International Journal of Heat and Mass Transfer 46 (2003) 25472556 www.elsevier.

com/locate/ijhmt

An experimental study of convective heat transfer in silicon microchannels with dierent surface conditions
H.Y. Wu, Ping Cheng
*
Department of Mechanical Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong Received 31 July 2002; received in revised form 9 January 2003

Abstract An experimental investigation has been performed on the laminar convective heat transfer and pressure drop of water in 13 dierent trapezoidal silicon microchannels. It is found that the values of Nusselt number and apparent friction constant depend greatly on dierent geometric parameters. The laminar Nusselt number and apparent friction constant increase with the increase of surface roughness and surface hydrophilic property. These increases become more obvious at larger Reynolds numbers. The experimental results also show that the Nusselt number increases almost linearly with the Reynolds number at low Reynolds numbers (Re < 100), but increases slowly at a Reynolds number greater than 100. Based on 168 experimental data points, dimensionless correlations for the Nusselt number and the apparent friction constant are obtained for the ow of water in trapezoidal microchannels having dierent geometric parameters, surface roughnesses and surface hydrophilic properties. Finally, an evaluation of heat ux per pumping power and per temperature dierence is given for the microchannels used in this experiment. 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Convective heat transfer; Microchannels; Roughness; Hydrophilicity

1. Introduction With the advances of MEMS technology, various silicon-base microsystems such as micro-heat sinks, micro-biochips, micro-reactors, and micro-fuel cells have been developed in recent years. Since microchannels of noncircular cross sections are usually integrated in these silicon-base microsystems, it is important to know the uid ow and heat transfer characteristics in these microchannels for better design of various microsystems. Existing experimental results of friction factors and Nusselt numbers in microchannels show that there is a great deal of discrepancies between classical values and the experimental data. Experimental data also appear to be inconsistent with one another. Various reasons have been proposed to account for these dierences. One

Corresponding author. Tel.: +852-2358-7181; fax: +8522358-1543. E-mail address: mepcheng@ust.hk (P. Cheng).

reason may be attributed to various surface conditions, which cannot be neglected in microsystems because of the large surface-to-volume ratio in these systems [1]. For example, the theoretical investigation by Yang et al. [2] and Mala et al. [3] showed that the liquid ow and heat transfer in the rectangular microchannels may be signicantly inuenced by the presence of the electric double layer near the solid-liquid interfaces.Wu and Little [4] measured the friction factors of gas laminar ow in the trapezoidal silicon/glass microchannels, and found that the surface roughness aected the values of the friction factors even in the laminar ow, which is dierent from the conventional macrochannel ow. Rahman [5] measured heat transfer in microchannel heat sinks, and obtained larger average Nusselt number than those predicted for larger size channels, which they attributed to the surface roughness associated with the etched channel structure. However, they did not quantify the inuence of surface roughness on microchannel heat transfer. Qu et al. [6,7] performed an experimental investigation on pressure drop and heat transfer of water

0017-9310/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0017-9310(03)00035-8

2548

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556

Nomenclature Ac Aw cp C1;2;3 Dh fapp fapp Re f1 , f2 h H H =Wt k k =Dh L L=Dh M N Nu Dp Pr Q qp cross-sectional area of the microchannel total area of channel bottom wall and side walls specic heat of water correlation factors depending on surface material hydraulic diameter of microchannel apparent friction factor dened in Eq. (1) apparent friction constant function average heat transfer coecient height of microchannel height-to-top width ratio surface absolute roughness surface relative roughness length of microchannel length-to-diameter ratio mass ow rate of water total number of microchannels Nusselt number dened in Eq. (8) pressure drop across the microchannel Prandtl number of water total heat removed by water heat ux per pump power and per temperature dierence Re Reynolds number of water T1 ; . . . ; T5 longitudinal wall temperature Tin inlet water temperature Tout outlet water temperature DTm mean temperature dierence between water and microchannel wall u  mean velocity of water bottom width of microchannel Wb Wt top width of microchannel Wb =Wt bottom-to-top width ratio Greek symbols k thermal conductivity of water m kinetic viscosity of water h channel angle q density of water Subscripts app apparent b bottom c cross-section in inlet out outlet t top w wall

