Dependence of Flow Boiling Heat Transfer Coefficient on Location and Vapor Quality in a Microchannel Heat Sink

Transcription

1 Purdue University Purdue e-pubs Birck and NCN Publications Birck Nanotechnology Center Dependence o Flo Boiling Heat Transer Coeicient on Location and Vapor Quality in a Microchannel Heat Sink Ravi S. Patel Birck Nanotechnology Center, Purdue University, patel49@purdue.edu Tannaz Harirchian Birck Nanotechnology Center, Purdue University Suresh V. Garimella Birck Nanotechnology Center, Purdue University, sureshg@purdue.edu Follo this and additional orks at: Part o the Nanoscience and Nanotechnology Commons Patel, Ravi S.; Harirchian, Tannaz; and Garimella, Suresh V., "Dependence o Flo Boiling Heat Transer Coeicient on Location and Vapor Quality in a Microchannel Heat Sink" (011). Birck and NCN Publications. Paper This document has been made available through Purdue e-pubs, a service o the Purdue University Libraries. Please contact epubs@purdue.edu or additional inormation.

2 Proceedings o the ASME 011 Paciic Rim Technical Conerence & Exposition on Packaging and Integration o Electronic and Photonic Systems InterPACK011 July 6-8, 011, Portland, Oregon, USA IPACK011- DEPENDENCE OF FLOW BOILING HEAT TRANSFER COEFFICIENT ON LOCATION AND VAPOR QUALITY IN A MICROCHANNEL HEAT SINK Ravi S. Patel, Tannaz Harirchian and Suresh V. Garimella Cooling Technologies Research Center, an NSF IUCRC School o Mechanical Engineering and Birck Nanotechnology Center Purdue University, West Laayette, IN USA ABSTRACT Experiments ere conducted to determine the inluence o local vapor quality on local heat transer coeicient in lo boiling in an array o microchannels. Additionally, the variation o local heat transer coeicient along the length and idth o the microchannel heat sink or given operating conditions as investigated over a range o lo parameters. Each test piece includes a silicon parallel microchannel heat sink ith 5 integrated heaters and 5 temperature sensors arranged in a 5x5 grid, alloing or uniorm heat dissipation and local temperature measurements. Channel dimensions ranged rom 100 µm to 400 µm in depth and 100 µm to 5850 µm in idth; the orking luid or all cases as the perluorinated dielectric liquid, FC-77. The heat transer coeicient is ound to increase ith increasing vapor quality, reach a peak, and then decrease rapidly due to partial dryout on the channel alls. The vapor quality at hich the peak is observed shos a strong dependence on mass lux, occurring at loer vapor qualities ith increasing mass lux or ixed channel dimensions. Variations in local heat transer coeicient across the test piece ere examined both along the lo direction and in a direction transverse to it; observed trends included variations due to entrance region eects, to-phase transition, non-uniorm lo distribution, and channel all dryout. INTRODUCTION Relationship Beteen Heat Transer Coeicient and Vapor Quality To phase microchannel heat sinks are capable o achieving high heat transer rates, ith the added advantage o minimizing temperature luctuations over the heat sink [1]. Hoever, investigations into the relationship beteen heat transer coeicient and vapor quality in microchannels have yielded idely varying results. Establishing the relationship beteen these to variables is critical to the design o tophase microchannel heat sinks []. Trends o variation in the heat transer coeicient ith vapor quality observed in past studies included an M-shaped double peak [3], a single peak [], continuously decreasing [4], and continuously increasing to very high vapor qualities and then dropping o sharply due to dry-out [5]. All o these studies commonly concluded that the decreases in heat transer coeicient are due to contact o vapor ith the channel alls; the value o the vapor quality at hich this decrease occurs depends strongly on the lo regimes present in the test piece. There is a variety o experimental techniques and approaches that can be utilized in order to obtain the results o interest, hich adds to the complexity o investigating relationships beteen vapor quality and heat transer coeicient. One approach involves choosing a ixed temperature measurement location and varying the heat input to the luid upstream o the heat sink or to the bottom o the heat sink in order to adjust the vapor quality at the point o interest. The ixed measurement location chosen is most commonly at the exit o the test piece. Alternatively, multiple measurement locations may be spaced along the lo direction hile applying a ixed heat input to the bottom o the heat sink. In this manner, it is possible to simultaneously obtain multiple temperature measurements over a range o vapor qualities. Hoever, even studies that have used similar approaches to obtain the desired range o vapor qualities have still yielded idely dierent results ith respect to the variation in heat transer coeicient. Spatial Variation o Heat Transer Coeicient The second objective o the present study as to examine the spatial variation o local heat transer coeicient across the idth and length o a microchannel heat sink. The practical implementation o to-phase microchannel technologies demands a more thorough understanding o phase change phenomena in such devices [6]. Several studies in the literature have examined the development o the heat transer coeicient 1 Copyright 011 by ASME

