400°C, 30min

(a)

(b)

(c)

20 25

(d)

Deposit & structure resist, deposit Ti / Pt / Au Etch Ti/ TiN Deposit Ti/ TiN barrier, fill W Ti/ TiN Etch nitride / oxide W

5 µm

Gain error [%]

Fig. 7: Schematic plot showing standard (left) and different extended CMOS process options used to realize CMOS DNA microarray chips aiming for electronic functionalization and/or readout. Lift -off

Au finger Nitride

Fig. 8: Extended CMOS process flow.

Etching artifact due to preparation

*

output

* test / cal. enable

Au

10-12

**

sensor bias voltage

10 15

0 5

Pt Ti Au

Nitride deposited for preparation

Si3N4 SiO 2 Al CMOS

.

10-11

GND

10-10 sensor

* *

-9 test / calibration 10 10-8 10-7 current input VSS VSS VSS

10-6

Test current [A]

-5

Fig. 10: Measured current gain1.E-09 as a 1.E-08 function of the1.E-06 test input 1.E-12 1.E-11 1.E-10 1.E-07 current of the test circuitInput shown in the inset after different Current [A] annealing conditions.

Aluminum 2 Etching artifact due to preparation

cross section

Aluminum 1

5 µm FOX

Tungsten Diffusion

HDD Spacer

3 µm top view

3 µm

3 µm

1 µm

50 µm

Gate

no annealing

350 °C, 30 min

400°C, 30 min

Fig. 11: SEM photos showing Au sensor electrodes without and with annealing steps performed at different temperatures after Au processing.

Fig. 9: SEM photo of the extended CMOS process in Fig. 8.

care must be taken that post-processing does not deteriorate the quality of the CMOS devices, in particular when sensitive analog circuitry is realized. As an example we consider a process used in [12-16] (Figs. 8 and 9): There, a Ti/Pt/Au stack (50 nm / 50 nm / 300-500 nm) is deposited and structured using a lift-off process. The basic CMOS technology is a 5 V, 6'' n-well process specifically optimized for analog applications with a minimum gate length of 0.5 µm and an oxide thickness of 15 nm. During the development phase, testchip circuits were designed to be operated with quasi DC sensor currents from 1 pA to 100 nA (cf. chapter 4, redox-cycling method). As shown in the inset of Fig. 10, a regulation circuit controls the bias voltage of the sensor, whose current is amplified by a factor of approximately 100 using two simple cascode current mirrors in series. A test input allows to calibrate the circuit [12]. Measured current gain as a function of the input current is shown in Fig. 10. Data are shown with and without additional annealing steps after the gold process module. If no annealing step is applied, a severe gain deviation is obtained for input currents below 10 pA. This effect coincides with an increase of the interface state density (> 2 × 1011 cm-2) measured on single devices using the charge-pumping characterization technique. These excessive values translate into worsened off-state characteristics and increased junction-to-substrate or

junction-to-well leakage currents, so that the obtained behavior reflects nothing but the effect of degraded transistor performance. The most crucial nodes of the test circuit are emphasized by asterisks in the figure. Annealing in forming gas (N2, H2 at 400 °C / 350 °C, 30 min) after gold processing significantly reduces the interface state density again to very good values below 1010 cm-2 so that also reasonable circuit transfer characteristics are obtained again. However, under 400 °C conditions sensor backend parameters are found to degrade (increase of the gold square resistance by 20 %, occurrence of a few shorts in interdigitated sensor structures due to grain building processes, Fig. 11). Consequently, as a condition where both device and electrode properties are sufficient, the 350 °C annealing condition is identified.

4. Readout 4.1 Optical and Electronic Readout Starting with the situation depicted in Figs. 2f) and g), the remaining demand is to make double-stranded DNA visible. In the widely used state-of-the-art optics-based readout technique [2], the target molecules are labeled with fluorescence molecules. After hybridization and a subsequent washing step, the whole chip is illuminated or scanned with monochromatic light with a wavelength matched to the absorption profile of the marker molecules. A camera system with a blocking filter for the excitation wavelength takes an image of the entire array chip and fluorescence light emitted at a considered position reveals successful hybridization.

a)

ELECTROLYTE

ELECTROLYTE electrochemical label molecule, e.g. ferrocene target DNA molecule (H10 C10Fe) not functionalized probe molecule reference electrode

working electrode

VOD 2 VI 1

counter electrode

-

+

VI 2

VOD

gMC 2

VV OUT

g MC

VVOUT

2

2

i(t)

WE1 WE1

WE2 WE2 VCM VCM S3 S3' S2' S2 S3 S3' S2' S2

1

+

VCM VCM

S1 S1

-

0 .5

+

g MD

in(t)

generator electrode

working electrodes

v step (t )

WE

Cdl

e-

electron transfer from label displacement current through double -layer capacitance Cdl

S1' S1'

S4' S4'

counter electrode substrate

DNA probe

I gen

step contribution from label contribution from double -layer capacitance

offset signal

e-

Q =! I dt

electrochemical label molecule

C’ C’

DNA target

Fig. 12: Principle of coulometric DNA detection. a) Basic setup. b) Configuration optimized for array operation. ELECTRO LYTE

C C

Fig. 14: Sensor site circuit topology used in [15] to realize a 384 site array for coulometric DNA detection.

b) i1(t)

+ PIXEL + PIXEL -

S4 S4

VOC

vstep (t) potentiostat

-

2

g MD

time

Fig. 13: Schematic representation of charge contributions in coulometric DNA detection setups.

Circumvention of the optical reader - which is the case in fully electronic methods - promises the advantages of decreased system costs and increased user friendliness. Under condition that sample preparation (cf. Fig. 3) is further optimized as well – an issue which is frequently ignored in the context of such discussions – fully electronic approaches also provide the opportunity of handheld devices and significantly increased flexibility. 4.2 Electrochemical Principles Coulometric Detection Coulometric DNA detection (Fig. 12) is based on electron exchange between an electrode and electrochemical labels (e.g. ferrocene) attached to the DNA target strand. In order to trigger the electrochemical reaction the potential between electrode and electrolyte has to be changed step-wise. Whereas the standard setup is usually described as shown in Fig. 12a), where the electrolyte potential is held constant by a potentiostat and the electrode voltage is changed, a configuration more suitable for arrays is shown in Fig. 12b) where the electrodes are operated at a constant potential and the potentiostat provides a voltage step. The method is suitable for measurements requiring only a limited dynamic range: As shown in Fig. 13, not only the labels contribute to the charge flowing through the electrode but also the electrode-to-electrolyte doublelayer capacitance. For that reason, in practice electrode surface blocking layers are used after probe molecule immobilization to decrease the effect of the double-layer capacitance, labels are sometimes applied providing more than one electron per oxidation / reduction, and/or labels should be used with a relatively low oxidation / reduction voltage (e.g. of order 200 mV).

Vgen = enzyme label red

I col

+

Vcol =

potentiostat

collector electrode

ox

reference electrode

Fig. 15: Schematic plot showing the redox-cycling sensor principle and the sensor layout. Left: Single sensor consisting of interdigitated working electrodes and potentiostat circuit. Right: Blow-up of a tilted sensor cross-section. For simplicity, probe and target molecule are only shown on one of the electrodes.

Using the sensor circuit topology shown in Fig. 14, an array with 384 sites was designed and successfully tested [15]. The circuit in Fig. 14 allows integration of the charge from the electrodes. Single-ended data readout is achieved in spite of operating only one fully differential opamp for two sites in parallel by using an input referred common mode feedback configuration. The switches are arranged in a way that leakage currents are minimized after readout (=integration), i.e. during the period the integrators are subsequently connected to an external reader unit. Redox-Cycling As compared to the principle described above this technique provides increased sensitivity and dynamic range at the cost of increased chemical and operating complexity. The principle as such [17, 18] is explained in Fig. 15: A single sensor consists of interdigitated gold electrodes (generator and collector). The target molecules are tagged by an enzyme label (Alkaline Phosphatase) that is capable to cleave a chemical substrate (paraAminophenylphosphate) which is applied to the sample after hybridization. Whereas the substrate itself is not electrochemically active, the generated species (paraAminophenol) after cleaving is. Applying simultaneously an oxidation and a reduction potential to the sensor electrodes, para-Aminophenol is oxidized to Quinoneimine at the one electrode, and Quinoneimine is reduced to para-Aminophenol at the other one. The activity of these electrochemically redoxactive compounds translates into a quasi-DC electron current at both gold electrodes (symbolized by Igen and Icol in the figure).

16x8 sensor array with in sensor -site A/D conversion

&

Fig. 16: Redox-cycling prototype chip with 16 x 8 sensor sites and simplified schematic of sensor site circuitry [13, 14].

A chip photo of a prototype chip with 16 x 8 sensor sites and a simplified schematic of the sensor site circuit are shown in Fig. 16. A global potentiostat is used. Selection of the sensor positions is done by x- and ydecoders and a multiplexer. A serial interface allows the chip to communicate with a reader unit using six pins only independent of its array magnitude. Circuitry for reference generation and sensor site calibration is provided on-chip as well. Within each sensor site direct A/D-conversion is realized using a current-to-frequency sawtooth generator concept: An integrating capacitor is charged by the sensor current. When the threshold of the comparator is reached, a reset pulse is generated which passes through a delay stage and discharges the capacitor again. The number of reset pulses within a given time interval is counted with a 24 stage digital counter. For readout, the counter circuit is converted into a shift register by a control signal and the data are provided to the output of the chip. Further Approaches A number of further variants are available in the literature providing various specific benefits and disadvantages. As an example the use of intercalator molecules is discussed here [19]. These molecules carry a number of electrochemically active labels and are bound in between hybridizing (i.e. double-stranded) DNA but not captured by single-stranded molecules. The advantage is the high signal strength achieved by the availability of numerous labels per hybridization event, a disadvantage results from the fact that intercalator molecules require to consider specific safety regulations. 4.3 Non-Electrochemical Labeling-Based Approaches There are further labeling-based approaches which rely on different principles as discussed above. Two examples are briefly discussed in this chapter. The first method is based on labeling the target molecules with Au beads and application of a subsequent silver precipitation step. The Au beads form a seed layer to bind Ag which forms an increasingly dense layer under continuous Ag provision. Besides pure optical detection methods, conductivity measurement between electrodes separated by an isolating layer has been proposed [20],

target probe

C

hybridization

C

sensor electrode

Fig. 17: Basic principle of impedance-based DNA sensors.

as well as impedance (or RF parameter) measurement between electrodes isolated by a dielectric layer from the electrolyte [21]. Also optical attenuation detected by a CMOS imager chip has been suggested [22]. Moreover, there are a number of investigations using magnetic bead labeling. Recently, a first CMOS array with integrated GMR sensors has been published [23]. There, the presence of labeled target DNA molecules changes the resistance of the sensing elements. 4.4 Completely Labeling-Free Approaches The idea to simplify the sample preparation and the entire chemical assay procedure has triggered a number of ideas to completely avoid labeling steps. The two most prominent examples are impedance-based and gravimetric approaches. Impedance-Based Detection V The basic principle V of impedance sensors is digital VA schematically depicted in Fig.counter 17. There, the electrical A

Iref

out

V ref

Iref

Vout

VB

V ref

V ref

digital counter

VA VB

VB

V out T

VA

Vinterdigitated ref

VB

electrodes

V out T

Fig. 18: Circuit used for impedance-based DNA detection [16]. V bias

Fig. 19: Left: Schematic cross section FBAR. Middle: Schematic cross section flip-chip setup. Right: Chip photo flip-chip setup [25].

When operated in liquids the related mechanical attenuation leads to a severe decrease of quality factor of the resonator element. In order to compensate for that appropriate oscillator designs are mandatory. A recently published example is shown in Fig. 20, biological measurements using less integrated solutions have been published earlier [26]. Fig. 20: FBAR Oscillator topology used in [25].

5. Summary

impedance between a sensor electrode and the electrolyte or in between interdigitated electrodes is characterized. V i V o FBA properties R C+ 0 Hybridization leads to a change of the electrical of the electrode-to-electrolyte interface, in particular to a decrease of the capacitive part of the impedance. Unfortunately, the impedance itself and the impedance change of such sensors are not only determined by a capacitive effect, but also by an ohmic contribution. Both contributions depend on the varying quality of the layer of probe molecules. For this reason, phase-sensitive or other RC-discriminating techniques are required to evaluate the measured sensor signal. A suitable circuit used within a 16 x 8 sensor array is shown in Fig. 18. A constant current source (Iref) is used to alternately charge / discharge the two terminals of the sensor consisting of interdigitated electrodes. Its voltage difference is monitored using a fully differential comparator. When the amplitude exceeds a defined threshold, all switches change their state so that the charged (discharged) sensor terminal is discharged (charged) again, and the threshold voltage changes polarity, respectively. As a consequence, a signal with approximately triangular shape is obtained at the sensor terminals. Deviation from a constant slew-rate is due to the resistive component of the sensor. An evaluation shows, that measurement of the resulting frequency f for different values of Iref allows to determine capacitive and resistive part independently. First results of a set of hybridization experiments prove feasibility [16].

CMOS-based DNA microarrays clearly have proven to open the way to novel and user-friendly solutions. However, to gain success not only on a demonstrator level the entire system (including packaging, storage, microfluidics, software, ...) must be considered from the user's point of view. Successful commercialization has not yet widely been achieved, but approaches aiming in this direction are on-going.

Gravimetric Detection Mass sensitive sensors as considered here are mechanical / electrical oscillating systems. Mass changes at the sensor surface change the oscillation frequency. Following the Sauerbrey equation [24], high sensitivity requires high oscillating frequency. For that reason, film bulk acoustic resonators (FBARs) operated in the GHz range are superior compared to conventional quartz-based sensors operating in the MHz...tens of MHz range. An example of such a sensor and a flip-chip arrangement together with a CMOS chip is shown in Fig. 19 [25].

References: [1] http://www.nature.com/ng/chips_interstitial.html [2] "DNA microarrays: a practical approach", M. Schena ed., Oxford University Press Inc., Oxford, UK, 2000 [3] E. Zubritsky, Anal. Chem., 2000, December 1, p. 761 [4] V. G. Cheung et al., Nat Genet., 1999, p. 15 [5] M. J. Heller, IEEE Engineering in Medicine and Biology Magazine, 1996, p. 100 [6] http://www.nanogen.com [7] S. P. A. Fodor et al., Nature, 1993, 364, p. 555 [8] http://www.affymetrix.com [9] http://www.nimblegene.com [10] K. Dill et al., Anal. Chimica Acta, 2001, 444, p. 69 [11] http://www.combimatrix.com [12] F. Hofmann et al., Tech. Dig. IEDM 2002, p. 448 [13] M. Schienle et al., J. Solid-State Circuits, 2004, p. 2438 [14] A. Frey et al., Proc. ISCAS 2003, p. V9 [15] M. Augustyniak, Tech. Dig. ISSCC 2006, p. 46 [16] C. Stagni et al., J. Solid-State Circuits, 2004, p. 2956 [17] A. J. Bard et al., Anal. Chem., 1986, 58, p. 2321 [18] R. Hintsche et al., "Microbiosensors using electrodes made in Si-technology", in 'Frontiers in Biosensorics I', F. Scheller, F. Schubert, and J. Fedrowitz ed., Birkhäuser Verlag Basel/Switzerland, 1997 [19] N. Gemma et al., Tech. Dig. ISSCC 2006, p. 460 [20] M. Xue et al., Tech. Dig. IEDM 2002, p. 207 [21] L. M. Hagelsieb et al., Proc. ESSDERC 2006, p. 125 [22] J. Li et al., Tech. Dig. IEDM 2004, p. 1005 [23] S.-J. Han et al., Tech. Dig. IEDM 2006, p. 719 [24] G. Sauerbrey, Zeitschrift für Physik 155, 1959 [25] M. Augustyniak et al, Tech. Dig. ISSCC 2007, p. 392 [26] R. Brederlow et al., Tech. Dig. IEDM 2003, p. 992

Molecular Electronic Circuits Paul D. Franzon North Carolina State University, Raleigh NC 27695 [email protected], +1.919.515.7351 Abstract This talk focuses on pathways to solving the challenges involved in adapting molecular switches so that they can be used in practical circuits. Addressing these issues result in detailed understanding of the interactions between molecular switch properties and application requirements, deep consideration of interconnect structures, and modeling of potential candidate structures. This talk is part tutorial and part research talk. The tutorial portion will discuss fundamental circuit issues, including CMOS scaling concerns, and competing approaches to nanoelectronics. The research talk will start with a treatment of the appropriateness of Spice as a circuit modeler before presenting our results in memory and logic applications.

1-4244-1245-5/07/$25.00 ©2007 IEEE

Session IB Bio-sensing for Healt Care

Selective quenching of unhybridized fluorescent probes by gold nanoparticles for rapid SNP genotyping using conventional PCR Huixiang Li1, Laura Ascroft2, Andrew I. Brooks3, Robert Russell4, Sheryl Wildt4 and Lewis Rothberg1 1 Department of Chemistry, University of Rochester, Rochester, NY 14627. (e-mail [email protected]) 2 Functional Genomics Center, University of Rochester, Rochester, NY 3 Bionomics Research and Technology Center, Enviromental and Occupational Health Science Institute, UMDNJ / RWJMS, Piscataway, NJ 08854. 4 Harlan Inc., Indianapolis, IA 46229. Abstract We detect single nucleotide polymorphisms (SNP) in genomic DNA that has been amplified using a standard PCR protocol where a short fluorescently tagged probe sequence and a single denaturation and annealing step have been added. We utilize the ability of ionically treated gold nanoparticles to selectively adsorb unhybridized probes in the modified PCR product and quench their fluorescence. The method eliminates the time and expense of gel electrophoresis while retaining the ability to use an ordinary thermal cycler for PCR amplification. We describe an application to genotyping Fatty Zucker rats and validate the method against fluorogenic realtime PCR with double blind studies. Introduction The use of polymerase chain reaction (PCR) amplification of DNA for genotyping, pathogen detection and forensics is widely practiced. Frequently, PCR is followed by gel electrophoresis to assess whether amplification was successful, reflecting the presence of the target sequence. Gel analysis of PCR products is expensive, time consuming and labor intensive. Furthermore, it is very difficult to screen for single nucleotide polymorphisms using gel electrophoresis since amplicons have nearly identical charge and mass irrespective of whether a single nucleotide polymorphism (SNP) is present. SNP screening has become increasingly important as it is now well-established that variations in DNA between individual organisms at the single base level are correlated to the propensity to develop many diseases (1-10). In some cases, the SNPs contained in an individual’s genome can also be used to predict the efficacy of certain drugs or therapeutic treatments (11, 12). Traditional methods for identifying SNPs, such as sequencing and restriction fragment length polymorphism mapping, are labor-intensive and expensive (13, 14). Real-time PCR can provide rapid and accurate screening but requires relatively expensive hardware and complex assay design compared to ordinary PCR with thermal cyclers (15, 16). Here we report a simple method appropriate to SNP genotyping that utilizes traditional PCR but replaces gel electrophoresis with a 5 minute post process based on fluorescence. The method is based on the observation that short single-stranded DNA (ss-DNA) oligomers (< 20 mers) adsorb rapidly on negatively charged gold nanoparticles in colloidal solution while double–stranded DNA (ds-DNA) does not (17, 18). We demonstrate the technique by genotyping HsdHlr:ZUCKER (“FA”) rats used in obesity research (19, 20) and validate the results against real-time PCR with double blind studies. 1-4244-1245-5/07/$25.00 ©2007 IEEE

Materials and methods Tail clips from HsdHlr:ZUCKER (“FA”) rats (wildtype, heterozygote and homozygote genotypes) were digested and genomic DNA extracted using the Promega Wizard SV genomic DNA extraction kit. The identity of the tail snips was coded and the results not decoded until the completion of the experiment. Samples were prepared and labeled arbitrarily by a non-participant so that the studies performed were double blind. The genomic DNA was amplified by PCR using the Promega PCR master mix kit. A thermal cycler (T-gradient, Biometra) was used to conduct PCR and create a temperature profile for DNA denaturation and probe annealing. The PCR protocol involved 39 cycles with denaturation at 94 °C for 30 seconds, primer annealing at 58 °C for 30 seconds and extension at 72 °C for 30 seconds but this protocol can be further optimized to reduce the assay time. Aliquots of the same samples were analyzed using Taqman chemistry for allelic discrimination on the Applied Biosystems 7900HT Sequence Detection System. The PCR primers for FA rat DNA amplification (Forward: CCA AAC AAA AGC ACC ATT TCC ACT T, Reverse: GCA GCC TCT CTT ACG ATT GTA GAA T) and the rhodamine red labeled probes for FA mutation detection (wprobe: ATA TC [A] GGT GAA ATA T, m-probe: ATA TC[C] GGT GAA ATA T) were synthesized by MWGBiotech. The 5’ end of all of the probes was labeled with rhodamine red. The reagents for PBS buffer solution, potassium phosphate (monobasic, anhydrous 99.999%) and sodium phosphate (dibasic, anhydrous, 99.999%), were obtained from Aldrich Chemical (Milwaukee, WI) and used as supplied. Sodium chloride crystals were purchased from Mallinckrodt (Hazelwood, MO). The gold nanoparticles were synthesized and suspended in a colloid according to literature methods (21). The reagents used in the nanoparticle synthesis, hydrogen tetrachloroaurate (III) (HAuCl4.3H2O), 99.99% and sodium citrate (Na3C6H5O7•2H2O), 99%, were purchased from Alfa Aesar and used without further purification. A microfuge tube containing 4 µL of 0.4 µM w-probe in 10 mM PBS and 0.3 M NaCl and 4 µL target (PCR product) was heated to 95ºC for 2 minutes and the solution was annealed for 1 minute at a temperature (42ºC) determined to be optimal for SNP discrimination. The resulting solution was added to 500 µL gold colloid and then 500 µL of a solution of

10 mM PBS and 0.2 M NaCl was added without delay. The fluorescence intensity of those mixtures was monitored with a fluorimeter (Fluorolog 3, Jobin-Yvon). Traces of photoluminescence versus time were recorded at 590 nm with excitation at 570 nm and slit widths corresponding to 4 nm bandwidth. The procedure is repeated with m-probe on a separate aliquot of PCR product. The determination of genotype involves comparing the amounts of fluorescence observed in the case where w-probe is used to that where mprobe is used. Assay rationale and design The detailed mechanism behind the different propensity for ss-DNA and ds-DNA to adsorb on ionically coated gold nanoparticles remains under investigation. Previous experimental work, however, indicates that the origins of the difference are electrostatic (17). The relatively neutral, hydrophobic bases in ss-DNA avoid exposure to water and can find it attractive to adsorb on negatively charged gold nanoparticles. In contrast, the phosphate backbone of the dsDNA duplex shields the bases and accumulates a charged electrostatic double layer that repels an analogous double layer formed around the gold nanoparticles. Since ds-DNA and short ss-DNA behave differently upon exposure to colloidal gold nanoparticle suspensions, we can design a very simple assay for identifying SNPs in genomic DNA as illustrated in Figure 1.

+ 95ºC and anneal at Ts

MATCH

MISMATCH

Luminescence

Luminescence quenched

Figure 1. Schematic of the strategy for SNP genotyping. First, the ds-DNA target (solid and dashed squiggles) is denatured at 95ºC. The tagged probe (curly line with star representing tag) is annealed with the target at a temperature Ts slightly above the melting temperature of the mismatched probe. After this procedure is applied to separate aliquots of target with wild-type and mutant probes, the resulting solutions are exposed to gold colloid (circles) and salt. Finally, the amount of fluorescence of the mixtures is measured and compared. A fluorescently labeled oligonucleotide probe complementary to the wild type sequence can be annealed into thermally denatured PCR product at a temperature selected so that the probe will bind to wild type PCR amplicons but not mutant amplicons where a SNP is present. Conversely, a probe

complementary to the mutant (m-probe) will bind to mutant amplicons while the wild type probe (w-probe) will not. When this trial solution is mixed with colloidal gold nanoparticles, unhybridized probes adsorb rapidly on Au-np and their dye tag’s photoluminescence is quenched by interactions with the metal. Thus, observation of photoluminescence indicates the presence of wild type target when the w-probe fluoresces and the presence of mutant target when the m-probe fluoresces. The assay is easily interpreted to indicate a wild type (dominant w-probe fluorescence), homozygous (dominant mprobe fluorescence) or heterozygous (both probes fluoresce) genotype. Genotyping of SNP containing rats Control experiments without fluorescently tagged probes exhibit a small background signal due to emission or inelastic scattering from colloidal gold nanoparticles and/or aggregates (Figure 2A). Experiments with fluorescently tagged w-probe and m-probe but no PCR product (Figure 2B) or PCR product from a genetically unrelated sample (18) also exhibit low backgrounds identical to those from gold colloid alone implying complete quenching of unhybridized probe fluorescence (Figure 2C, lower traces). We have tested for possible effects of unreacted primers and unconsumed dNTP on the results by mixing probes with PCR solution prior to amplification and applying the assay protocol without thermal cycling. A small additional fluorescent signal from unquenched probes was observed (Figure 2C, upper traces) but its amplitude is an order of magnitude less than from hybridized probe (Figures 2D-F). The small additional fluorescence results from competitive adsorption of unused primers and dNTP on the nanoparticles which modifies their electrostatic properties and concomitant propensity to adsorb the single-stranded fluorescently tagged probe. Figures 2D-F depict representative traces of photoluminescence in FA samples versus time after formulating the mixture. The data in Figure 2D are characteristic of the behavior of a wild-type sample, Figure 2E a homozygous mutant sample and Figure 2F a heterozygous sample. The relative intensity of fluorescence from w-probe and m-probe clearly determines the genotype. Figure 3 presents the results of double blind experiments on samples from 17 FA rats where the observed photoluminescence ratio of wild type fluorescence to mutant probe fluorescence is plotted for each sample. Conclusions from those ratios enable us to unambiguously determine genotypes (1-3 Wild-type, 411 heterozygous, 12-17: homozygous), all in agreement with assignments from real-time PCR. Considerations in choosing stringency tests As with the result of any stringency test, choice of temperature will affect the specific ratios observed in the experiment of Figure 3. In particular, we are working with a w-probe that has greater binding affinity for a mutant target than the m-probe has for the w-target. The free energies of each probe-target combination are presented in Table 1 and were calculated for standard 1 M Na+ using the program RNAstructure Version 4.2 by Mathews, Zuker and Turner.

90

PL(counts/sec)

B 30

60 90 Time (seconds)

D 0

30 60 90 Time (seconds)

120

50

C 30 60 90 Time (seconds)

120

1000

500

E 0 0

100

0 0

120

1000

500

0

0 0

120

1000

PL(counts/sec)

A 30 60 Time (seconds)

50

30 60 90 Time (seconds)

120

PL(counts/sec)

PL(counts/sec)

50

100

PL(counts/sec)

PL (counts/sec)

100

0 0

150

150

150

500

F 0 0

30 60 90 Time (seconds)

120

Figure 2. Photoluminescence (PL) signal (counts per second) versus time after mixing trial hybridization solution with gold colloid and salt/buffer solution. Red traces are for w-probe and blue for m-probe. A-C are control experiments. (A) Assay results with no fluorescently tagged probes. Two independent experimental traces (green and black) are shown to illustrate variability. (B) Results with fluorescently tagged probes but no PCR solution. (C) Results with fluorescently tagged probes and unrelated PCR product (lower traces) or with fluorescently tagged probes and PCR solution prior to amplification (upper traces). Panels (D)-(F) present typical results for different genotypes of Fatty Zucker rats. (D) Wild-type (w-PL>>m-PL). (E) Homozygous mutant (m-PL>>w-PL). (F) Heterozygous (w-PL,m-PL >> background). The zero of time is a few seconds after mixing when adsorption of unhybridized probes has already occurred. The probe annealing temperature we chose for this work was 42ºC, just high enough to achieve nearly complete suppression of binding between w-probe and m-target (Figure 2D, blue curve). The temperature is therefore necessarily low enough to allow slightly more binding between m-probe and w-target (Figure 2E, red curve). Choosing a higher temperature would reduce signal due to that pairing but would also begin to reduce signal due to binding of the perfect

matches, particularly that between w-probe and w-target. Indeed, our observed levels of fluorescence reflect the relative binding energies as computed in Table 1 and explain why the w-probe to m-probe ratio in wild type samples is greater than the m-probe to w-probe ratio in mutant samples (Figure 3). Nevertheless, discriminating these from heterozygous samples presents no difficulty.

Free energy

w-target

m-target

w-probe

-15.2

-12.5

m-probe

-11.7

-16.4

Table 1. Calculated probe-target binding free energy (Kcal/mol) in 1 molar Na+.

Comparison to a colorimetric approach We have previously shown that similar discrimination of SNPs in PCR product is possible using a colorimetric detection scheme based on salt-induced aggregation of gold nanoparticles [18]. In that case, the adsorption of unhybridized probe DNA on the nanoparticles reduces their propensity to aggregate when challenged with salt solutions. That detection scheme relies heavily on matching the amount of probe to the amount of target and is difficult to apply if amplification is variable or if excess primers and dNTP remain to disturb the assay. The colorimetric assay relies on discrimination of relative colors for different probe hybridization temperatures and therefore requires three tests per probe rather than a single one. Differentiation between samples with one and two mutant alleles is nontrivial. The fluorescent method described here solves these problems as it has the advantage of being nearly a null method where little signal is observed when the probe does not hybridize to the target. The assay is therefore robust and tolerates mixtures of DNA and the presence of excess dNTP or unamplified primers quite well. The fluorescent assay is also quantitative in the same sense as real-time PCR in that the observed fluorescence is proportional to the number of target amplicons present.

Conclusions We have demonstrated a simple way to analyze the results of PCR amplification that is faster, less expensive and easier to automate than gel electrophoresis. In addition, we have shown that it is easy to apply the method to detection of SNPs in the PCR product. While we have illustrated the case of SNP genotyping since it is the most challenging, it is even more straightforward to genotype transgenic, knockout and conditionally modified mice where much larger sequence differences distinguish species and strains. In addition, the mixture tolerant feature of the assay enables us to use it with multiplex PCR (cf. ref. 22) and simultaneously screen for many targets with probe sequences that each have distinct fluorescent tags. Perhaps most exciting is the prospect that the high sensitivity and mixture tolerance can be used to avoid PCR altogether in cases where the total amount of DNA in the analyte is sufficient. Acknowledgments We are grateful for encouragement of this work by Dr. Howard Federoff and to a grant from NYSTAR Contract C020049 for partial support of this work. References 1.

w-probe/m-probe Fluorescence

In addition to being more reliable due to its null format, the fluorescent assay is much more sensitive than the colorimetric approach so it can potentially be applied after many fewer PCR cycles. The present data demonstrate the method to be adequate for standard PCR protocols of 30 – 40 cycles but we have used large volumes of gold colloid solution simply for the convenience of using an existing fluorimeter that requires milliliters of solution. With smaller volumes and laser excitation, the number of cycles required could be reduced substantially.

10

8

6

4

2

0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17

Sample number

Figure 3. Ratio of wild type probe fluorescence to mutant probe fluorescence in double blind tests on 17 fa rat samples. 1-3 are wild-type, 4-11 heterozygous and 12-17 homozygous mutant as determined by real time PCR. Note that the sample number designations for each type were made contiguous for ease of comparison but were randomly distributed in the well plate originally containing the unknown analytes. Error bars derive from fluctuations of the fluorescence signal level (cf. Figure 2) and are propagated to the quotient reported here.

Risch, N. and Merikangas, K., “The future of genetic studies of complex human diseases”, Science 273, (1996), 1516-1517. 2. Brookes, A.J., “The essence of SNPs”, Gene 234, (1999), 177-186. 3. Cargill, M., Altshuler, D., Ireland, J., Sklar, P., Ardlie, K., Patil, N., Shaw, N., Lane, C.R., Lim, E.P. and Kalyanaraman, N., “Characterization of single-nucleotide polymorphisms in coding regions of human genes”, Nat. Genet. 22, (1999), 231-238. 4. Kruglyak, L., “Prospects for whole-genome linkage disequilibrium mapping of common disease genes”, Nat. Genet. 22, (1999), 139-144. 5. Housman, D. and Ledley, F.D., “Why pharmacogenomics? Why now?” Nat. Biotechnol. 16, (1998), 492-493. 6. McCarthy, J.J. and Hilfiker, R., “The use of singlenucleotide polymorphism maps in pharmacogenomics”, Nat. Biotechnol. 18, (2000), 505-508. 7. Roses, A.D., “Pharmacogenetics”, Hum. Mol. Genet. 10, (2001), 2261-2267. 8. Nielsena, R., “Estimation of population parameters and recombination rates from single nucleotide polymorphisms”, Genetics 154, (2000), 931-942. 9. Jorde, L. B., Watkins, W.S., Bamshad, D.J., Dixon, M.E., Ricker, C.E., Seielstad, M.T. and Batzer, M.A., “The distribution of human genetic diversity: a comparison of mitochondrial, autosomal, and Y-chromosome data”, Am. J. Hum. Genet. 66, (2000), 979-988. 10. Hacia, J. G., Fan, J. B., Ryder, O., Jin, L., Edgemon, K., Ghandour, G., Mayer, R. A., Sun, B., Hsie, L. and Robbins, C.M., “Determination of ancestral alleles for human single-nucleotide polymorphisms using highdensity oligonucleotide arrays”, Nat. Genet. 22, (1999), 164-167.

