Microfluid Nanofluid (2008) 4:427–437

DOI 10.1007/s10404-007-0196-x

RESEARCH PAPER

Analysis and experiment of capillary valves for microfluidics

on a rotating disk
Jerry M. Chen Æ Po-Chun Huang Æ Mou-Gee Lin

Received: 20 April 2007 / Accepted: 27 June 2007 / Published online: 27 July 2007

Springer-Verlag 2007

Abstract This paper presents an analytical expression of applications in analysis or synthesis for chemistry and
the pressure barrier in a capillary-burst valve for flow biology (Manz et al. 1990; Reyes et al. 2002; Auroux et al.
regulation in centrifugal microfluidics. The analysis con- 2002; Nguyen and Wereley 2002). In these microfluidic
siders variations of the interfacial energies at the meniscus systems, the capability of precisely controlling fluid flow of
of three-dimensional (3D) configuration in a rectangular small amount is essential. The large surface-to-volume
microchannel with a sudden expansion in cross-section. ratio in microfluidics makes it possible to control micro-
We derive a simple expression that predicts the critical fluidic flow through the use of capillary forces as a valve.
burst pressure or rotational speed to overcome the capillary Unlike the conventional diaphragm valves, capillary val-
valve. Experiments were carried out for capillary valves ving requires no moving parts (Oh and Ahn 2006).
that were integrated with microchannels on a rotating disk Capillary valves may be designed actively utilizing
having various cross-sectional dimensions (300 and different effects such as electrocapillary and thermocapil-
400 lm in width and 80–600 lm in depth) and wedge lary (Marangoni) to change surface tensions of the
angles (30–100) of sudden expansion. The flow visuali- meniscus (Lee et al. 2002; Nguyen and Huang 2005). More
zation of the meniscus development across the capillary often, they are designed passively as a check-valve simply
valve supports the assumptions made for the present by making an abrupt change of geometry in the hydro-
analysis. The measurements of burst rotational speeds for phobic or hydrophilic microchannels (McNeely et al. 1999;
the capillary valves are in good agreement with the pre- Duffy et al. 1999) and the trigger of flow can be achieved
dictions by the simple expression except that those with a by centrifugal force or other forces like electrics and
larger channel width and wider wedge angles are nearly pressure. Passive capillary valves have been frequently
10% lower than the predicted values. used to regulate liquid flow in the compact disk (CD)-based
centrifugal microfluidics to which the ease of implement-
Keywords Capillary valve
Rectangular microchannel
ing valves is vitally important (Madou et al. 2001a,b).
Burst pressure
Centrifugal microfluidics Centrifugal microfluidics has been demonstrated to provide
promising platforms for efficient mixing and high-
throughput screening, while its applications to sensing and
1 Introduction diagnostic tests require more severe technical challenges
(Gustafsson et al. 2004; Grumann et al. 2005; Haeberle
Integrated microfluidic systems have received rapidly et al. 2006; Madou et al. 2006). In this study we present the
increasing attention over the past decade for their novel analysis of three-dimensional (3D) model along with the
experiments on capillary-burst valves for regulation of
microfluidic flow on a rotating disk.
J. M. Chen (&)
P.-C. Huang
M.-G. Lin Figure 1 shows the schematic configuration of the
Department of Mechanical Engineering, CD-based hydrophilic microfluidics where the capillary-
National Chung Hsing University,
Taichung 402, Taiwan
burst valve is connected to a reservoir near the rotational
e-mail: [email protected] center through a straight microchannel. When the disk is at

123
428 Microfluid Nanofluid (2008) 4:427–437

1D theory of Eq. 2 may suffer significant error when the

microchannel geometry deviates largely from the assumed
circular shape. Duffy et al. (1999) modified the 1D theory
to obtain an expression of the form

