Integrated particle shape sensor

(1)

(2)

Integrated Partiele Shape Sensor

Peter Turmezei

May 26, 2006

(3)

f

Integrated Partiele Shape Sensor

Proefschrift

ter verkrijging van de graad van doctor, aan de Technische Universiteit Delft,

op gezag van de Rector Magnificus Prof.dr.ir. J.T. Fokkema, voorzitter van het College voor Promoties,

in het openbaar te verdedigen

op maandag 19 juni 2006 om 15:00 uur door

Peter Balazs Turmezei

(4)

^

Dit proefschi'ift is goedgekeurd door de promotor Prof.dr. P. J. French

Toegevoegd promotor:

Dr.ir. A. Bossche

Preface

Project history

The work presented in this PhD thesis is part of the STW project entitled "Shape Vision", which was started in 2000 at the Delft University of Technology by Prof. Michiel J. Vellekoop and Jeroen Nieuwenhuis. Afterthey moved to Vienna in 2001,

the work continued in parallel in Delft and Vienna, until Jeroen Nieuwenhuis de-fended his PhD. thesis successfuUy [47] in January 2005. In his thesis the following microsystems were developed and fabricated for partiele detection and

manipula-tl on;

A flow cell, using sheath flow, was designed for accurate partiele positioning in a microflow (Section 4.3.6).

An integrated Coulter counter with a liquid aperture was developed to detect partiele size in a microflow (Section 1.2.2).

A sorting module with an optimised electrode design was also created. This module can steer particles into multiple outlet channels.

Partiele size was also detected with an integrated photo-detector, which was designed using bipolar technology.

In Delft a parallel strategy was chosen which aimed to add more functionality to the chip and improve device performance. Alternative measurement solutions were investigated and new technologies were examined. The research spread over four

scientific areas including integrated circuit (IC) design, electromagnetism, computer siraulation and microfabrication. The following achievements were attained;

(7)

VI PREFACE

A flexible 3D simulation tooi was developed which can track partiele motion

and orientation in a microchannel.

An orientation module was designed which cnables quasi-three dimensional partiele shape detection.

A new wafer-Lo-wafer bonding technique was developed which allows

accu-rate alignraent of the wafers without the use of special alignment equipment.

Chapter 1

Despite the above achievements there are still tasks for the future. They have to aim at the integration of the microsystem modules into an application specific environ-ment and experienviron-mental studies, which characterise the impleenviron-mented modules and verify the simulations.

Introduction

Organisation of the thesis

The work is organised into five chapters. In chapter 1 the role of particles in our everyday life is introduced. Existing measurement techniques are described and the objective of the thesis is dcfined. The partiele shape sensor is introduced in chapter 2. The principle of the quasi-3D shape measurement is described together with the operation of the device. The chapter ends with a list of challenges and completed tasks. Chapter 3 contains the description of CMOS photodetector developed for the partiele shape analysis. A short theory and a brief comparison of the possible detection methods start the chapter. Based on this analysis the line-photodiode array type detector is selected for the partiele shape sensor. The design of this airay is discussed in the last section of the chapter. Chapter 4 covers the description of partiele raotion in microchannel. After the theoretical analysis of partiele behaviour

in AC electric field and shear flow, a simplified mathematical model was made, and combined with finite element calculations in our partiele tracking tooi. With this tooi the special case of the partiele shape sensor was examined and the feasibility of device operation was verified. In order to realize the partiele shape sensor, a new

wafer-to-wafer bonding technique had to be developed. It is described in chapter 5, together with the other fabrication techniques which were used during the device fabrication. The device holder makes the connection between the chip and the macro world. It is presented at the end of the chapter. The thesis ends with a short summary and acknowledgments. In the appendix additional calculations can be found.

The aim of this chapter is to introducé particles and point out their importance. First the symbiosis exisdng between particles and ourselves is revealed, then mea-surement techniques and Instruments for pardcle size and shape analysis wili be surveyed. At the end of the chapter the on-chip solutions and the objectives of the thesis are presented.

1.1 When and where do we encounter particles and what

do they "look" like?

Every day, everywhere and they have all kinds of different forms could be the short answer to the question posed in the title of this section. They influence our per-sonal life and our work even without us noticing it. Interaction with particles is inevitable. They can be found everywhere, not only in the world of engineers but in the everyday world as well. Certainly they are present in partiele technology and engineering but we can easily find them in biology, chemistry, medical science etc. Consequently the discipline of partiele technology cannot be considered as a standalone subject. To illustrate this let us consider for example process technology (often called chemical engineering). It has been estimated that 70% of the interme-diate products of Üie industries which are involved in chemical engineering (such as pharmaceuticals, minerals and food etc.) are in particulate form [14]. The product is sometimes called a paste, slurry, suspension, spray or emulsion, but the com-mon feature is that particles are involved in all cases. It is therefore impossible to talk about chemical engineering without mendoning particles. On the other hand chemical engineering has so much influence on the development of pardcle tech-nology that it defines many of its reseaich directions and appears in many sample applications and learning exercises. Of course we could have also taken another

1

(8)

CHAPTER l. INTRODUCTION _{1.2. PARTÏCLE ANALYSÏS}

3

engineering field as an example, because particles are present in electronics, civil engineering, mechanical engineering etc. Their effect is somctimes positive and helpful, but sometimes they hinder the process or even make it completely

impossi-ble.

Laser Sample

Some application areas where particles help and are central to the technology in-clude electrostatic copiers and printers, powder coating, colloidal suspensions, packed and fluidised bed reactors, powder coating machines, powder injection moulding, Chemical technology. DNAs and cells are also small objects, particles. The precise

handling and manipulation of these biological "building blocks" are the main target of flow cytometry and /yTAS (Micro Total Analysis Systems), also termed bio-chips or lab-on-chips.

Detector screen

Figure 1.1: Laser diffraction: The laser (directed from the left) scatters due to

tor(Sht) ' ' ' ' " " ' ' '''™'' ^ ' ' '''"'"^"^ ' ' " " " '' ''''''''' ^" ' ""'''''

1.2 Partiele analysis

In üther situations avoiding particles poses many challenges for engineers. The fabrication process for solid-state devices is extremely sensitive to particles and air filtration makes up a considerable percentage of the total cost of an integrated

cir-cuit chip. Engineers are constantly struggling to achieve the best possible collection and removal of particulate matter from combustion gases, with electrostatic pre-cipitators, packed bed filters etc. The preservation of water quality by removing particulate matter from industrial waste is also a challenge. Particles can also cause fire hazard. Explosion caused by airbome dust in certain polymer and

metallurgi-cal manufacturing operations is possible and has to be prevented. Particles are also responsible for mechanical wear and can be harmful to human health.

It is obvious that the detection and analysis of particles is crucial in all these sit-uations, however, the full picture of the partiele is never necessary. In most cases only a small set of the many paiameters which is of interest. For example, for elec-trostatic copiers the size, colour and electrical properties of the toner particles are

important, while for functionalised bcads (used for surface binding-based analysis in bio-chips) chemical composition and surface area have to be known. Narrowing down the number of parameters is necessary because no instrument can universally measure all of them. Furthermore, there might be ether limitations; size constraints for the instrument dimensions, maximum power consumption, minimum

resolu-tion, maximum amount of analyte etc. The next section contains a brief review of

the commercially available techniques for partiele shape and size measurements and methods which are under research in laboratories.