3 in the streamise direction; these studies included both numerical [7] and experimental [6] analyses. A study o the literature reveals an absence o inormation on the distribution o local heat transer coeicients transverse to the lo direction in a microchannel heat sink, especially in the presence o lo boiling. In the present study, variation o the heat transer coeicient across the heat sink, both along the lo direction and in the transverse direction, are investigated. Each test piece includes 5 temperature sensors and heaters abricated into a silicon microchannel heat sink, enabling the calculation o local heat transer coeicient, h (W/m K), and the application o a uniorm base heat lux, q b (W/m ). Additionally, the impact o vapor quality on heat transer coeicient is examined and compared to indings rom past studies. EXPERIMENTAL SETUP AND DATA PROCESSING Fig. 1. Photograph o the lo loop [8]. Flo Loop Fig. 1 shos the experimental acility used in the present study. The acility is equipped ith an expandable reservoir, a pump to circulate the orking luid through the test loop, a preheater to control the degree o sub-cooling at the test section inlet, and a heat exchanger ater the test section to cool the orking luid beore returning to the reservoir. The loop includes lo meters, pressure transducers, and thermocouples at important locations; a high-speed camera is used or visualizing the lo patterns ithin the microchannels. The pressure at the outlet o the microchannel heat sink is maintained at 1 atm throughout the tests. The setup is equipped ith several degassing ports in order to ully degas the FC-77 beore each test as it has an ainity or the absorption o atmospheric gasses that result in altered luid perormance, especially during to-phase operation. Fig.. (a) Photograph o test section, and (b) schematic diagram o the microchannel heat sink illustrating the integrated sensors/heater array and construct details o test pieces [1]. Copyright 011 by ASME

4 Test Section Each test piece consists o a silicon chip (1.7 mm 1.7 mm 650 µm in size) ith microchannels saed into one side. The top cover o the heat sink is transparent, as shon in Fig. a, enabling lo visualizations to be obtained during operation. Several test pieces ith dierent channel dimensions are tested ith depths ranging rom 100 µm to 400 µm, and idths ranging rom 100 µm to 5850 µm. The average roughness o the channel side alls is 0.1 µm, hile the average roughness o the bottom all o the channels is in the range o 0.8 to 1.4 µm depending on test piece [8]. On the other side o the silicon chip are 5 diode temperature sensors and resistor heaters arranged in a 5 5 grid as shon in Fig. b. The heaters ere connected in parallel and a voltage applied in order to generate a uniorm base heat lux, q b (W/m ). There is an inherent variation in heater resistance values due to manuacturing tolerances; hoever, this led to a maximum variation in base heat lux o only ±5.5% over the heat sink. A constant current as applied to the temperature sensing diodes; the measured voltage drop across the diodes thus corresponded to their temperature. The voltagetemperature correlation or the diodes as obtained via calibration. Fig. 3 shos the reerence axes used or determination o position across the heat sink. Data Processing In order to obtain local heat transer coeicients, each o the 5 heated sections is examined individually. The net heat transerred to the luid rom each heater is ound using: q = q q (1) net here q = V / R. Heater resistance values sho a slight temperature dependence that as accounted or through calibration tests perormed on each heater beorehand. The heat losses included in the q loss term are due to conduction rom the heat sink to the exterior suraces o the test section, as ell as convection and radiation to the surroundings. The relationship beteen the rate o heat loss and the operating temperature as obtained by running the test piece dry at a range o elevated temperatures and measuring the required poer inputs. The outcome o this procedure yielded a linear relationship o the orm q loss = c1 Td + c, indicating that heat loss is primarily by conduction. In order to obtain the heat transer coeicient, several heat sink perormance parameters are irst determined. The in eiciency, η, is calculated by assuming an adiabatic tip condition, using η = tanh( md ) / md, here m = h/ ksi [8]. An overall surace eiciency or the microchannel heat sink is then ound using: loss NA η0 = 1 (1 η ) () A The channel all temperature, T, is ound by adjusting the diode temperature reading to account or conduction through the silicon chip using: T q b ( t d) = Td (3) k si Fig. 3. Coordinate axes used or measurement o location. Ater calculating the overall heat sink surace eiciency and all temperature the heat transer coeicient can be ound using: q h = η ( ) 0 T Tsat (4) Fig. 4. Notation used or critical parameters in the microchannel heat sink [1]. here q is the heat lux through the etted microchannel suraces. Since heat transer coeicient appears in calculating the in eiciency, an iterative approach had to be employed to obtain the correct value or h. The initial value or h used in the iterations as obtained by assuming 100% eiciency or the heat sink. Using this procedure a local heat transer coeicient as ound at each o the temperature sensor locations. Saturation temperature as adjusted based on variations in pressure. Additionally, during single phase operation the local bulk luid temperature as used instead to ind h. Finally, the local vapor quality or each sensor could be ound as: 3 Copyright 011 by ASME

