Handheld Total Chemical and Biological Analysis Systems: Bridging NMR, Digital Microfluidics, and Semiconductors

Ka-Meng Lei · Pui-In MakMan-Kay Law · Rui Paulo Martins

Handheld Total Chemical and Biological Analysis SystemsBridging NMR, Digital Micro� uidics, and Semiconductors

Handheld Total Chemical and Biological Analysis Systems

Ka-Meng Lei • Pui-In Mak Man-Kay Law • Rui Paulo Martins

Handheld Total Chemical and Biological Analysis SystemsBridging NMR, Digital Microfluidics, and Semiconductors

ISBN 978-3-319-67824-5 ISBN 978-3-319-67825-2 (eBook)DOI 10.1007/978-3-319-67825-2

Library of Congress Control Number: 2017952196

© Springer International Publishing AG 2018This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer NatureThe registered company is Springer International Publishing AGThe registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Ka-Meng LeiState-Key Laboratory of Analog

and Mixed-Signal VLSIUniversity of MacauMacau, China

Man-Kay LawState-Key Laboratory of Analog

and Mixed-Signal VLSIUniversity of MacauMacau, China

Pui-In MakState-Key Laboratory of Analog

and Mixed-Signal VLSI and FST-ECEUniversity of MacauMacau, China

Rui Paulo MartinsState-Key Laboratory of Analog and

Mixed-Signal VLSI and FST-ECEUniversity of MacauMacau, China

Instituto Superior Técnico Universidade de Lisboa

Lisbon, Portugal

v

Preface

This book investigates the handheld total chemical and biological analysis system implemented with complementary metal–oxide–semiconductor (CMOS) based on nuclear magnetic resonance (NMR) technique. The global market for in vitro diag-nosis is expanding in both developed and developing countries ascribed to the grow-ing population and longer life expectancy. Conventional benchtop tools for disease diagnosis such as PCR (DNA amplification) are costly, bulky, and time-consuming and require trained technicians for operation, which confound their usages in the centralized laboratory.

CMOS is a promising alternative solution for rapid and quantitative diagnosis at a low cost. It overcomes the miniaturization of healthcare diagnostic tools, allowing low-cost and rapid detection of specific targets in tiny fluid samples. Among numer-ous possible solutions to POC sensing mechanism, NMR stands out as a trailblazing option since it is versatile and low-cost as it requires little processing on both the samples and interfacing hardware, i.e., the transducers. However, the reported NMR systems in literature encounter some issues such as bulky hardware, sample man-agements, and magnetic field shifting. So herein the materials presented in this book are focused on optimizing CMOS NMR platform for enhancing their applicability by bridging NMR, semiconductor chips, and microfluidic technique and promoting the application of NMR outside standard centralized laboratory with the aid of CMOS chips. The proposed miniaturized NMR systems in this project achieve (1) accurate and sensitive chemical/biological detection from microliter samples by the CMOS integrated circuits; (2) electronic-automated sample management scheme inside the space-limiting portable magnet, which significantly reduces the labors and turnaround time of the assay; and (3) robust operation against environmental

vi

variation such as temperature or displacement of the sample. The platforms show promise as robust and portable diagnostic devices for a wide variety of biological analyses and screening applications.

We hope the readers will enjoy the contents of this book.

Macao, China Ka-Meng Lei Pui-In Mak Man-Kay Law Rui Paulo Martins

July 2017

Preface

vii

1 Introduction ............................................................................................ 1 1.1 Overview ......................................................................................... 1 1.2 Global Necessities for In Vitro Diagnostic Tools ........................... 2 1.3 Nuclear Magnetic Resonance for In Vitro Diagnosis ..................... 4 1.4 Organization .................................................................................... 6References ................................................................................................ 7

2 State-of-the-Art CMOS In Vitro Diagnostic Devices .......................... 11 2.1 Introduction ..................................................................................... 11 2.2 Transducing Mechanisms of CMOS IVD Tools ............................. 11

2.2.1 Electrical-Based .................................................................. 12 2.2.2 Optical-Based ...................................................................... 18 2.2.3 Magnetic-Based .................................................................. 20 2.2.4 Mechanical-Based ............................................................... 21 2.2.5 NMR-Based ........................................................................ 25

2.3 In Vitro Diagnostic Applications ..................................................... 26 2.3.1 Immunoassay ...................................................................... 26 2.3.2 DNA Hybridization Assay .................................................. 28 2.3.3 Cell/Bacteria Diagnosis ...................................................... 29

2.4 Discussions and Selection Guide .................................................... 32 2.4.1 Integration Level ................................................................. 32 2.4.2 Labeling ............................................................................. 32 2.4.3 Hardware Preparation ......................................................... 33 2.4.4 Operation............................................................................. 33 2.4.5 Specificity ........................................................................... 34 2.4.6 Summary ............................................................................. 34

References ................................................................................................ 36

Contents

viii

3 Electronic-Automated Micro-NMR Assay with DMF Device ............ 41 3.1 Introduction ..................................................................................... 41 3.2 First Prototype: Primary Investigation on NMR–DMF .................. 42

3.2.1 Discrete Electronics and Back-End Signal Processing ....... 43 3.2.2 Magnet .............................................................................. 44 3.2.3 RF Coils ............................................................................. 44 3.2.4 DMF Device Fabrication and Actuation ............................. 46 3.2.5 Experimental Results .......................................................... 47

3.3 Second Prototype: CMOS Micro-NMR Platform with DMF ......... 51 3.3.1 Design and Implementation of CMOS TRX ...................... 53 3.3.2 Portable Magnet and RF Coil Codesign ............................. 58 3.3.3 DMF Device and Its Control Circuit .................................. 60 3.3.4 Experimental Results .......................................................... 60 3.3.5 Discussion and Outlook ...................................................... 67

3.4 Summary ......................................................................................... 68References ................................................................................................ 69

4 One-Chip Micro-NMR Platform with B0-Field Stabilization ............ 73 4.1 Introduction ..................................................................................... 73 4.2 Platform Design .............................................................................. 74

4.2.1 Micro-NMR Transceiver ..................................................... 75 4.2.2 Multifunctional Planar Coil ................................................ 76 4.2.3 Hall Sensor, Readout Circuit, and Current Driver .............. 76

4.3 Prototype and Experimental Results ............................................... 82 4.3.1 Experimental Setup and Electrical Measurements ............. 82 4.3.2 Biological/Chemical Measurements ................................... 84 4.3.3 Comparison and Discussion ................................................ 86

4.4 Summary ......................................................................................... 88References ................................................................................................ 89

5 Conclusion and Outlook ........................................................................ 91 5.1 Summary of Researches .................................................................. 91 5.2 Future Prospects .............................................................................. 92References ................................................................................................ 92

Appendix A: Modular NMR Electronic Components and Measurement .................................................................................................. 95

Appendix B: DMF Device and Electronics .................................................. 97

Appendix C: Software and Hardware Interface of Micro-NMR Platform .................................................................................... 99

Index ................................................................................................................ 101

Contents

ix

1/f noise Flicker noiseAC Alternating currentADC Analog-to-digital converterAIDS Acquired immune deficiency syndromeB0-field Static magnetic fieldB1-field Radio-frequency magnetic fieldBJT Bipolar junction transistorBW BandwidthCMOS Complementary metal–oxide–semiconductorCPMG Carr–Purcell–Meiboom–GillDAC Digital-to-analog converterDC Direct currentDMF Digital microfluidicDNA Deoxyribonucleic acidEC Eddy currentEIS Electrochemical impedance spectroscopyELISA Enzyme-linked immunosorbent assayEWOD Electrowetting-on-dielectricFoM Figure of meritFPGA Field-programmable gate arrayGBW Gain–bandwidth producthCG Human chorionic gonadotropinHIV Human immunodeficiency virushMAM Human mammaglobinIC Integrated circuitIDT Interdigital transducerIF Intermediate frequencyIgG Immunoglobulin GIgY Immunoglobulin YIIP3 Third-order intercept pointIRN Input-referred noise

List of Abbreviations

x

ISFET Ion-sensitive field-effect transistorITO Indium tin oxideIVD In vitro diagnosticLNA Low-noise amplifierLO Local oscillatorLOC Lab-on-a-chipLoD Limit of detectionLPF Low-pass filterLSB Least significant bitMNP Magnetic nanoparticleMOSFET Metal–oxide–semiconductor field-effect transistorMP Magnetic particleMUX MultiplexerNMOS N-channel MOSFETNMR Nuclear magnetic resonanceNW NanowirePA Power amplifierPBS Phosphate-buffered salinePC Personal computerPCB Printed circuit boardPLL Phase-locked loopPM Phase marginPMOS P-channel MOSFETPNIPAM Poly(N-isopropylacrylamide)PoC Point of carePoU Point of usePSS Pulse sequence synthesizerqPCR Quantitative polymerase chain reactionRBC Red blood cellRF Radio-frequencyRX ReceiverSAL Supercritical angle luminescenceSAW Surface acoustic waveSNR Signal-to-noise ratioSPAD Single-photon avalanche diodeTAT Turnaround timeTHD Total harmonic distortionTIA Transimpedance amplifierTRX TransceiverTX TransmitterUART Universal asynchronous receiver/transmitterVHS Vertical Hall sensorWHO World Health OrganizationXO Crystal oscillatorβ-LG β-Lactoglobulin

List of Abbreviations

xi

Fig. 1.1 World population from 1950 to 2050, with a medium variant estimation from 2015. Data collected from the United Nations World Population Prospects: The 2015 Revision [14]. More developed countries: countries in Europe and Northern America, plus Australia/New Zealand and Japan. Less developed countries: countries in Africa, Asia (except Japan), Latin America, and the Caribbean plus Melanesia, Micronesia, and Polynesia ................ 2

Fig. 1.2 The old age dependency ratio (solid line), which is defined as the ratio of population of 65+ years old to the population of 15–64 years old with medium variant estimation from 2015. The children dependency ratio (dotted line), which is defined as the ratio of population of 0–14 years old to the population of 15–64 years old, is also shown on the graph as reference. Data collected from the United Nations World Population Prospects: The 2015 Revision [14]. More developed countries: countries in Europe and Northern America plus Australia/New Zealand and Japan. Less developed countries: countries in Africa, Asia (except Japan), Latin America, and the Caribbean plus Melanesia, Micronesia, and Polynesia ................ 3

Fig. 1.3 (a) Macroscopic view of the non-zero spin nuclei. With an external magnetic field B0 applied to the nuclei, part of them will align with this magnetic field. (b) The effect of RF excitation on the nucleus under external magnetization. When excited by the RF magnetic field at fL, the nuclei precess around the magnetization. After this excitation, the nuclei still resonate and return to the equilibrium, with this relaxation recorded and analyzed ................................................................................. 4

Fig. 1.4 The state of the probe-functionalized MNPs. (a) Without the target, the MNPs stay monodisperse in the solution without any aggregation. (b) When the targets exist in the sample, the targets bind with the probe, and the MNPs aggregate to form micro-clusters ................................................................... 6

List of Figures

xii

Fig. 2.1 Architecture and operation of electrical-based detection CMOS biosensor. An extra layer of noble and biocompatible metal such as gold is deposited on the original built-in metal layer. The capturing probe is then immobilized on the gold electrode to capture the target. Upon hybridization the electrical properties such as impedance or charge are sensed directly by the readout circuit ............................................................................... 16

Fig. 2.2 Cell culturing and monitoring with CMOS capacitive sensing chip. (a) The photograph showing the overall chip with dual in-line package. A well encloses the cell culturing site, and the CMOS chip is at the center of the well. The polymer protects the bond wires of the chip. (b) Photomicrograph of the electrodes. Since the system measures only the capacitance of the single electrode, the built-in passivation layer such as silicon nitride and silicon dioxide can be preserved without further post-processing. This simplifies the hardware preparation steps for biosensing. (c) The experimental results for cancer cell MDA-MB-231 culturing. The capacitance at specific site increases due to the proliferation of the cancer cells ascribed to the increased number of cells, allowing real-time monitoring for the growth of the cancer cells (Reproduced with permission from [38]. Copyright 2008 Elsevier) .............................................................. 17

Fig. 2.3 Architecture and operation of optical-based detection CMOS biosensor. The capturing probe is immobilized on a solid substrate such as glass or the built-in passivation layer atop the CMOS chip. Then fluorescence-labeled or chemiluminescence-labeled target will bind with the probe, and other unbound biomolecules will be washed away. The CMOS photodetector, which is formed by the embedded PN-junction, transduces the optical signal to current for subsequent signal processing ........................................................ 18

Fig. 2.4 Lens-free cell/microparticle counting system with CMOS image sensor. (a) The overall platform of the digital cell counting device. (b) The micrograph of the microcavity array for cell trapping. The sample under analysis is put atop the microcavity array. Then the suspended cells/microparticles will be pulled toward and trapped in the cavities attributed to the negative pressure. This negative pressure is produced by peristaltic pump, which extracts the air inside the chamber. (c) Detection principle of the system. The light from the external UV light source will arrive at the CMOS image sensor through the unoccupied cavity, while the trapped cell on the cavity blocks the light from arriving at the CMOS image sensor. (d) The schematics of the expected CMOS image acquired from (c). Since the cell blocks the UV light from passing through the cavity, the pixels under those occupied cavity will report a darker region, while the pixels under the vacant

List of Figures

xiii

cavity will report a brighter result. Thus the number of cells on the microcavity array can be identified from the result of the CMOS image sensor (Reproduced with permission from [43]. Copyright 2014 Saeki et al.) ......................................................... 20

Fig. 2.5 Architecture and operation of magnetic-based detection CMOS biosensor. The capturing probe is immobilized on a solid substrate such as glass or the built-in passivation layer atop the CMOS chip. Then the sample labeled with MP will mix with the capturing probe. Matched target will be captured, and unbound objects will then be rinsed off. A magnetic transducer such as LC oscillator or Hall sensor will transduce the magnetism of the sample to electrical signals, which will be processed by the readout circuit subsequently ................................................................................. 21

Fig. 2.6 The magnetic-based handheld diagnostic device for antigen and nucleic acid detection. (a) The overall diagnostic device. The CMOS chip can be easily connected with the PCB by a cartridge. (b) The disposable cartridge with the CMOS chip. The CMOS chip is attached to the cartridge with silver epoxy and connected with bond wires to the carrier leads. This arrangement enables a disposable, low-cost, and multiplexed assay and simplifies the sample handling module such as microfluidic to manage the sample to the sensing sites. (c) The CMOS chip. It has 48 on-chip sensing sites together with 16 reference sensors. Each coil together with its own capacitor forms an LC oscillator, which has an oscillating frequency inversely proportional to the square root of the inductance of the coil. The surface of the chip is bio-functionalized for probe immobilization. The sample with the MP is then applied to the surface of the chip, followed by a washing step to rinse the unbound molecules and MPs. The bound MPs increase the inductance of the coils. Thus by detecting the oscillation frequency, the concentration of the target at the specific site can be selectively evaluated. (d) The experimental results for DNA detection. The frequency shift of the oscillation frequency is commensurate with the concentration of the target. With the novel magnetic freezing scheme, a limit of detection of 100 pM DNA can be achieved (Reproduced with permission from [22]. Published by the RSC 2014) .......................................................................... 22

Fig. 2.7 Architecture and operation of mechanical-based detection CMOS biosensor. (a) Mechanical-based detection with cantilever. A cantilever can be exploited to transduce the mass attached on it to electrical signals such as resistance. A gold layer is deposited on the cantilever for growing the capturing probe on it. In order to allow the cantilever to bend upon the biomolecule attached, the neighbor insulating dielectrics and the base of the cantilever are etched away. A piezoresistor can be adopted to transduce the bending force on the cantilever to resistance change, and the readout

List of Figures

xiv

circuit will detect this variation. (b) Mechanical-based detection with SAW transducer. A complete SAW transducer consists of three modules, input metal interdigital transducer (IDT), the piezoelectric delay line where the acoustic wave travels through, and the output metal IDT. The input IDT generates the SAW. Then the wave travels through the delay line to the output IDT, where the SAW is transduced back to the electrical signal. The bio-functionalized gold layer atop the delay line captures the entity under analysis. The increased mass here will affect the characteristics of the delay line, resulting in change of resonant frequency, amplitude, or phase shift on the SAW, which then can be detected on the output IDT ........ 23

Fig. 2.8 A CMOS cantilever-based biosensor for DNA detection. (a) The operation procedures of the biosensor. After post-processing to implement the cantilever on the CMOS chip, the capturing DNA is then immobilized on the Au surface of the cantilever. Then the cantilever is immersed in the PBS buffer, and the sample of interest is injected around the cantilever to allow hybridization of DNA. After washing unbound biomolecule, the cantilever is left to dry. After all of the water molecules are evaporated, the matched target DNA will stay on the Au surface. Their masses incur bending of the cantilever, and an embedded piezoresistor implemented by N+ polysilicon is entailed to sense this bending and transduce it to variation of its own resistance, causing a frequency shift on the ring-type oscillator. (b) The SEM image of the cantilevers. In order to allow the cantilever bending freely in air, the surrounding materials such as the insulating dielectrics and underneath the p-substrate have to be etched away, creating a suspending cantilever. (c) Experimental results for the biosensor. The resistance variation of the polysilicon piezoresistor attributed to the bending of the cantilever incurs in a deviation of the oscillating frequency. After DNA sample injection, washing, and drying steps, the final steady-state frequency can be measured to selectively quantify the concentration of the target DNA inside the sample with limit of detection of 1 pM from hepatitis B virus (Reproduced with permission from [19]. Copyright 2013 IEEE) .................................................................. 24

Fig. 2.9 Architecture and operation of NMR-based detection CMOS biosensor. NMR focuses on the measurement of the NMR signals from the samples. First, the MNP functionalized with the capture probes reacts with the sample under analysis. Then the mixture will be put atop the spiral sensing coil to perform NMR experiment. The existence of target inside the sample incurs in MNPs aggregation; thus a larger micro-cluster will be formed, changing the spin–spin relaxation time of the NMR signal from the sample ............................................. 25

List of Figures

xv

Fig. 2.10 The one-chip CMOS NMR-based biosensor. (a) The prototype of the platform. The system consists of a portable permanent magnet for magnetizing the 1H nuclei and the CMOS chip to excite the nuclei and receive the NMR signal from them. The samples are put directly on top of the CMOS chip without further post-processing. (b) The experimental results from the biological samples. Without the target the functionalized MNPs stay monodispersed, and the sample has a higher T2. With the target hCG cancer marker inside the samples, the hCG antibody binds with the hCG cancer marker, and they together form the micro-cluster. Thus the T2 of the sample decreased, and the concentration of the target can be identified from the NMR signal (Reproduced with permission from [28]. Copyright 2011 IEEE) .................................................................. 27

Fig. 2.11 Smart CMOS system-on-chip platform for rapid blood screening test of risk prediction. (a) The experimental procedure of the platform. Firstly, the blood under analysis is put atop the anodic aluminum oxide membrane. The biomarkers will be diffused to the mixing reservoir and separated from other blood cells (>1 μm). After the filtration, the filtered sample in the mixing reservoir together with the bio-functionalized magnetic bead will be pumped to the sensing site by the force from the electrolytic pumping. Upon capturing by the coated antibody at the surface of the CMOS chips, the target and the magnetic bead will be seized, while the unbound magnetic bead will be flushed away by the magnetic force from the on-chip coil. Thus the Hall sensor can sense the magnetic bead and identify the concentration of the targeted biomarker. (b) The photograph showing the electrolytic pumping and magnetic flushing. At first, the sample is on the right of the sensing reservoir. Then, voltage is applied to the electrolytic electrodes, and bubbles are formed consequently. The bubbles here induce gas force and pump the sample to the sensing reservoir. After the sample arrived at the sensing site, the immobilized antibodies capture the targets and the magnetic beads. Then the unbound magnetic beads will be flushed away by the on-chip coil. (c) The experimental result (TNF-alpha) of the immunoassay. The Hall sensor detects the target analyte from the magnetic beads on the sensing site. The system can detect 0.8 pg/mL–80 ng/mL of TNF-alpha and NT-proBNP from whole bloods (Reproduced with permission from [34]. Copyright 2015 IEEE) .............................. 28

Fig. 2.12 Integrated qPCR system on CMOS chip. (a) The CMOS chip and illustration of its functions. The chip has three main modules to enable on-chip qPCR. An electrowetting-on-dielectric device serves as an electronic-automated droplet management module to extract the target, PCR reagents, buffer, and intercalator dye

List of Figures

xvi

from the reservoirs, respectively, and guides them to different electrodes for mixing and subsequent operations by applying voltage on corresponding electrodes. A thermal module, which includes a resistive heater and temperature sensor, regulates the temperature of the droplets to perform thermal cycling for PCR. SPAD is embodied on the CMOS chip to detect the fluorescent emission from the target DNA in real time for qPCR. (b) Experimental results of the qPCR. The fluorescent signal from the sample increases with the PCR cycle. The qPCR system achieves a linear relationship between the cycle threshold and logarithm of initial DNA concentration from 1 to 10,000 copies per 1.2 nL of droplet, resulting in a 40,000-fold of reduction on reagent consumption (Reproduced with permission from [24]. Copyright 2014 RSC publishing) .................................................. 30

Fig. 2.13 CMOS multimodal sensor array for cell-based assay. (a) Schematic of the multimodal cell-based assay platform. The entire platform consists of 3 × 3 sensor array, and each pixel consisted of a photodiode, a temperature sensor (shared within a pixel group), a voltage amplifier, and an impedance detector for multimodal study and monitoring of the cultured cell exposed to drug or pathogen stimulation. (b) The micrograph of the CMOS cellular sensor chip. The chip contains 9 pixel groups for individual cell-based assay, and each pixel group further contains 16 individual pixels. Each pixel is formed by a gold-plated electrode for action potential and impedance reading with a photodiode. (c) Real-time experimental results from the bioluminescence experiment at 2 pixels. The human ovarian cancer cell emits luminescence upon the addition of luciferin, enabling verification of cell viability. The photodiode captures this bioluminescence, and the readout circuit processes the signal for subsequent analysis (Reproduced with permission from [48]. Copyright 2015 IEEE) .................................................................. 31

Fig. 2.14 A radar chart showing the conceptual requisites to perform the in vitro diagnosis on biomolecule targeting with different transducing mechanism ................................................................. 35

Fig. 3.1 The overall schematic and operations of the NMR–DMF system. (a) The placement of the DMF device, magnet, RF coil, and PCB in 3D view; (b) schematic of the NMR electronics; (c) the filtered results from the PCB are captured by the oscilloscope for easier demonstration purpose and then analyzed in MATLAB; (d) the photograph of the DMF device and its structure; (e) the detection mechanism of the NMR–DMF system. The target-specific MNPs, which act as probes, are placed on the sensing site initially (in purple). The sample at other electrodes (in cyan) will be transported to the sensing site and mixed with the probe to perform NMR assay .................................................................. 42

List of Figures

xvii

Fig. 3.2 Timing diagram of the pulses, including the excitation CPMG pulse sequence delivered to the TX to excite the nuclei and the response from the nuclei, which is picked up by the coil ............. 43

Fig. 3.3 (a) Geometry and limitation from the opening gap of the portable magnet. (b, c) The EM simulation of the magnetic field direction and strength from a spiral coil (with 14 turns) and a Butterfly-coil (with 7 turns on each spiral), respectively, with a flowing current of 1 A ................................................................................ 45

Fig. 3.4 Ratio of EC loss generated by the seven-turn (each loop) Butterfly-coil to coil magnetic energy against the thickness of the ITO. The figure was plotted based on (3.2) with f = 20 MHz, ρ = 1 × 10−6 Ωm, and A = 40 mm × 24 mm. The dotted line shows 0.5% level and corresponds to the ITO thickness of 80 nm ......... 47

Fig. 3.5 Nutation curve of the seven-turn (each loop) Butterfly-coil. The normalized amplitude from different durations of RF excitation signals was recorded and fitted to the sinusoidal wave. The estimated π/2-pulse width for the coil is 144 μs .................... 49

Fig. 3.6 (a) Received NMR signal from water. Inset shows the received NMR signal. The echoes were bounded by the gray-dotted trend line. (b) T2 of the samples versus concentration of CuSO4 solution, and results were shown on the graph (■). The trend lines were drawn together with their equation and 1/T2 value, together with error percentages (defined as half of 95% confidence level/true value) marked on the graph with dot lines where the values were displayed on the right axis ................................................... 50

Fig. 3.7 (a) Fabricated DMF device. For illustration, the electrodes are numbered 1–8; (b, c) operation of the DMF platform. The droplet was originally placed at electrode no. 1 (highlighted by the circle). By applying a signal on electrode no. 2 and then turning off electrode no. 1, the droplet moved to electrode no. 2. As such, the droplet can be transported to electrode no. 8, which is the NMR sensing site ........................ 51

Fig. 3.8 (a) Illustration of droplets mixing. The droplets at electrode no. 1 (samples) and no. 8 (probe) were driven to electrode no. 7 and mixed together. (b) The NMR assay results from the mixed droplets ................................................................ 52

Fig. 3.9 Portable electronic-automated micro-NMR system. It features a CMOS TRX and a PCB-based Butterfly-coil inside the magnet to transduce between magnetic and voltage signals. The analyte is placed inside a glass substrate DMF device atop the Butterfly-coil for sample management (only one electrode is shown for simplicity) ................................................................ 52

Fig. 3.10 Three-phase operation of the micro-NMR system: setup, sample preparation, and analysis .................................................. 53

List of Figures

xviii

Fig. 3.11 Block diagram of the micro-NMR TRX cooperated with the DMF device. It includes a CMOS micro-NMR TRX with a Butterfly-coil input, a DMF device, and DMF electronic. An electrode has the Butterfly-coil placed underneath for performing micro-NMR assays. An FPGA connected to a computer coordinates the hardware ....................................... 54

Fig. 3.12 Pulse-sequence synthesizer. FPGA commands control the logic gates to master the start and duration of the excitation signals with different phases as well as the switching between TX and RX modes ............................................................................... 55

Fig. 3.13 (a) Butterfly-coil-input LNA and its noise model. (b) Double-balance quadrature mixer with RF-sharing stage. (c) Source-follower-based tunable bandwidth LPF ...................... 56

Fig. 3.14 Simulation results of the mixer with LO = 20 MHz and input frequency = 20.002 MHz (i.e., IF = 2 kHz): (a) output against input for fundamental and third harmonic. (b) THD of the mixer at different input amplitudes ......................................................... 57

Fig. 3.15 (a) Simulated pole plot of the LPF. The sixth-order LPF implements a Butterworth filter (poles form a semicircle) with various cutoff frequencies obtained by changing only their bias currents. (b) Simulated THD of the LPF with an input frequency of 2 kHz and a cutoff frequency of 5 kHz for different input levels .................................................................................... 58

Fig. 3.16 The micro-NMR pulse sequence. It includes the CPMG pulse, filter current control, and micro-NMR output signal where the dead time of the RX is shown ...................................... 59

Fig. 3.17 Simulated SNR of the Butterfly-coil-input CMOS RX with different number of turns in the coils ................................... 60

Fig. 3.18 (a) Chip photo. (b) Measured performance summary of the micro-NMR TRX. The RX’s IRN, gain, and BW can only be assessed by simulations as the RX input has been tied to the Butterfly-coil ....................................................................... 61

Fig. 3.19 (a) Block diagram of the image-reject RX. (b) Measured RX output spectrum with an externally coupled magnetic field at 19.999 MHz and a LO of 20 MHz after image noise removal. (c) Cutoff frequency and settling time of the LPF versus the bias current. Working regions of the LPF at different modes are labeled ......... 62

Fig. 3.20 Measured B0 with and without calibration .................................... 63Fig. 3.21 The system hardware of the micro-NMR system. It is linked

with an FPGA (DE0-Nano) and a program implemented in C# which facilitates the system control, result collection, and displays ....... 63

Fig. 3.22 The pulses counted on the electrodes covered by air and water, respectively. As the permittivity of water is substantially larger than air (80:1), the capacitance of the electrode covered by water is higher, causing lower pulses to be counted, and thus the system can detect if the electrode is vacant .............................................. 64

List of Figures

xix

Fig. 3.23 Operation of the micro-NMR system. (a) Initial position of the sample and its projected path. (b) Droplet moves to the adjacent electrode. (c) Final position (micro-NMR sensing site) of the droplet. (d) Measured micro-NMR signal from water droplet excited by CPMG pulse sequence with 256 echoes and 4 ms interval. The envelope is extracted and fitted to a mono-exponential function, as shown in the inset ..................................................................... 65

Fig. 3.24 (a) The correlation of ΔT2−1 (with reference to 0 mM of CuSO4)

with the concentration of CuSO4. The echoes amplitude for the case of CuSO4 at 1 mM concentration is plotted above. One hundred twenty-eight echoes were collected for each single experiment. (b) The correlation of ΔT2 (with reference to 0 μM of avidin) with the concentration of avidin. The echoes amplitude for the case of avidin at 0.2 μM concentration is plotted above. Sixty-four echoes are collected for each single experiment....................................... 66

Fig. 3.25 (a) Illustration of the motions of the droplets for multistep multi-sample handling. T2 for the water sample: 256 ms; for avidin: 211 ms. (b) A Gantt chart of the operation of an individual droplet. The total time for the experiment is 2.2 min................................. 67

Fig. 4.1 Conceptual diagram of the proposed micro-NMR platform for PoU applications. Different samples such as protein and DNA can be put directly atop the CMOS chip for assays. A portable magnet is entailed to magnetize the nuclei inside the samples ................. 74

Fig. 4.2 System block diagram. The TX and RX transduce between magnetic and electrical signals with a thermal-controlled spiral coil. The B0-field sensor and calibrator automatically stabilize the bulk magnetization on the μL sample. No frequency synthesizer is required ..................................................................................... 75

Fig. 4.3 (a) Simulated 3D temperature distribution of the droplet at applied power of 8 mW in COMSOL Multiphysics®; (b) Simulated droplet average temperature at applied power from 0 to 20 mW .............. 76

Fig. 4.4 The cross section of a single VHS element and its current path. (a) Without lateral magnetic field; (b) with lateral magnetic field B0 .......................................................................................... 77

Fig. 4.5 Proposed current-mode fourfolded VHS arranged in Wheatstone bridge to sense the lateral B0-field and its readout circuit (spinning circuitry is omitted for simplicity). The latter features a nominal B0-field compensator to offset the strong nominal B0-field (0.46 T) for better sensitivity (3.75 mT). The green arrows highlight the current paths of IHall. Inset shows the timing diagram for the switches and overall operations ......................................... 78

Fig. 4.6 Illustration for the two-phase spinning technique on the VHS. The bias direction (U1 and U3) together with the output terminals (U2 and U4) of the VHS is swapped periodically to eliminate the 1/f noise and offset of the elements ......................................... 78

Fig. 4.7 Simulated frequency response of the TIA with various TINT ........ 79

List of Figures

xx

Fig. 4.8 Simulated channel resistance (RDS) and parasitic capacitance (CS+CD) of the MOS versus channel width .................................. 80

Fig. 4.9 Simulated output waveforms of the integrator. Without the nominal B0-field compensator, the integrator is saturated due to the large current induced by the nominal B0-field before it accumulates an adequate voltage difference. Whereas with the compensator, the nominal B0-field can be compensated; thus, the integration time can be prolonged to produce sufficient voltage differences at the output .................................................................................. 81

Fig. 4.10 (a) Chip photo of the fabricated chip in 0.18-μm CMOS. (b) Prototype of the micro-NMR platform with B0-field stabilization and lab-on-a-chip feasibility for multi-type biological/chemical assays, including (1) permanent magnet, (2) CMOS micro-NMR chip (inside magnet), (3) PCB, (4) FPGA, and (5) current driver. (c) Experimental setup. A program developed in C# is entailed for hardware control and visualizing the experimental results. The platform is powered by two batteries for portability ............. 82

Fig. 4.11 Timing diagram of the B0-field calibration and its frequency-domain illustration .................................................. 83

Fig. 4.12 (a) Measured hall sensor response; (b) B0-field with and without calibration. Actual B0-field is the sum of the B0-fields from the permanent magnet and the auxiliary coil driven by the current driver ...................................................................... 84

Fig. 4.13 (a) Measured power consumption and FoM of the XO at different supply voltages. (b) Measured phase noise of the XO (VDD = 0.9 V, f = 78.5 MHz). Compared with the LO generated from signal generator (Agilent 3350A), the XO shows a much better phase noise at low power .................................................... 84

Fig. 4.14 Experimental results from biological samples. (a) Target quantification from human IgG as target and chicken IgY as control. (b) Target quantification from Enterococcus faecalis- derived DNA together with single-base mismatch DNA .............. 85

Fig. 4.15 Experimental results from biological/chemical samples. (a) Protein (β-LG) state detection with different heating temperature. (b) Polymer (PNIPAM) dynamics with the solvent during heating from the on-chip heater ................................................................. 86

Fig. A.1 Measured gain of the NMR RX .................................................... 96

Fig. A.2 Measured output spectrum of the RX with a 100-nV, 20-MHz sinusoidal input ............................................................................. 96

Fig. B.1 Visualized waveform applied to the electrode before and after the droplet arrives at the electrode ................................. 98

Fig. C.1 The communication between the PC and the FPGA board to drive the micro-NMR relaxometer. It is done by adopting the TTL-232R_PCB module to interfacing between the PC and FPGA board, which mastered the hardware of the micro-NMR relaxometer .... 100

List of Figures

xxi

Table 2.1 Recent CMOS-based DNA-related biosensors............................. 13Table 2.2 Recent CMOS-based protein-related biosensors .......................... 14Table 2.3 Recent CMOS-based cell-related biosensors ............................... 15

Table 3.1 Summary of the measured and simulated coil parameters at 20 MHz ..................................................................................... 48

Table 3.2 Simulated noise summary of the LNA ......................................... 56Table 3.3 Comparison with the existing CMOS-based NMR system .......... 68

Table 4.1 Summary and benchmark with other CMOS-based PoU systems ................................................................................. 87

Table 4.2 Benchmark with previous CMOS NMR systems ......................... 88

List of Tables

1© Springer International Publishing AG 2018 K.-M. Lei et al., Handheld Total Chemical and Biological Analysis Systems, https://doi.org/10.1007/978-3-319-67825-2_1

Chapter 1Introduction

1.1 Overview

An essential part to evaluate the success of global health is the access to appropriate diagnostic tools [1]. A commendable diagnostic tool should be able to identify the disease occurred from the individuals rapidly. Especially for the infectious diseases, the turnaround time (TAT) for the diagnosis strongly affects their exacerbation level to the community. In vitro diagnostic (IVD) tool is aimed to offer a comfortable diagnosis for the patients, by taking only small specimens from the human body, e.g., blood, urine, or sputum, for analysis. Consequently, technologies enabling effective in vitro diagnosis become highly attractive for both developed and devel-oping countries [2]. Tremendous efforts have been geared toward developing clinical- level IVD tools. Despite achieving high accuracy, the resulting TAT can be too long for diagnoses of contagious diseases like Ebola and SARS in the rural area, and the requisite of skillful operators and sophisticated equipment to perform the assays can dramatically raise the cost of the assay.

