review of production of microfluidic devices

8/4/2019 Review of Production of Microfluidic Devices

1/12

Review of production of microfluidic devices:

material, manufacturing and metrology

Shiguang Lia, b, c

, Zhiguang Xua, b, c

, Aaron Mazzeoe, Daniel J. Burns

eGang Fu

a, d, Matthew Dirckx

e,

Vijay Shilpiekandulae, Xing Chen

a, d, Nimai C. Nayak

a, d, Eehern Wong

e

Soon Fatt Yoona, b, Zhong Ping Fangc, Kamal Youcef-Toumie, David Hardte, Shu Beng Tora, d,

Chee Yoon Yuea, d

, Jung-Hoon Chune

aSingapore-MIT Alliance, N3.2-01-36, 65 Nanyang Drive, Singapore 637460

bSchool of Electrical & Electronic Eng., Nanyang Tech. University, Singapore

cSingapore Institute of Manufacturing Tech., Singapore

dSchool of Mech. & Aerospace Eng., Nanyang Tech. University, Singapore

eDept. of Mechanical Engineering, MIT, USA

ABSTRACT

Microfluidic devices play a crucial role in biology, life sciences and many other fields. Three aspects have to beconsidered in production of microfluidic devices: (i) material properties before and after processing, (ii) tooling and

processing methodologies, and (iii) measurements for process control. This paper presents a review of these three areas.

The key properties of materials are reviewed from both the production and device performance point of views in this

paper. The tooling and processing methodologies considered include both the direct tooling methods and the mold basedprocessing methods. The response of material on the production parameters during hot embossing process are simulated

for process control and product quality prediction purpose. Finally, the measurements for process control aspect discuss

different measurement approaches, especially the defect inspection, critical dimensional measurements, bonding quality

characterization and checking functionality. Simulation and experimental results are used throughout the paper toillustrate the effectiveness of such approaches.

Keywords: microfluid, lab-on-a-chip, bio-MEMS, material, tooling and processing, measurement

1. INTRODUCTIONMicrofluidic devices are characterized by feature sizes in the micron- or sub-micrometer range and store or exchange a

small amount of fluid. Microchannels, reaction reservoirs, microvalves, sensors, and other related components are

conventional parts in such devices. One important application of microfluidic devices is in bio-chemical fields in whichthe solutions of cells, protein, DNA, or other bio-chemical substance lie [1, 2]. The microfluidic devices exhibit many

advantages compared to the larger reaction devices, e.g., consume less reagent and power, increased portability, shorter

turnaround time, and more accurate control. The manufacturing of microfluidic devices has received increasing attentionin last two decades due to its promising application in the above fields. Generally speaking, three aspects must be

improved to produce a high quality microfluidic device: (i) material properties and the prediction of its properties on

tooling and device performance, (ii) tooling and processing methodologies, and (iii) measurements for process control.

One program in the Singapore-MIT Alliance (SMA) which involves around 20 researchers at MIT, Nanyang

Technological University and National University of Singapore, systematically studied the production of microfluidic

devices since 2005 in these three areas [3]. This paper briefly reviews their representative results achieved to date, aswell as work done by other researchers. The simulated and experimental results throughout the paper are completed by

the SMA researchers except where otherwise noted.

2. MATERIAL CHARACTERIZATIONMaterial properties are the first consideration in production of microfluidic devices, simply because the chosen materialshould be fit for the device performance, e.g., bio-compatible to do biological experiments, resistant to acid or alkali to

do chemical experiments, etc. The second reason is that the interaction between the materials and tooling or processing

MEMS, MOEMS, and Micromachining III, edited by Hakan UreyProc. of SPIE Vol. 6993, 69930F, (2008) 0277-786X/08/$18 doi: 10.1117/12.781942

Proc. of SPIE Vol. 6993 69930F-1


2/12

methodologies (T.P.M.) significantly affects the product quality. It means the product quality could be very different

using different materials even with the same tooling or processing method. The material cost is another consideration forlarge-scale, high-volume manufacturing. It is therefore necessary to analyze the impact of material properties on the

product quality or device performance. There are many materials that have been mentioned to make microfluidic devices,

e.g., silicon [4], quartz and glass [4], polydimethylsiloxane (PDMS) [4], poly(methyl methacrylate) (PMMA) [5], SU-8

[6], cyclic-olefin-copolymer (COC) [4], polycarbonate (PC) [4], polyethylene (PE) [6], polytetrafluoroethylene (PTFE)

[7], polyethylene terephthalate (PET) [7], polyacrylate [8], polystyrene [9], cellulose acetate [10], and even ceramics [11].Table 1 summarizes some requirements for materials related to the device performance, T.P.M., and others. The polymer

materials obtain wide applications in microfluidic devices because of good bio-chemical performance and low cost.

Polymer based microfluidic devices will be emphasized in the following discussion.

Table 1. Some requirements for materials in production of microfluidic devices

Categories Applications Material requirements

Biological experiments Bio-compatibility [12]

Chemical experiments Chemical stability [6], i.e., resistance to acid, alkali, solvent, organics or

others.