5 1 X L q dx ( ) 0 net x = cp, Tsat Tinlet h g m RESULTS AND DISCUSSION Relationship Beteen Heat Transer Coeicient and Vapor Quality Fig. 5 shos the heat transer coeicient as a unction o vapor quality or to dierent microchannel dimensions (400 µm deep 50 µm ide and 400 µm deep 5850 µm ide); and our dierent mass luxes (50, 700, 1150, and 1600 kg/m s). The results obtained here reveal that the heat transer coeicient increases ith increasing vapor quality until it reaches a maximum value, and then decreased rapidly. Fig. 5. Heat transer coeicient as a unction o vapor quality at sensor #3 or heat sink ith microchannel dimensions: (a) 400 µm deep 50 µm ide, and (b) 400 µm deep 5850 µm ide. (5) The vapor quality and heat transer coeicient are both calculated at a ixed location near the exit o the test piece (sensor #3). A range o vapor qualities as obtained by increasing the heat input to the test piece. Tests ere terminated soon ater dry-out occurred at the channel alls. The decrease in heat transer coeicient led to increased heat sink temperatures that could have damaged the test piece. The vapor quality at hich the peak in heat transer coeicient occurs shos a strong dependence on mass lux, occurring at loer vapor qualities or increasing mass lux as seen in Fig. 5a and Fig. 5b. This trend is in agreement ith a study by Madrid et al. [9], and as attributed to the stability o the liquid ilm in slug and annular lo. Due to higher velocities accompanying the larger mass luxes, the liquid ilm gros unstable at loer vapor qualities and is unable to remain attached to the channel alls. This yields churn, or in extreme cases inverted annular, lo regimes resulting in contact beteen vapor and the channel alls. This partial dryout at the alls drastically loers the heat transer coeicient. Fig. 6 shos our distinct trends in the relationship beteen heat transer coeicient and vapor quality ound in the literature that diered rom the trends observed in the present study. The data in these plots ere obtained by digitizing the images in the original studies and plotting in a comparative manner. Table 1 identiies several key operating conditions and experimental parameters in the studies presented. In contrast ith the current study, Bertsch et al. [] ound that the vapor quality at hich the peak heat transer coeicients is observed as independent o mass lux and occurred at roughly the same vapor quality, as seen in Fig. 6a. The cause o this as traced back to the lo regimes present in the study. Due to the experimental setup and lo conditions used in that study, slug lo as the dominant lo regime present, ith annular lo occurring only beyond vapor qualities o 0.7 []. Fig. 6b shos the results o various other studies [10], [11] compiled by Bar-Cohen and Rahim [3] in hich a double peak trend as observed. Saitoh et al. [5] ound that h continuously increased until very high vapor qualities, at hich point it drastically dropped as seen in Fig. 6c. Finally, Steinke and Kandlikar [4] observed that there as a continual decrease in h ith increasing x, as shon in Fig. 6d. All our o these distinct trends dier rom those seen in the present study. The studies in the literature had many dierences rom the present study, and rom one another, including channel geometry, orking luid, and operating conditions. All o these actors contribute to dierences in the lo regimes present in the channels and the vapor quality at hich vapor comes into contact ith the heat sink alls. Trends in heat transer coeicient as a unction o vapor quality ere also examined over a range o channel dimensions. Fig. 7 shos h vs. x at sensor 3 or a ixed mass lux (50 kg/m s), channel depth (400 µm) and varying channel idths. At a ixed mass lux, the peak in heat transer coeicient is seen to occur at loer vapor qualities or a larger channel cross-sectional area. These trends cannot be compared to other studies in the literature as no detailed results related to the eect o cross sectional area on the relationship beteen h and x in microchannel heat sinks are available. 4 Copyright 011 by ASME