Recently, decentralized diagnostic solutions, namely, point-of-care (PoC) devices, have gained notable interests attributed to their fastness, small footprint, and tiny sample usage. Wide varieties of diagnostic platforms have been invented, such as the lateral flow assays [3–6] and pathbreaking lab-on-a-disc immunoassay [7–10] for PoC applications. Beyond them, PoC devices on complementary metal–oxide–semiconductor (CMOS) chips are particularly promising, as they can enjoy the maturity of microelectronics in manufacturing and its outstanding performances in both physical sensing and signal processing. While the mainstream lateral flow assay is confounded to provide merely qualitative or semiquantitative results [11], the CMOS biosensors can attain a quantitative result and are beneficial to rapid and low-cost assays. Especially for low-cost IVD applications, CMOS chips in a centi-meter scale can significantly miniaturize the diagnostic tools.

2

1.2 Global Necessities for In Vitro Diagnostic Tools

Decentralized healthcare systems are highly attractive for developing countries, as they typically suffer from lack of access to high-quality centralized diagnostic tools in the resource-limited area. Delay of diagnosis and treatment aggravates the health-care condition of their countries, then affecting the global health system. According to the World Development Report in 2004, the lack of access results in failure of the health services [12]. Without proper equipment for diagnosis, the clinicians could only diagnose diseases from the clinical symptoms in resource-limited regions. Yet, this may cause difficulties in the diagnosis of the patients when the symptoms are still unobvious. Especially for infectious diseases, the delay of treatment can worsen the situation of individuals and consequently the communities. According to the report of the World Health Organization (WHO), the leading infectious diseases (lower respiratory infections, HIV/AIDS, diarrheal diseases, malaria, and tubercu-losis) account for roughly one-third of all deaths in low-income countries [13]. Also, the strong growths of the population in those areas give rise to the demand of affordable IVD tools. By the end of 2050, the less developed countries are expected to have a population of 8.4 billion, as depicted in Fig. 1.1, where Africa and Asia contribute roughly 2.48 and 5.27 billion, respectively [14]. Thus, there is a rapidly growing market of low-cost PoC devices for developing countries.

0

2

4

6

8

10

12

1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050Year

More Developed Countries Less Developed Countries

More Developed Countries (Est.) Less Developed Countries (Est.)

Popu

latio

n (B

illio

n)

Fig. 1.1 World population from 1950 to 2050, with a medium variant estimation from 2015. Data collected from the United Nations World Population Prospects: The 2015 Revision [14]. More developed countries: countries in Europe and Northern America, plus Australia/New Zealand and Japan. Less developed countries: countries in Africa, Asia (except Japan), Latin America, and the Caribbean plus Melanesia, Micronesia, and Polynesia

1 Introduction

3

The aging problem of the developed countries also creates an enormous chal-lenge. A healthcare solution that can deal with the continuous increment of life longevity is of demand. As revealed in Fig. 1.2, the old age dependency ratio, which gives insight to the population of elderly (65+ years) as a share of those in working age (age 15–64 years), will be rising in the coming decades. The old age depen-dency ratio of more developed countries will reach 0.4 in 2034 and eventually 0.46 by the end of 2050 (i.e., increase by 72% from 2015). Thus, the burdens on the clini-cal resources in those areas will become tighter, especially for the patients in prox-imity to death [15]. An efficacious healthcare solution will benefit this situation and drive the growth of the market for IVD tools. To this end, the market for IVD tools should not be merely aimed at less developing countries but also toward efficient and convenient diagnosis in developed countries. In fact, according to the report from Forbes/Investing, the IVD market, valued at ~$50 billion in 2012, will expand to $70 billion by 2017 [16].

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050Year

Ratio

More Developed Countries Less Developed Countries

More Developed Countries (Est.) Less Developed Countries (Est.)

Fig. 1.2 The old age dependency ratio (solid line), which is defined as the ratio of population of 65+ years old to the population of 15–64 years old with medium variant estimation from 2015. The children dependency ratio (dotted line), which is defined as the ratio of population of 0–14 years old to the population of 15–64 years old, is also shown on the graph as reference. Data collected from the United Nations World Population Prospects: The 2015 Revision [14]. More developed countries: countries in Europe and Northern America plus Australia/New Zealand and Japan. Less developed countries: countries in Africa, Asia (except Japan), Latin America, and the Caribbean plus Melanesia, Micronesia, and Polynesia

1.2 Global Necessities for In Vitro Diagnostic Tools

4

1.3 Nuclear Magnetic Resonance for In Vitro Diagnosis

Nuclear magnetic resonance (NMR) is powerful to explore the sample information at the molecular level. The underpinning physics of NMR is the exchange of energy between the RF magnetic field and the spin of the non-zero spin nuclei (i.e., 1H, 13C, 17O, 31P, etc.) [17, 18]. Under the magnetization with an external magnetic field B0, parts of the nuclei align with this external magnetic field and have a spin-up state, while the others have a spin-down state and align in opposite direction (Fig. 1.3a). As the population ratio between the nuclei with spin-up and spin-down state is pro-portional to B0 and this difference determines the amplitude of the NMR signal, there exists a tradeoff between the portability and sensitivity of the system as dis-cussed later. With an RF magnetic field B1 orthogonal to B0 applied to the nuclei, they precess and tip away from the direction of bulk magnetization (Fig. 1.3b). The nuclei only accept RF excitation at Larmor frequency, defined as:

f BL = γ 0 (1.1)

Nuclei

Magneticmoment

Without external magnetization With external magnetization

B0

B0B1 (at fL)

(a)

(b)

Before RF excitation Right after RF excitation After RF excitation (relaxation)

Spin-upSpin-down

Fig. 1.3 (a) Macroscopic view of the non-zero spin nuclei. With an external magnetic field B0 applied to the nuclei, part of them will align with this magnetic field. (b) The effect of RF excita-tion on the nucleus under external magnetization. When excited by the RF magnetic field at fL, the nuclei precess around the magnetization. After this excitation, the nuclei still resonate and return to the equilibrium, with this relaxation recorded and analyzed

1 Introduction

5

with the gyromagnetic ratio of the nuclei γ. For a 0.46-T magnet, the fL of 1H is ~20 MHz. The nuclei do not precess if there is a mismatch on the excitation fre-quency and fL. After tipping the nuclei with 90° from the direction of bulk magneti-zation, the excitation is turned off. Then, the recovery of the magnetization in parallel with B0 is defined as the spin-lattice relaxation time T1 whereas the recovery of the magnetization perpendicular with B0 is defined as the spin-spin relaxation time T2. This T2 reveals the magnetic field decoherence information across the nuclei. Unfortunately, the unavoidable inhomogeneity of B0 from portable magnet causes spatial variation on the precession rate of the nuclei thus the T2 decays at a much faster rate T2

*. This blemish hinders the measurement on original T2. To sur-mount this, the spin-echo technique such as Carr–Purcell–Meiboom–Gill (CPMG) pulse sequence can be utilized to refocus this dephasing effect on the nuclei by flip-ping the nuclei 180° with interval τ; thus the spins are maximized again from B0 inhomogeneity [19, 20]. The envelope of the echoes responses allows the derivation of the resulting T2 with the following mathematical expression:

A n A e

n

T[ ] =−

02

τ

(1.2)

with the echoes amplitude for nth echoes A[n] and the initial amplitude A0. More importantly, the strength of B0 correlates to the signal-to-noise ratio (SNR) of the NMR signal and is described as [21]:

SNR KBNMR ∝

1

7

4

4

1 2

Fl f

fL

ξ∆πρ

(1.3)

with homogeneity factor K, magnetic field strength per unit current produced by the RF-coil orthogonal to the permanent magnetic field B1, the noise figure of the receiver’s forefront amplifier F, length of the RF-coil conductor l, the bandwidth of the system Δf, and the resistivity ρ of the RF coil. From (1.1) and (1.3), the SNR of the system is proportional to the power of 7/4 of B0, thus demanding a stronger B0 to enhance the SNR. Although there seem to be numerous ways to enhance B0 and its homogeneity (i.e., for higher resolution and stronger signal), the portability and power consumption of the system will be penalized due to the need for a heavier and bulkier magnet, not to mention a higher operating frequency that will be required for the electronics.

By exploiting functionalized magnetic nanoparticles (MNPs) as the probe, the NMR-based quasi label-free detection scheme can pinpoint a broad range of unpro-cessed biological targets such as DNA [22], protein [23], and virus [24] for in vitro diagnosis. These superparamagnetic MNPs have significant impacts on T2 of the samples according to their magnetization (Ms) attributed to their capability to per-turb the local magnetic field homogeneity. When the target is absent in the sample, the MNPs stay monodisperse inside the solution (Fig. 1.4a). Consequently, when

1.3 Nuclear Magnetic Resonance for In Vitro Diagnosis

6

the targets exist inside the samples, they will cross-link with the probe- functionalized MNPs, assembling nanoparticles micro-clusters (Fig. 1.4b). These micro-clusters, with a diameter dc depending on the concentration of the target biomolecules, have a different magnetization MC [25]:

M Md

dC Sc

s

f

=

−3

(1.4)

with the fractal dimension of the micro-cluster f and the diameter of a single MNP ds. Accordingly, T2 of the sample is commensurate with MC. In this respect, T2 is linked with the amount of target upon nanoparticle agglomeration and attainable for quantification. Unlike other sensing schemes, screening by NMR is rapid and low cost as it is quasi label-free for the samples and immobilization-free for the trans-ducers/electrodes. Such benefits render NMR-based detection as a promising solu-tion for PoC applications. Although NMR is known for its relatively low sensitivity, the MNP here provides inherent signal amplification to NMR since a single MNP micro-cluster can affect billions of adjacent water molecules [26].

Conventionally NMR equipment are bulky and have limited applicability for PoC diagnosis. Recently, researchers have been focusing on miniaturizing the mag-net for NMR and migrating the modular and complex electronics into CMOS chips [27–29]. With advanced circuit techniques to lessen the penalty of signal attenua-tion induced by the compact magnet (<7.3 kg) as stated in (1.3), these micro-NMR systems offer a trailblazing sensing method befitting more the PoC diagnosis.

1.4 Organization

The book is organized as below:Chapter 2 reviews the state-of-the-art CMOS biosensors for in vitro diagnosis

[30]. The first session focuses on introducing different transducing mechanisms for CMOS biosensors. According to the transducing mechanism, the CMOS biosensors can be categorized into electrical based, optical based, magnetic based, mechanical based, and NMR based. In the second session, in vitro diagnosis on different bio-

Target

MNP

Probe

(a) (b)

Fig. 1.4 The state of the probe-functionalized MNPs. (a) Without the target, the MNPs stay monodisperse in the solution without any aggregation. (b) When the targets exist in the sample, the targets bind with the probe, and the MNPs aggregate to form micro-clusters

1 Introduction

7

logical targets with CMOS biosensors and their applications is discussed. Subsequently, the strengths and weaknesses of different sensing mechanisms for these biosensors are compared in detail.

Chapter 3 illustrates the design and implementation of the palm-size micro-NMR relaxometer with the electronic-automated digital microfluidic device for sample management. Confounded by the limited inner space of the portable magnet, man-agement of the sample under assay remains a critical issue for micro-NMR. Herein the integration of micro-NMR with the digital microfluidic device is evinced to facilitate the sample management and attain a comparable assay result to the con-ventional approach [31–35].

Chapter 4 discloses the design of the micro-NMR relaxometer with B0-field sta-bilization module [36, 37]. Ascribed to the temperature instability of the magnet, calibration on either excitation frequency or Larmor frequency of the protons is nec-essary to safeguard the operation of the micro-NMR relaxometer. Herein, a vertical Hall sensor and relevant readout circuit are integrated with the micro-NMR trans-ceiver to sense the B0-field variation of the magnet for calibrating the B0 field of the portable magnet by injecting a current to its auxiliary coil. The closed-loop B0-field stabilization here can suppress the variation of the B0 field and ease the operation.

Finally, Chap. 5 concludes the book and discusses the potential future works.

References

1. D.C.H. Burgess, J. Wasseramn, C.A. Dahl, Global health diagnostics. Nature (suppl. 1) 444, 1–2 (2006)

2. P. Yager, G.J. Domingo, J. Gerdes, Point-of-care diagnostics for global health. Annu. Rev. Biomed. Eng. 10(1), 107–144 (2008)

3. M. Zuiderwijk, H.J. Tanke, R. Sam Niedbala, P.L.A.M. Corstjens, An amplification-free hybridization-based DNA assay to detect Streptococcus pneumoniae utilizing the up- converting phosphor technology. Clin. Biochem. 36(5), 401–403 (2003)

4. X. Fu, Z. Cheng, J. Yu, P. Choo, L. Chen, J. Choo, A SERS-based lateral flow assay biosensor for highly sensitive detection of HIV-1 DNA. Biosens. Bioelectron. 78, 530–537 (2016)

5. J. Li, J. Macdonald, Multiplex lateral flow detection and binary encoding enables a molecular colorimetric 7-segment display. Lab Chip 16(2), 242–245 (2016)

6. J.R. Choi, J. Hu, R. Tang, Y. Gong, S. Feng, H. Ren, et al., An integrated paper-based sample- to- answer biosensor for nucleic acid testing at the point of care. Lab Chip 16(3), 611–621 (2016)

7. M. La, S.M. Park, D.S. Kim, Centrifugal multiplexing fixed-volume dispenser on a plastic lab-on-a-disk for parallel biochemical single-end-point assays. Biomicrofluidics 9(1), 014104 (2015)

8. B.S. Lee, Y.U. Lee, H.-S. Kim, T.-H. Kim, J. Park, J.-G. Lee, et al., Fully integrated lab-on-a- disc for simultaneous analysis of biochemistry and immunoassay from whole blood. Lab Chip 11(1), 70–78 (2011)

9. J. Park, V. Sunkara, T.-H. Kim, H. Hwang, Y.-K. Cho, Lab-on-a-disc for fully integrated mul-tiplex immunoassays. Anal. Chem. 84(5), 2133–2140 (2012)

10. W.S. Lee, V. Sunkara, J.-R. Han, Y.-S. Park, Y.-K. Cho, Electrospun TiO2 nanofiber integrated lab-on-a-disc for ultrasensitive protein detection from whole blood. Lab Chip 15(2), 478–485 (2015)

References

8

11. G.A. Posthuma-Trumpie, J. Korf, A. van Amerongen, Lateral flow (immuno)assay: its strengths, weaknesses, opportunities and threats. A literature survey. Anal. Bioanal. Chem. 393(2), 569–582 (2009)

12. World development report: making services work for poor people (World Bank, New York, 2004)

13. The top 10 causes of death, Available: http://www.who.int/mediacentre/factsheets/fs310/en/index2.html. Accessed 10 June 2016

14. World population prospects: the 2015 revision (United Nations, New York, 2015) 15. A. Palangkaraya, J. Yong, Population ageing and its implications on aggregate health care

demand: empirical evidence from 22 OECD countries. Int. J. Health Care Finance Econ. 9(4), 391–402 (2009)

16. Z. Miller, Investing in the future of medicine: a investor’s guide to the in vitro diagnostics mar-ket, Available: http://www.forbes.com/sites/zackmiller/2014/02/12/investing-in-the-future-of-medicine-a-investors-guide-to-the-in-vitro-diagnostics-market/#9d6e5ea5feb4. Accessed 10 June 2016

17. B. Cowan, Nuclear magnetic resonance and relaxation (Cambridge University Press, Cambridge, 1997)

18. N.E. Jacobsen, NMR spectroscopy explained: simplified theory, applications and examples for organic chemistry and structural biology (Wiley, Hoboken, 2007)

19. H.Y. Carr, E.M. Purcell, Effects of diffusion on free precession in nuclear magnetic resonance experiments. Phys. Rev. 94(3), 630–638 (1954)

20. S. Meiboom, D. Gill, Modified spin-echo method for measuring nuclear relaxation times. Rev. Sci. Instrum. 29(8), 688–691 (1958)

21. D.I. Hoult, R.E. Richards, The signal-to-noise ratio of the nuclear magnetic resonance experi-ment. J. Magn. Reson. 24(1), 71–85 (1976)

22. L. Josephson, J.M. Perez, R. Weissleder, Magnetic nanosensors for the detection of oligo-nucleotide sequences. Angew. Chem. 113(17), 3304–3306 (2001)

23. J.M. Perez, L. Josephson, T. O’Loughlin, D. Hogemann, R. Weissleder, Magnetic relaxation switches capable of sensing molecular interactions. Nat. Biotechnol. 20(8), 816–820 (2002)

24. J.M. Perez, F.J. Simeone, Y. Saeki, L. Josephson, R. Weissleder, Viral-induced self-assembly of magnetic nanoparticles allows the detection of viral particles in biological media. J. Am. Chem. Soc. 125(34), 10192–10193 (2003)

25. C. Min, H.L. Shao, M. Liong, T.J. Yoon, R. Weissleder, H. Lee, Mechanism of magnetic relax-ation switching sensing. ACS Nano 6(8), 6821–6828 (2012)

26. H. Lee, E. Sun, D. Ham, R. Weissleder, Chip-NMR biosensor for detection and molecular analysis of cells. Nat. Med. 14(8), 869–874 (2008)

27. N. Sun, Y. Liu, H. Lee, R. Weissleder, D. Ham, CMOS RF biosensor utilizing nuclear magnetic resonance. IEEE J. Solid State Circuits 44(5), 1629–1643 (2009)

28. N. Sun, T.J. Yoon, H. Lee, W. Andress, R. Weissleder, D. Ham, Palm NMR and 1-Chip NMR. IEEE J. Solid State Circuits 46(1), 342–352 (2011)

29. D. Ha, J. Paulsen, N. Sun, Y.Q. Song, D. Ham, Scalable NMR spectroscopy with semiconduc-tor chips. Proc. Nat. Acad. Sci. (PNAS) 111(33), 11955–11960 (2014)

30. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, CMOS biosensors for in vitro diagnosis – transducing mechanisms and applications. Lab Chip 16(19), 3664–3681 (2016)

31. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, NMR–DMF: a modular nuclear magnetic reso-nance–digital microfluidics system for biological assays. Analyst 139(23), 6204–6213 (2014)

32. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A thermal-insensitive all-electronic modular μNMR relaxometer with a 2D digital microfluidic chip for sample management, in Proceeding 19th International Conference on Miniaturized System Chemistry and Life Science (MicroTAS), 2015, pp. 302–304

33. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A μNMR CMOS transceiver using a Butterfly- coil input for integration with a digital microfluidic device inside a portable magnet, in Proceeding IEEE Asian Solid-State Circuits Conference (A-SSCC), 2015, pp. 1–4

1 Introduction

http://www.who.int/mediacentre/factsheets/fs310/en/index2.html

http://www.who.int/mediacentre/factsheets/fs310/en/index2.html

http://www.forbes.com/sites/zackmiller/2014/02/12/investing-in-the-future-of-medicine-a-investors-guide-to-the-in-vitro-diagnostics-market/#9d6e5ea5feb4

http://www.forbes.com/sites/zackmiller/2014/02/12/investing-in-the-future-of-medicine-a-investors-guide-to-the-in-vitro-diagnostics-market/#9d6e5ea5feb4

9

34. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A palm-size μNMR relaxometer using a digi-tal microfluidic device and a semiconductor transceiver for chemical/biological diagnosis. Analyst 140(15), 5129–5137 (2015)

35. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A μNMR CMOS transceiver using a Butterfly- coil input for integration with a digital microfluidic device inside a portable magnet. IEEE J. Solid State Circuits 51(10), 2274–2286 (2016)

36. K.-M. Lei, H. Heidari, P.-I. Mak, M.-K. Law, F. Maloberti, R.P. Martins, A handheld 50pM- sensitivity micro-NMR CMOS platform with B-field stabilization for multi-type biological/chemical assays, in IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), 2016, pp. 474–475

37. K.-M. Lei, H. Heidari, P.-I. Mak, M.-K. Law, F. Maloberti, R.P. Martins, A handheld high- sensitivity micro-NMR CMOS platform with B-Field stabilization for multi-type biological/chemical assays. IEEE J. Solid State Circuits 52(1), 284–297 (2017)

References


Chapter 2State-of-the-Art CMOS In Vitro Diagnostic Devices

2.1 Introduction

In vitro diagnosis focuses on analyzing the targets from the biological specimens. Different biological objects, such as protein, DNA, or cell/bacteria, can be targeted for diagnosis. This chapter firstly introduces different transducing mechanisms of CMOS PoC devices for biological sensing and targeting. Then different CMOS IVD tools for various targets of interest are discussed in Sect. 2.3. Section 2.4 out-lines and compares the pros and cons of the transducing mechanisms to provide a systematic outlook of their distinct characteristics and limitations.

2.2 Transducing Mechanisms of CMOS IVD Tools

Rapid downscaling of CMOS technology has enabled the possibility of integrating billions of transistors onto a single chip, allowing ultrafast signal processing at low power and low cost. While their transformative effects on computers and mobile devices have been witnessed, the development of CMOS PoC devices for in vitro diagnosis only started recently. While the electronic noise of the CMOS circuits may affect the sensitivity of the result, innovative circuit techniques such as time averaging, frequency modulation, and differential operation can be entailed to sup-press the effects of noise. Along with these circuit innovations, high-sensitivity ana-log interfaces are ready for the CMOS biosensors. Thus, single-chip CMOS biosensors allow full integration of the transducers and circuits necessary for signal processing. This property aligns with the expectation of PoC devices such as low sample consumption, midget footprint, easy-to-use, and rapid and quantitative results. Attributed to the design flexibility of CMOS chips, multifarious CMOS bio-sensors have been reported. Further, the capability to include post-processing steps

12

(e.g., chemical deposition and micromachining) after the chip manufacturing can help broadening the versatility of CMOS biosensors [1–6].

According to the transducing mechanisms, CMOS PoC devices can be classified as (1) electrical-based, sensing directly the electrical properties of samples, such as impedance or electric charges (currents); (2) optical-based, transducing the optical properties from the samples such as fluorescent label or chemiluminescence to an electrical signal; (3) magnetic-based, transducing the magnetic properties of the samples arising from the labeled magnetic particles (MPs) to an electrical signal; (4) mechanical-based, transducing the mechanical properties, namely the mass of the samples, to an electrical signal; and (5) NMR-based, transducing the nuclear spin from the sample that is affected by the magnetic susceptibility, to an electrical signal. Details of each transducing mechanism and its corresponding example from Tables 2.1, 2.2, and 2.3 will be presented next.

2.2.1 Electrical-Based

Ascribed to the capability of the CMOS chip to seamlessly handle the electrical signals such as current, impedance, and capacitance, direct electronic detection bio-sensors are the most popular types of CMOS biosensors for in vitro diagnosis (Fig. 2.1). The electrical properties serve as a reliable measure to quantify the con-centration of analyte inside the samples. The electrode acting as the substrate for probe immobilization is usually formed by the top metal layer, which offers great feasibility for the downscaling of CMOS technologies, and is expandable to form an array for multi-targeting.

A popular approach is the electrochemical impedance spectroscopy (EIS), where the impedance between the electrodes is monitored over a range of frequencies. It provides details of the impedance changes of the solution and the electrode–solution interface. EIS has been involved in detecting the affinity-based biosensing, which explores the binding events between the probe (i.e., complementary oligonucle-otides, antibody/antigen) and the target within the samples. The probe can be immo-bilized on the electrodes with special chemical post-processing [50]. Upon hybridization between the probe and target, the double-layer capacitance of the elec-trode–electrolyte interface decreases due to the increased thickness together with the decreased dielectric constant, whereas the charge transfer resistance increases since the attached targets partially block the flow of the ions [51]. Thus, the system is capable of quantifying the analyte of interest by detecting the impedance between the electrodes. For instance, Manickam et al. showed a versatile EIS biosensor array on a CMOS chip [15]. This fully integrated and label-free detecting platform is capable of quantifying different DNA and protein targets inside the samples with correspond-ing immobilized probes on the electrodes. Alternatively, the capacitance between the electrodes can also be entailed to detect the target upon the hybridization of DNA [52]. Lee et al. reported a CMOS label-free capacitive biosensor for DNA detection from H5N1 virus [17]. Instead of observing the complex impedance of the elec-

2 State-of-the-Art CMOS In Vitro Diagnostic Devices

13

Tabl

e 2.

1 R

ecen

t CM

OS-

base

d D

NA

-rel

ated

bio

sens

ors

App

licat

ion

Sens

ing

para

met

ers

Tra

nsdu

cers

Lab

elin

gIm

mob

iliza

tion

Low

est a

mou

nt

repo

rted

Yea

r

DN

A d

etec

tion

Cap

acita

nce

Au

elec

trod

esL

abel

-fre

eY

esN

/A20

06 [

7]D

NA

det

ectio

nC

harg

eIS

FET

Lab

el-f

ree

Yes

0.1

mM

2006

[8]

DN

A d

etec

tion

Mag

netis

mSp

in-v

alve

sen

sors

Mag

netic

bea

d (t

arge

t)Y

es10

nM

2007

[9]

DN

A p

olym

eriz

atio

nC

harg

ePo

lym

er +

met

alL

abel

-fre

eY

esN

/A20

08 [

10]

DN

A d

etec

tion

Fluo

resc

ence

Phot

odet

ecto

rC

y-3

labe

l (ta

rget

)Y

es0.

