CMOS Integrated Lab-on-a-chip System for Personalized Biomedical Diagnosis E-Book

Hao Yu

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Wiley - IEEE
• Sprache: Englisch

Seitenzahl: 453

Veröffentlichungsjahr: 2018

Beschreibung

A thorough examination of lab-on-a-chip circuit-level operations to improve system performance

A rapidly aging population demands rapid, cost-effective, flexible, personalized diagnostics. Existing systems tend to fall short in one or more capacities, making the development of alternatives a priority. CMOS Integrated Lab-on-a-Chip System for Personalized Biomedical Diagnosis provides insight toward the solution, with a comprehensive, multidisciplinary reference to the next wave of personalized medicine technology.

A standard complementary metal oxide semiconductor (CMOS) fabrication technology allows mass-production of large-array, miniaturized CMOS-integrated sensors from multi-modal domains with smart on-chip processing capability. This book provides an in-depth examination of the design and mechanics considerations that make this technology a promising platform for microfluidics, micro-electro-mechanical systems, electronics, and electromagnetics.

From CMOS fundamentals to end-user applications, all aspects of CMOS sensors are covered, with frequent diagrams and illustrations that clarify complex structures and processes. Detailed yet concise, and designed to help students and engineers develop smaller, cheaper, smarter lab-on-a-chip systems, this invaluable reference:

• Provides clarity and insight on the design of lab-on-a-chip personalized biomedical sensors and systems
• Features concise analyses of the integration of microfluidics and micro-electro-mechanical systems
• Highlights the use of compressive sensing, super-resolution, and machine learning through the use of smart SoC processing
• Discusses recent advances in complementary metal oxide semiconductor-integrated lab-on-a-chip systems
• Includes guidance on DNA sequencing and cell counting applications using dual-mode chemical/optical and energy harvesting sensors

The conventional reliance on the microscope, flow cytometry, and DNA sequencing leaves diagnosticians tied to bulky, expensive equipment with a central problem of scale. Lab-on-a-chip technology eliminates these constraints while improving accuracy and flexibility, ushering in a new era of medicine. This book is an essential reference for students, researchers, and engineers working in diagnostic circuitry and microsystems. 

Leseprobe

Table of Contents

Cover

Title Page

Preface

1 Introduction

1.1 Personalized Biomedical Diagnosis

1.2 CMOS Sensor‐based Lab‐on‐a‐Chip for System Miniaturization

1.3 Objectives and Organization of this Book

References

2 CMOS Sensor Design

2.1 Top Architecture

2.2 Noise Overview

2.3 Pixel Readout Circuit

2.4 Column Amplifier

2.5 Column ADC

2.6 Correlated Sampling

2.7 Timing Control

2.8 LVDS Interface

References

3 CMOS Impedance Sensor

3.1 Introduction

3.2 CMOS Impedance Pixel

3.3 Readout Circuit

3.4 A 96 × 96 Electronic Impedance Sensing System

References

4 CMOS Terahertz Sensor

4.1 Introduction

4.2 CMOS THz Pixel

4.3 Readout Circuit

4.4 A 135 GHz Imager

4.5 Plasmonic Sensor for Circulating Tumor Cell Detection

References

5 CMOS Ultrasound Sensor

5.1 Introduction

5.2 CMUT Pixel

5.3 Readout Circuit

5.4 A 320 × 320 CMUT‐based Ultrasound Imaging System

References

6 CMOS 3‐D‐Integrated MEMS Sensor

6.1 Introduction

6.2 MEMS Sensor

6.3 Readout Circuit

6.4 A 3‐D TSV‐less Accelerometer

References

7 CMOS Image Sensor

7.1 Introduction

7.2 CMOS Image Pixel

7.3 Readout Circuit

7.4 A 3.2 Mega CMOS Image Sensor

References

8 CMOS Dual‐mode pH‐Image Sensor

8.1 Introduction

8.2 CMOS Dual‐mode pH‐Image Pixel

8.3 Readout Circuit

8.4 A 64 × 64 Dual‐mode pH‐Image Sensor

References

9 CMOS Dual‐mode Energy‐harvesting‐image Sensor

9.1 Introduction

9.2 CMOS EHI Pixel

9.3 Readout Circuit

9.4 A 96 × 96 EHI Sensing System

References

10 DNA Sequencing

10.1 Introduction

10.2 CMOS ISFET‐based Sequencing

10.3 CMOS THz‐based Genotyping

10.4 Beyond CMOS Nanopore Sequencing

10.5 Summary

References

11 Cell Counting

11.1 Introduction

11.2 Optofluidic Imaging System

11.3 Super‐resolution Image Processing

11.4 Machine‐learning‐based Single‐frame Super‐resolution

11.5 Microfluidic Cytometer for Cell Counting

References

12 Conclusion

12.1 Summaries

12.2 Future Works

Index

End User License Agreement

List of Tables

Chapter 03

Table 3.1 Extracted parameter from equivalent circuit model.

Chapter 05

Table 5.1 Design Parameters for CMUT.

Table 5.2 Measured transceiver front‐end IC performance.

Chapter 06

Table 6.1 Specifications summary of the vertically bonded MEMS/CMOS chip.

Chapter 07

Table 7.1 Performances summary of BSI‐CIS.

Chapter 08

Table 8.1 Dual‐mode sensor chip design specifications.

Table 8.2 Comparison of state‐of‐the‐art ISFET sensors.

Chapter 09

Table 9.1 Performance summary of the EHI imager

Table 9.2 Performance summary of the EHI imager circuit blocks in IM on 1 V supply

Table 9.3 Performance summary in EHM

Table 9.4 Comparison of proposed imager to other EHIs

Chapter 10

Table 10.1 Summary of protein and solid‐state nanopore parameters.

Table 10.2 Summary of DNA detection methods.

Chapter 11

Table 11.1 Pseudo code for ELMSR.

Table 11.2 Pseudo code for CNNSR.

Table 11.3 Measured counting results of mixed RBC and HepG2 sample.

List of Illustrations

Chapter 01

Figure 1.1 Global aging trends for the percentage of the total population at 60 years of age or over in 2012 and 2050 [1].

Figure 1.2 A typical microscope and its optical path with objective lens and eye piece. To reach high‐resolution imaging capability, bulky, expensive, and sophisticated lenses are required.

Figure 1.3 Diagrams of: (a) Optical‐based; and (b) Electrochemical‐based flow cytometry.

Figure 1.4 BD FACSCount™ cell cytometer system.

Figure 1.5 A typical optical DNA sequencer.

Figure 1.6 CMOS‐based LOC integration for biomedical instrument miniaturization.

Figure 1.7 Cross‐section view (left) and symbol (right) of CMOS technology components: NMOS and PMOS.

Figure 1.8 Process flow for the fabrication of an n‐type MOSFET on p‐type silicon.