6 a b c d Fig. 6. Results or h vs. x rom the literature shoing (a) the peak occurring at a ixed x [], (b) a double peak resulting in an M-shaped curve [3], [10], [11], (c) h increasing until very high vapor qualities, and then dropping rapidly [5], and (d) a continual decrease [4]. Table 1 Experimental parameters or studies rom literature presented in this study. Pressure data, here available, is listed at measurement location(s) unless otherise noted. Author Year Working Fluid D h (mm) q" (kw/m ) G (kg/m s) Inlet Condition Pressure (kpa) Steinke and Kandlikar 004 ater Sub-cooled Yang and Fujita 004 R Saturated Saitoh, Daiguji, and Hihara 005 R134a Saturated (Drop across test piece) Cortina-Diaz and Schmidt 006 n-octane Sub-cooled Bertsch, Groll and Garimella 008 HFC-134a Sub-cooled and Saturated Copyright 011 by ASME

7 Fig. 7. Heat transer coeicient as a unction o vapor quality or a ixed mass lux (50 kg/m s) and channel depth (400 µm), and varying channel idth (µm). Spatial Variation o Heat Transer Coeicient Local variations in heat transer coeicient ere examined along the length and idth o the microchannel heat sink. The test piece as subject to both single- and to-phase operating conditions, ith luid lo in the positive X direction. For streamise measurements, data ere gathered rom sensors 3, 8, 13, 18 and 3 (Fig. b); or measurements normal to the lo, measurements ere taken at sensors 11, 1, 13, 14 and 15. The coordinate axes used or reerence measurements are shon in Fig. 3. Fig. 8 shos the spatial variations in h or a ixed channel size (depth 400 µm idth 50 µm) operating at to dierent mass luxes. The variations are presented in the to directions mentioned previously: along the lo direction (X) and orthogonal to it (Y). For each case several values o base heat lux are considered. Several key eatures may be identiied hen examining the variations in heat transer coeicient normal to the lo direction (Y-direction). During single-phase conditions the heat transer coeicient is relatively constant; hoever, as heat lux a Base Heat Flux, q b" b Base Heat Flux, q b" c Base Heat Flux, q b" d Base Heat Flux, q b" Fig. 8. Spatial variation o local heat transer coeicient transverse to lo direction or (a) G = 700 kg/m s and (b) G = 1600 kg/m s; and along the lo direction or (c) G = 700 kg/m s and (d) G = 1600 kg/m s. 6 Copyright 011 by ASME

