Tamkang Journal of Science and Engineering, Vol. 5, No. 3, pp. 129-136 (2002) 129 Te Manufacture and Test of (110) Orientated Silicon Based Micro Heat Excanger Sung-Wen Kang, Yu-Tang Cen and Guang-Sang Cang Department of Mecanical and Electro-Mecanical Engineering Tamkang University Tamsui, Taiwan 251, R.O.C. E-mail: david1@mail.tku.edu.tw Abstract Te micro cross-flow eat excanger made of (110)-orientated silicon is fabricated by bulk micromacining tat is compatible wit semiconductor producing processes, and wafers wit undreds of ig aspect ratio cannels are bonded togeter by diffusion bonding wit aluminum as medium layers. Te core of te micro eat excanger is about 0.918 cm 3, and te density of te eat transfer area is 15,294 m 2 /m 3. Using pure water as te working fluid, te Reynolds Numbers sow tat te fluid field is always laminar flow, and as te maximum pressure drop reaces 2.47 bar, te flow rate is greater tan 4.5 L/min. Te eat transfer measured between ot and cold fluid is 5 kw; te log mean temperature is greater tan 30 K, and tat makes te overall eat transfer coefficient up to 24.7 kw/m 2 -K, corresponds to a volumetric eat transfer coefficient of 188.5 MW/m 3 -K. Except reacted on a few special cemicals, silicon as excellent properties in mecanics, eat transfer, and anti-corrosion, so te (110) silicon based micro eat excanger suits for te operations at ig temperature or in corrosive fluids. Te extremely small eat sink appears to ave a variety of current and potential applications in areas suc as micro-electronic cooling and biomedical processes were ig eat transfer power are required wit little weigt and small volume. Key Words: Micro Cross-Flow Heat Excanger, Bulk Micromacining, Heat Transfer Area, Pressure Drop, Heat Transfer Coefficient 1. Introduction Te development of micro eat excanger began wit solving te problem of eat dissipation in VLSI. In 1984, Tuckerman [1] used anisotropic etcing and precision mecanical sawing tecniques to manufacture about 280 µm deep microcannels in 500 µm tick (110) orientated silicon wafers. He reported a ybrid of direct contact and cold plate tecnologies using laminar water flow in te micropassages and absorbing eat flux of 150 W/cm 2. Bot Cross and Ramsaw [2] reported on a type of eat excanger called printed circuit eat excanger. Te cannels on te plates of te special eat excanger were macined by using cemical etcing metods, and te plates were bonded togeter by diffusion brazing. Te cannel size is 0.4 mm wide and 0.3 mm deep, and te volumetric eat transfer coefficient is 7 MW/m 3 -K.
130 Sung-Wen Kang et al. Hig-capacity micro eat excangers ave also been produced using diamond macining tecniques by Friedric and Kang [3], owever, te processes are not easy for batc-fabrication. In te present work, te cannel and complete device fabrication is using silicon wet etcing tecniques. Te typical process include mask-making, potoresist coating and baking, exposing, etcing and cleaning. Te anisotropic etcing of SCS was discovered in te early 1960s. In 1969, Lee [4] etced silicon wafers by organic etcants, and e found tat te etcing rates on (111) were very slow. Bean [5] reported tat te extremely ig ratio greater tan 650 to 1 were obtained in (110) orientation-dependent etcing. Bean reported tat te etcing rate of (110) silicon in 35% w.t. KOH was about 0.8 µm/min. at 80 ; Krause and Obermeler [6] proved tat te etcing rate and te aspect ratio of etced cannels would increase wit te concentration of KOH liquor in recent report. Studies of conventional compact eat excangers design teories were already completed by Kays and London [7]. Te metallic eat excanger [8] wit features tat range in size from 150-750 µm ave termal resistances ranging from 0.07 to 0.12 /W cm at flow rates of water of ~20 L/ and pressures of 3 8.6-83 10 N/m 2 and fabricated using