Figure 1.9 Diagrams of biochemical sensing structures; the sensing parts are: (a) p‐n junction; (b) Passivation; and (c) Metal electrode pair.

Figure 1.10 A typical architecture diagram of a CMOS sensor.

Figure 1.11 The concept of the optical and potential dual‐mode CMOS image sensor.

Figure 1.12 Schematic of the multimodal optical and pH CMOS image sensor.

Figure 1.13 A typical microfluidic channel.

Figure 1.14 Cross‐sectional schematic of a unit cell on digital microfluidic arrays.

Figure 1.15 Microfluidic channel masks designed by AutoCAD.

Figure 1.16 PDMS mold fabrication process.

Figure 1.17 Microfluidic channel fabrication process using the soft lithography technique.

Chapter 02

Figure 2.1 A typical CMOS sensor architecture.

Figure 2.2 Resistor thermal noise model.

Figure 2.3 MOSFET thermal noise model in strong inversion region.

Figure 2.4 MOSFET conductive channel cross‐section view in (a) triode and (b) saturation region.

Figure 2.5 MOSFET flicker noise model.

Figure 2.6 Diode shot noise model.

Figure 2.7 MOSFET noise model with: (a) separate; and (b) equivalent noise.

Figure 2.8 (a) Source follower with active load; and (b) The corresponding small signal circuit.

Figure 2.9 (a) Pixel readout structure of a 2 T source‐follower pixel; and (b) The corresponding transfer curve.

Figure 2.10 Pixel readout structures of (a) a 4 T SF pixel; and (b) a 1.75 T 4‐shared source‐follower pixel. (c) Timing diagram and transfer curve of (b).

Figure 2.11 Schematic of a sub‐threshold Gm integrator pixel readout structure.

Figure 2.12 The corresponding timing diagram and transfer curve.

Figure 2.13 Schematic of a 4‐shared sub‐threshold Gm integrator pixel readout structure.

Figure 2.14 (a) Schematic of capacitive transimpedance amplifier (CTIA); The corresponding; (b) Timing diagram; and (c) Current‐to‐voltage transfer curve.

Figure 2.15 Schematic of: (a) A simple 5‐T operational transconductance amplifier (OPA); and (b) Folded‐cascode OTA.

Figure 2.16 Schematic of an in‐pixel CTIA readout structure.

Figure 2.17 (a) Schematic of a multi‐gain switched capacitor amplifier. (b) The corresponding timing diagram and voltage transfer curve.

Figure 2.18 (a) Schematic of a single‐slope ADC; and (b) The corresponding timing diagram.

Figure 2.19 (a) Diagram of the 3‐stage comparator, including two (b) differential pre‐amplifiers; and one (c) Schmitt Trigger comparator.

Figure 2.20 Schematic of a ramp generator.

Figure 2.21 (a) Schematic of a simple hysteresis comparator; and (b) The corresponding hysteresis voltage transfer curves.

Figure 2.22 (a) Schematic of a 3‐bit bidirectional counter; and (b) The corresponding timing diagram.

Figure 2.23 (a) Quantization transfer curve with a continuous ramp as input signal and the corresponding quantization error during conversion; and (b) Block diagram of a linear model of quantizer.

Figure 2.24 Quantization noise power spectrum density distribution of Nyquist‐rate sampling and oversampling.

Figure 2.25 (a) A first‐order sigma‐delta modulator; and (b) The corresponding frequency domain model.

Figure 2.26 Noise shaping property of sigma‐delta modulator.

Figure 2.27 Schematic of continuous‐time: (a) Gm‐C filter; and (b) Active‐RC filter.

Figure 2.28 (a) Schematic of a discrete‐time switched‐capacitor filter; and (b) Its operations in different clock phases.

Figure 2.29 Implementation of: (a) A triode‐based transconductor; and (b) A 2‐stage operational amplifier.

Figure 2.30 Implementation of a first‐order sigma‐delta ADC.

Figure 2.31 A schematic example that CDS and CMS are performed.

Figure 2.32 Timing diagram of correlated double sampling (CDS).

Figure 2.33 Timing diagram of correlated multiple sampling (CMS).

Figure 2.34 Two ways to realize a 5‐input NAND logic block.

Figure 2.35 Diagram of a 3‐8 decoder.

Figure 2.36 Diagram of a hierarchy 7‐128 decoder.

Figure 2.37 Block diagram and timing diagram of a row driver.

Figure 2.38 Schematic of a typical level shifter.

Figure 2.39 Schematic of data readout blocks following ADC.

Figure 2.40 Schematic of a SRAM cell.

Figure 2.41 Schematic of Sense Amplifier (SA).

Figure 2.42 Diagram of a typical LVDS.

Figure 2.43 Differential input signals generation circuit.

Figure 2.44 (a) Diagram of a LVDS driver circuit; and (b) Simple common‐mode feedback circuit.

Chapter 03

Figure 3.1 (a) CMOS impedance sensor pixel array; (b) Cross‐sectional illustration of the sensing and addressing scheme of the CMOS impedance pixel; and (c) 96 × 96 impedance sensing electrode array.

Figure 3.2 (a) Sectional view of an electrode covered with an MCF‐7 cell. The total current is split into three pathways:

I

spread

,

I

cell

, and

I

seal

. The equivalent circuit of (b) Electrode without cells; and (c) Electrode with cells.

Figure 3.3 Block diagram of the CMOS impedance sensor readout circuit.

Figure 3.4 Block diagram of a transimpedance amplifier as readout circuit for the shared pixel array.

Figure 3.5 Circuit schematic of pixel‐level transimpedance amplifier.

Figure 3.6 Circuit schematic of the column‐level column‐readout amplifier.

Figure 3.7 Architecture of the 96 × 96 Electronic Impedance Sensing System.

Figure 3.8 Electrical‐impedance spectroscopy (EIS) platform for the detection and enumeration of CTCs: (a) Optical image of high‐density CMOS electrodes surrounded with silicone glob top to make a chamber. The ITO coated glass is positioned on top of the chamber to act as a counter electrode; (b) Zoomed optical image of post‐processed CMOS microelectrode array with exposed gold capping for bio‐sensing; and (c) Cross‐sectional scanning electron micrograph of a 22‐m wide microelectrode with 6 stacks of metal layers. A 0.1‐m thick Ti adhesion layer is sandwich between Au and Al (not shown).

Figure 3.9 (a) Impedance spectroscopy of electrode with cell attached in cell buffer; (b) Impedance spectroscopy of electrode without cell attached in cell buffer. The black circles represent impedance spectrum of electrode without cell in pure PBS; (c) Percentage of normalized impedance change as a function of measured frequency; (d) Statistical normalized impedance changes with the presence and absence of cell on microelectrode (n = 45 over three chip with cell, n = 47 over three chip without cell). The inset shows an optical image of microelectrodes with presence of cells (left column) and absence of cell (right column).