8 is increased, boiling causes variations in h in the transverse direction. These variations ere due to the non-uniorm incipience o boiling across the chip, as ell as uneven lo distribution due to maniold design. In Fig. 8a, or q b = kw/m, to dierent values or h are observed on to transverse sides o the heat sink, ith the loer value being close to the single-phase cases. This demonstrates that at a ixed location donstream rom the inlet, boiling has started in the channels along one side o the heat sink, resulting in an increase in the heat transer coeicient, hile single-phase liquid is loing in the rest o the channels. Uneven incipience o boiling in dierent microchannels as conirmed through visualizations as ell, as shon in Fig. 9. Additionally, during to-phase operation there as a maximum o 1.9% variation in h transverse to the lo direction. For the cases presented in Fig. 8 the test piece as operating at Re = 360 or G = 700 kg/m s, and Re = 80 or G = 1600 kg/m s. Jones et al. [1] demonstrated that at Re = 10 there as up to 30% more lo in the center channels than the channels near the sides o a microchannel heat sink, this maldistribution o lo as due to the design o the inlet and exit maniolds. For a suiciently high heat input, dry-out occurs at the channel alls, yielding drastically loer heat transer coeicients; this can be seen in Fig. 8a here the heat transer coeicient or q b = kw/m is less than even that o q b = kw/m. Trends in the variation o heat transer coeicient transverse to the lo direction in microchannel heat sinks are not readily available in the literature or comparison. As seen in Fig. 8c (q b = 64.96, 8.6 kw/m ) and Fig. 8d (q b = kw/m ) during single-phase operation, the heat transer coeicient in the streamise direction started at a maximum at the test section inlet and steadily decreased. This can be attributed to the development o thermal and momentum boundary layers ithin the channels, as ell as variations in luid properties as the luid temperature changes. These indings are consistent ith the results o a past numerical study by Li et al. [7] and an experimental study by Kuznetsov and Shamirzaev [6] investigating streamise heat transer coeicient variation in microchannel heat sinks. A sudden increase in the heat transer coeicient or a small increase in heat input can be observed hen boiling begins in the test piece. This is noticed in Fig. 8c beteen the to single-phase cases (q b = 64.96, 8.6 kw/m ) that lie on top o each other, and the to-phase case ith q b = W/m. Additionally, in Fig. 8d, it is evident that or q b = kw/m, the transition rom single-phase to to-phase lo occurs near X = 5mm here the heat transer coeicient suddenly increases. Even during to-phase operation it is apparent that the heat sink has an entrance region here boundary layers are developing, as indicated by the act that the heat transer coeicient is highest near the inlet and decreases in the streamise direction. In the present study luid entering the heat sink as alays sub-cooled. While Kuznetsov and Shamirzaev [6] obtained similar results using sub-cooled luid, they also considered a to-phase inlet condition and ound that the heat transer coeicient remained relatively constant along the heat sink and did not exhibit entrance-region trends. Finally, at the highest heat input values, the test piece experiences dry-out at the channel alls near the exit o the heat sink, resulting in a dramatic drop in the heat transer coeicient toards the exit o the channels, as seen in Fig. 8c or q b = kw/m. Fig. 9. The location o incipience o boiling (solid red lines), and ully developed boiling (dashed green lines) in dierent channels in a heat sink. CONCLUSIONS The dependence o local heat transer coeicient on location and vapor quality in a silicon microchannel heat sink has been investigated experimentally using FC-77 as the orking luid and compared to trends ound in past studies. There have been many studies investigating the relationship beteen heat transer coeicient and vapor quality ith varying results. The main issue identiied in the present study, as ell as that o Madrid et al. [9], as that there is a stronger relationship beteen h and the lo regime present ithin the test piece than vapor quality. Ultimately, the maxima ere associated ith the point just beore vapor came into contact ith the channel alls. In the present study, a single peak as observed, and this peak shoed a strong dependence on mass lux, occuring at loer vapor qualities ith higher mass lux. For a ixed mass lux it as observed that the peak also occurs at loer vapor qualities or increasing channel cross sectional area. When examining the variation o heat transer coeicient transverse to the direction o the lo it as observed that at a ixed donstream location boiling can occur in some channels hile not in others. When this occurred the heat transer coeicient diered by up to 36.6% across the test piece. Due to maniold design there as a slight maldistribution o lo 7 Copyright 011 by ASME