Dpb ¼ að4cla =Dh Þ þ b ð3Þ

with two empirical constants, a accounting for variation of
channel cross-section and b representing the pressure
required to wet the expanded volume beyond the valve.
The term b depends on the geometry of the volume to be
filled and the wettability of its walls. Man et al. (1998)
derived a 2D theory for the barrier pressure developed in
the capillary valve from the energy changes in the liquid–
air–solid interface system. However, the 2D theory that
Fig. 1 Schematic of microfluidics fabricated on a rotating disk
equipped with a capillary valve to regulate the centrifugal-driven was derived assuming large aspect (depth-to-width) ratio of
flow. A straight microchannel links the reservoir and the capillary the channel cross-section may not be valid for channels
valve having an aspect ratio near unity or smaller. Numerical
simulations of 3D multiphase flow to compute the barrier
rest, the liquid stored in the reservoir flows to fill the pressure have been carried out by Zeng et al. (2000) based
hydrophilic channel but stops at the inlet to the suddenly on a Navier–Stokes solver coupled with the volume of
expanded valve due to the capillary force (barrier pressure) fluids method. The 3D multiphase flow simulations often
created on the liquid front in the direction towards the require considerable computational time. Glière and Del-
center. As the disk is rotating the centrifugal force induces attre (2006) later approached the 3D simulations by solving
a pressure at the liquid front opposite to the barrier pres- the static momentum equations for the equilibrium free
sure. The centrifugally induced pressure Dpc, that is in the surface. However, the numerical solutions for this simpler
form of artificial gravity acting on fluid volume, for the approach could still become complicated and challenging
resting liquid plug simply yields dependent on the geometric parameters chosen for the
surface profile.
Dpc ¼ qX2 Drrm ð1Þ An analytical expression of the barrier pressure based on
where q is the liquid density, X is the angular frequency, a 3D configuration should be greatly helpful for design and
Dr = r2 – r1 is the length of the liquid plug with its fronts fabrication of the capillary-burst microfluidic valves. The
located r1 in the reservoir and r2 in the capillary-burst valve analysis would also offer a more transparent, in-depth
from the rotational center, and rm = (r2 + r1)/2. When the understanding of the physics occurring in the capillary
rotational speed of the disk exceeds a threshold value, the valve. Leu and Chang (2004) recently reported the analysis
centrifugal force becomes larger than the capillary force of pressure barrier based on the model of 3D meniscus.
causing the liquid to burst into the expanded volume of the However, the magnitude of burst pressure obtained in their
valve. Capillary valves can be designed with different wedge Fig. 6 is shown to decrease with increasing aspect ratio
angles b associated with the channel dimensions to create (AR) for AR > 1 appearing to be inconsistent with the 2D
different barrier pressures. It is important to calculate these results of Man et al. (1998). This decreasing trend of burst
two forces for manipulating the rotational microfluidic flow. pressure with aspect ratio in the large range will also be
Despite the simple expression of the centrifugally shown opposite to the measurements of the present study.
induced pressure, the barrier pressure on the other hand can Cho et al. (2007) most recently explained the working
be extremely complicated depending on the capillary valve principle of a capillary-burst valve based on the theory of
geometry and the liquid–air–solid interfacial properties. contact line movement in a circular channel meeting a
For capillary valves with axisymmetric cross-sections and suddenly expanded section and obtained the burst condi-
a sudden opening of b = 90, the maximum barrier pres- tion of the valves in both circular and rectangular channels.
sure at the liquid front termed the burst pressure Dpb is Nevertheless, a more rigorous analysis based on a 3D
given by (Zeng et al. 2000) structure is still desired to clarify the fundamental mech-
anism of a capillary-burst valve.
Dpb ¼ 4cla sin hc =Dh ð2Þ In the following sections, we first present the 3D mod-
eling and analysis of the capillary-burst microfluidic valves
where cla is the liquid–air surface tension, hc is the contact
for rectangular transport channels with various aspect
angle, and Dh is the hydraulic diameter of the channel. This

123
Microfluid Nanofluid (2008) 4:427–437 429

ratios. The analytical expression of burst pressure derived

from the 3D model is compared with the results of afore-
mentioned 1D and 2D theories. Then we describe the
experimental apparatus developed for visualization and
measurements of the liquid flow in the rotating microflui-
dics. The visualization experiments provide sequential
images of the microfluidic flow. The measurements of burst
rotational speeds for the capillary valves are used to verify
the analytical results.

2 Analysis of capillary valve

2.1 Pressure barrier

Figure 2 illustrates the 3D configuration of liquid injected

into a microchannel with a sudden opening. The channel
has a constant cross-section of h (depth) · w (width) from
an arbitrarily axial position xc = 0 to L and then is followed
by a divergent section with a wedge angle b in the width
direction while the depth remains unchanged throughout
the entire channel. In the present 3D model, the meniscus is Fig. 2 Schematic configuration and parameters defined for the
assumed to have two circular arcs in both the width and meniscus developed in a rectangular capillary channel. Only part of
the straight channel in Fig. 1 is shown here. The straight channel
depth directions with angles of 2aw and 2ah, respectively.
starts from an arbitrarily position xc = 0 and ends at the valve edge
The thermodynamics in terms of interfacial free energy xc = L
may be applied to obtain the pressure barrier (Kim and
Whitesides 1997; Man et al. 1998) for the solid–liquid–air
system. The total interfacial energy UT of the solid–liquid–
air system is given by The pressure in the liquid may also be determined using
the Laplace equation
UT ¼ Asl csl þ Asa csa þ Ala cla ð4Þ

1 1
where Asl, Asa and Ala are solid–liquid, solid–air and liquid– p ¼ cla þ ð7Þ
R w Rh
air interface areas, and csl and csa denote the solid–liquid
and solid–air surface tensions (surface energy per unit where Rw and Rh are the principal radii of curvature of the
area), respectively. After the use of Young’s equation, meniscus in the width and height directions. In fact, it is
which relates the contact angle to the surface tensions easier to obtain the liquid pressure using Eq. 7 than Eq. 6
involved, the total energy becomes for the straight microchannel regime (xc < L). By
substituting the geometric relations
UT ¼ U0  Asl cla cos hc þ Ala cla ð5Þ
w h
where hc is the equilibrium contact angle and Rw ¼ and Rh ¼ ð8Þ
2 cos hc 2 cos hc
U0 = cla(Asl + Asa) represents the constant energy
component due to constant sum of the solid–liquid and into Eq. 7, we arrive at the Young–Laplace equation of the
solid–air interfaces. The pressure p in the liquid can be form (Kim et al. 2002)
derived from the change of total interfacial energy of the

solid–liquid–air system with respect to the injected liquid 1 1
p ¼ 2cla cos hc þ ð9Þ
volume Vl as w h