Cell/molecular solutions Low porosity [6], low adsorption

High driving pressure

applications

High bonding strength [6], low surface roughness

Solvent dose needs

accurate control

Lower water absorption [12]

Performance

related

Reaction observation High optical transparency [13]

Laser direct tooling

technique*

High refractive index to achieve smaller feature size (laser point), less

optical non-linearity [14], high viscosity to achieve high manufacturing

precision, strong absorption at laser wavelength

Thermal based technique

(hot embossing, injection

molding, etc)

Less thermal expansion coefficient [4], matched thermal properties

between master/tools or tools/parts**

Master Excellent thermal/mechanical/geometry quality, easy to be peeled off

from tools

Tool Good thermal/mechanical/geometry replication quality, robust tosubsequent processing procedure, easy to be peeled off from master and

final part [15]

Device Easy to process, required bio-chemical/optical/mechanical/thermal/geometry/bonding quality

Tooling and

processing

related

Stiffness [7], mechanical stability [4], surface and interfacial tension, adhesive affinity of or between

materials [15, 16]. Find the most matched master/mold and mold/device pair.

High aspect ratio High viscosity for UV exposing processes [17], low viscosity for thermal

processes

High manufacturing

precision

Less optical non-linearity [14] and less shrinkage, laser direct tooling

method not applicable

Geometry

related

Sub-m feature size High refractive index for laser tooling approaches, low viscosity for

thermal processes, high viscosity for UV exposing processes [17]

Bonding One layer Bonding between feature layer and cover/substrate [6]



3/12

related Multilayer Bonding between same materials besides the above [6], ease of

interconnect [4]

Low material price PolymerCost related

Less time consumption Short processing time

*The definition of the direct tooling techniques and the other processing techniques are to be discussed in section 3.

**The part directly used to make molds is denoted as master, and the part used to make final parts is denoted as mold or working toolin this paper.

The researchers in SMA program investigated the interaction between some representative materials and the T.P.M. and

analyzed the possible influence on product quality or device performance, as listed in Table 2. The impact of material

characteristics on the product quality or device performance is evident. Generally speaking, several potential materials

should be chosen based on the device performance at first, and then the final material is determined by analyzing the

interaction between the materials and the specific tooling or processing method. The device geometry, material bondingand cost should be considered at the same time. The most popular polymers making microfluidic devices in literature are

PDMS, PMMA, SU-8 and COC for the mold based processing technologies, and the Foturan photosensitive glass for the

laser direct tooling technologies. The traditional silicon and glass are also widely applied in microfluidic devices despite

they are more expensive.

Table 2. Some material properties studies related to device performance and T. P.M.

Effect Material T.P.

M.

Measurement results Conclusions

Molecular

weight (MW)

on surfacequality [18]

PMMA

Laserablation

Surface morphology of laser cut

microchannel with lower MWs is

smoother and less porous. Withincrease in molecular weight, the

number of pores increases but the

pore sizes decrease too.

Materials with lower MW may be more

suitable for bio-chemical experiments

with laser ablation technique.

MW onthermal

bondingquality [19]

PMMA

H

ot

embo

ssing Bonding strength increases with the

increase of PMMA MW. The

bonding pair, whose MW is thegreatest, is the strongest.

Materials bonded with higher MW mayendure higher flow pressure with

thermal bonding technique.

Ease of

demolding of

molds from thesilicon master

and the

replication

precision [15]

Molds:

COP, PC,

PEEK,

PSU, PEI

Hotembossing

PEEK and PEI demold from the

master most easily. The adherence

between PEEK and silicon is thesmallest. The geometry repeatability

of PEEK is a little worse than that of

the other materials. PC and COP are

also demolded easily. COP is the

easiest to process. The master isbroken when the PSU tried to demold

from the master.

Perhaps PEEK is a good mold (tool)

material if the replication precision is

not high. PSU is not a good moldmaterial when using silicon master and

with hot embossing method. PC and

PEI are potential to be tool materials

with good geometry duplication

precision and demolding ease with hotembossing method. COP perhaps is

more suitable for final parts.

Replicationprecision when

PDMS is cast

on molds [16]

Molds:PC,

PMMA,

PTFE,

PCTFE,Al*

Casting

Casting PDMS achieves goodgeometry replication quality for all

the molds at room temperature.

Rough PTFE surface replicated in

PDMS.

The geometry deformation is smallwith the casting method. No observed

correlation between height/width of

microchannels and simple work of

adhesion estimates for PDMS againstthe tested molds.



4/12

PDMS

shrinkage and

distortion

with/without

rigid carrierscaffold when

it is cast onsilicon/SU-8master [20]

Scaffold

material:

Al,

Teflon,

PC.

Master is

silanized

to reduce

itsadhesion

to PDMSCasting

No essential shrinkage or distortion

when PDMS is cured at room

temperature. No significant

correlation between feature sizes and

shrinkage.