1.2.1 Size and shape measurements Laser diffraction

Laser diffraction measures the size distribution of the particles within a .sample vol-ume. It is based on the principle that Hght scattered by particles has a characteristic scattering patteni which depends on the size of the particles in the volume [1], Fig-ure 1.1 shows an example for such pattem. When more than one partiele size is present in the sample, the characteristic pattems are superimposed on each other. The pattern is also a function of the angle a between the incident light beam and the line, which connects the detector with the sample. The size distribution can be deduced when the pattern for different partiele sizes as a function of the a angle is known. The interaction between particles and light is mainly dependent on partiele size, but the shape, surface roughness and refractive indices of the material and dis-persing medium also affect the pattern. The conventional diffraction theory loses its applicability in the sub-micron range, but with enhanced methods equipment sensitive down to 0.04/im is also available on the market.

Photon Correlation Spectroscopy

(9)

4 CHAPTER 1. INTRODUCTION 1.2. PARTICLE ANALYSIS

Light source ₂₅₀

Time Plot b)

Figure 1.2: Photon con'elation spectroscopy. The sample is illuminated and the scattered light is detected a), while the particles in the sample move about due to Brownian motion (b). The motion results in the intensity variation of the signal from which size distribution can bc calculated (c).

to the Brownian motion the scattered light detected on a screen changes continu-ously. The speed of the change is inversely proportional to the partiele size (smaller particles cause faster change on the screen than larger ones). This is illustrated in

figure 1.2. Analysing the time-dcpendcnt pattem provides the average size of the particles. By enhancing the method with multiple angle measurement and advanccd

mathematical algorithms, information on the width of the distribution can also be inferred.

Image analysis

The most straightforward method of partiele analysis is using a microscope. When

the microscope is equipped with a high-resolution, low-noise digital CCD camera, advanced image analysis algorithms can be used to analyse the sample. As the

analysis is done on a set of individual particles, a larger set of parameters can be determined, such as thé Ferret diameter, area, perimeter, shape factor and aspect ratio. Figure 1.3 is an example for such analysis made with the WaveMetrics IGOR Pro software.

Time of Transition i

In case of the time of transition method [3] a moving light source is focused on

the measurement volume and a photo-detector is placed perpendicular to the beam.

When a partiele obscures the light, the signal of the photo-detector drops. As the

Figure 1.3: Example for image analysis: The outlines of the particles are detected by software. Paiticles are numbcred and grouped for further analysis.

velocity of the light movement is known, the size of the partiele can be calculated from the obscuration time. The strength of the method is that it measures paiticle size directly, rather than calculating it from a secondary property. In other words, the partiele size is independent of the refractive index, absorption, suiface texture, porosity or electrical conductivity. Usually particles falling into the range between 0.5/jm and iOOOfim can be measured with this method. Movement of the light beam can be provided by, for example, a rotating wedge prism as in the case of Galai CIS-100 from Ankersmid [4], where the focused spot size is l.ljxni, see figure 1.4.

Laser Doppler Velocimetry

Laser Doppler velocimetry is a non-intrusive method for measuring partiele veloc-ity and size. A laser light illuminates the sample volume which the particles pass through. Due to the motion of the particles the frequency of the scattered light will shift according to the Doppler effect. At the photo-detector this scattered light and a reference light are mixed, and the resulting so-called beat frequency will be propor-tional to the particle's velocity. Many optical arrangements exist to provide and mix the scattered and reference signals, figure 1.5 illustrates the differential set-up. By recording the light scattering pattern with a fast line scan sensor, the spatial

mod-ulation of the signal is also detected. This yields, in addition to partiele velocity, information about the size and morphology of the partiele [24].

Electrical Sensing Zone (Coulter principle)

(10)

I

6

CHAPTER l. ÏNTRODUCTION 1.2. PARTICLE ANALYSIS 1

A-He-Ne Laser Tube B-Wedge Prism

C-Wedge Prism Assembly D-Photodiode Detector E-Lens F-Collimator Lens G-Sample ^ G Laser Data Acquisition Gard

FiEure 1 4- Time of transition: Rotation of the wedge prism (B) moves the focus p mt of the laser beam m the sample (G). The s.ze of the partiele :s detenn,^d from

the speed of the Hght spot and the time the light was blocked by the partiele.

Transmitting/ receiving opties F!OW with particles Measurement volume Light intensity 2sin(Ö/2)

Figure 1.5: Set-up for size measurement with laser Dopplervelocimetry

Figure 1.6: SEM image of Ti pailicle sample, as used for analysis [20]

pumped through a small orifice, which has an electrode pair on its sidewall. When-ever a partiele passes through the orifice the impedance measured by the electrodes changes. The speed of the particles can be inferred from the length of the measured pulse, while the amplitude of the pulse is proportional to the volume of the partiele. For con-eet operation the particles have to have different electrical proporties than the medium and the partiele concentration has to be low (only one partiele at a time

should cross the orifice).

Scanning Electron Microscopy

Analysis of particles fixed to a sample plate is possible with scanning electron mi-eroseopy (SEM). The sample is scanned with a focused electron beam and the im-age is reconstructed with one of the many detection modes of the SEM. The most frequently used method is the secondary electron imaging. In this mode the high energy electron beam interacts with the loosely bound electrons in the surface of the sample and as a result low energy electrons are ejected. The image is recon-structed by capturing the ejected electrons with a detector. The width of the beam, which defines the resolution, is only a few nm so the size and shape can both be accurately assessed, but only for a relatively small population of particles (a few hundred at best). Figure 1.6 shows a SEM micrograph of Ti partiele sample as used for detailed surface texture characterisation [20].

Sedimentation

Sedimentation is the simplest method for determining partiele size. The settling rate reflects the partiele size through the Stokes' formula, equation 4.2. Nowadays the

(11)

^ ^

8

CHAPTER l JNTRODVCTION

ÏÏ:-ÏSC£====--"S-1.2.2 Lab-on-chip solutions

Scaling down enlarges the relative importance of some physical phenomena, while making some ethers unimportant. For example, the electrostatic force is usually

small in the macro world, but becomes the driving mechanism for MEMS devices such as micro-grippers, MEMS RE switches, microfabricated gyroscopes and

ac-celerometers. In the case of microfluidic devices, which are usually the platforms for partiele analysis, scaling down results almost exclusively in laminar flow.

Microfabricated devices also benefit from batch fabrication, where devices are made in parallel at the same time. The consequence of batch fabrication is that many dif-ferent modules can be integrated on a single chip without significantly complicating

the fabrication process and increasing the price. Small overall size is also a charac-teristic feature of these devices, although external equipment is usually required.

Integrated Coulter counter

The integration of the Coulter principle into a microchip is presented in [36, 47]. These devices use a microchannel with multiple inlets and integrated electrodes,

see figure 1.7. Due to the laminar nature of the flow the hquids coming from side channels do not mix with the main flow. Therefore, generation of sheath flow is

possible with the multiple inlets. In this flow the particles are inside a small core flow which is sun"ounded, fully or partially, by a sheath flow. The advantage of this liquid "aperture" is twofold. First, there is no rigid wall between the core flow

and the sheath flow and so when a large partiele enters the inner tube it "expands" automatically. This makes the devices less prone to channel blocking than their

macrofabricated counterparts. The second advantage is that there is a possibility of using different liquids for the sheath and core flows. If low conductivity liquid is

used for the sheath and high conductivity liquid is used for the core flow, the system only appears to be a small conducting channel electrically isolated with a moving

shield. That is, when a low conductivity partiele flows in the core flow, it almost entirely replaces the high conductivity liquid causing a large change in impedance.