 The above equation indicates that a positive, constant
dUT dAsl dAla
p¼ ¼ cla cos hc  ð6Þ pressure propels the liquid originated from an arbitrary
dVl dVl dVl position xc = 0 toward the sudden opening at the end of the
where only the liquid–air surface tension and solid–liquid straight microchannel. This positive, constant pressure can
and liquid–air interface areas are involved. be realized from Eq. 6. As the liquid floods the uniform

123
430 Microfluid Nanofluid (2008) 4:427–437

channel of hydrophilic surface (hc < 90), Ala is fixed and f ðaw ; ah Þ ¼ 0 ð12Þ
Asl increases linearly with the liquid volume Vl. The liquid
maintains constant pressure until its front reaches the The constraint governs the ‘‘trajectory’’ of meniscus
intersection of the sudden opening (xc = L), referred to as growth in the first stage. For a rectangular channel made
the transition regime. of the same material, it is reasonable to assume the
In the transition regime, the solid–liquid–air contact line constraint of awah = 0, which means an equal variation in
is fixed at xc = L where the cross section of the channel is aw and ah. For the channel constructed using different
discontinuous and an increase in the liquid volume requires materials, the equal variation constraint may be no longer
meniscus curvature to reduce through both the angles aw valid and one should consider a more complex relation
and ah. The total interfacial energy UT in this regime can be between aw and ah. With a known constraint on the
expressed as meniscus curvatures, the pressure in the first stage yields

      1 
oUT oUT of of
w2 aw oaw þ oah  oaw oa
UT ¼ U0  2cla cos hc ðh þ wÞL   cos aw dUT h
4sin aw sin aw p¼ ¼       1  ð13Þ
dVl oVl oVl of of


 oaw þ oah  oaw oah
h2 ah hwaw ah
  cos ah þ cla ð10Þ
4 sin ah sin ah sin aw sin ah
where
where p/2  hc > aw  p/2  hc  b and p/2  hc >
2 

ah  0, the second term on the right represents the oUT w cos hc hwah sin aw  aw cos aw
¼ cla þ ;
energy contribution from solid–liquid interface, and the last oaw sin aw sin ah sin2 aw
term represents the energy from the liquid–air interface. ð14aÞ
The liquid volume is approximated as (Leu and Chang

2 

2004) oUT h cos hc hwaw sin ah  ah cos ah
¼ cla þ ;


oah sin ah sin aw sin2 ah
hw2 aw ð14bÞ
Vl ¼ hwL   cos aw
4 sin aw sin aw

 

2
h waw ah oVl hw2 h2 w ah
  cos ah ð11Þ ¼ þ  cos ah
4 sin ah sin aw sin ah oaw 2 sin aw 4 sin ah sin ah


sin aw  aw cos aw
It should be noted that the variation of meniscus profile  ; ð14cÞ
in this regime may be divided into two stages based on sin2 aw
the changes in aw and ah. In the first stage, both aw and

oVl h2 waw sin ah  ah cos ah
ah vary from a positive value to zero (p/2  hc > ¼ ð14dÞ
aw  0 and p/2  hc > ah  0), that is from a oah 2 sin ah sin aw sin2 ah
concave meniscus to a flat one. If the wedge angle is In the second stage, it is assumed that only aw continues to
large enough, the meniscus will develop into the second change from zero to negative (0 > aw  p/2  hc  b)
stage, resulting in a convex meniscus to reconcile the while ah remains unchanged holding at ah = 0. The change
wedge angles along the width direction. When applying in aw brings the ‘‘effective’’ contact angle heff = p/2  aw
Eq. 6 to obtain the pressure in this regime, we need an b in the width direction to approach the equilibrium contact
additional constraint imposing on the meniscus angle hc until the burst of the liquid into the expanded vol-
curvatures in the form ume. The pressure in this stage can be derived as

8
9
>
> w2 cos hc sin aw  aw cos aw >>


1 > c
< la sin aw >
=
1
oUT oVl sin2 aw hw2 sin aw  aw cos aw
p¼ ¼
 
oaw oaw a ¼0 >
> hwah sin aw  aw cos aw > > 2 sin aw sin2 aw
h >
: þ cla >
; ð15Þ
sin ah sin2 aw
2c h w  i
¼ la cos hc þ sin aw
w h