At higher curing temperature (85oC),

the shrinkage variation is between

0~1.84% for Al carrier scaffolds.

Certain areas of the PDMS-Alinterface have been peeled. Average

distortion is less for the scaffold with

less thermal expansion coefficient.

When PDMS is released fromscaffold, additional 1~2% shrinkage

takes place.

Thermal contraction is the primary

factor leading to shrinkage and

distortion. The residual stress resulting

from the thermal contraction may peel

PDMS off the Al scaffold at highcuring temperature.

Residual stress lies in PDMS with

carrier scaffold, which may lead to

1~2% shrinkage when releasing.

Rigid carrier scaffold is potential to

transfer PDMS without distortion.Special surface treatment or mechanical

fixturing is needed to prevent PDMSfrom peeling off the scaffold if

necessary. Scaffold should choose the

material with less thermal expansion

coefficient.

*PCTFE: polychlorotrifluoroethylene. Al: Aluminum

3. TOOLING AND PROCESSING METHODOLOGYTooling and processing methodologies refer to the methodologies used to make the features in microfluidic devices

before post-processing like packaging, bonding, sensor integrating, etc. Again, there are many tooling and processing

techniques applied to make microfluidic devices in publications. From the simplest mechanical machining [21] to

complicated lithography [22] to the nanoimprinting technique [23], almost all kinds of micro-manufacturing approachesare applied, and these methods are mostly from the semiconductor, Micro-Electromechanical System (MEMS) or rapid

prototyping industry. This section will briefly list some representative tooling and processing methodologies. The

detailed description of these techniques and their strengths and weaknesses can be found in some classical overviewbooks [24, 25]. After a certain manufacturing method is chosen, process control or quality control has to be conducted to

make the desired parts by optimizing the production parameters, e.g., temperature, pressure, holding time, laser power,

etc. Process modeling plays an important role in this procedure because it helps researchers to discover the internal

relationships between the feature variation during processing and the production parameters or the material properties

[26, 27]. The product quality also can be predicted by the process modeling. This section will introduce some processmodeling work for process control as well.

The tooling and processing of polymer microfluidic devices are generally divided into two categories: direct tooling

techniques and the mold based processing techniques. The representative techniques are shown in Fig. 1[21-25, 28-37].

The direct tooling techniques refer to processes in which the final parts are directly shaped with some machining tools,e.g. diamond knife, laser beam, or other shaping tools. The mold based processing techniques refer to processes in which

the final parts are replicated from the molds. The direct tooling methods are frequently used for rapid prototyping or the

conditions where the complex structures are needed. They are also used as the assistant method to make the special

features on a mold based device, e.g. through holes, or irregular shapes. The mold based techniques are widely applied in

mass production of the polymer microfluidic devices as it provides fast speed and low cost manufacturing. The productsmade with direct tooling techniques and with some lithography techniques in Fig. 1 also can be used as molds for further

replications or as final parts.Micro embossing and injection molding techniques have wide applications to the production of microfluidic devices dueto the short processing time and low cost of the final parts. Loke et al. [38] successfully made a micro-mixer based on the

injection molding technique with PMMA material. Micro-mixers are widely applied in microfluidic systems regarding

biochemistry analysis, drug delivery and sequencing or synthesis of nucleic acids, whose structures are illustrated in Fig.

2(a). Amorphous metallic glass mold machined by high speed milling machine was used in the experiment. The product

quality at region 1, 2 and 3 in Fig. 2(a) is shown in Fig. 2(b). Yeo et al. [39] made a high aspect ratio nickel mold withelectroplating method and a polymeric micro-array by UV-embossing method. The SEM images of the mold and the



5/12

v

. . _ . I W W "

patterned array are shown in Fig. 3. This experiment was conducted for 64 times, and the polymeric array can be

demolded from the nickel mold successfully every time.

Direct tooling

technique

Mold based

processing

technique

Laser ablation (direct

writing): inside and

outside ablation

stereolithography

(Photo-, UV-, X-ray)

lithography, LIGA

Etching/DRIE*

micro- and nano-

imprinting

EDM*

E-beam forming,

lithography

Micro embossing

Mold

manufacturing

Mold

replication

Injection molding

Postprocessing

Material bonding, sealing,

welding (laser, thermal,

ultrosonic, glue...)

Sensor intergration,...