11. Mukohara, T, Engelman, J. A., Hanna, N. H., Yeap, B. Y., Kobayashi, S., Lindeman, N., Halmos, B., Pearlberg, J., Tsuchihashi, Z., Cantley, L. C., Tenen, D. G., Johnson, B. E. and Jänne, P. A., “Differential Effects of Gefitinib and Cetuximab on Non–small-cell Lung Cancers Bearing Epidermal Growth Factor Receptor Mutations”, J. of the National Cancer Institute 97, (2005), 1185-1194. 12. Lynch, T. J., Bell, D. W., Sordella, R., Gurubhagavatula, S., Ross A. Okimoto, R. A., Brannigan, B. W., Harris, P. L., Haserlat, S. M., Supko, J. G., Haluska, F. G., Louis, D. N., Christiani, D. C., Settleman, J. and Haber, D. A., “Activating Mutations in the Epidermal Growth Factor Receptor Underlying Responsiveness of Non–Small-Cell Lung Cancer to Gefitinib”, The New England Journal of Medicine, 350, (2004), 2129-2139. 13. Sanger, F., Air, G. M, Barrell, B. G., Brown, N. L., Coulson, A. R., Fiddes, C. A., Hutchison, C. A., Slocombe, P. M., Smith, M., “Nucleotide sequence of bacteriophage phi X174 DNA”, Nature 265, (1977), 68795. 14. Dowling, T., Moritz, C. and Palmer, J.D. (1990) Nucleic Acids II: Restriction Site Analysis. In: Molecular Systematics, eds. D.M. Hillis and C. Moritz, Sinauer Associates, pp. 250-317. 15. Syvänen, A. C., “Accessing genetic variation: Genotyping single nucleotide polymorphisms”, Nature Reviews Genetics 2, (2001), 930-942. 16. Mackay, I. M., Arden, K. E., and Nitsche, A., “Real-time PCR in virology”, Nucleic Acids Research, 30, (2002), 1292-1305. 17. Li, H. X. and Rothberg, L. J., “Colorimetric detection of DNA sequences based on electrostatic interactions with unmodified gold nanoparticles”, Proc. Natl. Acad. Sci. USA 101, (2004), 14039-14041. 18. Li, H. X. and Rothberg, L. J., “Label-free colorimetric detection of single nucleotide polymorphisms in amplified genomic DNA”, J. Am. Chem. Soc. 126, (2004), 10958-10961. 19. Enriquez, A., Leclercq, I., Farrell, G. C. and Robertson, G., “Altered Expression of Hepatic CYP2E1 and CYP4A in Obese, Diabetic ob/ob Mice, and fa/fa Zucker Rats”, Biochem. & Biophys. Res. Commun. 255, (1999), 300306. 20. Smoller, J. W., Truett, G. E., Hirsch, J. and Leibel, R. L., “A molecular genetic method for genotyping fatty (fa/fa) rats”, Am. J. Physiology 264, (1993), R8-R11. 21. Gradar, K. C., Freeman, R. G., Hommer, M. B. and Natan, M. J., “Preparation and characterization of Au colloid monolayers”, Anal. Chem. 67, (1995), 735-743. 22. Li, H. X. and Rothberg, L. J., “DNA Sequence Detection Using Selective Fluorescence Quenching of Tagged Oligonucleotide Probes by Gold Nanoparticles”, Anal. Chem., 76, (2004), 5414-5417.

Smart Sensors for Fast Biological Analysis. Massimo Lanzoni, Claudio Stagni, and Bruno Riccò DEIS, University of Bologna Viale Risorgimento 2, 40136 Bologna, Italy [email protected]

Abstract This paper presents an updated overview of present works concerning the realization of biological sensors based on electronic devices. In particular DNA sensors will be described and their main characteristics will be analyzed. As an example, a new sensor for DNA detection and analysis will be described in details. The proposed approach is based on the use of non-volatile memories and does not require sample treatment nor sensor surface functionalization. Moreover fabrication technology is widely compatible with standard CMOS process. Experimental results show that the sensor has sufficient accuracy and sensitivity, that can be further improved by suitable device engineering. Introduction Increasing research on electronic devices able to perform fast and accurate biological analysis is led not only by industrial interests in clinical applications but also by antiterrorism programs. In recent years a variety of devices has been developed thanks to new silicon technologies, such as micromachining and microfluidics allowing dramatic scaling of equipments usually bulky, difficult to use and requiring large amount of material for the analysis. Medicine, biology, food and ambient analysis require increasing DNA sequencing and this motivated a large development of new systems capable of answering such demand. All the proposed devices are based on the effects of possible hybridization between the single stranded DNA fragments to be analyzed (hereafter referred to as ”targets”) and known capture oligonucleotide sequences (called ”probes” or ”receptors”), immobilized on a substrate by means of suitable surface biological functionalization procedures (hereafter referred to as ”bio–functionalization”). In recent years, micro-fabricated devices, normally called DNA micro-arrays [1], [2] have been developed for this type of analysis. Micro-arrays offer significant advantages compared with conventional laboratory techniques, in that they allow large parallelism (i.e. a large number of tests can be performed at the same time) and are relatively simple to use in a laboratory. Furthermore, very little amounts of sample material and bio-chemical reagents are required. So far, in these devices DNA hybridization is detected by means of optical marker-molecules previously attached to the DNA targets. At present, DNA chips based on optical detection of the hybridization of labeled DNA [3] and capable

1-4244-1245-5/07/$25.00 ©2007 IEEE

of testing hundreds of thousands different probes in parallel are technologically mature devices already on the market. The targets are labelled with chromophores molecules and deposited on the array surface. Successively, all this material is removed and only the targets that have hybridized with complementary probes remain immobilized on their specific sensing sites. Finally, the device is observed optically to localize the chromophores, revealing where hybridization has taken place. Often, the resulting image is not immediately recognizable, because hybridization can involve many adjacent sites resulting in a complex optical pattern to be resolved by means of adequate image processing. This technology presents two main drawbacks: cost of the instrument to detect and resolve the optical signal; necessity of a labeling step, with the need of additional reagents and the possibility of sample pollution. This method, however, has several significant drawbacks, since: a) it requires preventive target manipulation to introduce the optical markers; b) requires expensive instrumentation for optical reading; c) does not allow real-time detection; d) can be limited in sensitivity by the non-homogeneity of marker distribution. Therefore, alternative approaches are being studied and, in this context, new label free approaches have been proposed. To eliminate the former problem, methods based on the generation of electrical signals upon hybridization have been developed. However, some of these techniques are not labelfree, as they use mediator elements to generate the electrical signals [4-7]. Morever, some innovative microarrays employ electrochemical labels resulting in an electrical current through sensor electrodes during readout in case of hybridization events ([8-10]). Naturally, label-free approaches featuring fully electrical reading techniques represent in principle the simplest, most direct, hence also best solution. Recently, a number of approaches have been proposed to this purpose based on mass changes ([11-15]) or electrical properties of electrode/solution interfaces induced by DNA hybridization ([16-19]). In particular, a simple label free approach based on electrode/solution interface has been investigated. This phenomenon has been studied extensively in the past, using a conventional electrochemical three electrodes system, by applying a 50 mV potentiostatic voltage step and measuring the electrode current [20]. In [21], biofunctionalized gold electrodes were characterized with impedance spectroscopy. Moreover, a simplified set up based on two gold electrodes and capacitance measurements has been proposed exploiting only micofabricated electrodes and standard instrument [22]. Moreover, an active CMOS sensor chips is required in applications where a large number of

analysis has to be performed in parallel. Then, not only the sensitivity of the single sensor sites is increased, but also the electrical interconnect to the outside world is reduced: two works have shown results in this direction ([23] [24]). Nevertheless, problems related to reproducibility and stability of measurements indicates that application out of research laboratories will still need much efforts. Optical detectors have been proposed too. In particular in this type of sensors some optical properties are modified by DNA hybridization with targets placed on some kind of optical surface. As a result intensity [25] or frequency response [26-27] of the structure can be changed depending on the sensor structure and characteristics. Finally, floating gate (FG) CMOS devices have been proposed as hybridization detectors [28]. In this work DNA hybridization on a part of the FG of a CMOS non volatile memory device can induce a threshold shift due to the electric charge associated to the DNA molecules that can vary in number and mass by effect of the reaction with targets fixed on the polysilicon FG exposed surface. In the following sections we describe a new DNA hybridization sensor based on UV detectors devices that presents significant differences with respect to those presented before. In particular DNA selective hybridization takes place on a quartz slide separate fom the sensor. In addition, FG devices are used as UV sensors. DNA hybridization system working principle At system level, our approach envisages an array of “sensing” sites, each consisting of a bio-layer of specific probes, positioned between an external UV source and a UV detector. The bio-layer is conventional, in that it is obtained by means of DNA probes, aligned with the underlying sensor and working as “selective glue” with respect to the DNA target molecules. Fig. 1 gives a schematic representation of the experimental single site detection system. As can be seen, DNA probes are not attached to the sensor but to a quartz support optically aligned to the UV detector.

Fig. 1. Experimental setup

In operation, all sites are simultaneously exposed to the targets molecules, that will selectively hybridize only with complementary probes. Since DNA has a significant UV absorption, the sites where hybridization has taken place will

present different UV absorption with respect to those where the same amount of DNA is in the non hybridized state. The radiation transmitted through such sites will be smaller in the first case, due to masking effect of the hybridized DNA. As already explained, the recognition of this difference implies that of complementary molecules. Among the advantages of this new approach the most relevant are: a) hybridization area is separate from the UV detector thus does not need to be cleaned or disposed after use, b) no pretreatment of the biological material to be analyzed is required, in particular no addition of optical markers is needed, c) UV detectors presented in this work are fabricated with standard CMOS technology and can be dimensionally scaled to low dimensions for microarray fabrication. In this work we focus on an implementation aimed at high performance devices, exploiting all the advantages of silicon integration. In particular, both the UV sensors and the electronics needed to address and read the individual sites could be integrated on the same chip and the resulting devices would be extremely compact, fast and suitable for highdensity arrays of sensing sites. In particular this work investigates the use of NV memory cells as UV sensors for DNA detection/recognition and describes the experimental characterization of single-poly EEPROM transistors, shown to be particularly suitable for the purpose. Because of the original solution adopted for the UV sensors, this paper represents a significant advance in the state-of-the-art in the field of DNA sensors/ detectors, with possible extension to protein analysis. CMOS FG devices as UV detectors The key point of this work is the use of NV memory cells as UV sensors. As known, these cells can be erased by means of UV radiation, particularly if designed so that their Floating Gate (FG) can be directly exposed to the incoming radiation. Since during erasing the cell threshold voltage shift (∆VTH = VTH0 -VTH , where VTH and VTH0 denote the present and initial value of the threshold voltage, respectively) increases with the UV dose, NV cells represent almost ideal UV dosimeters that can be exploited for the purpose of this work. Therefore, the DNA chip envisaged in this work features an array of suitable NV memory cells, all initially programmed at the same threshold voltage VTH0 (>> VTHN where VTHN is the cell threshold voltage with no charge on the FG). Each cell (or group of cells) is located below a specific bio-functionalized layer. After exposure to the target DNA the whole array is exposed to UV radiation for a fixed time. Consequently all cells are partially erased, but those covered by hybridized targets receive less radiation, hence exhibit a final higher value of VTH. (i.e a smaller ∆VTH ), compared with those where non DNA binding has occurred. Of course, all the electronics needed to individually select and read the cells in the array are essentially the same as in (multi-level) memories, hence can be integrated on the same chip as the cells. For the purpose of this work “single-poly EEPROM devices” are suitable for this application. In fact, in this technology, the CG is realized by means of a n+ diffusion under the FG, that extends outside the channel area (Fig. 2).

This feature has some very important benefits, namely: a) the FG is completely exposed to U.V. radiation (i.e. no “masking” CG is present); 2) the FG area is larger than normal (thus more UV radiation can be collected); 3) the cell can be easily fabricated by CMOS technology. In order to characterize system effectiveness liquid DNA solutions in quartz containers have been “superimposed” to memory cells. In this way DNA concentrations can be controlled with great accuracy. The DNA “sensing site” used in this work is shown in Fig. 2.

For each DNA concentration measurement is repeated 5 times to evaluate the standard deviation, hence the measurement reliability. Resulting data present very low standard deviation, thus no error bars have been plotted. Experimental results The results described in this Section are shown as a function of UV dose (a parameter equivalent to, but more appropriate than exposure time). Fig. 3 shows the typical behavior of ∆VTH as a function of the UV dose and, as can be seen, a difference is measured for different values of single-stranded DNA concentration. Similar results were obtained with double stranded DNA.

∆Vth [V]

2 1,6

Buffer solution TE 1x (1,5 ml) DNA - 900 nM single stranded DNA - 1950 nM single stranded

1,2

DNA - 3600 nM single stranded DNA - 5400 nM single stranded

0,8

Fig. 2. DNA detection principle 0,4

A quartz container with the DNA targets is placed between a UV lamp and the NV memory cell used as a dosimeter for the impinging UV radiation. A mechanical shutter allows to switch on and off the incident UV radiation. As for recognition of specific DNA sequences, the experiments of this work exploit the so called hypochromic effect, namely the fact that for the same number of DNA bases (nucleotides), the UV absorption of double-stranded molecules is smaller (by about 20%) than that of the single stranded form. Of course, for the purpose of this work a major question concerns the sensitivity of NV cells used as UV detectors and, from this point of view, a suitable device has to be used. Experimental Setup In the experimental set-up used in this work the source of UV radiation is a Xenon lamp characterized by a spectral distribution with high emission values in the range of interest (250 to 270 nm). The lamp filament is controlled by a special circuit which maintains the supplied power constant. However, since this does not guarantee radiation stability over long time periods, a specific circuit (based on commercial UV sensitive photodiode) has been used to measure the instantaneous radiation and compute the total dose. The sample data are postprocessed by a LabVIEW program in order to evaluate the radiation dose of the experiments. The same program controls the shutter and switches off the radiation once the dose has reached a desired value. The EEPROM VTH is measured, when the light source is disabled, by means of a HP4156 Semiconductor Parameter Analyzer (SPA) driven by the control PC. Resolution in VTH measurements is 1 mV.

0 0

50

100

150

200

250

Dose[A.U]

Fig. 3. Variation of EEPROM cell threshold voltage shift as a function of the UV dose for the case of single-stranded (i.e. non hybridized) molecules.

Since the significant parameter to be analyzed is the difference between the buffer solution and that containing a certain concentration of DNA, we introduce a new parameter denoted as Under Erasure (U.E.) defined as U.E. = ∆VTH_BUFFER - ∆VTH_DNA., where ∆VTH_BUFFER and ∆VTH_DNA represent ∆VTH measured in the case where only the buffer or the buffer containing DNA is interposed between the UV source and EEPROM cells, respectively (in this definition, of course,

∆VTH_BUFFER

merely represents a convenient common reference). The most relevant results obtained are illustrated in Fig. 4, where the experimental points are drawn in such a way that each value of concentration on the x-axis corresponds to the same number of single-stranded DNA bases (i.e. the elements actually absorbing UV radiation), so that the difference among the various curves is due only to whether or not the DNA is in the bound (i.e. hybridized) state. In particular, the points indicated with squares and diamonds represents the case where the same number (N) of DNA single strands are presented in double and single-stranded form, respectively. As can be seen, the measurements allow to easily recognize the case of hybridization, that is the key for the DNA sensor envisaged in this work.

Acknowledgments The authors are grateful to ST-Microelectronics (Milan, Italy) for providing the devices used in this work..

200 0,2 ∆Vth for DNA single stranded ∆Vth for DNA double stranded

U.E. [mV]

160 0,16

120 0,12

80 0,08

40 0,04

0 0

1000

2000

3000

4000

5000

6000

Base concentracion [nM]

Fig. 4. U.E. for the case of hybridized and non- hybridized DNA, lower and upper curves, respectively.

Discussion The results presented before clearly indicate that a DNA microarray based on the approach suggested in this work and realized in a single chip with standard CMOS technology seems possible, since (at least) single-poly EEPROM cells represent a suitable UV detectors. The results of Fig. 4 indicates that double and singlestranded molecules in 1 cm path length can be distinguished starting from a minimum concentration of 1.5 µM bases (1.5 nmoles/cm2). In the case of micro-fabricated devices, the height of the container should of course be reduced, for instance down to 100 µm, hence for the same number of bases along the optical path the DNA concentration should be increased to 15 µmoles/cm2. Thus, in a microwell containing for example 106 µm3 of solution, the absolute number of bases needed to distinguish between hybridized and non hybridized samples is 1.5 nmoles, a quantity definitely compatible with that usually employed for analysis. Conclusions This paper provides a significant contribution toward the realization of innovative DNA sensors fabricated in a single chip with standard CMOS technology and based on UV absorption of DNA molecules (not yet used for microfabricated devices). At this regard, the paper proposes to use Non Volatile EEPROM cells, that represent almost ideal UV dosimeters. More specifically, the suggestion is made to use single-poly memory cells that can be fabricated with standard CMOS technology, and can be easily integrated with all the required electronics (for addressing and reading) and present a number of significant advantages. The experimental results described in the paper indicate that such devices have adequate sensitivity to distinguish between cases where DNA (unknown) target molecules hybridize with (known) probe sequences and those where this is not the case, clearly the essential point for the use envisaged in this work.

References 1. M. C. Pirrung, “How to make a DNA chip,” Angewandte Chemie International Edition, vol. 41, pp. 1276–1289, 2002. 2. D. Gershon, “Microarray technology. an array of opportunities,” Nature vol. 416, pp. 885–891, april 2002. 3. G. Kennedy, H. Matsuzaki, S. Dong, W. M. Liu, J. Huang, G. Liu, X. Su, M. Cao, W. Chen, J. Zhang, W. Liu, G. Yang, X. Di, T. Ryder, Z. He, U. Surti, M. S. Phillips, M.T.Boyce-Jacino, S. Fodor, and K. Jones, “Large-scale genotyping of complex DNA,” Nature Biotechnology, vol. 21, no. 10, pp. 1233– 1237, 2003. 4. F. Hofmann, A. Frey, B. Holzapfl, M. Schienle, C. Paulus, P. Schindler-Bauer, D. Kuhlmeier, J. Krause, R. Hintsche, E. Nebling, J. Albers, W. Gumbrecht, K. Plehnert, G. Eckstein, and R. Thewes, “Fully electronic DNA detection on a CMOS chip: device and process issues,”in Proceedings of IEEE International Electron Devices Meeting, San Francisco (USA), 2002. pp. 488-491. 5. R. M. Umek, S. W. Lin, J. Vielmetter, R. H. Terbrueggen, B. Irvine, C. J. Yu, J. F. Kayyem, H. Yowanto, G. F. Blackburn, D. H. Farkas, and Y.-P. Chen, “Electronic detection of nucleic acids. a versatile platform for molecular diagnostics,” Journal of Molecular Diagnostics, vol. 3, pp. 100– 108, may 2001. 6. L. Moreno-Hagelsieb, P. Lobert, R. Pampin, J. Remacle, and D. Flandre, “DNA detection based on capacitive Al2O3/Al microelectrodes,” in Proc. IEEE International Conference on Micro Electro Mechanical Systems (MEMS’04), Maastricht, The Netherlands, Jan. 2004, pp. 308–311. 7. Shu-Jen Han et al. “CMOS integrated DNA Microarray Based on GMR Sensors“ , Proceedings of IEEE International Electron Devices Meeting, San Francisco (USA), Dec. 2006. pp. 1-4. 8. Motorola Life Sciences Inc - www.motorola.com. 9. C. Yu, Y. Wan, H. Yowanto, J. Li, C. Tao, M. D. James, C. L. Tan, G. F. Blackburn, and T. J. Meade, “Electronic detection of single-base mismatches with ferrocene-modified probes,” J. Am. Chem. Soc., vol. 123, no. 45, pp. 11 155–11 161, 2001 10. M. Schienle, C. Paulus, A. Frey, F. Hoffmann, B. Holzapfl, P. Schindler-Bauer, and R. Thewes, “A fully electronic DNA sensor with 128 positions and in-pixel A/D conversion,” IEEE journal of SolidState Circuits, vol. 39, no. 12, pp. 2438–2444, 2004. 11. R. Gabl, E. Green, M. Schreiter, H. D. Feucht, H. Zeininger, R. Primig, and D. Pitzer, “Novel integrated FBAR sensors: a universal platform for

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

bioand gas detection,” Proc. IEEE Sensors, vol. 2, pp. 1184 – 1188, 2003. R. Brederlow, S. Zauner, A. L. Scholtz, K. Aufinger, W. Simburger, C. Paulus, A. Martin, M. Fritz, H.-J. Timme, H. Heiss, S. Marksteiner, L. Elbrecht, R. Aigner, and R. Thewes, “Biochemical sensors based on bulk acoustic wave resonators,” IEDM Tech.Digest, pp. 992–994, 2003. Y. Li, C. Vancura, C. Hagleitner, J. Lichtenberg, O. Brand, and H. Baltes, “Very high q-factor in water achieved by monolithic, resonant cantilever sensor with fully integrated feedback,” Proc. IEEE Sensors, vol. 2, pp. 809–813, 2003. K.-U. Kirstein, Y. Li, M. Zimmermann, C. Vancura, T. Volden, W. H. Song, and J. Lichtenberg, “Cantilever-based biosensors in CMOS technology,” Proc. DATE05, vol. 2, pp. 210–214, 2005. Xiaomei Yu et al., “Electric-Field Assisted Immobilization and Hybridization of DNA Oligomers on Microcantilever Sensors“, Proc IEEE 1st Intl. Conf. On Nano/Micro Engineered and Molecular Systems, Zhuhai, China, Jan. 2006, pp.505-508. F. Uslu, S. Ingebrandt, S. Bcker-Meffert, D. Mayer, M. Krause, M. Odenthal, and A. Offenhusser, “Label-free fully electronic nucleic acid detection system based on a field-effect transistor device,” Biosensors and Bioelectronics, vol. 19, no. 12, pp. 1723–1731, 2004. J. Fritz, E. B. Cooper, S. Gaudet, P. K. Sorger, and S. R. Manalis, “Electronic detection of DNA by its intrinsic molecular charge,” Proceedings of the National Academy of Science, vol. 99, no. 8, p. 14142 -14146, 2000. E. Souteyrand, J. Cloarec, J. Martin, C. Wilson, S. Lawrence, I.and Mikkelsen, and M. Lawrence, “Direct detection of the hybridization of synthetic homo-oligomer DNA sequences by field effect,” J. Phys. Chem. B, vol. 101, pp. 2980–2985, 1997. F. Perkins, S. J. Fertig, K. A. Brownt, D. McCarthy, L. Tender, and M. Peckerar, “An active microelectronic transducer for enabling label-free miniaturized chemical sensors,” in Technical Digest. International Electron Devices Meeting (IEDM’00), San Francisco, USA, Dec 2000. pp. 407-410. C. Berggren, P. Stalhandske, J. Brundell, and G. Johansson, “A feasibility study of a capacitive biosensor for direct detection of DNA hybridization,” Electroanalysis, pp. 74–84, 1999. R. P. Janek, W. R. Fawcett, and A. Ulman, “Impedance spectroscopy of self-assembled monolayers on Au(111): Evidence for complex double-layer structure in aqueous NaClO4 at the potential of zero charge,” J. Phys. Chem. B, vol. 101, pp. 8550–8558, 1997. C. Guiducci, C. Stagni, A. Fischetti, U. Mastromatteo, L. Benini, and B. Ricc`o, “Microfabricated two-terminal capacitors for label-free

23.

24.

25.

26.

27.

28.

DNA sensing/recognition,” IEEE Sensors, in press 2006. C.Stagni; C.Guiducci; L.Benini; B.Ricco; S.Carrara; C.Paulus; M.Schienle; R.Thewes” A Fully Electronic Label-Free DNA Sensor Chip” Vol. 7, Issue 4, April 2007 pp. 577 – 585. C.Stagni C.Guiducci,; L.Benini,; B.Ricco,; S.Carrara,; B.Samori, C.Paulus,M.Schienle, M.Augustyniak, R.Thewes, “CMOS DNA Sensor Array With Integrated A/D Conversion Based on Label-Free Capacitance Measurement “ Vol.41, Issue 12, Dec. 2006, pp. 2956 – 2964. Y. Wang et al.,”A Single Chip Micro-DNA-Array System Based on CMOS Image sensor Technology”, IEEE ISCAS 2006, pp. 2137-2140. Chryssis et al.,”Detecting Hybridization of DNA by Highly Sensitive Evanescent Field Etched Core Fiber Bragg Grating Sensors”, IEEE Jourl. Of Selected Topics in Quantum Electronics, Vol. 11, NO. 4, July/August 2005, pp.864-872. T. Tokuda et al., “A CMOS optical/potential image sensor with 7.5 µm pixel size for on-chip neural and DNA spot sensing”, Proc. of the 27th IEEE Engineering in Medicine and Biology Conf., Shanghai, China, Sept. 2005, pp.7269-7272. M. Barbaro et al., “A CMOS, Fully Integrated Sensor for Electronic Detection of DNA Hybridization”, IEEE Electron Device Letters, Vol. 27, NO. 7, July 2006. pp.595-597.

Chemical Design, Synthesis and Thin Film Supramolecular Architecture for Advanced Performance Chemo- and Bio-Sensing Organic Field Effect Transistors

G.M. Farinolaa*, L. Torsia, b, F. Nasoa, c, P.G. Zambonina, L. Vallid, M.C. Tanesea, O. Hassan Omarc, G. Giancaned, F. Babudria,c, F. Palmisanoa a Dipartimento di Chimica Università degli Studi di Bari - Bari (Italy); b Centro di Eccellenza TIRES - Università degli Studi di Bari - Bari (Italy) c CNR ICCOM, Dipartimento di Chimica - Università degli Studi di Bari - Bari (Italy); d Dipartimento di Ingegneria dell’Innovazione - Università degli Studi di Lecce - Lecce (Italy); *

Dipartimento di Chimica Università degli Studi di Bari - (Italy) 4, via Orabona 70126 Bari-Italy email: [email protected]; phone: 0039 080 5442076

Abstract Organic thin film transistor (OTFT) sensors are capable of fast, sensitive and reliable detection of various classes of chemical and biological analytes with high selectivity, and display the additional advantage of being compatible with plastic electronic technologies. Their distinctive versatility is based on multilevel control of the properties, from molecular design up to device architecture. Here Phenylene-thiophene based semiconductors functionalized with bio-molecules have been synthesized to be used as active layers in sensing OTFTs. These materials, indeed, combine the recognition capability of bio-molecules with the electronic properties of the conjugated backbone. The resulting OTFTs have been used to perform chemical recognition of citronellol obtaining detection limit in the ppm range. Introduction Organic semiconductor thin-film transistors, OTFT, have seen a dramatic improvement of their performance in the last decade and recently they have been also exploited as gas sensors. These are semiconducting organic-based sensors that offer the advantage of remarkable response repeatability as standard deviations are within 2% for several hundreds subsequent exposure to an analyte [1]. The devices can be operated in the pulsed mode and full base-line recovery can be achieved, operating the OTFTs at room temperature, by strategic use of the gate bias. It has been also shown that it is possible to enhance OTFT response by properly choosing the imposed gate bias potentials. Selectivity and specificity is being pursued by choosing ad hoc chemically or biologically functionalized semiconducting organic active layers. In this respect, interesting is the recent work showing that it is possible to operate OTFT sensors in water and to integrate them to microfluidics [2]. They have also been proposed as bio-sensors for lactic acid, glucose and streptavidin as well as large-area flexible pressure sensors for artificial skin applications [3-5]. Scaling down OTFT dimensions to the nanoscale may be another way to further improve their response. It is however not clear at this point if this will result in a sensitivity improvement. Cost is a key driver as well, particularly for consumer-oriented sensor systems. In these regards, the advantages of organic electronics are well known. OTFT sensor newborn technology can take full advantage of the rapid developments occurring in the field of organic electronics where OTFTs have been already implemented in complementary-metal-oxide-semiconductor (CMOS) circuits 1-4244-1245-5/07/$25.00 ©2007 IEEE

and in flexible plastic displays. High specificity and sensitivity in OTFT sensors are however still open issues. Here we report on OTFT gas sensor with bio-substituted Conducting Polymer, CP, active layer for possible use as biosensors. Results and discussion This work focus on novel high performance OTFT sensors developed in our laboratories for chemical and biological analysis. The semi-conducting oligomers chosen as OTFT active and sensing layers, are designed to combine field-effect and chemical recognition properties. They are implemented in a novel sensing OTFT structure. Alkoxyphenylene-thienylene conjugated systems have been chosen for the active layer as they allow to covalently attach a wide variety of molecules as side groups, including bio-receptors [6-7], exhibiting also field-effect properties when deposited by the LangmuirSchäfer (LS) procedure [8]. Langmuir-Shäfer (LS) deposition technique has been demonstrated, indeed, to allow a control of the molecular conformation beneficial to charge transport properties. The following preliminary results demonstrate that combining field-effect detection with chemical and biological CP recognition properties is a powerful tool that allows to broaden the recognition toward a variety of chemical and biological analytes in the ppm range. The OTFT sensors here fabricated have been used to detect citronellol molecule. The bio-functionalization of OTFT CP active layers seems then to induce recognition towards molecules of biological interest such as the chiral citronellol. An example of the experimental OTFT responses is given in Fig.1. The figure shows the Ids-Vg OTFT transcharacteristics in pure N2, dotted line, and during citronellol exposure, solid line. The signal, ∆I, is the Ids shift between the analyte and the N2 base line signals. It is interesting to note that for Vg bias higher than -40 V in the figure, the two curves show similar Ids values. On the contrary for Vg bias lower than -40V, in other words for gate bias below the threshold voltage, the two curves show an Ids separation which increases with the gate bias. This means that a non zero response is clearly visible only when the device is operated in the transistor regime.

ds

I (nA)

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0

40 20 0 -20 -40 -60 -80 -100

Vg(V)

Figure 1: Ids-Vg OTFT transcharacteristics in pure N2, dotted line, and during citronellol exposure, solid line. Developments The present work will be continued by exploiting the sensing and bio-chemical recognition properties of newly substituted OTFT active layers in order to extend the chemical recognition also to other classes of bio-molecules. Conclusions New OTFT sensors with bio-substituted CP active layers have been here demonstrated to be sensitive to chiral citronellol. This shows the potential of this class of sensors in the field of the chemical and biological analysis. The aim of this research is to develop new bio-chemical sensors each one with specific responses towards a series of analytes to be implemented also in e-noses array for recognition of complex mixtures. References 1. 2.

3.

4. 5. 6. 7. 8.

L. Torsi, and A. Dodabalapur, Anal. Chem. 70, 381A (2005). T. Someya, A. Dodabalapur, A. Gelperin, H.E. Katz, Z. Bao Integration and response of organic electronics with aqueous microfluidics. Langmuir 18, 5299-5302 (2002). Z.-T. Zhu et al. A simple poly(3,4-ethylene dioxythiophene)/poly(styrene sulfonic acid) transistor for glucose sensing at neutral pH. Chem. Comm. 13, 1556-1557 (2004). A. Star, J.-C. Gabriel, K. Bradley, G. Grüner, Electronic detection of specific protein binding using nanotube FET devices. Nano Lett. 3, 459-463 (2003). T. Someya, T. Sekitani, S. Iba, Y. Kato, H. Kawaguki, T. Sakurai, PNAS, 101, 9966, 2004. Babudri, F., Farinola, G.M., Naso, F. J. Mater. Chem. 14, 11 (2004). Naso, F., Babudri, F., Colangliuli, D., Farinola, G.M., Quaranta, F., Rella, R., Tafuro, R., Valli, L. J. Am. Chem. Soc. 125, 9055 (2003). Tanese, M.C., Farinola, G.M., Pignataro, B., Valli, L., Giotta, L., Conoci, S., Lang, P., Colangiuli, D., Babudri, F., Naso, F., Sabbatini, L., Zambonin, P.G., Torsi, L, Chem. Mat. 18, 778 ( 2006).