123
Microfluid Nanofluid (2008) 4:427–437 431

Note that the term ah/sin ah in the above expression is Here only the change in xcL is considered for the variations
taken to be unity as ah?0. Moreover, the pressure in this of UT and Vl while the curvatures aw and ah of the
stage turns negative as the meniscus curvature aw becomes meniscus are assumed to be unchanging (Man et al. 1998).
large enough in the negative sense, i.e. sin(aw) > The pressure derived in this regime is more complex
(w/h)cos hc. depending on the geometry of the divergent section as well
In the third regime (xc > L), the liquid expands into the as the wetted length in this section.
divergent section of the channel. The total free energy in The pressure of the liquid in a 3D microchannel with
the expansion regime is expressed as sudden expansion as depicted in Fig. 2 can be computed
from Eqs. 9, 13, 15 and 18 for the corresponding regimes.
UT ¼ U0  2cla cos hc Figure 3 illustrates the pressure of the liquid as a function
2
3
h of the liquid volume for a square channel with a wedge
6 L ð h þ w Þ þ x cL þ w þ x cL tan b 7 angle in the width direction raging from b = 0–100. Note
6 cos b 7
6 7 that the constraint of awah = 0 is used in Eq. 13 to
6
7
6 ðw þ 2xcL tan bÞ 2
a 7 compute the pressure. In the straight channel regime, the
66
w
 cos aw 7 7
6 4 sin aw sin aw 7 ð16Þ constant positive pressure of 0.360 kPa drives the wicking
6 7 until the meniscus reaches the valve edge entering the
6
 7
4 h2 ah 5 transition regime. In the transition regime, the pressure
  cos ah
4 sin ah sin ah falls rapidly from the positive constant value. For small b
  the pressure due to interfacial tension is still in the positive
haw ah
þ cla ðw þ 2xcL tan bÞ range. This means that the capillary force propels the liquid
sin aw sin ah
forward across the valve, which is a ‘‘valveless’’ condition.
where xcL = xc  L denotes the horizontal position of the As b  50, the pressure tends to further decrease to reach
liquid–solid–air interface in the divergent section. The a negative and then rises gradually. In this situation, the
liquid volume is approximated as flow is blocked by the valve and its magnitude of negative
pressure (pressure barrier) varies with the geometry of the
Vl ¼ h wL þ wxcL þ x2cL tan b
capillary channel. The critical wedge angle at which the

#
ðw þ 2xcL tan bÞ2 aw liquid pressure turns negative can be evaluated from Eq.
  cos aw 15. Notably, the critical wedge angle for the 3D channel
4 sin aw sin aw
presented here is greater than the value of (p/2  hc)

 
h2 ah aw evaluated from the 2D model (Man et al. 1998) because the
  cos ah ðw þ 2xcL tan bÞ latter neglects the sidewall effects. The maximum magni-
4 sin ah sin ah sin aw
ð17Þ tude of the pressure barrier is referred to as the burst
pressure. It can be seen from Fig. 3 that a larger wedge
The liquid pressure can be obtained again from Eq. 6 as angle results in a larger burst pressure. The burst pressures



oUT oVl 1
p¼
oxcL oxcL
8 

 9
>
> 2cla 1 w 2xcL tan b w 2xcL aw >
>
< cos hc þ þ tan b  þ tan b  cos aw > >
=
w cos b h h sin aw h h sin aw
¼ ð18Þ
>
> 2c a a tan b >
>
>
: þ la w h >
;
w sin aw sin ah



1
2xcL tan b 2xcL aw h aw tan b ah
 1þ tan b  1þ tan b  cos aw   cos ah
w sin aw w sin aw 2w sin aw sin ah sin ah

123
432 Microfluid Nanofluid (2008) 4:427–437

are 0.265 and 0.290 kPa for b = 90 and 100, respec- The burst pressure may also be evaluated from Eq. 18 in
tively. Moreover, the present 3D model can be extended to the expansion regime by taking xcL = 0 and ah = 0:
take account of the non-uniform interface properties 2 3
resulted from the use of different materials for channel 2cla 4 w cos hc  asin
w sin b

structures as those reported by Melin et al. (2004) and Chen Dpb ¼  cos hc þ  aw
5
w h  cos b þ sin b aw  cos aw
et al. (2006). sin aw sin aw

ð20Þ

2.2 Burst pressure As the aspect ratio becomes infinite (w/h = 0), Eq. 18 will
reduce to the 2D result of Man et al. (1998). Note that there
Design of capillary-burst valve is based on the burst should be a ‘‘’’ sign placed in front the term cosb in the
pressure that depends on the valve geometry and channel– denominator of their Eq. 13. Also note that Eq. 20 can be
liquid–air interfacial properties. The burst pressure for the further reduced to a simpler form exactly identical to Eq.
3D model can be obtained either from the pressure 19 by substituting (p/2  hc  b) for aw.
expressions given in the transition or expansion regimes. Moreover, the burst pressure of Eq. 19 is essentially
When the liquid pressure turns negative in the transition identical to the result derived by Cho et al. (2007). The
regime, the meniscus will continue to change shape only by unique difference is the 3D structure analysis provided in
external forcing. As the external forcing is large enough, it the present study. In their result, the equilibrium contact
propels the liquid to overcome the pressure barrier of the angle hc is replaced by the advancing contact angle hA and
valve, causing the liquid to burst into the divergent section. the term (hc + b) in Eq. 19 is modified by substituting an
The burst pressure can be computed from Eq. 15 by taking angle defined as min{hA + b, 180} because 180 is the
aw = p/2  hc  b, at which aw reaches its negative maximum contact angle that a liquid can attain. Based on
maximum in the transition regime. The maximum magni- the modified theory for burst pressure Dpmb,
tude of pressure barrier then yields a very simple form as 2cla h w i
Dpmb ¼  cos hA  cosðmin fhA þ b; 180 gÞ
2cla h w i w h
Dpb ¼  cos hc  cosðhc þ bÞ ð19Þ ð21Þ
w h
where the first term in the bracket on the right represents Cho et al. (2007) show that the capillary valve can also
the liquid pressure contributed by the unchanged depth as work for a hydrophobic channel. When involving a liquid
in Eq. 9, and the second term represents the pressure plug in the channel, the receding contact angle hR at the
required to wet the expanded volume beyond the edge as rear liquid front may have to be considered in Eq. 21.
the term b in Eq. 3 proposed by Duffy et al. (1999).
1