Mechanical machining

Casting

Soft lithography

FIB etching, milling,

lithography

Fig. 1. Popular tooling and processing methodologies in production of microfluidic devices

* DRIE: Deep reactive ion etching. EDM: Electrical discharge machining

(a) (b)

Fig. 2. Micro-mixer made with injection molding technique. (a) Schematic diagram of the micro-mixer. (b) SEM images and

surface roughness at regions 1, 2, and 3. [38]

Fig. 3. Micro-arrays with aspect ratio 3 made with UV-embossing technique. Left: Nickel mold. Right: polymer arrays. [39](Courtesy of Yeo)

Region 1 2 3

SEMimage

Ra 223nm 211nm 239nm



6/12

E q u i v a l e n tP l a s t i c S t r a i n

1 9

. j j33 . 03 . 5 93 4 .: 1 . 3 93 . 2 9 .3 . 2 03 . 1 03 . 0 0

0E

U

d )

F

2 5

2 Q

' 61 0

1 2l 0 /

I ! I l 4 1 J L l J l i 1 1 1 1 1 l ? l IT i m e 5 )( a

I > 1 ! 4 ) J N I ) 5 I 1 1 : 1 1T i m e I s )

1 )

As far as the processing modeling, Henann et al. [40] simulated the geometry variation during hot embossing with a

large-deformation constitutive theory. The simulated sample is a piece of bulk metallic glass, and the mold is silicon. Along microchannel is to be molded on the glass. Four parameters are involved in the hot embossing process: molding

temperature, molding pressure, holding time and demolding temperature [15]. The production parameters in simulation

are tool pressure, holding time, as shown in Fig. 4(a), and molding temperature, i.e. 420oC. Demolding is not simulated

here because demold is not needed during the embossing of metallic glasses. Fig. 4(b) shows the consequent tool

displacement during processing, which is proportional to feature depth to be molded for a given pattern. Thedisplacement becomes almost stable after 20s. The glass will be deformed under pressure through time. The profile of

the metallic glass after 20s and 120s is shown in Fig. 4(c) (d), respectively. The mold is only partially filled after 20s,

Fig. 4. Geometry variation through time under pressure: 20MPa, temperature: 420oC. (a) Tool pressure variation. (b) Tooldisplacement variation. (c) Glass geometry at 20 seconds. (d) Glass geometry at 120 seconds. [40]

and is fully filled after 120s. The grey degree represents the equivalent plastic strain. The strain is the greatest at the

corner, and decreases gradually away the corner. According to the simulation, the tool pressure and holding time could be optimized, and the final geometry and strain distribution can be predicted. The experimental results are in good

agreement with the simulation. However, for the hot embossing technique, there is not only a concern with geometry

variation during processing, but also with predicting geometry variation after the removal of pressure. Based on the

framework of Anand [41], Chen et al. [42] deployed a thermo-visco-elastic-plastic constitutive model to describe theresponse of amorphous polymers under combined thermal and mechanical loads. This model could capture the major

features of the stress-strain mechanical response of an amorphous polymer. The material simulated is PMMA with a

glass transition temperature (Tg) of 110oC. Fig. 5 shows simulated results of stress-strain response of compression tests.

Fig. 5(a) indicated that for a processing temperature less than Tg, approximately 80%- 90% total strain remains as aresidual strain upon the removal of the applied force. Thus, there is permanent deformation, and this is conclusive for

embossing the necessary geometrical features. In contrast, if the processing temperature is above Tg, due to larger strain

recovery upon unloading, see Figure 5(b), geometrical features formed on embossing will not replicate well the features

of the mold. This implies that the demolding temperature should not be above Tg of PMMA. This is because the materialbehavior above Tg has transitioned to classical rubbery-like response, see Fig. 5(b). However, Figure 5 indicates that

with a decrease in embossing temperature, the higher loading force is required. Thus, simulation could provide



7/12

2 - ' o r 4 5 r _

0 2 0 4 0 6 1 ) 8 0 C M )C o m p I E i v e f l u e s t r a i n ( % I

0U

3 _ S3

2 52

1 5

0 _ S0

0 2 0 4 0 3 0 1 0 0C o m p r e i v e m a e f l a i r l ( c

information on the choice of processing parameters. In addition, it could provide deformation history and pattern if a

detail analysis is carried on the whole embossing process.

Fig. 5. Series of compressive true stress-strain curves at strain rate of 2x10-1s-1 below Tg (left) and above Tg (right) [42]

4. MEASUREMENTS FOR PROCESS CONTROLProcess control or quality control is a necessary step to make high quality products, which may involve process modeling

or prediction (discussed in section 3), sensing or measurements, and control algorithm development. In productionenvironments, the controllable variables are often production parameters. However, end users are concerned with other

parameters that may affect the device performance, e.g., critical dimensions, fluid leakage, bonding strength, etc. Process

control is necessary to translate between these two kinds of parameters, and to further control the production parameters

to make the desired devices. The measurements discussed here are mainly those of concern to the end user. Suchmeasurements not only provide feedback information for process control, but also characterize the quality of the final

part. The measurements of microfluidic devices are widely varied due to device feature size and structures, as well as the

instrumentation and fabrication methods. No universal approach exists, and the measurement methods must be

determined by specific requirements. However, this paper tries to categorize the measurements of microfluidic devices

into four general classes according to the measurement purposes, as discussed below.

4.1

Defect inspectionDefect inspection is the most popular measurement task as the unwanted defects tend to occur in each processing step.

