Noname manuscript No. (will be inserted by the editor) Wicking through a confined micropillar array Baptiste Darbois Texier · Philippe Laurent · Serguei Stoukatch · St´ephane Dorbolo 6 1 0 2 Received:date/Accepted:date b e F Abstract ThisstudyconsidersthespreadingofaNew- etal(2007)studiedexperimentallyhowawettingliquid 5 tonian and perfectly wetting liquid in a square array of propagates along solid surfaces decorated with a forest 2 cylindric micropillars confined between two plates. We of micropillars. This problem leads to numerous exper- ] show experimentally that the dynamics of the contact imentalderivations(Kimetal(2011);Maietal(2012); n line follows a Washburn-like law which depends on the Spruijtetal(2015);XiaoandWang(2011)),numerical y characteristics of the micropillar array (height, diam- contributions (Semprebon et al (2014)) and theoretical d - eter and pitch). The presence of pillars can either en- modelizations (Hale et al (2014a,b)). In the case of mi- u hanced or slow down the motion of the contact line. A crochannels,studieshaveinspectedtheeffectofsurface l f theoretical model based on capillary and viscous forces patterned by posts or pillars on the dynamics of liquid . s has been developed in order to rationalize our observa- insidethemicrochannel(Gamratetal(2008);Mognetti c i tions. Finally, the impact of pillars on the volumic flow andYeomans(2009);MohammadiandFloryan(2013)). s rate of liquid which is pumped in the microchannel is Besides, the liquid spreading into a confined micropil- y h inspected. lararrayisinvolvedinvariousapplicationssuchaslab- p on-a-chip chromatography (Op de Beeck et al (2012)), [ Keywords Microchannel wicking · micropillar array · isolation of tumor cells (Nagrath et al (2007)) or flow liquid impregnation 2 regulation in microfluidic (Vangelooven et al (2010)). v If the problem of liquid impregnation in a micropillar 6 array with a free surface has been extensively studied 1 Introduction 9 Bocquet and Barrat (2007), the effect of confinement 7 on this phenomenon is still an open question that we 0 Wicking can be defined as the spreading of a liquid address in this article. . in the tiny structures of a porous media due to capil- 1 0 lary forces. The pioneer work concerning the spreading 6 of a viscous liquid in a capillary tube is attributed to 1 Washburn(1921).Thewickingphenomenonhasalarge : v spectra of applications such as textile science (Kissa i (1996))orheatpipedesign(TienandSun(1971)).The X Inthisworkweconsiderthecaseofasquarearrayof problemwasrevisitedbecauseofmicroelectronicscom- r cylindric micropillars confined between two plates. We a ponents bonding (Schwiebert and Leong (1996)). One study experimentally the spreading of a perfectly wet- way to model the wicking process is to consider the ting liquid as a function of the pillars characteristics spreading of a liquid into a micropillar array. Ishino (height, pitch, diameter). The measurements are com- paredwithatheoreticalmodelbasedoncapillaryforces B.DarboisTexier ·S.Dorbolo Grasp,PhysicsDept.,ULg,Li`ege,Belgium andviscousresistance.Finally,theconsequencesofthis E-mail:[email protected] work for capillary pumping is discussed. In addition, P. Laurent·S.Stoukatch more complex micropillars arrangements are examined Microsys,MontefioreInstitute,ULg,Li`ege,Belgium such as non-square and non-uniform lattices. 2 BaptisteDarboisTexier, PhilippeLaurent,SergueiStoukatch,St´ephane Dorbolo (a) Fig.1 Sketchoftheexperimentalsetup.Thechannelheight is h, its width is (cid:96). The channel contains an array of pillars ofdiameterdspacedbyapitchp.Theheightofthepillarsis thesamethattheoneofthechannel.Thedynamicviscosity of the fluid which spreads in the channel is denoted η. The meanpositionofthecontactlineis x. (b) 1 Experiments Fig.2 Superimpositionoftop-viewpicturesofliquidimpreg- nationofmicrochannelsovertime.Thetimestepbetweentwo frames is 1 s. The fluid is Newtonian silicone oil V100. (a) Microchannels were obtained by negative SU-8 pho- Empty microchannel of height h=80µm. (b) Microchannel tolithography made on a silicon wafer (Del Campo and