sacrificial polymer mandrils. N. Saji, et al. [9] developed a compact laminar flow eat excanger wit stainless steel micro-tube, wic consists of 12 elements wit a total of 4800 stainless steel micro-tube. Eac element is formed wit 400 tubes tat ave an inner diameter of 0.5 mm, an outer diameter of 0.7 mm and is 310 mm long. In te major processes of fabricating te silicon based micro eat excanger, bonding is te key point. Direct bonding is one of te major bonding metods in semiconductor processes by te ydrate bonding on te wafer and eat treatments in ig temperature atmospere. Except anodic bonding and direct bonding, te diffusion bonding is te well-cosen metod to bond silicon wafer togeter in lower temperature. Because only a few studies in te micro eat excangers area ave been reported, researces are needed to investigate te basic design, fabrication and performance of te micro eat excangers to expand te developments and strengten teir reliability. Te objectives of tis researc were referred to design, fabricate and test a (110) silicon based micro cross-flow eat excanger. 2. Design A cross-flow arrangement was adopted in tis micro eat excanger design because tis arrangement greatly simplified te eader design at te entrance and exit of eac fluid. A generic design for a two-fluid, cross-flow plate eat excanger is sown in Figure 1. Figure 1. A generic plate-type cross-flow micro eat excanger Te purpose for using micro cannels is to acieve specific eat transfer performance, U, witin minimum weigt, material and volume constraints. Since te eat transfer in ot and cold fluids are Q = m c,m (T,i - T,o ) (1) Q c = m c c c,m (T c,o - T c,i ) (2) and te average eat transfer is Q m = 1/2 (Q + Q c ) (3) te overall eat transfer coefficient can be calculated from Q U = m A T F m (4) Were A is te total eat transfer area on one side, and F is te correction factor for a cross-flow eat excanger wit bot fluids unmixed. Te log mean temperature difference
Te Manufacture and Test of (110) Orientated Silicon Based Micro Heat Excanger 131 ( T - T )- ( T - T ),i c,i,o c,i T m = (5) T,i - Tc,i ln T,o - Tc,o Te total volumetric eat transfer surface coefficient could be expressed as U v = U B (6) In wic is te eat excanger eat transfer surface area per unit volume. Te overall eat transfer coefficient may be expressed as 1 UA 1 1 = + R w + (7) A A c Due to te ig termal conductivity of silicon and te small wafer tickness, a simple calculation sows tat te termal resistance of te wall (Rw) could be negligible compared wit te convection resistance. Because te cannels on te two sides ave almost te same geometry, mass flow rate, and eat transfer conditions, it is reasonable to assume tat te convection eat transfer coefficient must be te same on bot sides, terefore (8) c and simplify it sequentially to = 2U (9) Te Nusselt Number is of te form D Nu = (10) kf Te great advantage of using tin silicon wafers wit micro flow cannels is te ability to produce a very large ratio of eat transfer area per unit volume. In addition, te large termal resistance between te eat excanger plates is very low due to te tin, ig conductivity silicon walls. Te cross-flow eat excanger is made by stacking and bonding silicon wafers togeter wic ave ad flow cannels etced into tem by litograpy.te factors tat need to be addressed in te design of any of tese devices include eat transfer capacity, pressure drop, size and number of cannels, spacing between cannels, and internal deformation due to differential pressure and temperature. Any consideration of mecanical devices made from silicon must certainly take into account te mecanical beavior and properties of single-crystal-silicon. Te Young's modulus of silicon is 1.9 10 7 N/cm 2, and as a value approacing tat of stainless steel. Te Knoop ardness of silicon (850 kg/mm 2 ) is almost twice as