125

nmol

2009

[11

]D

NA

det

ectio

nFl

uore

scen

cePh

otod

etec

tor

Bio

tin a

nd

Qdo

t-65

5 (t

arge

t)Y

es4

nM20

09 [

12]

DN

A d

etec

tion

Mag

netis

mL

C o

scill

ator

Dig

oxig

enin

and

m

agne

tic b

ead

(tar

get)

Yes

1 nM

2009

[13

]

DN

A d

etec

tion

Cyc

lic v

olta

mm

etry

Au

elec

trod

esFe

rroc

ene

redo

x la

bel (

targ

et)

Yes

4 nM

2009

[14

]

DN

A d

etec

tion

Impe

danc

eA

u el

ectr

odes

Lab

el-f

ree

Yes

N/A

2010

[15

]D

NA

det

ectio

nIm

peda

nce

Au

elec

trod

esL

abel

-fre

eY

es5 μM

2012

[16

]D

NA

det

ectio

nC

apac

itanc

eA

u el

ectr

odes

Lab

el-f

ree

Yes

100

pM20

12 [

17]

DN

A d

etec

tion

Cha

rge

Au

elec

trod

esL

abel

-fre

eY

es10

0 pM

2012

[18

]D

NA

det

ectio

nM

ass

Can

tilev

erL

abel

-fre

eY

eslp

M20

13 [

19]

DN

A a

mpl

ifica

tion

and

dete

ctio

npH

ISFE

TL

abel

-fre

eN

o10

cop

ies

2013

[20

]

DN

A d

etec

tion

Cha

rge

Silic

on n

anow

ire

Lab

el-f

ree

Yes

3.2

pM20

13 [

21]

DN

A d

etec

tion

Mag

netis

mL

C o

scill

ator

Mag

netic

bea

d (p

robe

)Y

es10

0 pM

2014

[22

]

DN

A d

etec

tion

Cyc

lic v

olta

mm

etry

Au

elec

trod

esL

abel

-fre

eY

es10

aM

2014

[23

]D

NA

am

plifi

catio

n an

d de

tect

ion

Fluo

resc

ence

SPA

DE

vaG

reen

® d

yeN

o1

copy

2014

[24

]


14

Tabl

e 2.

2 R

ecen

t CM

OS-

base

d pr

otei

n-re

late

d bi

osen

sors

App

licat

ion

Sens

ing

para

met

ers

Tra

nsdu

cers

Lab

elin

gIm

mob

iliza

tion

Low

est a

mou

nt

repo

rted

Yea

r

C-r

eact

ive

prot

ein

Mas

sC

antil

ever

Lab

el-f

ree

Yes

1 μg

/mL

2009

[25

]A

vidi

nN

ucle

ar s

pin

Spir

al c

oil (

off-

chip

)M

agne

tic b

ead

(pro

be)

No

5 nM

2009

[26

]

hMA

M c

ance

r m

arke

rM

ass

SAW

tran

sduc

erL

abel

-fre

eY

es1.

5 μg

/mL

2010

[27

]

Prot

ein

GIm

peda

nce

Au

elec

trod

esL

abel

-fre

eY

esN

/A20

10 [

15]

hCG

can

cer

mar

ker

Nuc

lear

spi

nSp

iral

coi

lM

agne

tic b

ead

(pro

be)

No

5 nM

2011

[28

]

Hum

an s

erum

al

bum

inM

agne

tism

Hal

l sen

sor

Mag

netic

bea

d (p

robe

)Y

es1

ng/m

L20

13 [

29]

SLPI

can

cer

mar

ker

Mag

netis

mG

MR

spi

n-va

lve

sens

or (

off-

chip

)M

agne

tic b

ead

(pro

be)

Yes

10 f

M20

13 [

30]

Car

diac

trop

onin

I

prot

ein

Cha

rge

Silic

on n

anow

ire

Lab

el-f

ree

Yes

10 f

M20

13 [

21]

Cyt

okin

es (

IL8,

IF

N, T

NFα

)C

hem

ilum

ines

cenc

ePh

otod

etec

tor

Hor

sera

dish

pe

roxi

dase

Yes

3 pg

/mL

2014

[31

]

C-r

eact

ive

prot

ein

Cap

acita

nce

Au

elec

trod

esL

abel

-fre

eY

es0.

5 m

g/L

2015

[32

]St

rept

avid

inFl

uore

scen

cePh

otod

etec

tor

Qdo

t 800

flu

orop

hore

Yes

48 z

mol

2015

[33

]

TN

F-al

pha

and

NT-

prob

npM

agne

tism

Hal

l sen

sor

Mag

netic

bea

d (p

robe

)Y

esN

/A20

15 [

34]

Mul

ti-an

alyt

eC

hem

ilum

ines

cenc

ePh

otod

etec

tor

Hor

sera

dish

pe

roxi

dase

Yes

45 n

g/m

L

Ferr

itin

2015

[35

]

Mou

se I

gGM

agne

tism

Mic

roco

il (o

ff-c

hip)

Mag

netic

bea

d (p

robe

)Y

es10

0 pg

/mL

2016

[36

]


15

Tabl

e 2.

3 R

ecen

t CM

OS-

base

d ce

ll-re

late

d bi

osen

sors

App

licat

ion

Sens

ing

para

met

ers

Tra

nsdu

cers

Lab

elin

gIm

mob

iliza

tion

Low

est a

mou

nt

repo

rted

Yea

r

Bov

ine

aort

ic s

moo

th

mus

cle

cell

mon

itori

ngC

apac

itanc

eM

etal

ele

ctro

des

Lab

el-f

ree

No

20 ×

20 μm

220

07 [

37]

MD

A-M

B-2

31 b

reas

t ca

ncer

cel

l mon

itori

ngC

apac

itanc

eM

etal

ele

ctro

des

Lab

el-f

ree

No

20 ×

20 μm

220

08 [

38]

Bla

dder

can

cer

cell

dete

ctio

nN

ucle

ar s

pin

Spir

al c

oil

Mag

netic

bea

d (p

robe

)N

o17

.5 c

ells

/μL

2011

[28

]

MC

F-7

brea

st c

ance

r ce

ll de

tect

ion

Fluo

resc

ence

SPA

D5D

10 m

onoc

lona

l an

tibod

yN

oSi

ngle

cel

l20

12 [

39]

MC

F-7

brea

st c

ance

r ce

ll de

tect

ion

Impe

danc

eA

u el

ectr

odes

Lab

el-f

ree

No

Sing

le c

ell

2012

[40

]

RB

C fl

ow c

ytom

eter

Impe

danc

eM

etal

ele

ctro

des

Lab

el-f

ree

No

Sing

le c

ell

2012

[41

]C

ardi

ac p

roge

nito

r ce

ll m

onito

ring

Mag

netis

mL

C o

scill

ator

Mag

netic

bea

d (t

arge

t)N

oN

/A20

12 [

42]

HeL

a ce

ll co

untin

gO

ptic

sC

MO

S im

age

sens

orL

abel

-fre

eN

o25

cel

ls/m

L20

14 [

43]

Mou

se e

mbr

yoni

c fib

robl

ast c

ell

Mag

netis

mE

xcita

tion

and

pick

up

coils

Mag

netic

bea

d (t

arge

t)N

oSi

ngle

cel

l20

14 [

44]

Com

plet

e bl

ood

coun

tO

ptic

sC

MO

S im

age

sens

orL

abel

-fre

eN

oN

/A20

14 [

45]

HeL

a ce

ll co

untin

gC

yclic

vol

tam

met

ryA

u el

ectr

odes

Lab

el-f

ree

No

N/A

2015

[46

]M

CF7

, BE

AS,

and

K56

2 ca

ncer

cel

l mon

itori

ngC

apac

itanc

eA

u el

ectr

odes

Lab

el-f

ree

No

N/A

2015

[47

]

Hum

an c

ardi

omyo

cyte

s m

onito

ring

Mul

timod

alA

u el

ectr

odes

and

ph

otod

etec

tor

Lab

el-f

ree

No

N/A

2015

[48

]

Hep

G2

canc

er c

ell

mon

itori

ngO

ptic

sC

MO

S im

age

sens

orL

abel

-fre

eN

o50

00 c

ells

/mL

2015

[49

]


16

trodes, this capacitive biosensor only focuses on the measurement of the capacitance of the electrodes; thus the circuit design can be simplified. The capacitive sensing mechanism is also an adequate candidate for cell monitoring. Prakash et al. presented a CMOS platform for cancer cell MDA-MB-231 proliferation monitoring (Fig. 2.2a) [38]. Since the capacitance can be sensed without removing the passivation layer, post-processing on CMOS chip can be obviated (Fig. 2.2b). The underlying principle is the insulating nature of the cell together with the counterionic polarization in the surrounding aqueous medium when exposed to low-frequency electric fields, which counts for the capacitive behavior of the cell (Fig. 2.2c).

Alternatively, voltammetric, amperometric, and potentiometric techniques moni-tor the electrochemical reactions (i.e., charge transfer) within the samples directly. These electrochemical reactions are related to the concentration of the analytes. With a specific readout circuit such as a potentiostat, these electrochemical reac-tions can be measured, and the concentration of the biomolecule can be quantified. Levine et al. described a CMOS multiplexed electrochemical microsystem with a 4 × 4 electrode array for DNA detection [14]. By sweeping the voltage between the electrodes, the current produced by redox label on the target DNA sequence can be recorded by the on-chip electronics.

With the advance of post-processing in standard CMOS chips, special field- effect transistors (FET) biosensor, which senses the charges of the ions/electrons from the target such as ion-sensitive FET (ISFET) [8] and silicon nanowire (NW) [6, 21], also shows the promises for biomolecule detection. Compared to the stan-dard electrical-based biosensors shown in Fig. 2.1a, they have a different architec-ture and usually require further post-processing. Their operations are similar to

P-substrate

Readout circuit

Au electrode

Built-in metallayers and vias

Target

Capturingprobe

Insulating dielectrics

Fig. 2.1 Architecture and operation of electrical-based detection CMOS biosensor. An extra layer of noble and biocompatible metal such as gold is deposited on the original built-in metal layer. The capturing probe is then immobilized on the gold electrode to capture the target. Upon hybridization the electrical properties such as impedance or charge are sensed directly by the readout circuit

Fig. 2.2 (continued) The experimental results for cancer cell MDA-MB-231 culturing. The capaci-tance at specific site increases due to the proliferation of the cancer cells ascribed to the increased number of cells, allowing real-time monitoring for the growth of the cancer cells (Reproduced with permission from [38]. Copyright 2008 Elsevier)


17

Fig. 2.2 Cell culturing and monitoring with CMOS capacitive sensing chip. (a) The photograph showing the overall chip with dual in-line package. A well encloses the cell culturing site, and the CMOS chip is at the center of the well. The polymer protects the bond wires of the chip. (b) Photomicrograph of the electrodes. Since the system measures only the capacitance of the single electrode, the built-in passivation layer such as silicon nitride and silicon dioxide can be preserved without further post-processing. This simplifies the hardware preparation steps for biosensing. (c)


18

MOSFET: the target concentration alters the charges on the surface of the FET; thus the physical properties of the channel are altered. For instance, Huang et al. reported a CMOS wireless biomolecular sensing system based on polysilicon NW [21]. The resistance of the n-type doping polysilicon NW decreases upon the hybridization of complementary DNA on the NW, since the electrons in the NW are repelled from the surface. Thus, by detecting the resistance of the NW, the quantity of the DNA inside the samples can be identified.

2.2.2 Optical-Based

As the dominant detection modality for existing diagnostic tools is within the visible range of fluorescence spectroscopy, optical sensing still plays an important role in the CMOS biosensing platform since the conventional immunoassay protocols such as enzyme-linked immunosorbent assay (ELISA) can migrate to the CMOS chip. The principle of fluorescent biomolecule assay is the detection of the fluorescence signal from the labeled fluorescent tag after washing the unconjugated tag. Upon the excita-tion from an external light source, these fluorescent labels emit signals at longer wavelength due to Stokes shift. Then, the designated transducers such as photodetec-tor, which is implemented by an embedded PN-junction available in the standard CMOS process, can collect these signals and transduce them to the electrical domain (i.e., voltage or current) for signal amplification and conditioning (Fig. 2.3).


P-substrate

N-well

P+ active

Excitation light sources(for fluorescene)

Fluorescent/

Chemiluminescent tag

Target

Capturing probe

Optical filter/Faceplate (if necessary)

Readoutcircuit

Fig. 2.3 Architecture and operation of optical-based detection CMOS biosensor. The capturing probe is immobilized on a solid substrate such as glass or the built-in passivation layer atop the CMOS chip. Then fluorescence-labeled or chemiluminescence-labeled target will bind with the probe, and other unbound biomolecules will be washed away. The CMOS photodetector, which is formed by the embedded PN-junction, transduces the optical signal to current for subsequent sig-nal processing


19

In order to integrate the fluorescence module into a CMOS chip, several consid-erations have to be taken into account. Firstly, since the signals are in the optical domain, the metal layers above the photodetector should be avoided to prevent blocking the signal. Secondly, as the intensity of the excitation signal is usually much stronger than the fluorescent signal, its presence at the photodetector will saturate it thus causing system malfunctioning. Jang et al. proposed a CMOS fluorescent- based biosensor microarray for DNA detection [11]. To filter the excita-tion signal and prevent it from reaching the photodetector, a separate 20-layer thin- film long-pass optical filter has to be put atop the CMOS chip. This filter rejects the excitation signal at 532 nm by 98 dB while preserving a low passband attenuation of 1 dB. Further, to direct the fluorescence signals with the photodetector, a fiber- optic faceplate was fabricated to guide the signals along the vertical direction. These CMOS-incompatible modules stiffen the integration and raise the system cost. To this end, Huang et al. proposed an integrated time-resolved fluorescence detection CMOS array sensor suitable for DNA detection [12]. Instead of removing the exci-tation signal by an optical filter, the proposed system adopts a time-gated arrange-ment to monitor the fluorescent decay of the labels after the excitation. This scheme leads to a high signal-to-background ratio without any external filters. Lately, a fully integrated CMOS fluorescence biosensor with on-chip nanophotonic filter was pro-posed [33]. The design included an integrated optical filter obtained with back-end- of-line copper layers. This filter can suppress the excitation light without any external filtering/collimation, facilitating the integration of the entailed CMOS elec-tronics and optical filters for fluorescent assays.

Similar to fluorescent detection, the chemiluminescent signal can also be detected from the signaling tag with the photodetector. The chemiluminescent tag, which emits light signal based on the chemical reaction, averts the external light source to excite them. This prevents the saturation of the photodiode. Further, the avoidance of optical filter shortens the distance from the signaling tag to the surface of the photodetector, which fulfills the criteria for supercritical angle luminescence (SAL) and enables efficient signal detection [53]. Sandeau et al. proposed a large-area CMOS bio-pixel array for multiplex biosensing [31]. The high refraction index of the silicon substrate enables the SAL phenomenon, and thus the direction of light emission from the chemiluminescent tag is confounded to the surface, and the sen-sitivity is boosted by 100-fold compared to the standard microplate reader.

For larger biological objects such as cell and bacteria, direct optical detection without labeling tag is also feasible by using a high-speed CMOS image sensor. With the advance of CMOS imaging technologies, real-time cell monitoring and counting can also be achieved on a CMOS chip directly. The appearance of the objects blocks the photon from arriving the CMOS image sensor; thus the system can count and monitor the activity of the cells/bacteria. Different CMOS-based cell monitoring systems have been described in the literature. Saeki et al. proposed a lens-free cell-counting device for HeLa cells with a microcavity array (Fig. 2.4a, b) [43]. The cells are trapped on the microcavity array by applying a negative pressure underneath. Hence, the cells trapped inside the microcavity block the light from the light-emitting diode and create a shadow on the CMOS image sensor; thus the CMOS image sensor can enumerate the cells on the array (Fig. 2.4c, d).


20

2.2.3 Magnetic-Based

Magnetism has become a tantalizing mechanism for biomolecule detection recently. Ascribed to the absence of magnetic signal in the sample and background, the mag-netism originated from the labeled magnetic micro- or nanoparticles can be clearly detected from the samples. In addition, as the magnetic field can penetrate the insu-lating dielectric layers, direct contact between the samples and the transducers is avoided. This simplifies the hardware preparation of the CMOS chip before the assay. The basic operation is to grow an immobilization layer on a surface proximity to the transducer similar to the electrical- and optical-based biosensors. Then, the magnetic-labeled sample can be put on the transducer for hybridization. After a washing step to flush the unbound biomolecules and MPs, the target of interest with MPs will stay on the surface allowing the transducer to detect the corresponding signal if the target exists inside the sample (Fig. 2.5).

Fig. 2.4 Lens-free cell/microparticle counting system with CMOS image sensor. (a) The overall platform of the digital cell counting device. (b) The micrograph of the microcavity array for cell trap-ping. The sample under analysis is put atop the microcavity array. Then the suspended cells/mic-roparticles will be pulled toward and trapped in the cavities attributed to the negative pressure. This negative pressure is produced by peristaltic pump, which extracts the air inside the chamber. (c) Detection principle of the system. The light from the external UV light source will arrive at the CMOS image sensor through the unoccupied cavity, while the trapped cell on the cavity blocks the light from arriving at the CMOS image sensor. (d) The schematics of the expected CMOS image acquired from (c). Since the cell blocks the UV light from passing through the cavity, the pixels under those occupied cavity will report a darker region, while the pixels under the vacant cavity will report a brighter result. Thus the number of cells on the microcavity array can be identified from the result of the CMOS image sensor (Reproduced with permission from [43]. Copyright 2014 Saeki et al.)


21

There are diverse transducers to convert this magnetism to an electrical signal. For instance, Hall sensor can be adopted to sense the magnetic field and transduce it to voltage/current signals. The Hall sensor composed of an n-type silicon can offer a moderate mobility (i.e., sensitivity) yet fully compatible with the CMOS process. The current carriers in the sensor experience Lorentz force when there is a magnetic field orthogonal to the direction of current flow applied to them. This causes charge deflection and allows electronic detection. Gambini et al. proposed a CMOS Hall sensor array for immunoassays [29]. A Hall sensor was adopted to perform the relaxation measurement on the magnetic field and transduce this to a voltage signal. With the bound MPs, the relaxations of the magnetic field from the particles relax to zero according to the Néel relaxation mechanism. Thus, by detect-ing this relaxation time, the concentration of the MPs (i.e., the target concentration) can be detected. Alternatively, an inductor can also be employed to sense the mag-netism of the samples. As the inductance is heavily affected by the magnetic suscep-tibility in proximity, the target can be quantified by detecting its inductance. Pai et al. presented a CMOS magnetic biosensor array for DNA and protein detection using an embedded LC oscillator (L and C represent the inductor and capacitor), where the inductor implementation uses the top metal layer of the CMOS chip (Fig. 2.6a, c) [22]. The bound MPs are detected from the oscillation frequency of the oscillator. They demonstrated the detection of 100 pM DNA by a novel magnetic freezing technique to soothe the sensor noise (Fig. 2.6d).

2.2.4 Mechanical-Based

Mass is a fundamental parameter for analyzing the concentration of the target by utilizing an immobilized probe on the transducer without labeling process. The probes capture the targets in the samples, and thus the mass on the transducer increases. Then a specific transducer converts the mass variation to an electrical signal.

P-substrate

Readoutcircuit

Magnetic sensitivetransducers

Magnetic bead

Target

Capturing probe


Fig. 2.5 Architecture and operation of magnetic-based detection CMOS biosensor. The capturing probe is immobilized on a solid substrate such as glass or the built-in passivation layer atop the CMOS chip. Then the sample labeled with MP will mix with the capturing probe. Matched target will be captured, and unbound objects will then be rinsed off. A magnetic transducer such as LC oscillator or Hall sensor will transduce the magnetism of the sample to electrical signals, which will be processed by the readout circuit subsequently


22

Fig. 2.6 The magnetic-based handheld diagnostic device for antigen and nucleic acid detection. (a) The overall diagnostic device. The CMOS chip can be easily connected with the PCB by a cartridge. (b) The disposable cartridge with the CMOS chip. The CMOS chip is attached to the cartridge with silver epoxy and connected with bond wires to the carrier leads. This arrangement enables a disposable, low-cost, and multiplexed assay and simplifies the sample handling module such as microfluidic to manage the sample to the sensing sites. (c) The CMOS chip. It has 48 on- chip sensing sites together with 16 reference sensors. Each coil together with its own capacitor forms an LC oscillator, which has an oscillating frequency inversely proportional to the square root of the inductance of the coil. The surface of the chip is bio-functionalized for probe immobiliza-tion. The sample with the MP is then applied to the surface of the chip, followed by a washing step to rinse the unbound molecules and MPs. The bound MPs increase the inductance of the coils. Thus by detecting the oscillation frequency, the concentration of the target at the specific site can be selectively evaluated. (d) The experimental results for DNA detection. The frequency shift of the oscillation frequency is commensurate with the concentration of the target. With the novel magnetic freezing scheme, a limit of detection of 100 pM DNA can be achieved (Reproduced with permission from [22]. Published by the RSC 2014)


23

An example of this kind of mechanical transducer is the cantilever (Fig. 2.7a). It senses the concentration of the analyte by the mass attached to the cantilever and is designed to work in the static or dynamic mode [54]. In the static mode, only one surface of the cantilever is bio-functionalized and thus the bending occurs in one direction statically. While working in the dynamic mode, the entire cantilever oper-ates in the capture of the targets with its resonant frequency sensitive to the mass attached. By adopting special design and post-processing such as etching and addi-tional layer depositing, the CMOS chip is capable of embodying a cantilever and transduces the mass attached to electrical signals. If the targets are presented inside the samples, the probe will seize the targets raising the mass of the cantilever as a mechanical effect. This event can be detected from the resistance, resonant frequency, or optical deflection of the cantilever, depending on the design and criteria of the platforms. Chen et al. demonstrated silicon-based micro-cantilever to detect the C-reactive protein from the sample [25]. The deflection ascribed to the additional mass on the cantilever is detected by the displacement of the optical beam, which is reflected by the surface of the cantilever and detected by another photodetector chip. Yet, the separate modules and the inclusion of the optical gadgets (i.e., laser beam

P-substrate


Readoutcircuit

Au electrode

N+ poly

Etched

TargetCapturing probe

P-substrate

Readout circuit

Zinc Oxide

Delay line

Input IDT Output IDTBuilt-in metal

TargetCapturing probe

(a) (b)

Fig. 2.7 Architecture and operation of mechanical-based detection CMOS biosensor. (a) Mechanical-based detection with cantilever. A cantilever can be exploited to transduce the mass attached on it to electrical signals such as resistance. A gold layer is deposited on the cantilever for growing the capturing probe on it. In order to allow the cantilever to bend upon the biomolecule attached, the neighbor insulating dielectrics and the base of the cantilever are etched away. A piezoresistor can be adopted to transduce the bending force on the cantilever to resistance change, and the readout circuit will detect this variation. (b) Mechanical-based detection with SAW trans-ducer. A complete SAW transducer consists of three modules, input metal interdigital transducer (IDT), the piezoelectric delay line where the acoustic wave travels through, and the output metal IDT. The input IDT generates the SAW. Then the wave travels through the delay line to the output IDT, where the SAW is transduced back to the electrical signal. The bio-functionalized gold layer atop the delay line captures the entity under analysis. The increased mass here will affect the char-acteristics of the delay line, resulting in change of resonant frequency, amplitude, or phase shift on the SAW, which then can be detected on the output IDT


24

generator, lens, and mirror) stiffen the integration of the systems and their applicabil-ity outside the laboratory. Subsequently, their group proposed an integrated cantile-ver system-on-chip for label-free DNA of hepatitis B virus detection [19]. In this work the deflection of the cantilever is detected by embedded n+ polysilicon piezo-resistor, whose resistance varies upon the bending of the cantilever (Fig. 2.8a, b). Hence the optical gadgets can be eliminated allowing the integration of the cantilever with the readout circuitry. They demonstrated the detection of DNA from hepatitis B virus from the frequency deviation of the ring-type oscillator after washing and dry-ing of water molecules from the cantilever with sensitivity down to 1 pM (Fig. 2.8c).

Another novel type of mechanical transducer in CMOS is the surface acoustic wave (SAW) transducer (Fig. 2.7b) that detects the acoustic wave transmission pat-tern between two transducers. When the target binds with the immobilized probe in the acoustic wave transmission path (delay line), the properties of the output will change thus allowing mass detection [55]. Tigli et al. proposed a SAW biosensor in CMOS technology to detect the cancer marker human mammaglobin (hMAM) [27]. By adopting a post-processing step, a piezoelectric zinc oxide layer can be

Fig. 2.8 A CMOS cantilever-based biosensor for DNA detection. (a) The operation procedures of the biosensor. After post-processing to implement the cantilever on the CMOS chip, the capturing DNA is then immobilized on the Au surface of the cantilever. Then the cantilever is immersed in the PBS buffer, and the sample of interest is injected around the cantilever to allow hybridization of DNA. After washing unbound biomolecule, the cantilever is left to dry. After all of the water molecules are evaporated, the matched target DNA will stay on the Au surface. Their masses incur bending of the cantilever, and an embedded piezoresistor implemented by N+ polysilicon is entailed to sense this bending and transduce it to variation of its own resistance, causing a fre-quency shift on the ring-type oscillator. (b) The SEM image of the cantilevers. In order to allow the cantilever bending freely in air, the surrounding materials such as the insulating dielectrics and underneath the p-substrate have to be etched away, creating a suspending cantilever. (c) Experimental results for the biosensor. The resistance variation of the polysilicon piezoresistor attributed to the bending of the cantilever incurs in a deviation of the oscillating frequency. After DNA sample injection, washing, and drying steps, the final steady-state frequency can be measured to selectively quantify the concentration of the target DNA inside the sample with limit of detection of 1 pM from hepatitis B virus (Reproduced with permission from [19]. Copyright 2013 IEEE)


25

deposited on the CMOS chip. This zinc oxide layer affects the resonant frequency of the SAW along the delay line, according to the mass attached on it permitting the quantification of the hMAM biomarker.

2.2.5 NMR-Based

Recently, NMR has gained popularity for biomolecule detection. The underlying physics of NMR is the exchange of energy between the nuclear spins of the atoms and the radio-frequency magnetic field in the presence of a static magnetic field. Similar to the magnetic sensing scheme, NMR-based biosensors rely on the mag-netic labels to selectively detect the target. Yet, it indirectly detects these magnetic labels from the 1H NMR signals of aqueous samples. The spin–spin relaxation time of the NMR signals will be disturbed by the magnetization and the size of the MNPs [56]. The probe-decorated MNPs stay monodispersed before mixing with the target. After mixing with the target inside the samples, the probe binds to the target, and thus the MNPs gather and form micro-clusters. These micro-clusters have a larger size and different magnetization, and thus the relaxation time of the sample will change accordingly. By detecting the NMR signals from the samples, the concentra-tion of the target can be quantified from their spin–spin relaxation time. An NMR spiral sensing coil implemented by top metal layer transduces between the electrical signals from the readout circuit and the magnetic field generated by the nuclear spins (Fig. 2.9). Compared to the direct magnetic sensing scheme, NMR-based technique radically eases the hardware preparation since immobilization of probe on CMOS chip is not required.

P-substrate

Readout circuit

Spiral coil

Vias

Insulatingdielectrics

Magnetic bead

Capturing probe A

Capturing probe B

Target

Micro-clusters

Fig. 2.9 Architecture and operation of NMR-based detection CMOS biosensor. NMR focuses on the measurement of the NMR signals from the samples. First, the MNP functionalized with the capture probes reacts with the sample under analysis. Then the mixture will be put atop the spiral sensing coil to perform NMR experiment. The existence of target inside the sample incurs in MNPs aggregation; thus a larger micro-cluster will be formed, changing the spin–spin relaxation time of the NMR signal from the sample


26

The versatility and flexibility of NMR-based biosensing enable different kinds of biomolecule assays. While the NMR-based detection was originally performed with a discrete prototype [57–63], significant efforts have evolved to integrate the elec-tronics and detecting coil into a single CMOS chip for portable and low-cost assays. For instance, Sun et al. reported a one-chip NMR system with a portable permanent magnet (Fig. 2.10a) [28]. It equipped with a transceiver along with the sensing coil to perform the NMR experiment and enable LOC operation. The relaxation time of the samples containing the targets is sensitive to the amount of the MNP inside the samples, as explained in Sect. 1.3. Thus the system is capable of detecting avidin, human chorionic gonadotropin (hCG) cancer marker, and bladder cancer cell inside the 5-μL samples (Fig. 2.10b).

2.3 In Vitro Diagnostic Applications

This section introduces the assays with CMOS biosensors and their biological applications suitable for in vitro diagnosis according to Tables 2.1, 2.2, and 2.3.

2.3.1 Immunoassay

The prevailing assay for detecting protein-based targets is the enzyme immunoassay. It is based on enzyme-labeled antibody (antigen) to detect the target antibody (anti-gen). The most dominant method is ELISA [64]. By observing the visual signal such as color change from enzyme or fluorescent tag, the concentration of target inside the sample can be quantified. ELISA is referred as the gold standard in the clinical labora-tory for a broad range of applications, such as dengue [65], cancer marker [66], and H5N1 influenza [67]. Despite the popularity of ELISA in the conventional laboratory, it is labor-intensive and time-consuming; both limit its prevalence in the resource-limited area. In this regard, significant efforts have been endeavoring to migrate the immunoassay to CMOS chip for superior performance and small form factor. Klapproth et al. demonstrated a multi-analyte CMOS sensor to measure multiple sand-wich-ELISA reactions performed on the CMOS chip [35]. The chemiluminescence is recorded by the on-chip photodetector working in the reverse-bias region. It allows parallel detection of different biological targets such as immunoglobulin E and myo-globin and shows comparable results to the clinical protocols. Alternatively, Kuo et al. reported a smart CMOS assay system-on-chip platform for rapid blood screening with a Hall sensor to transduce the magnetic field generated by an on-chip magnetic coil to voltage signal (Fig. 2.11a) [34]. An anodic aluminum oxide membrane filters the bio-markers from the blood sample, which is then pumped to the Hall sensor array by the electrolytic electrodes for immunoassay (Fig. 2.11b). With the bound MPs, the mag-netic field generated by the coil will be enhanced. The system can detect the biological target such as tumor necrosis factor-α for antitumor response and N-terminal pro-brain natriuretic peptide for heart failure response from the whole blood (Fig. 2.11c).


27

Fig. 2.10 The one-chip CMOS NMR-based biosensor. (a) The prototype of the platform. The system consists of a portable permanent magnet for magnetizing the 1H nuclei and the CMOS chip to excite the nuclei and receive the NMR signal from them. The samples are put directly on top of the CMOS chip without further post-processing. (b) The experimental results from the biological samples. Without the target the functionalized MNPs stay monodispersed, and the sample has a higher T2. With the target hCG cancer marker inside the samples, the hCG antibody binds with the hCG cancer marker, and they together form the micro-cluster. Thus the T2 of the sample decreased, and the concentration of the target can be identified from the NMR signal (Reproduced with per-mission from [28]. Copyright 2011 IEEE)


28

2.3.2 DNA Hybridization Assay

Diagnoses with DNA biomolecule have a broad range of applications such as muta-tion in DNA sequence for cancer prediction [68] and pathogen detection [69]. Detection of the specific DNA is based on the DNA hybridization. DNA hybridiza-tion is a molecular biology technique to identify the sequence of interest, with a principle analogy to immunoassay. A single-strand DNA complementary to the tar-get DNA acts as a probe and is necessitated to selectively detect the target. If the target DNA is complementary to the designated probe, DNA hybridization occurs, and double-stranded DNA will be formed. Then the detector will detect the desig-nated signaling tag to quantify the target DNA inside the sample.