ofheighth=80µmfilledwithpillarsofdiameterd=400µm Greiner (2007)) prior to a PDMS molding (Folch et al separatedbyapitch p=800µm. (1999)).ThePDMScavitywasbondedonasubstrateof thesamematerialbyoxygenplasmaexposure.Byvary- 16 ing the design of a the mask used in the photolithogra- phy,wechangedthegeometryofthemicropillarsarray 14 inside the channel [Fig. 1]. The channel height h could 12 be changed between 50µm and 100µm, the pillar di- )10 ameter d from 80µm to 1000µm and the lattice pitch m m 8 pbetween150µmand1000µm.Thus,thepillaraspect ( 250 ratio h/d may vary between 0.05 and 1.2 and the ra- x 6 200 2m)150 tio p/d between 1.5 and 10. The transverse dimension 4 x2(m100 of the channel (cid:96) was chosen in order to keep the ratio 50 2 (cid:96)/h larger than 100. These PDMS microchannels were 00 10 20 30 40 50 0 t(s) put into contact with a liquid reservoir containing a 0 10 20 30 40 50 Newtonian silicone oil V100 of density ρ = 980kg/m3, t(s) dynamic viscosity η = 100mPa·s and surface tension Fig. 3 Mean position of the contact line x as a function of γ (cid:39) 23±0.3mN/m at a temperature T = 20◦C. The time t for a silicone oil V100 (ρ=980kg/m3, γ =23mN/m and η = 100mPa·s) which spreads in two different mi- mean position of the contact line was recorded from crochannels.Eachchannelhasthesameheighth=80µmbut above and reported in Fig. 2. The analysis of these ex- a different density of pillars. The blue dashed line shows the periments provides the position of the contact line x case of an empty microchannel and the red solid line stands over time as shown in Fig. 3. foramicrochannelwithpillarsofdiameterd=400µmsepa- rated by a pitch p = 800µm. The inset shows the square of Oneobservesthat,whatevertheconsideredmicrochan- the mean contact line position x2 as a function of time for nel,themeanpositionofthecontactlinepropagatesac- thetwopreviousexperiments. cording to the square root of time. This dependency is underlined by the linearity of x2 with time as shown in the inset of Fig. 3 and corresponds to a Washburn-like Thereafter,theeffectofthemicropillararrayonthe law expected for the spreading of a viscous fluid in- spreading dynamics is quantified. We measure the pre- √ duced by capillary forces (Washburn (1921)). Besides, factorDofthesquarerootlawx= Dtformicrochan- one notices the influence of the pillars array on the dy- nels as a function of the lattice properties (pillar diam- namicsofthecontactlinebyachangeofthepre-factor eter d, height h and pitch p). This pre-factor defines ofthesquarerootlaw.Intheconsideredcase,thepres- the diffusivity of the liquid in the micro-channel and is ence of pillars in the channel slows down the liquid dy- measured by fitting the slope of x2(t) by the way of namics. Additionally, the introduction of pillars in the a least mean squares method. The analogous diffusiv- microchannel induces a stick-slip behavior in the dy- ity of the liquid in the empty microchannel is denotes namics of the contact line which appears by the way of D . Symbols in Fig. 4a and 4b show the evolution of 0 steps in the red curves in Fig. 3. the experimental ratio D/D as a function of the pillar 0 Wickingthroughaconfinedmicropillararray 3 density φ=πd2/4p2 where d and h are kept constants ence of pillars enhances the spreading of the liquid in andonlypisvaried.Twosetsofexperimentshavebeen the channel. done for a different pillar aspect ratio of h/d=0.2 and h/d=1 and are plotted with dots. 2 Model (a) 2.1 Empty cavity 1 Thegoalofthissectionistodevelopatheoreticalmodel in order to rationalize the experimental observations. 