Optical detection using a photodiode array

Ontical detection of particles is possible with integrated photodiodes fabricated di-X I n d ™ ^ ^ ^ ^ ^ ^ The main advantage of this on-ch.p solutton rs that no

1.3. OBJECTIVES 9

Sheath

inlets _Electrodes

Figure 1.7: Photograph of an integrated Coulter counter [47]. The particles flow through the device in the (died) sample flow. The impedance change is measured by the electrodes with 4 point measurement.

complex optical system is necessary for the measurement (provided that the parti-cles are close to the detector). The design and measurements of a CMOS-integrated photodiode array is described in this thesis. The results obtained with bipolar pho-todiodes are presented in [47].

1.2.3 Measurement range of partiele sizing methods

The range of pardcle sizing methods is summarised in figure 1.8

1.3 Objectives

Before proceeding to the next chapter the objectives of this thesis have to be laid down. In the last section a short overview was given of the commercially avail-able partiele measurement techniques. Among these, even the compact ones are relatively large and expensive (for example the dimensions of the Analysette 22 COMPACT from FRITSCH GmbH are 64 x 52 x 32cm). There is a need for miniaturisation and, therefore, the objective of this thesis was to conduct scientific research to find new approaches for novel measurement methods in order to reduce the size and cost of partiele shape measurement devices. The development of the devices was motivated by a few possible applications.

Application No.1: In medical health care, shape analysis of red blood cells (RBC)

(12)

10 CHAPTER 1. INTRODUCTION

Scanning X-raV Disk Centrifuge Disd Centrifuge entometer

l

Normai irield Flow Fractionation

Photon

Dynamic Ligtit Scattering ton Correlaiion'sp'ecroscb: CaDiüJi Stati' Optical Micros

w

Steric& Hyperlayer FFF Eleqtrical &,Optca

Capiüary Hydrod 'nami

Ud it Scatterin ; R a c t Electron l\^icrbdcopiy Lk

i

one Sedimentètiör óunters Saïtibii onatior Time o

\AM

t

Bcanneirs Time of Trah

Acbuétiffi Spectroscopy

ition e JQl

r

100nm 1pm 10tJm

Figure 1.8: Range of partiele sizing methods [72].

1.3. OBJECTIVES

11 Apphcafon No.2: Most pharr^aeeuttcal manufacturing processes include a series of crystalltsation processes. The propert.es of the product after c^stal isaüon

are not on y important at final product stage buf a.so for the ^mS^Z he down-stream processes, such as fihrat.on, drytng, formulating etc U6

5 . The most tmportant properties include size and shape, which hav tö be measured m-s,tu for optimum performance. The aim with our inte' ated parfCe shape sensor ts to provide an altemative solution to the currentl us d sensor systems, which are mainly based on light scattenng.

scienffic fields. For example, the strength of fibre-re.nforced matenals can be mcreased rf the fibre particles are onented m the right dtrection In e ti

auons mto wood fibre and ceramrc fibre matenals were carried out m 2]

Ztmi^cTTf " " ° ' " " " ' ° " '"''"'' ''^ ^'^'^'^ -^«Cs ,s dtfïeretlt

cel s t s f r h ^ . ' ' = " ™ - ° " ^ f " ° " « " "e used as a tooi for detecting these cel s [68]. The onentat.on behaviour of bacterial flagellae [71] and DNA [701

molecules were also reported. J anu UINA L/UJ

t Ï Ï r o r i e ' f r " ' ' T " ' " " • ' ' ' '''''"' ""''^ '°' electro-onentation and par-tiele onentation can be determined using image analysis. The goal with the

mtegrated partiele shape sensor was to combine the electro-orTentad and image capturing in a single hybrid device.

a

::i°a?srÉT.T;;^"'''''' ^^^'^"'"-^

- '-^-^^

^p-'«-^- ^^ ^^^-^

(f-tor of 8 ' j " T h " ''^T^' ' " ' ' " ' " ' ^ " '^P'^^"^ 3^™ 'hick and have ü ame er of 8/im. Th.s size determined the smallest partiele s.ze, which was

tar-geted to be measured. Crystal particles in the m m range are not r^re but there 1 practical limitations for the maximum size that can be handled with n^ óde c

-w

rTr::r ^^'"

^"

T

^'^^^-^^^ °'''-

""^°"

*

-^-^ ^"'o

'^-sinn e 2 D ' ? ' ' i ' ' " ' ' * ' ^ ' " ' ' ' ^ ' " ^ ^'' ' ^ ^ « ^ ^ to be 250^.m. 2.) In a s ngle 2-D image it ,s not possible to distinguish a small spherical partiele from an

Z : 7 ' " ' " ' ' = ; " ' ' ' ^ ' P°'"'^ V^n>.n,icu^^r to the plane of the im'aTe.Ïi r l r e

(13)

12 CHAPTER 1. INTRODUCTION 1 Shape measurement of particles whieh are between 5^tm and

250/im in diameter. 3D measurement

6

7

Partiele coneentration measurement by counting. Possibility of in-line installation

Small size

Low cost device

Parallel processing

Table 1.1: Objeetives for the partiele shape sensor

parallel to each other. In-channel parallel data processing means that two or more particles can be delected at a time in one ehannel with no anomalous results.

r

In the following chapter our integrated partiele shape sensor will be described. A short overview will be given on the measurement prineiple and device operation. The challenges and completed tasks regarding design and fabrication will also be described.

Chapter 2 Integrated partiele shape sensor

In this chapter the integrated partiele shape sensor, the topic of this thesis, is intro-duced. The results and the reasons for the decisions are not explained here in detail Instead this is done in later chapters. The puqjose is only to provide a comraon Imk between chapters.

2.1 Selection of the measurement prineiple

f'

Instead of developing a new measurement method, reinventing the wheel in mi-cro scale is what should be the key concept for MEMS designers. Analysing the methods that exist in the macro world give the design pattem and it is the source for the "reinvention". The techniques of microfabrication supply the tools for die miniaturisation. This approach was foliowed during the design of the partiele shape sensor.

Amongst the measurement methods in the "macro world" shape analysis with imag-ing, see section 1.2.1, was chosen to be the device for "reinvention". The reason for this choice is simple, this method is the least complex to realize on-chip. Miniatur-isation of the SEM prineiple is not yet feasible because creating and controlling an electron beam in a sub-centimeter device is only now being researched [15]. Other methods including laser are efficiënt when there is a distance between partiele and detector, which is larger than what is available on-chip. The Coulter prineiple was investigated elsewhere [47] and sedimentation is difficult to implement in an in-line process due to the forces exerted by fluid motion.

(14)

14 CHAPTER 2. INTEGRATED PARTICLE SHAPE SENSOR

On-chip

Compact

Can be cost-effective

Continuous partiele flow Scanning

Resolution varies with the distance be-tween partiele and detector

Conventional

Large, heavy Expensive

Limited number of particles per mea-surement

Instant 2-D image

Resolution is by Abbe's formula

Table 2.1: Conventional versus on-chip image analysis systems

2.2 Conventional versus on-chip image analysis

There is a fundamental difference between conventional and on-chip image analysis. Conventional image analysis systems use a microscope to capture the image of the particles. This happens through a lens system. An extemal light souree, a CCD camera, and a computer for the calculations are peripheral devices necessary for the complete system, see figure 2.1 a). In the case of on-chip image analysis there is no space for a complex optical system and therefore only the diffraction pattern of the partiele is detected. Detection of the diffraction pattern rcquires the particles to be focused close (within few partiele distance away) to the detector. Saving the opties results in a compact device, such as the partiele shape sensor 2.1 b). This system also have to be surrounded with a computer and an external light source.