Figure 3.10 Calculated impedance magnitude as a function of frequency for 22 µm

2

electrode covered with cells: (a) Impedance evolution when R

seal

increases for cell covered electrode, and (b) Normalized impedance change as a function of frequency. Each line corresponds to values of R

seal

in the range of 0.14–0.84 MΩ.

Figure 3.11 (a) Optical image of cells seeded on a 12 × 12 microelectrode array; (b) Optical mapping; and (c) Impedance mapping of cells on the microelectrode array. The dark and gray square denote the status of an individual electrode; (d) The categories plot shows the accuracy of impedance mapping from an array of 144 pixels. A mapping accuracy of 90% was achieved.

Chapter 04

Figure 4.1 Application examples of THz spectroscopy and imaging: (a) Nondestructive detection of crack initiation in a film‐coated layer on a swelling tablet [5]; (b) Hydration state characterization in solution [6]; (c)

In‐vitro

breast cancer diagnosis [7]; and (d)

In‐vivo

skin cancer diagnosis [8].

Figure 4.2 (a) Stacked SRR unit‐cell designed by metal layers of M7 and M6; and (b) S21 simulation results with different stacking methods.

Figure 4.3 Geometries of resonators with slow‐wave shielding: (a) Differential T‐line loaded with stacked SRR; (b) T‐line based standing‐wave resonator; and (c) Cross‐section of BEOL.

Figure 4.4 T‐line‐based SRR excitation: (a) Single‐ended approach; (b) Differential approach; (c) Magnetic field distribution of single‐ended approach; and (d) Magnetic field distribution of differential approach.

Figure 4.5 Reflection coefficients of both single ended and differential resonators.

Figure 4.6 Simulated Γ and quality factor (Q) of SRR/T‐line unit cell at resonance.

Figure 4.7 On‐chip differential T‐line loaded with CSRR.

Figure 4.8 EM characterization of the proposed differential CSRR resonator.

Figure 4.9 Simplified equivalent circuit model of super‐regenerative amplifier.

Figure 4.10 Reflection coefficient of T‐line loaded with CSRR unit‐cells.

Figure 4.11 Reflection loss compensation by cross‐coupled NMOS pair with controlled tail current.

Figure 4.12 Layout for CMOS on‐chip implementation of DTL‐CSRR for 96 GHz SRX.

Figure 4.13 EM‐simulation based comparison of DTL‐CSRR and LC‐tank resonator for CMOS 96 GHz SRX design.

Figure 4.14 Schematic of CMOS 96 GHz SRX with DTL‐CSRR.

Figure 4.15 Layout for CMOS on‐chip implementation of DTL‐SRR for 135 GHz SRX.

Figure 4.16 EM‐simulation‐based comparison of DTL‐SRR and LC‐tank resonator for CMOS 135 GHz SRX design.

Figure 4.17 Impedance diagram of DTL‐SRR and LC‐Tank in Globalfoundries 65‐nm CMOS process: (a) Real and imaginary parts of

Z

Diff

; (b) Phase of

Z

Diff

; and (c) Phase of Z

Diff

when the ideal 16‐fF capacitor is included.

Figure 4.18 Schematic of CMOS 135 GHz SRX with DTL‐SRR.

Figure 4.19 (a) PCB integration of CMOS 135‐GHz SRX with antenna; and (b) THz imaging measurement setup with the proposed receiver chip integrated on PCB and object under test fixed on an X‐Y moving stage.

Figure 4.20 Images captured by imaging system with the proposed 135‐GHz SRX receiver: animal skin samples and Panadol pills.

Figure 4.21 Images captured by imaging system with the proposed 135‐GHz SRX receiver: various types of oil.

Figure 4.22 Absorption ratio of various types of oil detected at 135 GHz.

Figure 4.23 CMOS sub‐THz cell impedance detection system.

Figure 4.24 The simplified equivalent circuit and the route to extract the cell dielectric parameters using a SRR biosensor loaded with CTCs solutions.

Figure 4.25 The schematic diagram of SRR‐based‐oscillator sensor.

Figure 4.26 THz chip in the 65 nm process for CTC detection.

Figure 4.27 The outputs of oscillator and charge accumulator based on circuit simulation.

Figure 4.28 The voltage outputs under various capacitance values of loaded CTC solution. The inset shows the relationship between voltage output and loaded capacitance.

Chapter 05

Figure 5.1 Overall system diagram of an ultrasound imaging system with an AFE receiver integrating with the capacitive micro‐machined ultrasound transducer (CMUT).

Figure 5.2 (a) Diagram of 2‐D CMUT array; (b) One CMUT element; (c) One CMUT cell; (d) Cross‐section view of CMUT cell; and (e) Top view of CMUT cells.

Figure 5.3 Equivalent simulation model for CMUT.

Figure 5.4 Two‐channel analog front‐end IC unit cell consisting of two HV pulsers, two HV protection switches, and a shared low‐noise preamplifier.

Figure 5.5 Overall system floor plan for 2‐D multi‐channel analog front‐end IC.

Figure 5.6 Timing diagram of analog front‐end operation.

Figure 5.7 Block diagram of transmitter front‐end IC.

Figure 5.8 Schematic of 1.8‐to‐5 V level‐shifter.

Figure 5.9 Schematic of 5‐to‐25 V level‐shifter and HV output driver.

Figure 5.10 Schematic of the low‐noise preamplifier.

Figure 5.11 Simulated timing response of the HV transmitter front‐end.

Figure 5.12 Simulated: (a) closed‐loop frequency response; and (b) Input referred noise current.

Figure 5.13 Chip microphotograph of: (a) AFE array; and (b) 2‐channel AFE IC cell.

Figure 5.14 Measured transient response of the HV pulser.

Figure 5.15 Measured: (a) Frequency response; and (b) Input noise current of the preamplifier.

Figure 5.16 Acoustic transmission testing setup for the implemented ultrasound sensor and fabricated CMUT sample.

Figure 5.17 Measured hydrophone output voltage signal.

Figure 5.18 Acoustic pulse‐echo testing setup for implemented AFE IC and fabricated CMUT sample.

Figure 5.19 Measured pulse‐echo response.

Chapter 06

Figure 6.1 The MEMS capacitive accelerator is fabricated on an SOI wafer using DRIE and release. The single layer of patterned metal consists of an electrical contact pad and a hermetic seal is used. Electrical feed‐through is routed through the on‐chip interconnect in the CMOS chip.

Figure 6.2 System block diagram for the readout circuit.

Figure 6.3 Schematic diagram of the low noise, band‐pass gain stage, which consists of two single‐ended output amplifiers based on folded‐cascode architecture. Pseudo‐resistors are used in the feedback path of the amplifiers. Tunable feedback capacitance allows variable gain.

Figure 6.4 Die micrograph of the readout circuit fabricated through MPW (0.35 µm, 2P4M process).