0.8 First, we evaluate the dimensionless numbers associ- ated to previous experiments in order to determine the 0.6 DD0 mainforcesinpresence.Accordingtoexperimentaldata 0.4 presented in Section 1, the spreading velocity reaches x˙ ∼1mm/s in less than 20 ms. This allows to estimate 0.2 theReynoldsnumberassociatedtothefluidflowatthis particulartime:Re=ρhx˙/η ∼103×10−4×10−3/0.1∼ 00 0.2 0.4 0.6 10−3.Therefore,theliquidflowisdominatedbyviscous φ frictionandnoinertialtermswillbetakenintoaccount (b) in this study. The experimental Bond number, which compares the relative effect of gravity and surface ten- 1 sion, is Bo = ρgh2/γ ∼ 103×10×10−8/10−2 ∼ 10−2. Also, we estimate the capillary number Ca = ηU/γ (cid:39) 0.8 0.1×10−3/2×10−2 (cid:39) 5×10−3. The small value of the capillary number compared to one signifies that 0.6 DD0 thecontactangleremainsclosetoitsequilibriumvalue 0.4 (Thompson and Robbins (1989)). Finally, the experi- mental data presented in Section 1 will be approached 0.2 byamodelwhichbalancescapillaryandviscousforces. In the case where the microchannel is empty, the 0 0 0.2 0.4 0.6 surface wetted by the fluid at a positon x is equal to φ S =2xl (neglectingthechannelheighthwhichismore Fig. 4 Diffusivity ratio D/D0 as a function of the pillars than 100 times smaller than its width (cid:96)). The resulting density φ = πd2/4p2 for two different pillar aspect ratios capillaryforceinthexdirectionisf =γ∂S/∂x=2γl. h/d. Dots correspond to the flow of silicone oil V100 (ρ = γ 980kg/m3, γ = 23mN/m and η = 100mPa·s) into micro- The non-slipping boundary conditions of the fluid on channels of height h = 80µm with pillars of diameter d = the two microchannel plates impose a Poiseuille flow 400µm (h/d=0.2) in figure (a) and pillars of diameter d= which can be expressed as: v = v (cid:0)1−4z2/h2(cid:1) [Fig. x 0 80µm (h/d = 1) in figure (b). Solid lines correspond to the 5a]. The flow conservation imposes v = 3x˙/2. The theoretical model developed further in Section 2 and show 0 viscous force along the x-direction is equal to f = thesolutionofequation(12)fordifferentpillarsaspectratio x (h/d=0.2fortheredsolidlineinfigure(a)andh/d=1for η(cid:82)x(cid:48)=x(cid:82)y(cid:48)=l/2 (cid:82)z(cid:48)=h/2 ∆v dx(cid:48)dy(cid:48)dz(cid:48) which provides x(cid:48)=0 y(cid:48)=−l/2 z(cid:48)=−h/2 x thebluesolid lineinfigure(b)). f = −12ηxx˙l/h. Balancing the capillary force with x the viscous contribution provides As expected, when the pillar density tends to zero, γh x2(t)= t (1) thefluiddiffusivityD recoverstheoneofanemptymi- 3η crochannel D . In the case for which the pillars aspect 0 ratio h/d = 0.2, the diffusivities ratio D/D decreases This approach predicts the square root evolution of 0 monotonically with the pillar density φ. The situation thecontactlinepositionwithtimeobservedexperimen- ismorecomplicatedforthecaseofh/d=1becausethe tally in Section 1 and provides a theoretical value of ratio D/D first increases up to pillar density of about the diffusivity D = γh/3η in the case of an empty 0 0 0.06andthendecreasestoreachvalueslowerthanunity microchannel. Note that, in practice this law is used for large pillar densities (φ > 0.15). One remarks that to estimate the flowing time of the underfill process in in the situation where h/d=1 and φ<0.15, the pres- microelectronics (Wan et al (2008)). 4 BaptisteDarboisTexier, PhilippeLaurent,SergueiStoukatch,St´ephane Dorbolo (a) betweenmicropillars.Wesupposethatthefluidvelocity profilein-betweenpillarsisv =v (cid:0)1−4y2/(p−d)2(cid:1) xp 0p (cid:0)1−4z2/h2(cid:1)whereasthefluidvelocityprofileelsewhere remains v = v (cid:0)1−4z2/h2(cid:1) [Fig. 5b]. The flow mass x 0 conservation yields v =9px˙/4(p−d) and v =3x˙/2. 0p 0 Finally, the viscous force resulting from the fluid flow between two pillars reads (b) (cid:18) (cid:19) p h p−d f =−12ηx˙d + (5) xp p−d p−d h The total viscous force in-between pillars when the contact line is located in x can be approached by (cid:18) (cid:19) d p h p−d f =−12ηx˙xl + (6) xptot p2 p−d p−d h Fig.5 (a)Sketchoftheliquidflowinanemptymicrochannel The previous equation yields viewed from the side. (b) Sketch of the assumed liquid flow in-betweenfourmicropillarsviewedfromabove. (cid:115) 2 12ηx˙xl h φ f =− + (7) xptot h (cid:113) (cid:16)(cid:113) (cid:17)2 φ 2.2 Effect of the pillar array φm φm −1 m φ φ When the pillars are present in the microchannel, the withφm =π/4(cid:39)0.78themaximalpillardensitywhen surface of the cavity wetted by the fluid reads p=d.Theviscouscontributionresultingfromthefluid flow out of inter-pillar areas is (cid:18)d(cid:19)2 x l x l S =2xl−2π +πdh (2) (cid:18) (cid:19) 12ηx˙xl