ow in trapezoidal silicon microchannels. They attributed both the measured higher pressure drops and lower Nusselt numbers to the wall roughness, and proposed a roughness-viscosity model to interpret their experimental data. According to their model, however, the increase in wall roughness caused the decrease in Nusselt number that is contradictory to common sense. The review paper by Meinhart and Wereley [1] pointed out that the uid transport in nanotubes could be facilitated by decreasing the surface hydrophilic capability, although this novel concept has not been veried experimentally. Ma et al. [8] performed a numerical simulation of microchannel ow over a hydrophilic and hydrophobic surface. They showed that the liquid can remain more easily inside the well of hydrophilic inner surface than that of the hydrophobic inner surface. As far as the authors are aware, however, no published experimental data on the eect of surface hydrophilic property on the convective heat transfer in microchannels has been obtained. In addition to the various interfacial eects discussed above, the cross-sectional shape of the channel can have great inuence on the uid ow and heat transfer inside the noncircular microchannels as shown by Ma and Petersons analysis [9], and was conrmed experimentally by Wu and Cheng [10]. Peng and Peterson [11]

performed experimental investigations on the pressure drop and convective heat transfer for water ow in rectangular microchannels, and found that the crosssectional aspect ratio had great inuence on the ow friction and convective heat transfer both in laminar and turbulent ows. The eect of cross-sectional aspect ratio on the cooling characteristics of the uid ow in the rectangular microchannels was also observed by Choi and Cho [12]. Very little experimental data, however, has been published on the eect of the geometric parameters in heat transfer and pressure drop in the trapezoidal microchannels. For the silicon microchannels used widely in various microsystems, the geometric parameters, the surface roughness, the surface hydrophilic property, and the ow parameters are the important factors aecting the uid ow and heat transfer characteristics. Depending on the etching method and etching time, two kinds of microchannels (trapezoidal or rectangular microchannels) having dierent aspect ratios can be fabricated in silicon wafers [13]. Also, various surface roughnesses in the microchannels can be obtained depending on the concentration and temperature of etching solutions [14]. The surface hydrophobic/hydrophilic property can easily be changed by decreasing or increasing the thickness of

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556

2549

oxide layer on a silicon surface [15]. In this paper, the eects of the geometric parameters, the surface roughness, the surface hydrophilic property on pressure drop and heat transfer in microchannels will be claried. Based on the experimental data, correlations in terms of various aecting factors will be presented for the laminar convective heat transfer and pressure drop in the trapezoidal silicon microchannels. A comparison of heat ux per pumping power and per temperature dierence for dierent microchannels used in this experiment is also obtained.

Wt
microchannel

silicon wafer Wb

Fig. 2. Cross-section of microchannels.

2. Experimental description 2.1. Experimental setup Fig. 1 shows the experimental setup used in the present investigation. The deionized water in the highpressure tank, being pushed by the compressed nitrogen gas, owed successively through a valve regulator, a lter, a owmeter, to the test section, and nally was collected by a container. The collecting container was placed on a high precision electronic balance. The mass ow rate of water indicated by the owmeter was calibrated by the primary measurement of the water mass collected per unit time on the balance. The mass ow rate of water was regulated by the water valve. The schematic diagram of the test section is also shown in Fig. 1. The silicon chip having parallel microchannels was anodically bonded with a thin pyrex glass plate from the top. The chip was then placed in a teon base which was attached by a lm heater and a thermal-insulating layer at its bottom. The lm heater was energized by a DC power supply, which could be regulated in the ranges of 0220 V and 03 A. To measure the wall temperatures, ve type-T thermocouples (0.1 mm diameter), being equally spaced from the channel inlet to outlet, were attached at the bottom of the silicon chip. A layer of electric insulation with good heat conduction property was sandwiched between