Figure 6.5 Heterogeneous 3‐D CMOS‐on‐MEMS stacking with face‐to‐face direct metal bonding (no solder) that realizes electrical, mechanical, and hermetic bonds simultaneously.

Figure 6.6 (a) FIB/SEM image showing the on‐chip metal layers in the CMOS chip; and (b) Since no solder bumping is applied, the top passivation layer is intentionally recessed to create a standoff gap for the metal pads to facilitate direct metal bonding (thermo‐compression) with the matching pads on the MEMS chip.

Figure 6.7 (a) Cross‐sectional view of the bonded Al–Au layer from dummy structure; (b) High resolution TEM view; (c) EDX elemental mapping of the bonded layer; and (d) EDX line scan results.

Figure 6.8 (a) The alignment marks on MEMS and CMOS die are used for orienting the dies before stacking; and (b) The CMOS die is vertically stacked on top of the MEMS die.

Figure 6.9 The face‐to‐face bonded CMOS chip on MEMS chip. The bonded chip is then wire bonded to the package for electrical testing.

Figure 6.10 Shear strength test results with varying bonding temperature.

Figure 6.11 Shear strength test results with varying bonding temperature C‐SAM image of samples bonded at 290 °C.

Figure 6.12 Schematics of the formation of a sealed cavity for helium leak rate detection: (a) Forming of cavities, seal rings, and air channel using DRIE; (b) Sequential deposition of Cr layer and Au bonding layer; (c) Deposition of SiO

2

isolation and Al bonding layer; and (d) Au–Al thermo‐compression bonding of the cavity wafer to the capping wafer.

Figure 6.13 C‐SAM images of hermetic test dies.

Figure 6.14 (a) The band‐pass gain stage output (upper trace) during the standalone testing of the CMOS readout, when one of the carrier amplitude is 7.5 Vpp (lower trace) at 50 kHz frequency and 0 g acceleration; and (b) Fully rectified sinusoids at the synchronous demodulator outputs, verifying that the readout is working as intended.

Figure 6.15 The gain stage output (yellow, upper trace) when the excitation carrier amplitude is 1 Vpp (blue, lower trace) at 50 kHz frequency and the bonded chip is tilted at: (a) 0 g; (b) +1 g; and (c) –1 g orientation. Acceleration at a tilt angle θ is given by g sin(θ).

Figure 6.16 Variation in the mean demodulator output with g at various carrier frequencies for the bonded chip.

Figure 6.17 FFT spectrum showing fundamental peak at the carrier frequency and higher‐order harmonics at integer multiples of carrier frequency.

Figure 6.18 (a) SNR as a function of carrier frequency at various g orientations, when the carrier amplitude is 1 Vpp; and (b) Output voltage noise as a function of carrier frequency at various g orientations, when the carrier amplitude is 1 Vpp.

Figure 6.19 Setup for the 500 g shock test. The chip is subjected to a maximum vertical acceleration of 503.55 g for a short duration of 1.03 m for a total number of 10 times.

Figure 6.20 General control profile during the 500 g and 1 ms mechanical shock test.

Chapter 07

Figure 7.1 CCD read out architecture.

Figure 7.2 CCD charge transfer operations.

Figure 7.3 Passive and active pixel schematic and potential well diagram.

Figure 7.4 (a) Architecture of a typical digital imaging system with CCD sensor; and (b) CIS chip integrates pixel, analog readout and digital control.

Figure 7.5 FSI 4 T pinned PD pixel structure: (a) Pixel Schematic; and (b) CIS chip cross‐section (Samsung NX200).

Figure 7.6 (a) Potential well diagram during pixel readout; and (b) Timing diagram during pixel readout.

Figure 7.7 Backside illuminated (BSI) 4 T CIS pixel structure: (a) Schematic; and (b) Cross‐section of CIS chip (OmniVision).

Figure 7.8 Stacked CIS: (a) Structure of stacked CIS; and (b) Cross‐sectional view of stacked CIS.

Figure 7.9 Photon transfer curve (PTC) showing different noise sources at the CIS imaging system.

Figure 7.10 Small signal model for the noise analysis in CIS.

Figure 7.11 CIS readout circuit structure.

Figure 7.12 CIS Global readout circuit.

Figure 7.13 CDS implementation by integrating switch‐capacitor amplifier with column sample and hold circuitry.

Figure 7.14 Contact imaging using: (a) FSI sensor, where PDs are far from the cell: and (b) BSI sensor, where the PDs are close to the cell.

Figure 7.15 Detail 4‐way shared no‐row‐select pixel structure of the BSI image sensor. Four PDs form a unit cell.

Figure 7.16 System architecture of the 3.2‐Mega‐Pixels CIS with top‐bottom readout scheme and 4.4‐ column pitch column readout circuit.

Figure 7.17 Block diagram of column readout flow and circuit.

Figure 7.18 Timing diagram of 3.2 Mega BSI‐CIS.

Figure 7.19 Photos of: (a) Setup of the developed microfluidic cytometer system; (b) Die micrograph of BSI CIS; and (c) Microfluidic channel mounted on the ceramic packaged sensor.

Figure 7.20 Measurement results of: (a) ADC output codes under different input voltages; and (b) BSI‐CIS sensor sensitivity.

Figure 7.21 ADC outputs of a 2.8 V reset voltage, a 2.2 V pixel signal and the subtraction of the two voltages.

Figure 7.22 Captured raw images of (a) WBC and (c) RBC and PLT. Processed images of (b) WBC and (d) RBC and PLT.

Figure 7.23 Captured contact images of WBC, RBC, and PLT, with microscope images as references.

Chapter 08

Figure 8.1 Contact imaging principle: with light source illuminated from above; the contact shadow images of microbeads can be captured by the sensor underneath.

Figure 8.2 Schematic of: (a) 4 T‐CIS pixel; (b) Dual‐mode pixel; and (c) ISFET pixel.

Figure 8.3 (a) Cross‐section layout of the dual‐mode CMOS ISFET pixel; and (b) The top view of the dual‐model pixel.

Figure 8.4 CDS readout schematic for dual‐mode sensor.

Figure 8.5 CDS readout timing diagram for both: (a) Optical mode; pH mode; and (b) before and (c) after loading solution.

Figure 8.6 Top architecture of dual‐mode sensor.

Figure 8.7 The global core amplifier circuit.

Figure 8.8 12‐bit pipelined ADC architecture.

Figure 8.9 Micrograph photo of the dual‐mode sensor chip and testing setup.

Figure 8.10 Cross‐sectional structure of the ISFET, as modeled in Sentaurus TCAD. The electron concentration during ISFET operation is indicated by coloring.

Figure 8.11 ISFET device simulation results showing the V

T

change.

Figure 8.12 The global amplifier AC simulation results.

Figure 8.13 12‐bit pipelined ADC INL/DNL simulation results.