9 among the channels as made evident by the 1.9% maximum variation in heat transer coeicient transverse to the lo direction during to-phase operation. These luctuations can be minimized ith more careul maniold design at the inlet and exit o the heat sink. In agreement ith past numerical and experimental studies, it as observed that the heat transer coeicient as at a maximum value near the entrance o the channels. For singlephase lo, the heat transer coeicient as at a maximum at the inlet o the channels, decreased through the entrance region, and approached a ully developed value. In some cases, there as a large jump in h at a location donstream rom the inlet here boiling began. When the entire length o the channels as operating under to-phase lo, a distinct entrance region ith a higher heat transer coeicient as still present. Ultimately at high heat luxes the heat transer coeicient rapidly decreased toards the exit o the heat sink due to dryout at the channel alls. NOMENCLATURE A Heat sink base area, m b A Wetted area o a in, m A Total etted area o microchannels, m c1, c Constants in heat loss and temperature relation d Microchannel depth, m D h Hydraulic Diameter, mm G Mass lux, kg / m s h Heat transer coeicient, W/ mk h g Latent heat o vaporization or FC-77, J / kg k si Thermal conductivity o silicon, W / mk L Heat sink idth, microchannel length, m m Intermediate variable in in eiciency calculation, 1 m m Mass lo rate o luid, kg / s N Number o microchannels in a heat sink q Heat dissipated by heater resistors, W q Base heat lux, b W / m q loss Heat loss, W q Heat transerred to luid, W net q Wall heat lux, W / m R Heater resistance, Ω Re Reynolds number t Heat sink thickness, m T d Diode temperature, C T inlet Fluid temperature at heat sink inlet, C T sat Fluid saturation temperature, C T Wall temperature, C V x X Y Greek η η 0 Voltage applied to heaters, V Microchannel idth, m Fin idth, m Vapor quality Position in X direction, m Position in Y direction, m REFERENCES Eiciency o a in in heat sink Overall microchannel heat sink eiciency [1] Harirchian, T., and Garimella, S. V., 008, Microchannel Size Eects on Local Flo Boiling Heat Transer to a Dielectric Fluid, Int. J. Heat and Mass Transer, 51, pp [] Bertsch, S.S, Groll, E. A., and Garimella, S. V., 008, Rerigerant Flo Boiling Heat Transer in Parallel Microchannels as a Function o Local Vapor Quality, Int. J. Heat and Mass Transer, 51, pp [3] Bar-Cohen, A., and Rahim, E., 009, Modeling and Prediction o To-Phase Microgap Channel Heat Transer Characteristics, Heat Transer Engineering, 30(8), pp [4] Steinke, M. E., and Kandlikar, S. G., 003, Flo Boiling and Pressure Drop in Parallel Flo Microchannels, Paper No. ICMM , Proc. 1 st International Conerence on Microchannels and Minichannels, American Society o Mechanical Engineers, Rochester, NY, pp [5] Saitoh, S., Daiguji, H., and Hihara, E., 005, Eect o Tube Diameter on Boiling Heat Transer o R-134a in Horizontal Small-Diameter Tubes, Int. J. Heat and Mass Transer, 48, pp [6] Kuznetsov, V. V., and Shamirzaev, A. S., 009, Flo Boiling Heat Transer in To-Phase Micro Channel Heat Sink at Lo Water Mass Flux, Microgravity Science and Technology, 1(1), pp [7] Li, Z., Huai, X., Tao, Y., and Chen, H., 007, Eects o Thermal Property Variations on the Liquid Flo and Heat Transer in Microchannel Heat Sinks, Applied Thermal Engineering, 7, pp [8] Harirchian, T., and Garimella, S. V., 009, Eects o Channel Dimension, Heat Flux, and Mass Flux on Flo Boiling Regimes in Microchannels, Int. J. Multiphase Flo, 35, pp Copyright 011 by ASME

10 [9] Madrid, F., Caney, N., and Marty, P., 007, Study o a Vertical Boiling Flo in Rectangular Mini-Channels, Heat Transer Engineering, 8(8-9), pp [10] Yang, Y., and Fujita, Y., 004, Flo Boiling Heat Transer and Flo Pattern in Rectangular Channel o Mini-Gap, Paper No. ICMM , Proc. nd International Conerence on Microchannels and Minichannels, American Society o Mechanical Engineers, Rochester, NY, pp [11] Cortina-Diaz, M., and Schmidt, J., 006, Flo Boiling Heat Transer o n-hexane and n-octane in a Minichannel, Proc. 13th International Heat Transer Conerence, Begell House Inc., Sydney, Australia, 13. [1] Jones, B. J., Lee, P., and Garimella, S. V., 008, Inrared Micro-Particle Image Velocimetry Measurements and Predictions o Flo Distribution in a Microchannel Heat Sink, Int. J. Heat and Mass Transer, 51, pp Copyright 011 by ASME