ig as nickel and iron. SCS as a tensile yield strengt of 6.9 10 5 N/cm 2, wic is at least 3 times iger tan stainless steel wire. 3. Fabrication Litograpy tecnology is widely used for semi-conductor processes; it includes waferscleaning, growt of etcing mask, potoresist coating, exposure and develop, etcing, doping, etc., almost all of tem made a contribution to tis study. Tree etcant systems are of particular interest due to teir versatility: etylene diamine pyrocatecol (EDP) and water, KOH and water, HF, HNO 3, and acetic acid, (CH 3 COOH). In tis study, KOH system was used. KOH and water is orientation dependent and, in fact, exibits muc iger (110) to (111) etcing rate ratio tan EDP, and for tis reason it is especially useful for grooves etcing on (110) wafers since te large differential etcing ratio permits deep, ig aspect ratio grooves wit minimal undercut of te mask. Si 3 N 4 is te preferred masking material for ig aspect ratio, long KOH etcing cannels. We used 120 nm silicon nitride and 180 nm silicon dioxide as te etcing mask. Te concentration of etcant also influences te etcing rate effectively. Te etcing rate would ascend wit te quantity of ydrate in liquor. Generally speaking, te etcing rate of 40% w.t. would be larger tan tat of 25% w.t., because te ydrate in 25% w.t. KOH are more tan tat in 40% w.t. KOH. Temperature affects etcing rate significantly. Te etcing rate and temperature ave a relation tat could be expressed as R = R o exp(-e c /kt) (11) Te R o represents te Pre-Exponential Factor; in eq. (11), we understand tat te etcing rate will raise apparently wit temperature. K. E. Bean reported tat te (110) wafers were etced at about 0.8 µm/min. in 50% w.t. KOH at 80. In our study 40% w.t. KOH wit stirring at 65 was adopted, and it took about 400 minute to etc a 200 µm deep cannel; te misalignment was about 3 ~5, so tat te undercut beside te 40 mm wide cannel was about 5~10 µm. Figure 2 sows te micrograp of te cannels on te wafer.
132 Sung-Wen Kang et al. Figure 2. Te micrograp of te cannels on te wafer (wafer tickness: 370 µm, cross section of te cannels: 40 µm 200 µm, bottom tickness: 170 m, fin widt: 40 µm) Te pattern we designed is 15 mm wide and 25 mm long, and te cannel area is 10 mm wide by 10 mm long; tere are 125 cannels wit 40 µm widt and 200 µm dept in te cannel area. Temperature is te most influential variable during diffusion bonding. In any termally activated process an incremental cange in temperature will cause te greatest cange in process kinetics wen compared to most oter process variables. In addition, virtually all te mecanisms in diffusion bonding are temperature dependent. In general, te temperature at wic diffusion bonding will take place is above one alf of te absolute melting temperature. Time is closely related to temperature in wic most diffusion controlled reaction rates vary wit time. Experience indicates tat increasing bot te time and te pressure at bonding temperature increase joint strengt up to a limit. Beyond tis point no furter gains are acieved. Pressure is important just in te beginning wen te bonding takes place. Oxide layers will fracture and te grains will deform and contact one anoter between te working pieces if necessary pressure is given, but as te diffusion is ongoing, pressure is not as important as in te beginning. In tis researc low pressure is employed to avoid te tin walls on te silicon wafers being bulked, and te pressure of 240 kg/cm 2 was used for clamping te wafers Silicon as a ig melting point of more tan 1400, and te temperature at wic te diffusion bonding takes place would ig enoug to increase te complexity, for tis reason, a low melting point medium material will lower te bonding temperature, in addition, aluminum as a good diffusion property wit silicon, so we consider aluminum foils as te medium layers