The defects can be in the form of particles, pattern defects, process-induced defects, scratches, incomplete etches, and

many others. Defect inspection aims to find any possible defects in microfluidic devices and/or estimate their distribution

or overall sizes. Defect inspection does not need too much quantified data of the complex features; the measurementaccuracy and repeatability requirements therefore are not strict. However the measurement resolution and speed should

be high enough to distinguish the defects from the features frequently, or even piece-by-piece. There are wide varieties

of configurations to detect the defects in production, and almost all configurations are from semiconductor, MEMS andoptics industries. The common defect inspection technique is the combination of a camera or microscope together with

the bright field, dark field or tilt illumination [43, 44]. Figure 6(a) shows one side of a microchannel of a silicon mold

obtained with dark field microscopy. The edge is not straight. In recent years, the chromatic confocal microscopy iswidely applied in material inspection to provide some depth information in one image with colors, in which the red color

represents the deeper position and the blue color is the shallower position [45]. Figure 6(b) shows an image of one

surface obtained with chromatic confocal microscopy. The surface is not flat as judged by the non-uniform colordistribution. One particle at the right bottom corner is evident. To ascertain the smaller defects whose feature size is in

nano-scale, the scanning electric microscopy (SEM) is often applied [46]. As the microflduidic devices are often polymer

based, special SEM utilizing high water vapor pressure in the specimen chamber may be more applicable [47], in whichthe polymer material does not need coating. Some special measurement techniques are potential to measure some special

cases. For example, there are some parallel microchannels in microfluidic devices which look like an optical grating.

Evident diffraction occurs when monochromatic light transmits through the features. Taylor et al. [48] studied thedimensional variation in components according to the diffraction patterns with fast speed.



8/12

l e to w d e r \ / B i o d e r

L ) M i x i n gI n j e c t i o n m o I d i n g

D e b i n d i n g 1 1 1 1 1 1

S i n t e r i n g / M i o r o s t r u o t u r e s

Fig. 6. Detect defects with dark field microscopy (a) and chromatic confocal microscopy (b). (a) Image of one side ofmicrochannel of a silicon mold. (b) Image of one curved surface with one particle. Red color represents the deeper

position and the blue color represents the shallower position.

4.2 Critical dimensional measurementsThis kind of measurements is to quantify the critical dimensions (CDs) or the consequent parameters in microflduidic

devices, i.e., width, length, depth, sidewall slope, surface roughness, flatness, etc. Such measurements are necessary inproduction because there are often some internal relationships between them and the production parameters or material

properties. As the variation of CDs or CD-related parameters in two working conditions could be very small, for example,

the surface roughness variation could be only several tens of nanometer with two kinds of similar materials, the CDcharacterization has to be accomplished with high measurement resolution, accuracy and repeatability. Interferometry,

confocal microscopy, stylus profilemetry, AFM and critical dimensional scanning electron microscopy (CD-SEM) [49]

are conventional measurement techniques used to characterize the open features before bonding. The detailed

comparisons of these techniques can be found in Ref. [50]. Characterizing the CDs of the multi-layer devices after

bonding or the complex 3D devices made with the direct tooling methods will be discussed in Section 4.3.

As an example, Figure 7 shows an application of interferometry in a dimensional variation study for the production of amicro-pillars array by micro powder injection molding (PIM) [51].There are four processing steps inPIM: mixing,

Fig. 7. Study dimensional variation during PIM process with interferometry technique. (a) Processing steps ofPIM. (b)SEM image of the part after sintering. (c) Global profile of the sintered part. [51]

injection molding, debinding and sintering, as shown in Fig. 7 (a). The test pattern is an array of 2424 stand alone microcylindrical pillars, whose SEM image and the global profile are shown in Fig. 7(b), (c) respectively. These

microstructures are possibly applied as molds for micro plastic injection molding, e.g. microtiter-plates for DNA tests,

nerve stimulation needles, etc. By observing pillar heights, diameters, pitches, surface roughness and global curvaturethrough the processing steps, it was found that the shrinkage at the sintering stage was the largest due to the burning

(a)

(b) (c)

(b)(a)



9/12

away of material. Surface roughness increased gradually due to the removal of material in the debinding and sintering

processes. The bending direction of the molded parts was always the same, while the debinding tended to flatten thebending.

To speed up the CD characterization while not losing the feature details in production, it is desired to measure CDs in a

larger field-of-view and at a high measurement resolution. However, this requirement is difficult to achieve with a single

technique, or by a single instrument. Shilpiekandula et al. [52] investigated an approach to meet this requirement by

fusing the data from multiple instruments, namely, the AFM and interferometer, as shown in Fig. 8. Some critical

positions such as the sidewalls of the microchannels were measured with AFM, and the global profile was measured withthe fast speed interferometer. The two sets of data were fused together with a correlation-based alignment algorithm.

Figure 8(a) shows the combined 3D profile of a feature whose raw data are from two sets of high resolution data from

the AFM and one set of large range data from the interferometer. Different colors represent the variation in height.Figure 8(b) compares the cross sectional profile of the three sets of data. The data fusion algorithm also can enlarge the

measurement area by stitching a series of adjacent data sets of small areas.