A New Integrated Kinematic Sensor for the Classification of Sit-to-Stand Locomotion Task Giovanni Costantini 1,2 , Massimo Carota 1 , Giovanni Maccioni 3 , Daniele Giansanti 3 1 Department of Electronic Engineering, University of Rome “Tor Vergata” Via del Politecnico 1, Rome, Italy 2 Istituto di Acustica “O. M. Corbino” Via del Fosso del Cavaliere, 100 - 00133 Roma - Italy 3 Tecnology and Healt department, The national Healt Institute Viale Regina Elena 299, Rome, Italy Abstract In this paper we introduce a new kinematic sensor to evaluate the ability to rise from a chair by means of the sit-tostand locomotion task. It is based on the analysis of the acceleration assessed by a homemade accelerometric transducer. Preliminary results show the feasibility of discriminating the rising from a chair fixed to different heights and the capability of distinguishing between pathological and non pathological parkinsonian subjects. Introduction Human functional ability/disability is evaluated by means of simple clinical tests. The main limitation of these techniques is that they are based on clinical trials with qualitative or partially quantitative observations of motor responses up to simple and well standardized short motor tasks such as stand-to-sit, gait-initiation, standing a stair, sitto-stand [1-3]. The ability to rise from a chair by means of the so called sit-to-stand locomotion task is for example of fundamenta1 importance for the quality of life because it is largely connected to the functional independence. The sit-to-stand task is largely considered as the more mechanically demanding functional task of the daily activities [4,5] and essential for gait [6]. In order to assess these motor tasks, quantitative measurements should be introduced in the evaluation process. Optoelectronic or ultrasound equipments are not adequate for their costs and encumbrance; furthermore, they require an great number of markers which constrain the investigated movement itself and suffer of shadowing effect. Kinematic sensors could be a valid aid to the functional study; in fact, they add the necessary quantitative measurement to qualitative observations. The motion analysis, performed by means of kinematic sensors, is based essentially on the use of accelerometer sensors that directly furnish motion acceleration. ACcelerometers (ACs) are so small and light to be easily connected to a body segment, without hindering the execution of the motor tasks. ACs can be combined together into a single accelerometric assembly to be positioned in a single device or used as isolated motion sensors affixed in different body positions, in order to provide an integrated, practical method for long-term, ambulatory, monitoring of human movement. Especially in the past few years, advances in miniature devices and a growing interest for non-invasive patient monitoring have promoted a huge development of the use of 1-4244-1245-5/07/$25.00 ©2007 IEEE

these sensors as it is well documented by Mathie et al. in their recent review [7]. They showed that ACs can be successfully used for the human continuous monitoring, for the gait analysis, sit-to-stand and stand-to-sit analysis, postural sway, fall risk. However few works using accelerometers for assessment of the sit-stand-sit movement have been reported. Sit-to-stand can be automatically divided into phases [8], as well as classified by identifying the sequence of two different postures, the sitting one and the standing one [9]. More recently, we showed that pure accelerometric systems such as the Morris [10] and Padgaonkar [11] architectures could not assure the feasibility of the trajectory reconstruction [12]: this was essentially due to the errors in eliminating the component of gravity acceleration. We also showed that architectures with accelerometers and rate gyroscopes assure the feasibility of trajectory reconstruction for short locomotor tasks [13]. In the present paper, we introduce a novel method for the classification of the sit-to-stand based on the analysis of the acceleration assessed by a homemade transducer. The accelerometric transducer The transducer [13] consists of three mono-axial accelerometers (3031-Euro Sensors, US) and three rate gyroscopes (Gyrostar ENC-03J-Murata, Japan), assembled together and relatively oriented according to an orthogonal reference system. Figure 1 shows the relative orientations of the sensors.

Z

1,2,3: Accelerometers

6

4

3 5

2

4,5,6: gyrostars

O1

Y

X Figure 1: Relative orientation of the sensors: 1, 2, 3 for accelerometers; 4, 5, 6 for gyrostars. The actual body segment angular velocity vector (ωx, ωy, ωz) is obtained by multiplying the relevant calibration matrix

by the gyrostar tern output vector. The orientation of the segment is determined by means of the orientation matrix [R],

which is calculated by solving the following differential matrix-equation.

6 ch 12 b, A/D

acceler.

Tx 433Mhz FM

µC gyrostar

Powering & connecting unit 8b, D/A

Sensor unit

Figure 2: Block diagram of the wearable device. ⎡ 0 dR ⎢ -1 R * =⎢ω dt ⎢ z ⎢⎣− ω y

−ω

z

0

ω

x

⎤ y ⎥ −ω ⎥ x⎥ 0 ⎥ ⎦

ω

It was found convenient to express the segment orientation in nautical angles: pitch, roll and rotation. The real instantaneous linear acceleration vector is obtained by equation. ⎡ ax ⎤ ⎢a ⎥ = [R] ⎢ y⎥ ⎢⎣ az ⎥⎦

⎡a x' ⎤ ⎢a ⎥ - g ⎢ y' ⎥ ⎢⎣ a z ' ⎥⎦

2nd order Butterworth Low Pass Filter with a cut-off frequency optimizable by means of a test bench of 14 HZ. Figure 3 shows the device, as well as the powering and connecting unit. The sensor circuitry has been split in two separate boards: one for the rate gyrostar tern and the other for the accelerometer tern. The circuitry, also shown in Figure 3, was assembled with surface montage technology, taking special care with the positioning of the sensors and their

Transducer

where ax’ ay’ and az’ are the acceleration vectors in the reference system solid with the device. All the algorithms, drift zeroing inclused, were tested using the Matalb R12 software tool (The Mathworks, USA). The Kinematic Sensor Figure 2 shows the block diagram of the wearable device It is composed of two separate units: a sensor unit and a powering, connecting unit. The latter contains the battery and the circuitry for communication at a 433 MHz FM radio frequency. The sensor unit have been kept very small (4x5x2 cm by 250 g of weight), in order to limit the encumbrance to the subject. The circuitry of the wearable device consists of 7 signal conditioning chains (3 for the triaxial accelerometer, 3 for the gyrostars, 1 for the thermal sensor), a microcontroller that features a 12 b A/D converter and an 8 b D/A converter. Each conditioning chains comprise an amplifier (G=10), a Voltage Controlled Voltage Source (Sallen-Key cell), and a

Data storing and trasmission

R-gys

ACs

stability, so as to ensure the proper functioning and reliability of the device.

Figure 3: The device and the powering and connecting unit with the detail of accelerometer and rate gyroscopes mounting. acceleration (g) 1

M

0.9 0.8

M2

0.7 0.6 0.5 0.4 0.3

M1

0.2

M3

0.1 0

forward bending

-0.1 -0.2

active raising

passive raising

-0.3

downward setting

m1

-0.4

m

-0.5 -0.6 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

time (s)

Figure 4: The phases of the sit to stand.

where [L] is the matrix of the imposed quantities and [S] is the matrix of the assessed quantities.

The sit-to-stand of five healthy subjects was recorded, with a sample period of 50 ms, during 4 second trials in three different conditions: A) height of the chair fixed to the 90 % of the feet to knee distance B) height of the chair fixed to the 100 % of the feet to knee distance C) height of the chair fixed to the 110 % of the feet to knee distance. Three trials were performed for each condition; the order of the trials was randomised. The same protocol above described was performed on 3 parkinsonian subjects (at first stage of pathology). Figure 4 shows the main phases of the component az of the real acceleration vector. For each trial, the value and temporal position of the main important points (Mi, mi), corresponding to the phases of the sit-to-stand were registered. In order to detect the timing of these acts, we used the algorithms developed in [14] and used also in [13]. We followed an investigative approach based on the distribution of the timing and acceleration amplitudes of the sit-to-stand. In particular, we determined the value and the position of the absolute maximum and minimum in the waveform in Figure 4 (M, m). Figure 5 shows the acceleration and angular velocity processing scheme.

Protocol and Investigation

Tests and Results

A dedicate Testing Equipment was developed in order to calibrate the equipment and for the test bench. Based on the DMC-1410 controller and a step-by-step motor with encoder [15], this equipment permitted an error lower than 2*10-2 deg in the interval of interest [13]. The instrument was calibrated in two phases: one static, the other dynamic. The calibration of the accelerometer tern was carried out during the static phase. The sensor unit was subjected to different g vectors by positioning each of its six faces on a horizontal plane; the output signals were averaged over a 6-second time interval. The dynamic calibration was done with the Testing Equipment. A plate on which the device is affixed was rotated by means of the testing equipment, in order to impose known rotational time laws with the necessary accuracy. Angular velocities ranging from 10 to 60 degrees/s in both directions were imposed for each of the three orthogonal axes. In both cases the calibration matrixes were computed by the least squares method. [CA] is the matrix for the accelerometers channels, [CRG] is the matrix for the rate-gyroscopes channel. Each one of these matrix was obtained solving the following equation [C]=[L][S]T([S][S]T)-1

As stated above, the sit-to-stand of five healthy subjects was recorded, with a sample period of 50 ms, during 4 second trials in the three conditions A, B and C. In Figure 6 we report, for the three different chair heights, the minimum and maximum acceleration peak absolute mean values (in g), assessed over all the 15 trials (three for each subject); in Figure 7, instead, we report the minimum and maximum acceleration time mean values (in seconds).

Figure 6: Maximum and minimum acceleration peak (healthy subjects): mean value and standard deviation of the mean.

t(s)

1.6 1.2 0.8

⎡ 0 −ω ω ⎤ z y⎥ dR ⎢ R-1* =⎢ ω 0 −ω ⎥ x⎥ dt ⎢ z ω ω − 0 ⎥⎦ ⎢⎣ y x

W(t)

0.4 90%

90%

100%

100%

110%

110%

chair height

Figure 7: Maximum and minimum acceleration peak time (healthy subjects): mean value and standard deviation.

AI(t)

A(t)=[R(θ, ϕ , ψ ) ]=R* AI - g a(g)

A(t)

0.8 0.6 0.4 0.2 0

TC=50ms TW=2s

Timing & Peack Revelation

m

M

m

M

m

M

90%

90%

100%

100%

110%

110%

chair height

Tstart, Tstop, T M, Tmi |A M|, | Am |

Figure 8: Maximum and minimum acceleration peak (healthy subjects): mean value and standard deviation of the mean.

Figure 5: Signal processing: the peack acceleration and timing revelation. As we can observe in Figure 6 and 7, both time and peak mean values are suitable to the discrimination between the three different chair heights. The protocol described above was also performed on 3 parkinsonian subjects (at first stage of pathology). Figures 8 and 9 report the same results for the pathological subjects, showing that the distributions are completely different.

t(s) 2.4 2 1.6 1.2 0.8 0.4 90%

a(g)

90%

100%

100%

110%

110%

chair height 1

0.8

Figure 9: Maximum and minimum acceleration peak (healthy subjects): mean value and standard deviation of the mean.

0.6 0.4 0.2 0 m

M

m

M

m

M

90%

90%

100%

100%

110%

110%

chair height

Discussion and Conclusions The clinical tests more frequently exploited in the ability/disability discrimination are based on simple tasks

which are not always conjugated with a fesibility and/or simplicity of the parameters assessment. In this paper we concentrated on a simple, largely used task (the sit-to-stand) and we used a wearable device and a simple postprocessing procedure in order to characterize the kinematic properties of the sit-to-stand act in different situations of the real life corresponding to different chair heights. Results showed the effectiveness of the developed procedure for the determination of the principal characteristics of the sit-to-stand locomotory task. Preliminary results also prove that the mean acceleration peak and mean timing values are well suited to the discrimination between the three different chair heights for the non pathological subjects; furthermore, the used method also shows the preliminary feasibility of discriminating between pathological and non pathological subjects. Another main aspect emerged from the analysis of the potentialities of this system that is a promising starting point in the development of a classification procedure to be used in the discrimination of many locomotory acts – sit-to-stand and standing the stair included – as well as to build up a powerful diagnostic tool for the identification of kinematic pathologies in their first stage. The next step will be the widening of the set of parameters involved in the investigation, through the inclusion of other important phases of the specific act (besides mi, Mi ), as well as by taking into account different chair heights. Obviously, a boost to the power of the investigations can be given if we derive the above mentioned parameters from frequency domain, instead of time domain. Nevertheless, the most interesting and promising improvement relies in the exploiting of an automatic classification methodology based on adaptive systems, such as Neural Networks [16]. Both supervised and unsupervised learning algorithms are well suited for this application. In future works, they will be investigated and applied to locomotory act classification. Neural classifiers can be exploited to process huge amounts of data, as well as to develop knowledge bases for the identification of particular pathologies (such as Parkinson disease) in their early stage. References 1. J. M. Winters and P. E. Crago, Biomechanics and Neural Control of Posture and Movement, New York: Springer 2000. 2. M. J Adrian and J.M Cooper, Biomechanics of human movement, Indianapolis IN. Benchmark, 1989. 3. J. C. Masdeu, L. Sudarsky and L. Wolfson, Gait disorders of aging, PA:Lippincott-Raven, 1997. 4. Kerr K. M., “Analysis of the sit to stand movement cycle in normal subjects”, Clinical Biomechanics 12, pp 236-45, 1997. 5. Kraly A. “Analysis of standing up and sitting down in humans: definitions and normative data presentation”, Journal of Biomechanics, 23, pp. 1123-38, 1990. 6. Munro B.J, “A kinematic and kinetic analysis of the sit-tostand transfer using an ejector chair: implications for rheumathoid arthritic patients”, Journal of Biomechanics, 263-71.

7. J. M. Mathie, A. C. F. Coster, N. H. Lovell, B. G. Celler, “Accelerometry: providing an integrated, practical method for long-term, ambulatory monitoring of human movement”, Physiological Measurement, 25, pp R1-R20, 2004. 8. J. M. Mathie, P. E. Crago, B.G Celler, “Detection of daily physical activities using a triaxial accelerometer”, Med Biol. Eng. Comp., 41, pp 296-301, 2003. 9. K. Aminian, M. Depairon, D. Hayoz “ Physical activity monitoring based on accelerometry: validation and comparison with video observation”, Med Biol. Eng. Comp., 37, pp 304-308, 1999. 10. J. R. W. Morris. “Accelerometry - a technique for the measurement of human body movements”, Journal of Biomechanics, 6, pp. 729-736, 1973. 11. A. J. Padgaonkar, K. W., King, A. I., “Measurement of angular acceleration of a rigid body using linear accelerometers”, ASME Journal of Applied Mechanics, 42, pp. 552-556, 1975. 12. D. Giansanti, V. Macellari, Maccioni G., and Cappozzo A. “Is it feasible to reconstruct body segment 3-D position and orientation using accelerometric data?”, IEEE Trans. on Biomedical Engineering, Vol 50 N. 4, pp 476-83, 2003. 13. D. Giansanti, G. Maccioni, V. Macellari “The development and test of a device for the reconstruction of 3d position and orientation by means of a kinematic sensor assembly with rate gyroscopes and accelerometers”, IEEE Trans. on Biomedical Engineering, Vol 52, n. 7, pp. 1271-1277, 2005. 14. D. Giansanti, V. Macellari, G. Maccioni, M. Paolizzi, S. Cesinaro, “A wearable device for measurement of kinematic parameters of sit to stand”, Proceeding of International Symposium on Biomechanics (Zurigo, 2001) 15. DMC-1400 Series Manual By Galil Motion Control, Inc 16. Bishop C..M. Neural Networks for pattern recognition, OXFORD University Press.

Gold Nanoparticle Sensors For Environmental Pollutant Monitoring E. Ievaa *, K. Buchholtb, L. Colaiannia, N. Cioffia, I. D. van der Werfa, A. Lloyd Spetzb, P.O. Källc, L. Torsia a Dipartimento di Chimica, Università degli Studi di Bari, Bari - Italy * tel. +390805442019, fax +390805442026, email: [email protected] b Division of Applied Physics, Linköping University, Linköping – Sweden c Division of Physical and Inorganic Chemistry, Linköping University, Linköping - Sweden Abstract Gold nanoparticles (Au-NPs) have been synthesised using a sacrificial anode electrolysis in the presence of tetra-alkylammonium halides, employed as cationic stabilizers. Catalytic NPs have been then deposited on top of Field Effect (FE) gas sensing devices and subjected to mild annealing procedures. Transmission Electron Microscopy (TEM) shows that the NP average core diameter is around 5 nm. X-Ray Photoelectron Spectroscopy (XPS) and Scanning Electron Microscopy (SEM) have been applied to the surface characterization of the annealed NP films used as active sensing layers. Morphological and spectroscopic results demonstrate that the annealed inorganic nano-clusters are finely dispersed and maintain a metallic oxidation state. Au-NPs can be proficiently employed as gate material in Si-Field Effect Gas Sensors. Preliminary results show interesting selectivity and sensitivity sensing features towards NOx detection.

1. Introduction Some of the most diffused air pollutants, nitrogen monoxide (NO) and nitrogen dioxide (NO2), collectively referred as NOx, have been of great concern in the past decades due to their adverse health effect and their abundance in the high-density urban areas. Many toxicological and epidemiological studies establish adverse health effects by NOx. They are believed to aggravate asthmatic conditions and to react with many atmospheric species, like oxygen and hydrocarbons, thus producing ozone, peroxy-nitrates and high oxidation state species (N2O5, HNO3 etc.); the latter, when dissolved in atmospheric moisture, contributes to generate acidic rain. [1] Therefore, great efforts have been taken to reduce the NOx emissions and the regulations have become stricter. For instance, the Kyoto Treaty, ratified by 54 Nations in 1997, call for a substantial world wide reduction of greenhouse gases and pollutants, including nitrogen oxide. In the light of such a dramatic problem, sensor devices for the monitoring of NOx have attracted large interest due to the possible environmental applications. Several types of NOx sensors have been employed, most of them are based on changes of conductance of metal oxide (SnO2, TiO2, WO3) film or organic materials (such as porphyrins) during adsorption on NOx. [2] Recently, a new field effect gas sensor based on a nano or micro-structured thermally evaporated gold film has been found to be sensitive to NO2; better results corresponding to thinner gates and smaller grain sizes. [3] This result could be explained considering that the gas sensor response depends on the surface reaction between the gate material and the gas molecules. Nano-structured films are, in principle, expected to exhibit an increased sensitivity as well as faster response and recovery time compared to 1-4244-1245-5/07/$25.00 ©2007 IEEE

corresponding bulk materials, due to the large surface-to-mass ratio. [4] The high sensitivity of this type of sensors could be explained by the large surface area arising from the adsorption on nanometer-size particles. The decreasing size effects potentially affect all the material properties; in particular, a pronounced dependence of the sensitivity on the particle size is expected with enhanced sensitivity towards smaller particle sizes. [5] Recent investigations have demonstrated the large potential of monolayer-protected nanoparticles for gas sensing applications. [6] An important feature of these materials is expected to be the possibility to tune their properties by varying the size of the particle core and/or the composition of the organic shell. The present study just deals with the controlled synthesis of gold nanoparticles to be employed as catalytically active gate material in NOx sensing devices. A large number of techiques has been used for the preparation of gold nanoparticles: among these, the wet chemical reduction of gold salts by means of citrate, [7] photochemical processes, [8] thermolysis, [9] and electrochemical depositions [10] can be cited. In the present work, an electrochemical route has been successfully employed for the controlled Au-NPs synthesis in presence of cation surfactants.

2. Gold nanoparticle synthesis and characterization. The electrosynthesis of core-shell Au-NPs was carried out by means of the Reetz’s sacrificial anode electrolysis. [11] A three electrode cell holding a gold sacrificial anode, a Pt cathode and a Ag|AgNO3 reference electrode, was used. The electrolytic solution was composed of a 0,1 M tetra-alkylammonium chloride solution in tetrahydrofuran (THF) / acetonitrile (ACN) mixed solvent (mixing ratio = 3:1). The salt behaved both as supporting electrolyte and as NP stabiliser. The cell was kept under a nitrogen atmosphere. Au-NPs have a core-shell structure with a nano-sized metallic core surrounded by a monolayer of surfactant. The morphology of the electro-synthesized materials can be tuned by acting on simple experimental parameters. For instance, the thickness of the NP shell depends on the length of the surfactant alkyl chains, [12] while the core size decreases as the electro-synthesis parameters become more extreme. [11] In the present study the electrolysis charge was fixed to around 300 mC and it was slightly adjusted from case to case, in order to give rise to identical concentrations of Au-NPs in the colloidal solutions. The Au-NPs were synthesised by fixing the corrosion voltage at +1V. A potential step was applied to the anode and the current was monitored throughout the synthesis. The solution assumed an intense red-wine colour, due to the plasmon resonance absorption of the gold nano-clusters.

Process conditions, employed in the case of tetraoctylammonium chloride (TOAC), are summed up in Table 1. Working potential (V) +1

Mean Current Density (mA/cm2) 1.5

Circulated charge (C) 300

[Au]s (mol/l) 0.20

Table 1. Typical electrosynthesis parameters employed for the potentiostatic preparation of Au-NPs. [Au]s = concentration of nanoparticles in the colloidal suspension; this parameter has been determined by the weight variation of the anode.

Transmission Electron Spectroscopy (TEM) characterization of the colloidal particles revealed that their average core diameter is equal to 5 nm (see Figure 1). The shell thickness is, in the case of TOAC, approximately 1.2 nm; thus the overall diameter of these core-shell Au-NPs is around 7 nm.

Figure 1. TEM characterization of colloidal gold particles stabilized by TOAC.

UV-vis analysis of the sample showed the typical plasmon resonance absorption peak falling at 520 ± 2nm. This is a well known size-dependent feature: tuning the core-size diameter resulted in a peak-shift.

3. Surface chemical composition and morphology of the nanostructured active layers Surface chemical characterization of the nanomaterials was performed by means of X ray Photoelectron Spectroscopy (XPS) using a Thermo VG Theta Probe spectrometer equipped with a micro-spot monochromatized Al Kα source. Both survey and high-resolution spectra were acquired in fixed analyzer transmission mode with a pass energy of 200 eV and 150 eV, respectively. Au-NPs were deposited by drop-casting on Si/SiO2 susbtrates. Afterwards, the material was heated up to 200°C

for 1 hour to reproduce the same thermal treatment used before the gas sensing measurements. Typical results of the XPS surface elemental analysis are reported in Table 2.

atomic percentage

C 34.6

Au 1.3

N 3.2

Cl ≤0.1

O 36.1

Si 24.7

Table 2: Surface chemical percentage recorded by XPS on annealed Au-NPs in the presence of TOAC at +1 V. Error on the atomic percentages is 0.1% for gold, 0.3% for the other elements.

Carbon and nitrogen are mainly due to TOAC residuals, this surfactant is being used in large excess during the synthesis. On the contrary, chlorine is almost completely removed by the thermal treatment. Oxygen and silicon signals are due to the exposure of the Si/SiO2 substrate. Neverthless, the partial removal of the organic component allows the exposition of an appreciable amount of nanostructured gold to the gaseous atmosphere. Chemical composition Wide scan and high resolution XP spectra of annealed AuNP films are reported in Figure 2. The Au4f region is exclusively composed by the metallic gold doublet (BEAu4f7/2= 83.7 ± 0.1 eV). Noteworthy, no peak ascribable to oxidised Au is detected. The C1s region shows the presence of two chemical environment, the main component (BE= 284.8 ± 0.1 eV) is attributed to aliphatic carbon; the second one (BE= 286.2 ± 0.2 eV) to the C-N species. These signals are in agreement with the use of alkyl-ammonium salt. N1s core level spectra is composed by two main peaks, falling at 401.8 ± 0.1 eV and 399.0 ± 0.2 eV, respectively. The former is attributed to quaternary nitrogen, while the latter is attributed to aminic species formed by the Hoffmann degradation of the ammonium salt during the heating treatment. Morphology Surface morfological study was carried out on annealed samples by means of a LEO 1550 VP Field Emission Scanning Electron Microscopy (SEM). A typical micrograph of TOAC-stabilised Au-NPs after the thermal annealing is reported in Figure 3. It can be observed that the annealed material still maintains the nano-structured character and the spherical morphology of the particles, even if a small increase in the particle size is cleary detected, the mean diameter ranging from 10 to 40 nm. This size-increase represents a slight limitation in the heating treatment. XPS shows that higher temperatures can be helpful in enhancing the surface availability of nano-gold; on the contrary, SEM shows that high-temperature treatments tend to give rise to marked increases in the particle size.

200 nm

Figure 3. SEM characterization of nano-structured Au active layer

4. Device structure and gas sensing setup The sensors used for the gas sensing measurements were capacitors with a catalytically active layer of Au-NPs as the gate material. [4]. The ohmic backside contact consists of evaporated, annealed Al. As semiconductor p-doped Si is used. On this semiconductor surface a thermal oxide of SiO2 is grown. Bonding pads of evaporated Cr/Au are deposited on the insulator. The sensor chip, a ceramic heater, and a Pt-100 element are mounted on a 16-pin holder and electric contacts are made from the sensors to the pins with gold bonding. A constant volume of gold nanoparticles was drop deposited on the insulator of the sensor, partly on top of the bonding pad. Afterwards, the solvent was evaporated by means of a thermal treatment at 200°C. The annealing treatment also has the potential to enhance the stability of the nanostructured film on the sensor, and it is always carried out at higher temperature than the one used during the gas sensing measurements.

Figure 2. XPS characterization of nano-structured Au active layer. From the top to the bottom: overall spectrum and high resolution C1s, N1s, Au4f regions.

Measurement setup During measurements the holders were mounted in aluminum blocks connected to a gas flow line. A computercontrolled gas mixing system was used to flow the test gases over the sensor surface. The measurements were performed using an MCM, (Multi Capacitor Meter) equipment, which makes it possible to measure the voltage, at a constant capacitance, of several devices simultaneously. The output signal of the device is the voltage across the structure at the constant capacitance while exposing the sensor to gases. During the measurements the sensors were exposed to pulses of 250 and 500 ppm of the following gases: NO, NO2, NH3, CO, and H2. The sensors were also exposed to 100 and 200 ppm pulses of C3H6. Nitrogen was used as the carrier gas with a background level of 10% O2 present.

5. Gas sensitivity The Au-NP capacitors were exposed to the following series of test gases: C3H6, CO, H2, NH3, and NO. The same test series was performed with NO2 replacing the NO. Each test gas pulse was 1200 seconds long with an equally long pulse of background gas before and after each test gas. In Figure 4 a graph representative of the sensor responses toward the test gases, at 150°C, is shown. The sensor responses towards NO were equal in size to the responses for NO2. The recovery time was not sufficient for the sensors to reach the baseline in between the pulses of different concentrations of NH3, and NO2 but it can be observed that the sensors give the highest response for NOx with a considerably smaller response towards NH3, and an even smaller response towards H2.

1.

2.

3. -0,1

-0,6

4. 500 ppm NH3

250 ppm NH3

250 ppm H2

500 ppm H2

500 ppm CO

-0,5

250 ppm CO

-0,4

50 ppm C3H6

Voltage [V]

-0,3

100 ppm C3H6

-0,2

5. 500 ppm NO2

6.

-0,7

250 ppm NO2

-0,8 0

5000

10000

15000

20000

25000

Time [s]

7. Figure 4. Sensor response of Au-NP capacitor at 150°C.

6. Conclusions. The sacrificial anode electrolysis was successfully employed to synthesised gold nanoparticle having a controlled core-shell structure in presence of tetraalkylammonium salt as stabilizer. The pristine AuNPs colloidal samples show an average core diameter of 5 nm. SEM analyses have revealed the spherical morphology of the annealed materials, even if an appreciable size increase occurs, due to the partial thermal degradation of the organic shell. XPS was used to assess the surface chemical composition of the annealed AuNPs, highlighting the exclusive presence of nano-structured metallic gold as catalytical species in the active layer. The nanoparticle sensor was able to reproducibly detect NOx, with some sensitivity also to NH3.

8.

Acknowledgments Dr. N. Ditaranto is gratefully thanked for the skilled assistance during XPS analyses. Work carried out with the financial support of the Marie Curie Early Stage Research Training Programm MEST-CT-2004-504272.

12.

9.

10.

11.

References Morrow, P.E. J., “Toxicological data on NOx: an overview”, Toxicol. Environ. Health, Vol. 13 (1984), pp. 205-207;(b) Blaise, G.A.; Gauvin, D., Gangal, M.; Authier, S., “Nitric oxide, cell signaling and cell death”, Toxicology, Vol. 208 (2005), pp. 177-192. (a) Zhuiykov, S.; Miura, N., “Development of zirconiabased potentiometric NOx sensors for automotive and energy industries in the early 21st century: What are the prospect for sensors?”, Sensors and Acuators B, Vol. 121 (2007), pp. 639-651, and ref. therein; (b) Richardson, T.H.; Dooling, C.M.; Jones, L.T.; Brook, R.A., “Development and optimization of porphyrin gas sensing LB films”, Advances in Colloid and Interface Science, Vol. 16 (2005), pp.81-96. Filippini, D.; Weiβ, T.; Aragon, R.; Weimar, U., “New NO2 sensor based on Au gate field effect devices”, Sensors and Actuators B, Vol. 78 (2001), pp 195-201. Salomonsson, A.; Petoral Jr, R.M.; Udval, K.; Aulin, C.; Käll, P.-O.; Ojamäe, L.; Sanati, M.; Lloyd Spetz, A. J. Nanoparticle Research, Vol. 8 (2006), pp. 866-910. Franke, M.E.; Koplin, T.J.; Simon, U., “Metal and metal oxide nanoparticles in chemiresistors: does the nanoscale matter”, Small, Vol.1 (2006), pp. 36-50. Hanwell, M. D.; Heriot, S. Y.; Richardson, T. H.; Cowlam, N.; Ross, I. M., “Gas and vapor sensing characteristics of Langmuir-Schaeffer thiol encapsulated gold nanoparticle thin films”, Colloids and Surfaces A: Physicochemical and Engineering Aspects, Vol. 284 (2006), pp. 379-383. Turkevitch, J.; Stevenson, P.C.; Hillier, “A study on the nucleation anf growth processes in the synthesis of colloidal gold”, J. Discuss. Faraday Soc., Vol. 11 (1951), pp. 55-75. Mallick, K.; Wang, Z.L.; Pal, T. J., “Seed- mediated successive growth of gold nanoparticles accomplished by UV radiation: a photochemical approach for sizecontrolled synthesis”, Photochem. Photobiol. A: Chemistry, Vol. 140 (2001), pp. 75-80. Nakamoto, M.; Yamamoto, M.; Fukusumi, M., "Thermolysis of gold (I) thiolate complexes producing novel gold nanoparticles passivated by alkyl groups”, Chem. Comm.,Vol. 15 (2002), pp. 1622-1623. Bartlett, P.N.; Birkin, P.R.; Ghanem, M.A., ”Electrochemical deposition of macroporous platinum, palladium and cobalt films using polystyrene latex sphere template”, Chem. Comm. Vol.17 (2000), pp. 1671. Reetz, T.M.; Helbig, W., “Size-selective synthesis of nanostructured transition metal clusters”, J. Am. Chem. Soc., Vol. 116 (1994), pp.7401-7402. Reetz, T.M.; Helbig, W.; Quaiser, S.T.; Stimming, U.; Breuer, N.; Vogel, R., “Visualization of surfactants on nanostructured palladium clusters by a combination of STM and High-Resolution TEM”, Science, Vol. 267, pp. 367-369.