Of course there are some penalties of integration. ID scanning (line-by-line detec-tion) has to be used instead of 2D imaging. Particles have to be suspended in a fluid that is moving, the resolution is limited by the technology used, and the distance between the partiele and the detector has to be small, as will be seen later (section

3.1). Table 2.1 summarises the differences between conventional and integrated image analysis systems.

2.3 Quasi-3D partiele shape detection

Using two imaging modules in combination with two partiele orientation modules a quasi-3D rniage can be obtained about the particles. The principle of operation of the process is best demonstrated by a figure and a list of subsequent events. The partiele shape sensor is depicted schematically in figure 2.2 and shape is measured in the following steps:

1. Particles are pumped into a niicrochannel with the help of a syringe pump

2.3. QUASE3D PARTICLE SHAPE DETECTION

15

> i

a)

b)

S r i ' • ' " ' ' ' " ' ^ ' " ' '''''''''• ^'^ — ^ - n ^ l system, (b) integrated partiele shape sensor

Measurement 1 Measurement 2 After reconstruction

*! Flow direction

(15)

16 CHAPTER 2. ÏNTEGRATED PARTICLE SHAPE SENSOR

2. On-chip manipulation techniques (sheath flow, dielectrophoresis etc.) are

used to ensure a partiele position that is close to the bottom of the channel. 3. Paiticles are oriented such that their long axis is parallel to the flow

(electro-orientation at low frequency)

4. The shadow of the particles is scanned with a 1-D line-photodiode array. An external light source is used for illumination.

5. The long axis of the particles is oriented perpendicular to the

flow.(Electro-orientation at higher frequency) ,! 6. The shadow of the partiele is scanned again.

7. 3-D image reconstruction from the two 2-D images on a computer. 8. The partiele is disposed of or fed back to the system.

2.4 Description of the partiele shape sensor

In addition to the orientation modules, mentioned in the previous section, the device also include a dielectrophoretic focusing module and allows the generation of sheath flow through multiple fluid inlets, see figure 2.3 for the fabricated partiele shape

sensor'. The modules of the device are described below.

Sheath flow module

The microchannel is defined in a SU-8 (thick film photoepoxy) layer, which is pro-cessed on a siücon wafer. After sawing flie wafer Üie dies are capped with a glass chip. In these microchannels the development of sheath flow is possible (section 4.3.6). Sheath flow means that there are two flows in the microchannel; the core flow, also called the sample flow because the particles are suspended in that, and the buffer flow which surrounds the core. It is important that the two flows do not mix because microflows are in the laminar flow regime and it also means that the parti-cles stay within the sample flow, see figure 2.3 sheath flow module. There are three inlets for buffer flows, one inlet for a sample flow and one outlet. By changing the

flow rates at the buffer inlets positioning and shaping of the sample flow is possible.

In this way particles can be steered close to the detector, what is required for the detection of the diffraction pattern.

The photodiocle arrays have been l'abncated and Lested but not yet integrated to the device.

2.5. CHALLENGES AND COMPLETED TASK

Dielectrophoretic focusing

17

section 4.3^5), too. It focuses the particles into the bottom middie part of the flow in two steps. First centering is done than the particles are pushed down to the b o t L of the channel as depicted m figure 2.3 focusing module. As the dielectrophomk

h e T l ^ t S ' " " ^ ' " * ^ ''''''"' °^ * ^ ^'^^^"^ «^'^ '^ '^ -°- effectiTeto ₄

Electro-orientation module

Orientation of particles is done with the electro-orientation module depicted in fig-ure 2.3 electro-orientation module. The module consists of two electrode pairs fab-ricated on the top and the bottom of the channel (section 4.3.4). There are two of these modules for the two different orientations. Unlike dielectrophoresis, electro orientation is effective between electrodes as the orientation torques are generated by the electric field and not by the gradiënt of the electric field.

Photodiodes

Shape is raeasured by photodiode arrays. They will be placed in between the elec-trode pairs of the orientation modules. The minimum diode size is defined by the lechnology and is 2.4 x 2.4fj,m. In order to increase the fill factor the diodes were arranged in a shifted line array (section 3.2), see figure 2.3 schematic view.

Electrical and fluidic connections

As there are electrodes both on the bottom and the top side of the channel electrical connections are necessary to the silicon and the glass chips. The connection points are placed in two rows, one on the glass and one on the silicon chip, which, there-fore, have to be shifted, as figure 2.3 shows. The contacts are made through rubber connectors. The fluid connections are made via through wafer holes.

2.5 Challenges and completed task

(16)

18 CHAPTER 2. INTEGRATED PARTICLE SHAPE SENSOR co 03 O

E

C O

5

0) I m ^ ^ > ü co

E

0) o C/D 0) Ü • ^ ^ ^ > 0 • a " ü o co ü ra LL Dielectrophoretic focusing Push-down (sid^ viqw) _ ~ Centering co 03 CL

O

Sheath flow module f Buffer ^ flow

I

^Sample flow Front view Electro-orientation module side view Lowfrequency EO + E : : — \ -.EOA High frequency EO

sa

Push-down Front view Centering Glass , SU-8 j x Silicon Pad Photodiode array (Nol yet implemented)

Electrode contacts on the Silicon

Alignment structures

Through wafer

hole Electrode contacts _{on the glass}

Figure 2.3: The fabricated device

2.5. CHALLENGES AND COMPLETED TASK

Optical detection

19

From the available techniques photodetection with a line photodiode array was cho-sen to measure the diffraction pattem of the partiele (sectjon 3.2). Three types of photodiodes were tested and different levels of isolation between diodes were im-plemented.

Thennal, shot and l/f noise are present in all semiconductor circuits. With the exception of l/f noise, they can be filtered with integration of the signal with the penalty of loss in temporal resolution [44]. The measurements described in the next chapter show that the fabricated diodes perform well for the partiele shape sensor. Spatial quantization error is introduced as the array consists of pixels with finite size. The liniited minimum pitch, i.e. the minimum distance between diodes, defines the magnitude of this error The pitch size is defined by the fabrication technology and design layout. Optimising the layout and finding the best diode with the highest performance was the challenge in the design of the photodetector.

Partiele manipulation

There are many techniques for partiele manipulation in a microfluidic flow. Control with sheath flow, dielectrophoresis and eJectro-orientation were already introduced briefly. Understanding of these techniques is a challenging task. As analytical So-lutions are limited to special cases, a simulation tooi is inevitable for the design process. The development of such tooi had to be done because there were no com-mercial software available on the market that could offer partiele position and orien-tation tracking (section 4.3.1). Ourtool is capable of handling dielectrophoretic and electro-orientation forces, Brownian modon and motions induced by fluid shear and pressure. As all physical models, the model behind the simulation tooi also include idealisations. Therefore the validity of the tooi have to be verified with experiments in the future. The design and optimizadon of the partiele shape sensor was done with the help of this tooi.

Fabrication

(17)

20 CHAPTER 2. INTEGRATED PARTICLE SHAPE SENSOR

channel cross-section is also required and the process have to be CMOS compat-ible. In this thesis a new adhesive bonding technique was devcloped which uses

ahgnment stnictures defined in SU-8 (section 5.2.3).