1D Theory

0.8
Burst pressure (kPa)

0.6
2D Theory

0.4
3D Theory

0.2

0
0 1 2 3 4 5
Aspect ratio

Fig. 4 Comparison of calculated burst pressures based on the 1D, 2D

Fig. 3 Variation of liquid pressure with liquid volume for a square and 3D models for rectangular channels of various aspect ratios with
channel (w = h = 300 lm) with wedge angles raging from b = 0– fixed width w = 300 lm and wedge angle b = 90. Interfacial
100. Interfacial properties cla = 0.072 N/m and hc = 68 were used properties cla = 0.072 N/m and hc = 68 were used for the
for the calculations calculations

123
Microfluid Nanofluid (2008) 4:427–437 433

Figure 4 compares burst pressures calculated from the tested in the present study was within ±0.2%. The angular
1D, 2D and 3D models for a rectangular channel of various position of the microchannel on the disk was acquired and
aspect ratios. In the figure, the 1D results are obtained from controlled through an optical encoder (10,000 pulse/rev).
Eq. 2, the 2D results are computed from the theory derived The signals from the encoder were used to synchronize the
by Man et al. (1998) for the channels with a fixed width, image-capturing unit. The image unit consisted of a frame
and the 3D results are given by Eq. 19. The 1D theory that grabber (Metero-II/CL, Matrox) and a CCD camera
assumes sudden expansions in both of the width and depth (CV-M71CL, 768 · 576 pixels, JAI) in conjunction with a
directions as in the axisymmetric case predicts a rather microscope (2· or 10· objective, Mitsutoyo) above the
higher burst pressure for small aspect ratios (h/w << 1). disk to obtain flow images. The 2· microscope lens has a
The predicted value is then followed by a rapid decrease depth of focus of 91 lm and a resolution of 5 lm. The 10·
with increasing aspect ratio becoming closer to the 2D lens has a depth of focus of 3.5 lm and a resolution of
theory. At an infinitely large aspect ratio, Dh = 2w and the 1 lm. A halogen lamp (MHF-M1001, Moritex) served for
1D theory yields Dpb = 2clasinhc/w, which is exactly the illumination beneath the disk. In synchronization with the
value of 0.445 kPa calculated from the 2D theory. The rotational motion, the CCD camera having a maximum
present theory based on the 3D model obviously gives a frame rate of 60 fr/s can be triggered to allow one shot of
more reasonable prediction. The capillary valve becomes the targeted object on the rotating disk per revolution.
effective at an aspect ratio of h/w = 0.4 and then the burst The microfluidic structures were machined using a
pressure increases with the aspect ratio until it approaches micro-CNC in a PMMA (polymethylmethacrylate) disk of
the 2D value at a large aspect ratio as one can expect from 10 cm in diameter and 2 mm thick. The microstructured
Eq. 20. Note that the burst pressure can be converted into disk was then bonded with another PMMA disk on a hot
the corresponding burst rotational speed for capillary-burst plate at a temperature of 63C and a pressure of 65–70 kgf/
valves fabricated on a rotating disk. cm2 for 20 min. All of the centrifugal-based microfluidic
disks were machined having a channel width of 300 or
400 lm with various depths (80–600 lm) and wedge
3 Experimental setup angles (30–100). The fluid used in the experiments was
de-ionized water. For clear visualization, the de-ionized
The experiments were carried out using the image-captur- water was mixed with a small amount of red ink. The
ing unit in synchronization with the rotational motion. contact angle between the fluid and the PMMA surface was
Figure 5 illustrates the schematic arrangement of the cou- measured using a dynamic contact angle system (FTÅ200,
pled optic-rotation apparatus. The centrifugal pumping was First Ten Angstroms) at 68 ± 1.
achieved by employing a DC motor (AS46AA, Oriental
Motor) to drive the disk in which the microfluidics as
depicted in Fig. 1 was microstructured. The reservoir of 4 Results and discussion
8 mm in diameter is located at r1 = 27 mm and the capil-
lary valve is at r2 = 31.5 mm for the present experiments. The synchronized image-capturing system developed in the
The accuracy of the rotational speeds (100–1,000 rpm) present study enables us to visualize the liquid flow that
bursts out at the capillary valve to fill the expanded volume
as the rotational speed exceeds the threshold value. Fig-
ure 6 shows flow visualization images of the liquid blocked
by the capillary valve with b = 60 and channel cross-
section of w = h = 300 lm on a stationary disk. The
meniscus, which can be more clearly seen in the picture
taken with the 10· objective, is nearly flat at the valve edge
(xc = L) for the stationary case. As the disk rotates at a low
speed X = 240 rpm, it can be seen in Fig. 7 that the liquid
still stops at the valve edge with its meniscus curving
outward to balance the centrifugal force. Figure 8 displays
sequential images of the liquid flow that bursts into the
valve as the speed is increased to 300 rpm higher than the
predicted burst value of 260 rpm. Note that the black line
on the meniscus contour is conjectured a result of inclined
Fig. 5 Experimental arrangement for flow visualization of micro- and curved meniscus surface across the channel depth. The
fluidics on a rotating disk sequential images show that the liquid enters the expanded