A Two Electrode C - NiO Nafion® Amperometric Sensor for NO2 Detection A.Fort1, C. Lotti1, M. Mugnaini1, R. Palmerini2, R.Palombari2 , S. Rocchi1, L.Tondi1, V.Vignoli1 1 Dept. of Information Engineering – University of Siena via Roma 56, Siena 53100, Italy Phone: +39 0577 233608, Fax: +39 0577 233602 Email: [email protected], URL: http://www.dii.unisi.it 2 Dipartimento Chimica – University of Perugia

Abstract In this paper the authors propose a new and simple Nafion® based amperometric sensor and a dedicated measurement system and measurement protocol. The system has a linear response to NO2 concentration and a sensitivity up to 90nA/ppm. Moreover, the developed sensor shows a satisfactory repeatability and a low cross-sensitivity to some common interfering gases (such as CO and oxygen). Introduction The international industrial system has started to ask for reliable, fast, as well as portable measurement systems to keep under control dangerous environmental emissions such as the NOx combustion products (i.e. the exhaust gases with high contents of NO and NO2). For these applications, sensing systems based on metal oxide semiconductor sensors as well as on electrochemical sensors have been widely studied [1-7]. Nevertheless, several problems remains open, and the development of NOx measuring devices ensuring the required performance is still a research issue. In particular, semiconductor sensors present in general a high sensitivity but a low selectivity, and the crosssensitivity remains therefore an unsolved problem. As far as electrochemical sensors are concerned, potentiometric, amperometric and impedancemetric sensors have been proposed with a variety of materials for the solid electrolite (usually polymeric or ceramic ionic conductors, depending on the application) and for the electrodes, and a variety of electrodes geometry. [6][7]. Most of these sensors ensure a good sensitivity to NOx but often, again, a non negligible cross-sensitivity to other combustion products [6] such as CO, and to other gases including oxygen and water vapor. As a further issue, with the exception of the simplest nernstian sensors, an exhaustive knowledge of the chemical reactions responsible for the sensor behavior is still lacking, and therefore the sensors have to be experimentally characterized. Also the measurement protocol has to be experimentally tailored to the application, and its design is an important step of the development of the measurement system. In this work a new and simple two electrode amperometric sensor consisting of a Nafion® proton exchange membrane (PEM), a Graphite working electrode, and a NiO counter electrode is presented. The developed sensor presents a good sensitivity to NO2 and very small cross sensitivity to CO and O2. As a drawback, the sensor response is highly sensitive to 1-4244-1245-5/07/$25.00 ©2007 IEEE

both humidity and temperature (that have therefore to be controlled). A compact and flexible measurement system is also presented, together with a measuring protocol ensuring good performance in terms of repeatability. Method and Materials 1 Experimental The amperometric sensor described in this work is a two electrode cell made of a three layer stack: NiO/PEM/Graphite. The NiO layer acts as reference/counter electrode, whereas the graphite layer is the sensing electrode. The interface NiO/PEM was already investigated, and it was proved to posses a moderate reversibility; the exchange current density was found of the order of 2 µA/cm2, using Zirconium Phosphate as electrolyte [4]. This value allows its use as reference/counter electrode in electrochemical devices, provided that the current densities are maintained lower than some tens of nA/cm2 2 Construction of the sensor The NiO electrode is obtained by a pressed (40 kN/cm2) pellet (10 mm diameter, 0.15 mg) of a mixture of the oxide (Aldrich 99.99%) with 10% Nafion®. The powder to be pressed is obtained by drying in mild condition a mixture of the oxide with the proper amount of Nafion® solution (Aldrich w 20%) diluted in n.propanol. One side of the pellet is connected to the external circuit using silver glue, and it is fixed on an insulating support, while the other side is covered by a layer of Nafion®, that is obtained by drying at room temperature the above solution. This layer can be reinforced by a glass tissue imbued with it. The sensing electrode is prepared by mixing 20 mg of Graphite with 0.5 ml of a Nafion® solution diluted to 2.5 % w/w in n.propanol. The slurry is placed as a spot over the Nafion® layer and dried at room temperature. For this sensor the reactions of interests at the Graphite electrode are expected to be: 1 (1) H 2 O ↔ 2 H + + 2e + O2 2 NO2 + H + + e ↔ HNO2

(2)

where equation (2) is valid only in presence of NO2. At the NiO electrode, on the other hand, the following reaction must be considered: 2 NiO + H 2O ↔ Ni2O3 + 2 H + + 2e

(3)

3 Measurement set up The measurement system is sketched in figure 1. The sampling system is composed of a gas tank system feeding two PC controlled flow-meters (BronkHorst F-201C). Each controlled flow (carrier gas and mixture under test) passes through a Drechsel bottle (filled with water and placed in a thermostatic bath whose temperature is fixed, with 1°C accuracy), and through a 3-way valve. Also the 3-way valves are PC controlled: they allow switching each flow to the measurement chamber or to the ambient. The presence of the valves allow to reduce as possible the fluido-dynamic transients due to variations of the flowmeter set-points.

Figure 1: Measurement set-up. The system is composed by gas reservoirs, a flowmeter bench, a mixing control station and a front end electronics interfaced to a data acquisition and control system. The measurement chamber is placed in an oven kept at a reference temperature (1°C accuracy). If not stated differently, in what follows the experiments are performed keeping the sensor at 40°C. The thermostatic bath temperature is set at 25°C. The RH is also measured by a humidity sensor (Humirel HTS2330) inside the measurement chamber). An accurate humidity control is necessary when using Nafion® based sensors, because their behavior depends on the ionic conduction related on water content in the PME and on the H+ ion generation/combination phenomena taking place at both the sensor electrodes in presence of water [2],[6]. In other words, as it can be seen from eqs. (1) and (3), water provides at the electrodes an excess of H+ ions available for transport. Moreover water weakens the SO3--H+ bonds within the Nafion® structure, allowing the proton conduction phenomenon between the sensor electrodes (actually, a humidity excess can also induce flooding, which causes an anomalous sensor behavior [8]). The total proton current in the PME is in general the sum of two contributions: a diffusion current due to the different concentrations of H+ ions at the Graphite and the NiO electrodes (contribution whose sign and value vary as a function of the competition among the reactions in eqs. (1)(3)), and a drift current, that is present if a voltage difference is applied between the Graphite and the NiO electrodes (contribution whose sign and value vary with the sign and value of the applied voltage difference).

The developed measurement system is composed of the sensor, of a dedicated front end electronics (a I-V converter), and of a PC controlled acquisition and processing system. The system allows to apply a voltage difference between the electrodes in the range (-1V,1V), with a 1 mV accuracy. Moreover, it allows controlling, as mentioned before, both the flowmeter bench and the electro-valve system described in figure 1, and it allows measuring both the relative humidity (2% accuracy) and the temperature (by a J type thermocouple, 10-1 °C accuracy) in the measurement chamber. Summarizing, with this system it is possible to implement different measurement protocols: from tracking the sensor response to abrupt changes of the flow composition (with or without a voltage difference applied between the electrodes), to performing voltammograms, to evaluating sensor sensitivity as a function of different flows or different flow compositions. Results and Discussion The developed sensor was tested with different mixtures of nitrogen and NO2 with concentrations from 2 ppm to 10 ppm. The different mixtures were obtained by adding the two variable flows controlled by the two flowmeters in figure 1, keeping the total flow constant during a measurement. The measurement chamber temperature was kept at 40°C, and the Drechsel temperature was fixed at 25°C (RH =43%). In this work the current sign is chosen to be positive when protons move from the NiO to the Graphite electrode, that is, for currents entering the front end electronics. With the simple two electrode sensor arrangement it is of the utmost importance the selection of a suitable measurement protocol. In fact for this sensor the response to a constant concentration of NO2 eventually tends to a very small value due to the change of the electrode potentials with time. To obtain satisfactory measurements the average value of the current has to be kept low to maintain as constant as possible the average reaction potentials. To this purpose a pulsed measurement technique is used, i.e. the gas under test is delivered to the measurement chamber for short periods, during which the current rises, and after that the reference gas (humid nitrogen) is fed to the sensor in order to restore the reference conditions. To better illustrate this point, see the results shown in figure 2 and compare them with those of figure 3. In both cases the sensor was tested by keeping a constant total flow of 150 ml/min and delivering some pulses of nitrogen and NO2 mixtures with different concentrations and durations. The responses in figure 2 are obtained by flowing nitrogen for 20 minutes, the gas under test (nitrogen + NO2 , maximum NO2 concentration of 5 ppm.) for the following 4 minutes, and again nitrogen for the following 6 minutes. In figure 3 the sensor response to pulses with higher concentrations of NO2 followed by shorter recovery times in nitrogen is shown.

Figure 2: Two sensor response curves obtained in different days (black solid line and dashed red line) with a total flow of 150ml/min and with 2.3, 3.2, 4.2 ppm of NO2. It is possible to see in figure 3, that when the concentration rises over 9 ppm the baseline changes. This can be explained considering the reaction potentials (interface electrode/electrolyte) and the reaction expression: A+e ↔ D

(4)

where A is the reacting specie (acceptor), e is the electron and D is the donor. Actually the electrode potential (E) can be expressed using the following Nernst equation:

and where k1 and k2 are two reaction constants as in (5), and C0x indicates the equilibrium concentration of x. When the NO2 pulse is supplied, the working electrode potential grows higher than that of the NiO electrode, electrons entering the working electrode reduce NO2 to HNO2. At the same time NiO3+ ions are produced at the Nafion®-NiO interface and also the potential of this electrode starts to grow . The ratio CF grows with the concentration of NO2 and at the equilibrium, if the NO2 concentration is maintained, the two potentials tend to become equal and consequently the measured current tends to zero. In other words, the generation process of H+ ions taking place at the NiO electrode is no longer able to counterbalance the depletion of positive ions induced by the presence of the reacting species at the working electrode. When the NO2 pulse is removed, the NiO electrode has a higher potential with respect to the C electrode and this justifies a negative current. The excess of NiO3+ ions at the interface tends to diffuse slowly in the electrode bulk, and hence also the baseline is slowly recovered. This sets a specific limit to the use of such sensor, related to the NO2 maximum concentration. As a conclusion, if the NO2 pulse duration and the recovery time in humid nitrogen are not properly set as a function of the maximum NO2 concentration to be detected, both the baseline and the NO2 response amplitude drift. Nevertheless it was observed that for the concentration range of interest (lower than 10 ppm), the measurement protocol shown in figure 2 is a satisfactory trade-off in terms of repeatability and measurement time, even at higher sensor biasing voltages (up to –200 mV). x 10

E=

µ0 e

+

kT ⎛ C A ln⎜ e ⎜⎝ C D

⎞ ⎟⎟ ⎠

-8

23 ppm

12

(5)

17 ppm 10

Where CF is expressed by (7), ⎛ C 0 Ni 3+ CF = ⎜⎜ 0 2+ ⎝ C Ni

⎞ ⎟ ⎟ ⎠

(7)

14 ppm 8 I (A)

where µ0 is the standard chemical reaction potential, k is the Boltzmann constant, T is the temperature, e is the electron charge, and Cx is the concentration of the involved species x. Note that the use of such formulation implies a thermodynamic steady state. Nevertheless eq (5) can be used to approximately analyze the different experimental dynamic conditions. Initially, in the absence of NO2 at the equilibrium and with an external voltage difference applied across the electrodes of 0 V the sensor current is close to zero, that is the potentials of the two electrodes is equal (ENiO=EC, where EC is the working electrode potential, and ENiO is the counter electrode potential). The counter electrode potential can be written as: E NiO = k1 + k 2 ln(CF ) (6)

6

9 ppm

4

2

0 10

20

30

40

50 60 t (min)

70

80

90

100

Figure 3: Sensor response obtained with a total flow of 150ml/min and with 9, 14, 17 and 23 ppm of NO2 . In figure 4 the sensor responses (current peak values) obtained in the described experiments are shown as a function of the NO2 concentration, for different flows. Using the measurement protocol of Figure 2, the sensor response is found to be linearly related to the NO2 concentration (rms fitting error lower than 10%). The sensitivity varies from 4nA/ppm (50 ml/min) to 9 nA/ppm (200 ml/min) (see figure 5).

-8

These response variations are due both to the variation of the NO2 flow and to the variation of the amount of water present in the measurement chamber and available for conduction. Actually, as it can be seen in figure 6, with equal NO2 flow, the sensor response grows when decreasing the water flow. This can be explained by the competing water dissociation reaction occurring at the working electrode.

75 ml/min 6 100 ml/min 5

125 ml/min

4

150 ml/min

I (A)

x 10

x 10

-8

200 ml/min

3

200 ml/min 5

1

4 I (A)

2

150 ml/min

125 ml/min 0

100 ml/min

3

400

600 800 NO2 flow (nl/min)

75 ml/min

1000

Figure 6 Sensor response (peak values) obtained in the described experiments as a function of the NO2 flow, for different total flows.

2

1

-7

0

6

2

3

4 5 6 NO2 concentration (ppm)

7

x 10

8 5

Figure 4 Sensor response (peak values) obtained in the described experiments as a function of the NO2 concentration, for different flows.

4.2 ppm

4

7.7 ppm 3

6 ppm

-9

9

x 10

2 1

8

2.5 ppm

sensitivity (A/ppm)

0

7 -1 -0.5

6

-0.4

-0.3

-0.2

-0.1

0 0.1 bias (V)

0.2

0.3

0.4

0.5

Figure 7: Voltammograms performed with different concentrations of NO2 . Total flow 150 ml/min.

5 4 3 50

100

150

200

flow (ml/min)

Figure 5 Sensitivity to NO2 concentration as a function of the total flow. In figure 7 some voltammograms are reported for different concentrations of NO2. The sensor sensitivity becomes 90 nA/ppm at a bias voltage of -0.5V (total flow 150 ml/min). The higher sensitivity in this biasing condition has to be traded with a longer measurement time, as discussed before, and with a more complex measurement system. To conclude this investigation some experiments have been carried out with CO and oxygen as a first analysis of the sensor sensitivity to interfering gases. The experiments have been performed in the same manner as described before, with 0V biasing and 150 ml/min of total flow. The response peak is 4 nA for mixtures of nitrogen and 300 ppm of CO, and 5 nA for mixture of nitrogen and 3000 ppm of oxygen, showing sensitivities that are 2 and 3 order of magnitude smaller than that to NO2, respectively.

Conclusions A novel simple electrochemical two electrodes C/Nafion®/NiO amperometric sensor has been proposed and characterized. The detection limit of the sensor result close to 1 ppm, when a 150 ml/min constant flow of gas is used. A chemical sampling system and a front end electronics has been designed to acquire small current signals. The sensor showed good characteristics in terms of response repeatability, sensitivity, response time and cross-sensitivity to interfering gases, even with 0V biasing. References 1. Opekar, F., Stulík, K., “Electrochemical sensors with solid polymer electrolytes,” Analytica Chimica Acta, Vol. 385 (1999), pp. 151-162. 2. Hrncırova, P., Opekar,F., Stulık, K., “An amperometric solid-state NO sensor with a solid polymer electrolyte and a reticulated vitreous carbon indicator electrode,” Sensors and Actuators B, Vol. 69 (2000), pp. 199-204.

3. Opekar, F., Stulik, K., “Electrochemical sensors with solid polymer electrolytes, ” Analytica Chimica Acta, Vol. 385 (1999), pp. 151-162. 4. Palombari, R., Pierri, F. , “Ni(III) doped NiO as the electrode material for electrochemical devices employing protonic conductors,” Journal of Electroanalytical Chemistry, Vol. 433 (1997), pp. 213-217. 5. Alberti, G., Cherubini, F., Palombari, R., “Amperometric solid-state sensor for NO and NO2 based on protonic conduction,” Sensors and Actuators B, Vol. 37 (1996) pp. 131-134. 6. Fergus, J.W., “Materials for high temperature electrochemical NOX gas sensors,” Sensors and Actuators B, Vol. 121 (200), pp. 652–663. 7. Zhuiykov, S., Miura, N., “Development of zirconiabased potenziometric NOX sensors for automotive and energy industries in the early 21st century: What are the prospects for sensors?,” Sensors and Actuators B, Vol. 121 (2007), pp. 639-651. 8. Larminie, J., Dicks, A., “Fuel Cell Systems Explained” (Second Edition), Wiley, March 2006.

CHROMATOGRAPHIC SYSTEM BASED ON AMORPHOUS SILICON PHOTODIODES D. Caputo, G. de Cesare, C. Manetti*, A. Nascetti, R. Scipinotti Dept. of Electronic Engineering, University “La Sapienza”, via Eudossiana, 18 00184 Rome (Italy) *Dept. of Chemistry, University of Rome “La Sapienza”, Piazzale Aldo Moro 5, Rome (Italy) Abstract In this work, we present a novel thin layer chromatography system based on fluorescence detection by means of a linear array of amorphous silicon photodiodes. The photodiodes are optically coupled to a thin layer chromatography plate to monitor, in real-time, the separation of the components of a mixture during the chromatographic run. We designed a horizontal development chamber with UV transparent window and integrated eluent tank. The resulting system is extremely compact and ensures fast analysis using only small amounts of eluent. With our system, we analyzed different mixtures of fluorescent inks and other molecules such as fluorescein. Real-time data of sensors located in different position are used to detect the composition of the mixture. The various components can be determined from the time of the transit of the different species in front of the sensors. Furthermore, from the measured data it is possible to extract additional information as the transport properties of the stationary phase or the velocity profile along the run for each component of the mixture. Introduction Chromatography is a method of separating mixtures of two or more compounds [1]. The separation is accomplished by the distribution of the mixture between two phases: one that is stationary and one that is moving. Different compounds will have different solubility and adsorption to the two phases between which they are partitioned. Among the different chromatography methods the Thin Layer Chromatography (TLC) is a solid-liquid technique in which the stationary phase is silica gel, while the moving phase is constituted by solvents with different polarity [2, 3]. TLC involves spotting the sample to be analyzed near one end of a glass that is coated with a thin layer of silica gel. The glass substrate, which can be the size of a microscope slide, is placed in a covered jar containing a shallow layer of solvent. As the solvent rises by capillary action up through the adsorbent, differential partitioning occurs between the components of the mixture dissolved in the solvent along the stationary adsorbent phase. When the solvent front reaches the other edge of the stationary phase, the plate is removed from the solvent reservoir. The separated spots are typically visualized with a scanner by exciting their fluorescence with an ultraviolet light or by a suitable reaction procedure. In this work, we introduce a novel system for in-situ realtime analysis of the chromatographic run. The basic idea presented here is the integration of a set of photosensors on the TLC plate. This is possible by using the hydrogenated amorphous silicon (a-Si:H) technology that permits to deposit high-quality photodiodes on glass substrates [4].

1-4244-1245-5/07/$25.00 ©2007 IEEE

The structure of the proposed TLC plate with integrated photosensors is reported in Figure 1. One side of a glass substrate is covered with a layer of Indium Tin Oxide (ITO) that represents the transparent bottom-electrode of the photodiode. The photodiode is an amorphous silicon ntype/intrinsic/p-type stacked structure deposited by Plasma Enhanced Chemical Vapor Deposition (PECVD). The photodiode top-contact is a three metal Cr/Al/Cr stack. The structure is passivated with a 8 µm thick Cyclotene layer [5]. The other side of the glass substrate is coated with a 200 to 800 µm thick layer of a solid adsorbent (usually silica or alumina) that constitutes the stationary phase for the thin layer chromatography. By irradiating the TLC plate with a UV lamp during the chromatographic run it is possible to monitor, in real-time, the transit of the components of the mixture in front of the photodiode. Different components can be then distinguished from their arrival time. By using a linear array of photodiodes it is possible to observe the chromatographic run at different grade of separation of the mixture components. This gives the ability to optimize the development time in order to achieve the desired resolution of a given separation. This helps to shorten the analysis time by terminating the run when the components of interest have reached the desired degree of separation, eliminating the differences in homogeneity and surface activity from plate to plate. Furthermore, the width of each photocurrent peak, associated with the transit of a fluorescent component in front of a photodiode is related to the local velocity of that component leading to additional information as the transport properties of the stationary phase or the velocity profile along the run of each component of the mixture. As an additional advantage, the proposed system can be handled in the same way as the conventional plates from the spotting of the mixture to the possibility of scan with conventional systems. In particular, thanks to the low cost technology of the amorphous silicon, the TLC plates

UV EXCITATION Analyte

Stationary Phase

FLUORESCENCE

a-Si:H sensor

Figure 1: Schematic view of the TLC plate with integrated amorphous silicon photodiode

Figure 2: Prototype of the 16-pixel linear array of amorphous silicon photodiodes to be optically coupled to conventional TLC plates

presented here are disposable, making them attractive for practical applications. Experiment As a first step toward the development of the system described above, we deposited a linear array of amorphous silicon photodiodes on a glass substrate covered with a transparent conductive oxide. Then we optically coupled the sensor glass to a standard TLC plate. This modular approach allowed us to setup the entire system by tuning its single components individually. In Figure 2, a picture of the glass substrate with a linear

array of sixteen amorphous silicon photodiodes is reported. The pixel size (0.5x1mm2) and pitch (1 mm) were chosen according to the width of the chromatographic zone and the thickness of the glass substrate and of the TLC plate. The sensors are n-type/intrinsic/p-type a-Si:H stacked structures deposited by PECVD in our three chamber reactor. The deposition parameters have been chosen for minimizing the dark current and achieving a spectral response centered around 500 nm, to match the fluorescence emission of a wide range of substances including the fluorescein used in the following experiments. The current-voltage characteristic measured in dark condition is plotted in Figure 3. The dark current density is about 10-10 A/cm2 at -0.5 V.The quantum efficiency of the same sensor is reported in Figure 4. To test our system we designed a development chamber with a UV transparent quartz window to be able to illuminate the TLC plate during the chromatographic run. The chamber is made of PTFE (Teflon) to prevent chemical interaction with the moving phase (eluent). The development chamber includes the housing for the TLC plate and the sensor substrate ensuring an optimal optical coupling between them. The volume of the chamber is very small to guarantee fast saturation with the eluent vapors. Consequently, very small amounts (less than 5 ml) of eluent are needed to perform a chromatography run, reducing the costs and the risk associated with harmful eluents. A 10 ml eluent tank is included in the chamber. A drawing of the development chamber is reported in Figure 5. The experimental setup has been mounted on a optical bench. A mercury lamp filtered by a Jobin-Yvon H10 monochromator supplies the UV radiation at 254.3 nm. The incident power is 2.3 µW. The photocurrent of selected photodiodes is measured by a Keithley 236 Source-Measure unit. The development chamber is placed horizontally in the optical path and the incident radiation is focused with UV optics on the selected sensor. Results Experiments have been performed with several mixtures

1E-5

1.0

Quantum Efficiency

1E-6

Current (A)

1E-7 1E-8 1E-9 1E-10 1E-11 1E-12 1E-13 -1.0

-0.5

0.0

0.5

1.0

Voltage (V)

Figure 3: Current-voltage characteristic measured in dark conditions.

0.8

0.6

0.4

0.2

0.0 350

400

450

500

550

600

650

700

750

Wavelength (nm)

Figure 4: Quantum efficiency of the a-Si.H photodiode

TLC plate

Flex connector to readout electronics

Sensor substrate

Eluent tank

Figure 5: Prototype of the horizontal development chamber with UV transparent window and integrated eluent tank

of fluorescent molecules that have been spotted on conventional TLC plates. The same linear array has been used for all the experiments. In Figure 6 the experimental results achieved with a solution of a single fluorescent component (cumarine) is shown. The cumarine has been dissolved in water and ethanol has been used as eluent for the chromatography. The graph shows the real-time signal of a sensor located in the center of the array. The photocurrent peak detected after 6 minutes has to be ascribed to the transit of the cumarine in front of the selected photodiode. From system characterization we have estimated an average noise current around 0.1 pA [6], and then from figure 6 a very good signal-to-noise ratio can be

22 20 18

Current (pA)

UV Transparent window

appreciated. The tail observed after the transit of the cumarine is due to impurities contained in the solution. A second experiment has been performed on undiluted ink of a fluorescent marker using ethanol as mobile phase. In Figure 7a, the signal of a photodiode located in correspondence of the initial part of the chromatography run is reported. Although at this point the complete separation of the mixture is still not achieved, the system clearly identifies three fluorescent components present in the analyzed sample. The real-time signal is in agreement with the image of the developed TLC plate observed under UV illumination and reported in Figure 7b. The direction of the chromatographic run is from left to right in the picture. The fourth component observed in Figure 7b is not detected by the sensor because the corresponding band has still not reached the photodiode, whose position is indicated by the arrow in the figure. The rightmost peak in the picture corresponds to the photocurrent peak observed after 400 s. The higher velocity of this component with respect to the others is also evident from the narrower width of the corresponding photocurrent peak due to the lower transit time in front of the photodiode.

16 14 12 10 8 6 4 2 0 0

500

1000

1500

2000

2500

Time (s)

14

Current (pA)

12

a)

10

Selected sensor

8 6 4 2 0 0

200

400

600

800

1000

1200

Time (s)

Figure 6 Real-time signal of a photodiode during a chromatography run performed on a solution of cumarine.

b) Figure 7: a) Real time signal of a chromatography run performed on a solution of a highlighter ink. b) picture of the developed TLC plate under UV illumination. The position of the sensor, whose signal is reported in a) is indicated by the arrow

Conclusions An innovative TLC system based on the optical coupling between a thin layer chromatography plate and a linear array of a-Si:H photodiodes has been presented. The system results very compact thanks to a small, “ad hoc” designed horizontal development chamber, where a UV-transparent quartz window and the eluent tank are integrated. During the chromatographic run an UV radiation is impinging on the TLC plate and the photocurrent induced in the a-Si:H sensor by the fluorescence of the analytes can be monitored in real-time. Experimental results achieved by analyzing different mixtures demonstrate the suitability of our system for realtime chromatography. Acknowledgments Authors wish to aknowledge Ing. Domenico Pellegrino for helpful collaboration. References 1. Scott, R. P. W., Techniques and practice of chromatography, Marcel Dekker (New York, 1995). 2. Sherma, J., Fried, B. Handbook of Thin-Layer Chromatography, 2nd ed. Marcel Dekker (New York, 1996). 3. Hahn-Deinstrop, E.: Applied Thin-Layer Chromatography. Wiley-VCH Verlag, (Weinheim, 2000) 4. Street, R. A., Technology and Applications of Amorphous silicon, Springer (Berlin, 2000), pp. 147-156. 5. http://www.dow.com/cyclotene/index.htm 6. Caputo D., de Cesare G., Nascetti, A. Negri R.Scipinotti R., Amorphous Silicon Sensors for Single and Multicolor Detection of Biomolecules, in press on IEEE Sensor

A Readout Circuit in 0.35 μm CMOS Technology for Lab-on-a-Chip Applications. P. Delizia, S. D’Amico, A. Baschirotto. Department of Innovation Engineering, University of Salento, Italy pasquale.delizia/stefano.damico/[email protected]

Abstract. In this paper an integrated readout circuit for Lab-on-aChip applications is described. The system consist of a 320×320 array of capacitor sensors and actuators, the pitch of each site is 20μm. Sensors detect dielectric permittivity variation thanks to dielectrophoresis (DEP) process. Usually for this kind of applications the analog-to-digital conversion is carried out off chip but the major limitation in this case is the noise floor and then high signal to noise ratios are difficult to obtain. In fact, for these applications the noise floor specification (>10b) is more critical than linearity (≈8b). In order to reduce the amount of noise coupled to the signal at the chip pad and on the board, an on-chip analog-todigital conversion is here implemented. The electronic interface is composed by two main blocks: a pre-amplifier with programmable gain and an algorithmic analog-to-digital converter with a 1.5-bit/stage architecture. Each one is realized by fully differential switched capacitor technique. The proposed A/D converter has 11b resolution, a sampling rate of about 100ksample/s and an input full-scale range of 1Vp-p differential. Simulation results show a SNR=65.7 dB and an ENOB value of 10.6b. Readout chain is implemented in 0.35μm CMOS technology with a 3.3V supply voltage.

address

mux

1-4244-1245-5/07/$25.00 ©2007 IEEE

column decoder column circuits

control

row circuits row decoder

I. Introduction Significant research efforts in biology, chemistry and engineering have been recently aimed at pursuing the advantages of miniaturization for cheaper, better and faster sample analysis. Micro Total Analysis Systems (μTASs) were envisioned in the late 1980s as miniaturized, highly integrated chemical analysis systems. The advent of DNA microarrays, propelled by genomic research, captured the attention of researchers and investors alike. Although the field was generally indicated as that of biochips, the word “Lab-On-AChip” (LOAC) entered the jargon to differentiate between passive microarrays and microanalytical systems sporting some degree of integration, programmability or microfluidic capabilities. LOAC systems are typically based on a capacitor sensor array. The system under development consists of a CMOS chip covered by a conductive glass lid which is separated by the chip by about 100μm [1]. The aim is to detect variations in dielectric permittivity due to the presence of particles in the region above superficial electrodes, which affects the coupling capacitance with the lid. This is possible thanks to dielectrophoresis (DEP). In the system under development the elementary unit is replicated to form a 320×320 array where the pitch of each site is 20μm. Each microsite is supplied by an actuator circuit that senses the dielectric permittivity variation and provides an output voltage, Voarr. The expected

die size is 8×8mm2. A block diagram of the chip is shown in Fig.1, each microsite is addressed by specific control signals generated by the row and column decoders through row and column circuits. This paper describes the design of the electronics read-out channel. The main requirements regard the signal-to-noise level (in excess of 10b in order to detect any small event) and the area occupancy (to be minimized due to the large number of read-out channels). The linearity specification is less critical since the read-out input signal is provided by a source follower operating at low bias current. With respect to the previous realizations, the innovation of this design consists in realizing on-chip the A/D conversion. In the previous design the chip output signal was analog and then additional noise was collected at pad and board level. In this design, at the cost of design complexity and die area, the A/D conversion is onchip and this guarantees significantly larger noise immunity at pad and board level. The paper is organized as follow: in Section II the readout circuit architecture is reported; in Section III the pre-amplifier is described; in Section IV the A/D converter is reported, all blocks are described and noise considerations are reported. Section V reports simulations results and Section VI draws conclusions.

320x320

Voarr

sensor array

Figure 1. Block diagram of the chip. II. Readout chain architecture. The readout chain architecture is made up of a preamplifier and an A/D converter. The overall scheme is shown in Fig.2. Voutp

Voarr

pre-amplifier

A/D Voutm

Voff Gain

Figure 2. Readout architecture.

11 bits

Due to the mixed-mode chip nature of this design a fullydifferential philosophy was adopted. Fully differential circuits are less susceptible than their single-ended counterparts to common-mode noise, such as noise on the power supplies that is generated by digital circuits that are integrated on the same substrate as the analog circuits. The pre-amplifier amplifies the single-ended input voltage Voarr from the sensor array and provides a differential output voltage to the analog-to-digital converter. The sensor input signal may be unipolar. Input signals control Voff and Gain allowing pre-amplifier to obtain more sensibility and to optimally allocate the unipolar signal within the ADC input range. The analog-to-digital converter provides a resolution of 11 bits, it is realized by a fully-differential structure such as the pre-amplifier. The sampling frequency is about 100kHz. The input full-scale range of the A/D converter is about 1Vp-p differential.

The use of Voff doubles the output range: the differential output voltage can then even be negative (which would otherwise be impossible since the pixel reset voltage is always higher than the voltage at SIG1 phase). The doubled output swing can, thus, be used to increase the pre-amplifier gain. Another feature of Voff is to modify the unipolar input signal common mode voltage and allocate it within the ADC input range. To achieve the target signal-to-noise ratio > 10b, an accurate noise analysis of the circuit is proposed. At first only kT/C noise is taken into consideration and then the opamp noise contribution is analysed. During PHI1 the noise charge stored onto Ci, Coff and Cf is:

Qn2,Ci = kT ⋅ Ci Eq.2

III. Pre-amplifier The pre-amplifier schematic is shown in Fig.3. PHI2

Coff

Qn2,Cf = kT ⋅ C f

Cf

PHI2

Voff

During PHI2 the output noise due to the previous charge noise and switches active in this phase is about:

PHI1 Vcm_out

Ci

SIG1

PHI1

Voarr

PHI2

Vcm_out

-

PHI1

+

Voutm

SIG2

PHI1

PHI2 Coff

Cf

kT + Ci

⎛C ⋅⎜ i ⎜C ⎝ f

Voutp − Voutm = Voff ⎛ ⋅ ⎜⎜Voarr @ SIG 2 − Voarr @ SIG1 − 2 ⎝

2

⎞ ⎟ + kT ⎟ C off ⎠

⎛C ⋅ ⎜ off ⎜C ⎝ f

⎞ ⎟ ⎟ ⎠

2

The opamp noise contribution is about:

It is realized by a fully-differential switched-capacitor structure and uses the correlated-double-sampling technique for low frequency noise reduction. It provides an accurate transfer function due the accurate integrated capacitance matching. All the switches are CMOS pass-gate with WNMOS=0.5μm, WPMOS=1.5μm and with L=0.35μm for both transistors. The input capacitance Ci is fixed to be four times a unit capacitor Cu≅270fF, capacitance Coff is fixed to Cu, while the feedback capacitance Cf is implemented as a bank of four unit capacitors each in series with CMOS switches. The switches are set by using the startup configuration word so as to fix variable gains in the set of X={1, 4/3, 2, 4}. Then, in the following Cf will be represent by XCu. During SIG1 the input voltage Vaorr is sampled across Ci, it represents the result of the dielectrophoresis process of the addressed microsite. While PHI1 is still high a RESET signal is activated to the same microsite and the reset voltage is sampled during SIG2 on the other Ci. The output differential voltage is provided during PHI2 according the following relationship:

Ci Cf

kT kT ⋅ Ci kT ⋅ Coff + + + Cf C 2f C 2f

PHI2

Figure 3. Pre-amplifier schematic.