Figure 8.14 The correlated maps of distributed microbeads: (a) Contact images; and (b) pH values.

Figure 8.15 Measurement results: pH sensitivity of dual‐mode ISFET sensor.

Figure 8.16 The comparison with commercial pH meter for bacteria (

E. Coli

) culture solution with glucose at different time intervals.

Figure 8.17 Measurement results: spatial FFT of readout voltage variations: (a) with CDS; and (b) without CDS read out.

Chapter 09

Figure 9.1 Dual‐mode CMOS EHI APS pixel: (a) Circuit schematic; (b) Energy harvesting mode (EHM) configuration; and (c) Imaging mode (IM) configuration.

Figure 9.2 Proposed EHI CMOS APS pixel: (a) Layout; and (b) Cross‐section.

Figure 9.3 (a) Global and column level analog readout circuits; and (b) Current mirror OTA schematic used in GCA.

Figure 9.4 The schematics of the 10‐bit SAR ADC and two‐stage dynamic comparator.

Figure 9.5 (a) Ultra‐low power CIS architecture; and (b) Chip photo.

Figure 9.6 Global, row, and pixel level circuits for reset and select boosting in 3 T CMOS APS imagers.

Figure 9.7 Schematic of current‐mirror OTA.

Figure 9.8 Low‐power dynamic comparator in SAR ADC.

Figure 9.9 (a) Circuit diagram of the PMS; and (b) Operation principle of the PMS.

Figure 9.10 Simulation result of the full operating cycles of the PMS.

Figure 9.11 Power consumption of the EHI imager for varying frame rates (a) and varying supply voltages (b).

Figure 9.12 FoM performance of the imager under different frame rates and supply voltages.

Figure 9.13 Measured output voltage vs. output current (a) and output power of the solar cell array (b).

Figure 9.14 Measured efficiency of boost converter and overall PMS block.

Figure 9.15 Measured storage capacitor voltage (V

OUT

) and V

START

signals.

Chapter 10

Figure 10.1 The relationship between DNA sequence cost and Moore’s law: Data from the NHGRI Genome Sequencing Program (GSP): Available at: www.genome.gov/sequencingcosts.

Figure 10.2 A typical DNA molecule structure.

Figure 10.3 ISFET‐based sequencing principle.

Figure 10.4 DNA sample preparation process.

Figure 10.5 (a) PCR components; and (b) PCR process.

Figure 10.6 pH‐based DNA sequencing diagram (left); Cross‐section of ISFET during sequencing (right).

Figure 10.7 Output voltage in one microwell during the DNA sequencing procedure.

Figure 10.8 pH changes of different multi‐base incorporations.

Figure 10.9 Dual‐mode sensor to deal with pH crosstalk (left); Cross‐section of dual‐mode pixel with microbead contact imaging and ion sensing (right).

Figure 10.10 Diagram of THz‐based genotyping.

Figure 10.11 Realization diagram of metamaterial based THz‐sensing system.

Figure 10.12 Structure of (a) α‐HL protein; and (b) MspA protein;

Figure 10.13 Structure and cross‐section view of a Si

3

N

4

nanopore.

Figure 10.14 A typical nanopore sequencing principle.

Figure 10.15 Equivalent nanopore model with a pair of electrodes.

Figure 10.16 A typical nanopore ion current sensing diagram.

Figure 10.17 Post process to deposit Ag/AgCl on read electrode array.

Figure 10.18 A typical procedure to assemble proteins and a sensor chip.

Figure 10.19 A typical Si

3

N

4

nanopore array fabrication process.

Figure 10.20 Si

3

N

4

nanopore array and IC readout chip integration.

Chapter 11

Figure 11.1 General lensless cell counting system setup based on CMOS image sensor (CIS): (a) Lensless cell imaging principle; (b) Cross‐sectional view of the lensless system; and (c) Concept of the machine‐learning based single‐frame super‐resolution (SR) processing.

Figure 11.2 Resolution model for lensless microfluidic imaging system.

Figure 11.3 Working principle of multi‐frame SR processing.

Figure 11.4 Working principle of the proposed single‐frame SR algorithm.

Figure 11.5 Structure of the ELM model.

Figure 11.6 ELM‐SR processing flowchart. The training is performed off‐line to generate a reference model that can map the interpolated LR images with the HF components from the HR images; and the testing is performed on‐line to recover an SR image from the input LR image with the reference model.

Figure 11.7 CNNSR processing flow including one off‐line training and one on‐line testing step.

Figure 11.8 Contact‐imaging based microfluidic cytometer for flowing cell recognition and counting, (a) Process of bonding with PDMS chip; and (b) Microfluidic cytometer system diagram.

Figure 11.9 Hardware design of the microfluidic cytometer: (a) PCB schematic; (b) PCB hardware with size of 5.6 cm × 5.6 cm; and (c) Control GUI of the microfluidic cytometer.

Figure 11.10 Flowing cell recognition flowchart. The detected LR image is processed with ELM‐SR to obtain SR images according to different off‐line trained models. Then, the SR images are compared with typical HR cell images in the library with cells categorized to one type that has the largest MSSIM.

Figure 11.11 Comparison of concentration measurement results for 6 µm microbead solution between the developed microfluidic cytometer and the commercial flow cytometer.

Figure 11.12 Comparison of counting results of different microbead concentration solutions between the developed microfluidic cytometer and the commercial flow cytometer: (a) Measurement results correlate well between the developed system and the commercial one (y = 0.97x – 8, correlation coefficient = 0.996); (b) The Bland‐Altman analysis of the measurement results between the developed one and the commercial one show a mean bias of –13.6 uL

−1

, the lower 95% limit of agreement by –61.0 uL

−1

, and the upper 95% limit of agreement by 33.8 uL

−1

.

Figure 11.13 Example images of HepG2, RBC, and WBC in ELMSR and CNNSR training image libraries: (a) Original HR images with all HF details in ELMSR library; (b) Down‐sampled LR images with HF information lost in ELMSR library; (c) Interpolated LR images whose HF cannot be recovered in ELMSR library; (d) HF components that are lost during down‐sampling in the ELMSR library; (e) Original HR images with all HF details in the CNNSR library; (f) Down‐sampled LR images with HF information lost in the CNNSR library; (g) Interpolated LR images whose HF cannot be recovered in the ELMSR library; and (h) HF components that are lost during down‐sampling.

Figure 11.14 Example of HepG2, RBC, and WBC images in ELMSR and CNNSR testing: (a) Raw LR images captured by FSI CIS with pixel pitch 2.2 um; (b) Interpolated LR images; (c) ELMSR recovered HR images; (d) Raw LR images captured by BSI CIS with pixel pitch 1.1 µm; (e) Interpolated LR images; and (f) CNNSR recovered HR images, showing better performance in resolution improvement.