Fig. 2.11 Smart CMOS system-on-chip platform for rapid blood screening test of risk prediction. (a) The experimental procedure of the platform. Firstly, the blood under analysis is put atop the anodic aluminum oxide membrane. The biomarkers will be diffused to the mixing reservoir and separated from other blood cells (>1 μm). After the filtration, the filtered sample in the mixing reservoir together with the bio-functionalized magnetic bead will be pumped to the sensing site by the force from the electrolytic pumping. Upon capturing by the coated antibody at the surface of the CMOS chips, the target and the magnetic bead will be seized, while the unbound magnetic bead will be flushed away by the magnetic force from the on-chip coil. Thus the Hall sensor can sense the magnetic bead and identify the concentration of the targeted biomarker. (b) The photo-graph showing the electrolytic pumping and magnetic flushing. At first, the sample is on the right of the sensing reservoir. Then, voltage is applied to the electrolytic electrodes, and bubbles are formed consequently. The bubbles here induce gas force and pump the sample to the sensing res-ervoir. After the sample arrived at the sensing site, the immobilized antibodies capture the targets and the magnetic beads. Then the unbound magnetic beads will be flushed away by the on-chip coil. (c) The experimental result (TNF-alpha) of the immunoassay. The Hall sensor detects the target analyte from the magnetic beads on the sensing site. The system can detect 0.8 pg/mL–80 ng/mL of TNF-alpha and NT-proBNP from whole bloods (Reproduced with permission from [34]. Copyright 2015 IEEE)


29

Conventionally the DNA detection relied on fluorescence-based detection simi-lar to immunoassay [70]. It shares similar drawbacks with the immunoassay such as the bulky optical instrument. Thus, considerable efforts have been geared toward implementing the DNA assay tool with CMOS chip for multifarious biological pur-poses. For instance, Jafari et al. proposed a nanostructured CMOS ultra-wideband label-free PCR-free DNA detection system [23]. It supports 54-channel fast-scan cyclic voltammetry DNA analysis. The electrical-based transducing mechanism here enables the complete integration of the platform including the transducers (electrodes), current readout circuit, waveform generator for cyclic voltammetry, and transmitter for communication. The system demonstrated label-free detection of prostate cancer synthetic DNA without DNA amplification, featuring a detection range from 10 aM to 10 μM. In another work, Norian et al. implemented an inte-grated CMOS quantitative polymerase chain reaction (qPCR) LOC platform by integrating a thermal module, a digital microfluidic module, and a single-photon avalanche diode (SPAD) on a single CMOS chip (Fig. 2.12a) [24]. The proposed platform achieved fully integrated qPCR instrumentation, with proof-of-concept detection on DNA from Staphylococcus aureus with detection limit down to 1 copy per 1.2 nL droplet, reducing the reagent consumption by 40,000x (Fig. 2.12b).

2.3.3 Cell/Bacteria Diagnosis

Cell-level diagnosis also plays a significant role in the IVD field. An example is the counting of CD4+ T cell in the human whole blood to spot the human immunodefi-ciency virus (HIV) infection. CD4+ T cell is a kind of white blood cell essential to the human immune system. Upon infection by the HIV virus, it causes depletion of CD4+ T cells, and thus the immune system degenerates. An efficient approach to diagnose and monitor the HIV infection is CD4+ T-cell counting. Flow cytometry is the gold standard for counting the CD4+ T cells [71]. The basic principle is to enu-merate the CD4+ T cells passing through the detector during the continuous flow of the sample. The appearance of the CD4+ T cells will alter the parameters of interest, such as impedance, or light beam on the photodetector such that the number of cells inside the samples can be counted. Yet, conventional flow cytometer requires bulky detection tools and large volume of sample. In this regard, Lee et al. proposed a CMOS impedance cytometer to monitor the flow inside the PDMS microchannel [41]. They utilized the cytometer to diagnose the rigidity of the red blood cell (RBC). A rigid RBC opposes to deformation caused by shear stress in the medium and is related to distinctive diseases, which offers a potential for microcirculation study.

Another key application of cell-level diagnosis is the cell monitoring for growth, cytotoxicity, and virus detection inside the cell. Monitoring the cellular activity (i.e., cell division, apoptosis, and necrosis) using electronics enables real-time automated assay on designated cells when compared with the traditional microscopic approach. There have been efforts to monitor the cells using different electronic-based tech-niques such as impedance sensing and imaging for different diagnostic purposes.


30

For instance, Chi et al. proposed a CMOS 3 × 3 multimodal cell-assay platform for cellular assay (Fig. 2.13a, b) [48]. Benefitting from the high integration level of the CMOS technology, different kinds of sensors such as photodiode, temperature sen-sor, and impedometer are integrated within a single pixel, rendering it a promising platform for joint-modality cellular physiological monitoring. Various cells such as ovarian cancer cell and human cardiomyocytes have been entailed for biological experiments and verification (Fig. 2.13c). Recently, Laborde et al. reported a 256 × 256 nanocapacitor array for real-time imaging of microparticles and living

Fig. 2.12 Integrated qPCR system on CMOS chip. (a) The CMOS chip and illustration of its func-tions. The chip has three main modules to enable on-chip qPCR. An electrowetting-on-dielectric device serves as an electronic-automated droplet management module to extract the target, PCR reagents, buffer, and intercalator dye from the reservoirs, respectively, and guides them to different electrodes for mixing and subsequent operations by applying voltage on corresponding electrodes. A thermal module, which includes a resistive heater and temperature sensor, regulates the tempera-ture of the droplets to perform thermal cycling for PCR. SPAD is embodied on the CMOS chip to detect the fluorescent emission from the target DNA in real time for qPCR. (b) Experimental results of the qPCR. The fluorescent signal from the sample increases with the PCR cycle. The qPCR sys-tem achieves a linear relationship between the cycle threshold and logarithm of initial DNA concen-tration from 1 to 10,000 copies per 1.2 nL of droplet, resulting in a 40,000-fold of reduction on reagent consumption (Reproduced with permission from [24]. Copyright 2014 RSC publishing)


31

Fig. 2.13 CMOS multimodal sensor array for cell-based assay. (a) Schematic of the multimodal cell-based assay platform. The entire platform consists of 3 × 3 sensor array, and each pixel con-sisted of a photodiode, a temperature sensor (shared within a pixel group), a voltage amplifier, and an impedance detector for multimodal study and monitoring of the cultured cell exposed to drug or pathogen stimulation. (b) The micrograph of the CMOS cellular sensor chip. The chip contains 9 pixel groups for individual cell-based assay, and each pixel group further contains 16 individual pixels. Each pixel is formed by a gold-plated electrode for action potential and impedance reading with a photodiode. (c) Real-time experimental results from the bioluminescence experiment at 2 pixels. The human ovarian cancer cell emits luminescence upon the addition of luciferin, enabling verification of cell viability. The photodiode captures this bioluminescence, and the readout circuit processes the signal for subsequent analysis (Reproduced with permission from [48]. Copyright 2015 IEEE)

2.4 Discussions and Selection Guide

32

cells on CMOS chip [47]. By simultaneously exploring the benefits of CMOS tech-nologies, they achieved label-free and high-throughput monitoring of different can-cer cells with attofarad resolution in the sub-micrometer scale.


Since there is a high diversification of the characteristics and properties of CMOS in vitro biosensors as shown in Tables 2.1, 2.2, and 2.3, this section aims to summa-rize a selection guide and provide a radar chart of each transducing method. Herein we analyze and discuss the properties, requisites, and limitations of the CMOS bio-sensors based on their transducing mechanisms, evaluated in terms of the integration level, labeling scheme, hardware preparation, operation steps, and specificity.

2.4.1 Integration Level

For cost and size reduction, it is desirable to integrate all necessary hardware of the biosensor into a unified platform. Electrical-based detection is an ideal solution for both sensing small biomolecules (DNA and proteins) and large biological objects (cell and bacteria). External nonelectronic gadgets (e.g., light source, optical filter or magnet, etc.) can be avoided. Similarly, mechanical-based detection is a promis-ing way for biosensing, except for the case of utilizing the laser beam for detecting the cantilever deflection [25]. Traditionally, magnetic-based detection involves either a permanent magnet [22] or external coil [30] to magnetize and sense the MP. Yet, there are certain efforts to eliminate these external gadgets by implement-ing the coil on the chip [34] or adopting the LC oscillator [13] to sense the MP. Optical-based biomolecule detection, especially for fluorescence, shares the same drawback of centralized benchtop assay such as the need for excitation light sources. The NMR-based detection, in this perspective, is unpleasant for integration since it involves a large permanent magnet (typical size of the magnet for NMR: 8 cm in diameter; 5.5 cm in height) for performing the NMR experiments. Further, the magnetic field generated by the permanent magnet is temperature dependent, and calibration is required to ensure proper operation [61].

2.4.2 Labeling

The labeling process determines the efforts and difficulties to prepare the sample and corresponding probes. The label-free assay is preferable, which refrains from complicated signal tag such as fluorescence, redox-active molecule, and MP label-ing process on the sample and probe to detect the target. Electrical-/mechanical- based detection is superior in this area since they both support label-free detection.


33

While certain works entailed redox tag for signaling [14], most of the electrical-/mechanical-based biosensors are label-free, easing the sample preparation before the assay. On the other hand, NMR-based biosensors utilize MP functionalized with the capturing probe for the detection. The labeling process stiffens the experiments when compared with the electrical-/mechanical-based biosensors. Nevertheless, since only the probe entails the surface functionalization instead of the target, the sample under assay can still remain unprocessed. For optical-/magnetic-based bio-sensors, they involve labeling of fluorescent, luminescent, or magnetic label on the target for assaying, which substantially increase the efforts and costs for the assay. Yet, sandwich-based bioassay technique can be applied to circumvent from labeling on the target to lessen the sample preparation steps.

2.4.3 Hardware Preparation

Hardware preparation indicates the procedures required to prepare the hardware after receiving the standard CMOS chip from the foundry and before the assay, which includes CMOS post-processing: etching, chemical depositing, or immobili-zation of capturing probe. Certainly, an ideal biosensor should refrain from these processes to reduce the cost and simplify the assay. NMR-based biosensor is prom-ising in this regard as it does not entail any surface modification, plating, nor probe immobilization step (i.e., chip is used as is). While the optical-/magnetic-based bio-sensors do not necessarily involve post-processing on CMOS chip since there is no direct contact, immobilization of probe on the substrate is still entailed. This sub-strate may be an external epoxy container, glass plate, or on-chip Au electrode. The electrical-/mechanical-based biosensors involve complex post-processing steps to deposit biocompatible Au layer on the predefined electrodes and immobilize the necessary probe. Especially for the cantilever, the underneath silicon substrate needs to be etched away to allow bending or oscillating of the cantilever. These post-processing steps greatly raise the cost and difficulty to prepare the biosensors.

2.4.4 Operation

The operation denotes the procedures required to perform the assay, e.g., washing and drying. An ideal CMOS biosensor should involve only a sample loading step to detect the target. For small biomolecule sensing, NMR-based biosensor stands out as a tantalizing solution since it does not encompass any washing step after the mixing of the probe and sample. Still, for cellular detection, a washing step is included to rinse off unbound MNP. On the other hand, electrical-based detection has a facile operation procedure for diagnosis. Despite certain works involving a washing step after hybridization to remove the unbound target on the surface, the change in electri-cal properties upon hybridization can be detected without further washing steps. For fluorescent- and magnetic-based biosensors, they encompass a washing step to


34

remove the unbound molecule, similar to the conventional ELISA. Nevertheless, they are promising for cellular sensing since the washing step is omitted. The mechanical-based biosensor is confined by the operation procedures since it entails both washing and drying steps to rinse off the unbound molecules and drying the surface of the cantilever. These steps are time-consuming and require considerable labors.

2.4.5 Specificity

The archetypal and ultimate goal of CMOS biosensor is to detect the desired biomol-ecule from the sample. Yet, for complex biological media (e.g., blood plasma, serum, etc.), the sample matrix effect will incur nonspecific binding and confound the detec-tion limit of the system. For instance, the albumin, the most abundant protein inside the blood plasma, may exist with a concentration of 600 μM (i.e. ~1 billion times above the desired detection limit) [72]. The nonspecific bindings of albumin with the probe lead to a background biological noise floor and create a false positive on the output of the diagnostic tool. Additionally, other physical parameters such as tem-perature or pH of the sample may affect the diagnosis. While these interferences can possibly be suppressed by hardware techniques such as differential detection (i.e., compare the experimental result from a reference result), the additional measurement increases the hardware cost and the sample consumption. The optical-based and magnetic-based detections are transcendent in this perspective attributed to the label-ing process and multistep protocols. Any undesired signaling tag will be removed from the sensing region after the rinsing step. In contrast, the electrical-based detec-tion is prone to the sample matrix effect due to the nonspecific absorption of other biomolecules [73, 74]. Especially for EIS, not only is the impedance affected by the nonspecific absorption from the sample matrix but also the constitution of the media (i.e., the conductivity of the media and permittivity of the cells). This poses a detri-mental effect on the measurement. The mechanical-based detector is sensitive to nonspecific binding of the biomolecule and ambient temperature interference. Yet, the washing and drying steps soothe the influence from nonspecific binding. Finally, the NMR-based detection, similar to magnetic-based detection, shows high specific-ity ascribed to the adoption of surface functionalized MNPs. Yet, the spin–spin relax-ation time of the sample also correlates with the viscosity as well as the composition and state of the media (i.e., oxygenation of the blood sample [75]) and the concentra-tion of the MNPs. This stiffens the direct measurement from complex media.

2.4.6 Summary

The electrical-based transducing mechanism is generally the most widespread and favorable detection scheme for CMOS IVD tools since it can provide label-free biomolecule detection and cellular monitoring without bulky external components such as optical filters or permanent magnet. Further, the operation easiness (i.e.,


35

without washing and drying steps) promises fast assays and befits the PoC applica-tions. Yet, the hardware preparation on CMOS chip and necessity for direct contacts pose certain limits for electrical-based detection, especially when interfacing with other microfluidic networks [76]. Further, the external influence requires delicate systematic design to achieve better specificity. Mechanical-based biomolecule detection, constrained by its complexity on hardware preparation and experimental procedure, shows limited exposures for CMOS IVD tools.

In contrast, despite the labeling step and non-integrated gadgets, optical-based detection is still popular for CMOS IVD tools since the conventional laboratory pro-tocols such as ELISA and qPCR can be transferred smoothly to the CMOS chip. This increases the consistency of the experimental results of the CMOS IVD tools to the centralized benchtop assay. Its capability of detecting light transmittance also renders it a promising scheme for cellular assay. Alternatively, magnetic-based detection is similar to optical-based detection for biomolecule targeting, whereas both require labeling and several mixing and washing steps. The lack of magnetic substance inside the biological sample enables sensitive and specific biomolecule quantification for magnetic-based detection. Further, the high specificity of the labeling schemes for optical-based and magnetic-based detections guarantees the robustness of the sys-tems against matrix effects from complex biological media. NMR-based detection is a promising solution for CMOS IVD tools when the preparations of the CMOS chip and sample beforehand have to be minimized. Also, the contactless property of NMR facilitates the integration of the NMR electronics with the sample management net-works [59]. Yet, the relatively weak NMR signal limits the sensitivity of NMR-based detection and entails delicate design on the readout circuits. Further, the permanent magnet hinders the miniaturization of the overall platform.

A conceptual radar chart is plotted in Fig. 2.14, showing, in general, the overall strengths and weaknesses of the above characteristics of the biosensors using dis-tinct transducing mechanisms. Obviously, there are many applications for in vitro

Labelingeasiness

Labelingeasiness

Labelingeasiness

Labelingeasiness

Labelingeasiness

Hardwarepreparationeasiness





Operationeasiness

Operationeasiness

Operationeasiness

Operationeasiness

Operationeasiness

Integrability

Integrability

Integrability Integrability

Integrability

Selectivity/Specificity





Electrical Optical Magnetic

Mechanical NMR

Fig. 2.14 A radar chart showing the conceptual requisites to perform the in vitro diagnosis on biomolecule targeting with different transducing mechanism


36

diagnosis, especially for cellular level study (cell counting, monitoring, or flow cytometry); thus Fig. 2.14 is evaluated mainly on their performance for biomolecule (DNA and protein) targeting. Also, there are exceptional cases. For instance, Peng et al. reported an NMR-based label-free platform (with discrete electronics) for detecting malaria in whole blood by detecting the variation on spin–spin relaxation time of NMR signals from paramagnetic hemozoin particles [77].

References

1. J. Schmitz, Adding functionality to microchips by wafer post-processing. Nucl. Instrum. Methods Phys. Res., Sect. A 576(1), 142–149 (2007)

2. C.G. Jakobson, U. Dinnar, M. Feinsod, Y. Nemirovsky, Ion-sensitive field-effect transistors in standard CMOS fabricated by post processing. IEEE Sensors J. 2(4), 279–287 (2002)

3. A.H.D. Graham, S.M. Surguy, P. Langlois, C.R. Bowen, J. Taylor, J. Robbins, Modification of standard CMOS technology for cell-based biosensors. Biosens. Bioelectron. 31(1), 458–462 (2012)

4. J.M. Rothberg, W. Hinz, T.M. Rearick, J. Schultz, W. Mileski, M. Davey, et al., An integrated semiconductor device enabling non-optical genome sequencing. Nature 475(7356), 348–352 (2011)

5. A. Gao, N. Lu, Y. Wang, T. Li, Robust ultrasensitive tunneling-FET biosensor for point-of-care diagnostics. Sci Rep 6, 22554 (2016)

6. J. Lee, J. Jang, B. Choi, J. Yoon, J.-Y. Kim, Y.-K. Choi, et al., A highly responsive silicon nanowire/amplifier MOSFET hybrid biosensor. Sci Rep 5, 12286 (2015)

7. C. Stagni, C. Guiducci, L. Benini, B. Ricco, S. Carrara, B. Samori, et al., CMOS DNA sen-sor array with integrated A/D conversion based on label-free capacitance measurement. IEEE J. Solid State Circuits 41(12), 2956–2964 (2006)

8. M. Barbaro, A. Bonfiglio, L. Raffo, A. Alessandrini, P. Facci, I. Barák, Fully electronic DNA hybridization detection by a standard CMOS biochip. Sens. Actuators B 118(1–2), 41–46 (2006)

9. S.J. Han, H. Yu, B. Murmann, N. Pourmand, S.X. Wang, A high-density magnetoresistive bio-sensor array with drift-compensation mechanism, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2007, pp. 168–169

10. E. Anderson, J. Daniels, H. Yu, T. Lee, N. Pourmand, A label-free CMOS DNA microarray based on charge sensing, in Proceedings of International Instrumentation and Measurement Technology Conference, 2008, pp. 1631–1636

11. B. Jang, P. Cao, A. Chevalier, A. Ellington, A. Hassibi, A CMOS fluorescent-based biosen-sor microarray, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2009, pp. 436–437

12. T.C.D. Huang, S. Sorgenfrei, P. Gong, R. Levicky, K.L. Shepard, A 0.18-μm CMOS array sensor for integrated time-resolved fluorescence detection. IEEE J. Solid State Circuits 44(5), 1644–1654 (2009)

13. W. Hua, C. Yan, A. Hassibi, A. Scherer, A. Hajimiri, A frequency-shift CMOS magnetic bio-sensor array with single-bead sensitivity and no external magnet, in IEEE International Solid- State Circuits Conference (ISSCC) Digest of Technical Papers, 2009, pp. 438–439

14. P.M. Levine, P. Gong, R. Levicky, K.L. Shepard, Real-time, multiplexed electrochemical DNA detection using an active complementary metal-oxide-semiconductor biosensor array with integrated sensor electronics. Biosens. Bioelectron. 24(7), 1995–2001 (2009)

15. A. Manickam, A. Chevalier, M. McDermott, A.D. Ellington, A. Hassibi, A CMOS electro-chemical impedance spectroscopy (EIS) biosensor array. IEEE Trans. Biomed. Circuits Syst. 4(6), 379–390 (2010)


37

16. H. Jafari, L. Soleymani, R. Genov, 16-channel CMOS impedance spectroscopy DNA analyzer with dual-slope multiplying ADCs. IEEE Trans. Biomed. Circuits Syst. 6(5), 468–478 (2012)

17. K.H. Lee, S. Choi, J.O. Lee, J.B. Yoon, G.H. Cho, CMOS capacitive biosensor with enhanced sensitivity for label-free DNA detection, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2012, pp. 120–122

18. M. Barbaro, A. Caboni, D. Loi, S. Lai, A. Homsy, P.D. van der Wal, et al., Label-free, direct DNA detection by means of a standard CMOS electronic chip. Sens. Actuators B 171–172, 148–154 (2012)

19. Y.J. Huang, C.W. Huang, T.H. Lin, C.T. Lin, L.G. Chen, P.Y. Hsiao, et al., A CMOS cantilever- based label-free DNA SoC with improved sensitivity for hepatitis B virus detection. IEEE Trans. Biomed. Circuits Syst. 7(6), 820–831 (2013)

20. C. Toumazou, L.M. Shepherd, S.C. Reed, G.I. Chen, A. Patel, D.M. Garner, et al., Simultaneous DNA amplification and detection using a pH-sensing semiconductor system. Nat. Methods 10(7), 641–646 (2013)

21. C.-W. Huang, Y.-J. Huang, P.-W. Yen, H.-H. Tsai, H.-H. Liao, Y.-Z. Juang, et al., A CMOS wireless biomolecular sensing system-on-chip based on polysilicon nanowire technology. Lab Chip 13(22), 4451–4459 (2013)

22. A. Pai, A. Khachaturian, S. Chapman, A. Hu, H. Wang, A. Hajimiri, A handheld magnetic sensing platform for antigen and nucleic acid detection. Analyst 139(6), 1403–1411 (2014)

23. H.M. Jafari, K. Abdelhalim, L. Soleymani, E.H. Sargent, S.O. Kelley, R. Genov, Nanostructured CMOS wireless ultra-wideband label-free PCR-free DNA analysis SoC. IEEE J. Solid State Circuits 49(5), 1223–1241 (2014)

24. H. Norian, R.M. Field, I. Kymissis, K.L. Shepard, An integrated CMOS quantitative- polymerase- chain-reaction lab-on-chip for point-of-care diagnostics. Lab Chip 14(20), 4076–4084 (2014)

25. C.H. Chen, R.Z. Hwang, L.S. Huang, S.M. Lin, H.C. Chen, Y.C. Yang, et al., A wireless bio- MEMS sensor for C-reactive protein detection based on nanomechanics. IEEE Trans. Biomed. Eng. 56(2), 462–470 (2009)


27. O. Tigli, L. Bivona, P. Berg, M.E. Zaghloul, Fabrication and characterization of a surface- acoustic- wave biosensor in CMOS technology for cancer biomarker detection. IEEE Trans. Biomed. Circuits Syst. 4(1), 62–73 (2010)

28. N. Sun, T.J. Yoon, H. Lee, W. Andress, R. Weissleder, D. Ham, Palm NMR and 1-chip NMR. IEEE J. Solid State Circuits 46(1), 342–352 (2011)

29. S. Gambini, K. Skucha, P.P. Liu, J. Kim, R. Krigel, A 10 kPixel CMOS hall sensor array with baseline suppression and parallel readout for immunoassays. IEEE J. Solid State Circuits 48(1), 302–317 (2013)

30. D.A. Hall, R.S. Gaster, K.A.A. Makinwa, S.X. Wang, B. Murmann, A 256 pixel magnetoresistive biosensor microarray in 0.18μm CMOS. IEEE J. Solid State Circuits 48(5), 1290–1301 (2013)

31. L. Sandeau, C. Vuillaume, S. Contie, E. Grinenval, F. Belloni, H. Rigneault, et al., Large area CMOS bio-pixel array for compact high sensitive multiplex biosensing. Lab Chip 15(3), 877–881 (2015)

32. C. Sapsanis, S. Sivashankar, H. Omran, U. Buttner, K.N. Salama, Capacitive immunosen-sor for C-reactive protein quantification, in Proceedings of the ICEE International Midwest Symposium on Circuits and Systems, 2015, pp. 1–4

33. L.Y. Hong, S. McManus, H. Yang, K. Sengupta, A fully integrated CMOS fluorescence bio-sensor with on-chip nanophotonic filter, in Proceedings of Symposium on VLSI Circuits, 2015, pp. C206–C207

34. P.H. Kuo, J.C. Kuo, H.T. Hsueh, J.Y. Hsieh, Y.C. Huang, T. Wang, et al., A smart CMOS assay SoC for rapid blood screening test of risk prediction. IEEE Trans. Biomed. Circuits Syst. 9(6), 790–800 (2015)

35. H. Klapproth, S. Bednar, J. Baader, M. Lehmann, I. Freund, T. Brandstetter, et al., Development of a multi-analyte CMOS sensor for point-of-care testing. Sens. Bio-Sens. Res. 5, 117–122 (2015)

References

38

36. Y. Zheng, N. Shang, P.S. Haddad, M. Sawan, A microsystem for magnetic immunoassay based on planar microcoil array. IEEE Trans. Biomed. Circuits Syst. 10(2), 477–486 (2016)

37. S.B. Prakash, P. Abshire, On-chip capacitance sensing for cell monitoring applications. IEEE Sensors J. 7(3–4), 440–447 (2007)

38. S.B. Prakash, P. Abshire, Tracking cancer cell proliferation on a CMOS capacitance sensor chip. Biosens. Bioelectron. 23(10), 1449–1457 (2008)

39. E.P. Dupont, E. Labonne, Y. Maruyama, C. Vandevyver, U. Lehmann, M.A.M. Gijs, et al., Fluorescent magnetic bead and cell differentiation/counting using a CMOS SPAD matrix. Sens. Actuators B 174, 609–615 (2012)

40. Y. Chen, C.C. Wong, T.S. Pui, R. Nadipalli, R. Weerasekera, J. Chandran, et al., CMOS high density electrical impedance biosensor array for tumor cell detection. Sens. Actuators B 173, 903–907 (2012)

41. K.H. Lee, J. Nam, S. Choi, H. Lim, S. Shin, G.H. Cho, A CMOS impedance cytometer for 3D flowing single-cell real-time analysis with ΔΣ error correction, in IEEE International Solid- State Circuits Conference (ISSCC) Digest of Technical Papers, 2012, pp. 304–306

42. H. Wang, A. Mahdavi, D.A. Tirrell, A. Hajimiri, A magnetic cell-based sensor. Lab Chip 12(21), 4465–4471 (2012)

43. T. Saeki, M. Hosokawa, T. Lim, M. Harada, T. Matsunaga, T. Tanaka, Digital cell counting device integrated with a single-cell array. PLoS One 9(2), e89011 (2014)

44. P. Murali, I. Izyumin, D. Cohen, J.C. Chien, A.M. Niknejad, B. Boser, A CMOS micro-flow cytometer for magnetic label detection and classification, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2014, pp. 422–423

45. M. Roy, G. Jin, D. Seo, M.-H. Nam, S. Seo, A simple and low-cost device performing blood cell counting based on lens-free shadow imaging technique. Sens. Actuators B 201, 321–328 (2014)

46. K. Niitsu, S. Ota, K. Gamo, H. Kondo, M. Hori, K. Nakazato, Development of microelectrode arrays using electroless plating for CMOS-based direct counting of bacterial and HeLa cells. IEEE Trans. Biomed. Circuits Syst. 9(5), 607–619 (2015)

47. C. Laborde, C. Pittino, H.A. Verhoeven, S.G. Lemay, L. Selmi, M.A. Jongsma, et al., Real- time imaging of microparticles and living cells with CMOS nanocapacitor arrays. Nat. Nanotechnol. 10(9), 791–795 (2015)

48. T. Chi, J.S. Park, J.C. Butts, T.A. Hookway, A. Su, C. Zhu, et al., A multi-modality CMOS sensor array for cell-based assay and drug screening. IEEE Trans. Biomed. Circuits Syst. 9(6), 801–814 (2015)

49. K.T. Chang, Y.J. Chang, C.L. Chen, Y.N. Wang, Multichannel lens-free CMOS sensors for real-time monitoring of cell growth. Electrophoresis 36(3), 413–419 (2015)

50. J.C. Love, L.A. Estroff, J.K. Kriebel, R.G. Nuzzo, G.M. Whitesides, Self-assembled monolay-ers of thiolates on metals as a form of nanotechnology. Chem. Rev. 105(4), 1103–1169 (2005)

51. A. Manickam, C.A. Johnson, S. Kavusi, A. Hassibi, Interface design for CMOS-integrated electrochemical impedance spectroscopy (EIS) biosensors. Sensors 12(11), 14467 (2012)

52. C. Berggren, P. Stalhandske, J. Brundell, G. Johansson, A feasibility study of a capacitive biosensor for direct detection of DNA hybridization. Electroanalysis 11(3), 156–160 (1999)

53. J. Enderlein, T. Ruckstuhl, S. Seeger, Highly efficient optical detection of surface-generated fluorescence. Appl. Opt. 38(4), 724–732 (1999)

54. M. Alvarez, L.M. Lechuga, Microcantilever-based platforms as biosensing tools. Analyst 135(5), 827–836 (2010)

55. H. Wohltjen, R. Dessy, Surface acoustic wave probe for chemical analysis. I. Introduction and instrument description. Anal. Chem. 51(9), 1458–1464 (1979)


57. L. Josephson, J.M. Perez, R. Weissleder, Magnetic nanosensors for the detection of oligo-nucleotide sequences. Angew. Chem. 113(17), 3304–3306 (2001)

58. J.M. Perez, L. Josephson, T. O’Loughlin, D. Hogemann, R. Weissleder, Magnetic relaxation switches capable of sensing molecular interactions. Nat. Biotechnol. 20(8), 816–820 (2002)



39

60. I. Koh, R. Hong, R. Weissleder, L. Josephson, Sensitive NMR sensors detect antibodies to influenza. Angew. Chem. 47(22), 4119–4121 (2008)

61. D. Issadore, C. Min, M. Liong, J. Chung, R. Weissleder, H. Lee, Miniature magnetic resonance system for point-of-care diagnostics. Lab Chip 11(13), 2282–2287 (2011)

62. M. Liong, A.N. Hoang, J. Chung, N. Gural, C.B. Ford, C. Min, et al., Magnetic barcode assay for genetic detection of pathogens. Nat. Commun. 4(1752), 1–9 (2013)

63. C.M. Castro, A.A. Ghazani, J. Chung, H.L. Shao, D. Issadore, T.J. Yoon, et al., Miniaturized nuclear magnetic resonance platform for detection and profiling of circulating tumor cells. Lab Chip 14(1), 14–23 (2014)

64. E. Engvall, P. Perlmann, Enzyme-linked immunosorbent assay (ELISA) quantitative assay of immunoglobulin G. Immunochemistry 8(9), 871–874 (1971)

65. R.W. Peeling, H. Artsob, J.L. Pelegrino, P. Buchy, M.J. Cardosa, S. Devi, et al., Evaluation of diagnostic tests: dengue. Nat. Rev. Microbiol. 8, S30–S37 (2010)

66. N. Scholler, M. Crawford, A. Sato, C.W. Drescher, K.C. O’Briant, N. Kiviat, et al., Bead- based ELISA for validation of ovarian cancer early detection markers. Clin. Cancer Res. 12(7), 2117–2124 (2006)

67. S. Velumani, H.-T. Ho, F. He, S. Musthaq, M. Prabakaran, J. Kwang, A novel peptide ELISA for universal detection of antibodies to human H5N1 influenza viruses. PLoS One 6(6), e20737 (2011)

68. C. Kandoth, M.D. McLellan, F. Vandin, K. Ye, B. Niu, C. Lu, et al., Mutational landscape and significance across 12 major cancer types. Nature 502(7471), 333–339 (2013)

69. E.A. Ottesen, J.W. Hong, S.R. Quake, J.R. Leadbetter, Microfluidic digital PCR enables mul-tigene analysis of individual environmental bacteria. Science 314(5804), 1464–1467 (2006)

70. M. Schena, D. Shalon, R.W. Davis, P.O. Brown, Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270(5235), 467–470 (1995)

71. H. Shafiee, S. Wang, F. Inci, M. Toy, T.J. Henrich, D.R. Kuritzkes, et al., Emerging technolo-gies for point-of-care management of HIV infection. Annu. Rev. Med. 66(1), 387–405 (2015)

72. J.L. Arlett, E.B. Myers, M.L. Roukes, Comparative advantages of mechanical biosensors. Nat. Nanotechnol. 6(4), 203–215 (2011)

73. T. Bryan, X. Luo, P.R. Bueno, J.J. Davis, An optimised electrochemical biosensor for the label- free detection of C-reactive protein in blood. Biosens. Bioelectron. 39(1), 94–98 (2013)

74. J.T. Kirk, N.D. Brault, T. Baehr-Jones, M. Hochberg, S. Jiang, D.M. Ratner, Zwitterionic polymer-modified silicon microring resonators for label-free biosensing in undiluted human-plasma. Biosens. Bioelectron. 42, 100–105 (2013)

75. K.R. Thulborn, J.C. Waterton, P.M. Matthews, G.K. Radda, Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biochim. Biophys. Acta 714(2), 265–270 (1982)

76. E. Ghafar-Zadeh, M. Sawan, A hybrid microfluidic/CMOS capacitive sensor dedicated to lab- on- chip applications. IEEE Trans. Biomed. Circuits Syst. 1, 270–277 (2007)

77. W.K. Peng, T.F. Kong, C.S. Ng, L. Chen, Y. Huang, A.A.S. Bhagat, et al., Micromagnetic reso-nance relaxometry for rapid label-free malaria diagnosis. Nat. Med. 20(9), 1069–1073 (2014)

References


Chapter 3Electronic-Automated Micro-NMR Assay with DMF Device

3.1 Introduction

The miniaturization of the NMR system enhances its portability and facilitates the application outside the laboratory. Regrettably, it is inflexible to pipeline multiple samples to the micro-NMR sensing region for higher throughput and real-time result comparison (e.g., concentration of the analytes) due to the small inner space and limited NMR sensing region of the portable magnet (0.46 T, 1.25 kg). The operation of tiny samples beforehand, which can involve multistep multisite treat-ments, relies heavily on the human efforts, degrading the throughput and consis-tency of diagnostic results while raising the chance of sample contamination. To address this issue, certain efforts have been undertaken to facilitate sample manipu-lation in NMR systems like capillary electrophoresis [1] and microfluidic channels [2, 3]. Still, these methods involve several laboratory accessories (e.g., pumps and pressure generators) and fixed fluidic paths/pipes that have low portability and reconfigurability and are inadequate for PoC application.