the wall thermocouples and the lm heater. Finally, a polycarbonate plate with liquid inlet and outlet holes was aligned and secured to the teon base, leaving a leak-proof contact with the pyrex glass through two o-rings. The temperature and pressure of the deionized water at the inlet and outlet of the microchannels were measured by two type-T thermocouples (0.5 mm diameter) and two pressure transducers, respectively. All these measurements along with the measurements of the wall thermocouples were collected and displayed by a PC through a data acquisition system. The trapezoidal microchannels were formed by wet etching of the h1 0 0i silicon wafers in the solution of KOH. Here, the trapezoidal microchannels were chosen because of its wide application in various microsystems. Moreover, by controlling the concentration, temperature and time of the wet etching, microchannels having dierent surface roughness and geometric parameters can easily be fabricated, which is essential for the present experimental investigation. Fig. 2 shows the typical cross-sectional shape of the microchannels in the h1 0 0i silicon wafers by wet etching. It can be seen that the triangular channel or rectangular channel is a special trapezoidal channel with Wb 0 or h 90, respectively. Using a surface proler, the top width Wt , bottom width Wb , and depth H of the trapezoidal microchannels were measured. Also, the surface roughness could be determined using the atomic force microscope (AFM). Table 1 lists the geometric parameters and the surface relative roughness k =Dh of the 13 microchannels used in this
PC Data Acquisition System

gas valve N2

filter
H2O

flowmeter
Tw Ti

Pi

Po T o

Tw

Test section

water valve pressure tank polycarbonate plate microchannel film heater

collecting container

teflon base
Fig. 1. Experimental setup.

balance

2550

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556

Table 1 Geometric parameters and surface properties of microchannels Channel no. #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 Wt (106 m) 1473.08 770.48 549.83 423.2 248.83 157.99 437.21 171.7 429.99 168.03 555.00 251.5 158.12 Wb (106 m) 1375.86 672.63 454.71 327.4 153.03 61.62 270.19 0 262.30 0 459.54 155.7 62.30 H (106 m) 56.22 56.34 56.33 56.13 56.24 56.28 110.7 110.8 109.1 108.9 56.77 56.50 56.49 Wb =Wt 0.934 0.873 0.827 0.774 0.615 0.390 0.618 0 0.610 0 0.828 0.619 0.394 H =Wt 0.0382 0.0731 0.1024 0.1326 0.2260 0.3562 0.2532 0.6453 0.2537 0.6481 0.1023 0.2247 0.3573 k =Dh 9.85 105 7.59 105 5.69 105 8.63 105 7.64 105 4.30 105 3.26 105 3.62 105 5.87 103 1.09 102 9.76 105 5.71 105 6.94 105 L=Dh 285.41 298.67 310.46 325.11 370.85 453.79 191.77 362.35 195.34 369.29 307.90 368.11 451.40 Surface material Si Si Si Si Si Si Si Si Si Si SiO2 SiO2 SiO2

experiment. To study the eect of surface hydrophilic property on the ow and heat transfer, some silicon microchannels were deposited by a thermal oxide layer  to increase their surface hydrophilic capabilof 5000 A ity. Surface materials for dierent microchannels are also listed in Table 1. 2.2. Data reduction and error analysis 2.2.1. Flow friction In this paper, we use the apparent friction factor to evaluate the total pressure drop of the deionized water owing through the microchannels which is dened as, fapp Dp Dh 1 u2 L 2q 1

fapp Re

DpNAc D2 h 2mML

2.2.2. Heat transfer In this paper, the heat transfer coecient for the deionized water owing through the microchannels is dened by h Q NAw DTm 5

where Dp is the pressure drop measured by the inlet and outlet pressure transducers, q is the water density based on the mean value of the inlet and outlet temperatures, Dh and L are the hydraulic diameter and length of the microchannels, and  u is the average velocity of water obtained from the following equation M  u N qAc 2

where Aw is the heat transfer area which is the total area of the side walls and the bottom wall of the microchannel, but it does not include the area of the pyrex glass from the top because it is assumed to be an adiabatic condition. Q is the total heat removed by water which is calculated by Q Mcp Tout Tin 6