Building a flexible measurement set-up which allows fluidic and elcctric contacts is also important and challenging the device holder is described in section 5.4.

Chapter 3 Optical detection

This chapter concentrates on the photodetector design of the partiele shape sen explamed. A shoU summary on photodetection and the available photodiode struc-h ^ T p struc-h o i r o d " ' - ' ' ' ' ' r ' - - ^ - ^ - ^'^ - - p a n s o n of tstruc-hese structur Tstruc-he Ï r h H H ""'"' """' ^"'"^ '° ^' '^' ""''' ^^^'^^ f«^ ^he partiele shape sensor The photodiode array is described in last section.

3.1 Lensless optical detection

In the case of conventional image analysis the image of the object (cells, crystal partiele etc.) is usually registered by a CCD camera. The image is obtained by the lens system of a microscope and the quality of this system determines the maxi-mum resolution. Abbe's formula 3.1 gives a mathematical limit to the maximaxi-mum resolution.

R = 0.61 X

NA' (3.1)

where R is the resolving power, A is the wavelength and A''^ is the numerical aper-ture. In the case of a microfluidic chip photodiodes can be fabiicated directly on the bottom of the channel. In these devices the distance between the partiele and the de-tector is only a few micrometers. In this condition the measurement of the shadow projected by the partiele becomes possible. Actually, in this case the diffraction pattem of the partiele is registered. The reason why the distance have to be small is shown in fignre 3.1. The diagram illustrates how light scatters at two points on the partiele surface. These points act as sources of outgoing spherical waves, which overlap and interfere. It can be seen that when the light intensity is measured with a line detector the received signal depends on the distance between the partiele and

21

(18)

22 CHAPTER 3. OPTÏCAL DETECTION

Detector signal 1 Detector signal 2

Figure 3.1: Propagation of light diffracted on a partiele. The detected light intensity pattern (simplified in the diagram) changes with the particle-detector distance.

the detector. The distortion of the signal limits the detection distance to few partiele length. In 3D the diffraction pattern is formed by the combination of the scattered light from the whole partiele suiface. Figure 3.2 shows two diffraction patterns

sim-ulated with the LightPipes Optical ToolBox under Matlab 6. The original shape of the partiele, figure 3.2 a), is a rectangle with long and short sides of 20^ni and

lO^m, respectively. Figure 3.2 b) and c) shows the diffraction pattern detected 20 and lOOfiin distance away from the partiele. The distortion of the original image is obvious in these patterns. When the distance is 20jxm the shape of the partiele is stil] visible, only the corners are roundcd. When the distance is increased to 100/.im the shape is almost completely lost.

Figure 3.2 d) and e) show the images that can be reconstructed from b) and c), respectively, with a threshold function using a grid of 2.4 x 2.4/i7?i and resolution

of 1-bit. Although the shape information is lost in e) the aspect ratio still gives useful information. (It has to be noted that the reconstruction of the exact shape of the partiele is possible and a simple algorithm is presented in [47].)

The resolution that can be detected without a lens can be approximated from the equation which is used to define the optical resolution of photolithography equip-ments [66],

(3.2)

where R is the resolution, D is the distance between the partiele and the projection screen and A is the wavelength of the light. If we apply it for a distance of 10/.im with a wavelength of 400nm, we see that distortion of 3/j,ï7i can be expected. This result makes it obvious that D has to be kept as small as possible and the shortest possible wavelength has to be used. Keeping D small is achieved with the dielec-trophoretic focusing module and the sheath flow.

3.1. LENSLESS OPTICAL DETECTION

(19)

24 CHAPTER 3. OPTÏCAL DETECTÏON 3.2. PARTICLE SHAPE MEASUREMENT 25

3.2 Partiele shape measurement

p

P'

3.2.1 Photodetection Photodetectors

Light inlensity can be measured in many different ways. However, most coniraonly semiconductor devices are used. When a photon with appropriate energy hits an atom in tlie semiconductor crystal, an electron from the valence band jumps to the conduction band leaving a conductive positive "hole" behind. When an electric field is present in the material, these photo-generated electrons and holes start to move

toward the electrodes. Those electrons and holes which reach the electrodes (which do not recombine) are coUected and form the photocuiTent signal. This current is proportional to the light intensity and is used for the measurement.

One of the most important properties of a photodetector is its overall quantum ef-ficiency rj. It is defined as the probability that an impinging photon generates an

electron-hole pair which is coUected on the electrodes. This property depends on many parameters a few of which are explained below.

• The quantum efficiency is wavelength dependent, through the penetration depth l/cv where a is the absorption coëfficiënt. The penetration depth

de-fines the depth in the material where the light intensity drops by a factor of e (to 37%). The relation between the penetration depth and the wavelength is shown in figure 3.3.

• The quantum efficiency depends on the diffusion length Lp, L^ in the material

{Lp and Ln are the average lengths that a hole and electron, respectively, can

travel without recombining).

• rj also depends on the voltage bias across the device.

• The optical power reflectance U has obvious influence on rj. The more light is reflected the less can be coUected.

• The dielectric layers on top of the silicon wafer also affect the quantum effi-ciency. This phenomenon is shown in figure 3.4 [31]. When there is no layer

on the silicon wafer, the optical transmittance (1 — 3R) increases monotoni-cally with the wavelength. However, when a layer of oxide is present on the wafer, the transmittance has peaks at certain wavelengths. If is important to note that, despite the troughs and crests in the function, the transmittance is

always larger than that of the pure silicon. This can be explained by the more continuous transition of the refractive index through the interface.

Ü. 1 0 '

b-200 ₄₀₀ _{600 800 1000}

Wavelength [nm]

1200

Figure 3.3: Penetration depth of light in silicon

• The overall quantum efficiency also depends on the number of doped regions, their location in the silicon bulk and their doping levels.

Many types of photodetector structures can be implemented in silicon with the com-bination of doped layers. Such devices include pn-photodiodes, CCDs (Charge-Coupled Devices), phototransistors, avalanche photodiodes, p-i-n photodiodes, pho-totransistors, Schottky photodiodes and silicon photomultipliers. The two most common of these are CCDs and CMOS pn-photodiodes. These two technologies compete with each other on the market of digital imaging and today there is no line dividing the types of applications each can serve. In university research, how-ever, the CMOS process has advantages over CCD. In CMOS, the fabrication of the photosensitive part and the readout electronics is possible in the same process. The accumulated knowledge on CMOS is more extensive than on CCDs. Tools and device models are more accessible. Because of these advantages, in this thesis only the pn-photodiode structure implemented in CMOS will be discussed further.

PN-photodiode

(20)

-26 CHAPTER 3. OPTICAL DETECTION 3.2. PARTICLE SHAPE MEASUREMENT ₂₇ 1.0 d =O.Vm ^ 0.8 h 0) o c: ca « p ^ H

E

m c co 0.6 0.4 0.2 SiO^/Si 400 500 600 700 800 900 1000 1100 Wavelength [nm]

Figure 3.4: Transmittance of sihcon versus wavelength

where q is the elementary charge, A is the active area of the detector, Pphoto is the irradiance on the surface, A is the wavelength of the light, h is the Planck's

constant and c is the speed of light. The quantum efficiency i] can be calculated analytically [61], but the necessary parameters, for example the diffusion lengths

Lp and L^ and doping levels are often not available. Therefore, an approximate

value was used for 77 in order to estimate the magnitude of the photocurrent of the partiele shape sensor. In [39] the quantum efficiency of a Standard 0.35/^7?^ CMOS process was investigated. In this process no anti-reflection layer was applied and

a single oxide layer covered the p-type epitaxial layer. The line with data points in figure 3.5 shows their measurement results of the external quantum efficiency of n"*" — pepj photodiode. The solid line shows the curve received with their

semi-empirical quantum efficiency model, where the substrate diffusion length was varied to receive the best fit. What can be seen in this figure is that r] = 0.5 is a safe approximate value. If the pixel size is 2.4 x 2.4/im, substituting this ?] value and

lOmW/crn^ optical power to 3.3 gives a photocurrent of about lOOpA, a value that

can be easily handled with CMOS electronics.