123
434 Microfluid Nanofluid (2008) 4:427–437

Fig. 6 Liquid front stops at the sudden opening of the capillary valve
(w = h = 300 lm and b = 60) on a resting disk. The left picture (a)
was taken using ·2 objective (resolution 5 lm) and the right one (b)
using ·10 objective (resolution 1 lm) focusing at the meniscus

volume at a slow flow rate taking 3.2 s for the meniscus to

advance to halfway in the divergent section. Then, a quick
flooding (t = 3.2–3.4 s) occurs and the liquid fills the
whole valve. The liquid gradually occupies the divergent
section with a liquid front of convex shape but having a
hydrophilic contact angle (less than 90) with the side
walls. The ‘‘effective’’ contact angles with the side walls
may be estimated from the images that are applicable. It is
found that the ‘‘effective’’ contact angle varies from
60 ± 2 at t = 0.2–0.4 s, to 65 ± 3 at t = 0.6–1.6 s, then Fig. 8 Sequential images of the liquid flow burst into the capillary
70 ± 3 at t = 1.8–3.0 s, and finally rises rapidly to valve (w = h = 300 lm and b = 60) at a rotational speed of 300 rpm
for the time t = 0.2–3.4 s
86 ± 2 at t = 3.2 s. During the early development

(t = 0.6–3.0 s) of the meniscus in the expansion regime,

the measured ‘‘effective’’ contact angle is in accordance
with the assumption (heff = hc = p/2  aw  b) made to
derive the burst pressure of Eqs. 19, and 20. As the
meniscus reaches halfway in the divergent section at
t = 3.2 s, where the quick flooding occurs and momentum
effects become significant, the dynamic contact angle may
have to be considered in the analytical modeling. More-
over, the meniscus movement during t = 2.0–3.2 s in
Fig. 8 appears to retain its shape, which is consistent with
the derivation leading to Eq. 18 where only the change in
xcL (the marching distance along the channel axis in the
divergent section) is considered for the variation.
The burst conditions for various capillary channel
dimensions were carefully measured by comparing the
rotational speed and the microscopic image of the flow. For
each of the data points presented, the burst rotational speed
was repeatedly examined at least three times. The uncer-
tainty of the burst rotational speeds presented in the fol-
Fig. 7 Liquid front stops at the sudden opening of the capillary valve
(w = h = 300 lm and b = 60) on a rotating disk at a speed of lowing figures is approximately within ±4%. Figure 9
240 rpm compares the measured burst rotational speeds with the 3D

123
Microfluid Nanofluid (2008) 4:427–437 435

500 450

450 400
w = 300 m
Burst rotational speed (rpm)

Burst rotationa l speed (rpm)

400
350 w = 300 µm

350
300
w = 400 m
300
250
250 w = 400 µm

200
200

150 150

100 100
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 50 60 70 80 90 100
Aspect ratio β (deg)
Fig. 9 Comparison of the predicted burst rotational speeds from the Fig. 10 Comparison of the predicted burst rotational speeds from the
3D theory (cla = 0.072 N/m, hc = 68, solid line for w = 300 lm and 3D theory (cla = 0.072 N/m, hc = 68, solid line for w = 300 lm and
dashed line for w = 400 lm) with the measurements for the channels dashed line for w = 400 lm) with the measurements for the square
of w = 300 lm (circle) and 400 lm (triangle) with various aspect channels of w = h = 300 lm (circle) and 400 lm (triangle) with
ratios (h/w = 0.42–2.0). The uncertainty of the measured burst various wedge angles (b = 50–100). The uncertainty of the
rotational speeds is approximately ±4% measured burst rotational speeds is approximately ±4%

theory of Eq. 19 for capillary valves with channel cross- and increased for a hydrophobic surface. In other words,
sections of various aspect ratios while the wedge angle is roughness makes hydrophilic surface more hydrophilic and
fixed at b = 90. Both the theory and experiment show that hydrophobic surface more hydrophobic. When examining
the burst rotational speed is higher for the valve with the effect of roughness on Eq. 19, one discovers that
smaller channel width and larger aspect ratio. The 3D roughness tends to cause a decrease in burst pressure.
theory agrees well with the measurements, in which even Quantitatively, a decrease of 3 in hc will result in a
the largest deviations are found to be less than 6%. Fig- decrease of 6.4% in the burst rotational speed for a square
ure 10 shows a similar comparison for the variation with channel with b = 90. However, the dependence of burst
wedge angles (b = 50–100) for square channels. Note pressure on contact angle could be much more complicated
that the wedge angle was purposely widened over 90 to than such a simple argument. Considering the meniscus is
examine the validity of the theory. The capillary valve in motion, Cho et al. (2007) suggest the use of advancing
becomes effective at b = 50 as predicted by the theory. contact angle hA instead of the equilibrium contact angle hc
The agreement on burst rotational speed between the the- for computing the burst pressure from Eq. 19. As a result,
ory and experiment is good particularly for the valves with the burst pressure computed using hA is greater than that
a smaller channel width (w = 300 lm) including the case using hc because hA > hc. Therefore, the surface roughness
of b = 100. For the valves with a larger channel width and the associated contact angle could have a mixed impact
(w = 400 lm), the measured speeds are generally lower on the burst pressure.
than the prediction yielding a deviation in the range of 6– The effect of momentum on the burst pressure was
10% for b = 80–100. This may be partly due to uncer- previously addressed by Zeng et al. (2000) in terms of a
tainties on geometry and surface quality and partly due to fluidic capacitance that takes into account of the energy
the fact that the liquid momentum is considered negligible loss due to meniscus oscillation and viscous dissipation.
throughout the entire derivation to obtain the pressure. Alternatively, the momentum effect may be evaluated
The roughness of the machined channel surface was based on the average velocity of the liquid flow in the
measured using a stylus profilometry (SE3500K, Kosaka channel. For the channel with w = h = 400 lm and
Lab) at 1.4–2.4 lm Ra. How the surface roughness affects b = 80, which yields the worst deviation of 10%, an
the burst pressure may be understood from its effect on average velocity uo = 0.13 m/s was estimated for
contact angle as were discussed by a number of studies X = 300 rpm from the time needed to empty the reservoir.
(Wenzel 1936; Lambert et al. 2003; Katoh 2004). These The velocity in this case could contribute as much as 17 N/
studies generally suggest that the contact angle of a smooth m2 in momentum difference (qu2o) across the capillary
surface is decreased by roughness for a hydrophilic surface valve corresponding to a deviation of 5.5% below the