The mean temperature dierence DTm between the wall and the water is evaluated by 1 1 DTm T1 T2 T3 T4 T5 Tin Tout 5 2 7

where M is the mass ow rate of water, N is the number of microchannels in the chip, and Ac is the cross-sectional area of a microchannel. The apparent friction constant is dened as the product of the apparent friction factor fapp and the Reynolds number Re, i.e., fapp Re D p D2 h 2qm uL 3

where, T1 ; T2 ; . . . ; T5 are the wall temperatures measured by the ve wall thermocouples, and Tin , Tout are the inlet and outlet bulk temperatures of water. The Nusselt number based on the heat transfer coecient is given by Nu hDh k 8

where Re  uDh =m. Substituting Eq. (2) into Eq. (3), we obtain the apparent friction constant in terms of various measurements as

where k is the thermal conductivity of the uid. Substituting Eqs. (5) and (6) into Eq. (8), we obtain the Nusselt number in terms of various measurements as Nu Mcp Dh Tout Tin N kAw DTm 9

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556 Table 2 Measurement errors Parameters Dh L Ac Aw M Dp Tout Tin DTm Maximum errors (%) 1.88 0.82 1.34 1.53 0.98 1.11 0.73 2.37 Parameters q m cp k Re fapp Re Nu Maximum errors (%) 0.02 2.35 0.02 0.31 6.58 10.36 7.82

2551

The eect of each parameter inside the brackets on the Nusselt number and the apparent friction constant can be shown if other parameters are the same. In the following sections, we will discuss the eects of the surface roughness, surface hydrophilic property, geometric parameters and Reynolds number separately. 3.1. Eect of surface roughness It is well known that the surface roughness does not aect the laminar ow in macrochannels. However, as the channel size decreases to the order of a few microns, the eect in the surface roughness on the ow and heat transfer of the microchannels becomes important. Figs. 3 and 4 give the experimental results of two pairs of the silicon microchannels with dierent surface roughness. Microchannels #7 and #9 are two trapezoidal channels having the same geometric parameters (Wb =Wt 0.618 0.610, H =Wt 0.25320.2537, L=Dh 191.77195.34) but dierent relative surface roughness (3:26 105 5:87 103 ), while microchannels #8 and #10 have the triangular cross-section with same geometric parameters but widely dierent relative surface roughness (3:62 105 1:09 102 ). It is shown that at the same Reynolds number, the Nusselt number and apparent friction constant of the trapezoidal microchannel #9 are larger than those of the trapezoidal microchannel #7, which has much lower surface roughness than microchannel 9# . Similar increase in apparent friction constant and Nusselt number is observed for the triangular microchannel #10 (k =Dh 1:09 102 ) which has a higher surface roughness when compared to the triangular microchannel #8 (k =Dh 3:62 105 ). In fact, the surface

The mean bulk temperature of water Tin Tout =2 was used as the characteristic temperature to evaluate the physical properties of the water involved in all above calculations, including the density q, kinetic viscosity m, thermal conductivity k, and specic heat cp , which are assumed to be independent of pressure. 2.2.3. Error analysis According to Eq. (4), the error in determining the apparent friction constant fapp Re came from the measurement errors of Dp, Ac , Dh , M , L, and m. Also according to Eq. (9), the error in determining the Nusselt number Nu came from the errors of Aw , Dh , M , Tout Tin , DTm , cp and k. Performing the standard error analysis, the maximum uncertainties in determining Re, fapp Re and Nu are presented in Table 2, which also lists the maximum error of various parameters in this study. It can be seen that the errors of fapp Re and Nu due to measurement errors in this study were no more than 10.3% and 7.8%, respectively.

3. Experimental results and discussions From Table 1, it can be conjectured that the Nusselt number and the apparent friction constant are a function of the geometric parameters (Wb =Wt , H =Wt , L=Dh ), surface roughness k =Dh , and the surface material. In addition to these factors, the Nusselt number is also a function of the Reynolds number and the Prandtl number of the uid. Thus,   Wb H L k Nu f1 ; ; ; ; surface material; Re; Pr Wt Wt Dh Dh 10a The apparent friction constant is a function of the geometric parameters, surface roughness, surface material, and the Reynolds number, i.e.,   Wb H L k ; ; ; ; surface material; Re fapp Re f2 Wt Wt Dh Dh 10b

40
#9, k/Dh=5.87E-3 #7, k/Dh=3.26E-5 #10, k/Dh=1.09E-2 #8, k/Dh=3.62E-5

30

fapp Re

20

10

0 0 400 800 1200 1600

Re

Fig. 3. Eects of surface roughness on apparent friction constant.