in te bonding processes. In tis researc, te aluminum foils were cleaned in a 5% ydrocloric acid for 5 minutes wit ultrasonic saking to remove alumina on te foils, and te silicon wafers were cleaned by te same way in litograpy. Te foils and wafers are ten stacked and clamped in a grapite clamp, and tey were placed in a diffusion bonding furnace, and were eated to 600 for one our, and te vacuum was approximately 15 µtorr. A micrograp of an edge of te micro eat excanger is sown in Figure 3. Figure 4 sows te packaged micro eat excanger by a 10-dollar coin. Figure 3. A micrograp of an edge of te micro eat excanger Figure 4. Te micro eat excanger (cannel eat transfer lengt: 0.9 cm, number of cannel on eac wafer: 125, number of wafer layers: 26, number of cannels on one side: 1625, free flow area: 13 mm 2, free flow area/ frontal area: 12.7 %, total eat transfer area on one side: 70.2 cm 2, total eat transfer volume: 0.918 cm 3, total eat transfer area/ total volume: 15295 m 2 /m 3 ) 4. Test
Te Manufacture and Test of (110) Orientated Silicon Based Micro Heat Excanger 133 Figure 5 sows a scematic of te experimental apparatus used to test te micro eat excanger. To minimize te amount of equipment required, measurements were made using only DI water as te working fluid on bot ot and cold side of te eat excanger. A test setup contains separate measuring devices for measuring te flow, as well ot fluid as pressure and temperature at bot inlets and outlets of te micro eat excanger. Te ot and cold water was supplied in a close circuit from ot and cold water tank. Te same flow rate was used for bot streams by adjusting te control valves. Te water pressure and temperature were measured at te inlets and outlets respectively. cold fluid pump pump out out ot fluid tank filter filter cold fluid tank flow meter flow meter in Pressure transducer P c,1 T c,1 in P,1 P,o T,1 T,o Micro eat excanger P c,o T c,o Te data obtained in te experiments were pressure drop, flow rates, inlet and outlet temperatures from te bot sides of te eat excanger; te flow rates of bot sides were kept te same by controlling te valves. Te experimental procedures consisted of varying te flow rates and measuring te pressures, temperatures, and te flow Figure 5. Scematic of te experimental apparatus rates. Te calculations were carried out according to equations based on a single-pass, cross-flow eat excangers wit bot te ot and cold water unmixed. From te measurements of te mass flow rate, inlet and outlet temperatures, Reynolds number (Re), te eat transfer (Q), te
134 Sung-Wen Kang et al. corresponding overall eat transfer coefficient (), volumetric eat transfer coefficient (U v ) and could be calculated. Figure 6 sows te measured pressure drop on te ot side and cold side as a function of te water flow rate. We found te maximum pressure drop reaced 2.47 bar wen flow rate is 4.5 L/min. As we kept te inlet temperatures bot te ot and cold side in te same region, and only adjusted te flow rates, we could obtain corresponding log mean temperature differences, and tey are sown in Figure 7. 3.00 Cold Hot for te maximum flow rate measured were 820 on te ot side, and 440 on te cold side, tat is, only laminar flow occurred over te test range. Figure 9 sows te Nusselt number variation wit Reynolds number for bot te ot and cold sides. Te difference between te two curves is tat te Reynolds number range on te ot side is greater tan tat of te cold side due to te variation of te viscosity. 