Fig. 8. Combine the data sets from AFM and interferometry. (a) Combined 3D profile by combining two sets of data fromAFM and one set of data from interferometry. (b) Comparison of the cross sectional profile of the three data sets. [52]

4.3 Bonding quality characterizationA special and critical procedure in the production of microfluidic devices is the bonding of a cover or substrate to a

featured layer or the bonding of several layers together. The bond should be strong enough to prevent separation, freefrom air gaps around channels (to prevent fluid leakage), conform to dimensional specifications after deformation, etc.

Again, the camera and microscope are the conventional inspection techniques used to observe the bonding related issues,

e.g., the 2D geometry of the buried features after bonding, air gaps around the channels, air bubbles, etc. The SEM isalso applied to observe the small details after the device is sectioned. Additionally, the confocal Raman and two-photon

microscopes have been used to characterize the UV bonding quality in some publication [53].The interferometer with a

2000 2500 3000 3500 4000

8

10

12

14

16

18

20

22

24

Microvalvethickness(m)

Spin-coating speed (rpm)

Before bonding

After bonding

Fig. 9. Measurements of microvalve thickness with white light confocal microscopy. (a) 3D volume image of a microvalve

region. h is the microvalve thickness, h~10m in the image. (b) Comparison of microvalve thickness before and afterbonding. [56]

compensator [54] and the confocal microscope (whose objective is specially designed by considering the covers

thickness and refractive index) are often used to characterize the one layer samples bonded to a transparent cover. The

h

(a)(b)



10/12

CD characterization of the multilayer devices after bonding is rarely discussed. However, such characterization is

necessary because the geometry tends to distort after bonding especially with thermal bonding methods. In thesemiconductor industry, multilayer devices are often characterized with X-ray tomography [55]. However, X-ray

tomography is not applicable for microfluidic devices because the polymer device often exhibits low x-ray absorption.

Fortunately, most microfluidic devices are optically transparent, and the optical confocal technique has the potential to

form a 3D volume image of the multilayer devices. Li et al. [56] obtained the 3D volume image of a bi-layer device with

white light confocal microscopy. The microvalve thickness in this device was quantified, as shown in Fig. 9(a). The twobright layers represent the two surfaces of the microvalve. The lower layer is the open layer and the higher layer is the

buried layer. h represents the microvalve thickness. The microvalve thickness before and after bonding was measured

across the range of spin-coating speeds, as shown in Fig. 9(b). The spin-coating speed is the controllable production

parameter when making a microvalve. There is consistent and evident thickness shrinkage during bonding. Note that thelaser confocal microscopy is not applicable to directly reconstruct the 3D volume image of the multilayer microfluidic

devices as there are numerous interference fringes between multiple surfaces. The conventional ellipsometer is also notapplicable because the beam size is much greater than the microvalve dimensions: 100x100m2.

4.4 Functionality characterizationFunctionality characterization is the last measurement to check the feasibility of the microfluidic devices after thedevices are fabricated. It may include: examining if the fluid flows as expected, noting if the biological or chemical

reaction occurs, determining if the integrated sensor works, or measuring if the microvalves close or open the channels.

The fluid flow measurements are the most representative functional measurements in microfluidic devices. It can bedivided into two categories: fluid self imaging and fluid velocity imaging, as shown in Table 3. Fluid turbulence tends to

Table 3. Comparisons of fluid related characterization techniques

Category Sub-category Strengths Limitations

Conventional fluorescent

imaging technique [57]

Fast speed 2D measurements

Laser scanning confocal

microscopy [58]

3D measurements Slow. Depth resolution is

limited by diffraction

Multi-photon excitationfluorescent microscopy (multi-

photon microscopy) [59]

3D measurements. No pinhole.Little florescence bleaching

High power pulsed laserFluid imaging

technique

Optical coherence tomography

(OCT) [60]

3D measurements. Compact

equipment

Slowest. Depth resolution is

limited by wavelengthbandwidth of light source

Particle image velocimetry

(PIV) [61]

Fast speed 2D measurements

Fluid velocity

imaging

technique

Doppler optical coherence

tomography (Doppler OCT, or

optical Doppler tomography,

ODT) [62]

3D measurements Lower scanning speed

Laser Doppler velocimetry

[63]

Fast speed. Compact

equipment

2D measurements

appear when fluid flows in microchannels. Bonesi et al. at Cranfield University studied the turbulence close to the junction of a T shaped microchannel with Doppler OCT [62]. Fig. 10(a) shows the microchannel structure. The

turbulent cross-sectional velocity profile at the T junction is shown in Fig. 10(b). The blue color represents the lower

velocity and the red color represents the higher velocity.

In general, there are four kinds of measurements in the production of microfluidic devices. The measurement techniques

are widely varied due to the varied measurement purposes. Defect inspection and CD measurement techniques are



11/12

0 . 5 1 . 0 1 . 5L a t e r a l s h i f t , m m

5EEn a0 ) 2 .

mostly from the semiconductor, MEMS and optics industries, and some reviews can provide good references [50, 64].