Unlike conventional channel microfluidics, digital microfluidics (DMF) is highly amenable to co-integration, electronic automation, and reconfiguration. It has gained tremendous research attention recently [4–10]. This biocompatible platform has been adopted in a wide variety of biological applications, including cell cultur-ing [4, 11, 12], DNA amplification [13–15], and single protein molecule capturing [16]. Microdroplets (e.g., <10 μL) in the DMF device can be transported over an electrode array by modifying the surface tension of the electrode utilizing the prin-ciple of electrowetting-on-dielectric (EWOD). Such distinct microdroplet control-lability renders the DMF a promising droplet management platform for PoC devices, particularly for co-integration with NMR. In addition, as the DMF device is planar, all droplets can be preloaded in the device before routinely executing the reaction or screening, enhancing the consistency of the experiments.

42

In this respect, this chapter proposes the design of a micro-NMR platform for chemical/biological assay. It equips with a DMF device for electronic-automated sample management. We carried out this project in two phases. First, we firstly per-formed a primary investigation of the NMR and DMF to study the characteristics and explore the difficulties with discrete electronics (Sect. 3.2). Afterward, the NMR electronics are integrated on a single CMOS chip to extend the capability and enhance the performance, with the thorough analysis of the TRX for relaxometry application (Sect. 3.3).

3.2 First Prototype: Primary Investigation on NMR–DMF

Figure 3.1a shows the overview of the NMR–DMF system. It was designed to drive the droplets under detection and the target-specific probes, if any, to the desired location for NMR assays. The movements of the droplets are handled by an elec-trode array, which was fabricated on the glass substrate. The RF coil functions as a transducer to transform the magnetic field to voltage (or vice versa). The electronics transmit the excitation signal to the RF coil and receive the NMR signal from it. In this prototype, an oscilloscope is responsible for collecting the NMR results of the samples. A software algorithm derives the relaxation times of the NMR signals. The DMF platform and NMR coils are customized in size to befit the limited inner vol-ume of the magnet. The design considerations of the system are described in detail as follows.

Fig. 3.1 The overall schematic and operations of the NMR–DMF system. (a) The placement of the DMF device, magnet, RF coil, and PCB in 3D view; (b) schematic of the NMR electronics; (c) the filtered results from the PCB are captured by the oscilloscope for easier demonstration purpose and then analyzed in MATLAB; (d) the photograph of the DMF device and its structure; (e) the detection mechanism of the NMR–DMF system. The target-specific MNPs, which act as probes, are placed on the sensing site initially (in purple). The sample at other electrodes (in cyan) will be transported to the sensing site and mixed with the probe to perform NMR assay

3 Electronic-Automated Micro-NMR Assay with DMF Device

43

3.2.1 Discrete Electronics and Back-End Signal Processing

The electronics of the NMR–DMF system were mainly built by discrete compo-nents for fast prototyping and greater flexibility to integrate with the DMF electron-ics. Figure 3.1b depicts the schematic of the electronics, which mainly consist of two paths: transmitter (TX) and receiver (RX). Before the nuclear spins can induce signal to the coil, they have to be excited by the coil at fL. CPMG pulse sequences, as shown in Fig. 3.2, are entailed to excite the nuclei and refocus the spin from B0- field inhomogeneity. Both in-phase (I) and quadrature (Q) waveforms are collected for image rejection to circumvent from additional 3-dB noise figure degradation. By applying frequency division, a pair of equal-pulse-width I and Q signals can be generated under a clock-signal frequency four times that of fL. To prevent the SNR of the system from degrading by the 1/f noise of the electronics, the clock frequency was chosen at 4x (fL + fIF). Although the excitation frequency (fEXC) now shifts to fL + fIF instead of fL, the nuclei can still be excited if fIF is small enough, which also facilitates the design of the electronics [17]. The excitation signal here also serves as the LO signal for the mixer. This scheme allows the RX and TX to share the LO circuitry and ease the synchronization of their signals. Comparing with the case of feeding an external LO at 2x (fL + fIF), the use of 4x (fL + fIF) secures an accurate phase over the process, voltage, and temperature variations.

Since the RF coil serves both the TX and RX, switches have to be used to isolate the excitation signals from leaking to the RX. For the TX, output buffers were uti-lized to boost up the driving capability. A field-programmable gate array (FPGA) controls the operating phases of the switches and buffers. The RX amplifies the weak NMR signal coupling from the RF coils. The weak amplitude of the induced NMR signal is at a level of 100 nV–40 μV [18]. The first amplification is based on

TX Pulses

RX Output

Single p /2 pulsewith F= 0∞

Multiples re-phasingp-pulse with F= 90∞

T2

Excitation pulse Refocused echoes from the atoms

Excitation

Relaxation

Fig. 3.2 Timing diagram of the pulses, including the excitation CPMG pulse sequence delivered to the TX to excite the nuclei and the response from the nuclei, which is picked up by the coil


44

an LC tank to provide a passive gain of Q2 1+ to the signal at its resonant fre-quency 1 2/ ≠ LC , where Q is the quality factor of the RF coil and L and C are its inductance and capacitance, respectively [17–19]. Hence, this gain will suppress the noise figure of the RX, and the overall SNR will be increased according to (1.3). This scheme is feasible attributed to the narrowband behavior of the NMR signals (<5 kHz for this platform). The input impedance of the forefront amplifier should be adequately large to prevent its loading effect from the LC tank, which otherwise deteriorates the sensitivity.

After signal amplification, the signal is driven to the I and Q mixers for downcon-version to fIF for baseband signal processing. Appendix A describes the details about the discrete electronic components. The resulting signals are then low-pass filtered and further amplified before driving into a digital oscilloscope (Fig. 3.1c) to reduce out-of-band noise and high-frequency mixing product. According to (1.3), the band-width (BW) of the filter should be set carefully to prevent excessive out-of-band noise. Finally, the signals are collected from the oscilloscope for back-end process-ing such as I/Q demodulation and T2 derivation. The echoes’ amplitudes decay exponentially, and an algorithm written in MATLAB was built to derive and fit the result to the exponential curve, as depicted in (1.2).

3.2.2 Magnet

The portable permanent magnet is responsible for magnetizing the nuclei of the atoms. From (1.1 and 1.3) the SNR of the system is proportional to the power of 7/4 of B0. Although there seem to be numerous ways to enhance the B0 (i.e., for higher resolution and lower noise), the portability and power consumption of the system will be penalized due to the need for a heavier and bulkier magnet, not to mention a higher operating frequency that will be required for the electronics. To balance the performance of the system with its portability, a 0.5-T permanent magnet (PM-1055 from Metrolab, Switzerland) was chosen. It is 1.25 kg in weight and 1005 cm3 in volume, as shown in Fig. 3.3a. The corresponding fL of the hydrogen nucleus under this magnet strength is ~21 MHz.

3.2.3 RF Coils

The RF coil is the interface between the droplet samples and electronics. It trans-duces between the B1-field produced by the nuclei and the voltage for the TRX. There are different kinds of coils such as saddle-shaped coil [20], solenoid [17, 18, 20], and planar coil [17, 19, 21]. The saddle-shaped coil and solenoid either required hand-wrapping or extra fabrication process. In fact, the planar coil (this work) on a


https://doi.org/10.1007/978-3-319-67825-2_BM1

45

low-cost, two-side printed circuit board (PCB) is appealing for its consistency of parameters and disposability under volume production.

Typically, the planar coil is a circular spiral as shown in Fig. 3.3b. The dominant magnetic field of this coil is in its axial direction (z-direction). This spiral coil is common due to its high sensitivity. However, since B1-field should be orthogonal to B0-field (z-direction) for NMR experiment, B1-field has to be in either the x- or the y-direction. For solenoids and spiral coils, the circular planes of the coils need to be in the x-z plane in order to generate B1-field in the y-direction. This restricts the usable space and thus the number of DMF electrodes inside the magnet since the width is 2.3x longer than the height. To resolve this, a PCB Butterfly-coil containing two square spiral loops connected in series with different rotation (i.e., clockwise and counter-clockwise) entails the generation of the plane-parallel B1-field (x- direction) (Fig. 3.3c). Routed with square loops, the Butterfly-coil can effectively concentrate the magnetic field between the centers of the two loops (~40% stronger than its circular loop counterpart). By inserting this Butterfly-coil in the portable magnet (x-y plane), B1-field travels orthogonally to B0-field, easing the integration of the DMF device with a higher number of electrodes inside the portable magnet. Yet, as the amplitude of the micro-NMR signals is commensurate with B1, the sys-tem will have a lower SNR ascribed to the lower B1 of the Butterfly-coil when com-pared with the spiral counterpart (0.5x). Nevertheless, it has been proven that the Butterfly-coil is less susceptible to environmental couplings (e.g., powerline cables, equipment, and RF interference) since they appear as a common-mode noise [22–24].

Fig. 3.3 (a) Geometry and limitation from the opening gap of the portable magnet. (b, c) The EM simulation of the magnetic field direction and strength from a spiral coil (with 14 turns) and a Butterfly-coil (with 7 turns on each spiral), respectively, with a flowing current of 1 A


46

3.2.4 DMF Device Fabrication and Actuation

DMF is an LOC technology based on the principle of EWOD. Droplets inside the DMF device are manipulated via surface-tension modulation induced by an electric field, enabling electronic-automated chemical/biological reactions with low sample volumes. When compared with the conventional channel microfluidic devices, DMF can avoid separate and complex networks of connections, laboratory gadgets such as pumps and valves and the device is reconfigurable in such a way that the droplets can move freely over a surface of electrode matrix.

The rudiment of DMF is to change the contact angle between the droplets and substrate by electric fields, where the relationship can be expressed as [25]:

cos coslg

θ θγFr

tV( ) − ( ) =0

0 2

2

(3.1)

with the final contact angle θF, initial contact angle (i.e., without driving voltage) θ0, relative dielectric permittivity of the liquid ϵr, permittivity of free space ϵ0, liquid–gas surface tension γlg, and thickness of the dielectric layer and the driving voltage V. As depicted in (3.1), the contact angle of the droplet is altered according to V. By creating an unbalance contact angle on the droplet, there exists a force causing the movement of the droplets, and thus the droplets can be controlled electronically by driving the electrodes with a desired voltage signal.

Figure 3.1d shows the structure of the DMF device, which is composed of an electrode array. The one with chromium plating is patterned by lithography and wet etching to achieve customized electrodes array, followed by Ta2O5 and parylene-C deposition to enhance the EWOD force. Indium tin oxide (ITO) coated on another glass plate (thickness, 0.5 mm) serves as a ground plane for all of the electrodes. A hydrophobic Teflon® layer covers both plates. To integrate the NMR and DMF, the DMF device is sized for inserting into the magnet for magnetizing the samples, and the coil is located over the sample for high-sensitivity sensing. However, as the electrode (chromium) and ITO are made by conducting materials, the Q of the coil degrades when the DMF device is placed near the coil, due to the eddy current (EC) formed in the conductors. The energy loss per cycle in the conductors by EC (when the skin effect does not take place) is given as [26]:

W

dfBAdloss =

( )πρ

2

2

6 (3.2)

with magnetic field generated by the RF coil on the conductor B2, thickness of the conductor d, cross-section area of the conductor A, frequency of the magnetic field f, and the resistivity of the conductor ρ. By the law of energy conservation, the energy losses in the conductors are generated from the RF coil, causing Q degrada-tion of the coil. To maintain Q of the coil, the EC loss in the nearby conductor must be minimized. The EC loss in chromium is negligible in most cases, as it has sheet


47

resistance >100 Ω sq−1. However, for ITO, its sheet resistance can vary from 5 to 500 Ω sq−1. ITO with higher sheet resistance may lead to slower droplet motion, but it will decrease the EC loss. Moreover, the SNR of the NMR signal will be reduced with diminishing B1, and the EC generated by the conductor will diminish the effec-tive B1. From (3.2), the thickness of the ITO should be optimized to reduce the EC loss as it will rise proportionally to the cube of d. Figure 3.4 shows the simulated EC loss on the ITO generated by a unit current passing through the Butterfly-coil against the ITO thickness. The EC loss should be suppressed to less than 0.5% of the mag-netic energy generated by the coil, which was marked on Fig. 3.4. The thickness is limited to 80 nm, which is equivalent to a sheet resistance of 12.5 Ω sq−1. ITO with sheet resistance of 100 Ω sq−1 was chosen for the system. Moreover, to ensure the droplet movement and reduce the required driving voltage, silicone oil is injected to fill the gap within the DMF device.

The driving signal of each electrode is a 50% duty cycle square wave, which is generated from the signal generator and then amplified by a step-up transformer. To avoid the overstress conditions, the driving signals are clipped by diode protection scheme to limit its amplitude below 40 V. To move a droplet to the desired elec-trode, the driving signal will be connected to the correlated pads, while the neigh-boring electrode will be grounded to prevent excessive charges stored in the electrodes that otherwise might cause dielectric breakdown. A square wave with peak-to-peak voltage 40 V and frequency 1 kHz drives the electrodes. The size of the electrodes is 3.5 × 3.5 mm2, and the gap between the top and bottom planes is 0.45 mm.

3.2.5 Experimental Results

Electrical Measurements

The magnet’s field strength was first measured to approximately locate the fL of the hydrogen nuclei. The B0 at the center of the magnet is around 0.4615 T at 20 °C, which corresponds to an fL of 19.65 MHz.

ITO Thickness (nm)0 100 200 300R

atio

of

Ed

dy

curr

ent

Lo

ssto

Co

il M

agn

etic

En

erg

y

0.5%

100

10-1

10-2

10-3

10-4

10-5

Fig. 3.4 Ratio of EC loss generated by the seven- turn (each loop) Butterfly- coil to coil magnetic energy against the thickness of the ITO. The figure was plotted based on (3.2) with f = 20 MHz, ρ = 1 × 10−6 Ωm, and A = 40 mm × 24 mm. The dotted line shows 0.5% level and corresponds to the ITO thickness of 80 nm


48

The fabricated Butterfly-coils were then characterized in both simulation and measurement. The fabricated coils were measured by an impedance analyzer and compared with the simulations in COMSOL® with a three-dimensional (3D) model to confirm the parameters. Table 3.1 shows two sets of parameters for the coil with seven and nine turns spiral loop. The difference between the simulation and mea-surement results is sufficiently small (<7%). Such error could be originated from the leads of the coils in the measurements, the coarse meshing in 3D simulation, and the thickness deviation of PCB copper traces. Nonetheless, the accuracy is adequate here as only the trend and B1-field direction are decisive. This also offers a system-atic study of the RF coils when compared with the solenoid and saddle-shaped coils. Another advantage of the PCB coil is the reproducibility. From Table 3.1, the repro-ducibility of the PCB coils was adequately high as the standard deviations are <3% of the nominal values. This feature makes PCB coil attractive, as the fabrication only relies on machines, minimizing the variation between the coils.

The measured gain of the RX is 95.7 dB within 5 kHz of IF. The output signal of the system with a 100 nV sinusoidal input is 30 dB above the noise floor, and thus it can detect a signal amplitude down to 100 nV. Appendix A reveals the data about the electronic measurement.

NMR Systems

The pulse width of the Butterfly-coil coil was determined by varying the RF excita-tion pulse duration and observing the corresponding NMR signal. Figure 3.5 shows the plot of such a nutation curve. The π/2-pulse for the seven-turn Butterfly-coil is estimated as 144 μs.

Before the integration of the NMR–DMF system, the NMR part was tested sepa-rately to verify its own functionality. According to the study, CuSO4 will affect the T2 of water [27, 28]. We prepared CuSO4 with different concentrations for the exper-iment. A seven-turn Butterfly-coil was selected for the NMR system. Figure 3.6a shows the received NMR signal of water. T2 derived by the algorithm is 343.6 ms. The glitches appearing between the echoes correspond to the excitation signal. They can be prevented by adding switches at the RX front-end. Yet, since this will con-tribute noise to the signal, they are left together with the echoes. Nevertheless, it will not affect the derivation of T2 since the algorithm will ignore the excitation signal.

Table 3.1 Summary of the measured and simulated coil parameters at 20 MHz

Turns Resistance Inductance Quality factor

7 (sim.) 1.44 Ω 347.80 nH 30.297 (meas.) 1.52 ± 0.04 Ω 373.4 ± 2.7 nH 31.0 ± 0.69 (sim.) 2.39 Ω 646.58 nH 34.019 (meas.) 2.48 ± 0.03 Ω 687.0 ± 4.4 nH 34.8 ± 0.3

For measurement data, four coils are measured for each case. Sims simulation and meas measure-ment. Parameters: copper thickness = 26 μm, FR-4 relative permittivity = 4.5


https://doi.org/10.1007/978-3-319-67825-2_BM1

49

Further, the amplitude of the first echo is smaller when compared with the second and third echoes. This is caused by the superposition of different coherence path-ways [29]. Since the first echo is severely interrupted by this effect, leading to an amplitude smaller than expected, it will be excluded from the curve-fitting algo-rithm to prevent the error. Figure 3.6b shows the relation between the concentrations of CuSO4 and T2. T2

−1 increases linearly with the concentration of CuSO4 (0.9866 mM−1 s−1).

DMF Device Integration

Figure 3.7a shows the fabricated DMF chip. It consists of only eight electrodes in a row for this prototype, and thus only one assay can be performed each time. Yet, this system can still demonstrate the idea of integrating NMR assays with the DMF platform. Figure 3.7b, c shows the original and final positions of the droplet. The droplet was transported from electrode no. 1 to no. 8, which is the NMR sensing site, by applying voltage signals on the electrodes properly. The velocity of the droplets is ∼1.8 mm s−1. No obvious distinction of droplet movement was observed with and without a strong magnetic field, as expected, since the DMF works with the electric field for droplet manipulation.

The proposed system integrates two core technologies: NMR and DMF. Integration with the DMF system can enable NMR to be performed in a more automatic and controllable way. Herein the system was operated to test the presence of avidin in the water sample using biotinylated MNP as the probe [2, 17]. The sample under assay was placed at electrode no. 1, and the probe (droplets with MNPs) was placed at electrode no. 8. These two droplets will combine at electrode no. 7 to form a

0

0.2

0.4

0.6

0.8

1

1.2

Nutation Curve of Butterfly coil

Pulse width (µs)

No

rmal

ized

Am

plit

ud

e Fitted sine curve

Measured point

0 50 100 150 200 250 300

Fig. 3.5 Nutation curve of the seven-turn (each loop) Butterfly-coil. The normalized amplitude from different durations of RF excitation signals was recorded and fitted to the sinusoidal wave. The estimated π/2-pulse width for the coil is 144 μs


50

larger droplet, as shown in Fig. 3.8a. To ensure thorough mixing of the droplet, it was shuffled between electrode no. 6 and no. 8 several times. Then, the droplet was moved to electrode no. 8 for NMR sensing. Figure 3.8b shows the received NMR signal with and without avidin. Without avidin in the sample (i.e., sample only contains water), T2 of the droplets is 181.5 ms. When avidin presents in the target droplet, it will bind to the biotin and form micro-cluster. Consequently, the T2 of the droplets decreases to 86.13 ms with ΔT2 of −52.55%. The T2 of the NMR signals decreases proportionally to the concentration of the targets in the samples, as sug-gested in previous research [2]. These results show that the NMR–DMF platform can successfully detect the existence of a specific target in the samples in a fully automatic manner. The results promise that this system is a low-cost NMR-based diagnostic tool with high portability and electronic automation.

Measured data Trend lines

95% confidence level Error percentage

0

10

20

30

40

0 5 10 15 20 25 30

10

20

5

15

25

y = 0.9866x + 2.1009R2=0.9838

1/T

2(s-

1 )

Concentration (mM)

Erro

r Percen

tage

b0 0.05 0.1 0.15 0.2 0.25 0.3

-1

-0.5

0

0.5

1

Time (s)

Different coherence pathway

Excitationsignal-1.5

a

0 10 20 300

0.1

0.2

0.3

0.4

0.5

0.6

T2=343.6 ms

No. of Echoes

Am

plit

ud

e (V

)

Am

plit

ud

e (V

)

Fig. 3.6 (a) Received NMR signal from water. Inset shows the received NMR signal. The echoes were bounded by the gray-dotted trend line. (b) T2 of the samples versus concentration of CuSO4 solution, and results were shown on the graph (■). The trend lines were drawn together with their equation and 1/T2 value, together with error percentages (defined as half of 95% confidence level/true value) marked on the graph with dot lines where the values were displayed on the right axis


51

3.3 Second Prototype: CMOS Micro-NMR Platform with DMF

After verifying the functionality from the first NMR–DMF prototype, we figured out that the discrete electronics stiffened the integration of NMR system and the DMF device. Further, the discrete electronics are prone to interference and thus affecting the diagnostic sensitivity. Hence, integrating those discrete electronics on a single CMOS IC will reduce the overall dimensions of the module while improv-ing the sensitivity. In addition, the fL of the samples were discovered to relate with the ambient temperature. This hinders the applicability of the platform outside the laboratory and thus calls for a calibration scheme on either the fL or fEXC.

The second prototype developed here—CMOS micro-NMR platform with closed-loop DMF device—combines relaxometry measurement and electronic- automated sample management in a handheld scale. As shown in Fig. 3.9, there are four major parts: (1) a portable magnet for magnetizing the nuclei; (2) a PCB-based Butterfly-coil for transduction between magnetic fields from the nuclei and electri-cal RF signals for the TRX, inside the space-limited magnet; (3) a CMOS TRX to excite the nuclei with specific pulse sequences and collecting their responses; and

Fig. 3.7 (a) Fabricated DMF device. For illustration, the electrodes are numbered 1–8; (b, c) operation of the DMF platform. The droplet was originally placed at electrode no. 1 (highlighted by the circle). By applying a signal on electrode no. 2 and then turning off electrode no. 1, the droplet moved to electrode no. 2. As such, the droplet can be transported to electrode no. 8, which is the NMR sensing site


52

(4) a DMF device for closed-loop sample management. Its operation involves three phases: pre-diagnosis setup, sample preparation, and analysis (Fig. 3.10). Before the experiment, all of the samples and probes (if any) under assay are preloaded into the DMF device. Frequency matching between the fEXC and fL requires magnetic field calibration. This obviates the challenge from ambient temperature variation

Fig. 3.8 (a) Illustration of droplets mixing. The droplets at electrode no. 1 (samples) and no. 8 (probe) were driven to electrode no. 7 and mixed together. (b) The NMR assay results from the mixed droplets

Fig. 3.9 Portable electronic-automated micro-NMR system. It features a CMOS TRX and a PCB- based Butterfly-coil inside the magnet to transduce between magnetic and voltage signals. The analyte is placed inside a glass substrate DMF device atop the Butterfly-coil for sample manage-ment (only one electrode is shown for simplicity)


53

and secures the function of the micro-NMR assay. Then, the DMF device scans and delivers, individually, the samples to the micro-NMR sensing site, with closed-loop automation, thus tracking the experimental status. Upon arrival, with the micro- NMR assay triggered and the CMOS TRX taking results, a customized software analyzes and performs target identification. For multi-sample analysis, the DMF device keeps transporting the samples to the micro-NMR sensing site for assay. These modules cooperate in a closed-loop manner to culminate in a PoC device capable of identifying analyte inside μL droplets with multistep reaction and timing control.

3.3.1 Design and Implementation of CMOS TRX

Figure 3.11 illustrates the electrical schematic of the system, which includes a Butterfly-coil, a CMOS micro-NMR TRX, and a DMF device with its DMF elec-tronics, plus an FPGA with onboard ADCs that delivers the assayed results as well as receives command to/from the PC for easy execution and upgrade of protocols

Portable Magnet with Calibration Scheme

Sample preloading

Magneti cfieldcalibration

Detect dropletslocation

Paths calculationand validation

Transport dropletsto destination

Trigger NMRexperiment

Data acquisition &result derivation

YesNo

Droplets Manipulation

Result Analysis

T2:87 ms

Micro-NMR Sensing SiteElectrode

Target Droplet 2

Target Droplet 1

Setup

SamplePreparation

Analysis

Analyze allthe samples?

Reach NMRsensing site?

No

Auxiliary coil(75 mT/A)

Temperaturedependent variation(T.C. -1000 ppm/˚ C)

Static magnetic field(nominal 0.458 T

@ 24˚C)

Portable NdFeB magnet (1.2 kg)

∑

Temperatureindependent

magnetic field

Magnetic fieldcalibrator

Ambienttemperature

Fig. 3.10 Three-phase operation of the micro-NMR system: setup, sample preparation, and analysis


54

written in software. One electrode has the Butterfly-coil placed underneath as the micro-NMR sensing site. The design of the CMOS micro-NMR TRX focuses on the TX’s flexibility and power efficiency, RX’s input-referred noise (IRN), and the interface with its Butterfly-coil.

For micro-NMR excitation, a specific CPMG pulse sequence at fL (~20 MHz), as revealed in Fig. 3.2, is required to apply on the sample to precess the protons and refocus the incoherent magnetization across the samples. To realize this, a TX con-sisting of a pulse-sequence synthesizer (PSS, Fig. 3.12) and a switching power amplifier (PA) is utilized. An external signal generator LOref feeds the TX with a frequency 4x the fL + fIF, and the frequency dividers generate a four-phase LO matched with fL + fIF, with its output serving as excitation pulses as well as the LO signal for the quadrature mixers in the RX.

The logic gates mastered by the command from the FPGA flexibly generate the control signals with specific pulse width and interval for the switches of the whole TRX and multiplexer (MUX). The MUX governs the excitation signals delivered to the subsequent PA and Butterfly-coil. The PA is constituted of tapped inverter chains

CMOS micro-NMR Transceiver

Receiver

5V

40VBoostConverter

0V

40V

-20V20V

T

LOref

I/Q MixersLPFIF Amp.

fRF ≈19.6MHz

DMF Electronics

Capacitance-to-DigitalSensing Module

4-Phase LO

SwitchArray

T

FPGA

µNMRSensing

Site

Butterfly Coil

DMFDevice

Oscillator(1kHz)

MoveDroplet

1. Detect Droplet Positioning2. Move Dropletto Sensing site3. Perform micro-NMR screening

HPF

12

3PA

LNA

ADC

ADC

Actuation of Electrode

Electrode Array

15

Capacitance (CElec) of Electrode

TransmitterPulse Sequence

Synthesizer

Cext

CElec

CElec

Fig. 3.11 Block diagram of the micro-NMR TRX cooperated with the DMF device. It includes a CMOS micro-NMR TRX with a Butterfly-coil input, a DMF device, and DMF electronic. An electrode has the Butterfly-coil placed underneath for performing micro-NMR assays. An FPGA connected to a computer coordinates the hardware


55

to ensure a high dynamic power efficiency when driving up the Butterfly-coil to excite the protons.