2552
5

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556
28

4
24

#11, Thermal oxide #3, Silicon #12, Thermal oxide #5, Silicon #13, Thermal oxide #6, Silicon

fappRe
2 1
#9, k/Dh=5.87E-3 #7, k/Dh=3.26E-5 #10, k/Dh=1.09E-2 #8, k/Dh=3.62E-5

Nu

20

16

0 0 400 800 1200 1600

12 0 200 400 600 800

Re

Re

Fig. 4. Eects of surface roughness on Nusselt number.

Fig. 5. Eects of surface hydrophilic property on apparent friction constant.

Nu

roughness of microchannels #7 and #8 are so small (having order 105 ) that they can be regarded as smooth when compared to the microchannels #9 and #10, whose relative roughness is of the order 103 or more. It is also found that the Nusselt number and the friction constant of high roughness microchannel increase faster than those of low roughness microchannel with increasing Reynolds number.This is because that the disturbance in the boundary sublayer caused by the roughness is more signicant at high Reynolds numbers. 3.2. Eect of surface hydrophilic property The eects of surface hydrophilic property on the convective heat transfer and ow friction in the microchannels are shown in Figs. 5 and 6, where three sets of data (microchannels #3 and #11, microchannels #5 and #12, microchannels #6 and #13) for the apparent friction constant and the Nussult number are compared. It is pertinent to note that the microchannel pairs of #3 and #11, #5 and #12, #6 and #13 have the same order of roughness respectively. The solid symbols represent the data of the microchannels in the silicon wafer without thermal oxide coatings, while the open symbols represent the data of the microchannels whose silicon surfaces have been coated with a layer of thermal oxide. Note that the thermal oxide surface has a stronger hydrophilic capability than the bared silicon surface [15]. Figs. 5 and 6 show that the Nusselt number and the apparent friction constant of microchannels #11, #12 and #13 (whose surfaces were coated with a thermal oxide layer), are larger than those of the microchannels #3, #5 and #6 (whose surfaces were not coated), respectively. It can be concluded from these experimental results that the increase in the surface hydrophilic ca-

2.5

1.5

1
#11, Thermal oxide #3, Silicon #12, Thermal oxide #5, Silicon #13, Thermal oxide #6, Silicon

0.5

0 0 200 400 600 800

Re

Fig. 6. Eects of surface hydrophilic property on Nusselt number.

pability gives rise to the increase in heat transfer and ow friction inside the microchannels. This is consistent with the analysis by Meinhart and Wereley [1] who showed that the viscous resistance inside nanotubes can be decreased by weakening the surface hydrophilic property. 3.3. Eect of geometric parameters It is known that the height-to-width ratio has great eect on the ow friction and heat transfer in the rectangular microchannels [10,11]. For the trapezoidal mi-

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556

2553

crochannel, its cross-sectional shape is determined by two aspect ratios, the height-to-top width ratio H =Wt and the bottom-to-top width ratio Wb =Wt . Therefore, there are three geometric parameters including Wb =Wt , H =Wt and length-to-diameter ratio L=Dh , which aect the friction and heat transfer in the trapezoidal microchannels. Figs. 7 and 8 show the experimental results of microchannels #1 to #6. Since these microchannels were etched under the same conditions in the same silicon wafer, they have approximately the same order of surface roughness (9:85 105 4:30 105 ) and same surface hydrophilic property. Thus, the dierences in the

apparent friction constant and Nusselt number of these microchannels can be attributed to the dierences in their geometric parameters Wb =Wt , H =Wt and L=Dh . The results of microchannel #8 were also included in Figs. 7 and 8 to be compared with the data of microchannel #5. It is found that even if the length-to-diameter ratio L=Dh and other conditions are the same, the apparent friction constant and Nusselt number of the microchannels #5 and #8 which have dierent cross-sectional aspect ratios (Wb =Wt and H =Wt ) were much dierent. 3.4. Eect of Reynolds number From Figs. 38, it can be observed that: (1) the Nusselt number and apparent friction constant increased with the increase of Reynolds number; (2) at very low Reynolds numbers (0 < Re < 100), the Nusselt number increased drastically with the increase of the Reynolds number; but this increase became much more gently for Re > 100; (3) The changes in the Nusselt number and the apparent friction constant due to different geometric parameters, surface roughness and surface hydrophilic property were more obvious at large Reynolds numbers than at low Reynolds numbers. When comparing with the experimental results with those of Peng and Wang [16], it was found that the Nusselt numbers from this experiment have the same order of magnitude as those reported by Wang and Peng [16] although the former are a little higher than the latter. This is probably due to the dierent denitions of the DTm were used for the denitions of the Nusselt number, and dierent shape of the microchannels were used. 3.5. Correlations of laminar convective heat transfer