The bonding inspection is a special but critical measurement task in production of microfluidic devices. Thefunctionality measurements widely appeared in publications because the functional measurements are the most direct

way to evaluate a final microfluidic device, and to observe the reactions in microfluidic devices. When measuring the

microfluidic devices, the commercial techniques based on monochromatic light should be used with caution due to the

serious light interference between multiple surfaces. However, the interference or diffraction pattern is also possible to

provide much useful information with fast speed in some cases. The x-ray or electronic based techniques should also beused with caution due to the high transparency to x-ray and serious charging effects to e-sources of the polymer devices.

Fig. 10. Study on fluid turbulence near the T junction in microchannel. Left: Microchannel structure. Right: Cross-sectional

velocity profile at the junction. [62]

5. CONCLUSIONSThree aspects related to the production of microfluidic devices, namely material, tooling and processing methodologies,and measurements, have been reviewed in this paper. There are abundant potential materials to make microfluidic

devices. The choice of materials should be based on the device performance, response to manufacturing methods, device

structures, and others. The conventional tooling and processing methodologies were briefly reviewed. It is possible tointegrate the different techniques to make one microfluidic device. Prediction of the behavior of the material during

processing is a critical work for not only the design of the processing equipments but also for process control. Some

process modeling work shows its capability to predict the final part quality. Measurements of microfluidic devices are

necessary for defect detection, bonding quality inspection, and critical dimensional and functionality characterization.The measurement results provide the feedback information for process control and the product quality characterization.

6. ACKNOWLEDGEMENTSThe authors would like to thank Dr. Ivan Reading in Singapore Institute of Manufacturing Technology for the early

discussion about this work. The first three authors checked the whole paper. Authors 4-6 checked the measurements,

tooling and processing methodology, and material sections, respectively. The other co-authors checked the work done bythem in this paper and David L. Henann made some comments on his work.

REFERENCES

[1] Stepanek, J., Pribyl, M., Snita, D., et al., Biomicrofluidics, 1(2), 024101/1-11 (2007).[2] Llopis, S.L., Osiri, J., and Soper, S.A., Electrophoresis, 28(6), 984-93 (2007).[3] http://web.mit.edu/cpmweb/index.html

[4] Gartner, C., Becker, H., Anton, B., et al., SPIE 4982, 99-104 (2003).[5] Narasimhan, J., and Papautsky, I., SPIE 4982, 110-119 (2003).[6] Peng, Z.-C., Ling, Z.-G., Goettert, J., et al., SPIE 5718, 209-215 (2005).[7] Sandison, M. E., and Morgan, H., J. Micromachanics and Microengineering, 15(7), S139-144 (2005).[8] Zhou, W. X., and Chan-Park, M. B., Lab on a Chip, 5, 512-518 (2005).[9] Bubendorfer, A., Liu, X. M., and Ellis, A. V., Smart Materials and Structures, 16(2), 367-71 (2007).[10] Mohamed, H., Russo, A. P., Szarowski, D. H., et al., Separation Science and Technology, 42(1), 25-41(2007).[11] Le, Hue. P., Le, Hoi. P., Elder, D. B., et al., US patent 20060050109

inlet

outletoutlet

1mm



12/12

[12] Lee, J. H., Peterson, E. T. K., Dagani, G., et al., SPIE 5718, 82-91 (2005).[13] Fujii, T., Microoelectronic Engineering, 61-62, 907-914 (2002).[14] Fisette, B., and Meunier, M., SPIE 5578, Part 2, 677-686 (2004).[15] Dirckx, M., Mazzeo, A. D., and Hardt, D. E., ISNM 5, conf126a34-r39-1-6 (2008).[16] Mazzeo, A. D. and Hardt, D. E., ISNM 5, conf126a26-r30-1-6 (2008).[17] Jurischka, R., Blattert, C., Tahhan, I., et al., SPIE 5718, 65-72 (2005).[18]

Nayak, N. C., Sinha A. T., Yue., C. Y., et al., ISNM 5, conf126a124-r122-1-4 (2008).[19] Nayak, N. C., Tan, Y. L., Yue, C. Y., et al, ISNM 5, conf126a138-r122-1-4 (2008).[20] Wong, E., Hong, V., Mazzeo, A., et al., ISNM 5, conf126a28-r25-1-5 (2008).[21] Bai, Y. L., Koh, C. G., Juang, Y. J., et al., AIChE Annual Meeting, 847-852(2004).[22] Mali, P., Sarkar, A., and Lal, R., Lab on a Chip, 6(2), 310-15(2006).[23] Horsley, D. A., Talin, A. A., and Skinner, J. L., J. Nanophotonics, 2(1), 021785-1-11(2008).[24] Madou, M. J., [Fundamentals of microfabrication: the science of miniaturization], 2nd ed., CRC Press, 2002.[25] Rai-Choudhury, P., [Handbook of microlithography, micromachining, and microfabrication], SPIE Optical