=

Eq.3

Vcm_out

Voff

Vn2 =

Voutp PHI1

+

Vcm_in

Ci PHI1

Eq.1

Qn2,Coff = kT ⋅ Coff

⎞ ⎟ ⎟ ⎠

Eq.4

V =V 2 n

2 n , opamp

⎛ C + Coff ⋅ ⎜⎜1 + i Cu ⎝

⎞ ⎟⎟ ⎠

2

Therefore, substituting the capacitance values and supposing uncorrelated noise sources the total output noise of the pre-amplifier is about:

Eq.5

2 ⎛ kT ⎛ 5⎞ 5⎞ ⎞ ⎛ ⋅ ⎜ 6 + ⎟ + Vn2,opamp ⋅ ⎜1 + ⎟ ⎟ Vn2 = 2 ⋅ ⎜ ⎜ Χ ⋅ Cu ⎝ X⎠ ⎝ Χ ⎠ ⎟⎠ ⎝

In the previous formula the factor 2 is due to the fully differential structure.

Vdd=3.3V

PHI

Vin

M3

Vbias1

+

S/H PHIC

ADC

Vin+

M1

DAC

Vin-

M2

2 bits 11 bits

digital correction circuit

Vbias2

M4

M5

M6

M7

M8

M9

Vout-

Figure 5. Algorithmic A/D converter with 1.5-bit/stage architecture. Vout+

Vbias3

Vbias4

M10

2x

-

M11

Figure 4. Fully-differential telescopic cascode schematic. The OTA design in the pre-amplifier schematic is a fully differential telescopic cascode with continuous time common mode feedback using triode devices, as shown in Fig.4. This choice is due principally to its minimum occupied area and the simplicity realization; the different input and output voltage common mode values are not critical in switched capacitor circuits. Its performance is as follows: bias current=60μA, dc-gain≈ 78dB, unity-gain-bandwidth≈12MHz and phase margin ≈ 86° (Cload=3pF). These features allow obtaining an 11 bits settling precision in half period time. Table I shows the size of the transistors used. Table I. Fully-differential telescopic cascode transistor size. Transistor W [μm] L[μm] M1, M2 18 0.6 M3 18 0.6 M4,M5 36 1.2 M6,M7,M8,M9 10 1.2 M10,M11 1.2 0.6

The main blocks are a sample&hold (S/H), a multiply-bytwo amplifier, a sub-ADC, a DAC and a digital correction circuit. The converter works as follows: only during the first conversion cycle the input signal goes through the S/H while during the other conversions the output signal of the multiplyby-two amplifier goes through the S/H, each cycle resolves two bits with the sub-ADC, subtracts this value from its input and amplifies the resulting residue by a gain of two. The input signal ranges from +Vref to –Vref (Vref is about 500mV for this work), the sub-ADC thresholds are +Vref/4 and –Vref/4, the DAC levels are –Vref, 0, +Vref. Therefore the residue transfer function is:

Vo = Vi +1 = Eq.6

⎧2 ⋅ Vi − Vref ,Vi > Vref / 4, b1b0 = 10 ⎪ = ⎨2 ⋅ Vi,−Vref / 4 < Vi < Vref / 4, b1b0 = 01 ⎪2 ⋅ Vi + Vref ,Vi < −Vref 4, b b = 00 1 0 ⎩

In the previous formula b1 and b0 indicate respectively the most significant and less significant sub-ADC output bits. At the end of the N-cycles, the output code is achieved as follow by the digital correction block: Cycle -1 Cycle -2

b1

+

b0 b1

b0

+ +

… Cycle -N-1

b1

Cycle -N

b0

+

b1 b0

=

+

output code

IV. Analog to digital converter architecture. The essential point of this work is to design the electronic interface and the sensor array on the same chip. For this reason an algorithmic A/D converter was chosen where the number of components is drastically reduced in comparison with other structures and the accuracy is much higher [2]. Furthermore, a 1.5-bit/stage architecture, shown in Fig.5, has been chosen for relaxed component requirements. The internal clock frequency needed is about 1.2MHz to obtain a sampling frequency of 100ksample/s. Although a single-ended configuration is shown for simplicity the actual design is fully-differential. A common switched capacitor implementation was chosen.

For this architecture the accuracy requirements of the subADC are greatly reduced, in this case a maximum offset of Vref/4 can be tolerated before bits error occurs. This allows converter to neglect comparator offset performance. Otherwise any offset cancellation (CDS or large device area) would increse the ADC area occupancy. IV.1 Sub-ADC implementation. The two-bits ADC block is implemented by two fullydifferential clocked comparators and a switched network; each one schematic is depicted in Fig.6. All switches designed are CMOS switches with WNMOS=0.5μm, WPMOS=1.5μm and L=0.35μm for both transistors. During PHI1, the input signal is sampled across C by two CMOS switches. During the next phase PHI2, the comparison

between the sampled input voltage and the reference voltage takes place. Finally, during the latch phase, the output is available. The capacitor C is set to 0.1pF. The comparator schematic is shown in Fig.7.

Vdd=3.3V R1

S1 S2

PHI1C R2

C

PHI1

PHI2

-

Vinp PHI2

PHI1

Vrefp

+

PHI1

PHI1

R3

Vcm_out

R4

Voutm

+

PHI2

PHI2

Vout

S1 S2

-

Vinm

S1

Voutp

S1

C

Vrefm

R5

PHI2C

Figure 6. Half sub-ADC schematic. Figure 8. Two bits R-string D/A converter schematic

Vdd=3.3V M3

Vbias

PHI2C

M12 M9

M10 M11

Vout+ Vin+

M1

M2

Vin-

PHI2C Vout-

PHI2C

M7

PHI1C

M8

PHI2C

M6

M4

M5

Figure 7. Comparator schematic The comparator was designed according to [3]. The transistors sizes are reported in Table II.

IV.3 Multiply-by-two and sample-&-hold Although shown as separate functions in Fig.5, the S/H and the multiply-by-two blocks are combined into one switched capacitor circuit to obtain a more compact structure and to minimize the occupied area. A simplified schematic and a timing diagram are shown respectively in Fig.9 and Fig.10 [4]. The fully differential OTA here used has the same schematic as Fig.4 but different transistors size (Table III) due the different clock frequency with respect to the pre-amplifier circuit. Its performances are: bias current ≈ 37 μA, gain ≈ 78dB, unity-gain-bandwidth ≈ 12MHz and phase margin≈ 86° (Cload=2pF). Vinp

Vcm_in

Table II. Fully-differential comparator transistor sizes Transistor W [μm] L [μm] M1, M2 5 0.5 M3 5 0.5 M4,M5 2.5 0.5 M6,M7,M8 2 0.5 M9,M10,M11,M12 6.5 0.5

Cf

PHIS

PHIH

PHIH Cs

PHI1 Voutp PHI2M PHI2Z PHI2P Vrefm Vrefz Vrefp PHI2M PHI2Z PHI2P Voutm

PHI1

PHI2

-

Voutp +

Vcm_in PHI1

-

PHI2

Voutm

+

PHI1 Cs PHIH

PHIH PHIS

Cf

Vcm_in

IV.2 Two bits DAC implementation. The two bits DAC implementation is based on the classical R-string D/A converter. It is realized by two different single-ended structures due to the fully differential philosophy of the entire interface, each one is depicted in Fig.8. The resistors’ values are the following: R1=5.1kΩ, R2=R3=R4=850Ω, R5=3.75kΩ. The input bits, S1 and S2, different for each implementation, are achieved by a logic circuit driven by the sub-ADC output bits b1, b0. For this application to have two ron switches in series is not critical due to low clock frequency of internal ADC. On the other hand, this approach allows a chip area saving with respect to the AND implementation in the digital domain.

PHIS

PHIS Vinm

Figure 9. S/H and multiply-by-two schematic PHIS PHIH PHI1 PHI2

Figure 10. Timing diagram Table III. Fully-differential telescopic cascode transistor size Transistor W [μm] L [μm] M1, M2 11 0.6 M3 11 0.6 M4, M5 22.5 1.2 M6, M7, M8, M9 6.25 1.2 M10, M11 0.75 0.6 Phases PHIS and PHIH control sampling of the input signal onto Cf capacitors and its duration is one cycle clock;

phases PHI1 and PHI2 control the cyclic gain during the conversion process. Very important are the capacitors values and the operational amplifier input referred noise for the signal-to-noise ratio. If Cs=Cf the input referred noise of the multiply-by-two circuit for each cycle is given by: Eq.7

Vn2,i −cycle =

kT 5 2 + ⋅ Vn ,OTA C 4

The total input referred noise of the algorithmic stage using the above circuit is given by:

Eq.8

V =V 2 n

2 n ,1

+

Vn2, 2 4

+

Vn2,3 16

+

Vn2,3 64

+ ...

With the previous formula the A/D has been designed to obtain an ENOB as near as possible to 11 bits. Therefore, Cs=Cf=1pF. Like in the previous schematics, all switches are CMOS pas-gate with WNMOS=0.5μm, WPMOS=1.5μm and L=0.35μm for both transistors. V. Simulation Results. In Fig.10 the readout output voltage for different gain value is reported. The input values at phase SIG1 and SIG2 are set respectively at 0.1V and 0.4V, whereas Voff is set to 0.1V (capacitive output load of about 2pF). The resulting gains are in the set X={1, 4/3, 2, 4}. For each gain value the output voltage results in the range of 11 bits precision.

Figure 10. Readout output voltage for different gain values in the set X={1, 4/3, 2, 4}. Fig.11 shows the output spectra of the ADC. The input signal voltage is set to 1Vp-p i.e. the full-scale range, while the input frequency signal is about 20kHz which is about the maximum input signal frequency. The sampling frequency is 100ksamples/s. The samples number is 512. The SNR value obtained is about 65.7 dB and the relative ENOB value obtained is 10.6 bits.

Figure 11. ADC output spectra for fs=100ksample/s, fin≈20kHz and Vin=1Vp-p (differential). MATLAB simulations show that capacitance mismatch affects only linearity performance but not noise floor. This result respects the signal-to-noise ratio constraint that is a must for this design. Table IV summarizes the performance obtained for this A/D converter design. Table IV. A/D Converter Performance Summary. Process 0.35μm CMOS Supply 3.3V Sampling rate 100ksample/s Resolution 11b Full-Scale Input

1Vp-p (differential)

Power Consumption SNR SNDR

2mW 65.7dB 65.6dB

ENOB

10.6b

Due the pitch site, the layout of the pre-amplifier circuit and the A/D converter must respect the constraint to be developed in a strip of 20μm. A detail of the layout is shown in Figure 12. It is the preamplifier operational amplifier layout. Its width results of about 15μm, instead its length is about 120μm. The rest of the 20μm is allocated for clock, reference and supply voltage bus. Since the fully-differential design philosophy, the layout has vertical symmetry axes.

120um

15um VI. Conclusions In this paper an integrated readout circuit for Lab-on-aChip applications has been described. Integrating the A/D conversion on-chip allows readout circuit to obtain better performance in term of signal-to-noise ratio with respect to the off-chip solution. However, this requires designing the readout chain with the constraint of minimal occupied area and good performance at the same time. Therefore special attention to the choice of the architecture has been specified. The main constraint of the design is the achievement of a >10b signal-to-noise ratio with a small die area at a 100kHz sampling rate. The Lab-on-a-Chip approach with mini sensors array and electronic processing circuits on the same chip allow fast and low cost biological analysis making the consequent information immediately available to the user. A fully-differential switched-capacitor architecture has been adopted. It is made up of a pre-amplifier circuit with programmable gain in the set X={1, 4/3, 2, 4}, and an algorithmic 1.5-bit/stage architecture A/D converter. This choice allows designing relaxed components requirements and saves die area. The achieved results show a good resolution for the readout circuit for any gain value and good A/D converter performances: SNR=65.7dB, SNDR=65.6dB, ENOB=10.6b. Currently, time sharing techniques are investigated to use only one operational amplifier in the readout chain. This will allow to save 40% of chip area and 40% of power consumption. VII. Acknowledgments The authors acknowledge Silicon Bio Systems s.r.l., for the collaboration and technical informations on sensor array. This project is partially funded by the Regione Puglia Explorative Project “Integrated microsystem for rapid biological cell detection (Lab-on-a-Chip)”. The authors acknowledge Regione Puglia for this support. References. 1. N. Menaresi, A. Romani, G. Medoro, L. Altomare, A. Leonardi, M. Tartagni, R. Guerrieri, “A CMOS Chip for Individual Cell Manipulation and Detection,” IEEE J. Solid-State Circuits, Vol.38, No.12, December 2003, pp. 2297-2305. 2. R. Van De Plassche, “Integrated Analog-to-Digital and Digital to Analog Converters ”, 1994 Kluwer Academic Publichers. 3. G. M. Yin, F. Op’t Eynde, and W. Sansen, “ A HighSpeed CMOS Comparator with 8-b Resolution,” IEEE J. Solid-State Circuits, Vol.27, No.2, February 1992, pp. 208-211. 4. J. McNeill, M. C.W.Coln, and B. J. Larivee, “Split ADC Architecture for Deterministic Digital Background Calibration of a 16-bit 1-MS/s ADC,” IEEE J. Solid-State Circuits, Vol.40, No.12, December 2005, pp. 2437-2445. Figure 12. Operational amplifier layout of the readout circuit.

Modeling and Design of a Microdisk Photonic Sensor for Biological Applications Vittorio M. N. Passaro, Biagio Casamassima, Francesco De Leonardis *, Francesco Dell’Olio and Francesca Magno Photonics Research Group, Dipartimento di Elettrotecnica ed Elettronica, Politecnico di Bari, via E. Orabona n. 4, 70125 Bari, Italy e-mail: [email protected], URL page: http://dee.poliba.it/photonicsgroup * Photonics Research Group, Dipartimento di Ingegneria dell’Ambiente e per lo Sviluppo Sostenibile, Politecnico di Bari, viale del Turismo n. 8, 74100 Taranto, Italy Abstract In this paper, we present the modeling and design of a new approach to detect an analyte for biochemical and biological sensing applications. It is based on a whisperinggallery-mode optical resonator coupled with a Fabry–Perot cavity in silicon-on-insulator technology. The theoretical model of the whole architecture includes the influence of all electrical and optical parameters, such as thin oxide thickness, silicon and poly silicon doping concentration, optical losses, wavelength and amplitude characteristics of the architecture, charge accumulation effects, and thermal effects. The very high sensitivity of this device, demonstrated by simulations, is due to the simultaneous influence of the two coupled resonators and the metal-oxide-semiconductor structure. Introduction Since 1960, whispering-gallery-mode (WGM) resonators have been studied as important components of many optical systems. They have numerous applications in sensing, linear and nonlinear optics. Recently, micro-resonators have been proposed for optical sensing because they offer a unique advantage of reducing the device sizes by orders of magnitude without sacrificing the interaction length, due to their high quality factor Q. Microsphere resonators have been demonstrated capable to detect adsorption of a monolayer of protein [1], microring resonators have been used to detect small concentration changes of glucose [2] and microdisk resonators have been investigated to provide sensitivity enhancement in comparison with straight waveguide sensors [3]. In this paper, we propose a new micro-disk resonator based on integrated optical sensor in which sensing area is a portion of the straight waveguide. In addition, part of the straight waveguide includes a metal-oxide-semiconductor (MOS) structure in Silicon-on-Insulator (SOI) technology. Theory Our microdisk resonator works as a resonant cavity and is closely coupled with one straight bus waveguide (see Fig. 1). Under monomodal propagation in the resonator and for lossless coupling, the interaction between microdisk and waveguide in absence of gratings (by setting E1b = 0 and E3b = 0) can be described by a matrix relation [4]:

⎡ E3 ⎤ ⎡ t ⎢ E ⎥ = ⎢k * ⎣ 4⎦ ⎣

k ⎤ ⎡ E1 ⎤ t * ⎥⎦ ⎢⎣ E2 ⎥⎦

1-4244-1245-5/07/$25.00 ©2007 IEEE

(1)

where the complex mode amplitudes Ei are normalised, E1 = 1 , while t and k are coefficients of transmission and coupling, respectively.

E1b

E3b

Figure 1: Schematic architecture of sensor. The transmission along the disk is taken into account imposing that [4]: E2 = α ⋅ e jϕ ⋅ E4 (2) where α is the loss inner circulation factor and ϕ is phaseshift acquired by optical mode during its propagation in the disk. If inner circulation is lossless, then α = 1 . The transmission intensity is given by [4]:

E3 E1

2

T=

E2 E1

2

Y=

α 2 + t − 2 ⋅ α ⋅ t ⋅ cos ( β ) 2

=

1 + α 2 ⋅ t − 2 ⋅ α ⋅ t ⋅ cos ( β ) While the total circulating power is given by:

=

2

(

α 2 ⋅ 1− t

2

)

1 + α 2 ⋅ t − 2α ⋅ t ⋅ cos ( β ) 2

(3)

(4)

where: t = t ⋅ e jτ and β = ϕ + τ , being τ the phase variation in microdisk, as [4]:

τ = 2π ⋅ neff ( λ ) ⋅

R

λ

(5)

where R is the microdisk radius and neff is the effective index of microdisk guided mode. The microdisk resonance condition is given by: 2π ⋅ neff ( λ ) (6) λ = R⋅ 2π ⋅ m − ϕ For β = ϕ + τ = 2π ⋅ m (with m an integer number) the total circulating power is maximum. In fact, from Eq. (4) we have:

Ymax =

(

α 2 ⋅ 1− t

2

)

(7)

1 + α 2 ⋅ t − 2α ⋅ t 2

2⋅c

π ⋅ neff ( λres ) ⋅ R

1−α ⋅ t



α⋅t

(8)

(18)

where

⎡ −1 − r ⎤ (19) ⋅⎢ 1 ⎥⎦ j ⋅ 1− r ⎣ r is the scattering matrix of Fano configuration [6] and r is the grating reflection coefficient. 1

⎡⎣ R ⎤⎦ =

The bandwidth, defined as Y ( Δω ) = 0.5 ⋅ Ymax , is written:

Δω =

K = ⎡⎣ R ⎤⎦ × [ M ] × [ N ] × [ M ] × ⎡⎣ R ⎤⎦

2

Moreover, the quality factor is given by: Q=

2 ωres π ⋅ neff ⋅ R α ⋅ t = Δω λres 1−α ⋅ t

(9)

It is important to observe that for small variations of resonant wavelength, the following approximation holds: neff ( λ ) ≅ neff ( λres ) (10) Due to the presence of the gratings, in case of our architecture it is convenient to use the complete matrix scattering formalism [4]: ⎡ E3 ⎤ ⎡ a11 ⎢ E ⎥ = ⎢a ⎣ 3b ⎦ ⎣ 21

a12 ⎤ ⎡ E1 ⎤ a22 ⎥⎦ ⎢⎣ E1b ⎥⎦



τ

=



4π τ coup



τ coup

+



τ loss

0 ⎤ ⎥ e ⎦ − jθ

(20)

and

2

⎛ 4π ⎞ ⎡⎣ j ⋅ (ω − ωres ) + A⎤⎦ − ⎜ ⎟⎟ ⎜τ coup ⎠ ⎝ a11 = 2π ⎞ ⎛ ⋅ ( A + j ⋅ (ω − ωres ) ) ⎜ j ⋅ (ω − ωres ) + τ ⎟⎠ ⎝ 2π j ⋅ (ω − ωres ) + τ a22 = ⎡⎣ A + j ⋅ (ω − ωres ) ⎤⎦ 4π 1 a12 = − τ coup ⎡⎣ A + j ⋅ (ω − ωres ) ⎤⎦

τ

⎡e jθ

[M ] = ⎢

⎣0

2



For r = 0, the transmittivity line shape is a Lorentzian. Moreover, the scattering matrix of phase variations is:

(11)

where:

A=

Figure 2: 2D view of waveguide (Fano configuration).

L ⎛ ⎞ (21) + neff , met ⋅ D ⎟ 2 ⎝ ⎠ is the relevant phase variation. Matrix [ N ] of Eq. (18) is

θ = k0 ⋅ ⎜ neff , sen ⋅

(12)

(13) (14)

calculated by Eq. (11). The transmittivity coefficient is given by [6]: 1 (22) T= 2 K 22 Fig. 3 shows the transmittivity versus wavelength shift for r = 0, 0.2, 0.5 , being L + 2 D = 1.3 mm and L = 100 μm.

(15) (16)

a21 = −a12 , 1 τ coup is the contribution to resonator decay rate due to the power exchange with the straight waveguide and 1 τ loss is the contribution related to resonator intrinsic optical losses. If A = 0 , a perfect matching condition occurs, i.e. τ loss = τ coup , known as traveling-wave operation. The transmittivity coefficient is [5]: T=

1 a22

2

(17)

Sensor modelling Our architecture is based on two coupled microcavities, as shown in Fig. 1. The device consists of a WGM disk resonator, a straight waveguide coupled to the disk through a coupled power fraction k (depending on the gap g), and two partial reflectors (gratings) appropriately placed at the ends of the waveguide to form a Fabry-Perot (FP) cavity(see Fig. 2). The scattering matrix is written as [6]:

Figure 3: Transmittivity versus wavelength shift for r = 0, 0.2, 0.5 ( L + 2 D = 1.3 mm and L = 100 µm). In case of r = 0.5, the lineshape is an asymmetric Fano resonant curve with an increasing slope between zero and unity transmission, compared with the Lorentzian shape. It is possible to demonstrate that the optimum distance between two reflectors to increase the slope with respect to the conventional disk resonator [5] is given by:

D=

λ ⎛ 2 ⋅ m + 0.5 ⎞ ⋅⎜ ⎟ 4 ⎜⎝ neff , w − met ⎟⎠

(23)

with m an integer number. The Fano resonance is due to the spectrum interference between FP cavity and disk resonator. In fact, compared with a conventional disk resonator where waves propagate in only one direction, the partially reflecting elements introduce backward propagating waves that perturb the phase of the wave transmitted, and hence lead to a complex interference.

Sensitivity For photonic devices for biomedical applications, generally two sensing mechanisms are used: homogeneous sensing and surface sensing. In homogeneous sensing, the device is typically surrounded by an analyte solution, which can be regarded as the top cladding of the waveguide. By this way, analyte will modify the bulk refractive index of the solution and this will change the guided mode effective index. In surface sensing, the photonic device is pre-treated to have receptors or binding sites on the sensor surfaces which can selectively bind the specific analyte. Because of the presence of a larger evanescent field at the waveguide boundary to probe the analytes, this method can provide a higher sensitivity. To transduce the analyte amount in a detectable signal, two sensing schemes could be used in our device: the resonant wavelength-shift and the intensity variation. In the former, the sensitivity [7] is: ∂λ (24) S = res ∂neff The detection limit is defined as the smallest detectable change of the waveguide parameters caused by analytes, which is directly related to the smallest analyte amount that a sensor can detect. In the case of the resonant wavelength-shift monitoring, the detection limit is proportional to the smallest detectable index change of the top cladding for the case of homogeneous sensing, and to the smallest detectable thickness change of the adsorbed analyte film in case of surface sensing. The absorption of analyte molecules determinates the transmittivity and such the device sensitivity. Thus, decay rate τ loss depends on the fraction of absorbed light . To take into account this absorption, it is necessary to make some observations [8]. The absorption cross sections of biological molecules are unlikely to change dramatically as a consequence of the weak binding associated with an antigen or an antibody. We indicate with ℑ the cross-section of the mode excited within the sensible zone of device, that in first approximation is [9]: 2

where neff

⎛ λ ⎞ (25) ℑ=⎜ ⎜ neff ( λ ) ⎟⎟ ⎝ ⎠ is the effective index of the sensible region of

device, σ the absorption cross section of a biological molecule, η the efficiency factor that accounts for the fact that the molecule, located on the surface of the disk, does not

experience the maximum intensity of the guided optical mode. Then, the fraction of the light absorbed in interacting with this molecule is [9]:

ψ =η

σ

(26) ℑ Then, the loss factor due to the absorption is: [9] (27) α abs = N ⋅ σ where N is the molecular density. For example, if σ = 2 ⋅10−16 cm 2 (in case of dopamine with concentration 20 μM), we have: N = 20 μ × 6 ⋅1020 molecules cm −3 and α abs = 2 ⋅10−16 × 20μ × 6 ⋅1020 = 2.4 cm −1

Losses The total optical losses in the microdisk resonator can be expressed as: (28) α tot = α bend + α leak + α scatt + α prop The leakage loss to the substrate, αleak , is negligible because a buffer thick oxide layer between guiding layer and substrate layer is introduced in both resonator and straight waveguide (assuming SOI technology). The bending loss coefficient, αbend , can be considered negligible because of the strong confinement in the resonator and its large radius, required to increase the sensor sensitivity (> 50 μm). The scattering loss, α scatt , is caused by sidewall roughness, especially if the sizes are very small. The sidewall imperfections can be usually described by a random function having a Gaussian distribution for its self-correlation function as [4]: ⎛ ⎛ s − s ' ⎞2 ⎞ (29) corr ( s − s ' ) = σ c2 ⋅ exp ⎜ − ⎜ ⎟ ⎟ ⎜ ⎝ Lc ⎠ ⎟ ⎝ ⎠ where s is a curvilinear coordinate, Lc is the correlation length, and σ c is the relevant standard deviation. Moreover, α prop is the propagation loss. It is important to note that if the analyte is deposited on the disk, then: α prop = α optical − prop + α abs

(30)

We have calculated the decay rate due to losses. In general, the decay rate is proportional to quality factor Qloss of the losses [7] by: 1 (31) τ loss = ⋅ Qloss f

where f is the frequency [Hz]. It can be determined by: 1 1 1 1 1 = + + + Qloss Qbend Qleak Qscat Q prop

(32)

Particularly, it results [4], [7]: ⎛ αscatt ⋅π ⋅ Reff ⎞ exp⎜ − ⎟ 2⋅π ⋅ Reff ⋅ neff ,disk 2 ⎝ ⎠ Qscat = ⋅ λ 1−exp( −αscatt ⋅ Reff ⋅π ) 2

(33)

⎛ ⎞ λ Qprop = ⎜ ⋅ α prop ⎟ ⎜ π ⋅ neff , disk ⎟ ⎝ ⎠

−1

(34)

technology). Moreover, the bending loss coefficient is negligible due to the strong confinement in the resonator, and to the large radius required to increase the sensor sensitivity.

where Reff ≅ 0.98 ⋅ R (for microring it occurs Reff ≅ R ). Moreover, the quality factor of coupling loss is: Qcoup =

λ 4 ⋅ π ⋅ neff ⋅ Reff 2

k

(35)

1 (36) ⋅ Qcoup f where k is the power coupling coefficient between disk and straight waveguide.

τ coup =

Design of components Straight waveguide The straight waveguide including the MOS structure has been designed [5] in order to support only the quasi-TE fundamental mode (see Fig. 4), assuming a thin oxide sandwiched between them, 12 nm thick. A buried oxide buffer layer, 1 μm thick, has been considered. Moreover, the structure includes two strongly doped polysilicon and silicon thin layers (100 nm) to ensure ohmic contacts with the metallized regions, one on the rib top and the other for ground contact (see Fig. 4).

Figure 4: Cross section of SOI straight waveguide with MOS structure. Fig. 5 shows the cross section of the SOI straight waveguide in sensible zone. Here bio-receptors are present, but not aluminium contact neither thin oxide. Microdisk structure The disk cross section has been designed [5] in order to be birefringence free and with low optical loss due to doping. It can be noted that the same parameters as the straight waveguide are used whereas possible, to ensure good power coupling through the gap and high technological compatibility. Perfect matching condition requires: (37) k = 4 ⋅ π ⋅ R ⋅ α disk where k is the coupled power fraction. However, k is a function of the gap: thus, to obtain a perfect matching condition, we have fixed the gap to g = 2.5 μm. In this case, the propagation loss contribution is α prop = 0.145 cm −1 for the

disk and α prop = 0.176 cm −1 for the straight waveguide. The scattering loss contribution measures is α scatt = 0.048 cm −1 . The leakage loss to the substrate is negligible because we introduce in the resonator structure the buffer thick oxide layer between the guiding layer and substrate layer (SOI

Figure 5: Cross section of the SOI straight waveguide in sensible zone.

Finally, to model the electrical modulation of effective index [4] we can write: ∂neff ,w−met ∂neff ,w (38) C (V ) ⋅V Δneff ,w−met = ΔTtemp + ∂T ∂Q where V is applied voltage, C(V) is the capacitance of the MOS structure (depending on V), Q is the capacitor charge induced by external electric field through its voltage V and ΔTtemp is the thermal variation. Performance

The operation of the sensor is based on the analysis of the transmitted spectrum. In the sensitive zone, the presence of the analyte causes a variation of the refractive index and the relevant change of effective index. Nevertheless, the microdisk resonance wavelength doesn't change because it depends by the constant effective index of microdisk. However, the variation of the effective index of the sensitive zone causes a variation in the lineshape of transmitted spectrum. This variation can be evaluated from the variation ΔT = Tλ =1.5501nm ( Δneff = 0 ) − T1.5501nm ( Δneff ≠ 0 ) . Since a refraction index could be associated to each biological substance (and, then, the effective index will change according), then it is possible to derive the particular analyte concentration from transmittivity measurements ΔT . The sensor is very sensitive if a large variation of ΔT occurs for small changes of Δneff . For this reason, we have calculated ΔT in absence or in presence of the electrodes, for applied voltages of 2 and 4 V. Without electrodes, we estimate a transmittivity variation: ΔT = 0.0220 (for Δneff = 10−3 ) ΔT = 0.0015 (for Δneff = 10−4 )

In presence of electrodes, as in our device, we have for λ = 1550.1 nm: ΔT = 0.053 for Δneff = 10−3 ( V = 2 V, see Fig. 6) ΔT = 0.1044 for Δneff = 10−3 ( V = 4 V, Fig. 6) ΔT = 0.0140 for Δneff = 10−4 ( V = 2 V, see Fig. 7) ΔT = 0.0368 for Δneff = 10−4 ( V = 4 V, Fig. 7)

In case of absorption due to analyte, the matrix in Eq. (17) changes as: T = 1 K 22 ⋅ exp ( −2 ⋅ν ) 2

(39)

where ν = α abs ⋅ L . Assuming an analyte concentration of 5μ M ( Δneff = 10 ), we have obtained for λ = 1550.1 nm: −3

concentration of 10 μM and (iii) with absorption produced by an analyte concentration of 20 μM.

ΔT = 0.0546 ( V = 2 V)

Table 1. Results in terms of ΔT for different cases.

ΔT = 0.1114 ( V = 4 V).

V=2V 4V 2V 4V

0.05 nm 0.05 nm 0.15 nm 0.15 nm

Δneff = 10-5

Δneff = 10-4

Δneff = 10-3

0.29766 6.9069 10-3 0.37147 4.9151 10-3

0.29771 6.89535 10-3 0.37147 4.7675 10-3

0.29809 6.79571 10-3 0.37148 3.4981 10-3

Table 2. Results in terms of ΔT for C=10μM and absorption

2V 4V 2V 4V

0.05 nm 0.05 nm 0.15 nm 0.15 nm

cross section of 2 10-16 cm2. Δneff = 10-5 Δneff = 10-4 0.2656 0.2657 25.8 10-3 25.7 10-3 0.34978 0.34978 34.8 10-3 35.0 10-3

Δneff = 10-3 0.2661 24.6 10-3 0.34978 36.3 10-3

Figure 6: Transmittivity versus wavelength shift for V = 0, 2, 4 V and Δneff = 10−3 .

Finally, assuming an analyte 10 μ M ( Δneff = 10−3 ), we have obtained:

concentration

of

ΔT = 0.0592 ( V = 2 V, see Fig. 8) ΔT = 0.1187 ( V = 4 V, Fig. 8).

It is important to note that, for this small variation of effective index, not larger than 10−3 , the coupling coefficient doesn't change and the sensor still works under perfect matching condition. Table 1 summarizes the results in terms of ΔT for various waveguide effective index changes, voltages and wavelength shifts, Δλ = 0.05-0.15 nm. In Table 2 results are given for C=10μM and absorption cross section of 2 10-16 cm2.

Figure 7: Transmittivity versus wavelength shift for V = 0, 2, 4 V and Δneff = 10−4 .