40 36 32 28

fapp Re

24 20 16 12 8 0 200 400 600 800 1000

Re

Fig. 7. Eects of geometric parameters on apparent friction constant.

With all factors (H =Wt , Wb =Wt , L=Dh , k =Dh , Re, Pr, and surface material) taken into consideration, the laminar convective heat transfer and friction data for the deionized water owing through the 13 dierent trapezoidal microchannels can be correlated in the forms of Eq. (10). The Nusselt numbers are correlated in two separate regimes: Re < 100 and 100 < Re < 1500 as follows: 10 < Re < 100 The correlation equation for the Nusselt number in this range of the Reynolds number is 3:547  3:577  Wb Wt Nu C1 Re0:946 Pr0:488 1 Wt H  0:041  1:369 k Dh L Dh

Nu
1

#1, #2, #3, #4, #5, #6, #8,

Wb/Wt=0.934, H/Wt=0.0382, L/Dh=285.41 Wb/Wt=0.873, H/Wt=0.0731, L/Dh=298.67 Wb/Wt=0.827, H/Wt=0.1024, L/Dh=310.46 Wb/Wt=0.774, H/Wt=0.1326, L/Dh=325.11 Wb/Wt=0.615, H/Wt=0.2260, L/Dh=370.85 Wb/Wt=0.390, H/Wt=0.3562, L/Dh=453.79 Wb/Wt=0, H/Wt=0.6453, L/Dh=362.35

0 0 200 400 600 800 1000

11a

Re

Fig. 8. Eects of geometric parameters on Nusselt number.

where C1 is a factor depending on the surface hydrophilic property: C1 6:7 for silicon surfaces and

2554

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556

C1 6:6 for thermal oxide surfaces. Thus, the eect of hydrophilic property on Nusselt number is small for 10 < Re < 100. Eq. (11a) is valid for 4:05 6 Pr 6 5:79, 0 6 Wb =Wt 6 0:934, 0:038 6 H =Wt 6 0:648, 3:26 104 6 k =Dh 6 1:09 102 , and 191:77 6 L=Dh 6 453:79. 100 < Re < 1500 The Nusselt number in this Reynolds number range is correlated as 0:908  1:001  Wb Wt Nu C2 Re0:148 Pr0:163 1 Wt H  0:033  0:798 k Dh Dh L

3.6. Correlation of the apparent friction constant The apparent friction factor is correlated by  4:359  4:444 Wb Wt fapp Re C3 Re0:089 1 Wt H  0:028  1:023 k Dh Dh L

12

11b

where C2 47:8 for silicon surfaces and C2 54:4 for thermal oxide surfaces. It can be seen that the surface hydrophilic property has signicant eects on the Nusselt number for 100 < Re < 1500. Eq. (11b) is valid for 4:44 6 Pr 6 6:05, 0 6 Wb =Wt 6 0:934, 0:038 6 H =Wt 6 0:648, 3:26 104 6 k =Dh 6 1:09 102 , and 191:77 6 L=Dh 6 453:79. Fig. 9 gives the comparison of Eqs. (11a) and (11b) with experimental data. It is shown that the correlations are in reasonable agreement with experimental data. The range of deviations of Eqs. (11a) and (11b) from experimental data are 20.3% and 19.8%, respectively. It is pertinent to note that the vertical coordinates in Fig. 9 are dierent for the two ranges of the Reynolds number.