After the excitation pulses, the same Butterfly-coil picks up the magnetic field produced from the nuclei spin. Similarly, a capacitor Cext is placed in parallel with the Butterfly-coil to form an LC resonator with resonant frequency at fL to achieve passive amplification. Afterward, this diminutive signal (<1 μV) is recorded by the RX. A multistage NMOS–PMOS complementary differential-pair low-noise ampli-fier (LNA) heads the RX to maximize its sensitivity and reject the common-mode noise (Fig. 3.13a). The IRN for this LNA (single stage) is:

vkT

g gM M

2 8

1 3

noise LNA, =+γ

(3.3)

with the Boltzmann constant k, the temperature in Kelvin T, and noise coefficient of MOSFET γ (assuming the same γ for PMOS and NMOS). gM1

(M1) and gM3 (M3)

are the transconductances assuming matched input pairs (i.e., M1 = M2 and M3 = M4). When compared with PMOS-/NMOS-only input stage where the loads (either active or resistive) inevitably contribute noise to the output, this current- reuse topol-ogy can reduce the noise more effectively (simulated, 0.51 nV/√Hz) by raising the bias current (i.e., higher gM1

and gM3). This topology usually suffers from limited

voltage headroom since both M1 (M2) and M3 (M4) concurrently bound the input and output voltages. Nevertheless, ascribed to the fixed input bias (0.9 V) and tiny dif-ferential micro-NMR voltages, this topology particularly suits micro- NMR sensing. To reduce the substrate coupling noise from the digital circuits of the TX, the LNA is isolated from the p-type substrate by the deep N-well. Three stages of this ampli-fier are cascaded to achieve a gain of 87.6 dB with an overall simulated IRN of

QQ

QQ

QQ

CLK

CLK

CLK

4-Phase LO(to mixers)

DifferentialExcitation Pulses

with Specific PhasesLOref

Commands(FPGA)

ModeSwitching

MUX

Div-by-2

Logic Gates

Fig. 3.12 Pulse-sequence synthesizer. FPGA commands control the logic gates to master the start and duration of the excitation signals with different phases as well as the switching between TX and RX modes


56

0.63 nV/√Hz, dominated by the forefront LNA (Table 3.2). These stages are AC coupled to avoid the DC offset of the LNA from saturating the RX.

Subsequent to the LNA, a double-balanced active quadrature mixer based on the Gilbert cell (Fig. 3.13b) is employed to downconvert the RF signal to the baseband (BW <5 kHz), whereas the quadrature signals can be processed at the software level to reject the image noise. This active mixer provides additional gain (12 dB), relax-ing the noise requirement of the subsequent stages. The RF-sharing stage (M5 and M6) improves the I/Q matching, and the cascode transistors (M7 and M8) improve the isolation among the ports. The downconversion is accomplished by the switches (S1–S8), which are controlled by the square LO from the TX for better noise perfor-mance. To prevent the hard-switching interference affecting other sensitive circuits,

(a)

ωLcoil

Butterfly-Coil Noise Model

CextVEMF

VDD

Lcoil

Rcoil

Q = Rcoil

V2noise,coil V2noise,coil

Mb1

M3

CMFB

M4

Vout+Vout-

M1 M2

(b) (c)

Vin+Vin-

M9 M10

VDD

M11 M12M15

C4/2

C2/2

C1/2

C3/2C6/2

C5/2

M13 M14

M16

M17

M19 M20

M18

Vout+Vout-

R1

LOl+ LOl+LOQ+

Vb,i Vb,i

Vin+ Vin-

IFQ+IFQ-IFl+IFl-

LOl- LOQ-LOQ+

R2

S1 S2 S3 S4 S5 S6 S7 S8

R3 R4

M5

M7

M6

VDD

M8

Fig. 3.13 (a) Butterfly-coil-input LNA and its noise model. (b) Double-balance quadrature mixer with RF-sharing stage. (c) Source-follower-based tunable bandwidth LPF

Table 3.2 Simulated noise summary of the LNA

Input-referred noise (nV/√Hz)

Noise contribution (%)

First-stage LNA PMOS pairs 0.30 22.76NMOS pairs 0.41 42.35

Thermal noise from back end (via polysilicon and metal)

0.36 32.91

Others (second- and third-stage LNAs) 0.09 1.88Total noise for the RF stage 0.63 100


57

all transistors in this mixer (including the switches) are PMOS with individual N-well. As the RX only has to process a single tone signal at fL, typical linearity indicators such as IIP3 cannot fully reflect its performance. Since only the ampli-tudes of the signals influence the relaxometry, the distortion on the downconverted monotone signals is critical for the RX design. From the simulations, for an input voltage amplitude of interest (<100 mVpp), the third harmonic of the mixer is 40 dB below the fundamental signal (Fig. 3.14a). In terms of distortion, the mixer has a total harmonic distortion (THD) below 1% (assuming the high-frequency mixing products are filtered and irrelevant) for an input voltage within 100 mVpp, facilitating the consummation of frequency downconversion for the NMR signals (Fig. 3.14b).

A baseband low-pass filter (LPF) is essential to remove the high-frequency mix-ing products as well as the out-of-band noises of the signals after mixing. Two sixth- order Butterworth I/Q LPFs are implemented by cascading three source-follower-based biquads (Fig. 3.13c). Such biquad topology is area efficient as it can realize poles at low frequency by limiting the transconductance of the transistors, thereby averting large capacitors [30, 31]. Furthermore, it is power efficient as it does not entail power-hungry operational amplifier and is capable of synthesizing complex poles in a single branch by using a transistorized positive feedback (e.g., M11 and M12), which also benefits the linearity [31]. The change in bias current (i.e., transconductance) easily tunes its BW, preserving its Butterworth response against BW variations, i.e., the poles form a semicircle on the s-plane (Fig. 3.15a). Similar to the mixer, the simulated THD of the LPF is within 1% for an input voltage within 400 mVpp (Fig. 3.15b). The simulated integrated IRN (0.1–100 kHz) of the LPF is 177.6 μV, and it is designed with a power consumption of 212.4 nW per channel. The total capacitance of the LPF (per channel) is 49.2 pF. The figure of merit of the LPF is 128 fJ at a cutoff frequency (f-3dB) of 5 kHz and is comparable to the state of the art [30].

0

0.2

0.4

0.6

0.8

1

1.2

1.4

10.1 10 100 100010.1 10 100 1000

Mixe

r THD

(%)

Input Freq.: 20.002 MHz Input Freq.: 20.002 MHzLO Freq.: 20 MHz LO Freq.: 20 MHzFundamental (2 kHz)

3rdHarmonic (6 kHz)

100

1

10m

100µ

1µ

10n

Outp

ut L

evel

(mV p

p)

Input Level (mVpp) Input Level (mVpp)

(a) (b)

Fig. 3.14 Simulation results of the mixer with LO = 20 MHz and input frequency = 20.002 MHz (i.e., IF = 2 kHz): (a) output against input for fundamental and third harmonic. (b) THD of the mixer at different input amplitudes


58

Theoretically, f-3dB should match with the BW of the signal (i.e., 5 kHz). Yet, the exciting pulses from TX will saturate the RX, particular in the LPF as it has a slow recovery time attributed to the low f-3dB. This dead time (tdead), which is dominated by the LPF and inversely proportional to its f-3dB, ineluctably limits the shortest interval between the echoes of the CPMG pulses and hence the number of achiev-able echoes within the echoes train. For the samples with short T2, tdead limits the sensitivity of the experiment [32]. To minimize tdead, a dynamic-baseband-BW tun-ing scheme is applied for the source-follower-based LPF (Fig. 3.16). It leverages the compromise between fast recovery time from the excitation pulses in the TX mode and minimum BW in the RX mode to reduce the in-band noise. The TX signals dynamically tune the LPF’s bias current, slightly delayed from the exciting pulses for fully recovering. After the exciting pulses, the LPF’s BW returns to low for higher rejection of the out-of-band noise. The simulated IRN of the overall RX is 0.92 nV/√Hz before image noise rejection. Operated by batteries, the entire TRX powers down automatically when idling.

3.3.2 Portable Magnet and RF Coil Codesign

This platform shares the same portable magnet with the first prototype. However, the ambient temperature severely affects the fL of the protons since from (3.1), the fL of the protons shifts with the temperature-dependent B0 (temperature coefficient,

-20k -10k 0

-20k

-10k

0

10k

20k

5 kHz11 kHz16 kHz21 kHz

OverallBandwidth

Re(s)

Im(s

)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

101 100 1000LP

F TH

D (%

)Input Level (mVpp)

Input Freq.: 2 kHzCutoff Freq.: 5 kHz

(a) (b)

Complex s-planeLocus Path

(Increase Current)

Poles

Fig. 3.15 (a) Simulated pole plot of the LPF. The sixth-order LPF implements a Butterworth filter (poles form a semicircle) with various cutoff frequencies obtained by changing only their bias cur-rents. (b) Simulated THD of the LPF with an input frequency of 2 kHz and a cutoff frequency of 5 kHz for different input levels


59

−1000 ppm/K). Without calibration, LO frequency deviation from fL will paralyze the system due to improper fEXC on the nuclei. To resolve this, a magnetic field cali-brator consisting of a temperature sensor and a current driver is proposed, as shown in the inset of Fig. 3.10. A temperature sensor MAX6612 detects the ambient tem-perature. Then the current driver delivers the appropriate amount of current (±50 mA) into the auxiliary coil of the magnet. This coil can trim B0 accordingly (75 mT/A) to compensate the shifting of B0. Thus, a stable B0, which is the sum of the magnetic field from the permanent magnet and the auxiliary coil, is attainable. The injected current necessitated can be calculated from the temperature (i.e., shifting of B0 shift from the nominal value). This calibrator allows a fix fEXC, which is beneficial to the robustness of the system as a PoC diagnostic device. Furthermore, for this prototype, the optimization of the SNRs is not only at the circuit level but also in the codesign of the coil and RX to soothe the weak NMR signal. The voltage induced on the Butterfly-coil (VEMF) increases with the number of turns (N) for each spiral coil attributed to the higher B1. Yet, the thermal noise of the undesired coil resistance ( v2

noise coil, ), as depicted in Fig. 3.13a, also increases with the turn’s number. Together with the passive amplification network and the RX, the SNR of the output will become:

SNROUTEMF

noise coil noise LNA

=+

+( ) +( ) ⋅V Q

v Q v f

2

2 2 2

1

1, , ∆

(3.4)

where v2noise LNA, is assumed dominating the IRN of the RX and ∆f is the BW of the

signal. Revealed from (3.4), the SNROUT is affected at both coils’ geometric level (VEMF, v2

noise coil, , and Q) and circuit level ( v2noise LNA, ). After settling the RX noise

level limited by the process and power budget, the geometry of the Butterfly-coil can be finalized by performing simulation on VEMF, v2

noise coil, , and Q with the finite element analysis software (COMSOL Multiphysics). The SNROUT maximized at 23.2 dB with seven turns on each spiral coil (Fig. 3.17) is a reference value for opti-mization, as the magnitude of VEMF will vary in practice.

Dead time for the receiver torecover from the excitation pulse

TX Pulses

RX Output

TX EN

T2

Filter CurrentControl

Excitation pulse

Fig. 3.16 The micro-NMR pulse sequence. It includes the CPMG pulse, filter current control, and micro-NMR output signal where the dead time of the RX is shown


60

3.3.3 DMF Device and Its Control Circuit

To obviate laboring efforts on samples during the experiments (also the chance of sample defilement), a DMF device is introduced to enable electronic-automated sample management inside the portable magnet similar to the first prototype. Compared to the first prototype, the DMF electronics are improved to avoid the usage of external signal generator and transformer to produce the actuation voltage signal. To manipulate the droplets, the DMF electronics (Fig. 3.11) control the over-all DMF device. For the actuation of droplets, a high-voltage inverter (0–40 V) driven by an oscillator (~1 kHz) is utilized to generate a square wave for driving the electrodes, with the high voltage generated by a DC-DC boost converter. A switch array controls the on/off patterns of the electrodes. Further, a capacitance-to-digital module is included to sense the position of the droplets in real time, as the high- permittivity water droplets (80x of air) affect the capacitance of the electrodes (CElec) [33]. These electronics are integrated on the same PCB with the micro-NMR system to enhance the integration level and facilitate the operations. An FPGA mas-ters the operations including sample locating and transporting. This entire DMF module forms a closed-loop control for complete automated sample management, where the FPGA controls the path of the droplets together with the operations of the micro-NMR assays. To prevent cross talk from appearing on the micro-NMR results, during the micro-NMR assay, the DMF module switches off. The designed DMF device has 15 electrodes, each with 3.5 × 3.5 mm2. Appendix B includes the completed description about the DMF electronics.

3.3.4 Experimental Results

Electrical Measurements

The micro-NMR TRX fabricated in 0.18-μm CMOS has a die area of 1.6 × 1.3 mm2, dominated by the capacitors of the LPF in the RX (Fig. 3.18a). The operating volt-age is 1.8 V, and the power consumption is 6.6 mW in TX and 23.7 mW in RX

18

20

22

5 7 9 11No. of turns (per spiral)

SNR O

UT (d

B)

SNROUT =v2noise,coil (Q2+1) + v2noise,LNA

24

Q2+1VEMF

Fig. 3.17 Simulated SNR of the Butterfly-coil-input CMOS RX with different number of turns in the coils


https://doi.org/10.1007/978-3-319-67825-2_BM1

61

(Fig. 3.18b). The TX has a power efficiency of 27.5%, and the effective B1 applied to the sample is 5.3 Gauss. For the RX, the forefront LNA dominates the power consumption (68%). When the micro-NMR system turns off after each assay, the power consumption lowers to 100 μW. Assuming that the micro-NMR experiments, performed with two AA batteries and an external DC-DC converter with a 90% efficiency, are with 256 echoes and 4 ms of echoes interval (~1.1 s), they allow approximately ~1 million assays. The Butterfly-coil is fabricated on the same PCB containing the CMOS TRX to prevent parasitics and enable a higher level of integration.

As the TRX is linked with the Butterfly-coil, the RX cannot be evaluated by typi-cal 50-Ω equipment. To address it, another Butterfly-coil (same dimension) is con-nected with a signal generator to couple an RF magnetic field on the original Butterfly-coil (separation, 10 mm), thereby emulating the RF magnetic field induced by protons’ spinning. With the external RF magnetic field with a frequency of 19.999 MHz and the LOref with frequency of 80 MHz (i.e., LO for mixer at 20 MHz), signals with a frequency of 1 kHz are revealed at the quadrature outputs. The digi-tized I channel is processed with the Hilbert transform in the digital domain (soft-ware level) to achieve a 90° phase shift (Fig. 3.19a). The resultant signal is then added to the signal from the Q channel to achieve image rejection. The SNR of the signal improved by 40% (3 dB) after image noise removal matches with the theo-retical derivation (Fig. 3.19b). Besides, to substantiate the fact that the DMF device has a negligible impact on the Butterfly-coil, a repetition of the above measurements is necessary with the DMF device put atop the Butterfly-coil. There is no manifest

1.3 mm

1.6 m

mµNMR Receiver

Front-End

µNMRTransmitter

Receiver IFAmplifier & LPF

Process CMOS 0.18 μm 6M1PVoltage 1.8 V

Power Consumption

RX: 23.7 mWTX: 6.6 mW

TX efficiency 27.5 %RX Input-

Referred Noise* 0.92 nV/√Hz

RX Gain and Passband Bandwidth*

87.6 dB

2 to 33 MHz

LPFBandwidth: 2 to 20 kHz

Settling Time: 0.1 to 0.4 ms

*Extracted from simulation as the RX is interfacedwith the Butterfly-Coil

(a) (b)

Fig. 3.18 (a) Chip photo. (b) Measured performance summary of the micro-NMR TRX. The RX’s IRN, gain, and BW can only be assessed by simulations as the RX input has been tied to the Butterfly-coil


62

impact on the signal (both signal amplitude and SNR) at the output of the RX, vali-dating that the DMF device is uninfluential on the Butterfly-coil-input CMOS TRX.

Figure 3.19c presents the f-3dB and settling time (ts, 5% from the steady state) of the LPF from a step response (0–100 mV) measured at different bias currents. The f-3dB can be dynamically tuned from 2 to 20 kHz by switching a single bias current (from 1.5 to 21.5 μA) shared by the three biquads of the LPF. Consequently, ts of the LPF from the step response decreases from 0.4 to 0.1 ms. Thus, the RX can be set at a higher f-3dB (20 kHz) to swiftly recover from the excitation pulses (ts: 0.1 ms) in the TX mode, to facilitate a short tdead. On the other hand, a lower f-3dB (5 kHz) is set in the RX mode to limit the in-band noise of the RX, perceiving the contrariety between tdead and in-band noise.

The measurement of B0 of the magnet against different temperatures (calibrated inside a temperature chamber) verifies the operation of the magnetic field calibrator. As shown in Fig. 3.20, the magnetic field stabilization scheme suppresses B0 varia-

(b) (c)

0.1

1m

10µ0.1k 1k 10k

Frequency (Hz)

Outp

ut V

olta

ge (V

)

0.1

0.2

0.3

0.4 Settling Time (m

s)

0 5 10 15 20 250

5

10

15

20

0

Cuto

ff Fr

eque

ncy (

kHz)

Reference Current (µA)

RX mode TX mode10 25 0.5

(a)

ADC HilbertTransform

ADC

On-chip Off-chip PC

sinωLOtcosωLOt

RFinReconstructed

output

LPF&Driver

LPF &Driver

I

Q

Fig. 3.19 (a) Block diagram of the image-reject RX. (b) Measured RX output spectrum with an externally coupled magnetic field at 19.999 MHz and a LO of 20 MHz after image noise removal. (c) Cutoff frequency and settling time of the LPF versus the bias current. Working regions of the LPF at different modes are labeled


63

tion from 5.54 mT to 70 μT (3 kHz in terms of frequency) at nominal B0 (0.458 T) with temperature variation from 18 to 30 °C. Thus, with the fEXC fixed at 19.5 MHz, it eases the calibration of the device for PoC applications.

System Verification

Figure 3.21 illustrates the micro-NMR system, including the DMF device and its control circuit, integrated on a single PCB for compactness and better reproducibil-ity. An FPGA (DE0-Nano) monitors the schedule of sample movement and micro- NMR signal acquisition and digitization. A software program in C# visualizes the micro-NMR assays results and shows the execution status of the experimental pro-tocol in real time. Appendix C explains in detail about the software and hardware interface. Samples are preloaded inside the DMF device before the experiment, with their transportation and mixing together with the triggering of micro-NMR

Temperature (˚C)

Magn

etic

field

(mT)

With Calibration

18 20 22 24 26 28 30455456457458459460461

Without Calibration

(Error bars blocked by the symbol)

Fig. 3.20 Measured B0 with and without calibration

Fig. 3.21 The system hardware of the micro-NMR system. It is linked with an FPGA (DE0-Nano) and a program implemented in C# which facilitates the system control, result collection, and displays


https://doi.org/10.1007/978-3-319-67825-2_BM1

64

experiments executed electronically. The platform firstly starts with detecting the location of the samples inside the DMF device. As shown in Fig. 3.22, the average pulses counted on the occupied and vacant electrodes are 277.5 and 757.7, respec-tively. This 2.73× difference is sufficient to identify whether the electrodes are occu-pied by the droplets. This sensing module is critical in this two-dimensional (2D) DMF device as the relaxometer entails handling of multiple droplets. Moreover, with this sensing module adopted, the system is under a closed-loop control to attain an efficient droplet management scheme. The operations of the droplets including the trigger of the NMR experiment, stabilization of the hydrogen nuclei before the NMR experiment, and antimerging droplet paths can be manipulated and optimized by the software within the shortest time to boost the efficiency and throughput of the system.

After the identification of the droplet location, the program starts to transport the droplet to the micro-NMR sensing site. The droplets are guided to the destination gradually with their positions tracked in real time to ensure successful movement. To visualize these movements, the movement of the droplets was recorded outside the magnet. Figure 3.23a exhibits an example using a water droplet (8 μL) routed to the micro-NMR sensing site. Figure 3.23b shows the progressive movement of the droplet. The DMF platform guides the droplet to the corresponding electrode through the application of a voltage signal progressively to achieve sample transpor-tation with an average velocity of 1.17 mm/s. The elevation of the actuation voltage improves the velocity. Yet, this burdens the electric fields on the Ta2O5 layer and deteriorates the reliability of the DMF device. Thus, a moderate voltage of 40 Vpp was chosen, as the velocity is not a critical issue for the application.

Upon the arrival on the micro-NMR sensing site, the system triggers the micro- NMR assay automatically (Fig. 3.23c). The parameters of the micro-NMR experi-ment can be tuned from the program. After removing the image noise for the quadrature signals, the acquired micro-NMR results are analyzed and displayed in the program. The T2 from the water droplet (1.16 ± 0.03 s) can be derived from

0200400600800

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15Electrode No.

No. o

f Pul

ses C

ount

edon

eac

h el

ectro

de Covered by air Covered by water

Fig. 3.22 The pulses counted on the electrodes covered by air and water, respectively. As the permittivity of water is substantially larger than air (80:1), the capacitance of the electrode covered by water is higher, causing lower pulses to be counted, and thus the system can detect if the elec-trode is vacant


65

fitting the envelope of the echoes response to a decay exponential function (Fig. 3.23d). To enhance the SNR, the experiments are repeated 8x, and their results are averaged to suppress the background noise.

Similar samples from the first prototype are tested with this platform. As para-magnetic CuSO4 ions have a high magnetic susceptibility, it will perturb the local field of the surrounding protons and shorten the T2, and thus it is used as the test agent in the first experiment. As shown in Fig. 3.24a, the micro-NMR relaxometer can detect the CuSO4 concentration with respect to T2

−1.The second experiment demonstrates the capability of the system to pinpoint

specific biological targets with predesigned probe-decorated MNPs similar to the first prototype. Iron MNPs with biotin labeling were used as a probe to quantify the avidin in the samples. Figure 3.24b depicts the experimental results and shows that the T2 value decreased proportionally to the concentration of avidin with an achieved sensitivity of 0.2 μM from 8 μL of samples. The sample volume can be further reduced, with the trade-off of worse sensitivity. These experiments evince that the micro-NMR relaxometer is capable of handling and quantifying chemical and bio-logical targets.

Fig. 3.23 Operation of the micro-NMR system. (a) Initial position of the sample and its projected path. (b) Droplet moves to the adjacent electrode. (c) Final position (micro-NMR sensing site) of the droplet. (d) Measured micro-NMR signal from water droplet excited by CPMG pulse sequence with 256 echoes and 4 ms interval. The envelope is extracted and fitted to a mono-exponential function, as shown in the inset


66

Multistep Multi-Sample Droplet Operation

One unique feature of this micro-NMR relaxometer is the capability to handle dis-tinct samples and perform micro-NMR experiments on them sequentially. This is attributed to the expanded 2D electrodes, which is benefitted from the compact CMOS TRX. This feature is demonstrated by placing two stationary targets and two identical probe-decorated MNP droplets inside the DMF device at the same time. Since the relaxometer has to handle multiple samples, it is crucial to distinguish the droplets and project individual paths for them without the risk of fortuitous mixing. As the capacitance-to-digital module can track the location of the droplets, their individual paths can be procured at the software level. As shown in Fig. 3.25a, the first probe-decorated MNP droplet (7.5 μL) is firstly guided to a stationary target (2.5 μL) for mixing and then to the micro-NMR sensing site to extract the T2. Concurrently, the second probe-decorated MNP droplet is guided to another target for mixing. The first mixture is led away from the micro-NMR sensing site after finishing the diagnosis in 48 s, allowing the second mixture to enter. The assays are completed after the second mixture finished micro-NMR screening, and the raw data are processed using the PC for concentration quantification (T2 for the water sample, 256 ms; for avidin, 211 ms). Two or more probes and target pairs can be

Fig. 3.24 (a) The correlation of ΔT2−1 (with reference to 0 mM of CuSO4) with the concentration

of CuSO4. The echoes amplitude for the case of CuSO4 at 1 mM concentration is plotted above. One hundred twenty-eight echoes were collected for each single experiment. (b) The correlation of ΔT2 (with reference to 0 μM of avidin) with the concentration of avidin. The echoes amplitude for the case of avidin at 0.2 μM concentration is plotted above. Sixty-four echoes are collected for each single experiment


67

placed inside the DMF device for enhancing the throughput of the relaxometer, depending on the geometry of the DMF device and its electrodes.

Figure 3.25b depicted the timing diagram of the micro-NMR relaxometer operation. With the droplet movement controlled by the program automatically and their positions tracked in real time, the optimization of the route and timing man-agement can be done at the software level. The fastest protocol can be found and applied to the DMF device. This 2.2-min experiment validates the entire system as being capable of transporting, mixing, and analyzing multiple distinct samples in real time without challenging human operations while reducing the labor (error) and the risks of defilement.

3.3.5 Discussion and Outlook

Table 3.3 compares this work with other NMR platforms equipped with CMOS TRX/RX for different functionalities. Ascribed to the 2D DMF device, this micro- NMR system for target pinpointing is the first platform to accomplish electronic- automated multi-sample management, which is a promising feature for PoC devices [34].

ProbeNo.1

ProbeNo.2

Move to thetarget

Stay

Leave

Move the micro-NMRsensing site Perform micro-NMR

ElectrodeMicro-NMR Sensing SiteProbe (Biotinylated Nanoparticles)Target 1 (Water)Target 2 (Avidin)

1 2

3 4 5

Micro-NMR Micro-NMR

move& mix

Leave

move

move

Stay

4 steps, 12 s 4 steps, 12 s Repeat 8x, 48 s in total 6 steps, 18 s

3 steps, 9 s 58 s

Perform micro-NMR

6 steps,18 s Repeat 8x, 48 s in total

Move tothe target

move&mix

Stay

Stay

43 s

(a)

(b)

Move the micro-NMR sensing site

Fig. 3.25 (a) Illustration of the motions of the droplets for multistep multi-sample handling. T2 for the water sample: 256 ms; for avidin: 211 ms. (b) A Gantt chart of the operation of an individual droplet. The total time for the experiment is 2.2 min


68

This DMF device not only locates and transports the droplet to the micro- NMR sensing site, but it also supports pre- and post-experiment processes such as mixing and splitting of the droplets for broader applications. Besides, this DMF device can be integrated with the heating module to perform thermal profiling on the samples with micro-NMR assay [35]. This versatility of NMR and DMF integration is tantalizing to a micro total analysis system. Yet, the degraded sensitivity of the coil and the distance from the coil to the samples ascribed to the thickness of the DMF device worsen the SNR and thus the limit of detection of the system. To resolve this, a complete study of the geometry of the coil (multilayer or asymmetric spiral) and investigation on the size, magnetic core materials, and valency of the MNPs can be performed to ameliorate the limit of detection. Despite this, the electronic- automated sample micro-NMR assay still renders the proposed platform as a promising tool for PoC applications.

3.4 Summary

This chapter presents two NMR relaxometers for biological/chemical assays. The first one is a modular integration of NMR and DMF [36]. The geometrical limita-tions among the traditional spiral coil, planar DMF chip, and portable magnet are

JSSC’11 [3.17] CICC’12 [3.18] A-SSCC’13 [3.19] This Work

System PerspectiveFunctionality Relaxometry Spectroscopy Microscopy Relaxometry

Coil Style On-chip Spiral Off-chip Solenoid On-chip Spiral Off-chip PCB Butterfly

Magnet Portable (1.2 kg, 0.5 T)

Not Specified(5 T)

Large-scale (Bruker 7 T)

Portable (1.2 kg, 0.5 T)

One One One Multiple

No(Only micro-NMR)

No (Only micro-NMR)

No (Only micro-NMR)

Yes(Mixing/Splitting +

micro-NMR)Circuit Perspective

Process (Voltage) 0.18 μm (1.8 V) 0.13 μm (1.5 V) 0.13 μm (1.5 V) 0.18 μm (1.8 V)Operation Frequency 21 MHz 5-300 MHz 300 MHz 20 MHz

Integration Level TRX RX RX TRXIRN (w/ & w/o coil) 0.93/1.26 nV/√Hz --/3.5 nV/√Hz 0.26/--nV/√Hz 0.16/0.92 nV/√Hz

Power N/A 18 mW 18 mW RX: 23.7 mW/ TX: 6.6 mW

LPF Integration No Yes No Yes (Dynamic BW Tuning)

Pre-/Post-Sample Reaction

Supportability

Sample Handling per Experiment

Table 3.3 Comparison with the existing CMOS-based NMR system


69

overcome by introducing a Butterfly-coil, which could be fabricated with low-cost PCB with high reproducibility. It transduces between magnetic signals from nuclei spins and voltage signals of the TRX without physically contacting the samples (i.e., noninvasive) or affecting the sample movement managed by the DMF device (i.e., fully compatible). Extensive studies and experiments verified the compatibility and co-functionality of the two distinctive technologies: NMR (assay tool) and DMF (droplet sample operation). After the modular validation, a micro-NMR CMOS TRX for biological/chemical assays has been codeveloped with a glass sub-strate DMF device with closed-loop capacitive feedback for electronic-automated multi-sample management plus relaxometry measurement [37–40]. The TRX, which reduces the overall dimensions of the module while improves the sensitivity, primar-ily includes (1) a TX consisting of a PSS together with an inverter-based PA to excite the protons and (2) a RX consisting of a multistage NMOS–PMOS differential- pair LNA, an active double-balance quadrature mixer, and a source- follower- based dynamic-baseband-BW sixth-order Butterworth LPF that culminated in low in-band noise in the RX mode and short dead time in the TX mode to process the micro-NMR signals with negligible distortion. The functionality of this micro-NMR platform was evinced from the micro-NMR experiments with copper (II) ions and avidin detec-tion with biotinylated MNPs by measuring T2 of the samples. When compared to conventional microchannel NMR systems, this work offers a more flexible and elec-tronically automated method to handle multistep multi- sample diagnostic protocols inside the space-limiting magnet effectively.