123
436 Microfluid Nanofluid (2008) 4:427–437

predicted burst rotational speed (310 rpm). It should be apparatus was developed for visualization and measure-
noted that the average velocity of laminar flow in a cen- ments of the liquid flow in the CD-based microfluidics
trifugal-based channel is approximated to be proportional equipped with a capillary valve. The microfluidics was
to the cross-sectional area and to the square of rotational machined in a PMMA disk and sandwiched with another
speed (Ducrée et al. 2006). On the other hand, the burst PMMA disk. The visualization images clearly show the
pressure (or the square of the burst rotational speed) pre- development of the meniscus shape during the burst of the
dicted from Eq. 19 appears to be inversely proportional to liquid from the capillary channel into the expanded vol-
the channel dimension. Accordingly, the momentum effect ume. The meniscus development observed in the experi-
on the burst pressure is expected to diminish rapidly as the ments also supports the assumptions made in the present
channel dimensions become smaller. Further studies need analysis for the derivation of pressure barrier in a capillary
to clarify the actual role of momentum in a capillary valve valve. Both the theory and the experiment consistently
as well as the details of the effects due to surface quality show that the burst rotational speed is higher for the valve
and the associated contact angle variation. with a smaller channel width, larger aspect (depth-to-
In the present experiments, the measurements of burst width) ratio and wider wedge angle. The 3D theory is in
pressure were carried out for channels having widths of good agreement with the measurements of burst rotational
300 and 400 lm. In the range of smaller dimensions, Cho speed particularly for the valves with a smaller channel
et al. (2007) have made similar measurements for centrif- width (w = 300 lm). For the valves with a channel of
ugal microchannels of widths w = 15–150 lm with fixed larger square cross-section (w = h = 400 lm), the mea-
h = 150 lm and found reasonable agreement between the sured rotational speeds are found to be significantly lower
measured burst rotational speeds and the modified theory than the predictions by nearly 10% for those with larger
of Eq. 21, which considers the advancing and receding wedge angles (b = 80–100) which require relatively
contact angles for the liquid plug in the experiments of Cho higher burst values (310–350 rpm) than those with smaller
et al. (2007). However, for very narrow channels wedge angles. This discrepancy may be attributed to the
(w < 40 lm) the measured data are noticeably lower than surface quality and the associated contact angle variation as
the theoretical results, indicating the increasing sensitivity well as the neglect of liquid momentum in the derivation
of the burst rotational speed to the channel geometry of leading to the present 3D theory. Nevertheless, the present
small sizes. In general, the measured burst rotational 3D theory is shown to be more accurate than previous 1D
speeds in the smaller dimension range (w = 36–150 lm) and 2D theories in evaluating the burst pressure for capil-
scale as *(1/w)m with m & 0.4, slightly less than the one- lary valves. The simple expression of burst pressure
half power approximated by the present 3D theory. derived here provides a convenient tool for design of
capillary valves.
5 Conclusions Acknowledgments The authors are grateful for the financial sup-
port for this study from National Science Council of Taiwan under
An analytical study has been carried out to investigate the Contract Number NSC 94-2212-E-005-017.
pressure barrier in a capillary-burst valve with a sudden
expansion in channel cross-section. The analysis of the
References
internal liquid pressure is based on the surface energy for
3D meniscus shape in a rectangular channel where the Auroux P-A, Iossifidis D, Reyes DR, Manz A (2002) Micro total
sudden expansion is made with a wedge angle in the width analysis systems. 2. Analytical standard operations and applica-
direction only. An analytical expression of simple form has tions. Anal Chem 74:2637–2652
been found to predict the burst pressure required for a Chen JM, Lin M-G, Huang P-C (2006) Design of capillary valves for
microfluidics on a rotating disk. In: Proceedings of 2nd
capillary valve. This simple expression relates the burst international symposium on micro and nano technology, ISM-
pressure to the channel dimensions (w and h) and the valve NT-2, Hsinchu, pp 100–103
geometry (b) as well as the surface property (cla and hc). Cho H, Kim H-Y, Kang JY, Kim TS (2007) How the capillary
The burst pressure derived from the 3D model will reduce microvalve works. J Colloid Interface Sci 306:379–385
Ducrée J, Haeberle S, Brenner T, Glatzel T, Zengerle R (2006)
to the 2D analytical expression of Man et al. (1998) when Patterning of flow and mixing in rotating radial microchannels.
assuming a channel depth much greater than its width. Our Microfluid Nanofluid 2:97–105
expression of burst pressure is also essentially identical to Duffy DC, Gillis HL, Lin J, Sheppard NF Jr, Kellogg GJ (1999)
the result derived by Cho et al. (2007). However, the Microfabricated centrifugal microfluidic systems: characteriza-
tion and multiple enzymatic assays. Anal Chem 71:4669–4678
present study provides a more thorough 3D structure Glière A, Delattre C (2006) Modeling and fabrication of capillary stop
analysis. Capillary valves are commonly used to regulate valves for plannar microfluidic systems. Sens Actuators A 130–
liquid flow in centrifugal microfluidics. Experimental 131:601–608