Figure 11.15 The mean structural similarity (MSSIM) results for on‐line recover cell images.

Chapter 12

Figure 12.1 Smart multimodal CMOS sensor based LOC for personalized diagnosis.

Guide

Cover

Table of Contents

Begin Reading

Pages

iii

iv

x

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

CMOS Integrated Lab‐on‐a‐Chip System for Personalized Biomedical Diagnosis

Hao Yu

Southern University of Science and TechnologyChina

Mei Yan

ConsultantChina

Xiwei Huang

Hangzhou Dianzi UniversityChina

This edition first published 2018© 2018 John Wiley & Sons Singapore Pte. Ltd.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.

The right of Hao Yu, Mei Yan, Xiwei Huang to be identified as the authors of this work has been asserted in accordance with law.

Registered Office(s)John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USAJohn Wiley & Sons Singapore Pte. Ltd, 1 Fusionopolis Walk, #07‐01 Solaris South Tower, Singapore 138628

Editorial Office1 Fusionopolis Walk, #07‐01 Solaris South Tower, Singapore 138628

For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.

Wiley also publishes its books in a variety of electronic formats and by print‐on‐demand. Some content that appears in standard print versions of this book may not be available in other formats.

Limit of Liability/Disclaimer of WarrantyWhile the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

Library of Congress Cataloging‐in‐Publication Data

Names: Yu, Hao, 1976– author.Title: CMOS integrated lab‐on‐a‐chip system for personalized biomedical diagnosis / Hao Yu, Southern University of Science and Technology, China, Mei Yan, Consultant, China, Xiwei Huang, Hangzhou Dianzi University, China.Description: Hoboken, NJ : Wiley, 2018. | Series: Wiley ‐ IEEE | Includes bibliographical references and index. |Identifiers: LCCN 2017049248 (print) | LCCN 2017050886 (ebook) | ISBN 9781119218357 (pdf) | ISBN 9781119218340 (epub) | ISBN 9781119218326 (hardback)Subjects: LCSH: Medical instruments and apparatus–Research. | Metal oxide semiconductors, Complementary.Classification: LCC RA856.4 (ebook) | LCC RA856.4 .J53 2018 (print) | DDC 610.28/4–dc23LC record available at https://lccn.loc.gov/2017049248

Cover Design: WileyCover Image: © e‐crow/Gettyimages

Preface

Considering the current aging society, the future personalized diagnosis requires portable biomedical devices with miniaturization of bio‐instruments. The recent development of lab‐on‐a‐chip (LOC) technology has provided a promising integration platform of microfluidic channels, microelectromechanical systems (MEMS), and multi‐modal sensors, which allow non‐invasive and near‐field sensing functions. The standard complimentary metal‐oxide semiconductor (CMOS) process allows a low‐cost system‐on‐chip (SOC) solution to integrate sensors from multimodal domains, which has raised many new design challenges, such as how to develop multimodal sensors for system integration; how to integrate with MEMS and microfluidic channels from the device technology perspective; as well as data fusing and smart processing of multiple domains from system application perspective.

This book will report on the recent progress in CMOS integrated LOC system for personalized diagnosis. The book is organized into 12 chapters. After a background discussion on personalized diagnosis, LOC, and CMOS‐compatible multimodal sensors in the first Chapter of introduction, Chapter 2 to Chapter 7 discuss CMOS sensor design and several CMOS sensor technologies, namely CMOS electronic impedance sensor, CMOS Terahertz sensor, CMUT ultrasound sensor, CMOS 3D‐integrated MEMS sensor, and CMOS optical sensor for microfluidic contact imaging. Then two dual‐mode sensor are illustrated in Chapters 8 and 9, such as dual‐mode chemical/optical image sensor and dual‐mode energy‐harvesting image sensing. Finally, based on the aforementioned sensors, two important applications of DNA sequencing and cell counting are elaborated on. Chapter 10 focuses on DNA sequencing application, Chapter 11 covers cell imaging and counting application, and Chapter 12 briefly summarizes the book.

The authors would like to thank their colleagues Y. Shang, S. Manoj, X. Liu, Hantao Huang, Zichuan Liu, and Hang Xu. The authors also acknowledge with gratitude discussions with Professors Krishnendu Chakrabarty, Tsung‐yi Ho, Zhihua Wang, Yong Lian, Guoxing Wang, Mohamad Sawan, Pantelis Georgiou, Dongping Wu, Chenjun Huang, Paul Franzon, Xin Li, and Kenneth Shepard. Their support was invaluable to us during the writing of this book. Finally, the authors acknowledge TSMC’s contribution to sensor chips fabrication in the MPW tapeout shuttle.

Hao Yu, Mei Yan, and Xiwei Huang

1Introduction

1.1 Personalized Biomedical Diagnosis

1.1.1 Personalized Diagnosis

The world’s older population continues to grow at an unprecedented rate. The proportion of people aged 60 years and over grows faster than any other age group, as shown in Figure 1.1 [1]. According to the expectation of the World Health Organization [2], between 2000 and 2050, the world’s population of over 60 years of age will double from about 11% to 22%, and the absolute number of such people will also increase from 605 million to 2 billion. Among the aging countries, the most dramatic changes are now taking place in low and middle income countries with limited biomedical infrastructure, incomplete healthcare systems, and shortage of funds and resources. The current aging society also comes with special healthcare challenges due to limited hospital resources, doctors and related facilities. A portable and low‐cost biomedical diagnosis instrument is thereby in high demand to meet the needs of the growing aging population in the form of personalized biomedical diagnosis.

Figure 1.1 Global aging trends for the percentage of the total population at 60 years of age or over in 2012 and 2050 [1].

Over the past several decades, biomedical diagnostic techniques such as the microscope [3], ultrasound [4–6], flow cytometry [7–9], and genetic sequencing [10–12], have improved the accurate monitoring of existing diseases, and also the understanding of the underlying causes of those diseases. However, to obtain highly sensitive measurements, current diagnosis instrument systems are usually bulky and expensive, their complicated operation requiring professional personnel. As such, they are usually only available in established hospitals or clinics, and hence are not flexible for multiple functionality diagnoses for on‐site personalized diagnosis. These problems pose significant challenges for the personalized healthcare of aging populations, especially in low‐income developing countries. As such, portable and affordable biomedical instruments that can be miniaturized are required to provide a point‐of‐care (POC) diagnosis [13–15].

A POC biomedical instrument is meant to perform the diagnosis at the site of patient care by a clinician or by the patient without the need for clinical laboratory facilities. The tests are rapid, portable, non‐invasive, and easy‐to‐use, with timely testing results, which allow rapid clinical decision‐making and also mitigate treatment delay. The development of these POC diagnostic and monitoring instruments thereby allow individuals, especially these older people, to monitor their own health. It leads to a paradigm shift from conventional curative medicine, to predictive and personalized diagnostics [16]. Moreover, because factors such as medication adherence, genetics, age, nutrition, health, and environmental exposure can vary, and also the extent of biomedical treatment and drug response of each individual, people are becoming more interested in exploring the biology of the disease and its treatment at his or her own individual level. For example, people can use the information about his or her own genes, proteins, metabolites, etc. at the molecular level, and the leverage with the existing environment to prevent, diagnose, and treat the disease at the individual level. Therefore, such personalized biomedical diagnostics, with the existing supporting instrument platform, has emerged as a significant need for the coming aging world.