References

1. J.D. Trumbull, I.K. Glasgow, D.J. Beebe, R.L. Magin, Integrating microfabricated fluidic systems and NMR spectroscopy. IEEE Trans. Biomed. Eng. 47(1), 3–7 (2000)


3. C. Massin, F. Vincent, A. Homsy, K. Ehrmann, G. Boero, P.A. Besse, et al., Planar microcoil- based microfluidic NMR probes. J. Magn. Reson. 164(2), 242–255 (2003)

4. I. Barbulovic-Nad, H. Yang, P.S. Park, A.R. Wheeler, Digital microfluidics for cell-based assays. Lab Chip 8(4), 519–526 (2008)

5. J. Gao, X.M. Liu, T.L. Chen, P.I. Mak, Y.G. Du, M.I. Vai, et al., An intelligent digital micro-fluidic system with fuzzy-enhanced feedback for multi-droplet manipulation. Lab Chip 13(3), 443–451 (2013)

6. F. Lapierre, M. Harnois, Y. Coffinier, R. Boukherroub, V. Thomy, Split and flow: reconfigurable capillary connection for digital microfluidic devices. Lab Chip 14(18), 3589–3593 (2014)

7. M.G. Pollack, A.D. Shenderov, R.B. Fair, Electrowetting-based actuation of droplets for integrated microfluidics. Lab Chip 2(2), 96–101 (2002)

8. M.H. Shamsi, K. Choi, A.H.C. Ng, A.R. Wheeler, A digital microfluidic electrochemical immunoassay. Lab Chip 14(3), 547–554 (2014)

9. V. Srinivasan, V.K. Pamula, R.B. Fair, An integrated digital microfluidic lab-on-a-chip for clinical diagnostics on human physiological fluids. Lab Chip 4(4), 310–315 (2004)

10. A.R. Wheeler, Chemistry—putting electrowetting to work. Science 322(5901), 539–540 (2008)

11. I. Barbulovic-Nad, S.H. Au, A.R. Wheeler, A microfluidic platform for complete mammalian cell culture. Lab Chip 10(12), 1536–1542 (2010)

References

70

12. G.J. Shah, A.T. Ohta, E.P.Y. Chiou, M.C. Wu, C.-J. Kim, EWOD-driven droplet microfluidic device integrated with optoelectronic tweezers as an automated platform for cellular isolation and analysis. Lab Chip 9(12), 1732–1739 (2009)

13. R. Sista, Z. Hua, P. Thwar, A. Sudarsan, V. Srinivasan, A. Eckhardt, et al., Development of a digital microfluidic platform for point of care testing. Lab Chip 8(12), 2091–2104 (2008)

14. Y.-H. Chang, G.-B. Lee, F.-C. Huang, Y.-Y. Chen, J.-L. Lin, Integrated polymerase chain reaction chips utilizing digital microfluidics. Biomed. Microdevices 8(3), 215–225 (2006)

15. Z. Hua, J.L. Rouse, A.E. Eckhardt, V. Srinivasan, V.K. Pamula, W.A. Schell, et al., Multiplexed real-time polymerase chain reaction on a digital microfluidic platform. Anal. Chem. 82(6), 2310–2316 (2010)

16. D. Witters, K. Knez, F. Ceyssens, R. Puers, J. Lammertyn, Digital microfluidics-enabled single- molecule detection by printing and sealing single magnetic beads in femtoliter droplets. Lab Chip 13(11), 2047–2054 (2013)

17. N. Sun, T.J. Yoon, H. Lee, W. Andress, R. Weissleder, D. Ham, Palm NMR and 1-Chip NMR. IEEE J. Solid State Circuits 46(1), 342–352 (2011)

18. J. Kim, B. Hammer, R. Harjani, A 5–300MHz CMOS transceiver for multi-nuclear NMR spec-troscopy, in Proceeding IEEE Custom Integrated Circuits Conference (CICC), 2012, pp. 1–4

19. J. Anders, P. SanGiorgio, G. Boero, A fully integrated IQ-receiver for NMR microscopy. J. Magn. Reson. 209(1), 1–7 (2011)

20. D.I. Hoult, R.E. Richards, The signal-to-noise ratio of the nuclear magnetic resonance experiment. J. Magn. Reson. 24(1), 71–85 (1976)


22. P. Andreani, K. Kozmin, P. Sandrup, M. Nilsson, T. Mattsson, A TX VCO for WCDMA/EDGE in 90 nm RF CMOS. IEEE J. Solid State Circuits 46(7), 1618–1626 (2011)

23. T. Mattsson, Method of and inductor layout for reduced VCO coupling, US Patent US 7,151,430, 19 Dec 2006

24. M. Nagata, H. Masuoka, S.I. Fukase, M. Kikuta, M. Morita, N. Itoh, 5.8 GHz RF transceiver LSI including on-chip matching circuits, in 2006 Bipolar/BiCMOS Circuits and Tech. Meeting, 2006, pp. 263–266

25. F. Mugele, J.C. Baret, Electrowetting: from basics to applications. J. Phys. Condens. Matter 17(28), R705–R774 (2005)

26. F. Fiorillo, C. Beatrice, Energy losses in soft magnets from DC to radiofrequencies: theory and experiment. J. Supercond. Nov. Magn. 24(1–2), 559–566 (2011)

27. W.K. Peng, L. Chen, J. Han, Development of miniaturized, portable magnetic resonance relax-ometry system for point-of-care medical diagnosis. Rev. Sci. Instrum. 83(9), 095115 (2012)

28. J.M. Pope, N. Repin, A simple approach to T2 imaging in MRI. Magn. Reson. Imaging 6(6), 641–646 (1988)

29. B. Blumich, J. Perlo, F. Casanova, Mobile single-sided NMR. Prog. Nucl. Magn. Reson. Spectrosc. 52(4), 197–269 (2008)

30. T.T. Zhang, P.I. Mak, M.I. Vai, P.U. Mak, M.K. Law, S.H. Pun, et al., 15-nW biopotential LPFs in 0.35-μm CMOS using subthreshold-source-follower biquads with and without gain compensation. IEEE Trans. Biomed. Circuits Syst. 7(5), 690–702 (2013)

31. S. D’Amico, M. Conta, A. Baschirotto, A 4.1-mW 10-MHz fourth-order source-follower- based continuous-time filter with 79-dB DR. IEEE J. Solid State Circuits 41(12), 2713–2719 (2006)

32. J. Watzlaw, S. Gloggler, B. Blumich, W. Mokwa, U. Schnakenberg, Stacked planar micro coils for single-sided NMR applications. J. Magn. Reson. 230(1), 176–185 (2013)

33. J. Gong, C.J. Kim, All-electronic droplet generation on-chip with real-time feedback control for EWOD digital microfluidics. Lab Chip 8(6), 898–906 (2008)

34. V. Gubala, L.F. Harris, A.J. Ricco, M.X. Tan, D.E. Williams, Point of care diagnostics: status and future. Anal. Chem. 84(2), 487–515 (2012)

35. P.Y. Keng, S.P. Chen, H.J. Ding, S. Sadeghi, G.J. Shah, A. Dooraghi, et al., Micro-chemical synthesis of molecular probes on an electronic microfluidic device. Proc. Nat. Acad. Sci. (PNAS) 109(3), 690–695 (2012)


71

36. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, NMR–DMF: a modular nuclear magnetic resonance–digital microfluidics system for biological assays. Analyst 139(23), 6204–6213 (2014)

37. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A thermal-insensitive all-electronic modular μNMR relaxometer with a 2D digital microfluidic chip for sample management, in Proceeding 19th International Conference on Miniaturized System Chemistry and Life Sciences (MicroTAS), 2015, pp. 302–304

38. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A μNMR CMOS transceiver using a Butterfly- coil input for integration with a digital microfluidic device inside a portable magnet, in Proceeding IEEE Asian Solid-State Circuits Conference (A-SSCC), 2015, pp. 1–4

39. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A palm-size μNMR relaxometer using a digital microfluidic device and a semiconductor transceiver for chemical/biological diagnosis. Analyst 140(15), 5129–5137 (2015)

40. K.-M. Lei, P.-I. Mak, M.-K. Law, R.P. Martins, A μNMR CMOS transceiver using a Butterfly- coil input for integration with a digital microfluidic device inside a portable magnet. IEEE J. Solid State Circuits 51(10), 2274–2286 (2016)

References


Chapter 4One-Chip Micro-NMR Platform with B0-Field Stabilization

4.1 Introduction

As demonstrated in Chap. 3, the B0-field shifting from the permanent magnet caused by the environmental variation can paralyze the NMR experiment if the fL shifts away from the fEXC. This instability calls for a calibration scheme to enhance the robustness of the system for point-of-use (PoU) application. Conventional fre-quency stabilization techniques are based on the measured NMR signals [1–3]. However, if the B0-field fluctuates large enough such that the excitation pulses can-not excite the nucleus effectively, those calibration schemes may not work properly. Chapter 3 has revealed the calibration on the B0-field to stabilize the fL of the pro-tons by detecting the ambient temperature and modulating the magnetic field from the auxiliary coil of the magnet. This method, however, suffers from a delay between the ambient temperature variation and the response on the B0-field of the magnet due to its high heat capacity (time constant > 10 min). Moreover, the off-chip tem-perature sensor complicates the design of the platform. This stiffens the design of the platform and prevents its applicability outside the laboratory.

To circumvent the above challenges, a trailblazing B0-field stabilization scheme for the portable magnet is proposed here. As the primary influence on the operation is B0, sensing the B0 directly can provide information for calibration immediately. Herein a handheld high-sensitivity micro-NMR CMOS platform utilizing a portable magnet is reported, as illustrated in Fig. 4.1. A Hall sensor with low-noise readout circuit is embedded, for the first time in the literature, with a CMOS micro-NMR TRX to achieve better robustness. The current driver stabilizes the B0-field of the magnet against ambient variation. The stabilized B0 avoids the need of a frequency synthesizer to tune the local oscillator (LO), and an untuned LO can be generated directly by the crystal oscillator (XO). To minimize the B0-field offset error between the sample and the Hall sensor, the samples under assay are loaded on the on-chip planar coil by a handheld pipette. The sensing coil also serves as a sample heater for thermal profiling. Further, benefitting from the versatility of the NMR assays, this

74

handheld tool unifies multi-type assays such as target detection, protein state analy-sis, and solvent-polymer dynamics in a platform, rendering it suitable for health-care, food industry, and colloidal applications.

The first section of this chapter discloses the detailed design and engineering of the platform, which is constituted of the design of micro-NMR TRX with multi-function sensing coil and the Hall sensor with its readout circuit. The experimental setup, electrical measurements, and biomolecule experimental results are then man-ifested. Finally, the discussion and compendium for this platform are revealed.

4.2 Platform Design

Figure 4.2 shows the schematic of the proposed micro-NMR platform. It includes B0-field stabilization to enhance the robustness and simplify the hardware. The Hall sensor and the readout circuit, together with an off-chip current driver, manage the lateral B0-field and stabilize the bulk magnetization and thus fL of the nuclei. The dynamic B1-field transduction between the nuclei and the electronics is based on an on-chip planar coil driven by a TX/RX together with the matching capacitor CM, to

Fig. 4.1 Conceptual diagram of the proposed micro-NMR platform for PoU applications. Different samples such as protein and DNA can be put directly atop the CMOS chip for assays. A portable magnet is entailed to magnetize the nuclei inside the samples

4 One-Chip Micro-NMR Platform with B0-Field Stabilization

75

excite/obtain the magnetic signal to/from the droplet samples (2.5 μL) normal to the chip surface. Furthermore, the on-chip NMR sensing coil can act as a sample heater for thermal profiling. A BJT-based temperature sensor aids both the Hall sensor thermal correction and thermal profiling of the samples.

4.2.1 Micro-NMR Transceiver

The micro-NMR TRX of this work has a similar structure to the TRX reported in Chap. 3. The details design and consideration can be referred therein. The main dif-ference in this work is that a XO is integrated with the PSS to generate the desired pulse sequence for excitations. This XO, with an oscillation frequency 4x (fL+ fIF), directly serves as the LO for both the TX and RX. The XO with an inverter-based amplifier and an off-chip crystal (78.5 MHz) features low-power low-noise opera-tion. Together with the proposed B0-field stabilization, the TX avoids the need of a frequency synthesizer to generate a variable LO to track the unstable fL. The signal from the XO is divided by four to provide the four-phase LO for the PSS and mixers in the RX mode.

Fig. 4.2 System block diagram. The TX and RX transduce between magnetic and electrical sig-nals with a thermal-controlled spiral coil. The B0-field sensor and calibrator automatically stabilize the bulk magnetization on the μL sample. No frequency synthesizer is required

4.2 Platform Design

76

4.2.2 Multifunctional Planar Coil

An on-chip planar octagonal coil (23 turns; conductor width, 40 μm; spacing, 1.85 μm) transduces between the magnetic signals from the nucleus spin in the axial direction and the electric signals for the TRX. This tiny coil miniaturizes the plat-form and sample consumption compared with the off-chip coil. A conventional octagonal coil is adopted since the microfluidic module is not included for this plat-form. To improve the quality factor of the coil (simulated Q = 1.9 at 19.6 MHz), a thick top metal (2.5 μm) process and a stacked-metal-layer technique are applied. The coil in parallel with a capacitor forms an LC tank offering passive amplification for the signal [4]. Furthermore, to extend the functionality of this micro-NMR plat-form, the NMR sensing coil is time-multiplexed to serve as a sample heater for thermal profiling. Modeled in COMSOL Multiphysics®, the entire temperature of the droplet uniformly raises (ΔT <0.3 °C) according to the applied current on the coil, which the power is dissipated as heat to raise the temperature of the droplet (Fig. 4.3a). The heating temperature is easily controllable by tuning the duty cycle of the injected current (Fig. 4.3b). A BJT-based temperature sensor monitors the sample temperature. It has a static temperature offset when reading the droplet temperature due to the physical separation, but the relative value adequately indicates the droplet temperature rise/drop activities. The design temperature rise (<20 °C) of the IC has a negligible effect on the performance of the TRX, as verified by simulations.

4.2.3 Hall Sensor, Readout Circuit, and Current Driver

The micro-NMR TRX integrates a Hall sensor to sense the B0-field variation of the permanent magnet, which must be appeared orthogonally to the B1-field of the pla-nar coil. The detected B0-field variation modulates the current passing through the

Fig. 4.3 (a) Simulated 3D temperature distribution of the droplet at applied power of 8 mW in COMSOL Multiphysics®; (b) Simulated droplet average temperature at applied power from 0 to 20 mW


77

auxiliary coil of the magnet to stabilize the resultant B0-field, which allows system- level calibration.

Hall sensors can detect the magnetic field normal [5–7] or parallel [8, 9] to the chip surface. As the B1-field generated by the planar coil is normal to the chip sur-face and thus the B0-field has to be in the lateral direction, the latter solution is required, which can be achieved by a vertical Hall sensor (VHS). Each VHS sub- element contains an N-well as the substrate and three N-diffusions as contacts (Fig. 4.4a). P-diffusions are embedded between the N-diffusions to avert the current flowing at the surface, soothing the 1/f noise. This architecture renders the device with full standard CMOS compatibility. Working in the current mode implies the input is a constant current source (IBias), and the outputs are two current terminals (IP and IN). When no external magnetic field exists, there is an equal split of IBias between IP and IN.

When considering B0 in the lateral direction (i.e., normal to the cross section of the VHS device), the charge carriers from the input terminal will experience a Lorentz force deflecting the charges, leading to nonidentical magnitudes of IP and IN (Fig. 4.4b) expressed by:

I

II BP = + ( )BiasHall2 0

(4.1)

I

II BN = − ( )BiasHall2 0

(4.2)

where IHall(B0) is the induced Hall current on each output terminal commensurate with B0. Thus, the measurement of IHall can determine B0. Yet, IHall can be much smaller than IBias. Such prodigious bias component stiffens the measurement on IP and IN. To circumvent this measurement barrier, four identical VHS sub-elements U1–4 are arranged to form a Wheatstone bridge (Fig. 4.5). The Wheatstone bridge prunes IBias at the output, and only the B0-dependent term IHall appears at the Wheatstone bridge’s output. Furthermore, this configuration not only features a fully differential architecture but also doubles the output Hall current improving the sensitivity of the VHS. A common-mode feedback circuit regulates the tail of the Wheatstone bridge, where two-phase spinning eliminates the effect of mismatch between the VHS sub-elements by periodically interchanging the output and supply terminals of the Wheatstone bridge (Fig. 4.6).

IBias IP IPINIBiasIN

B0

(a) (b)Current Path Current PathN-Well

P+ diffusion

N+ diffusion

Fig. 4.4 The cross section of a single VHS element and its current path. (a) Without lateral mag-netic field; (b) with lateral magnetic field B0

4.2 Platform Design

78

Hall Sensor Equivalent Model

+- +

-

To TIA

ResetPhase

IIN,P & IIN,N

ResetPhase

ReadingPhase

Measuring Phase(Integrating)

S1 - S2S3 - S4

S5 - S8

S9 - S12

S13 - S16

fCLKCCAL(VTUNE-VCM)

IBIAS

Current Path(w/o B0-Field)Current Path(w/ B0-Field)

CMFB

-2+IBIAS

IBIAS

IBIAS IHALL

IHALL IHALL IIN,P =2-IHALL

IIN,N =2-IHALLIHALL

IBIAS

IBIAS

2

-2 +2

1/fCLK

Nominal B0-FieldCompensator

Timing Diagram

R0: Intrinsic resistanceα: magnetic resistance coefficient

U1

U4

U3

U2

S7

S3

S1

CF

CF

CCAL

CCAL

S2

S13

VCM VCM

VTUNE

VOUT,P

VOUT,N

VCM

VOUT,P(VOUT,P)

VCM

VCAL+

VCAL-ICAL-

ICAL-ICAL+

ICAL+

VCAL+

VCAL-

ICAL+ ≈ ICAL- =

S10

S15 S16

S12

S11S14S9

S4

S5

S6

S8

Fig. 4.5 Proposed current-mode fourfolded VHS arranged in Wheatstone bridge to sense the lat-eral B0-field and its readout circuit (spinning circuitry is omitted for simplicity). The latter features a nominal B0-field compensator to offset the strong nominal B0-field (0.46 T) for better sensitivity (3.75 mT). The green arrows highlight the current paths of IHall. Inset shows the timing diagram for the switches and overall operations

IIN,PIIN,N

IBias IBias

IBiasIBias

Ø1

U4

U1

U3

U2 U4

U3

U2

U1

IIN,P IIN,N

Ø2

Fig. 4.6 Illustration for the two-phase spinning technique on the VHS. The bias direction (U1 and U3) together with the output terminals (U2 and U4) of the VHS is swapped periodically to eliminate the 1/f noise and offset of the elements


79

The induced differential currents from the Wheatstone bridge (IIN , P and IIN , N) are then converted to voltages for recording. Among multifarious options of tran-simpedance amplifier (TIA), the current integrator formed by a high-gain amplifier with shunt integrating capacitor (CF) appears as a promising solution, since it inher-ently offers low-pass filtering on the outputs without passive noise sources (e.g., feedback resistors for resistive feedback TIA), leading to a better noise performance [10–12]. Furthermore, the variation of the integration time TINT can alter the gain of the fully differential TIA, providing conversion flexibility (Fig. 4.7). The core of this TIA is a two-stage amplifier with a telescopic first stage, where the DC gain (ADC, 100 dB) guarantees accurate and stable operation (GBW = 100 MHz, PM = 50° at 15-pF loads), with a power budget of 2 mW excluding the bias circuit. A chopper deals with the offset and 1/f noise of the amplifier with a chopping fre-quency of 1 MHz. This chopping technique reduces the 1/f noise corner by 10,000× (from 200 kHz to 20 Hz) in an open-loop configuration. During the reset phase, S1–4 nullify the residual voltages on CF, while opening S5–8 prevents the current to flow from the Wheatstone bridge into the TIA. In the measuring phase, IIN , P and IIN , N flow through S7–8 and charge CF, causing complementary voltage ramps at the outputs of the TIA, expressed by (assuming an ideal TIA):

V V V

I T

CFOUT OUT P OUT N

Hall INT= − =, ,

4

(4.3)

For an ideal TIA, its input resistance (RIN , TIA) should be zero to accurately mea-sure its input current. The finite gain of the amplifier and the intrinsic resistances of the switches (RSW), however, ineluctably exacerbate the performance of the TIA. Considering a TIA with n switches to regulate the current flow from the Wheatstone bridge and the spinning circuitry in its front, the average RIN , TIA below the resetting frequency (1/TINT) will be:

0

80

120

160

Tran

sim

peda

nce

Gai

n (d

BW

)1 10 100 1k 10k 100k 1M

Frequency (Hz)

40

CF = 8 pF

TINT = 100 µsTINT = 50 µsTINT = 25 µs

-20 dB /dec

Fig. 4.7 Simulated frequency response of the TIA with various TINT

4.2 Platform Design

80

R

T

A Cn R

FIN TIA conv

INT

DCSW, , = + ⋅

(4.4)

assuming a negligible reset time relative to TINT. The first term on the right of (4.4) is the input impedance of the integrator, whereas the second term emerges from the switches. While a multistage architecture can boost the gain of the amplifier at the expense of power consumption [13], the value of RSW is limited to a certain range in a particular process since a large switch inextricably induces parasitic capacitance on the signal paths (Fig. 4.8). To this end, it is mandatory to diminish the effect of RSW on the TIA. To achieve this, S7–8 manage to guide the current passing through the feedback paths, while avoiding switches between the TIA and Wheatstone bridge. S5–6 provide connections between the input of the amplifier and the Wheatstone bridge. As long as there is no current (except leakage currents) flowing through S5–6, the operation of the TIA will not be deteriorated. RIN,TIA in the pro-posed architecture is given by:

R

T

A C

n R

AFIN TIA prop

INT

DC

SW

DC, , = +

⋅

(4.5)

Where, in this switching scheme, ADC represses the impact of the switches and allows the realization of a low input impedance TIA without creating a large para-sitic capacitance at the signal paths. For example, RIN , TIA , prop = 171 Ω (simulated) with CF = 8 pF, TINT = 100 μs (gain, 250 V/μA), and RSW = 240 Ω. When compared with the general approach that places four switches (two for mastering and two for spinning) before the TIA [7], the suppression of RIN , TIA is now 85%, while the TIA absorbs ~20% more current (output impedance from Wheatstone bridge, 5 kΩ). The IRN of the TIA is 10 pA/√Hz at 1 Hz, which is 79× better than the case without the chopper. After the measuring phase, the Hall currents from the Wheatstone bridge cease by switching off S5–8, while turning on S3–4 allows the reading of the voltages on CF. The three-phase operation can be repeated, and the

10k

1k

100

Cha

nnel

Res

ista

nce

(W)

0.5Channel Width (µm)

Channel length: 0.18µmVGS= 0.9V

1000

10

1

Total Capacitance (fF)

505

100

10

Fig. 4.8 Simulated channel resistance (RDS) and parasitic capacitance (CS+CD) of the MOS versus channel width


81

results can be averaged to reduce the background noise. To calibrate the tempera-ture error on the VHS, a temperature sensor is entailed to provide information for calibration off-chip.

The extension of TINT produces an acceptable output level (e.g., LSB of the exter-nal ADC, 0.8 mV) induced by the B0-field variation (3.75 mT). Yet, ascribed to the prodigious nominal B0-field, a typical TIA can saturate and fail to sense this tiny B0-field variation. A nominal B0-field compensator implemented by a pair of switched-capacitor based current source with alternating phase (i.e., opposite cur-rent flow) is necessary to obviate this contrariety. During the measuring phase, the compensator nullifies the Hall current induced from the nominal B0-field flowing into the TIA by generating and injecting the currents (ICAL+ and ICAL−) into the TIA. Thus, there is only the integration of the variable part from the VHS at the output of the TIA (Fig. 4.9), and by TINT extension, more gain is achieved without saturating the TIA.

The current driver receives the converted results after digitization and subse-quently drives the auxiliary coil of the portable magnet and trims the B0-field. Based on the Biot–Savart law, this injected current to the coil generates an additional mag-netic field (75 mT/A). It modulates and stabilizes the actual B0-field, which is the sum of the B0-field from the permanent magnet and auxiliary coil [14], with the applied current according to the results from the Hall sensor. The current driver implemented off-chip with discrete components prevents its interference (on-chip) due to the related high power consumption.

Fig. 4.9 Simulated output waveforms of the integrator. Without the nominal B0-field compensator, the integrator is saturated due to the large current induced by the nominal B0-field before it accu-mulates an adequate voltage difference. Whereas with the compensator, the nominal B0-field can be compensated; thus, the integration time can be prolonged to produce sufficient voltage differ-ences at the output

4.2 Platform Design

82

4.3 Prototype and Experimental Results

4.3.1 Experimental Setup and Electrical Measurements

Fabricated in the 1P6M 0.18-μm CMOS process, the chip occupies an area of 2.0 × 3.8 mm2, dominated by the planar coil (dimension, 2.0 × 2.0 mm2; L, 506 nH; Q, 1.84) as shown in Fig. 4.10a. The dimension (14 × 6 × 11 cm3) and weight (1.4 kg) of the system are dominated by the 0.46-T portable magnet (weight, 1.25 kg; diameter, 8 cm; height, 5.5 cm). In addition to the CMOS chip, there are a system PCB, a commercial FPGA (DE0-Nano), and a current driver (Fig. 4.10b). A customized program in the PC controls the platform, simplifying the operation and

Fig. 4.10 (a) Chip photo of the fabricated chip in 0.18-μm CMOS. (b) Prototype of the micro- NMR platform with B0-field stabilization and lab-on-a-chip feasibility for multi-type biological/chemical assays, including (1) permanent magnet, (2) CMOS micro-NMR chip (inside magnet), (3) PCB, (4) FPGA, and (5) current driver. (c) Experimental setup. A program developed in C# is entailed for hardware control and visualizing the experimental results. The platform is powered by two batteries for portability


83

visualization of the assay results (Fig. 4.10c). The computer extracts and evaluates the T2 of the samples from the acquired signals. During the micro-NMR experi-ments, the RX mode consumes 22.1 mW, dominated by the forefront LNA for high sensitivity. For TX mode, the power consumption is 51.0 mW, dominated by the PA that has a power efficiency of 31.6%. The VHS together with the readout circuit draws 8.4 mW. Both the TRX and Hall sensor parts switch into the idle mode after each assay to prolong the lifetime of the battery. With two AA batteries of 2.5 Wh capacity and a DC-DC converter efficiency of 90%, the system can finish up to 500,000 assays, assuming that the VHS is on for 1 s, then the NMR RX is on for 1 s (TX pulses are short such that the power of TX can be neglected in this case), and finally the entire system is idle for 9 s.

Before each micro-NMR assay, the Hall sensor is turned on first with the VHS sensing the B0-field (Fig. 4.11). As the B0-field may shift away from its nominal value due to environmental variation (e.g., temperature and sample-to-magnet posi-tion), an untracked fL can be easily off-center from fEXC (BW, 16.7 kHz, equivalent to 0.5 mT in terms of B0-field), and the operation of the platform can be paralyzed. With the proposed calibration scheme, the VHS and readout circuit track the B0- field variation; they show a sensitivity of 4.12 V/T (Fig. 4.12a). The eventual B0- field is then balanced by modulating the auxiliary coil of the magnet with a particular magnitude of DC current, according to the result from the VHS and readout circuit. Thus, fL can be reset to match with fEXC to proceed the micro-NMR assay. Associated with signal-averaging performed in the frequency domain to suppress the back-ground noise on the Hall sensor, the proposed calibration improves the B0-field stability by 13× (from 2 to 0.15 mT) at 0.46 T nominal B0-field (fL = 19.6 MHz), corresponding to a variation on fL of 6.9 kHz (Fig. 4.12b). This variation is smaller than the BW of the excitation pulse, ensuring a secure micro-NMR operation. While this calibration scheme can handle large magnetic field fluctuation, the calibration for small magnetic field fluctuation such as those proposed in [3] can be performed in parallel to enhance the resolution if necessary.

µNMRExperimentB0-field w/oCalibration

B0-field w/Calibration

Hall SensorOutput

Off OnOn

Proton's LarmorFrequency

FrequencyDomain

Timing Diagram of the B0-Field Calibration

Off-center from fEXC

Off-center from fEXC

Drifting (e.g., temperature variation)Cal. on

<1s >10s <1s Spectrum of theexcitation pulses

Effective BW »16.7 kHz

Fig. 4.11 Timing diagram of the B0-field calibration and its frequency-domain illustration


84

The stabilized fL allows the use of a simple XO as the LO, which measures low power (79 μW at VDD of 0.9 V) and low phase noise (−116 dBc/Hz at 1 kHz offset), as shown in Fig. 4.13a, b, respectively.

4.3.2 Biological/Chemical Measurements

One of the crucial aims of this platform is to detect and quantify the biological target inside the samples for disease screening. Human immunoglobulin G (IgG), which protects the body from infections by binding themselves to different pathogens, was exemplified for measurement. Protein A, which specifically binds human IgG, is used as a probe to detect the existence of IgG inside the samples by functionalizing them on the water-soluble MNPs ([Fe2O3], 10 μg/mL; Ø, 25–30 nm). As illustrated

Measured data

Fitted line

B0-Field (mT)

Volta

ge (m

V)

457 458 459 460 4610

4

8

12

16

456

y = 4.12x – 1880.5R2= 0.999

20Hall Sensor Response

458

460

462

458.6 459 459.4 459.8 460.2Permanent Magnet B0-Field (mT)

Act

ual B

0-Fi

eld

(mT)

461

459

460.6

w/o Calibration

w/ Calibration Operable region

(a) (b)

Fig. 4.12 (a) Measured hall sensor response; (b) B0-field with and without calibration. Actual B0- field is the sum of the B0-fields from the permanent magnet and the auxiliary coil driven by the current driver

0

200

400

600

800

-200

-195

-190

Phas

e N

oise

(dB

c/H

z)

-185

-180

Pow

er C

onsu

mpt

ion

(µW

)

FoM @

1kHz (dB

c/Hz)

VDD(V)(a)

0.9 1 1.1 1.3

f = 78.5 MHz

1.2-160

-140

-120

-100

-80

1 10 100 1000

Offset Frequency (kHz)

CrystalOscillator(VDD = 0.9V)

SignalGenerator(33250A)

(b)

Fig. 4.13 (a) Measured power consumption and FoM of the XO at different supply voltages. (b) Measured phase noise of the XO (VDD = 0.9 V, f = 78.5 MHz). Compared with the LO generated from signal generator (Agilent 3350A), the XO shows a much better phase noise at low power


85

in Sect. 1.3, these superparamagnetic MNPs have significant impacts on T2 of the samples according to the target concentration of the samples. When IgG is absent in the sample, the MNPs stay monodisperse inside the solution, yielding a T2 of 258 ms. Consequently, when there is IgG in the samples, the human IgG cross-links with the protein A on the MNPs, assembling nanoparticle micro-clusters. In this respect, T2 of the sample is linked with the amount of IgG upon nanoparticles agglomeration. For instance, T2 of the sample drops to 232.2 ms when the concen-tration of IgG is 12.5 nM inside the sample (Fig. 4.14a). To evince the selectivity of the NMR-based biomolecule screening, chicken immunoglobulin Y (IgY), which does not conjugate with the Protein A, is also tested with the platform. T2 from the samples have negligible change (<2%) from varying concentration of chicken IgY (5–50 nM), thus validating the selectivity of the assay.