Fig. 9 shows the transmittivity variations versus refractive index of the analyte for three different cases: (i) negligible absorption, (ii) with absorption produced by an analyte

Figure 8: Transmittivity versus wavelength shift for V = 0, 2, 4 V and C = 10 μM.

Figure 9: Transmittivity variation |ΔT| versus analyte refractive index in three cases: (i) negligible absorption (squares), (ii) not negligible absorption and analyte concentration of 5 μM (diamonds), (iii) not negligible absorption and analyte concentration of 10 μM (triangles).

In all cases, the applied voltage is supposed to be 4 V and the wavelength is fixed as 1.55015 μm. It is possible to observe that, when the analyte refractive index increases, the transmittivity variation is reduced. A good linearity behaviour can be observed by these plots. Then, results demonstrate the potential of such photonic sensor to detect even small analyte concentrations. Although the architecture is similar to that presented in literature [5], in this paper a completely different use for chemical and biological sensing is theoretically proved. Conclusions In this paper, a very detailed modeling of a novel photonic sensor for detecting biological substances has been presented. The sensor is based on two coupled resonators, a Fabry-Perot and a WGM resonators, and on the physical effects induced by an applied voltage on a metal-oxide-semiconductor capacitor in SOI technology. Since in this sensor the resonant wavelength does not change, the detecting of the analyte is not associated to the shift of the resonant wavelength but to the spectrum lineshape. The relevant advantages of this configuration are: the resonance wavelength is constant, i.e.: 1.

λres (Δneff = 0) ≅ λres ( Δneff ≠ 0 ) ;

2.

3.

the detection instrumentation is very simple. In fact, it is enough to measure the spectrum for a fixed wavelength, λ = λres + Δλ , rather than to execute a wavelength scanning; larger sensibility is possible for biological substances showing a significant optical absorption. In fact, if losses for optical absorption are present due to the analyte, the transmittivity change ΔT

increases (see Table 2), because the absorption reduces the lineshape of transmitted spectrum in comparison with lossless case. The resulting transmittivity spectrum is particularly sensitive to vary small changes of analyte refractive index. Good results have been obtained by detecting the intensity (or its variation in comparison with the reference value) of the transmitted power at fixed wavelength (amplitude interrogation). In our architecture, a very important role plays both the grating and the voltage applied to the electrodes. Otherwise, it should not be possible to distinguish ΔT for small index changes, Δn = 10-5,10-4,10-3. These advantages could be useful in the realization of sensors for commercial use. References 1. Vollmer, F. et al., “Protein detection by optical shift of resonant microcavity”, Appl. Phys. Lett., Vol. 80, No. 21 (2002), pp. 4057-4059. 2. Chao, C.-Y., and Guo, L. J., “Biomedical sensors based on polymer microrings with sharp asymmetrical resonance”, Appl. Phys. Lett., Vol. 83, No. 8 (2003), pp 1527-1529. 3. Yang, J. and Guo, L. J., “Optical Sensors Based on Active Microcavities”, IEEE J. Sel. Top. Quantum Electron., Vol. 12, No. 1 (2006), pp. 143-147.

4. Yariv, A. “Universal realizations for coupling of optical power between microresonators and dielectric waveguide”, Electron. Lett., Vol. 36, (2000), pp.321-322. 5. Passaro, V.M.N. and De Leonardis, F., “Modeling and Design of a Novel High-Sensitivity Electric Field Siliconon-Insulator Sensor Based on a Whispering-Gallery-Mode Resonator ”, IEEE J. Sel. Top. Quantum Electron., Vol. 12, No. 1 (2006), pp. 124-133. 6. Fan, S. “Sharp asymmetric line shapes in side-coupled waveguide-cavity systems,” Appl. Phys. Lett., Vol. 80, No. 6 (2002), pp. 908–910. 7. Chao, C.-Y., and Guo, L. J. “Design and Optimization of Microring Resonators in Biochemical Sensing Applications”, J. Lightwave Technol., Vol. 24, No. 3 (2006), pp.1395-1402. 8. Boyd, R.W., and Heebner, J.E., “Sensitive disk resonator photonic biosensor”, Appl. Opt., Vol. 40, No. 31 (2001), pp.5742-5747. 9. Djordjev, K. et al., “Active Semiconductor Microdisk Devices”, J. Lightwave Technol., Vol. 20, No. 1 (2002), pp. 105-113.

A Configurable Architecture for the Detection of DNA Sequences based on a E2PROM device Luca Abbati1, Pisana Placidi1, Andrea Scorzoni1, Massimo Lanzoni2 1 DIEI, via Duranti 93, I-06125 Perugia, Italy 2 DEIS, viale Risorgimento 2, Bologna, Italy [email protected] Abstract The authors present a novel approach to perform the readout of a genetic sensor using a configurable architecture with extremely low drift with temperature and low sensitivity to the drift of mobile gate oxide charges. The proposed architecture, able to detect small changes of U.V. absorbance, calculates the analog weighted difference between the gate / source voltages of a floating gate and a reference nMOS transistor fabricated using a single poly technology. The theoretical dependence of the obtained results on both device and circuit parameters has carefully been analyzed. Experimental results show that the circuit offers a valid solution to decrease the error due to temperature variation during operation. Moreover, the measurement procedure is sufficiently quick for neglecting threshold voltage shifts due to mobile charges. Finally, the proposed approach has a particular advantage in integrated system design, since it is compatible with high-density MOS logic. Introduction In several applications (point of care, air or water control, food testing) the recognition of the DNA is performed by sensing the effects of possible hybridization between the single-stranded DNA fragments to be analyzed and known capture sequences. To this purpose DNA microarrays, fabricated on glass and requiring external optics for the analysis, have been developed. Despite the feasibility of new generation readers with lower cost and reduced foot print, microarrays require fluorescence readers and remain costly and not transportable. These features limit the use of conventional DNA chips for both decentralized testing and as a routine tool in the field of diagnostics. Therefore alternative, low cost, label-free and possibly “electrical only” methods for DNA detection are being investigated. [1-3] In this scenario silicon technology could usefully be exploited to produce low cost devices. Moreover, silicon memories are well known as the most cost effective silicon devices, being their packed architecture best suited for saving silicon area. In particular, nonvolatile memories (NV), like EPROM, E2ROM, or Flash, are able to balance the less-aggressive (with respect to SRAM and DRAM) programming and reading performances with nonvolatility, i.e., with the capability to keep the data content even without power supply. Thanks to this feature, the nonvolatile memories offer system designers a different opportunity [4-6] and cover a wide range of applications,

1-4244-1245-5/07/$25.00 ©2007 IEEE

from entertainment and automotive to computing and communication. [7-9] The different non-volatile memory families can qualitatively be compared in terms of flexibility and cost. Flexibility means the possibility to be programmed and erased many times on the system. [10, 11] Cost means process complexity and in particular silicon occupancy, density or cell size and use of technological simplifications like, e.g., “single poly” FLOTOX, which is fully compatible with standard digital CMOS processes. On the other hand, a common feature of both DNA and non-volatile memories is their sensitivity to UV radiation. In particular, hybridized DNA features a different absorption coefficient to UV compared with single stranded DNA. [12, 13] The interposition of hybridized or single stranded DNA between a UV lamp and a nonvolatile memory will therefore yield a different erase rate and could effectively be employed for hybridization detection. In this paper we present a circuit architecture suitable for performing the readout of a genetic sensor with a resolution sufficient, in principle, to discriminate between hybridized and non-hybridized DNA. [14] The architecture is highly flexible and scalable and could be considered as a proof concept for a future chip implementation. Compared to state-of-the-art, the proposed solution aims at higher performance, being suitable for large arrays of sensing sites, in turn leading to high parallelism, high throughput DNA analysis. The U.V. Sensor and the detection principle Our aim is to investigate an innovative solution, where the sensor is essentially represented by a silicon chip realized in CMOS technology featuring the option of NV E2PROMs. In principle, single-poly E2PROM cells operate as U.V. sensors. In fact, these cells can also be erased by means of U.V. radiation. Since the process of electron extraction from the floating gate is continuous, for fixed radiation time an E2PROM cell could operate as a U.V.sensor or dosimeter. [15] In principle, the measurement of the cell transistor threshold voltage shift can provide a simple means for electrical sensor read-out. The envisaged sensor houses an array of “sensing sites”, each formed by a bio-functionalized layer, positioned between an external U.V. source and a U.V. sensor measuring the radiation transmitted through the biolayer. The bio-functionalization of each site is obtained by means of specific DNA probes (single-stranded sequences) immobilized on the surface in a position aligned with the

underlying sensor and acting as "selective glue" respect to the DNA target molecules. In operation, all sites are simultaneously exposed to the target molecules selectively hybridizing only with complementary probes. Consequently, after cleaning the device, only the sites where hybridization has taken place will present additional DNA between the U.V. source and the sensors. Since DNA has a significant U.V. absorption, [13] the radiation transmitted through such sites will be smaller than in the opposite case: thus the recognition of this difference implies that of hybridization (hence the recognition of the target sequence). The proposed solution presents a number of significant advantages. Most of the electronics required for site addressing and read-out are already available in commercial products, excluding the analog readout. Another advantage is the reduced physical dimension of the cells, which could allow designers to dedicate more cells (sensors) to each sensing site in order to implement simple schemes for error detection and correction. The architecture of the adopted NV memories (E2PROM device shown in Figure 1) relies on a SinglePoly floating-gate nMOS transistor, i.e., a transistor with a gate completely surrounded by dielectrics, the floating gate (FG), and electrically governed by a capacitively coupled n+ diffusion acting a control gate (CG). A simple array of E2PROM devices manufactured with relatively old technology was adopted as an U.V. sensor prototype (Figure 2). Four programmable FG memory cells are available (governed by CG_065, CG_070, CG_075 and CG_080, where 0.65, 0.70 etc. are the coupling factors αG defined later) and one nMOS device, called reference or “equivalent” transistor TE (governed by the terminal CG_TE). The bit lines of each memory cell are activated by a common signal TG_COM and every device provides two different terminals: the control gate (CG_#) and the drain (D_#). In our sensor the common source (SOURCE_COM) and the BULK pins have been connected to ground.

temperature T variation and VT,ion the contribution due to mobile and trapped gate oxide charges.

Figure 2. Schematic of the considered U.V. sensor In this device the voltage VFG at the FG connection with respect to an arbitrary potential (due to capacitive coupling CCF between the FG and the CG and due to the intrinsic parasitic capacitances of the MOSFET devices CD , CS and CB, see Figure 3) can be expressed as: C C C C Q (2) V FG = CF VCG + S V S + D V D + B V B + CT CT CT CT CT where CT = CCF + CD + CS + CB and Q is the charge stored in the floating gate electrode. For the TE where Q = 0 we adopt a similar representation with a short-circuit between the FG and the CG (due to the buried contact between the FG and the CG which means VFGS = VCGS = VGS). CCF CG

CD

CS S

FG +

n

D

CB

n

+

Bulk

Figure 3. Simplified scheme of the capacitive coupling in a FG device Because VB = VS =ground the expression (2) can be written as:

Figure 1. Cross section of a FLOTOX cell featuring a single poly technology [16] For a FG device the expression of the threshold voltage VT can be written as (1) VT = VT 0 + VT ,Q + VT ,T (T ) + VT ,ion in which VT0 represents the threshold voltage for zero bulksource bias @ 25 °C, VT,Q the term due to charge storage on the floating gate, VT,T(T) the contribution due to the

⎛ Q ⎞ ⎟ (3) V FGS = α G ⎜⎜VCGS + fV DS + C CF ⎟⎠ ⎝ where αG = CCF/CT and f = CD/CCF. Now, if we define VT,Q as the voltage VCGS required to create the channel between source and drain (with VDS= 0) and VTFG as the threshold voltage seen by the floating gate electrode when no charge is stored (Q=0), we can write:

VT ,Q =

VTFG Q . − α G CCF

(4)

In principle, the change in threshold voltage while the E2PROM cell is under a U.V. radiation is an indirect measurement of the charge Q stored in the FG device. The

system to be designed should extract the rate of threshold voltage change before and after DNA insertion. In addition, considering a first-order MOS theory, for the FG device in the pinch-off region the current between source and drain can be expressed as 2 k FG FG I D = n VFGS − VT 2 (5) 2 knCG = α G (VCGS + fVDS − VT ,Q ) 2 in which the transconductance parameter knFG depends on the technological parameters and on physical dimensions of the device, and V DS > α G VCGS + fV DS − VT ,Q

To this purpose a system based on the circuit reported in Figure 4 has been designed. This circuit, inspired in principle to the classical ref. [17], but with a very different feedback circuit, relies on a differential architecture, where the input stage has been built using the devices housed in the U.V. sensor (M1 is one of the FG devices whereas M2 is the TE device). From now on, the subscripts 1 and 2 will identify the M1 and M2 devices, respectively. Because the devices are fabricated on the same die we can assume the devices share the same temperature. Therefore in the proposed circuit we have

knCG = αG knFG . Whereas for the TE transistor in the same region we have 2 k FG I D = n VGS − VTFG (6) 2 where the threshold voltage is simply VTFG and

where the term +/-2|Voffset| represents the operational amplifier offset voltage and can be forced to become zero. Therefore considering the expression (5) and (6) we obtain k1VCGS 1 − ⎡⎣ k2VTFG + VTQ (Q, VDS ) ⎤⎦ = VGS 2 − VTFG (8)

(

(

)

)

VDS > VGS − VTFG . In order to characterize the behavior of each memory cell housed in the sensor an operative definition of the threshold voltage VTFG has been adopted. Therefore in the remainder of the paper, VTFG will indicate the voltage VGS provided to the CG terminal in order to obtain a drain current IDS of 5 µA, keeping the drain voltage VD at a constant value of 1 V. The proposed system architecture Our aim is to design a system that is temperature independent and that doesn’t produce any drift in threshold voltages due to mobile and trapped gate oxide charges in order to evaluate the Q charge stored in the FG device. Experimental analysis demonstrated that the influence of both temperature variation and oxide charge allocation due to the bias of the cell can induce fatal error in the reading of VTFG. Therefore a suitable system devoted to measure ΔVT,Q has to be temperature independent and should be fast enough to be independent of charge drift.

Figure 4. Simplified schematic showing the basic circuit devoted to the measurement of the threshold voltage (M2 has been used as a reference device)

VREF 191 − R1I D1 = VREF 191 − R2 I D 2 ± 2Voffset

where

we

assumed

k1 = α G

VTQ (Q, V DS ) = − k1 ( fV DS + Q / CCF ) .

(7)

R1 , R2 If FG T

contributions due to the temperature ( V

R1 , R2

k2 = we

split

the

(T ) ) and due to

FG mobile and trapped gate oxide charges ( VTFG , ion1 and VT , ion 2 )

we have k1VCGS 1 − VGS 2 = ( k 2 − 1) VTFG (T ) + VTQ (Q, VDS ) +

(

FG + k2VTFG , ion1 − VT , ion 2

)

(9)

Since in our design αG =0.8 and R1=R2, eq. (9) becomes

(

FG α GVCGS1 − VGS 2 = VTQ (Q,VDS ) + VTFG ,ion1 − VT ,ion 2

)

(10)

where the dependence on the temperature is cancelled. FG Now, remembering that VTFG , ion1 and VT , ion 2 depend on the bias values of the devices and on the time spent to bias the devices, it is of primary importance to take into account these contributions. With the aim to investigate the contributions of the VT,ion1FG and the VT,ion2FG we performed several measurements of the threshold voltage under different bias and temperature conditions. For example, biasing the cell at constant temperature during the time delay between two consecutive measurements we observed a significant drift of the threshold voltage. Such drift exhibits an asymptotic behavior and therefore fixing the bias of the cell and measuring the threshold voltage we can observe a saturation of that variation. In order to get a saturation the total time to wait is about 1 h that is too much for use in research and lab practice. One example of measured behavior is reported in Figure 5. In the figure we reported twelve measurements and the wait time among two consecutive measurements is 10 minutes. During the wait time in the first seven measurements VGS=5 V and VDS=1 V, whereas during the wait time of the last five measurements the control gate and the drain terminals were connected to ground. For example, when keeping the device biased for the first 10 min we find a threshold voltage drift of about 8 mV which is too much with respect

to the expected resolution of the system, i.e. about 1 mV [14]. Therefore an algorithm devoted to control the measurements set up and which is able to connect the cell to the bias circuit just only during the measurement time (thus for less than 1 s) has been adopted. This approach allows us to neglect the VT,ion1FG and the VT,ion2FG terms and therefore the expression (10) can be simplified as:

α GVCGS1 − VGS 2 = VTQ (Q,V DS ) .

(11)

Figure 5. Measured threshold voltage at constant temperature under different bias conditions during the wait time between two consecutive measurements (VGS=5 V and VDS=1 V from measurement #1 to #7 and VGS = VDS=0 V from measurement #8 to #12) Experimental Results In order to implement the proposed approach a dedicated system (Figure 6) has been designed using offthe shelf components. The INA114 instrumentation amplifier (whose connections INA+ and INA- are shown in the figure) connected to an external voltmeter has been used for differential measurements, by directly implementing the weighted difference Vdiff = (αGVCGS1VGS2). In our prototype system, threshold voltages are also separately measured through remotely controlled voltmeters interfaced to a PC through the GPIB interface. In order to establish / remove the connections to the bias circuit, a digital control tailored on a PIC16F876A μcontroller architecture has been used. Such device housed also the interface circuit between the sensor and a remote control running on a PC, operated by means of a graphical interface. The choice of the hardware has been performed comparing the required performance in terms of speed and the “recurring” / “nonrecurring” costs. To this purpose it should be noted that considering the required speed of the system for the threshold voltage measurement, the main bottle necks are the transient of the system after the bias circuit connection (about 400 ms) and the integration time of the differential signal, which should be accomplished in at least one power line cycle (20 ms) in order to limit the influence of the noise in the system.

The V_enable pin is presently connected to an external power supply and is exploited for both enabling the transfer gates of the cells and for programming. The E2PROM cell is electrically programmed or erased by Fowler-Nordheim tunneling through 14 V pulses at the control gate or at the drain of the cell, while the other terminal is connected to the ground. The programming subsystem is embedded in the circuit of Figure 6. The digital controlled analog switches ‘program_drain’, ‘program_CG’ and ‘drain_2_gnd’ are driven by the microcontroller. The actual pulse length is about 800 µs regardless of the target threshold voltage. During the normal operations as a DNA hybridization detector, the initial threshold voltage value should be set at a relatively high value: preliminary characterizations with DNA material [14] showed that a convenient value is 5.5 V. In this range of threshold voltages the system has been shown to work properly. In fact the minimum threshold voltage is 1.5 V, while the maximum is 5.5 V. In order to verify the performance of the proposed system as a function of the temperature, tests in a thermostatic cell at two different temperatures (25 °C and 40 °C) have been performed. After the temperature value was set, before proceeding with the measurement it was necessary to wait for about 30 min to get an approximately flat temperature. Despite this settling time, the temperature still had variations of 2-3 °C. Three are the measured values: VCGS1, VGS2 and the differential voltage Vdiff = (αGVCGS1-VGS2). Table 1 shows the experimental results, values are in volt. Table 1. Measured values: four different measurements were taken at both 25°C and 40°C T=25°C

aver. 6σ T=40°C

aver. 6σ

VCGS1

VGS2

Vdiff

2.9544 2.9543 2.9542 2.9540 2.9542 0.0010

1.3125 1.3124 1.3123 1.3123 1.3124 0.0006

1.0511 1.0511 1.0511 1.0509

2.9518 2.9515 2.9515 2.9515 2.9516 0.0009

1.3101 1.3099 1.3101 1.3100 1.3100 0.0006

1.0513 1.0513 1.0511 1.0512

aver. 6σ

1.0511 0.0008

Each measurement of the VCGS1, VGS2 and Vdiff, has been calculated by averaging 30 values obtained with a sampling rate of 2 samples/s. At higher temperatures we always measured a lower average threshold voltage. The average threshold voltage drift measured between 25°C and 40°C is about 3 mV. However, the 6σ (where σ = standard deviation) of Vdiff, calculated on all the measurements at both 25°C and 40°C (Table 1), is 0.8 mV, smaller than 1 mV confirming the validity of the proposed approach, at least in a typical laboratory environment situation [14].

Figure 6. Schematic of the system. The sensor is highlighted by a dashed line. Conclusions In this paper a configurable circuit based on a differential topology and devoted to investigate the performance of a sensor dedicated to detect hybridized DNA has been presented. Changes of U.V. absorbance are detected by measuring the analog weighted difference of the threshold voltages of an E2PROM cell and a reference transistor. Experimental results show that the circuit offers a valid solution to decrease the error due to temperature variation during operation. Fluctuations of 3 mV in the threshold voltage for 15°C temperature changes are reduced to less than 1 mV, that is more than an acceptable value for the particular application envisaged. Moreover, the measurement procedure is sufficiently quick (> Zl. As stated above, inertial scavengers may also use rotating masses. Typically these are unbalanced (e.g. semi-circular) so that they may be driven by linear motion. In [7] an analysis is presented which shows that the power limit of such a device, for a semi-circular proof mass m of radius R, is given by:

0.15 0.1 0.05 0 1996

1998

2000

2002

2004

2006

year

Figure 3 - Normalised measured power Pn vs. year of publication. From [8]. Also in [8] it was shown that higher normalized power levels are generally achieved for larger devices, and for lower operating frequencies. For the former, it is likely that practical constraints of integrated microengineered devices reduce the power extraction efficiency. For the latter factor, it is generally the case that high frequency sources have low

excitation amplitudes, and thus require resonant enhancement within the extraction device to achieve optimum power. However, the higher the resonant enhancement (and thus the mechanical Q) needed, the greater is the impact of parasitic damping mechanisms such as viscous drag on the proof mass motion. In such cases the inherent Q (i.e. the quality factor excluding the effect of the transduction damping) becomes the limiting factor on achievable power. A final conclusion from [8] is that reported results do not show any clear differences between the three transducer types in terms of normalised power, and each has been investigated over a wide range of both device size and operating frequency. Recently, commercial inertial energy scavenging devices have begun to appear. These have mostly been based on piezoelectric cantilever designs, with device size in the cm range. For example, the Midé Technology Corp. advertises a piezo scavenger [9] of about 40 cm3 and 50 g in size. This device is reported to provide 2.4 mW at 1 g acceleration, for a drive frequency of 50 Hz. If we use the given dimensions to estimate the internal motion range and proof mass as 2.5 mm and 25 g respectively, then (2) gives a maximum theoretical power at 50 Hz and 1 g of about 60 mW. It should be noted that this device is not optimized for these specific operating conditions, and that (2) does not include any consideration of the efficiency of the power conditioning circuit. With these factors in mind, the device comes reasonably close to what is possible, while not precluding significant future improvement. One possible area for improvement is in transducer damping strength. High frequency devices, requiring high mechanical Q, do not require strong damping by the transduction mechanism, but at lower frequencies the damping force needed to maximize power may well be more than can be practically achieved. This is for different reasons in each of the transduction cases. In piezoelectric devices, the output impedance of the piezo element is dominated by its capacitance, which is too large be tuned out with inductance at the frequencies of interest. This means that the optimum load is the one that matches the magnitude of the capacitative impedance 1/ωC, which is far from matching the real component of the output impedance, as would be optimum if the capacitance were not present or could be compensated. Consequently, a number of groups are looking at improved circuits to get higher power extraction (and stronger damping) from piezoelectric scavengers, e.g [10]. For electromagnetic devices, strong damping forces require a high time rate of change of linked flux. This is inevitably more difficult at low frequencies and small device size, since the slow relative movement of the proof mass demands a very high spatial flux gradient, and a large number of coil turns. The latter is difficult to achieve in microengineered form, and leads to undesirably high coil resistance owing to the high length to diameter ratio of the coil windings. For electrostatic devices, the holding force (and thus the damping strength) depends on the applied voltage and on the spatial rate of change of capacitance. Unfortunately, high absolute capacitance, and thus high capacitance variation, is difficult to achieve in a mechanically variable capacitor compared to a fixed device of similar size. Furthermore, dealing with high voltages is undesirable in a micro-engineered device, and the need for a pre-charge or

priming voltage in these devices is already a disadvantage, which is exacerbated if this voltage is high. 4.

A Non-Resonant Electrostatic Energy Scavenger

Most reported inertial linear energy scavengers have used a resonant mechanical mounting for the inertial proof mass. This is necessary for high frequency devices, where the internal motion range is likely to be greater than the excitation amplitude. However, it necessarily limits effective operation to a narrow range of source frequencies. For low frequency operation, such as for body motion excitation, the internal motion enhancement is not required, and operation across a wide range of excitation frequencies and waveforms is essential for a practical application. For that reason we pioneered an electrostatic device which has a non-resonant proof mass mounting, whose internal motion is non-linear and discontinuous [11]. This is illustrated in Fig. 4. The mass is pre-charged in one position, where it is held in place until the external acceleration is enough to overcome the electrostatic force. At that point the mass accelerates across to the other side of the frame, where it discharges its energy. Thus it can operate equally effectively for a wide range of input motions. Since the pre-charge voltage sets the holding force, this parameter can in principle be used to dynamically optimize the power for different motion amplitudes.

Figure 4 Electrostatic energy scavenger for low frequency applications (from [11]). This device also illustrates some of the issues discussed in section 3. The moving plate dimensions are ≈ 11 × 11 × 1 mm, and the maximum capacitance is about 150 pF. This required a pre-charge voltage of 30 V to generate 120 nJ/cycle. Increasing the starting capacitance would allow reduction of the pre-charge voltage without loss of output power, and would lessen the effects of parasitic capacitances. The output is in the form of high voltage pulses, which creates considerable demands on the power conditioning circuitry [12]. 5.

Sensor Node Power Requirements

Crucial to the practical exploitation of energy scavenging devices is the identification of applications whose power requirements are within the range such scavengers can achieve in the environment determined by the application. Solar power has been the most successful scavenging technology to date, benefiting from a well developed technology, strong compatibility with electronic integration, and reasonable cost. However, solar cells are dependent on a strong and reliable source of light, and must be correctly oriented and free of obstructions. In [13], solar cells were used to power wireless senor nodes only 16 mm3 in size, with on-board (passive) optical data communication, two sensors, and some processing and control circuitry. However, the light source for powering was a remote laser rather than ambient light. Vibration –powered energy harvesters have also been used to demonstrate fully autonomous self-powered sensor nodes. In [14], a wireless temperature sensor is reported which was powered by piezoelectric transduction from vibration present on a staircase to which the device was attached. The scavenger provided 30 µW under continuous stairway traffic, enough to power the sensor electronics and short range data transmission. Wireless sensor arrays are attracting great interest in many application domains, and appear to be the most attractive application for energy scavenging, as they often have low power requirements, combined with a need for low cost and size, and ease of maintenance. The three main power requirements in such devices are the sensors themselves, the signal conditioning circuitry, and the wireless data communication. For sensor types generating modest amounts of data and requiring low sampling and transmission rates, total power requirements in the micro-watt level range are realistic [15], and are becoming increasingly so with advances in low power analogue and digital circuitry. 6.

Conclusions

Motion or vibration energy scavenging is an attractive approach to powering wireless electronic devices, particularly sensor nodes. While achievable power levels are modest, they are sufficient for many applications, and reported devices continue to advance towards realizing power output near the ultimate limits. References [1] T. Starner and J. A. Paradiso, "Human Generated Power for Mobile Electronics," in Low-Power Electronics Design, C. Piquet, Ed.: CRC Press, 2004, pp. 1-35. [2] S. Roundy, D. Steingart, L. Frechette, P. Wright, and J. Rabaey, "Power sources for wireless sensor networks," in Wireless Sensor Networks, Proceedings, vol. 2920, Lecture Notes in Computer Science, 2004, pp. 1-17. [3] N. S. Shenck and J. A. Paradiso, "Energy scavenging with shoe-mounted piezoelectrics," IEEE Micro, vol. 21, pp. 30-42, 2001.

[4] P. D. Mitcheson, T. C. Green, E. M. Yeatman, and A. S. Holmes, "Architectures for vibration-driven micropower generators," Microelectromechanical Systems, Journal of, vol. 13, pp. 429-440, 2004. [5] P. D. Mitcheson, E. K. Reilly, T. Toh, P. K. Wright, and E. M. Yeatman, "Performance Limits of the Three MEMS Inertial Energy Generator Transduction Types," J. Micromechanics & Microengineering, vol. in press, 2007. [6] P. D. Mitcheson, P. Miao, B. H. Stark, E. M. Yeatman, A. S. Holmes, and T. C. Green, "MEMS electrostatic micropower generator for low frequency operation," Sensors and Actuators A: Physical, vol. 115, pp. 523529, 2004. [7] E. M. Yeatman, "Energy Scavenging from Motion Using Rotating and Gyroscopic Proof Masses," submitted to J. Mech. Eng. Sci., 2007. [8] P. D. Mitcheson, E. K. Reilly, T. Toh, P. K. Wright, and E. M. Yeatman, "Transduction Mechanisms and Power Density for MEMS Inertial Energy Scavengers," presented at Power MEMS 2006, Berkeley, CA, 2006. [9] M. T. Corp., "Volture - Piezoelectric Energy Harvester," http://www.mide.com/prod_energy_harvester.html#, 2007. [10] D. Guyomar, A. Badel, E. Lefeuvre, and C. Richard, "Toward energy harvesting using active materials and conversion improvement by nonlinear processing," IEEE Transactions on Ultrasonics Ferroelectrics and Frequency Control, vol. 52, pp. 584-595, 2005. [11] P. Miao, P. D. Mitcheson, A. S. Holmes, E. M. Yeatman, T. C. Green, and B. H. Stark, "MEMS inertial power generators for biomedical applications," Microsystem Technologies, vol. 12, pp. 1079-1083, 2006. [12] B. H. Stark, P. D. Mitcheson, P. Miao, T. C. Green, E. M. Yeatman, and A. S. Holmes, "Converter Circuit Design, Semiconductor Device Selection and Analysis of Parasitics for Micropower Electrostatic Generators," Power Electronics, IEEE Transactions on, vol. 21, pp. 27-37, 2006. [13] B. A. Warneke, M. D. Scott, B. S. Leibowitz, L. Zhou, C. L. Bellew, J. A. Chediak, J. M. Kahn, B. E. Boser, and K. S. J. Pister, "An autonomous 16 mm/sup 3/ solarpowered node for distributed wireless sensor networks," presented at Sensors, 2002. Proceedings of IEEE, 2002. [14] E. S. Leland, E. M. Lai, and P. K. Wright, "A SelfPowered Wireless Sensor for Indoor Environmental Monitoring," presented at 2004 Wireless Networking Symposium, University of Texas at Austin Department of Electrical & Computer Engineering, 2004. [15] E. M. Yeatman, "Advances In Power Sources For Wireless Sensor Nodes," presented at 1st International Workshop on Body Sensor Networks, April 6-7, London, 2004.