where C3 508:7 for silicon surfaces and C3 540:5 for thermal oxide surfaces. Thus, the thermal oxide surfaces have a higher friction factor than a silicon surface. Eq. (12) is valid when 10 < Re < 1500, 0 6 Wb =Wt 6 0:934, 0:038 6 H =Wt 6 0:648, 3:26 104 6 k =Dh 6 1:09 102 , and 191:77 6 L=Dh 6 453:79. Fig. 10 is a comparison of the correlation equation given by Eq. (12) with experimental results. It is found that the correlation deviates from experimental data by no more than 19.3%. 3.7. Heat ux per pumping power and per temperature dierence According to the experimental results, the increase of Nusselt number is usually at the expenses of the increase in apparent friction constant. In this paper, we use the heat ux per pumping power and per temperature difference, qp , to evaluate the performance of dierent microchannels, qp Q=N Aw P DTm 13

10

100

1000 10

Nu/(C1(1-Wb/Wt)3.547(Wt/H)3.577(k/Dh)0.041(Dh/L)1.369Pr0.488)

1000

Nu/(C2(1-Wb/Wt)0.908(Wt/H)1.001(k/Dh)0.033(Dh/L)0.798Pr0.163)

100 Channel #1 Channel #2 Channel #3 Channel #4 Channel #5 Channel #6 Channel #7 Channel #8 Channel #9 Channel #10 Channel #11 Channel #12 Channel #13 Eq.(12a) Eq.(12b)

10

1 10 100 1000

0.1

Re
Fig. 9. Comparison of heat transfer correlations with experimental data.

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556
1E+008

2555

1E+007

q p ( m -2 .

1E+006

Channel #1 Channel #2 Channel #3 Channel #4 Channel #5 Channel #6 Channel #7 Channel #8 Channel #9 Channel #10 Channel #11 Channel #12 Channel #13

C )
1E+005

Fig. 10. Comparison of frictin correlation with experimental data.

where P is the pumping power needed to drive the deionized water through the microchannels, P DP Ac  u 14

Re
Fig. 11. Heat ux per pumping power and per temperature dierence for dierent silicon microchannels at mean temperature of 30 C.

Substituting Eqs. (3), (5), (8), and (14) into Eq. (13) and rearranging, we can obtain qp Nu 1 D3 k 2 h fapp Re Re LAc 2qm3 4. Concluding remarks 15 Experimental data of Nusselt number and apparent friction constant for the laminar ow of deionized water through the thirteen trapezoidal microchannels having dierent geometric paramerters, surface roughness, and surface hydrophilic properties, have been obtained in this paper. The following conclusions are obtained: 1. the bottom-to-top width ratio, the height-to-top width ratio and the length-to-diameter ratio have great eect on the laminar Nusselt number and apparent friction constant of the trapezoidal microchannels; 2. the laminar Nusselt number and apparent friction constant of the trapezoidal microchannels increase with the increase of surface roughness. This increase is more obvious at large Reynolds numbers than that at low Reynolds numbers; 3. the Nusselt number and apparent friction constant of the trapezoidal microchannels having strong hydrophilic surfaces (thermal oxide surfaces) are larger than those having weak hydrophilic surfaces (silicon surfaces). This suggests that convective heat transfer can be enhanced by increasing the surface hydrophilic capability at the expense of increasing pressure drop.

Substituting Eqs. (11a), (11b) and (12) into Eq. (13), the values of qp for dierent microchannels at various Reynolds numbers can be obtained. Fig. 11 is the comparison of heat ux per pumping power and per temperature dierence of the 13 microchannels at the mean water temperature of 30 C. It can be seen that (1) the performance of microchannels decreased with the rise of Reynolds number in the laminar region; (2) the microchannels #7 and #9 with smaller length-to-diameter ratios had higher values of qp than other microchannels; (3) by comparing the results of microchannels #7 and #9, as well as microchannels #8 and #10, the surface roughness did not obviously aect the thermodynamic performance; (4) by comparing the results of microchannels #3 and #11, microchannels #5 and #12, and microchannels #6 and #13, the increased in the surface hydrophilic capability caused a small increase in the performance; (5) by comparing the results of all microchannels, it was found that the eect of the geometric parameters (Wb =Wt , H =Wt , L=Dh ) on the performance was much more signicant than the eects of surface roughness and surface hydrophilic property.