d 2 pp pp f =− 1− (8) xtot h p where x/p is the mean number of pillar rows wetted by the fluid when its interface is located in x and l/p the This equation reduced to meannumberofwettedpillarrowsalongthetransverse (cid:32) (cid:115) (cid:33) 12ηx˙xl φ directionofthemicrochannel.Theglobalapproachused f =− 1− (9) hereisjustifiedbecauseweconsiderthemeandynamics xtot h φm of the contact line over a distance x much larger than Balancing the capillary and viscous forces, f + the characteristics of the pillar array (p and d). xptot f +f =0, gives The resulting capillary force is xtot γ (cid:34) (cid:35) f =2γl(cid:20)1− πd2 + hπd2(cid:21) (3) 12ηxx˙l 1+ h2 = γ 4p2 d 2p2 h (cid:113)φm (cid:16)(cid:113)φm−1(cid:17)2 (10) φ φ (cid:2) (cid:0) (cid:1) (cid:3) 2γl 1+ 2h−1 φ which reduces to (cid:2) (cid:0) (cid:1) (cid:3) This relation simplifies as f =2γl 1+ 2h−1 φ (4) γ (cid:0) (cid:1) with φ = πd2/4p2 the pillar density and h = h/d the x˙2 = γh 1+ 2h−1 φ (11) pillar aspect ratio. 3η 1+h2/(cid:113)φm (cid:16)(cid:113)φm −1(cid:17)2 Thepresenceofpillarsinsidethemicrochannelmod- φ φ ifies the fluid velocity profile inside the cavity and thus Inpresenceofpillars,theliquidstillfollowsaWashburn- √ theviscousforceexperiencedbythefluid.Variousstud- like law (x= Dt) but the diffusivity now depends on iesinspectedexperimentallyandtheoreticallythepres- thepillararraycharacteristics(φandh).Wedefinethe suredropresultingofaliquidflowthroughasquarear- ratio of the diffusivity D of the channel in presence of raysofcylindersembeddedinsideamicrochannel(Gunda pillars and the diffusivity D = γh/3η for an empty 0 et al (2013); Sadiq et al (1995); Tamayol and Bahrami channel. This ratio is expressed as (2009); Tamayol et al (2013)). However, these studies (cid:0) (cid:1) only consider the situation where the vertical confine- D 1+ 2h−1 φ = (12) ment is negligible. In this paper, we estimate the vis- D0 1+h2/(cid:113)φm (cid:16)(cid:113)φm −1(cid:17)2 cous force by assuming a Poisseuille flow in the space φ φ Wickingthroughaconfinedmicropillararray 5 2.3 Comparison with experiments 3 Discussion PredictionsofEq.(12)arecomparedwithexperimental 3.1 Fluid pumping data in Fig. 4 for different pillars density φ and vari- In our experiments, the capillary forces pump the fluid ousnormalizedpillarheighthbythewayofsolidlines. inside the microchannel. The liquid flow rate is defined Thegoodagreementbetweenexperimentsandtheoret- as Q = dV/dt with V the volume of fluid inside the ical predictions validates the assumptions made for the microchannel. When the contact line has the position profileoftheflowinthemodeldevelopedinsection2.2. x, the volume of fluid inside the cavity is V = lhx− Thereafter,weconsiderthepredictionsofthemodel √ π(cid:0)d(cid:1)2hl x. Substituting the relation x = Dt in the in a broader range of parameters compared to experi- 2 pp previous definition leads to ments. Figure 6 shows D/D as a function the pillar 0 density φ and the normalized height h as predicted by hl(1−φ)(cid:114)D Eq. (12). One notices that whatever the pillar aspect Q= (13) 2 t ratio, the diffusivity ratio D/D tends to one for small 0 The ratio of the flow rate Q in a cavity with pillars pillar densities (φ → 0) and falls to zero for large pil- and the flow rate Q in an empty cavity is lar densities (φ → φ ). If h > 0.5, the diffusivity ra- 0 m tio is slightly larger than one for small density values Q (cid:114) D (φ (cid:46) 0.15) as delimited by the white solid line in Fig. Q =(1−φ) D (14) 0 0 6. The analytical determination of the domain where Inserting equation (12) in the previous ratio yields D/D > 1 is presented in the Appendix A. The maxi- 0 mal value of the diffusivity ratio D/D0 increases with Q (1−φ)(cid:0)1+(cid:0)2h−1(cid:1)φ(cid:1)1/2 thenormalizedaspectratioh.Forh=3,D/D reaches = (15) a maximal value of about 1.1 for φ=0.06. Thi0s behav- Q0 (cid:18)1+h2/(cid:113)φm (cid:16)(cid:113)φm −1(cid:17)2(cid:19)1/2 ior can be understood by the fact that when the liquid φ φ wetsapillar,thegainofsurfaceenergyisγπdhwhereas iftherewasnopillarthegainofenergywouldhavebeen γπd2/2.Ifh>d/2,thepresenceofpillarsincreasesthe 3 1 energetic benefit of the liquid impregnation and fasten its dynamics. 