SENSOR NETWORKS FOR INDUSTRIAL APPLICATIONS A. Flammini, P. Ferrari, D. Marioli, E. Sisinni, A. Taroni University of Brescia - Department of Electronics for the Automation Via Branze 38 - 25123 Brescia – [email protected] Tel: +39 030 3715627 – Fax: +39 030 380014 Web: www.ing.unibs.it/~wsnlab Abstract Industrial applications are moving from centralized architectures towards distributed ones, thanks to cost effectiveness, better flexibility, scalability, reliability and diagnostic functionalities. The use of sensors in industrial communications improves overall plant performance since sensor information can be used by several equipments and shared on the Web. A communication system suitable for computers and PLCs, that exchanges a large amount of data with soft real-time constrains, can be hardly adapted to sensors, especially to simple and low-cost ones. In fact, these devices typically require a cyclic, isochronous and hard realtime exchange of few data. For this reason, specific fieldbuses have been widely used to realize industrial sensor networks, while high-level industrial communication systems take advantage of Ethernet/Internet and, more recently, wireless technologies. In these years, Ethernet-based solutions that meet real-time operation requirements, called Real-Time Ethernet, are replacing traditional fieldbuses and research activities in real-time wireless sensor networking are growing. In this paper, following an overview of the state-of-art of real-time sensor networks for industrial applications, problems and possible approaches to solve them are presented, with particular reference to methods and instrumentation for performance measurement. I. INTRODUCTION Sensor networks Traditionally, sensors output is furnished by simple electronic circuits that provide a standard analog interface (e.g. 0-5V, 4-20mA, and so on). Thanks to availability of low cost microcontrollers, a new generation of sensors, normally called “smart sensors”, is growing [1]. They provide improvements in terms of linearity, signal-to-noise ratio and diagnostic features; in many cases, network connectivity is also supported. Unfortunately there is not an unique communication standard for sensor networking. In fact, to support computer connectivity there is no choice; Ethernet [2], together with Internet protocols, is the universally recognized solution. On the contrary, sensor networking requires very simple and low-cost protocols to be supported by a 8-bit low cost microcontroller and several incompatible solutions are competing each other for market leadership in a particular application. Sensor networking, in fact, is the objective of many application fields: military, agriculture, environment monitoring, home automation, health and welfare, automotive, industrial applications. Each field has its own requirements: for instance, health and welfare need sensor compactness and wireless [3], while low cost is imperative for automotive and home automation [4]. More 1-4244-1245-5/07/$25.00 ©2007 IEEE

generally, a sensor network is evaluated with respect to some characteristics as: transmission range, that could be greatly affected by the physical mean; compactness; mobility, that implies wireless sensors with autonomous power source [5]; cost; performance, that is roughly represented by bit rate but that could depend on general timing requirements, like latency and jitter; least but not last robustness, that is safety and security [6]. The industrial scenario As regards sensor networks for most of industrial applications, sensor compactness and mobility are not critical requirements. In fact sensors are usually placed in a fixed place and power supply availability is practically everywhere. On the contrary, robustness is a key factor, as strong electromagnetic power sources (welders, smelting furnaces, motors and so on) [7] can sensibly affect transmission quality. In addition, it is very important to transfer information within a small, fixed and known time and therefore performance is a crucial point. The best way to respect deadline in information transfer is a centralized architecture; sensors are read when needed and event reaction time, that is the delay between an input event and the related output actuation, is minimal and well-known. Even in this employment, smart sensors offer several advantages; in fact, the term smart transducer is widely used to define a transducer whose output is something more than raw measurement data. A formal definition can be found in the standard IEEE 1451.2 [8]: “a Smart Transducer provides functions beyond those necessary for generating a correct representation of a sensed or controlled quantity. This functionality typically simplifies the integration of the transducer into applications in a networked environment”. A sensor is smart if it can be managed regardless peculiarities due to its vendor or the adopted interfacing protocol. Transducers become PLUG & PLAY eliminating errors due to manual configuration and data entering: they can be installed, upgraded, replaced or moved with minimum effort. A smart transducer implements a general model for data, control, timing, configuration and calibration and it contains a standardized Transducer Electronic Data Sheets (TEDS) with manufacture-related data. The simplified block diagram of such a device is given in Figure 1. The key aspect is the network capability; analog point-to-point interfaces (e.g. 4-20mA, 0-10V…) can be substituted by a single, low cost, and reliable digital field area network; this is the first step towards a real distributed system. Advantages of distributed architectures are countless, including increased flexibility, improved performances, cost reduction due to cabling diminution, easiness of installation and maintenance.

n e t w o r k

μC

+ -

Sensing Analog Actuating Conditioning Element

Logic and network interface

Figure 1: Smart transducer block diagram. Unfortunately, distributed architectures imply transmission delays that could heavily affect performance. For the sake of clarity, an example is analyzed. If we suppose a simple distributed architecture, as depicted in Figure 2, then a simple program as “if A>B then immediately actuate C”, that properly works in a centralized architecture, could present some problems.

Controller: “if A>B then immediately actuate C” Network

Sensor A

Sensor B

Actuator C

Figure 2: Distributed architecture. In fact, A and B quantities could be sampled in different instants, that is A=A(t0) and B=B(t1) and therefore can be hardly compared. Even if we suppose t0≈t1, it could be difficult to exactly estimate the transmission time td,A and td,B of A and B to controller; typically they can be approximated by known limits ( Tmin < td,A ≠ td,B < Tmax ). Even if we suppose that elaboration starts as soon as sensor messages arrive and the actuator actuates C as soon the controller message arrives, elaboration takes time telab and the controller message takes time td,act to reach the actuator. Consequently there is a delay time Td that, if we neglect sampling and actuating time, is equal to Td = max(td,A , td,B) + telab + td,C. Td could be significant and variable, because td,A , td,B and td,C could depend on network traffic. This simple example shows how performance of a distributed architecture could be affected by network and application behavior; in fact the above Td expression is simplified due to strong hypothesis we have done. In addition, terms usage in industrial communications could be quite confusing, as expressions like “real-time” or “determinism” are often misused. According to International Electrotechnical Commission (IEC) [9], real-time is the ability of a system to provide a required result in a bounded time, that is maximum latency is “a priori” known. Consequently, a real-time communication system is able to transfer data in real-time. In some cases it is used a distinction between “soft real-time”, with a statistical real-time behavior, and “hard real-time”, where the maximum latency shall be respected in all cases, as in the IEC 61784-2 definition.

Determinism is related to the ability to set an imposed and invariable latency, that is the required result is provided in a fixed, known and repeatable time. However, it is often used in substitution of hard real-time, that is a less stringent constrain. Isochrony refers to the ability to be strictly repetitive in time; an isochronous communication system imposes that each data transfer takes action in a strictly cyclic way with a very low jitter, where jitter is intended as the difference between the maximum and the minimum value of cycle time. Some industrial applications, as packaging, manufacturing, wood machining or plastic extrusion, require high performance systems to achieve a cost reduction [10]. Data exchange must be fast, reliable and deterministic, that is latency times must be in the order of hundreds of microseconds to correctly close control loops between twin drives, while jitter times must be one order of magnitude lower. In our example, traffic between controller and transducers can be organized in a cyclic way, as shown in Figure 3, i.e. the controller periodically (every cycle time) exchanges information organized in time slots with field devices. If the chosen physical and medium layers ensure the respect of time slot bounds, then communication is real-time, deterministic and isochronous, that is frames are sent with a constant inter-arrival time. However, the system behavior, characterized by time Td, could show a considerable jitter, because time between event (A>B) and reaction (C actuated) depends on sensor sampling time and, more generally, on synchronization among nodes application tasks. Slot A

B

C Cycle k

t0



A

B

C

Cycle k+1



A

B



C

Cycle k+2 Time

t1

Figure 3: Cyclic traffic exchange. For this reason, industrial communication protocols often provide some synchronization services, as input/output synchronization commands (e.g. global read, global write) or application synchronization utilities in order to achieve determinism. In fact, if all the nodes share a common sense of time (i.e. they have synchronized clocks) and the isochronous scheme of Figure 3 is adopted, then determinism could be reached simply modifying the system program into: - sample A and B at time t0 (i.e. the start of the cycle k) - if A(t0)>B(t0) then actuate C at time t1 where time interval between t0 and t1 must be greater than three times the cycle time, supposing the elaboration is synchronized with the start of the (k+1)th cycle and the elaboration time is less than the cycle time in order to deliver message to C in (k+2)th cycle. Obviously, if the sensor network does not imply actions to take in real-time, but is only used to collect data from sensors, the only need is to “accurately” reconstruct the temporal sequence of data (data timestamping), because time accuracy/resolution affects data value accuracy/resolution.

This can be achieved by a timestamping mechanism in every node and a good synchronization among nodes. In conclusions, due to hard constrains in terms of performance, robustness and cost, sensor networks for industrial applications, usually called fieldbuses, are often “tailored” solutions. Fieldbuses are used in most of industrial plants to digitally link subsystems and to transfer few data in real-time. They are typically characterized by a cyclic behavior, synchronization utilities, a quite low data rate (Mbit/s), good efficiency (number of data bit with respect to transmitted bit), a good transmission range (100m), low cost and a special attention to safety [11,12,13,14]. They are similar to proprietary technologies; they reach satisfactory performances but proposals of different vendors typically can not coexist. There are several open standards with pros and cons describing these networks; for instance, DeviceNet or PROFIBUS [15] are quite simple and can be easily integrated in low-cost microcontrollers, reducing the need for external components (e.g. some 8-bit Freescale or Microchip microcontrollers provide a CANbus 2.0B interface). Nowadays, fieldbuses support the most of sensor networks in industrial applications, although many industrial fields [16], with few, simple and close sensors, still adopt traditional centralized architectures. II. FROM FIELDBUS TO ETHERNET AND RTE As high level communication systems adopt solutions based on TCP/IP [17], as for instance OPC (Ole for Process Control [18]), fieldbuses can be hardly integrated [19]. In addition, fieldbuses are often poor as regard diagnostic and self-configuring tools. Ethernet is the most common used physical layer of widespread TCP/IP-based solutions and it is widely used in industrial plants at PLC (Programmable Logic Controller) and SCADA (Supervisory Control And Data Acquisition) level, where it is called “Industrial Ethernet” [20]. The idea to use it even at the field level took place in the last years thanks to the more efficient switch-based architecture, to the increased transmission rate and to the availability of low-cost devices [21]. Ethernet seems unsuitable for real-time applications because the a priori estimation of the maximum transmission time of a data packet is impossible [22]. This is mainly due to the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) for the Medium Access Control (MAC) and to unpredictable delays introduced by switches, that depends on network topology, traffic conditions, switch technology (“Store&Forward”, “Cut-Through”, [23]), and so on. An important incentive to the diffusion of Ethernet in industrial plants comes from IEC61784-1 [15], that describes and acknowledges some commercial solutions of industrial Ethernet that in some cases could be used down to sensor level, as HSE (Fieldbus Foundation for Ethernet), Ethernet/IP, PROFINET. These protocols does not guarantee performance suitable for most of real-time control applications, therefore other solutions are emerging, called Real-Time Ethernet (RTE), as Powerlink [24], PROFINET IO [25], EtherCAT [26], MODBUS-RTPS [27] and so on, including dedicated solutions [28]. These technologies allow more powerful performances if compared to traditional fieldbuses, taking advantage from the high

performance of Ethernet (e.g. 100Mbit/s or more instead of typical 1-10Mbit/s of high-performance fieldbuses). Real-Time Ethernet (RTE) is defined by IEC61784-2 as the ISO/IEC 8802-3-based network that includes real-time communication [2]. RTE solve the non-determinism problem of Ethernet modifying media access rules by means of software protocols (e.g. master-slave protocols based on Time Division Multiple Access -TDMA-) or thanks to ad hoc switches or network interfaces. There is not a universally acknowledged single RTE protocol and the above cited solutions differ on the way they achieve determinism. Synchronization among nodes, that is all the nodes follow a common clock (master clock), takes a very important role in RTE. Synchronization methods can vary from simple proprietary protocols [29], as broadcast triggering messages, to the use of standard solutions, as Network Time Protocol (NTP) or Precision Time Protocol (PTP) described in standard IEEE1588 [30,31]. At the present, standard IEEE1588 seems the most promising synchronization method because it is independent from technology and it allows fullsoftware realizations but, in that case, it strictly depends on the application level. In addition, in industrial plants, star topologies are considered unsuitable, so if many switches are cascaded [32], propagation delay of a frame is asymmetric and IEEE1588 could yield to considerable estimation errors. Hardware-software solutions can be used in order to increase performance of IEEE1588 achieving an accurate timestamping of frames; by this way a synchronization in the order of 100 ns can be reach, but RTE protocols are not supported [33]. Besides IEEE1588-based approaches, new ideas have been proposed to synchronize nodes. For instance a suitable use of GPS (Global Positioning System) to obtain a Universal Coordinated Time (UTC) reference is described in [34]. Another common aim of an RTE network is to be compatible with TCP/IP traffic. In fact, industrial communications should support at the same time, on the same media, a fast (isochronous) real-time data exchange and, if an event occurs (alarm or diagnostic or configuration activity for a certain node), complex and acyclic communication. Some research activities have been carried out to quantify RTE performance reduction as a function of bandwidth dedicated to TCP/IP traffic [35], but methodologies for the test environment setup, like load generation and profile, are rather rough [36,37]. As an example of an RTE, PROFINET IO is briefly described [38]. PROFINET IO performance is described with the class number: RT_Class 1 (RT, Real-Time) is used in systems requiring cycle time down to tenth of milliseconds; RT_Class 2 (also called IRTflex, Isochronous Real-Time with Flexible network topology) and RT_Class 3 (also called IRTtop) are used with applications requiring isochrony and cycle time shorter than 1 ms. The PROFINET concept is to support on the same media hard real-time traffic, soft realtime traffic and non real-time traffic as TCP/IP. PROFINET IO defines IO-Controllers (i.e intelligent devices which carry out automation tasks), IO-Devices (i.e. field devices like sensors, actuators, IO module etc.) and IO-Supervisors for configuration and diagnosis purposes.

PROFINET IO data exchange is based on a highly repeatable cycle as described in IEC61158-5-10 [39] and illustrated in Figure 4. A synchronization message (sync frame) signals the cycle start. Several phases can be recognized in a cycle: • RED phase: during this phase, only RT_Class 3 messages are sent on a time scheduled basis through “a priori” defined path. This means that all PROFINET IO RT_Class 3 devices know when, and on which physical port, they are allowed to talk or listen to. • ORANGE phase: only RT_Class 2 frames are sent in this phase. Also RT_Class 2 has a time base schedule but the physical path is not defined. • GREEN phase: this phase is composed of Ethernet message managed using the Ethernet priorities (IEEE 802.1Q). GREEN phase communication is used by: RT_Class 2 devices with extra frame to send; RT_Class 1 (RT) devices which are not synchronized each other; all the rest of the IP based communication (TCP, UDP). It should be noted that, as a result, RT_Class 1 frames may suffer from low but unpredictable delay. • YELLOW phase: this is a transition phase used for the same type of traffic of the GREEN phase. During this period, only frames which can be completely transferred within the end of the YELLOW phase are transmitted. A relevant portion of the cycle (in the GREEN phase) is left for non real-time communication (NRT) such as TCP or UDP. Such traffic has low priority tags and very variable delays. Indeed, IP traffic is used for big data transfers, and the only important thing is the bandwidth. In PROFINET IO, NRT phase occupies at least the 40% of the total bandwidth.

Tcycle TProfinet ≤ 60%Tcycle S y n c

RT Class 3 RED (optional)

RT Class 2 ORANGE (optional)

Prio Prio 6 Prio Prio … Prio 7 RT Class 2/1 6 5 0 GREEN (mandatory)

YELLOW (optional)

Figure 4: PROFINET IO cycle. No collisions, no delays can happen within RT_Class 3 since the scheduling sequence in each cycle is “a priori” known and always identical. The engineering (network configuration) tool calculates the trip for every frame of a cycle and downloads the schedule in the network infrastructure. This means that the network infrastructure must be PROFINET IO compliant; in fact special switches must be used, that thanks to a powerful ASIC [40], forward RT_Class 3 frames looking only at the time schedule, without MAC address check. On the contrary, RT_Class 2 frames are forwarded using MAC addresses, as usual. In case traffic bursts, RT_Class 2 frames can be buffered and delayed to the next GREEN phase. RT_Class 2 exhibits a jitter higher than RT_Class 3. Normally these PROFINET IO compliant switches are integrated directly into RT_Class 3 nodes; the introduction of a normal switch (i.e. a “Store&Forward” [23] switch) could seriously affect overall performance.

III. METRICS AND INSTRUMENTS FOR RTE RTE networks are an example of emerging technology where scientific research and industrial interest converge. RTE networks are a new topic and recently some workshops are appearing [41,42,43]. Besides a lack of widespread knowledge, a general absence of measurement methods and instruments characterizes RTE-based applications. Particularly, a complete set of suitable parameters to characterize an RTE-based application is not defined; moreover, even if a feature derived from the Information and Communication Technology (ICT) field seems adequate, measurement methodologies and test environments are often not available. For instance bandwidth and latency are wellknown and widely used in Ethernet [44] and Internet [45] and they appear correct for RTE also. However, real bandwidth measurement is rather difficult, because it depends on data and on the state of linked nodes; actually, the peak value is often considered in the best case. As regard latency, it is usually measured in an empirical way thanks to instruments that measure the normally called “roundtrip delay”, that is defined as the time interval between the transmission of special frames and the receipt of the related acknowledge [46]. The most famous method is the Ping command that is based on ICMP (Internet Control Message Protocol). Obviously, this method does not support the resolution required by RTE networks. The above cited IEC61784 suggests some performance indicators: • delivery time: the time needed to convey application data from one node (source) to another node (destination). • time synchronization accuracy: the maximum deviation between any two node clocks. • non-time-based synchronization accuracy: the maximum jitter of the cyclic behavior of any two nodes when such cyclic behavior is established by means of periodical events over the network. For instance, this is the case of some RTE protocols that use a network message to signal the start of a cycle. In such protocols the sharing of a common clocks reference is not required. • redundancy recovery time: “the maximum time from failure to become fully operational again in case of a single permanent failure”. • throughput RTE: the total amount of RTE application data (by octet length) on one link per second. • non-RTE bandwidth: “the percentage of bandwidth, which can be used for non RTE communication on one link”. The total link bandwidth shall also be specified, since they are related to each other. Furthermore, several other indicators can be used [47,48]. For instance “Stack Traversal Time” is the time required by data to pass through the communication stack from top (application layer) to bottom (physical layer). “Event Reaction Time” is the time required by the system to acknowledge an external event (e.g input change) generating a response action. This time is very important in practical applications and it significantly depends on application level implementation. As the experimental evaluation of these

indicators is quite difficult on an industrial plant, the research activity is focused on provide simulation tools. Obviously network simulators like OPNET [49] or OMNET++ [50] do not natively support RTE protocols, therefore a great effort is spent to develop an effective model of an RTE node [51]. As regards measurement instrumentation, in the ICT field some instruments are used to associate a time reference to Ethernet frames: from the PC-based instruments like WireShark (formerly Ethereal), a well known network analyser software, [52, 53] with resolution in the order of 0.1 ms, to the high-performance network analyzers. The latter allows a time resolution in the order of tenths of nanoseconds that could be suitable for RTE networks; on the other hand, limits are the compactness, the cost and the robustness typically needed by industrial environments. As an example of new instruments designed for ICT, WAND group [54,55] of University of Waikato in New Zealand has developed a new instrument based on programmable logic devices that adds timestamps to every Ethernet packet. At present about 100 of these instruments, synchronized by GPS, are used all over the world to perform statistical analysis of Internet traffic. As software-based instruments are not adequate and network analysers allow only a costly and localized measurement, often RTE performance characterization is done looking at input and output signals, for instance measuring the event reaction time with respect to external event and reaction. In order to develop instruments tailored to RTE networks, a multi-probes approach must be considered taking advantage from recent developments in the FPGA technology and in the availability of network processors [56]. In fact, by means of a multi-probe architecture, it is possible to experimentally measure delay times (i.e. through a switch) and verify synchronization among nodes. A new instrument A new, low-cost, multi-probe instrument has been recently proposed [57]. General architecture is shown in Figure 5. The instrument can be viewed as a network of probes designed to simultaneously log Ethernet traffic in different links of a target RTE network. This “parallel” network, called measurement network, conveys data (logged by probes) toward a supervision equipment, called monitor station. Probes are requested to associate a reliable timestamp to every frame that transits on the Ethernet link they monitor. This results in a special probe architecture that enables RTE full-duplex logging together with strict time synchronization among probes. Monitor station must store and elaborate all the incoming data, thus the only critical point is the system bandwidth, that is the ability to manage all the data without dropping frames. In fact, logged frames and timestamping related data must be transferred, resulting in quickly growing bandwidth requirements. Generally, if 100BaseT high-bandwidth RTE protocols are considered, the measurement network should work with 1000BaseT or more. In the realized implementation, a 1000BaseT measurement network has been used.

RTE Station

1000BaseT Measurement Network 1000BaseT Switch

Probe

Monitor Station RTE Network Switch

Switch

Probe

RTE Station

Probe

RTE Station

RTE Station

RTE Station

Figure 5: General architecture of the new multi-probe instrument. One of the objectives during the development of the new instrument architecture was cost limitation. This led to have single-chip FPGA-based probes and a single Monitor Station implemented using a PC. Moreover, the monitor station can use open source user interface programs like the above cited WireShark. The probe local time is constantly synchronized with a reference clock despite local crystal oscillator variations (temperature, aging, etc). Synchronization Unit can operate using multiple synchronism sources: 1-PPS signal from an external source or IEEE1588 Sync Message coming from the measurement network. The local time is synthesized with an adder structure [58]. Briefly, an increment step is summed to the time register at every clock period of the local oscillator. Drift and offset can be compensated adjusting the increment step with a suitable control algorithm. The increment step is refreshed each time a synchronism event happens; it means 1 s with 1-PPS and 2 s with IEEE1588 PTP. A two-probes prototype has been experimentally characterized comparing performance to a powerful singleprobe network analyzer: Endace NinjaCapture 1500 [59]. The test network is the PROFINET IO Class 1 system shown in Figure 6. IO-Controller

Switch TAP 1

317-2PN/DP

A

B

IO-Device TAP 2 A

B ET200S PN

Monitoring Port

Monitoring Port

Figure 6: Experimental setup (PROFINET IO Class1 network). Two Ethernet TAPs have been inserted in the network in order to capture traffic before and after the switch. A TAP can duplicate full-duplex traffic, so it has two monitoring ports (A and B) one for each direction. The metric to be compared is the experimental evaluation of the propagation delay of the switch. Propagation delays can be measured in the following modes:



Connecting an input of the Ninjacapture to the monitoring output A of TAP 1 and the other to output A of TAP 2. Delay along a single direction can be measured, since NinjaCapture has two logging inputs. Delay in the reverse direction can be measured connecting the NinjaCapture to output B of the TAPs. • Using a single probe of the proposed instrument that has two input ports. Port 1 must be connected to the monitoring output A of TAP 1 and Port 2 to output A of TAP 2. As in the previous case, two separate measurements are needed to characterize the two traffic directions. • Using two probes of the proposed instrument. Now the instrument has four logging inputs. The two inputs of Probe 1 are connected to outputs A and B of TAP 1; inputs of Probe 2 are connected to outputs A and B of TAP 2. A single acquisition campaign is sufficient for estimation of switch behavior in both traffic directions. The measure of the propagation delay of this “Store&Forward” switch has been carried out with 64-byte long frames. Measurements results have been reported in Table 1. Generally, they are comparable even if the conditions are different. In particular, the proposed instrument can reduce measurement times, since a single setup is enough, and the user can save money since it is cheaper than NinjaCapture. Table 1: Switch propagation delay. (IOC: IO controller, IOD: IO device). Direction Ninja IOC → IOD Capture IOD → IOC Single IOC → IOD probe IOD → IOC Two IOC → IOD probes IOD → IOC

Switch propagation delay (ns) Max. Ave. Std. dev. 15 567

12 363

1670

12 483

1650

15 572

12 404

1653

15 563

12 496

1658

15 592

12 381

1693

15 601

12 460

1679

15 576

IV. THE WIRELESS OPPORTUNITY As previously stated, traditional networking offers many advantages but requires cables to interconnect devices. This leads to high installation and maintenance costs, e.g. due to low scalability and high failure rate of connectors. For this reason, wireless technologies gained an enormous success in the consumer goods industry in the last few years. In addition, the adoption of wireless solutions at the sensor level offers other advantages as continuous, high resolution, ubiquitous sensing, provides support for mobility, adds redundancy and takes advantage of MEMS technology. In particular, besides high power consumptions, high area coverage and high cost solution such as the well known and mature mobile phone technologies (GSM, GPRS and UMTS just o cite few of them), two standards have monopolized the market of the Local/Personal Area Networks: IEEE802.11 [60] and IEEE802.15.1 [61]. The former is the wireless

counterpart of the Ethernet standard and implements lower levels of WiFi [62], while the latter constitutes lower levels of the proprietary Bluetooth (BT) [63] solution. The main attractive of both of them is that they do not require any sort of frequency licensing because operate in the ISM (Industrial, Scientific and Medical) radio frequency region. However, WiFi and BT have been designed to address requirements of office/personal communication, and cannot efficiently be used to realize Wireless Sensor Networks (WSNs), as better explained in next sections. Obviously, advantages due to the absence of cables could be usefully exploited in several fields and many efforts have been done in this direction. For example, in the past, novel trends [64] have emerged in the agricultural sector converged in the so called “precision agriculture”, that concentrates on providing the means for observing, assessing and controlling agricultural practices. In this way, it would be possible to detect parasites on the field and automatically choose the right type and amount of insecticide. Another field where wireless technologies have been widely used is “environmental monitoring”; just to mention some applications, it is possible to monitor air quality in real-time by means of unattended stations or collect data in places that discourages human presence. Another interesting application is in the field of “smart structures”, that comprises home and building automation; in this case, a wireless sensor and actuator network is integrated within a building to improve living conditions and reduce overall energy consumption. Also “medical and health care” are fields where WSNs have been successfully employed; e.g. it is possible to ensure patients continuous monitoring without limiting their mobility. Wireless Sensor Networks As stated in previous section, wireless communications are an effective and reliable solution in home and office automations. Generally speaking, several medium can be exploited, including light and ultrasound, but considerations regarding data size, rates and area coverage make RF links more attractive. Many standards have been proposed to satisfy requirements of the consumer world, as proved by the IEEE802 subgroups that cope with these topics (refer to Figure 7), but the most interesting for WSN applications are probably those comprised in the IEEE802.15 [65] working group, whose effort focuses on the development of Personal Area Networks or short distance wireless networks (≈10 m). In particular, here is defined the concept of Personal Operating Space, a spherical region that surrounds a wireless device with a radius of 10 m. Even if originally designed for portable and mobile computing devices such as PCs, Personal Digital Assistants (PDAs), cell phones, pagers, and consumer electronics, it may be successfully applied to WSNs. However, it must be underlined that large scale applications in the sensor networking area are yet in the development stage. First of all, it is important to distinguish between the idea behind the WSN concept and implications related to an industrial scenario, better described further. From a general point of view, a WSN is made up of a large number of tiny

devices (sensors), which are densely deployed and collaborate to monitor and analyze a phenomenon of interest [64]. IEEE 802: LAN/MAN Stds IEEE 802.1: High Level I/F

IEEE 802.15 Wireless Personal Area Network (WPAN) Working Group

IEEE 802.3: ETH IEEE 802.11: WLAN

Task Group 1: WPAN/Bluetooth™

IEEE 802.15: WPAN

Task Group 2: Coexistence

IEEE 802.16: WMAN

Task Group 3: WPAN High Rate IEEE 802.18: Radio Reg. IEEE 802.19: Coexistence IEEE 802.20: Mobile BWA

Task Group 4: WPAN Low Rate/ZB Task Group 5: WPAN Mesh

IEEE 802.21: Media Independent Handoff IEEE 802.22: Wireless Regional Area Networks

Figure 7: IEEE802 family wireless standards. Due to cost and dimension constrains they have limited computational resources; power consumption must be as low as possible to ensure a true autonomous activity. In addition, sensors could be randomly positioned thus requiring localization and self-organizing capability. Besides issues considered in this paper, there are other questions that must be considered in a wireless system. In particular, security could be a key aspect; air is an open medium and it is easy for an attacker to malicious alter transmissions or make the link unreliable injecting jam sequences. The block diagram of a wireless sensor node is represented in Figure 8.

Localization

Transducer

ADC DAC

Power Source

CPU Storage

Synchronization

TX/RX

Power Manager

Figure 8: Wireless transducer block diagram. As every smart sensor, a wireless transducer consists of three main parts: a sensing unit, a processing unit and a transceiver unit. In addition, a power manager is present to handle on board power sources, such as electrochemical batteries or more exotic power scavenging units. Moreover, most of the performed tasks require also the knowledge of positions and time, furnished by proper localization and synchronization units. Many researchers are currently engaged in the design of proprietary schemes that fulfil such requirements, each one with its pros and cons. However, according to authors the most promising solution is to adapt standard solutions, that are already available on the market

and can exploit huge volume production and mature technologies, to the application under investigation. In the following, aspects regarding power consumption, localization and communication architecture will be detailed, while in next section the industrial scenario will be considered. Power consumption Power consumption of a wireless sensor node can be divided into three different domains: sensing, processing and communicating. The first one is strictly related to the application and in most of cases can be neglected. As regards data processing, usually processor consumption during the active phase decreases by an order of magnitude or less if compared with that needed in the communication phase. As explained in [66], supposing a Rayleigh fading and a forth order loss law, the energy required to transmit 1KB over a distance of 100 m is approximately the same as that for executing 3 million of operations by a 100 MIPS/W processor. From another point of view, it is convenient to implement complex algorithms if this result in shorter data packets and/or in a more robust data link that requires less retransmissions. It is a well known results that consumption is proportional to the voltage supply (Vdd), and to the operating frequency (f), i.e. P ∝ Vdd,f [67]. This relationship suggests two strategies to lower consumption: dynamic scale voltage, i.e. reducing the supply voltage Vdd as low as possible, and changing the CPU clock frequency f according to the computational load (usually, microprocessor uses a low and a high frequency oscillator in the idle and active phase respectively). The most demanding unit is thus the transceiver. If we consider short range (≈10 m) systems operating in the GHz range with low radiation power (≈0 dBm), energy required to transmit is almost the same as that required in data reception. Obviously, devices spent most of their time doing nothing; this means that low duty cycle strategies, probably the most diffused solution in battery supplied nodes, can be applied. What really matters is the average current consumption Icc,mean of the wireless sensor. For this reason becomes fundamental evaluate not only the active power but also consumption in standby mode and startup phase duration. If we consider a simple node that: • wakes-up every T seconds (it depends on the Medium Access Protocol implemented), • takes Ta [s] at Ia [A] to start-up and measure quantities (processing phase) • takes TRF [s] at IRF [A] to transmit and receive (transceiver phase) information by means of the RF link • requires Isleep [A] in the standby phase the Icc,mean can be computed as shown in equation (1):

I cc,mean =

I a ⋅ Ta + I RF ⋅ TRF + Isleep ⋅ (T − Ta − TRF ) T

(1).

Designer can adapt Ta and TRF (e.g. shortening the measuring phase or choosing a very simple protocol and/or a high

transfer rate) so that Icc,mean remains in the order of Isleep, as shown in equations (2):

Ta , TRF αn): Vin+

Vin-

Figure 3: Schematic circuit of temperature sensor

]

   

2

(6)

When the temperature variation is small, the relation (2) is simplified in (7):

Vout ≈ I 0 R0

∆T (α p + α n ) 2

constant current I=1mA through a stabilized external power source, have a stable output voltage greater of 1mV/K in all the range of measure with a good level of linearity, as it is shown in figure 5a). In the diagram of figure 5b) and 5c) the linearity level of the output characteristic sensor by the linear regression, respectively in a range of 100 ÷ 220 °C and 240 ÷ 340 °C is evidenced. Caratteristica SAT @ Ialim=1mA (segnale non amplificato)

(7)

Sensor packaging The device, as can be seen from the design of the complete package of fig. 4, is encapsulated in a stainless steel AISI 304 tube in with external diameter of approximately 3 mm ,the sensitive membrane is placed in head to the cylindrical covering and contact with a sheet in steel that watertight closes the tube (with thermal brazing). Inside the steel tube a ceramic four bore tube is inserted that carries out the twofold function of mechanical support and electrical isolation of the gold conductors that are going through the ceramic tube bores.

500,0 450,0 400,0 350,0 V [mV]

α − δR p = 1 −  n α  p

300,0 250,0 200,0 150,0 100,0 50,0 0,0 0

50

100

150

200

250

300

350

T [°C] Valore mediato

a) Linearity (T = 100-220 °C) 370,00 350,00

y = 1,2375x + 74,131

330,00

Vout [mV]

310,00 290,00 270,00 250,00 230,00 210,00 190,00 90

100

110

120

130

140

150

160

170

180

190

200

210

220

230

T [°C] Vout (06/02/06)

Lineare (Vout (06/02/06))

b) Linearity (T = 240-340 °C) 510,00

490,00

y = 1,1969x + 81,251

470,00

Vout [mV]

450,00

430,00

410,00

390,00

370,00

350,00 230

240

250

260

270

280

290

300

310

320

330

340

350

T [°C]

Figure 4: Some steps of the sensor packaging for high temperature applications Experimental measures Below the sensor output characteristic in the range 17 ÷ 320 °C, obtained as average of five consecutive measures that were carried out in the same conditions in tubular furnace with standard thermal controller and standard reference thermocouple, are presented . The sensor, supplied with

Vout (06/02/06)

Lineare (Vout (06/02/06))

c) Figure 5: Output characteristics of temperature sensor: a) range of 17÷320°C, b) range of 100÷220°C, c) range of 240÷340°C with line of linear regression. Conclusions The prototype of the sensor is manufactured and tested. It can find application in the field of temperature and gas and liquid flow rate measurements and can be used for contact

and non contact measurements, and for the realization of the remote infrared detectors , given the elevated sensibility shown in the experimental tests. The advantages of this device comparing with the sensors currently in commerce, permits to use the sensor in extreme conditions of measurement, guaranteeing its elevated output level without the aid of front-end electronics and, at the same time, an elevated sensibility and stability of output signal in a wide range of measure from the cryogenic temperatures until more than 900°C. References 1. 2.