3.2.2 Photodiode structures

DIMOS, the CMOS process of DIMES (Delft Institute of Microelectronics and Submicron-technology) was chosen for the fabrication of the photodiodes. The ad-vantages of this process include higher quantum efficiency than modern processes,

o <D t^^m o ii= 0) E CD 13 O 420 470 520 570 620 670 Incident Wavelength (nm) 720 770 820

Figure 3.5: Quantum efficiency of a Standard 0.3^fj.m CMOS process [39]

\

n-well psuö n+ psub p+ n-well

psub

Figure 3.6: pn-junctions in the DIMOS process.

full wafer processing, possibility for small modifications in the process and the ac-cumulated design experience in our laboratory. The pn-junctions and diode arranse-ments available in DIMOS are described below.

Junctions

There are three pn-junctions available in DIMOS. They are substrate/ n-well p-substrate/ shallow-n and n-well/ shallow-p. The shallow-n and shallow-p layers are highly doped, while the n-well and the p-substrate (actually the epitaxial layer on the p substrate) are hghtly doped. The junctions are depicted in figure

2D photodiode matrix:

ine most convenient way to measure partiele shape seems to be to use a 2D pho-todiode matrix. With this sensor, all phopho-todiodes sample the light intensity at the

(21)

28 CHAPTER 3. OPTICAL DETECTION 3.2. PARTICLE SHAPE MEASUREMENT 29

Image: Image: X ^ o image Line Diode Array Shifted diode array

Ideal diode array a)

Possible closest arrangement

b)

Figure 3.7: Image of blood cell with an ideal a) and a real b) 2D photodiode array. The real photodiode array misses the cell with a 70% chance

same moment. The shape of the partiele can be directly obtained by mapping each photodiode to one pixel in the image. The problem with this method is that the

pitch, the minimum distance between diodes, and the minimum pixel size can not be lowered indefinitely because of the minimum resolution of the fabrication

pro-cess. Although, the so-called feature size of DIMOS is l.Gp.m, it does not mean that diodes can be fabricated this far from each other. It only means that the shortest gate length that can beused fora CMOS transistor is 1.6/.im. Each fabrication step hasits own minimum dimensions, for example, the minimum size of the metallisation layer

is 2.4/im and the distance between two n-well regions has to be larger than Sfim. Adding all the design restrictions together, the minimum pitch size for a 2.4 x 2.4/im diode implemented in the shallow-p/ well junction is larger than 5.6^m and in n-well/ p-substrate junction is approximately 16/i?n. The latter means that no shape

Information can be measured, even there is approximately a 70% chance of missing a blood cell, which is only 8/im in diameter. This is represented in figure 3.7.

Scanning:

The principle of scanning is to measure part by part. Scanning of particles in a microchannel can be used in the following way.

1. The particles have to be moved at a uniform speed over the detector by a fluid flow.

2. 3.

Each pixel of a line-photodiode array is sampled with a known frequency.

The image is reconstructed by placing the sampled pixels after each period into the image at a distance of Z = x; x At far from the previous pixel. Here

V is the velocity and At is the time between two measurements on the same

pixel.

Flow

direction _directionFlow

t

Measurement at "t"

Measurement before "t" Measurement to be done

Reconstructed shape

Figure 3.8; ID photodiode array for scanning

a

/ : ^

t

L'^^^^^^^^^^^^^^^^^^^^^^H.

• H^lLJ. 1 ^

F iiJ-rW

\ ^ f r V .f

Figure 3.8 a) shows this re construction process in the case of an ideal photodiode array where the pixels ai-e directly next to each other. However, the positions of the pixels are not forced into one continuous line. They can be shifted relative to each other in the direction of the flow. The reconstniction process in the case of such shifted arrangement is represented in figure 3.8 b). If we look at the figure we can see that every pixel of the partiele is read, only the order of reading has changed. It also has to be noted that only 1 bit resolution is used in figure 3.8 (the partiele either covers the pixel or not). There is, however, no barrier in front of an extension to an analogue, grey-scale method. In that case, the resolution can go sub-pixel.

The scanning method is important because with the shifted aixangement 100% fill factor can be achieved in one direction, despite the design rule constraints. Further-more, the length of the array is not limited and it is allowed for two particles to cross the detector at the same time (provided that they do not overlap each other), this we call in-channel parallel processing.

Strip photodiode:

(22)

30 CHAPTER 3. OPTICAL DETECTION 3.2. PARTICLE SHAPE MEASUREMENT ₃₁ Diode strip t Diode 1 t=ti ShallowP Diode 2 b) Current 1 Current 2

Figure 3.9: Strip photodiode

constant frequency. The result is a sequence of lines that represents the width of the partiele at different sections 3.9 a). The exact shape can be reconstructed if a

symmetrie partiele is assumed 3.9 a). However, this is not the only possibility, fig-ure 3.9 b) shows three different possible shapes. This problem can be somewhat remedied by using two strips as shown in 3.9 c). Tlie diode on the top measures a

larger current drop because a larger area is covered by shadow. Combination of the two signals provide the partiele shape. Adding lines in the same way further reduces the indeterminacy of the shape and in the limiting case the an'ay structure received. The strip photodiode array does not provide in-channel parallel processing, because there is no distinction between a large partiele and two particles crossing

simultane-ously.

Position sensilive device:

Normally position sensitive device (PSD) is used to determine the position of the centre of a light spot which illuminates the device. PSD works also in inverted mode, that is, it is able to determine the position where the otherwise uniform light

is blocked. Siich one dimensional PSD can be used in a similar way as the strip photodiode to determine partiele shape.

The construction of a PSD is a simple photodiode which has electrical contacts on both ends and one of its doped layer is high ohmic. The position is determined by

using the pai'asitic resistance of the high ohmic layer as figure 3.10 shows. The pho-tocurrent generated at the large arrow finds an easier path through the right contact because the resistance is smaller over a shorter distance, consequently the current

through the right contact will be larger than on the left. The position of the spot is determined from the difference signal between the two end electrodes. In the case of the inverted PSD cuiTent is generated at each point along the line except where

Figure 3.10: Position sensitive devi ce

Figure 3.11: Position sensitive device in inverse mode

the partiele blocks the light. The sum of the currents measured at the end contacts is proportional to the width of the partiele in the way that larger pailicle causes smaller signal. The difference between the currents at the two ends gives informa-tion about the posiinforma-tion of the partiele, see figure 3.11. The partiele is closer to the contact where smaller current can be measured. With the position measurement the PSD can provide accurate shape information but it also does not give the feature of in-line parallel processing because of the same reason as the strip photodiode structure.

r

The problem with the PSD method is that it requires significant sheet resistance in one of the implanted layers. The Standard DIMOS process does not provide such a layer. Even when the highest resistivity n-Well layer was pinched (made shallower) with a shallow-p layer (figure 3.12), the resistance is too low for proper device operation.