2.5 0.8 2 0.6 3 1 /d1.5 h 2.5 0.4 0.8 1 2 0.2 0.5 0.6 d /1.5 h 0.4 0 0.2 0.4 0.6 1 φ Fig. 7 RatioQ/Q asafunctionofthepillardensityφand 0.2 0 0.5 thenormalizedheightofthechannelh=h/d. 0 0.2 0.4 0.6 Figure7presentstheflowrateratioQ/Q asafunc- φ 0 tion of the pillar density φ and the pillar aspect ratio Fig. 6 RatioD/D0 asafunctionofthepillardensityφand h. Whatever the ratio h/d, if the pillar density is small thenormalizedheightofpillarsh=h/daspredictedbyequa- (φ → 0), the flow rate ratio Q/Q tends to one. In tion(12).Thewhitesolidlineseparatestheregionwherethe 0 diffusivity ratio D/D is larger than one. The white dashed the opposite situation where the pillar density is large 0 line shows the pillar density φmax which maximizes the ra- (φ→φm),theratioQ/Q0fallstozero.Onenoticesthat tio D/D0 for a given aspect ratio h larger than 0.5. The the presence of pillars only slows down the flow rate Q whitedottedlineindicatesthepillaraspectratioh which max inacavitywithpillarscomparedtothecaseofanempty maximes the ratio D/D for a given pillar density. The the- 0 oretical predictions of these three lines are detailed in the cavity. Thus, the addition of micro-structures in a mi- AppendixA. crochannel does not enhance its pumping properties. 6 BaptisteDarboisTexier, PhilippeLaurent,SergueiStoukatch,St´ephane Dorbolo This conclusion is of first importance for the develop- tion pillars, presence of defect or random distribution ment of capillary pumps in the context of autonomous of pillars. microfluidic.Insuchasituation,thepresenceofmicro- structures in the capillary pump increases its efficiency by inducing large interface curvatures which produce 3.3 Model limitations a large overpressure and drive the fluid (Zimmermann We propose in this section to discuss the limitation of et al (2007)). According to our conclusions, this effect the model developed in Section 2. This latter is based works only if the hydraulic resistance of the system is on a basic assumption of the fluid velocity profile in- imposed by an element placed before the pump. In the betweenpillars.Thisvelocityprofileisclearlynon-physical context of autonomous microfluidic, the determination because it does not respect the non-slipping conditions of the hydraulic resistance of the capillary pump com- of the fluid velocity along walls. In order to solve ex- pared to the one of upstream elements is essential to actly the problem of the fluid flow inside a confined maximize the fluid flow rate. micropillarsarray,theStokesequationhastobesolved with the corresponding boundary conditions and thus the resulting viscous force can be deduced. We expect 3.2 Other lattices ourmodeltobeclosefromtheexactsolutioninthecase of low pillar densities but to be less accurate for large Thissectionaimstodiscusshowtheconclusionsdrawn pillardensitieswheretheconfinementmodifiesthefluid previously for square lattices of cylindrical pillars de- flow far from the assumed profile. Such an effect may pend on the pillars geometry and arrangement. With explain the discrepancy between experiments and the- the procedure described in Section 1 for microchannels oreticalpredictionsobservedforlargepillardensitiesin fabrication,wecreateoriginalpillarsarrangementssuch Fig. 4b. as square lattices of square pillars [Fig. 8a], square lat- Also, we assumed in Section 2.2 that the fluid flow tices of cylindric pillars with defects [Fig. 8b] and ran- adopts everywhere a fully developed velocity profile. domlydistributedlatticesofcylindricmicropillars[Fig. The time needed for the fluid to reach such a state can 8c]. beestimated.Whenanon-slippingboundarycondition For all these cases, we recorded the wicking of a is imposed to a fluid flow, it spreads in a lateral direc- √ Newtonian silicone oil V100 inside the microchannels. tion z according the relation: z = νt where ν = η/ρ The diffusivity coefficient D was measured and com- is the kinematic viscosity of the liquid. In the case of pared with the one of a reference case that we choose the