3. 4.

5.

6.

7.

8.

9.

10.

S. M. Sze, Semiconductor sensors, John Wiley & Sons, Inc. (New York, NY 1994), pp. 331-374; S. M. Sze, VLSI Technology second edition, McGraw-Hill Book Co.(Singapore 1988), pp. 223-267, 375-413; Marc Madou, Fundamentals of Microfabrication, CRC Press LLC, (Boca Raton, Florida 1997), pp. 378-399; Kuntner J., Jakoby B., Beigelbeck R., Kohl F., “Analysis of Spurious Effects in Membrane-Based Micromachined Thermal Conductivity Sensors”, Proc. 46th IEEE Conference on Sensors, Hyatt Regency Irvine, CA, Oct 31 – Nov 3 2005, pp. 11451148. M. von Arx, O. Paul, and H. Baltes, “Process-Dependent Thin-Film Thermal Conductivities for Thermal CMOS MEMS,” Journal of Micromechanical Systems, vol. 9, no. 1, pp. 136–145, March 2003. Z. Fan, J.M. Engel, J. Chen, and C. Liu, “Parylene Surface Micromachined Membranes for Sensor Applications,” Journal of Micromechanical Systems, vol. 13, no. 9, pp. 484-490, June 2004. J. Kuntner, F. Kohl, and B. Jakoby, “Micromachined Thermal Conductivity Sensor for the Liquid Phase,” Proceedings of the 13th International Conference on SolidState Sensors, Actuators and Microsystems, pp. 571-574, Seoul, Korea, June 5-9, 2005. F. Kohl, R. Fasching, F. Keplinger, R. Chabicovsky, A. Jachimowicz, and G. Urban, “Development of miniaturized semiconductor flow sensors,” Measurement, vol. 33, pp. 109–119, March 2003. J. Kuntner, F. Kohl, and B. Jakoby, “Simultaneous Thermal Conductivity and Diffusivity Sensing in Liquids Using a Micromachined Device,” Sensors and Actuators A, accepted for publication. Mancarella F., Roncaglia A., Tamarri F., Pizzochero G., Cardinali G., Severi M., “Fabrication of Pt-polysilicon thin-film thermopiles: a preliminary study”, Proc. 46th IEEE Conference on Sensors, Hyatt Regency

11. 12. 13.

Irvine, CA, Oct 31 – Nov 3 2005, pp. 11411144. www.goodfellow.com www.accumet.com www.coorstek

Session IIA High Energy Physics Detectors

Electronic system trends and challenges in present day particle experiments W. Snoeys, CERN Geneva, CH Abstract Present day particle physics experiments are large systems imposing stringent requirements on electronics in terms of timing, data reduction, radiation tolerance, and speed-power-noise performance. Many modern technologies were successfully adopted to meet all these requirements: custom integrated circuit design in commercial deep submicron CMOS technology, optical data transmission, field programmable gate arrays, etc… The paper will give a system overview through some examples. The emphasis will be on practical experience in the construction of the LHC experiments to try to point out some of the trends and challenges ahead.

1-4244-1245-5/07/$25.00 ©2007 IEEE

Highly Integrated System-On-Chip Circuits for the Readout of High-Energy Physics Detectors L. Musa, CERN Geneva, CH Abstract The very high particle rates and multiplicities that will be produced by the ultra-relativistic hadron collider at LHC, CERN, have set new demands on the readout electronics in terms of resolution, density, speed, complexity and power consumption. These requirements, which are beyond the present capability of commercial-off-the-shelf components, could only be met by highly integrated systems implemented with very deep submicron CMOS technologies. These technologies not only offer speed, density, and computational power, but also radiation tolerance. The potential upgrade of the LHC and other newly planned facilities, e.g. the International Linear Collider (ILC), set even more demanding requirements, which call for ASICs that embed in a single chip the circuits to amplify, digitize, process, compress and store the information of a high number of channels. This paper reviews some examples of highly integrated single-chip circuits developed for the LHC Experiments, and will discuss the role that future silicon processes may play on the on-detector readout and signal processing electronics for future experiments.

1-4244-1245-5/07/$25.00 ©2007 IEEE

Current-Mode Front-End Electronics for Silicon Photo-Multiplier Detectors F. Corsi*, M. Foresta*, C. Marzocca*, G. Matarrese*, A. Del Guerra° *Department of Electronics and Electrical Engineering, Politecnico di Bari and INFN Bari, Italy °Department of Physics, Università di Pisa and INFN Pisa, Italy Contact: [email protected] Abstract Silicon Photo-Multiplier (SiPM) detectors represent an attractive solution for the detection of low energy photons in several fields of both high energy physics and medical imaging. Here we review a recently proposed electrical model for this kind of detectors, which can be conveniently used to perform reliable simulations at circuit level and allows to reproduce accurately the waveform of the signal generated by the SiPM when coupled to the front-end electronics. This is particularly useful in order to choose the most suitable frontend architecture for SiPM detectors. In particular, we propose a front-end architecture based on a current buffer as input stage, featuring small input impedance and large bandwidth due to the use of a current feedback. Moreover the currentmode approach enhances the dynamic range as it does not suffer from possible voltage limitations due to deepsubmicron implementations. Two alternative circuit solutions have been designed and manufactured in a 0.35μm CMOS process. We report the first measurement results obtained by coupling the two prototypes to a SiPM detector excited by a pulsed infrared laser. The measurements allow to validate the functionality of the proposed front-end architecture and demonstrate its capability of managing large current signals with good linearity. 1 Introduction Photo-Multiplier (SiPM) detectors can be considered the most promising candidates to replace vacuum photomultiplier tubes (PMT) in several low-level light detection applications, in both fields of high energy physics and medical imaging [1,2]. High quantum efficiency, high gain, operation at low bias voltages, insensitivity to magnetic fields, excellent time resolution, robustness and compactness are among the well known interesting features of these devices. Much effort is currently devoted to further improve the remarkable performance of these detectors. In particular the technology development is aimed at producing SiPMs with characteristics well suited for different applications, for instance in terms of quantum efficiency at certain wavelengths. Reduction of the dark rate count, optimization of the active sensitive area and mitigation of the optical cross-talk are also very important goals towards the realization of reliable and effective SiPM detectors. As far as the front-end electronics is concerned, in the literature there is a very limited number of circuits specifically designed to read-out a SiPM. The vast majority of these circuits is derived from previous discrete or integrated implementations developed for PMTs [3-5]. In case hundreds of SiPM channels must be read-out, a multi-channel ASIC represents the only viable solution to realize a compact and reliable detection system. Among the other requirements, the 1-4244-1245-5/07/$25.00 ©2007 IEEE

dynamic range and the frequency response of the circuit are the most critical ones, due to the excellent gain and timing characteristics of the SiPM, whereas noise does not represent a major concern. A multi-channel ASIC expressly designed for a SiPM based calorimeter is reported in [4]. The front-end is structured according to the classical CSA+shaper front-end approach, which is able to guarantee the best noise performance but shows serious dynamic range limitations, especially when deep-submicron, low-voltage technologies are employed. Since noise is not the main problem in the SiPM case, due to the remarkable amount of charge generated by an event, different architectures can be possibly adopted, therefore a preliminary study about the most suitable approach for an optimal integrated front-end circuit is mandatory. The availability of an accurate electrical model of the SiPM is of great help for this purpose, since reliable simulations at circuit level of the detector coupled to the front-end electronics can be performed. Thus the main characteristics of the waveform of the achieved signals can be conveniently related to both SiPM parameters and front-end characteristics. In section 2 we review a recently proposed model of the SiPM [6] which is very similar to the classical one [7], but includes parasitic elements which have been estimated to be not negligible at all with simple considerations based on the typical values of technological and geometrical parameters. These parasitics must be added to the model to correctly account for all the time constants which characterize the waveform of the generated signals. Using this model of the SiPM with realistic parameters, the advantages and the limitations of different front-end architectures have been identified, in terms of dynamic range, frequency response, compactness and simplicity, together with a simplified analytical expression of the signal achieved by the SiPM coupled to the front-end electronics, which has been derived for some of the considered front-end structures. Among the possible front-end solutions, the one which exploits a current buffer as input stage has been addressed in this paper. Small input impedance and large bandwidth can be easily achieved, by means of the application of suitable current feedback techniques, resulting into insensitivity of the detector bias voltage with respect to the signal amplitude and fast timing. Moreover, in case a standard CMOS deepsubmicron technology is used to implement the circuit, the low value of the available supply voltage does not represent a serious limitation to the signal dynamic range, thanks to the current-mode approach. Two alternative circuit solutions for the current buffer have been designed and manufactured in a 0.35μm CMOS process. The first solution is characterized by a larger

bandwidth, which can be adjusted by setting a reference voltage, while the second one exhibits a larger dynamic range and allows to vary the SiPM bias voltage (and thus its gain) by tuning the value of the input voltage. In section 3 an experimental comparison of the two circuits is given in terms of linearity, dynamic range and resolution. Measurements obtained by coupling the two circuit prototypes to a real SiPM detector excited by a pulsed infrared laser are also reported and demonstrate the capability of the circuits to process the large amount of charge delivered by the detector in response to an event. 2 SiPM model and choice of the front-end electronics The SiPM structure is composed by the parallel connection of hundreds of micro-cells, each consisting of a Geiger-mode operated photodiode passively quenched by a large series resistor [1,2]. The model of the single micro-cell, besides the diode capacitance Cd and the quenching resistor Rq, contains also a small parasitic capacitor in parallel to Rq, which works as a fast path for the charge delivered during the avalanche [8]. A metal routing is used to connect in parallel hundreds of these micro-cells sharing the same substrate. A further parasitic capacitance Cg between the terminal of the whole device is introduced, due to the presence of this metal grid which spans over the entire surface of the SiPM. As an example, if a SiPM of 1mm2 area is considered, assuming that the metal routing grid covers a reasonable 35% of the total surface, a value of about 11pF can be estimated for the capacitance Cg, considering only the contribution due to a typical metal-to-substrate capacitance per unit area of 0.03 fF/μm2. A greater value can be envisaged for Cg, due to the fringe capacitance of the metal lines, the bonding pad etc. These considerations suggest that an accurate electrical model of the SiPM, able to reproduce its behavior when an event triggers the avalanche in one or more micro-cells, must include this capacitance Cg. Fig. 1 represents the linearized, small-signal equivalent circuit of the whole device, in case only one out of the total N micro-cells is interested by a Geiger discharge, as happens, for instance, when a single dark count event is generated.

is the charge delivered by a single fired micro-cell and ΔV is the applied overvoltage, i.e. the difference between the bias voltage Vbias and the breakdown voltage Vbr. This assumption holds true as long as all the time constants introduced by the circuit are much larger than the ones associated to the avalanche phenomenon, which is a realistic hypothesis. Thanks to the superposition principle, the same circuit of fig. 1 can be also used to model the case in which more than one micro-cell is interested by an avalanche event, simply considering IAV formed by more Dirac’s delta pulses, distributed in time accordingly to the arrival of the events. The charge delivered by a micro-cell after a Geiger discharge can be easily measured considering single dark count pulses read-out by means of a simple front-end channel with known gain. For instance the current provided by the SiPM can be converted into a voltage by means of a linear resistor and the voltage obtained can be amplified by a voltage amplifier, assembled with discrete components. We performed dark pulse charge measurements with this approach, while varying the bias voltage Vbias applied to the SiPM, and the results are shown in fig. 2 for a SiPM manufactured by ITC-irst. According to eq. (1), the slope of the obtained curve relating Q to Vbias provides the total micro-cell capacitance Cd+Cq, The complete characterization procedure needed to evaluate all the parameter values of the model in fig. 1, which fit a real device, has already been presented in [6] and is beyond the scope of this paper. Table I summarizes the results of the extraction procedure applied to two SiPM detectors produced by different manufacturers.

Figure 2: Charge associated with a single dark current pulse as a function of the bias voltage for the ITC-irst SiPM.

Figure 1: Equivalent circuit of the SiPM, including the grid parasitic capacitance Cg. In this model the waveform of the current source IAV can be considered a Dirac’s delta pulse Qδ(t), where: Q=ΔV(Cd+Cq)

(1)

The availability of a realistic model for the detector and the knowledge of the values of the parameters involved can help the designer in the choice of a suitable read-out architecture. Among the different front-end architectures which can be considered to read-out a SiPM, the charge sensitive amplifier (CSA), in which the charge delivered by an event is collected onto a feedback capacitor as in fig. 3, is able to guarantee the best noise performance. However, in case several front-end channels must be integrated in a single chip to read-out an array of SiPMs, the CSA dynamic range can represent a

serious limitation. For instance, if the maximum allowed voltage swing, due to supply voltage limitations, is ΔV=3V, the feedback capacitance required to collect a charge Qtot=50pC, corresponding to the signal delivered by about 300 micro-cells, should be at least Cf=Qtot/ΔV≅16.7pF, which is a quite large value for an integrated capacitor. For deepsubmicron technologies and increasing amounts of charge the Cf required would be even larger, thus impractical. Moreover, the output stage of the amplifier used to implement the CSA must be able to drive the large capacitive load given by the series of the feedback capacitance and the terminal capacitance of the SiPM. To avoid stability issues and to achieve the speed constraints required in many applications, the output stage must be biased with a large current, thus increasing also the power consumption.

converted into a voltage by means of the resistor RS and a voltage amplifier provides the desired output signal level. In this approach, the amplifier input is not a virtual ground, thus the value of RS must be small enough to avoid variations of the SiPM bias under the signal. In several applications the information associated to the total charge released by the detector is needed, thus the output voltage VOUT must be integrated. This involves a further voltage to current conversion. To avoid multiple conversions between current and voltage signals, the current signal of the detector can be directly read by means of a current buffer with low input impedance. This current buffer provides an output current which is a scaled replica of the detector signal at high impedance, according to the principle schematic in fig. 5. Vbias

Table I. Results of the parameter extraction procedure applied to two SiPM from different manufacturers. SiPM ITC-irst N=625, Vbias=35V

SiPM Photonique N=516, Vbias=63V

Rq

393 kΩ

774 kΩ

Vbr

31.2 V

61 V

Q

175.5 fC

127.1 fC

Cd

34.6 fF

40.8 fF

Cq

12.2 fF

21.2 fF

Cg

27.8 pF

18.1 pF

Model parameter

Vbias

CF

SiPM

+

VOUT

Figure 3: Typical configuration of a charge sensitive amplifier.

SiPM

RS

IS

kIS=IOUT

Figure 5: SiPM read-out by means of a current buffer. This output current can be easily replicated, by means of current mirrors, and sent to a simple integration stage, to extract the charge information, and/or to a current discriminator, to obtain a fast timing signal. The circuit is inherently fast and the current mode of operation enhances the dynamic range, since it does not suffer from possible voltage limitations due to deep-submicron implementation. When the SiPM is coupled to a voltage or a current amplifier as in fig. 4 or 5, the model previously described can be used to find the signal waveform when a single micro-cell of the detector undergoes a Geiger discharge. If the SiPM model represented in fig. 1 is loaded with the resistor RS, a qualitative study of the resulting circuit can be carried out with reference to a simplified schematic depicted in fig. 6, which is a good approximation of the real circuit, provided that RS is much lower than Rqtot.

Vbias SiPM

RS

+ -

VOUT

Figure 4: SiPM read-out by means of a resistor Rs and a voltage amplifier. Since, as already mentioned, noise does not represent a major issue, these area, dynamic range and power dissipation drawbacks can be overcome resorting to the architecture outlined in fig. 4, in which the current signal of the SiPM is

Figure 6: SiPM coupled to the input impedance of the frontend electronics Rs: simplified circuit valid if RS108 Ω·cm were grown by the THM method and supplied by Acrorad, Ltd. (Japan). They were (111)-oriented, 1×1 cm2 large and 1 mm thick. Systematic inspection of CdTe substrates evidenced a very rough surface and the occurrence of arrays of large surface scratches, their density varying from wafer to wafer and from point to point within the same wafer; this makes the asreceived material unsuited for epitaxy. Furthermore, as CdTe is mechanically weaker than most compound semiconductors, the usual cutting/lapping procedures of the material slabs from the ingot result in lattice damages, distortion and/or stresses within a nearby-surface region of the crystal that need to be removed before epitaxy. To this purpose the crystals were etched in Br2-methanol before MOVPE growth. The wafers were first degreased in boiling acetone and isopropanol for several minutes, cleaned in isopropanol vapours for 1 h and dried under pure N2; after that they were etched at RT in a 2% Br2-methanol solution, thoroughly rinsed in methanol and dried under N2. It was found that a 1 min etch, corresponding to a removal of a ~18-µm thick surface layer, allowed to eliminate most scratches and smooth down the crystal surface, achieving a rms roughness around 1.3 nm [6]. Immediately before growth the substrates were insitu (i.e. inside the reactor) annealed under 1.0 sl/min pure H2 flow for 10 min at 350°C, as this temperature proved sufficient to remove any trace of residual oxides from the crystal surface [7] and to further reduce its roughness [6].

mTorr and an RF power of 200 W were used [8]. These conditions allowed us to remove ~1.25 µm of the CdTe:I layer, thus obtaining a mesa structure around the Al contact. Fig. 1 reports a schematic of the final Pt/i-CdTe/n-CdTe:I/Al structure. The electrical properties of the as-fabricated device structures were studied by RT current-voltage (I-V) measurements performed by a 237 Keithley SMU in a K. Suss probe station.

Figure 1: Schematic of the fabricated Al/n-CdTe:I/iCdTe/Pt device structure. The n-CdTe:I layer below the Al electrode is 2 µm thick. 3. Results and discussion 3.1 − Properties of the n-CdTe:I contact layers (111) (333)

Intensity (A.U.)

CdTe layers were grown at 330°C using dimethylcadmium (Me2Cd), di-isopropyltelluride (iPr2Te) and ethyl-iodide (EtI), all supplied by Epichem Ltd. (UK), as Cd, Te and I precursors, respectively. For all growth experiments the Me2Cd molar flow rate was kept fixed at 55 µmol/min, while the iPr2Te molar flow was varied between 18.3 µmol/min and 27.5 µmol/min, corresponding to a Te:Cd relative concentration in the vapour between 0.33 and 0.50, respectively; for the growth of CdTe:I layers the molar flow rate of EtI was kept fixed at 2.2 µmol/min for all runs. The surface morphology of CdTe layers was investigated by field emission gun scanning electron microscopy (FEGSEM) using a JEOL model JSM 6500 F with an electron beam energy of 5 kV. X-ray diffraction (XRD) measurements were performed to check the degree of epitaxy of the material by using a highresolution X-ray diffractometer (HRD3000 Ital Structures) in parallel beam optic configuration (Max-FluxTM Optical System). A fine focus X-ray tube equipped with a copper target (λCu = 0.154056 nm) was employed as X-ray source. The electrical properties of as-grown CdTe:I samples were determined by resistivity and Hall effect measurements carried out at RT in the Van der Pauw configuration on relatively thick (4 µm) CdTe epilayers. To this purpose, Au and Al contacts were e-beam evaporated on undoped and Idoped layers, respectively. I atoms incorporation in the layers was also studied by low temperature photoluminescence (PL) measurements performed on samples mounted on the cold finger of a closedcycle He cryostat. The material was excited by the 532 nm line of a solid state laser at a power density of ~3 W/cm2. The detection system consisted of a 0.55 m monochromator equipped with a cooled GaAs photomultiplier tube and a lockin amplifier. As-grown n-CdTe:I/i-CdTe samples were used to fabricate a M-i-n device structure, a preliminary technological step towards the final p-i-n diode. To this purpose the CdTe:I layer was covered with wax and the sample lightly etched in a 0.2% Br2-methanol solution to remove any trace of material unintentionally deposited on either the substrate edges or its back surface; this procedure was adopted to avoid the occurrence of short-circuit paths in the device. The wax was afterwards removed by thorough rinsing of the sample in acetone and metal thin films were deposited on both its sides as electrical contacts. In particular, a 200-nm thick and 0.8x0.8 cm2 wide Al electrode was sputter deposited on the nCdTe layer, while a 200-nm thick Pt contact was e-beam evaporated on the entire back face (substrate side) of the nCdTe:I/i-CdTe structure. To improve the electrical insulation between the back and front electrodes and reduce any possible surface leakage path we removed part of the CdTe:I layer around the Al contact area by performing a reactive ion etching (RIE) treatment on the sample. Noteworthy, the front contact acts as a mask since Al is not affected by the dry-etch process. The dry-etch was performed using a H2/Ar gas mixture, with a 17 scm3/min H2 flow and a 8 scm3/min Ar flow; a chamber pressure of 200

10

(a)

(222)

(b)

20

30

40 50 60 70 Angle 2θ (degrees)

80

90

100

Figure 2: X-ray diffraction spectra (Bragg geometry) recorded for (a) a 2-µm thick CdTe layer grown under a precursor Te:Cd molar flow ratio in the vapour equal to 0.50, and (b) a bare (i.e., not deposited) (111)B-oriented CdTe substrate. The combination of substrate treatments (as described in Sec. 2) and the use of Cd-rich vapour conditions during MOVPE allows to grow CdTe layers having fairly good surface morphology and crystalline quality [6].

Fig. 2a shows a typical XRD pattern of a 2-µm thick undoped CdTe layer sample grown under a Te:Cd molar flow ratio in the vapour of 0.50; as a comparison, Fig. 2b shows the XRD pattern of a bare (i.e., not deposited) substrate. While the substrate XRD pattern exhibits only the relatively narrow (111), (222) – very weak – and (333) diffraction peaks of CdTe, as expected for a (111)-oriented single crystal, no extra peaks beyond those related to the (111) planes appear in the diffraction spectrum of the deposited sample, demonstrating that a fully epitaxial relationship holds between overgrown layer and substrate. The surface morphology of the epitaxial layer is reported in Fig. 3 and shows a peculiar surface texture (on the submicron scale length) attributed to the material inherent growth mode under present MOVPE conditions.

Tab. I reports the variation of the I-doped layer electrical properties with growth conditions: the material resistivity steadily decreases with Te:Cd vapour stoichiometry during growth, while electron concentrations remain at around 1016 cm-3 for values of the precursor molar flow ratio in the 0.33−0.50 interval. However, decreasing the Te:Cd ratio in the vapour seems to favour the growth of a less electrically compensated material, as it appears by the larger Hall mobility value found for the samples grown under the lowest Te:Cd value in Tab. I. Fig. 4 shows the 7K PL spectra of both a detector-grade CdTe crystal and a 3-µm thick CdTe:I layer (Te:Cd molar flow ratio equal to 0.33). The PL spectrum of the bulk sample (Fig. 4a) is dominated by the near band-edge emission: a narrow and intense line at 1.591 eV followed by a weaker line at 1.593 eV, both ascribable to a neutral donor bound exciton recombination (D0,X).

Figure 3: Plan-view FEG-SEM micrograph (15,000× magnification) of 2-µm thick CdTe layer (same sample as in Fig.2a). Addition of I dopant during the MOVPE process does not change either the material growth rate nor its surface morphology, while the CdTe:I layers retain a (111)-oriented texture. These layers turn out to be n-type with a resistivity of the order of a few Ω·cm, indicating a good electrical activation of I donors. Tab. I: Values of RT resistivity (ρ), electron concentration (n) and Hall mobility (µH) of CdTe:I layers grown on detector grade CdTe crystals under different values of precursors vapour stoichiometry (Te:Cd). h represents the layer thickness.

Sample Te:Cd CT37a CT38 CT62 a

1.0 0.50 0.33

ρ

µH

h (µm)

(Ω⋅cm)

n (cm-3)

(cm2/Vs)

4 4 3

34 6.9 1.3

− 4.8×1016 1.4×1016

− 18.6 359

Substrate annealed in H2+Me2Cd [ref. 6]

Figure 4: PL spectra recorded at 7K from (a) a detectorgrade CdTe crystal, and (b) a 3-µm thick CdTe:I layer grown under a Te:Cd molar flow ratio in the vapour equal to 0.33 (same sample as in Tab. I). Also a weak shoulder on the high energy side of the latter bound exciton line appears at 1.595 eV, ascribable to free exciton (FX) recombination. Longitudinal optical (LO) phonon replicas (ħωLO ~21 meV) of both free and bound exciton lines appear between 1.570 eV and 1.576 eV. A broad and weak band due to a donor-acceptor pair (DAP) recombination appears in the 1.37−1.50 eV interval, its zerophonon DAP emission being detected at ~1.475 eV followed by its LO-phonon replicas at lower energy. For comparison, the PL spectrum of the I-doped layer (Fig. 4b) shows only a broad and intense emission band at around 1.473 eV; no exciton-related features appear in the near band-edge region of the spectrum, as expected for a heavily doped material due to the coulomb screening effect.

The broad band can be ascribed to a DAP recombination involving a shallow I donor on a substitutional Te site (ITe) and the so-called A-centre acceptor complex, the latter formed by pairing the I donor with a nearest neighbour Cd vacancy (VCd-ITe) [9]. The latter attribution may well explain the somewhat lower electrical compensation of I dopant observed for CdTe:I layers grown under very Cd-rich conditions (Tab. I), as a consequence of the material lower proclivity to form A-centres for reduced concentration of Cd vacancies. 3.2 − Electrical characteristics of the M-i-n device The I-V characteristics of the Pt/i-CdTe/n-CdTe:I/Al device structure have been measured before and after RIE treatment. This treatment turned out to be effective in reducing the current flowing through the device by more than one order of magnitude. -3

10

V>0 on Al contact V 0.6, above 4.2 K up to the niobium transition, the junctions are fully non-hysteretic, and the behaviour is SNS like

It should be noted that a transition from the hysteretic to the non-hysteretic regime is induced in these junctions in two ways. For a given set of niobium and aluminum film thickness, a reduction of the oxidation exposure of the AlOx layer causes the shunting of the capacitance and an overdamped response at temperatures of 4.2 K and above. On the other hand, this junction becomes hysteretic at low temperatures. Applications: voltage standard and innovative devices Considering the application to measurement, an interesting feature is the increment of both Jc and Vc when temperature is approaching the transition of the metallic aluminum film, up to twice the values obtained at 4.2 K, providing Jc values has high as 100 kA/cm2 and Vc near 1 mV around 2 K. Such high values look promising for the possible realization of structures for quantum computation. In fact two different stable states corresponding to the two states of the qubit, related to the 0 or p phase of the supercurrent flowing in the junctions, can happen in junctions made of superconductor and ferromagnetic films [11]. Since however the presence of the ferromagnetic layer reduces strongly the current density of these junctions compared to the equivalent SIS structure, the possibility of using our overdamped SIS, which at the temperatures where the transition between the two phase states happens are hysteretic, can be useful, since they feature Jc values more than 10 times the SIS junction. Concerning the application to voltage standard metrology, even more interesting appears the fact that these junctions show at temperatures above liquid helium Jc and Vc values still compatible with an employ in voltage standard circuits: a feature suitable for the realization of compact and less expensive measurement systems involving cryocooler setup. As a matter of fact, a major problem when thinking of using series arrays circuits for programmable voltage standards or for AC synthesis is the necessity of both the maximization of the quantized step amplitude and minimization of the power dissipated in the helium bath. The analysis carried out by Kautz in [12] states that these conditions are achieved when the characteristic junction frequency of the device, related to the characteristic voltage by the Josephson relationship Fc = 2e/h Vc , has about the same value as the RF drive frequency F of the signal radiating the array. F/Fc = 1. For F/Fc < 1 the step amplitudes is reduced, while for F/Fc > 1 there is a considerable increment in the dissipated RF power. Since, as mentioned above, the optimal drive frequency lies in the K band, about 75 Ghz, this implies to use junctions with characteristic voltage about 150 mV. Therefore, while the majority of SNS junctions have too small values of Vc even at 4.2 K and the SINIS on the other hand show too small quantized step amplitude to work above

4.2 K, our Nb/Al-AlOx/Nb overdamped junctions having high values of both these parameters, allow us to work at temperatures approaching their transition temperature. First experiments carried out on junctions with a characteristic voltage of 250- 400 µV at 4.2 K, have shown [13] that ample and stable voltage stesp can be obtained up to a temperature corresponding to 0.85 T/Tc. In figure 3 the profile of the n=1 step measured with submicrovolt accuracy at 5.4 K is compared to the same step at 4.2 K. Despite the reduction, due to the decrement of Ic, the step is about 1.5 mA wide at the higher temperature.

References

1. C. A. Hamilton, “Josephson voltage standards,” Rev. Sci. Instrum., vol. 71, pp. 3611-3623, Oct. 2000. 2. EUROMET iMERA Project, Technical Committee Electricity and Magnetism, Quantum Standards Roadmap. 3. J. Kohlmann, F. Muller, R. Behr, D. Hagedorn, J. Niemeyer, “SINIS junction series arrays for the Josephson arbitrary waveform synthesizer”, IEEE Trans. on Applied Supercond., Vol. 15, no. 2, pp. 121-124, June 2005. 4. A. Sosso, V. Lacquaniti, D. Andreone, R. Cerri, A. Klushin, “Study and operating conditions of HTS Josephson arrays for metrological application”, Physica C: Superconductivity, Volume 435, Issue 1-2, Pages 125-127. 5. A. Shoji, H. Yamamori, H. Sasaki, and M. Ishizaki, “NbN-based digital to analog converters for a programmable jpsephson voltage standard”, Proc 2006 CPEM Conf, June 2006, pp. 444-445. 6. V. Lacquaniti, C. Cagliero, S. Maggi, R. Steni, “Overdamped Nb/Al-AlOx/Nb Josephson junctions,” Applied Physics Letters, n. 86, 042501, 2005. 7. R. Behr, J. M. Williams, P. Patel, T. J. B. M. Janssen, T. Funck, and M. Klonz, “Synthesis of precision waveforms using a SINIS Josephson junction array,” IEEE Trans. Instrum. Meas., vol. 54, no. 2, pp. 612–615, Apr. 2005. 8. Y. Chong, P. D. Dresselhaus, and S. P. Benz, “Electrical properties of Nb-MoSi -Nb Josephson junctions,” Appl. Phys. Lett., vol. 86, pp. 232505-1–232505-3, Jun. 2005. 9. Burm Baek, Paul D. Dresselhaus, and Samuel P. Benz, “Co-Sputtered Amorphous NbxSi1-x Barriers for Josephson-Junction Circuits”, IEEE Trans. on Applied Supercond., Vol. 16, n. 4, Dec. 2006 10. V. Lacquaniti, D. Andreone, N. De Leo, M. Fretto, S. Maggi, A. Sosso, M. Belogolovskii, “Analysis of the temperature stability of overdamped Nb/Al-AlOx/Nb Josephson junctions”, to appear in IEEE Trans. on

Comparison of the amplitude of the step n = 1 (153 at 4.2 K and 5.4 K for a single junction radiated at 75 GHz, step voltages (x axis) are slightly shifted for clarity. It must be observed that, due to the almost similar thickness of the two films, the transition temperature of the Nb/Al bilayer is reduced respect to the bulk value of 9.1 K. These results indicate also a direction for the optimization of the fabrication process of these junctions This implies to use electrodes with a thicker niobium and a thinner aluminum since, to increase the working temperature we must enhance the transition temperature of the junction, as near as possible to the bulk value. Of course this must be realized maintaining the main aspect Applied Supercond. for the realization of these junctions, that is leading to the 11. M. Weides, M. Kemmler, E. Goldobin, D. Koelle, R. shunted electrical characteristic previously outlined. Kleiner, H. Kohlstedt, and A. Buzdin, “High quality Therefore the oxide barrier must be thinned, avoiding ferromagnetic 0 and π Josephson tunnel junctions”, Appl. microshorts. Phys. Lett., 89, 122511 (2006). Conclusions 12. R.L. Kautz, “Shapiro steps in large-area metallic-barrier We have reported on some properties of Nb/Al-AlOx/Nb Josephson junctions”, Journal of Applied Physics, Vol. overdamped junctions realized at INRIM which can be useful 78, Issue 9, pp. 5811-5819, 1995. for their employ in circuits for precision measurement. 13. Lacquaniti et al., in print In particular the high values of critical current density and characteristic voltage, together with their temperature dependence look promising for the realization of improved devices for programmable and AC voltage standard, where a simple fabrication process, a moderate number of junctions and the possible use of cryocoolers with reduced power dissipation can open the employ of these standards to a more widespread public usage other than the primary metrological institute. Figure 3:

µV)