4 3 Uv U 25 20 pressure drop, bar pressure drop, bar 2.00 1.00 U, kw/m U, kw/m2-k 2 -K 2 1 15 10 5 Uv Uv,, MW/m3-K 3 1.00 2.00 3.00 4.00 5.00 flow rate,, l/min L/min Figure 6. Measured pressure drop as a function of te flow rate 8 T,i i 8 10 20 30 40 50 Re, cold side Figure 8(a). U and U v as a function of Re on bot cold sides 4 3 Re, cold side Uv U 20 temperature, C Temperature, 6 4 LMTD calculated T,oT,o T c,0 Tc,o 6 4 LMTD.C, C U, U, kw/m2-k 2 2 1 15 10 5 Uv Uv,, MW/m3-K 3 2 T Tc,o c,o 2 1.00 2.00 3.00 4.00 5.00 flow rate,, l/min L/min Figure 7. Relationsip of te temperature and te flow 20 40 60 80 100 Re, ot side Re, ot side Figure 8(b). U and U v as a function of Re on ot sides In Figure 8 te overall eat transfer coefficient (U) and te volumetric eat transfer coefficient (U v ) calculated from te measured data as been plotted versus te Reynolds number on te bot sides. Te igest measured U, 24.7 kw/m 2 -K, corresponds to an U v of 188.5 MW/m 3 -K. Te Reynolds number
Te Manufacture and Test of (110) Orientated Silicon Based Micro Heat Excanger 135 Nu Nu 8.00 7.00 6.00 5.00 4.00 3.00 2.00 Cold Hot 25 50 75 100 Re Figure 9. Nu as a function of Re on bot ot and cold sides 5. Conclusion Micro manufacturing tecniques suc as litograpy processes and vacuum diffusion bonding were adopted to fabricate te silicon based micro eat excangers, wic ave dimensions an order of magnitude smaller tan in conventional compact eat excanger. Potential applications for te eat excanger are seen in areas suc as microelectronics cooling, aircraft, aerospace cooling and biomedical processes were ig eat transfer power are required wit little weigt and small volume. A test system was developed to measure micro eat excanger flow-friction and termal performance. Using DI water as te working fluid, about 5 kw of energy could be transferred in a cubical eat transfer volume of 0.918 cm 3 wit a log mean temperature difference of about 30 K. Tis corresponds to a overall eat transfer coefficient of more tan 24.7 kw/m 2 -K, and volumetric eat transfer coefficient of 188.5 MW/m 3 -K. A B c D E c F k Nomenclature eat excanger total transfer area on one side ratio of total eat transfer area on one side of te excanger to total volume of te excanger specific eat ydraulic diameter reactive activation energy eat excanger correction factor eat convection coefficient Boltzmann constant k f m Nu Q R R w T U U v c i m o termal conductivity of working fluid mass flow rate mass flow rate Nusselt Number eat transferred reactive rate termal resistance of wall temperature overall eat transfer coefficient volumetric eat transfer coefficient Subscripts cold fluid ot fluid inlet condition mean condition outlet condition References [1] Tuckerman, D. B., Heat-Transfer Microstructures for Integrated Circuits, P.D. Tesis, Department of Electronic Engineering, Stanford University, U.S.A. (1984). [2] Cross, W. T. and Ramsaw, C., Process Intensification Laminar Flow Heat Transfer, Cemical Engineering Researc & Design: Transaction of te Institute of Cemical Engineers, Vol. 64, pp. 258-294 (1986). [3] Friedric, C. R. and Kang, S. W., Micro Heat Excangers Fabricated by Diamond Macining, Precision Engineering, Vol. 16, pp. 56-59 (1994). [4] Lee D. B., Anisotropic Etcing of Silicon, Journal of Applied Pysics, Vol. 40, pp. 4569-4574 (1969). [5] Bean, K. E., Anisotropic Etcing of Silicon, IEEE Trans. Electron Devices, Vol. ED-25, pp. 1185 (1978). [6] Krause P. and Obermeler E, Etc and Surface Rougness of Deep Narrow U-Grooves in (110)-Orientated Silicon, J. of Micromec. Microeng., Vol. 5, pp. 112-115 (1992). [7] Kays, W. M. and London A. L., Compact Heat Excangers, 3rd. ed. McGraw Hill Co. N.Y., U.S.A. (1964). [8] Arias, F., Oliver, S. R. J., Xu, B., Holmlin, R. E. and Witesides, G. M., Fabrication of Metallic Heat Excangers Using Sacrificial Polymer Mandrils, Journal of Microelectromecanical System, Vol. 10, pp. 107-112 (2001). [9] Saji, N., Nagai, S., Tsucyia, K., Asakura, H. and Obata, M., Development of a compact
136 Sung-Wen Kang et al. laminar flow eat excanger wit stainless steel micro-tubes, Pysica C 345, pp. 148-151 (2001). Manuscript Received: Mar. 21, 2002 and Accepted: Jun. 26, 2002