2556

H.Y. Wu, P. Cheng / International Journal of Heat and Mass Transfer 46 (2003) 25472556 with electrokinetic eects, Int. J. Heat Fluid Flow 18 (5) (1997) 489496. P.Y. Wu, W.A. Little, Measurement of friction factors for the ow of gases in very ne channels used for microminiature JouleThomson refrigerators, Cryogenics 24 (8) (1983) 273277. M.M. Rahman, Measurements of heat transfer in microchannel heat sinks, Int. Commun. Heat Mass Transfer 27 (4) (2000) 495506. W.L. Qu, G.M. Mala, D.Q. Li, Pressure-driven water ows in trapezoidal silicon microchannels, Int. J. Heat Mass Transfer 43 (3) (2000) 353364. W.L. Qu, G.M. Mala, D.Q. Li, Heat transfer for water ow in trapezoidal silicon microchannels, Int. J. Heat Mass Transfer 43 (21) (2000) 39253936. K.T. Ma, F.G. Tseng, C.C. Chieng, Numerical simulation of micro-channel ow over a well of hydrophilic and hydrophobic surface, in: 39th AIAA Aerospace Sciences Meeting and Exhibition, AIAA 2001-1014, Reno, NV, January 2001. H.B. Ma, G.P. Peterson, Laminar friction factor in microscale ducts of irregular cross-section, Microscale Thermophys. Eng. 1 (3) (1997) 253265. H.Y. Wu, P. Cheng, Friction factors in smooth trapezoidal silicon microchannels with dierent aspect ratio, Int. J. Heat Mass Transfer, doi:10.1016/S0017-9310(03)00106-6. X.F. Peng, G.P. Peterson, Convective heat transfer and ow friction for water ow in microchannel structures, Int. J. Heat Mass Transfer 39 (12) (1996) 25992608. M. Choi, K. Cho, Eect of the aspect ratio of rectangular channels on the heat transfer and hydrodynamics of paran slurry ow, Int. J. Heat Mass Transfer 44 (1) (2001) 5561. M. Madou, Fundamentals of Microfabrication, CRC Press LLC, Boca Raton, FL, 1997, pp. 337339. S.W. Kang, J.S. Chen, J.Y. Hung, Surface roughness of h1 1 0i orientation silicon based micro heat exchanger channel, Int. J. Mach. Tools Manufact. 38 (56) (1998) 663668. S.A.S. Asif, K.J. Wahl, R.J. Colton, The inuence of oxide and adsorbates on the nanomechanical response of silicon surfaces, J. Mater. Res. 15 (2) (2002) 546553. B.X. Wang, X.F. Peng, Experimental investigation on liquid forced-convection heat transfer through microchannels, Int. J. Heat Mass Transfer 37 (Suppl. 1) (1994) 7382.

The enhancement in heat transfer is more signicant at large Reynolds numbers; 4. the laminar convective heat transfer shows two dierent characteristics at low and high Reynolds number ranges. For very low Reynolds number ow (0 < Re < 100), the Nusselt number increases acutely and almost linearly with the rise of Reynolds number. However, the increase of the Nusselt number after Re 100 is gentle with the rise of the Reynolds number; 5. based on the 168 experimental data points, correlation equations of the Nusselt number and the apparent friction constants in terms of appropriate dimensionless parameters are obtained for the deionized water owing through the 13 microchannels; 6. heat ux per pumping power and per unit temperature dierence is proposed to evaluate the performance of the 13 microchannels used in this experiment. A comparison of results shows that the geometric parameters have more signicant eect on the performance than the surface roughness and surface hydrophilic property. Acknowledgements The authors gratefully acknowledge the support of this work through grant no. HIA98/99.EG04 and HKUST6014/02. References
[1] C.D. Meinhart, S.T. Wereley, Fluid mechanics issues at the microscale, in: 39th AIAA Aerospace Sciences Meeting and Exhibition, AIAA 2001-0720, Reno, NV, January 2001. [2] C. Yang, D.Q. Li, J.H. Masliyah, Modeling forced liquid convection in rectangular microchannels with electrokinetic eects, Int. J. Heat Mass Transfer 41 (24) (1998) 42294249. [3] G.M. Mala, D.Q. Li, C. Werner, Flow characteristics of water through a microchannel between two parallel plates

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13] [14]

[15]

[16]