1.1.2 Conventional Biomedical Diagnostic Instruments

In the following section, three of the most widely‐used traditional biomedical diagnosis instruments, namely the high‐resolution optical microscope, flow cytometer, and DNA sequencer, are discussed as the starting point for comparison.

1.1.2.1 Optical Microscope

A microscope is an optical instrument that produces a magnified image of the biomedical object under inspection, compared with what the naked human eye can observe, using visible light with lenses. The optical microscope was invented more than 400 years ago by two Dutch spectacle makers, Hans and Zaccharias Janssen, then improved by Galileo and Antonie van Leeuwenhoek, and is the leading high‐resolution visualization tool and the gold standard for biomedical imaging at the cellular level [17].

To achieve micrometer or sub‐micrometer resolution, almost all microscopes require precise and expensive optical lenses, as well as a large distance between the object lens and eyepiece lens for the light to travel and reshape, as shown in Figure 1.2. The object to be observed is illuminated by a light source. As light passes through the object, the objective lens (i.e. the lens closest to the object) produces the corresponding magnified object image in the primary image angle. The eyepiece (i.e. the lens that people look into) acts as a magnifier that produces an enlarged image by the objective lens. The overall magnification of the microscope system is the multiplication of both the object and the eyepiece. The principle of magnification is based on the thin lens approximation as follows:

(1.1)

where Li and Lo are the image distance and object distance, F is the focal length of the objective lens, M is the magnification factor of the objective lens, and H1 and H2 are the sizes of object and image respectively. Therefore, a significant space Li is usually required to produce a large microscopic magnification, which is the main difficulty for the minimization of the optical microscope.

Figure 1.2 A typical microscope and its optical path with objective lens and eye piece. To reach high‐resolution imaging capability, bulky, expensive, and sophisticated lenses are required.

Compared with the earliest compound microscope, the current design has evolved to incorporate multiple lenses, filters, polarizers, beam‐splitters, sensors, illumination sources, and a host of other components, aimed at improving resolution and sample contrast. However, this basic microscope design has undergone very few fundamental changes over the centuries, so it bulky, expensive, and complicated, hence not suitable for the desired POC diagnosis.

1.1.2.2 Flow Cytometer

Flow cytometer is another widely‐used biomedical instrument for applications in, for example, blood cell counting and sorting. Based on basic working principles, there are two types of flow cytometry methods, optical‐based and electrochemical‐based, as shown in Figure 1.3. With the help of sheath fluid and fluid dynamics effects, the blood cells are injected and passed through the measuring tunnel one at a time. For the optical‐based cytometry shown in Figure 1.3(a), each cell along the path interacts with the laser beam respectively and the light intensity of the scattering is measured by the forward and orthogonal optical detectors. The measurement results depend on the size of the cell and its internal complexity, which can be used for cell differentiation and counting. Whereas, for the electrochemical based cytometry shown in Figure 1.3(b), pairs of electrodes are placed on both sides of a narrow orifice and a low‐frequency current is applied. When blood cells are driven through, the impedance values are measured, which vary with the cell size and composition.

Figure 1.3 Diagrams of: (a) Optical‐based; and (b) Electrochemical‐based flow cytometry.

As an example, the FACSCount, as shown in Figure 1.4, is one commercial optical‐based flow cytometer system that can provide absolute and percentage counting results of various types of cells, such as red blood cells (RBCs), leukocytes, and CD4 T‐lymphocytes. Clinicians rely on this system to diagnose the stage progression of HIV/AIDS, guide the treatment decision for HIV‐infected persons, and evaluate the effectiveness of Antiretroviral Therapy (ART) [18]. However, it is only available in laboratory settings due to the limitations of bulky desktop size, prohibitive equipment cost ($27,000), high maintenance and reagent costs ($5–$20), low throughput (30–50 samples/day), and the need for an experienced operator, etc. [19]. Thus, it cannot meet the needs of personalized diagnosis.

Figure 1.4 BD FACSCount™ cell cytometer system.

1.1.2.3 DNA Sequencer

DNA sequencing, which detects the nucleotide order in DNA strands, enables the study of metagenomics and genetic disorders for diseases in aging people on an individual basis. Therefore, it plays an important role in personalized diagnostics. The first widely‐applied DNA sequencing technique was the Sanger sequencing in the 1970s. This technique employs DNA polymerase to synthesize double‐stranded DNA (dsDNA) from a primed single‐stranded DNA (ssDNA) template. Four standard deoxyribonucleoside triphosphates (dNTPs), adenine (A), cytosine (C), guanine (G), and thymine (T), are used to extend the DNA, whereas four radioactively labeled di‐dNTP (ddNTP) elements (ddATP, ddGTP, ddCTP, and ddTTP) are used to cease DNA extension. That is, once a ddNTP is attached to the DNA template, polymerase synthesis of this strand is invalid and no more dNTPs (or ddNTPs) can be added.

During sequencing, four chambers are employed for DNA synthesis. Each chamber is loaded with ssDNA templates, primers, polymerases, all four dNTPs, and one type of ddNTPs respectively. With sufficient dNTPs and a certain volume of labeled ddNTP, DNA extension is randomly stopped, thereby producing a set of dsDNAs of various lengths. Locations of terminating ddNTPs in all four chambers can be visualized with electrophoresis gel. Since ddNTP in each location is complementary to the nucleotide in the template, the DNA sequence can be obtained after combining all the ddNTP types and location information. Sanger’s contribution accelerated the process of DNA sequencing from roughly 10 base pairs (bp) per year to about 100 bp/day [20]. Later, the adoption of fluorescent labeling [21] and its associated optical detection hardware further improved Sanger’s sequencing efficiency. In this technique, polymerase synthesis can be finished in one chamber by using four types of fluorescent labels. With the help of an optical detection system, each type of labeled ddNTP terminator can be identified based on the different color or wavelength they emit. These improvements have greatly helped the automation of the DNA sequencing process.

Since 2005, the next‐generation sequencing (NGS) techniques substantially advanced the sequencing methods, targeting high‐throughput and low‐cost sequencing. A typical NGS machine is shown in Figure 1.5. Important examples of NGS technologies include the Solexa/Illumina bridge amplification method, which dominates this field currently [22], as well as Roche’s pyrosequencing method [23] with bead‐emulsified polymerase chain reaction (PCR) [24]. However, most of these methods require bulky and expensive optical instruments, which greatly restrict their use at the patient scale, such that only hospitals and research centers can afford them. In order to reduce the cost and size, new technologies targeted at personalized sequencing should be explored.