LOC

ShallowP

Figure 3.12: Pinching N-well with shallow-p. the depth of n-well region is educed by the shallow-p region.

(23)

32 Method 2D array Scanning array Strip photodiode PSD

CHAPTER 3. OPTïCALDEÏtCT JON

Objectives in table l.I

Ü o • I-H 'S O vn — - — —

+ + +

4—^ a TU S u _f i—i ^ ^ r J C/:J cd Cü

a

Q 1 m

+

+ +

++ + + +

+

+ + +

c/5 cd

s

a o - » — ' ca -i—k CU o c: o U

+ + +

c o - *-H 13 T ^w 4 _ l C/ï Ö <U C • >-H ( a

+ + +

++ + + +

+

CU O É ^ ^ 1) X3 ^ ^ F "o cd D H

e

o U

+ + +

+ + + + + +

O i-i OH 13 Ï3 OH ,^ . CU ^^^^ Ö jz: o

+ + +

++

^ -- — A^^ in o u o

+

++

+ + +

1 1 1 1 L 1

3.3. PHOTODIODE ARRAY DESIGN ₃₃

Table 3.1: Comparison of the detection methods

Comparison

The question: "Which method is the best for the partiele shape sensor?", can be answered by considering the objectives of this thesis listed in table 1.1. According to this table a clear decision can be made about which method is the most attractive. The 2-D array cannot measure the targeted size range with the DIMOS technology

and more complex readout is necessary. The strip and PSD methods are inefficiënt in measuring concentration and are unable to perform in-channel parallel

process-ing. PSD has additionally fabrication issues, too. The best choice remains the method with a ID line-photodiode array implemented in CMOS technology^ It

can measure the targeted size range, can be equipped with the 3D orientation mod-ule discussed in chapter 4, multiple particles do not disturb the measurement (unless

they overlap each other) and it can also measure higher concentration level. Small size, low cost and in-line operation are inherent characteristics of the microfluidic approach.

3.3 Photodiode array design

In this section the shifted line-photodiode arrays designed for the partiele shape sensor is presented.

'Implementation in bipolar technology also has promising features. Nieuwenhuis investigaled these alternatives in [47].

Junctions

In the previous section the three pn-junctions available in DÏMOS were introduced. All of them can be used forphotodetection, but only the shallow-p/ well and the n-well/ p-substrate photodiodes were implemented. The n-n-well/ p-substrate junction differs from the other two in that it resides much dceper in the silicon, as shown in figure 3.6. As will be seen later many diode arrangements were implemented in the design. In order to reduce the number of possible combinations only the shallow-p/ n-well photodiode was tested from the two possible shallow junctions. The reason for choosing the shallow-p/ n-well junction is shown in figure 3.13 a). This diode is protected against the noise coming from the bulk silicon by the reversed bias n-well/ p-substrate diode. It is especially important in designs where on-chip digi-tal electronics is present, becau.se the switches inject spur noise into the substrate. Another advantage of the shallow-p/ n-well photodiode is that misdiffusion (lateral cross-talk) between pixels is expected to be lower. The reason is the lower diffusion length [65] in the n-well region than in the p-epi layer (above the bulk p-substrate). As can be seen in the figure 3.13 b) misdiffusion took place in the shallow-n/p-substrate, that is an electron was captured by a neighbouring diode. This did not happen in the shallow-p/ n-well diode, because the hole (in the n region holes are the minority carriers) recombined earlier due to the shorter diffusion length. The diffusion lengths for the DIMOS process were estimated to be 173.2 - 10~^cm and 3480.1 • lO^^cm for the n-well and p-epi, respectively [23].

Size

Light can generale electron-hole pairs everywhere in the silicon. In order to min-imise the generation and so the collection of charge carriers from outside the diode,

the detector area was fully covered with a metal shield. This shield was open only at the active areas. Figure 3.14 shows two times two photodiodes. In 3.14 a) the size of the diode is 2.4 x 2.4/im and it is the smallest allowed in DIMOS. The left one is shielded, while the right one is open. The situation is the same in figure 3.14 b) but the diode size is 15 x 15/im. The lines around the pixels are for electrical isolation and are explained in the next paragraph.

Cross talk considerations

(24)

-%

34 CHAPTER 3. OPTICAL DETECTION

n+/ psub diode

p+/ n-well diode

a)

Electron is collected by neighbouring diode

Recombination occurs before colleting.

Minority charge carrier generated under diode 1.

n+/ psub diode p+/ n-well diode

b)

Figure 3.13: Comparison of the shallow junctions: a) substrate noise is collcctedby the n+/psub diode, b) missdiffusion in the

n+Zp^^^-Shielded pixels

guard rings for electrical isolation

Active pixels a)

b)

j

Fi^ure 3.14: Shallow-p n-weli photodiode pixel, a) 2.4 x 2Afim, b) 15 x 15/^m

i

t

3.3. PHOTODIODE ARRAY DESIGN

35 Design rule

Minimum width: n-well

Value Minimum distance between n-wells

Sfj,m Minimum width: First interconnect

Distance between n-well and

n-+-8fim

2.4y7m Minimum width: p(n)+

Minimum width: Second interconnect Minimum distance between metal

4.8/im

1.6/j.m.

2.4//.m 2.4/im

Width of shal]ow-p(n)/ metal contact ^^^IST^ Table 3.2: Main design rules of DIMES

Figure 3.15 depicts the different arrangements, In the simplest case there is no separation, figure 3.15 a). All the photodiodes share the same n-well and the mini-mum pitch between the diodes is determined by the distance between metal contacts which is 2A(im + 4/zm = 6.4/j,m. In the second type the diodes were placed into separate n-wells, figure 3.15 b). The pitch in this case is detennined by the width of the n-well and the distance between n-wells. Adding these two values results in 16/im pitch. In the last design the wells were additionally surrounded with a p-f-guard implantation ring, at the oost of an extra 11.2/im microns in pitch, figure 3.15 c). Finally figure 3.15 d) depicts an n-well p-substrate diode array. The pitch of this array is also 16/.im.

It is important that increasing the pitch has the negative consequence that higher voltage have to be applied for the electro-orientation module. Additionally, as the partiele crosses the array in longer time there is more chance for random lateraJ movement during the measurement, which introduces error in the shape measure-ment.

Protection against spur noise

(25)

IT'

36 CHAPTER 3. OPTICAL DETECTÏON

n

f-'- - ^ ; UVgyJTmT^

- ' >1.w.i»*trf< I L _

-^T—^ r "-" 1. ' J ^ T T I ' j .

a) Diodes share a common n-well b) Diodes sit in their own n-well

c) Diodes sit in their n-well and are fully separated with a shallow-p ring

d) N-well p substrate photodiode array

Figure3.15; Photodiode types

TilBTif^

/ '

.*-noise collected by the guard rings

Figure 3.16: Photodiode types

3.3. PHOTODIODE ARRAY DESIGN

37

Shfefded diodes

Active diodes

Figure 3.17: Test array

Measurements

From all types of photodiode layout 4 size variations were made, 2.4 x 2.4fj,Tn,

4 X 4fini, 8 X 8/2771 and 15 x 15/^m. On each array 4 diodes were measured for comparison. The first diode (DDD) was completely shielded with metal, so it was

expected to give only the dark current. Additionally, both of its neighbouring diodes are also shielded, as figure 3.17 shows. The second diode (DDL) is also dark, but one of its neighbouring diode is open. The third diode is open with one dark diode next to it (DLL). Finally the fourth is open with two open diode neighbouring it (LLL).