microchannels studied previously, the non-slipping to be a square lattice of cylindric micropillars with the conditions on the top and bottom walls spread in the z same pillar density. For lattices that were non-square (cid:112) direction according z = ν/D x. The fully developed 0 with cylindrical pillars, the density of pillars is defined velocity profile is reached when z = h/2 which occurs by the ratio between the cross section of all the pillars (cid:112) for a critical traveled distance x /h = D /ν/2. min 0 and the total section of the microchannel. The results For η =100mPa·s, ρ=980kg/m3, γ =23mN/m and of these experiments are listed in Table 1. h = 80µm we estimate that x /h ∼ 10−3. As the min First, one notices that in the three situations con- microchannel is much longer than its height in our ex- sidered previously, the diffusivity modification induced periments,theassumptionofafullydevelopedvelocity by a change of pillar geometry or lattice arrangement profile is justified. are moderate. Indeed, the relative variation of the dif- fusivity stays below 3 % in the three situation consid- ered above. Thus, the dynamics of the contact line in 3.4 Comparison between open and confined a microchannel is mainly ruled by the micro-structure micropillar arrays density φ and their aspect ratio h/d. In details, we ob- serve that the change in the geometry of the pillars Thepresenceofatopwallmodifiestheliquidimpregna- from cylindrical to square section reduces the dynam- tionbehaviorcomparedtothecaseofanopenmicropil- ics of the liquid impregnation. The presence of defects lar array. Such a wall introduces two counteracting in- in a square lattice of cylindric micropillars enhances gredients on the capillary flow: an additional surface slightlythefluidwicking.Finally,arandomdistribution towetandanotherno-slipboundarycondition.Follow- ofcylindricmicropillarsisabitmoreefficientrelatively ing the same procedure as in Section 2.2 with a free toasquarearrangement.Obviously,asystematicstudy boundary condition at the top, we deduce a theoreti- asafunctionofthelatticeparametershastobeleadin calpredictionforthediffusivityD inthecaseofan open order to conclude on the precise impact of square sec- openmicrochannel.Theratiobetweenthediffusivityin Wickingthroughaconfinedmicropillararray 7 (a) (b) (c) Fig. 8 Examples of the progression of a Newtonian silicone oil of dynamic viscosity η = 100mPa·s in original lattices of micropillars: (a) Square lattice of square pillars of width d = 250µm. The pitch between pillars is p = 750µm. (b) Square lattice of cylindric micropillars of diameter d = 250µm with defects. The pitch of the lattice is p = 600µm. (b) Lattice of randomly distributed cylindric micropillars of diameter d = 250µm. In these three experiments, the time step between two positionsofthecontact lineis1s. Lattice d(µm) p(µm) φ h(µm) D(m2/s) D (m2/s) D/D ref ref square section pil- 250 750 0.11 80 5.82×10−6 5.94×10−6 0.98 lars square lattice of 250 600 0.12 80 6.06×10−6 5.88×10−6 1.03 cylindric pillars with defects random lattice of 250 - 0.05 80 6.2×10−6 6.07×10−6 1.02 cylindric pillars Table 1 Measurements of diffusivity coefficients D for a Newtonian silicone oil V100 (ρ = 980kg/m3, γ = 23mN/m and η = 100mPa·s) in original lattices (with square section pillars, square lattice with defects and random lattice of cylindrical pillars). The diffusivity coefficient is compared with a reference case defined as a square lattice of cylindrical pillars with the samedensityofpillars. an open micropillar array and the diffusivity in a close 3 one reads Dopen =(cid:32)1+(cid:0)4h−1(cid:1)φ(cid:33) 1+h2/(cid:113)φφm (cid:16)(cid:113)φφm −1(cid:17)2 (16)2.5 1 D 1+(cid:0)2h−1(cid:1)φ 1+2h2/(cid:113)φm (cid:16)(cid:113)φm −1(cid:17)2 2 0.9 φ φ d Figure 9 shows D /D as a function the pillar /1.5 0.8 open h density φ and the normalized height h as predicted by 1 Eq. (16). One notices that depending on the pillar as- 0.7 pect ratio and the pillar density, the diffusivity ratio 0.5 D /D is lower or larger than one. Thus, the two 0.6 open antagonistic effects, i.e. capillary