Alternatively, by selecting the corresponding probe functionalized MNP, this versatile NMR-based screening scheme is capable of sensing widespread biomole-cules. This is manifested by the detection of DNA for life-threatening bacteria screening. With a similar sensing mechanism to the case for detecting human IgG, the platform quantifies the synthesized DNA derived from Enterococcus faecalis by pair of probe-decorated MNPs (Ø, 100 nm). The limit of detection (LoD) of DNA for this platform, which is defined as ΔT2 = 3σ above the blank sample, is estimated to be <100 pM (Fig. 4.14b). Additionally, the dynamic range of the detection can be impelled to 125 nM of DNA by varying the concentration of the MNP (from 6.25 to 10 μg/mL). The response of the assay to single-nucleotide polymorphism is indis-tinguishable to T2 baseline (<4%), substantiating the possibility of differentiation of a single-base mismatch DNA. The reason why DNA assay has a better sensitivity compared to the IgG assay may attribute to the different size, number of binding sites, and relaxivity of the MNP. In fact, by optimizing these parameters, the detec-tion limit of the micro-NMR platform can be further decreased to ~fM level [15].

One of the unique features of the NMR is its capability to probe the molecular structure of the samples, which renders this micro-NMR platform a versatile tool for

Immunoglobulin G E. faecalis derived DNA

(a) (b)

0

10

20

30

10 20 30 40 500

Human IgG

Chicken IgY

[Ig] (nM)

MNP (Ø:30nm)

Protein A

Human IgG

5

15

25

- DT 2

(%)

6.25 10[Fe

2O

3] (µg/mL)

Matched DNASingle-Base

mismatch DNA0.1 1 10 100 1000

-40

0

40

80

120

[DNA] (nM)

Nor

mal

ized

- DT

2 (%

)

MNP (Ø:100nm)Target DNA

Probe

0.01

Fig. 4.14 Experimental results from biological samples. (a) Target quantification from human IgG as target and chicken IgY as control. (b) Target quantification from Enterococcus faecalis- derived DNA together with single-base mismatch DNA


86

different areas of application. For instance, protein β-lactoglobulin (β-LG), which is a major protein found in milk products, has significant impacts on the texture as well as the nutritional quality of the food [16]. β-LG exists as a native dimer at room tem-perature. When they experience heat >60 °C, β-LG denatures and aggregates as glob-ules irreversibly. This state transformation decreases T2 of the sample, as the transverse relaxation of protons is sensitive to the aggregation. In fact, T2 from the samples drops from 40% to 74% when heated to 95 °C according to their concentration (Fig. 4.15a). Thus, this T2 deviation can indicate the state of the whey proteins of foods.

Similarly, the adoption of this platform allows the detection of the dynamics between the solvent and polymer for the colloidal application. Poly(N- isopropylacrylamide) (PNIPAM), which is widely used as the advanced sensor and drug delivery carrier, demonstrates temperature-induced volume phase transition at lower critical solution temperature (~33 °C) in the water [17]. This phase transition affects the local environment in terms of solvent confinement and subsequently T2 of the solvent. To diversify the capabilities of this platform, the NMR sensing coil is utilized as a thermal heater to perform thermal profiling on the PNIPAM. When the applied power to the heater is low (duty cycle ≤2%), the temperature of the sample does not rise and the PNIPAM stays in a swollen state. If the power is increased gradually (duty cycle ≥4%), the PNIPAM starts to undergo a volume phase transi-tion. This collapsed state of PNIPAM confounds the mobility of bound water mol-ecule then resulting in the T2 decrement of the solvent (~9%) (Fig. 4.15b).

4.3.3 Comparison and Discussion

Table 4.1 compares this work with recent CMOS PoU tools. It supports multi-type assays in one unified platform while achieving high sensitivity and selectivity for DNA, as well as other proteins targeting capability inside 2.5 μL sample with cor-responding probe-decorated MNPs. The NMR-based sensing scheme eases the

1.4

1.5

1.6

1.7

0 4 8 12Collapsed

Duty Cycle (%)20 40 60 80 100

0.4

0.8

1.2

1.6

0

T 2 (s

)

T 2 (s

)

Temperature (˚C)

Control (water)10mg/mL20mg/mL40mg/mL

Poly(N-isopropylacrylamide)β-lactoglobulin

n

NHO

H3C CH

3

Heating

Cooling

NativeDimer

Globule

Heating

Swollen

(a) (b)

Fig. 4.15 Experimental results from biological/chemical samples. (a) Protein (β-LG) state detec-tion with different heating temperature. (b) Polymer (PNIPAM) dynamics with the solvent during heating from the on-chip heater


87

hardware and sample preparation for the assays (i.e., post-processing and probe immobilization on CMOS, tag labeling on target, and washing of MPs) when com-pared to others such as fluorescent/capacitive sensing. While compared to the previ-ously reported CMOS NMR systems, the B0-field stabilization module here facilitates the electronics and enhances the robustness of the platform (Table 4.2). Moreover, the micro-NNR platform includes a thermal module, which is an added feature to explore the thermal profile of the samples and opens up a broad range of possibilities for on-chip thermal analysis [18, 19]. Yet, due to the large sensing coil, the imple-mentation of microarray for multiplex sensing is unfeasible. In addition, limited by the homogeneity of the magnet, NMR spectroscopy, as demonstrated by Ha et al. [1], is not supported by this platform. Still, the NMR relaxometry applications demon-strated here render this platform as a promising candidate for PoU tool. A compari-son with a commercial benchtop relaxometer shows that this platform consumes 120× fewer samples and is 96× lighter, 175× smaller, and 16× cheaper when evalu-ated against the selling price of [20], with a trade-off of lower tuning range on sam-ple temperature. Together with the proposed B0-field stabilization scheme that facilitates the operations, this system is in the vanguard of CMOS PoU tools. Limited by the maximum rating of the current injected to the auxiliary coil, the temperature range of the operation is confounded. By entailing multiple XO with different oscil-lating frequency or even with PLL, the operating range can be increased. In addition,

Table 4.1 Summary and benchmark with other CMOS-based PoU systems

Sensing mechanism

No. of sensors

Demonstrated detection

Probe immobilization Labeling

Technology (chip area)

This work

NMR relaxometry

1 DNA and protein detection

No Quasi label- free

0.18 μm (7.6 mm2)

[21] Fluorescent sensing

56 DNA detection Yes Cy3- label

0.35 μm (9.0 mm2)

[22] Capacitive sensing

54 DNA detection Yes Label- free

0.35 μm (20.0 mm2)

[23] Cyclic voltammetry

12 DNA detection Yes Label- free

0.13 μm (9.0 mm2)

[24] Magnetic sensing

64 Protein detection

Yes Quasi label- free

0.35 μm (8.9 mm2)

[25]a Magnetic sensing

256 Protein detection

Yes Quasi label- free

0.18 μm (7.29 mm2)

[26]a NMR relaxometry

1 Protein detection


0.18 μm (3.8 mm2)

[4] NMR relaxometry

1 Protein and cell detection


0.18 μm (11.3 mm2)

aOff-chip sensors


88

a systematic study on the VHS including the temperature dependency and chip-to-chip variation would be beneficial to the robustness of the system.

4.4 Summary

By incorporating the Hall sensor and micro-NMR TRX, this chapter described a handheld CMOS micro-NMR platform with effective closed-loop B0-field stabiliza-tion and evinced superiority of the NMR-based sensing for PoU assay from differ-ent points of view [27, 28]. In terms of applications, this platform exhibited the utilization of a single CMOS chip to attain versatile sensitive chemical/biological assay from diverse unprocessed samples such as protein, DNA, and polymer. The LoD of the platform for Enterococcus faecalis-derived DNA is <100 pM from a 2.5-μL sample using functionalized MNPs. The developed thermal module also extends the potential applications (e.g., polymer dynamics study). In terms of hard-ware, the micro-NMR platform reveals high robustness confirmed by the explora-tion of the current-mode Hall sensor and readout circuit in conjunction with a current driver to stabilize the B0-field variation attributed to the environments. This handheld micro-NMR platform is anticipated to popularize the application of NMR and magnify the advantages of NMR to our daily lives.

Table 4.2 Benchmark with previous CMOS NMR systems

This work JSSC’09 [26]a JSSC’11 [4]

Application Specificity 1. Target detection2. Solvent-polymer dynamics3. Protein state analysis

Target detection

Target detection

Demo target DNA/human IgG Avidin hCG cancer marker/bladder cancer cell

Detection limit

<100 pM (DNA) 5 nM (avidin) 5 nM (cancer marker)

Sample handling limit

2.5 μL 5.0 μL 5.0 μL

Hardware Physics NMR relaxometry + thermal management

NMR relaxometry

NMR relaxometry

LO generation

Crystal oscillator (off-chip crystal)

On-chip/off-chip

Off-chip

B0-field calibration

Yes No No

CMOS tech. 0.18 μm 0.18 μm 0.18 μmChip area 7.6 mm2 3.8 mm2 11.3 mm2

aOff-chip sensors


89

References

1. D. Ha, J. Paulsen, N. Sun, Y.Q. Song, D. Ham, Scalable NMR spectroscopy with semiconduc-tor chips. Proc. Natl. Acad. Sci. 111(33), 11955–11960 (2014)

2. G.A. Morris, H. Barjat, T.J. Horne, Reference deconvolution methods. Prog. Nucl. Magn. Reson. Spectrosc. 31(1), 197–257 (1997)

3. E. Kupce, R. Freeman, Molecular structure from a single NMR sequence (fast-PANACEA). J. Magn. Reson. 206(1), 147–153 (2010)

4. N. Sun, T.J. Yoon, H. Lee, W. Andress, R. Weissleder, D. Ham, Palm NMR and 1-chip NMR. IEEE J. Solid State Circuits 46(1), 342–352 (2011)

5. J. Jiang, K. Makinwa, A hybrid multipath CMOS magnetic sensor with 210μTrms resolu-tion and 3MHz bandwidth for contactless current sensing, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2016, pp. 204–205

6. J.F. Jiang, W.J. Kindt, K.A.A. Makinwa, A continuous-time ripple reduction technique for spinning-current Hall sensors. IEEE J. Solid State Circuits 49(7), 1525–1534 (2014)

7. H. Heidari, E. Bonizzoni, U. Gatti, F. Maloberti, A CMOS current-mode magnetic Hall sensor with integrated front-end. IEEE Trans. Circuits Syst. I, Reg. Papers 62(5), 1270–1278 (2015)

8. G.M. Sung, C.P. Yu, 2-D differential folded vertical Hall device fabricated on a p-type sub-strate using CMOS technology. IEEE Sensors J. 13(6), 2253–2262 (2013)

9. C. Sander, M.C. Vecchi, M. Cornils, O. Paul, From three-contact vertical Hall elements to symmetrized vertical Hall sensors with low offset. Sens. Actuators A 240, 92–102 (2016)

10. K.M. Lei, H. Heidari, P.I. Mak, M.K. Law, F. Maloberti, Exploring the noise limits of fully-differential micro-watt transimpedance amplifiers for Sub-pA/√Hz sensitivity, in 11th Conference on Ph.D. Research in Microelectronics and Electronicss (PRIME), 2015, pp. 290–293

11. D. Kim, B. Goldstein, W. Tang, F.J. Sigworth, E. Culurciello, Noise analysis and performance comparison of low current measurement systems for biomedical applications. IEEE Trans. Biomed. Circuits Syst. 7(1), 52–62 (2013)

12. M. Crescentini, M. Bennati, M. Carminati, M. Tartagni, Noise limits of CMOS current inter-faces for biosensors: a review. IEEE Trans. Biomed. Circuits Syst. 8(2), 278–292 (2014)

13. K.N. Leung, P.K.T. Mok, Analysis of multistage amplifier-frequency compensation. IEEE Trans. Circuits Syst. I, Fundam. Theory Appl. 48(9), 1041–1056 (2001)

14. Datasheet of NMR permanent magnet PM-1055, Available: http://metrolab.com/wp-content/uploads/2015/07/PM1055_broch.pdf. Accessed 30 Jan 2016


16. L. Indrawati, R.L. Stroshine, G. Narsimhan, Low-field NMR: a tool for studying protein aggre-gation. J. Sci. Food Agric. 87(12), 2207–2216 (2007)

17. B. Sierra-Martin, J.R. Retama, M. Laurenti, A.F. Barbero, E.L. Cabarcos, Structure and poly-mer dynamics within PNIPAM-based microgel particles. Adv. Colloid Interf. Sci. 205, 113–123 (2014)

18. D.H. Gultekin, J.C. Gore, Temperature dependence of nuclear magnetization and relaxation. J. Magn. Reson. 172(1), 133–141 (2005)

19. L. Vermeir, M. Balcaen, P. Sabatino, K. Dewettinck, P. Van der Meeren, Influence of molecular exchange on the enclosed water volume fraction of W/O/W double emulsions as determined by low-resolution NMR diffusometry and T-2-relaxometry. Colloids Surf. A Physicochem. Eng. Asp. 456(1), 129–138 (2014)

20. Bruker minispec contrast agent analyzer, Available: https://www.bruker.com/products/mr/td-nmr/minispec-mq-series/mq-contrast-agent-analyzer/overview.html. Accessed 30 Jan 2016

21. B. Jang, P. Cao, A. Chevalier, A. Ellington, A. Hassibi, A CMOS fluorescent-based biosen-sor microarray, in International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2009, pp. 436–437

References

http://metrolab.com/wp-content/uploads/2015/07/PM1055_broch.pdf

http://metrolab.com/wp-content/uploads/2015/07/PM1055_broch.pdf

https://www.bruker.com/products/mr/td-nmr/minispec-mq-series/mq-contrast-agent-analyzer/overview.html

https://www.bruker.com/products/mr/td-nmr/minispec-mq-series/mq-contrast-agent-analyzer/overview.html

90

22. K.H. Lee, S. Choi, J.O. Lee, J.B. Yoon, G.H. Cho, CMOS capacitive biosensor with enhanced sensitivity for label-free DNA detection, in International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2012, pp. 120–122

23. H. Jafari, L. Soleymani, R. Genov, 16-channel CMOS impedance spectroscopy DNA analyzer with dual-slope multiplying ADCs. IEEE Trans. Biomed. Circuits Syst. 6(5), 468–478 (2012)

24. P.H. Kuo, J.C. Kuo, H.T. Hsueh, J.Y. Hsieh, Y.C. Huang, T. Wang, et al., A smart CMOS assay SoC for rapid blood screening test of risk prediction. IEEE Trans. Biomed. Circuits Syst. 9(6), 790–800 (2015)

25. D.A. Hall, R.S. Gaster, K.A.A. Makinwa, S.X. Wang, B. Murmann, A 256 pixel magnetoresis-tive biosensor microarray in 0.18μm CMOS. IEEE J. Solid State Circuits 48(5), 1290–1301 (2013)


27. K.-M. Lei, H. Heidari, P.-I. Mak, M.-K. Law, F. Maloberti, R.P. Martins, A handheld 50pM- sensitivity micro-NMR CMOS platform with B-field stabilization for multi-type biological/chemical assays, in International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2016, pp. 474–475

28. K.-M. Lei, H. Heidari, P.-I. Mak, M.-K. Law, F. Maloberti, R.P. Martins, A handheld high- sensitivity micro-NMR CMOS platform with B-field stabilization for multi-type biological/chemical assays. IEEE J. Solid State Circuits 52(1), 284–297 (2017)



Chapter 5Conclusion and Outlook

5.1 Summary of Researches

In retrospect, this book firstly reviewed recent CMOS biosensors for in vitro diag-nosis according to their transducing mechanisms (Chap. 2). Among these, NMR- based detection is a promising transducing mechanism as it averts complex post-processing on the CMOS chip and surface modification process to immobilize the probe. Still, there exist some challenges such as magnetic field instability and arduous sample management inside the space-limiting magnet. Thus, the essences of the researches around this book are to implement the NMR system on CMOS chips and streamline the experimental procedure.

In Chap. 3, two portable NMR relaxometers capable of handling the sample under NMR assay were demonstrated. The first one is a modular platform for proof of concept and validation for the integration of NMR (diagnosis) and DMF (sample management). The second platform equips with a CMOS TRX to reduce the overall dimensions of the module while improving the sensitivity. Inside the magnet, the electronic-automated DMF device with closed-loop capacitive feedback manages multiple droplet samples in real-time and can be reconfigured by software. The micro-NMR relaxometer is competent to achieve the real-time quantification of chemical/biological analytes in sub-10-μL samples, capable of manipulating mul-tiple samples automatically and performing multistep experiments inside the space- limiting magnet effectively.

In Chap. 4, a micro-NMR system compatible with multi-type biological/chemi-cal assays was reported. It is unified in a handheld CMOS platform with limit of detection < 100 pM for Enterococcus faecalis-derived DNA. It features a dynamic B1-field TRX, a thermal-controlled spiral coil, and a VHS to aid stabilizing the static B0-field fluctuation (<0.15mT). The system demos (1) target detection, (2) protein state analysis, and (3) solvent–polymer dynamics for healthcare, food, and colloidal applications, respectively.

92

5.2 Future Prospects

Driven by the world booming and longevous population in the coming century, there will be a huge market for efficient, low-cost, and easy-to-use PoC devices for rapid yet sensitive in vitro diagnosis in both developed and developing countries. The CMOS technology offers a platform to implement those IVD tools attributed to its capability for monolithically integrating the transducers and the signal process-ing units on the same chip, which can be massively produced to minimize the cost. Further, benefitting from the continuous improvement of the CMOS technology and post-processing techniques, a multifaceted variety of biosensors have been pro-posed with superior performance. There have been different ways to achieve molec-ular and cellular biosensing on CMOS chips. The necessity and complexity of each transducing mechanism entail a concern or balance of different design goals (e.g., cost) and constraints. Thus, an ideal CMOS biosensor should require profound knowledge not only in microelectronics but also in biology and chemistry to secure a successful multidisciplinary research.

The advances in CMOS biosensors will continue with the downscaling of CMOS technology, offering a higher integration level, better signal processing for lower detection limit, and supporting more complex operation such as next generation sequencing [1]. Yet, the post-processing of the CMOS device should be limited, and the labeling of the signaling tag should be avoided to suppress the manufacturing cost and the preparation time of the experiments, in order that the overall cost can be competitive with other PoC technologies such as test-strip devices. Further, it is also of importance to manage the sample matrix effects of the biosensors from com-plex media and include different functions such as multiplexing [2] and DNA amplification [3, 4] in an integrated platform along with the CMOS biosensors. In addition, as sample management is still a challenge for CMOS chips, extra efforts should be entailed to seamlessly integrate the sample management network and the CMOS biosensor with the aid of channel microfluidic network [5–7] or digital microfluidic array [8], rendering these CMOS biosensors as truly LOC platforms. In our opinion, these CMOS biosensors will provide ultimately the global patients an efficient and powerful IVD solution to improve their living quality.

References

1. D.A. Hall, J.S. Daniels, B. Geuskens, N. Tayebi, G.M. Credo, D.J. Liu, et al., A nanogap transducer array on 32nm CMOS for electrochemical DNA sequencing, in IEEE International Solid-State Circuits Conference (ISSCC) Digest of Technical Papers, 2016, pp. 288–289

2. L. Sandeau, C. Vuillaume, S. Contie, E. Grinenval, F. Belloni, H. Rigneault, et al., Large area CMOS bio-pixel array for compact high sensitive multiplex biosensing. Lab Chip 15(3), 877–881 (2015)

3. H. Norian, R.M. Field, I. Kymissis, K.L. Shepard, An integrated CMOS quantitative- polymerase- chain-reaction lab-on-chip for point-of-care diagnostics. Lab Chip 14(20), 4076–4084 (2014)

5 Conclusion and Outlook

93

4. C. Toumazou, L.M. Shepherd, S.C. Reed, G.I. Chen, A. Patel, D.M. Garner, et al., Simultaneous DNA amplification and detection using a pH-sensing semiconductor system. Nat. Methods 10(7), 641–646 (2013)

5. Y. Huang, A.J. Mason, Lab-on-CMOS integration of microfluidics and electrochemical sen-sors. Lab Chip 13(19), 3929–3934 (2013)

6. R.R. Singh, L. Leng, A. Guenther, R. Genov, A CMOS-microfluidic chemiluminescence con-tact imaging microsystem. IEEE J. Solid State Circuits 47(11), 2822–2833 (2012)

7. X. Liu, L. Li, A.J. Mason, High-throughput impedance spectroscopy biosensor array chip. Philos. Trans. R. Soc. London, Ser. A 372(2012), 20130107 (2014)

8. G. Wang, D. Teng, Y.T. Lai, Y.W. Lu, Y. Ho, C.Y. Lee, Field-programmable lab-on-a-chip based on microelectrode dot array architecture. IET Nanobiotechnol. 8(3), 163–171 (2014)

References

95© Springer International Publishing AG 2018 K.-M. Lei et al., Handheld Total Chemical and Biological Analysis Systems, https://doi.org/10.1007/978-3-319-67825-2

Appendix A: Modular NMR Electronic Components and Measurement

The forefront amplifier of the modular NMR RX is VCA2615 from Texas Instruments (Dallas, TX). It features a high input impedance of >100 kΩ and is with variable-gain control of 52-dB range. An operational amplifier OPA842 from Texas Instruments (Dallas, TX) is employed to provide additional gain and convert the differential signal into single-ended for frequency downconversion. Mixers TUF- 3HSM+ from Mini-Circuits (Brooklyn, NY) are chosen as the downmixing module. The NMR TX is constructed by simple digital electronics (flip-flops, switches, and buffers) which are products from Texas Instruments (Dallas, TX).

The electrical performances of the NMR electronics were characterized before sample measurements. Figure A.1 shows the gain of the RX measured at different RF frequencies. Sinusoidal RF signals with frequency from 19.9975 to 20.0075 MHz were injected to the RX, and a reference LO signal of 20.0025 MHz was provided for the mixer. The gain of the overall system is stable around 95.7–95.8 dB within ±5 kHz of IF. The gain can be further boosted by increasing the gain of the IF low- pass filter.

For the sensitivity of the RX, the spectrum of the received signal was plotted in Fig. A.2. A 100-nV sinusoidal signal at 20 MHz was injected to the RX. This ampli-tude is similar to the amplitude of the NMR signal. The frequency of the LO was set at 20.0025 MHz and resulted in an IF of 2.5 kHz. The resulting signal contains a fundamental tone with an amplitude of −20 dBm at 2.5 kHz. The noise floor is 30 dB below the injected signal. Thus, the RX is capable of detecting signal with amplitude down to 100 nV.

96

Con

vers

ion

Gai

n (d

B)

Input Frequency (MHz)

Intermediate Frequency (kHz)

95.68

95.7

95.72

95.74

95.76

95.78

95.8

95.82

19.9975 19.9995 20.0015 20.0035 20.0055 20.0075

-5 -3 -1 1 3 5

LO frequency:20.0025 MHz

Fig. A.1 Measured gain of the NMR RX

Frequency (Hz)

-80

-60

-40

-20

0

1500 2000 2500 3000 3500 4000

Am

plitu

de (d

Bm

)

30 dB abovenoise floor

Spectrum of the receiverFig. A.2 Measured output spectrum of the RX with a 100-nV, 20-MHz sinusoidal input

Appendix A: Modular NMR Electronic Components and Measurement


A step-up voltage-to-voltage boost converter built up with LM3478 switching con-troller from Texas Instruments Inc. was used to generate a sufficiently high-voltage signal for electrode actuation. The input power is directly drawn from the FPGA board at 5 V; this act avoids the need of another high-voltage supply for better porta-bility. An oscillator built up with timer ICM7555 from Intersil (Milpitas, CA) is used to generate a square wave of 1 kHz. This square wave is amplified into a 40-V peak-to-peak voltage by a switch pair and then high-pass filtered to remove the DC level for actuating the electrodes. A switch array mastered by the FPGA was used to control the on–off pattern of the electrodes. To reduce the RMS-voltage stress on the electrode so as to minimize the possibility of dielectric breakdown, the driving voltage on an occu-pied electrode is modulated with on (off) duty cycle of 10% (90%). Exemplified in Fig. B.1, after continuous square wave of 3 s acting on the electrode, the pulse acting on the electrode with the droplet is modulated with a turn on–off pattern of 1:9. This modulation technique allows the electrode to strap the droplet and prevents the dielec-tric breakdown of the electrode caused by the long-term voltage stress.

The location of each droplet sample is determined by scanning the derived capacitance CElec of each electrode. As the capacitance between two parallel plates is proportional to the permittivity of its insulating medium, a droplet-occupied elec-trode will increase the capacitance on the corresponding electrode when compared with the air. In this work, a timer ICM7555 working in the astable mode is used to sense the electrode capacitance. The oscillation frequency of the timer is inversely proportional to CElec. Thus, the identification of droplet position can be done by counting the pulses available in a fixed period on each electrode.

Appendix B: DMF Device and Electronics

98

Vext

Vext

3 s 0.9 s 0.1 s 0.9 s 0.1 s

Droplet moving Droplet still

Fig. B.1 Visualized waveform applied to the electrode before and after the droplet arrives at the electrode

Appendix B: DMF Device and Electronics


To facilitate the setting of micro-NMR parameters and route optimization of DMF, a graphic-user-interface program implemented in Visual C# was adopted to master the whole micro-NMR relaxometer, including (i) setting the micro-NMR parame-ters, (ii) displaying the micro-NMR results, (iii) reading the ambient temperature and calibrating the DAC output for magnetic field calibrator, (iv) controlling the switch array for the DMF device, and (v) displaying the vacancy of the electrodes. To this end, an interface is entailed for communications between the FPGA DE0- nano (for hardware control) and the PC (for software computing).

The TTL-232R_PCB module from Future Technology Devices International Limited (United Kingdom) is used to interface between the PC and FPGA. It can read/transmit data from/to FPGA board using the UART (universal asynchronous receiver/transmitter) signals, and the PC will process the data from the module. This protocol can ease the design for both hardware and software levels. As shown in Fig. C.1, the PC sends data to the FPGA using the TTL-232R_PCB module with a unique address. The module will process the command and convert it to a readable format for the FPGA. The FPGA board with a defined address will send the corre-sponding command to the appropriate module. For instance, if the PC set one of the electrode to “ON” state for the DMF device, the FPGA will recognize this com-mand and set the corresponding output to a high level, which will set the accompa-nying switch and drive the electrode for droplet actuation. With the PC, all the necessary control of the micro-NMR relaxometer can be simplified into the soft-ware level. This can eliminate the use of cumbersome hardware such as the micro-processor, providing a neat platform for controlling the micro-NMR relaxometer.

Appendix C: Software and Hardware Interface of Micro-NMR Platform

100

µNMRT2:87ms

DMF TTL-232R_PCB

FPGABoard

Micro-NMR Parameters

Micro-NMR Results

Electrode Actuation

Droplet Sensing

Temp. Sensing

DAC Setting

USBSignals

UARTSignals

Fig. C.1 The communication between the PC and the FPGA board to drive the micro-NMR relax-ometer. It is done by adopting the TTL-232R_PCB module to interfacing between the PC and FPGA board, which mastered the hardware of the micro-NMR relaxometer

Appendix C: Software and Hardware Interface of Micro-NMR Platform


BB0-field stabilization scheme, 73Biot–Savart law, 81

CCarr–Purcell–Meiboom–Gill (CPMG), 5Chemiluminescent signal, 19CMOS-based NMR system, 68CMOS IVD tools

cell-related biosensors, 15DNA-related biosensors, 13electrical-based, 12mechanical-based, 12mechanisms, 11–26NMR-based, 12optical-based, 12protein-related biosensors, 14

CMOS PoU tools, 87Complementary metal–oxide–semiconductor

(CMOS), 1biosensors, 92cantilever, 24capacitive sensing chip, 16–17electrical-based detection, 16Hall sensor array, 21magnetic biosensor array, 21multimodal sensor array, 31

DDC-DC converter, 83Decentralized healthcare systems, 2Digital microfluidics (DMF), 41DMF device integration, 49–51

EElectrical-based detection, 32Electrical-based detection CMOS, 16Electrochemical impedance spectroscopy

(EIS), 12Electronic automation, 41, 50Electronic-automated micro-NMR assay

Boltzmann constant, 55chemical/biological assay, 42coil parameters, 48DMF, 46, 60downconversion, 56electrical measurements, 60–63magnet, 44micro-NMR relaxometer,

66, 67micro-NMR system, 63NMR–DMF system, 42NMR systems, 48–49planar coil, 45RF coil, 42, 44RX and TX, 43system verification, 63–65

Electrowetting-on-dielectric (EWOD), 41

Enzyme-linked immunosorbent assay (ELISA), 18

FFabricated DMF device, 51Field-effect transistors (FET)

biosensor, 16Field-programmable gate array (FPGA)

controls, 43

Index

102

GGilbert cell, 56

HHall sensors, 74, 76, 77, 83Hardware preparation, 33HeLa cells, 19

IIn vitro diagnostic (IVD) tool, 1, 2, 11, 92IVD applications

cell-level diagnosis, 29chemiluminescence, 26DNA hybridization assay, 28–29immunoassay, 26integration level, 32labeling process, 32operation, 33specificity, 34

Indium tin oxide (ITO), 46

LLow-noise amplifier (LNA), 55, 56, 59, 60,

69, 83Local oscillator (LO), 73Low-pass filter (LPF), 57

MMagnetic-based detection CMOS biosensor, 21Magnetic nanoparticles (MNPs), 5Magnet’s field strength, 47Magnetism, 20Mechanical transducer, 23Mechanical-based detection, 23Micro-NMR assay, 83Micro-NMR pulse sequence, 59Micro-NMR system, 53

Micro-NMR TRX, 75Multifunctional planar coil, 76

NNMR-based biosensors, 25NMR-based detection CMOS biosensor, 25NMR–DMF prototype, 51NMR–DMF system, 43NMR equipment, 6Nuclear magnetic resonance (NMR), 4Nutation curve, 49

OOn-chip planar octagonal coil, 76Optical-based detection CMOS, 18

PPoint-of-use (PoU) application, 73Protein β-lactoglobulin (β-LG), 86

SSensing coil, 73Supercritical angle luminescence (SAL), 19Surface acoustic wave (SAW) transducer, 24

TTransimpedance amplifier (TIA), 79Transceiver (TRX), 51

VVertical Hall sensor (VHS), 77

WWheatstone bridge, 79

Index

Documents

Handheld Total Chemical and Biological Analysis Systems: Bridging NMR, Digital Microfluidics, and Semiconductors