Engineering Press, Bellingham, 1997.[26] Mitra, A., [Fundamentals of quality control and improvement], 2nd ed., Prentice Hall, 1998.[27] Wadsworth, H. M., Stephens, K. S., [Modern methods for quality control and improvement], 2nd ed., Wiley, 2002.[28] Andrieux, G., Eloy, J. C., and Mounier, E., SPIE 5718, 60-64(2005).[29] Chung, C.K., Lin, Y. C., and Huang, G. R., J. Micromechanics and Microengineering, 15(10), 1878-1884(2005).[30]

Kang, H.-W., Lee, I. H., and Cho, D.-W., J. Mfg. Sci. & Eng., Transactions of the ASME, 126(4), 766-771(2004).[31] Yao, P., Schneider, G. J., and Prather, D. W., SPIE 5718, 73-81(2005).[32] Singleton, L., Detemple, P., Frese, I., et al., SPIE 4984, 29-37(2003).[33] Bertsch, A., Jiguet, S., Bernhard, P., et al., Materials Research Society Symposium - Proceedings, 758, 3-15 (2003).[34] Sugioka, K., Masuda, M., Hongo, T., et al., Applied Physics A, 79(4-6), 815-817(2004).[35] Fu, C., Hung, C., and Huang, H., Journal of Physics: Conference Series, 34(1), 330-5(2006).[36] Romanato, F., Tormen, M., Businaro, L., et al., Microelectronic Engineering, 73-74, 870-5(2004).[37] Malaquin, L., Vieu, C., Genevieve, M., et al., Microelectronic Engineering, 73-74, 887-892(2004).[38] Loke, Y. W., Tor, S. B., Chun, J.-H., et al., Antec 2007 - SPE Annual Technical Conference, 1980-1984 (2007).[39] Yeo, L. P., Lam, Y. C., Chan-Park, et al., ISNM 5, conf126a56-r54-1-6 (2008).[40] Henann D. L., and Anand L., Acta Materiallia, submitted (2008).[41] Ames N. M., Srivastava V., Lele S.P. and Anand L., ICOMM 1, No.50 (2006).[42] Chen, X., Lam, Y. C., and Yue, C. Y., ISNM 5, conf126a97-r102-1-6 (2008).[43] Dimalanta, E. T., Lim, A., Runnheim, R., et al., Analytical Chemistry, 76(18), 5293-5301(2004).[44] Lee, C.-Y., Lee, G.-B., Lin, J.-L., et al., J. Micromechanics and Microengineering, 15(6), 1215-1223(2005).[45] Baumann, H.-G., US patent 6038066, filed 1997.[46] Cheng, G. J., Pirzada, D., and Dutta, P., J. lithography, fabrication and systems, 4(1), 04033, 1-8(2005).[47] SEM: http://dentistry.umkc.edu/microscopy/esem.htm[48] Taylor, H. K., and Boning, D. S., ISNM 5, conf126a41-r42-1-6 (2008).[49] Frase C. G., Buhr E., and Dirscherl K., Meas. Sci. Technol., 18, 510519 (2007).[50] Novak, E., SPIE 5716, 173-181 (2005).[51] Li, S. G., Fu, G., Reading, I., et al., Appl. Phys. A 89, 721-728 (2007).[52] Shilpiekandula V., Burns D. J., Youcef-Toumi K., et al, ICOMM 2, No.35, 323-328 (2007).[53] Schrof W., Beck, E., Etzrodt, G. et al., Progress in Organic Coatings, 43, 19 (2001).[54] http://www.veeco.com/products/details.php?cat=1&sub=5&pid=263&option=3&detail=1862&template=2navIII[55] Tkachuk, A., Duewer, F., Cui H.-T., et al., Z. Kristallogr., 222, 650-655 (2007).[56] Li, S. G., Thorsen, T., Xu, Z. G., et al., ISNM 5, conf126a53-r37-1-4 (2008).[57]

Li, P.C.H., de Camprieu, L., Jia, C., et al., Lab on a Chip, 4(3), 174-80(2004).[58] Peach, J. P., Hitt, D. L., and Dunlap, C. T., 2000 ASME Int. Mech. Eng. Congress & Exposition, 2, 497-503 (2000).[59] Dittrich, P. S., Schwille, P., Analytical Chemistry, 74(17), 4472-4479(2002).[60] Pierce, M.C., Joo, C., Cense, B., et al., SPIE 5692, 174-8(2005).[61] Van Steijn, V., Kreutzer, M.T., and Kleijn, C.R., Chemical Engineering Science, 62(24), 7505-14 (2007).[62] Bonesi, M., Churmakov, D. Y., Ritchie, L. J., et al., Laser Phys. Lett., 14 (2006).[63] Chuang, H-S., Lo, Y-L., Optical Engineering, 46(2), 024301-1-9 (2007).[64] Osten, W., [Optical inspection of Microsystems], Taylor & Francis Group, 2006.

P f SPIE V l 6993 69930F 12

review of production of microfluidic devices

Documents