pumping and viscous 0 0.2 0.4 0.6 resistance,causedbytheintroductionofatopwallcan φ overcomeeachotherregardingthepropertiesofthemi- Fig. 9 Ratio D /D as a function of the pillar density φ cropillar array. For dense arrays of slender pillars, the open andthe normalizedheight ofpillarsh=h/daspredicted by liquid impregnation in a confined micropillar array is equation(16).Thewhitesolidlineseparatestheregionwhere faster than in a similar open array. In such situations, thediffusivityratioD /D islargerthanone. open the confinement enhances the liquid spreading. the presence of pillars can either enhance or slow down Conclusions the dynamics of the contact line. However, the pillars only slow down the flow rate of the liquid which pen- The wicking of a micropillar array confined between etrates into the microchannel. A model based on the two plates by a perfectly wetting and viscous liquid estimationofcapillaryforcesandviscousfrictioninside has been investigated experimentally. We showed that the microchannel was developed. This latter provides 8 BaptisteDarboisTexier, PhilippeLaurent,SergueiStoukatch,St´ephane Dorbolo a fair agreement with experimental data. In the end, maximizes the diffusivity ratio D/D for a given pillar 0 this study predicts the fall of the efficiency of the liq- density verifies uid impregnation with increasing pillar density. This result has implications for the underfill process in mi- φ(cid:18)(cid:113)φm (cid:16)(cid:113)φm −1(cid:17)2+h2 (cid:19)= φ φ max (18) croelectronics packaging where the liquid filling time is (cid:0) (cid:1) h 1+(2h −1)φ a limiting factor of the production (Wan et al (2008)). max max A perspective of this work is to refine the basic The solution of Eq. (18) is presented by a white model proposed to describe the liquid dynamics. The dotted line in Fig. 6. exactdeterminationoftheviscousforceinthemicrochan- nel can be addressed by the way of numerical simu- lations. Besides, this study only considers pillars with References cylindricalandsquaresection.Itcouldbeinterestingto varythisshapeandintroduceconcavegeometriesinor- Op de Beeck J, De Malsche W, Tezcan D, De Moor P, der to study the effect on the liquid dynamics and the Desmet G (2012) Impact of the limitations of state- possible entrapment of air bubbles. Furthermore, this of-the-art micro-fabrication processes on the perfor- study only considers the ideal situation where the liq- mance of pillar array columns for liquid chromatog- uid perfectly wets the entire surface of the microchan- raphy. Journal of Chromatography A 1239:35–48 nel. The modification of our predictions in the case of BocquetL,BarratJL(2007)Flowboundaryconditions a liquid which partially wets the solid is an open ques- from nano-to micro-scales. Soft matter 3(6):685–693 tionwhichhastobeinvestigated.Finally,theimpactof Del Campo A, Greiner C (2007) Su-8: a photore- the non-Newtonian behaviors of a liquid on its wicking sist for high-aspect-ratio and 3d submicron lithogra- dynamics through a micropillar array can also be the phy. Journal of Micromechanics and Microengineer- scoop of future studies. ing 17(6):R81 Folch A, Ayon A, Hurtado O, Schmidt M, Toner M Acknowledgements Thisresearchhasbeenfundedbythe (1999) Molding of deep polydimethylsiloxane mi- Inter-university Attraction Poles Programme (IAP 7/38 Mi- crostructures for microfluidics and biological ap- croMAST) initiated by the Belgian Science Policy Office. plications. Journal of Biomechanical Engineering SD thanks the FNRS for financial support. We acknowledge Mathilde Reyssat, Tristan Gilet and Pierre Colinet for fruit- 121(1):28–34 ful discussions and valuable comments. We are grateful to Gamrat G, Favre-Marinet M, Le Person S, Baviere R, St´ephanie Van Loo for precious advices concerning the real- AyelaF(2008)Anexperimentalstudyandmodelling izationofthePDMSmicrochannels. ofroughnesseffectsonlaminarflowinmicrochannels. Journal of Fluid Mechanics 594:399–423 Gunda N, Joseph J, Tamayol A, Akbari M, Mitra S Appendix A (2013) Measurement of pressure drop and flow resis- tance in microchannels with integrated micropillars. Equation (12) allows to determine the critical pillar Microfluidics and nanofluidics 14(3-4):711–721 density φ in order to have D = D with φ (cid:54)= 0. The HaleR,BonnecazeR,HidrovoC(2014a)Optimization c 0 c critical density verifies the relation of capillary flow through square micropillar arrays. International Journal of Multiphase Flow 58:39–51 (cid:115) (cid:32)(cid:115) (cid:33)2 2 Hale R, Ranjan R, Hidrovo C (2014b) Capillary flow φ φ h φ m m −1 = (17) throughrectangularmicropillararrays.International c φc φc 2h−1 Journal of Heat and Mass Transfer 75:710–717 Ishino C, Reyssat M, Reyssat E, Okumura K, Quere which has a solution only if h > 0.5. Equation (17) is D(2007)Wickingwithinforestsofmicropillars.EPL solved numerically as a function of h and the solution (Europhysics Letters) 79(5):56,005 is indicated in Fig. 6 by a white solid line. Kim S, Moon MW, Lee KR, Lee DY, Chang Y, Kim HY (2011) Liquid spreading on superhydrophilic mi- For h > 0.5, the diffusivity ratio D/D reaches a cropillararrays.JournalofFluidMechanics680:477– 0 maximal value for a pillar density φ which can be 487 max calculated by deriving Eq. (12) relatively to φ keep- Kissa E (1996) Wetting and wicking. Textile Research ing h constant. Such a calculation is performed numer- Journal 66(10):660–668 ically and the solution is indicated in Fig. 6 by a white Mai T, Lai C, Zheng H, Balasubramanian K, Leong K, dashed line. Finally, the pillar aspect ratio h which LeeP,LeeC,ChoiW(2012)Dynamicsofwickingin max Wickingthroughaconfinedmicropillararray 9 siliconnanopillarsfabricatedwithinterferencelithog- Washburn EW (1921) The dynamics of capillary flow. raphyandmetal-assistedchemicaletching.Langmuir Physical review 17(3):273 28(31):11,465–11,471 Xiao R, Wang EN (2011) Microscale liquid dynamics Mognetti B, Yeomans J (2009) Capillary filling in mi- and the effect on macroscale propagation in pillar crochannels patterned by posts. Physical Review E arrays. Langmuir 27(17):10,360–10,364 80(5):056,309 Zimmermann M, Schmid H, Hunziker P, Delamarche Mohammadi A, Floryan J (2013) Pressure losses E (2007) Capillary pumps for autonomous capillary in grooved channels. Journal of Fluid Mechanics systems. Lab on a Chip 7(1):119–125 725:23–54 Nagrath S, Sequist LV, Maheswaran S, Bell DW, Ir- imia D, Ulkus L, Smith MR, Kwak EL, Digumarthy S, Muzikansky A, et al (2007) Isolation of rare circu- lating tumour cells in cancer patients by microchip technology. Nature 450(7173):1235–1239 Sadiq T, Advani S, Parnas R (1995) Experimen- tal investigation of transverse flow through aligned cylinders. International Journal of Multiphase Flow 21(5):755–774 Schwiebert MK, Leong WH (1996) Underfill flow as viscous flow between parallel plates driven by capil- lary action. Components, Packaging, and Manufac- turing Technology, Part C, IEEE Transactions on 19(2):133–137 Semprebon C, Forsberg PSH, Priest C, Brinkmann M (2014) Pinning and wicking in regular pillar arrays. Soft Matter Spruijt E, Le Guludec E, Lix C, Wagner M, Qu´er´e D (2015)Liquidfilmificationfrommenisci.EPL(Euro- physics Letters) 112(1):16,002 Tamayol A, Bahrami M (2009) Analytical determina- tion of viscous permeability of fibrous porous me- dia.InternationalJournalofHeatandMassTransfer 52(9):2407–2414 Tamayol A, Yeom J, Akbari M, Bahrami M (2013) Low reynolds number flows across ordered arrays of micro-cylinders embedded in a rectangular mi- cro/minichannel. International Journal of Heat and Mass Transfer 58(1):420–426 ThompsonP,RobbinsM(1989)Simulationsofcontact- line motion: slip and the dynamic contact angle. Physical Review Letters 63(7):766 TienC,SunK(1971)Minimummeniscusradiusofheat pipewickingmaterials.InternationalJournalofHeat and Mass Transfer 14(11):1853–1855 Vangelooven J, De Malsche W, De Beeck JO, Eghbali H, Gardeniers H, Desmet G (2010) Design and eval- uation of flow distributors for microfabricated pillar array columns. Lab on a Chip 10(3):349–356 Wan J, Zhang WJ, Bergstrom D (2008) Experimental verificationofmodelsforunderfillflowdrivenbycap- illary forces in flip-chip packaging. Microelectronics Reliability 48(3):425–430