123
Microfluid Nanofluid (2008) 4:427–437 437

Grumann M, Geipel A, Riegger L, Zengerle R, Ducrée J (2005) Madou MJ, Lu Y, Lai S, Koh CG, Lee LJ, Wenner BR (2001b) A
Batch-mode mixing on centrifugal microfluidic platforms. Lab novel design on a CD disc for 2-point calibration measurement.
Chip 5:560–565 Sens Actuators A 91:301–306
Gustafsson M, Hirschberg D, Palmberg C, Jörnvall H, Bergman T Man PF, Mastrangelo CH, Burns MA, Burke DT (1998) Microfab-
(2004) Integrated sample preparation and MALDI mass spec- ricated capillary-driven stop valve and sample injector. In:
trometry on a microfluidic compact disk. Anal Chem 76:345– Proceedings of 11th annual international workshop on micro
350 electro mechanical systems, Heidelberg, pp 45–50
Haeberle S, Brenner T, Zengerle R, Ducrée J (2006) Centrifugal Manz A, Grabber N, Widmer HM (1990) Miniaturized total chemical
extraction of plasma from whole blood on a rotating disk. Lab analysis systems: a novel concept for chemical sensing. Sens
Chip 6:776–781 Actuators B 1:244–248
Katoh K (2004) Measurement of contact angle and surface tension. McNeely MR, Spute MK, Tusneem NA, Oliphant AR (1999)
In: Hartland S (ed) Surface and interfacial tension: measurement, Hydrophobic microfluidics. In: Ahn CH, Frazier AB (eds)
theory and applications. Dekker, New York, pp 375–546 Proceedings of SPIE—microfluidic devices & systems II, vol.
Kim DS, Lee K-C, Kwon TH, Lee SS (2002) Micro-channel filling 3877, pp 210–220
flow considering surface tension effect. J Micromech Microeng Melin J, Roxhed N, Gimenez G, Griss P, van der Wijngaart W,
12:236–246 Stemme G (2004) A liquid-triggered liquid microvalve for on-
Kim E, Whitesides GM (1997) Imbibition and flow of wetting liquids chip flow control. Sens Actuators B 100:463–468
in noncircular capillary. J Phys Chem B 101:855–863 Nguyen N-T, Huang X (2005) Thermocapillary effect of a liquid plug
Lambert P, Pierre Letier P, Delchambre A (2003) Capillary and in transient temperature fields. Jpn J Appl Phys 44:1139–1142
surface tension forces in the manipulation of small parts. In: Nguyen N-T, Wereley ST (2002) Fundamentals and applications of
Proceedings of the 5th IEEE international symposium on microfluidics. Artech House, Boston
assembly and tasks planning (ISATP), Besançon, pp 54–59 Oh KW, Ahn CH (2006) Topical review: a review of microvalves. J
Lee J, Moon H, Fowler J, Schoellhammer T, Kim C-J (2002) Micromech Microeng 16:R13–R39
Electrowetting and electrowetting-on-dielectric for microscale Reyes DR, Iossifidis D, Auroux P-A, Manz A (2002) Micro total
liquid handling. Sens Actuators A 95:259–268 analysis systems 1. Introduction, theory, and technology. Anal
Leu T-S, Chang P-Y (2004) Pressure barrier of capillary stop valves Chem 74:2623–2636
in micro sample separators. Sens Actuators A 115:508–515 Wenzel RN (1936) Resistance of solid surfaces to wetting by water.
Madou M, Zoval J, Jia G, Kido H, Kim J, Kim N (2006) Lab on a CD. Indust Eng Chem 28:988–994
Annu Rev Biomed Eng 8:601–628 Zeng J, Deshpande M, Greiner KB, Gilbert JR (2000) Fluidic
Madou MJ, Lee LJ, Daunert S, Lai S, Shih C-H (2001a) Design and capacitance model of capillary-driven stop valve. In: MEMS
fabrication of CD-like microfluidic platforms for diagnostics: proceedings of ASME international mechanical engineering
microfluidic functions. Biomed Microdevices 3:245–254 congress and exposition, Orlando

123