(26)

38 Q) i : o Cü 0,OOE+00 -5.00E-11 -1.00E-10 -1.50E-10 -2.00E-10 Fï -2.50E-10 "D O - I — * O CL -3.00E-10 -3,50E-10 -4.00E-10 •4.50E-10 O.OOE+00 -1.00E-10

f

0) h u 03 TS O -2.00E-10 -3.00E-10 -4.00E-10 O Q. -5.ooe-io -6.00E-10 -7.00E-10

CHAPTER 3. OPTICAL DETECTION

1V reverse bias

Light source on level 2

DDD_2 DDL_2 DLL_2 LLL 2 Measurement set 1V reverse bias

Light source on level 3

10 ₁₅ ₂₀ ₂₅ DDD_3 DDL_3 DLL_3 LLL 3 Measurement set

Figure 3.18; Measurement results on a 2.4 x 2.4^m shallow-p n-well photodiode array. The lateral cross talk (Lc=DDL-DDD and Lc=LLL-LLD) is apparent but the covered and open signals are still clearly distinguishable.

\

3-4. SUMMARY

3.4 Summary

39

b

(27)

40 CHAPTER 3. OPTICAL DETECTION

Chapter 4 Partiele behaviour in

mieroehannels

In the previous chapter partiele size and shape detection with a line-photodiode array was described. The measurement is based on scanning, which needs accurate partiele transport over the detector and constant partiele onentation. This chapter describes how these partiele manipulation tasks are solved in a microchannel. The first two sections survey the different manipuladon techniques, while section 4.3 present the resulls.

4.1 Partiele behaviour in microflows^

This section explains the characteristic behaviour of microflows and partiele motion in shear flow. Both analytical and numerical results found in the literature will be summarised.

4.1.1 Characterisation of microflows

Before investigating the forces which are exerted on particles by the surrounding fluid, it is necessary to determine the characteristics of the fluid and the behaviour of the flow itself.

The buffer material in which the particles are suspended is mostly water-based and can be treated as a Newtonian Hquid, that is the dynamic (absolute) viscosity fi

'The definition of microflow is not straightforward and varies from aulhor to author. In this thesis microflow will refer to flows where the characteristic length (diameter) of the channel falls into the micrometer range.

(28)

42 CHAPTER 4. PARTICLE BEHAVIOUR IN MICROCHANNELS _{4A. PARTICLE BEHAVIOUR IN MICROFLOWS}

43

Figure4.1: Flow regimes: (a) laminar, (b) laminarseparated flow, (c) turbulent flow is constant at all velocities. Furthermore, the pressure and temperature conditions present in the channel are always such thatthe fluid can be treated as incompressible. Other liquids of interest, such as isopropanol, have the same properties.

In general there are three flow regimes classified in fluid dynamics. Tlie dimension-less Reynolds number determines which regime the flow falls into. The Reynolds number is expressed by the ratio of inertial forces over viscous forces and is defined by

Dvp Re =

^

(4.1) where Re is the Reynolds number, D is the characteristic length, v is the mean velocity and p is the density of the liquid. When Re is lower than 2000, the flow is called laminar. The streamlines" run along side each other and the adjacent fluid layers do not mix. In turbulent flow the Reynolds number is higher than 4000. In this region eddies form, which intermix the fluid layers (mixing occurs). The

region between the laminar and turbulent flows is called the transition region, where characteristics of both laminar and turbulent flows can be observed. Figure 4.1

I

depicts the different flow regimes. Reynolds defined the border between the linear and transition region at Re = 2000. However, this value seems to be incorrect for microflows, mostly because of the non-circular channel cross section. Obot

and Jones developed a method [26] which takes care of the non-circular channel cross section by substituting the characteristic length with the "laminar equivalent" diameter, _D]e, defined by: '

D

4>

In fluid dynamics, a streamline is a line which is tangent to the velocity of the flow at all points.

where 0 is the geometry function, which is, for a fixed cross section geometry, a function of the aspect ratio only. Although the Obot-Jones method was developed for macrochannels, its validity for microflows can be concluded from the experi-mentaJ work in [73, 2, 21], An important consequence of scaling down is that the Reynolds number almost exclusively stays small in microflows, it means that the flow can be treated as laminar in all circumstances. This fact greatly simplifies the investigation of partiele dynamics and will be our assumption throughout this chap-ter for both theoretical and numerical analyses. In addition, when the Reynolds number is very small, particles can be assumed to adapt to the flow conditions im-mediately and so acceleration can be canccled out, as will be seen later.

Another important assumption which will also be used is the non-slip boundary con-dition. It simply expresses the fact that the fluid velocity at the solid fluid interfaces is always equal to the velocity of the solid body. Therefore the velocity of the liquid is equal to zero at the channel wall and the flow velocity at each point around the partiele is equal to the velocity of the particle's surface at that point. In our analy-sis the non-slip boundary condition breaks only when an electrical field generates fluid flow at the channel wall. This will be considered in section 4.1.2 where the electro-osmotic principle will be explained.

4.1.2 Pumping

Fluid is brought into motion by the action of pumping. Many different methods were developed in the past; using extemal pumps provided an immediate solution, but later a vast variety of on-chip methods were also developed. A broad overview is given of the different principles in [46]. In this thesis extemal pumping and electro-osmotic pumping will be explained.

External pumping

Using an extemal pump is the least troublesome method for supplying bio-chips with the analyte and buffer solutions. Despite its simplicity there are some parame-ters that need to be considered:

Complicated, long chamiels make pressure drop calculation difficult and so pressure driven pumping is not generally applied forprecise flow rate control. Vibration of the system can cause flow fluctuation to build up in the mi-crochannel.

Integrated particle shape sensor

Integrated Partiele Shape Sensor

Peter Turmezei

May 26, 2006

Integrated Partiele Shape Sensor

Contents

Preface

Project history

Chapter 1

Introduction

Organisation of the thesis

1.1

When and where do we encounter particles and what

do they "look" like?

tor(Sht) ' ' ' ' " " ' ' '''™'' ^ ' ' '''"'"^"^ ' ' " " " '' ''''''''' ^" ' ""'''''

1.2 Partiele analysis

I

1.3 Objectives

10

CHAPTER 1. INTRODUCTION

l

w

i

\AM

r

a

::i°a?srÉT.T;;^"'''''' ^^^'^"'"-^

- '-^-^^

^p-'«-^- ^^ ^^^-^

-w

rTr::r ^^'"

^"

T

^'^^^-^^^ °'''-

""^°"

*

-^-^ ^"'o

Chapter 2

Integrated partiele shape sensor

2.1 Selection of the measurement prineiple

f'

2.2 Conventional versus on-chip image analysis

2.3 Quasi-3D partiele shape detection

2.4 Description of the partiele shape sensor

2.5 Challenges and completed task

E

5

E

O

I

sa

Chapter 3

Optical detection

3.1 Lensless optical detection

3.2 Partiele shape measurement

p

P'

E

t

a

t

• H^lLJ. 1 ^

F iiJ-rW

+ + +

a

+

+ +

++ + + +

+

+ + +

s

+ + +

+ + +

++ + + +

+

+

e

+ + +

+ + +

+ + + + + +