Figure 1.5 A typical optical DNA sequencer.

1.2 CMOS Sensor‐based Lab‐on‐a‐Chip for System Miniaturization

1.2.1 CMOS Sensor‐based Lab‐on‐a‐Chip

For bio‐instrument miniaturization towards personalized diagnosis, effective solutions can only be derived from technologies that can resolve the scaling. One proved technology from the semiconductor industry is based on the complimentary metal‐oxide semiconductor (CMOS) process. CMOS technology has been utilized in digital, analog, and mixed‐signal integration circuits (ICs), such as microcontrollers, microprocessors, cell‐phones, amplifiers, data converters, and transceivers, etc. More recently, the CMOS process has been developed in optical sensing, charge detection, temperature measurement, and electrochemical sensing applications, which can be realized by integrating readout circuits with on‐chip sensing devices, such as photodiodes, ion‐selected field effect transistors (ISFET), and microelectrodes. The main advantages of CMOS are the scalability in integration but also the variability in sensing. Large numbers or multiple functions of sensing devices (also called multi‐modal sensors) can be integrated on one chip, so as to create mass‐produced, low‐cost, diversified, and miniaturized biomedical sensors for personalized diagnosis.

In most applications, biomedical samples are prepared and detected in an aqueous environment, which means that sensors usually need to be physically interfaced with fluidic samples, so as to realize a portable size. Thus, the emerging Lab‐on‐a‐Chip (LOC) technique can combine a microfluidic system with the CMOS sensor. A microfluidic structure directs fluid samples towards the sensing sites, which are then detected by the underlying CMOS sensing devices. These devices, on the one hand, may act as voltage or current sources for actuating purpose. On the other hand, they may convert interesting biological information into an electric signal for sensing purposes.

As is implied by the name, Lab‐on‐a‐Chip means a minimized chip‐scale (a few square centimeters in size) device that has the ability to perform one or several laboratory functions [25–29]. Since LOCs need to deal with chemical or small biological objects, analyses in extremely small fluid volumes (microliters or even picoliters), it greatly relies on microfluidics for sample preparation, delivery, and on‐chip processing. The recent advancement of LOC technology has provided a promising solution to integrate microfluidics [14], micro‐electro‐mechanical systems (MEMS), and CMOS multimodal sensors [30–33] on one platform, which allows a miniaturized biomedical sensing system without bulky mechanical components. For such a CMOS sensor‐based LOC method, the reduced sample volume, portability, low cost, and the possibility to integrate multiple analytical devices, are key advantages over the traditional laboratory‐scale instruments.

A diagram of the CMOS sensor‐based LOC is shown in Figure 1.6. The sensor chip is fabricated through the standard CMOS process for the scaling benefit of low cost and mass production. It usually consists of a large amount (millions) of sensing nodes, or CMOS sensing pixels, to improve the sensing throughput. Thus, it can detect a great number of biological reactions simultaneously. In order to increase sensor density, the pixel design not only needs to consider the performance optimization, but also the scaling capability of continuously shrinking the pixel size. The CMOS sensor array can be further equipped with a fluidic package to carry and separate biomedical samples, so that a large volume of samples can be detected in parallel by their respective underlying pixels. After the sensor chip is fabricated, post processed, and packaged, it can be integrated with the microfluidic channel with fluidic packages to realize a completed LOC integration. Thus, such CMOS sensor‐based LOC microsystems can create tremendous opportunities for future personalized diagnosis applications.

Figure 1.6 CMOS‐based LOC integration for biomedical instrument miniaturization.

Building such CMOS‐based LOC platforms can bring great potential and opportunities. It enables the development of multimodal CMOS sensors. The multimodal sensing types can vary from electrochemical, electric impedance, terahertz, and optical, etc., which will be mainly addressed in this book. Note that CMOS‐based LOC platform can further integrate the data analytics part as well.

1.2.2 CMOS Sensor

1.2.2.1 CMOS Process Fundamentals

CMOS circuit design was invented in 1963 by Frank Wanlass at Fairchild Semiconductors. In 1968, a group led by Albert Medwin at RCA invented the first commercial CMOS integrated circuits that consist of 4000 series of CMOS logic gates. During the 1970s, CMOS technology was used in computing processor development, which led to the explosive growth of personal computers. It is estimated that the CMOS transistor is the most manufactured device in the history of mankind. Now, more than 95% of integrated circuits are fabricated in CMOS. One important reason behind this dominance is that CMOS is scalable for system integration.

A metal oxide semiconductor (MOS) technology can be classified as PMOS (P‐type MOS), NMOS (N‐type MOS), and CMOS (Complementary MOS). The basic physical structure of MOS devices consists of a semiconductor, oxide, and a metal gate. Nowadays, polysilicon is more widely used as the gate (G). Voltage applied to the gate controls the current between source (S) and drain (D). Since very low power is consumed, MOS allows very high integration.

Figure 1.7 shows the components of a modern CMOS technology. Basically, it includes NMOS and PMOS transistors on the same substrate. For NMOS, the first letter “N” indicates the kind of carrier that carries the current flow between source and drain when the threshold voltage is reached. Thus, NMOS stands for transistors where negatively charged electrons are the current carriers between source and drain. PMOS stands for transistors where positively charged holes are the current carriers. The CMOS circuit uses both NMOS and PMOS transistors for pulling down to ground and up to power supply. The transistors are arranged in a structure formed by two complementary networks, Pull Up Network (PUN) and Pull Down Network (PDN). PUN is a network composed of PMOS transistors between output and power source, whereas PDN is an NMOS‐type network between output and ground. Typically, in the CMOS circuit, only one of PUN and PDN is switched on and the other one is switched off with the corresponding gate capacitor to hold the charge information. This physical basic can be utilized in sensing.

Figure 1.7 Cross‐section view (left) and symbol (right) of CMOS technology components: NMOS and PMOS.

CMOS technology requires fabrication of two different transistors – NMOS and PMOS on a single chip substrate. Actually, NMOS and PMOS need substrates with different kinds of doping in one integrated circuit. A silicon wafer always has one doping type and doping level. Hence, to accommodate both NMOS and PMOS transistors, CMOS technology requires creation of special regions, called wells, by impurity implantation in the substrate. The semiconductor type in these regions is opposite to the base substrate type. A p‐well is created in an n‐type substrate or, alternatively, an n‐well is created in a p‐type substrate. Thus, if the base substrate is p‐type, the NMOS transistor is created in the p‐type substrate, while the PMOS transistor is created in the n‐well built into the p‐type base substrate.

Historically, fabrication started with p‐well technology, but now has completely shifted to n‐well technology. This is due to the lower sheet resistance of the n‐well